SEMEME, POLYSEMY AND PROTOTYPE THEORY

(="Sémème, polysémie et théorie du prototype", in: BSL 91.1, 1996, 77-96)

Abstract - We would like to show that there is no incompatibility between the notion of “sememe” of some structural semantics and the inescapable notion of polysemy. But, to do it, it is important not to call “seme” each semantic feature, and not to see the sememe as a mere intersection between all the meanings of a given lexical unit (or lexeme).

If one studies semantics and wants to describe the meanings of lexical units in a rigorous and systematic way, one has to resort to the notions of seme (minimal distinctive semantic feature) and sememe, which allow to contrast the meanings of different lexical units with one another. The seme is usually defined as a distinctive semantic feature, namely a component of meaning which is peculiar to a lexical unit and notably allows contrast with another lexical unit. In this case the sememe is then, if we take up one of the first definitions of Bernard Pottier, “the set of the semantic features (or semes) falling within the definition of the substance of a lexical unit” (adapted from Pottier, 1963, 8).

But if one studies lexicology and wants to explain the real meaning of lexical units in the sentence, one is obliged to note the importance of polysemy, that is to say the fact that a single entry in a dictionary should nearly always have different meanings, namely several separate semantic definitions.

The problem that arises in that case is how to make theoretically compatible the theoretical demands of semantics and the practical demands of lexicology, which appear to be, at first sight, divergent, or even contradictory.

The lexicographers’ attitude, when they bring up the problem, may consist in considering the various semantic definitions dictionaries give for a single word as many distinct lexical units. This is what Alain Rey seems to admit. According to him, “the definition [of dictionaries] corresponds to an attempt to do a semantic analysis, and tends to draw for each meaning significant features or semes” (adapted from Rey, 1977, 17). This, he specially specifies, is an “analysis of the content (or signifier) into differential features or semes structured into sememes” (adapted from Rey, 1977, 40). It amounts to saying that a single word will have as many separate sememes as semantic definitions or separate meanings. So he adds that the intransitive verb descendre has “in the DFC […] 5 sememes or values” but “in Le Petit Robert it has 7 (concerning uses with animate subjects) + 4 (inanimate subjects) = 11 (adapted from Rey, 1977, 40). As a consequence, the first dictionary mentions that this verb has 5 separate meanings, numbered from 1 to 5, and for the second dictionary 11 separate meanings, numbered from 1 to 7 on the one hand, and from 1 to 4 on the other. However these two dictionaries do not always give an explicit definition of the meanings recognized as being separate by different numbers and by their own examples.

This is also what makes such a semantician as Robert Martin, when, to illustrate this type of polysemy corresponding to a “narrowing of meaning”, he acknowledges that the lexical unit femme (“1. woman, “2. wife”) consists of two sememes, namely:

S¹ : “female /s¹₁/ human being /S¹/

S² : “female /s²₁/ human being /S²/ who is or was married /s²₂/”

(adapted from Martin, 1983, 64; 1972, 128), the second sememe being distinguished from the first one by the presence of the additional seme /s²₂/ “who is or was married”.

Such a view brings up a rather serious problem concerning the theory of linguistic signs and even that of polysemy. It is indeed tempting to define polysemy as is proposed in la grammaire d’aujourd’hui:

“The term polysemy is used to describe the fact that a lexical unit correspond to various meanings; at the level of the sign, we will say that a single expression (signifier) is connected with various contents (signified)”.

(adapted from Arrivé, Gadet, Galmiche, 1986, 534)

But logically, this in fact means that the polysemic word corresponds to several homonymous signs since there are several inevitably different combinations of distinct contents with an identical expression. As a result polysemy can no longer be distinguished from homonymy: indeed in both cases distinct contents are associated with identical expressions. The only slight difference between polysemy and homonymy would lie in the fact that the meanings of two alleged homonymic lexical units would have nothing common, whereas those of two polysemic lexical units would be, in a way, connected to one another. Such seems to be Patrick Charaudeau’s point of view, at least when he defines polysemy in these terms:

“In a narrow sense, polysemy refers to, at the synchronic level, the fact that a single expression should cover various contents between which, unlike homonymy, there is a semantic intersection”.

(adapted from Charaudeau, 1992, 63)

Such a definition of polysemy may appear too restrictive. It would certainly be better to postulate that, if homonymy really corresponds with two different linguistic signs, polysemy refers to the various meanings of a single linguistic sign. But this implies that these meanings, though different from one another, represent the content of this linguistic sign and consequently one and the same content. A problem comparable to that of the squaring of the circle faces us then; because how can we reconcile the necessary unity of the sememe and the (not less) unquestionable diversity of meanings of the lexical unit?

The first solution that comes to mind is to define the sememe as the semantic intersection of the various meanings of the lexical unit. Such is Mariana Tutescu’s view when she writes:

“Polysemy only exists in accordance with the existence of a semantic nucleus common to polysemic elements, with an invariant which has been actualized in the values of uses” (adapted from Tutescu, 1978, 137)

As an example she quotes the word guide, whose two major meanings are defined by the DFC as follows:

“1° The one who leads, who shows the way (in the mountain, in a museum) […]

4° A book that gives classified information on a subject.” (adapted from DFC 585).

and “where the common semantic nucleus is “whose purpose is to give useful information (adapted from Tutescu, 1978, 137), which may be diagrammed as follows:

“whose purpose is to give useful information”

Fig.1

This description, which, in this particular case, is perhaps not the best one, presupposes that the two separate meanings of the lexical unit guide should appear according to whether, in a sentence, the sememe is applied to a person or to a book. So the content would really be the same, what would be different would only be the name/designation with which this content is connected.

We could support the same view concerning the two meanings of the word femme (“1. woman, “2. wife”) “if we say that its sememe is the one that Robert Martin considers as the first sememe, namely:

S¹ : “female /s¹₁/ human being /S¹/

and that the second meaning corresponds to a narrowing of meaning which adds to this sememe, not a “specific seme”, as says Robert Martin (adapted from Martin, 1983, 65), but only, in our opinion, a semantic feature which would be due to the linguistic context where this lexical unit occurs. It is indeed when we speak about a person’s wife (“la femme de quelqu’un”) or when we say that a man takes a wife (“prend femme”), to confine ourselves to the two types of examples of the DFC, that the lexical unit femme has the underextended meaning of wife. The context of an of-construction or that of a subject of the verb prendre (“take”) refering to a “male human being” indeed leads the hearer to add to the sememe “female human being” the semantic feature “who is in exclusive contact with the said male human being”, and consequently in French society, “who is or was married with the said human being”. If we speak in this way about the addition of the contextual semantic feature, and not about the seme it allows us to say that this lexical unit has in both cases the same central meaning, and thus that it indeed has a single sememe while being polysemic.

Unfortunately, each case of polysemy cannot be solved in such a satisfactory way, from a theoretical point of view. It is often difficult to find semantic intersection between the separate meanings of a single lexical unit without shrinking this lexical unit away. What does it happen, for instance, when polysemy corresponds to what dictionaries usually term a broadening of meaning? Robert Martin illustrates this phenomenon with the word minute (“minute”), which primarily means “one sixtieth of an hour” and “short period of time” (adapted from DFC). Robert Martin more technically rewords both meanings admitting the two following sememes:

S¹ : “Period of time /S¹/ corresponding to one sixtieth of an hour /s¹₁/”

S² : “Short /s²₁/ period of time /S²/

which have the archisememe “period of time /S/” in common (adapted from Martin, 1983,65). After noting that the seme “s¹₁ implies the seme ‘short’, one sixtieth of an hour [being] a short period of time or at least a period of time which may be considered as such”, and that the seme “short” (noted s²₁) is thus contained, potentially at least, within [the sememe] S¹” (adapted from Martin, 1983, 66). Robert Martin therefore concludes that the relationship of extension of meaning is a “deletion of specific features” (Martin, 1983, 66). If now, for theoretical reasons, we want this lexical unit to have, not two but a single sememe, and if we note that one seme which is or may be subtracted cannot be a distinctive semantic feature, then the sememe of the lexical unit minute is limited to the sole semantic feature common to the two meanings involved, namely “space and time”. But it is obvious that if such a semantic definition is distinctive, it not at all allows one to identify in an exclusive way the lexical unit minute. It is thus not an accurate semantic definition of this lexical unit.

And how shall we behave when, instead of a simple broadening of meaning, we are faced with a shift in meaning which no longer seems to have anything common with the so-called primary meaning, as is the case of the Latin verb sentire? This verb may indeed be either a verb of perception, or a verb of opinion, which the Gaffiot dictionary expresses in the following phrases:

“1. perceive by senses”

“2. perceive by intelligence”

which might be clarified, from Paul Morillon’s thesis, by the next two semantic definitions:

“1. to have a sensory impression”

“2. to become aware (possibly through sensory impressions).”

If we relied on Gaffiot’s wording, it would be tempting to postulate as sememe the intersection of both meanings, namely the sole semantic feature “percevoir” (“perceive”) but it would only be a device of wording. Indeed how would it be possible to explain the meaning of “percevoir” (“perceive”), which is allegedly common to both separate meanings of sentire, other than by something like “seize or apprehend by senses”. This would show well that the notion “perceive” is not common to both leanings, since it implies the notion of “sensory impression”, specific to the primary meaning of the verb. In brief, the explanation of the semantic broadening or the meaning transfer by a distinctive semantic intersection seems to ruin, more or less completely the very notion of sememe, by making it as it were flimsy. And if we rely on Paul Morillon’s wording, we can see that there is no longer anything common between the two meanings since the semes of the first meaning are only optional semantic features, thus not distinctive, of the second meaning.

There is another instance of polysemy in what dictionaries call metaphorical meanings of words. Some also tried to define metaphor as a semantic intersection between the two lexical units related by the metaphorical construction, and thus by an inclusion of this common seme within the sememe of the lexical unit which takes a metaphorical meaning, namely by a semantic intersection between the sememe of this lexical unit and the meaning it takes when it is in context. Such is Michel Le Guern’s view, according to whom:

“Metaphor can be explained at the level of logical communication by the subtraction, or more exactly by parenthesising part of the semes which make up the lexical unit.” (Le Guern, 1963, 15)

or Robert Martin’s, when he proposes the French word cuirasse (“1. breastplate” “2. armour”) as an example, thus taking up the two meanings admitted by the DFC:

S¹: “a piece of armour /S¹/ that protect /s¹₁’/ -ed /s¹₁’’/ the chest /s¹₂/

S²: “moral attitude /S²/ that protects /s²₁’/ from wounds to one’s self esteem, suffering /s²₂/

He explains that “there is a similarity” (adapted from Martin, 1963, 69) between the metaphorical meaning S² and the primary meaning S¹ of the lexical unit. He adds that this similarity “consists in the identity of at least one of the specific semes.” (adapted from Martin, 1963, 69), in this case the semes /s¹/ “that protect”. The metaphor would thus make disappear semes of the sememe and would only keep some of them, or even a single one. This poses the problem of the distinctiveness of those of the so-called semes that may be subtracted, which would lead us to think that only the semantic feature(s) that remain are semes. But in addition to this, for a good number of metaphors, it is impossible to find a common seme between the metaphorical meaning and the usual meaning of the lexical unit.

Let us take the traditional example of metaphor that Joëlle Gardes and Jean Molino give to illustrate traditional grammar which, from Quintilien on, defines metaphor as “an abridged comparison, similitudo breuior” (Molino-Gardes, 1987, 1, 161), that is to say a comparison without an “as” of comparison:

Achille est un lion (“Achilles is a lion”).

If, on the basis of the lexicological definitions given by the DFC, Le Nouveau Petit Robert and Le Petit Larousse Illustré, we propose as sememe for the word lion:

“a large wild carnivorous mammal, having a tawning yellow coat with a shaggy mane”

and if, following Searle’s proper nouns theory (cf. Searle, 1972, 218-227), we admit, concerning Achilles, the following “identifying description”:

“son of King Pélée, central character of the Iliad, whose wrath…

we must acknowledge, from the one hand that there is no seme common to these two lexical units, and on the other that the metaphorical meaning of “courageous person” that these three dictionaries attribute to the word lion apparently has no common seme with what seems to be the sememe of this word. Besides, this metaphorical meaning is different from that of this very word in the phrase

la part du lion. (“the lion’s share”)

which means, according to the DFC, “the largest portion” that a person gives oneself because he is the strongest” and also has no common seme with the semantic definition of the word lion. Under these conditions, it seems more accurate to say with Jean Molino, Françoise Douay-Soublin and Joëlle Gardes-Tamine, that “metaphor does not correspond to the highlighting of a common seme, it corresponds to the possibility of finding new common semes” (Molino-Soublin-Tamine, 1979, 30). It would be more accurate, from the terminological and theoretical points of view, to say that metaphor entails the appearance, namely just as well the discovery or the making up of new semantic features between lexical units that are brought together. Indeed, the features “courageous” or “the largest” are not at all distinctive semantic features of the lexical unit lion.

If, for theoretical reasons, one wants to keep the notion of sememe, despite, in particular polysemy and metaphor, which seem to make it shrink away, one must perhaps consider the sememe as a fuzzy set of features or as a set of semes which changes with the wind. This is more or less what could say some of those who are tempted to apply to semantics what cognitive psychologists call prototype theory. The American psychologist Eleanor Rosch indeed claims that

“Natural categories have an internal structure consisting of a prototype (clear-cut instances, best examples) of the category and non typical members in an order which goes from best instances to less good ones.” (adapted from Rosch, 1975, 544)

In this way, to take up an example that will be seen as classical or even a little repeated, the sparrow or the robin would have a higher degree of representativeness of the category bird than the ostrich or the chick, the latter being themselves less marginal in the category than the penguin or the kiwi are. The most marginal instances sharing fewer characteristics with typical instances than less marginal instances do.

If we transpose this psychological theory and apply it to semantics, we will consider the meaning of a lexical unit as “the mental representation or concept of its prototype-object” (adapted from Kleiber, 1990, 61). We will also be able to say that “the word ‘polysemic’ only represents […] one category whose prototype consists of the primary, basic or central meaning, whose other meanings are more or less remote instances” (adapted from Kleiber, 1990, 100-101). Let us consider how this would be applied to the meaning of the word oiseau (“bird”). If, in the first place, we compare the semantic definitions given by the three above-mentioned French dictionaries, it seems that we can consider as the sememe of this word the definition of the DFC, which is the less wordy and encyclopaedic one:

“oviparous vertebrate animal with feathers, wings, two legs and a beak, which can ordinarily fly”. (adapted from DFC, 1966, 791)

but we must however subtract the last feature, which is not really distinctive, since it is not always significant. And if, after that, we interpret semantically the typical structure corresponding to the extension of the concept bird, such as D. Geeraerts represents it in figure 2 (adapted from Kleiber, 1990, 56), it leads us to say that the typical definition of the lexical unit bird consist of seven features which may be termed “(proto)typical”:

“1. able to fly”, “2. with feathers”, “3. S-shaped”, “4. with wings”, “5. not domesticated’, “6. oviparous”, “7. with a beak”.

We can note that all these features except for the 3^rd and the 5^th ones perfectly correspond to the semantic definition of the word bird, or more exactly to the definition the DFC proposes concerning the primary meaning of this word. The prototypical definition is thus comparable to, or even identical with, the sememe of the lexical unit bird. But what makes things considerably more complicated is the fact that the said

typical features may not be used in some use or other or in marginal meanings, as can be seen in:

The chick or the ostrich are birds, although they cannot fly.

Or else:

The penguin is a bird, whereas according to Le Nouveau Petit Robert, it “only has rudiments of wings” (1993, 1245)

Consequently, if the prototypical definition comes close to the sememe, it differs from it insofar as it consists of a more or less distinctive set of features. In fact, only two of these typical features, which are always present, can be considered as real semes or distinctive features: the next two typical features, namely “6. oviparous” and “7. with a beak”. What is worse, if the seme “with a beak” may really be considered as specific to the lexical unit bird, the seme “oviparous”, on the contrary, is not at all peculiar to it, since fish or snakes are oviparous. This strongly reduces the significance of the very notion of distinctive feature; because things turn out as if the most typical features such as “1. able to fly” were finally the most distinctive ones, since they are often absent, and the most distinctive features such as “6. oviparous” and “7. with a beak” were the less typical ones.

Some semanticians, advocates of prototype theory, even proposed a reformulation of the said theory that Georges Kleiber precisely calls “extended version of prototype theory”, which removes any relevance and substance from the notion of seme. It is in fact a return to the famous theory of “family resemblance” formerly developed by Wittgenstein, whose great novelty and difference relative to the standard version of prototype theory is, as Georges Lakoff says, “the idea that members of a category may be connected to one another without their having a property in common that defines the category” (adapted from Lakoff, 1987, 12). If this is applied to semantics, it means that there is usually no semantic feature common to the set of meanings of a lexical unit. Wittgenstein produced evidence of this with the word jeu (“1. game”, “2. play”, “3. gambling”), which he described in the following way:

“Let us consider for instance the process we name “jeux”. I mean draughts, chess, games of card, games of ball, sporting events. What do they have in common? – Don’t say: they must have something common, or they wouldn’t be called “jeux”, – but first try to find if they have something in common – Because if you consider that it is right, you certainly won’t see what is common to all of them, but you will see analogies, affinities, and you will note a whole series of them. Consider for instance draughtboard games with their various affinities. Then switch to games of card: there you will find certain conformity with the former class, many common features disappear whereas other appear. If we switch to ball games, there still remains something common, but much disappears. – Are all these “jeux” ‘entertaining’? Compare chess with hopscotch. Or is there for all of them a way of losing or winning, or an event of layers? Think of games of patience. In games of ball you lose or you win; but when a child throws a ball against a wall and catches it, this characteristic disappears. Consider the role played by skill and luck. And how different is skill at chess from skill at tennis. […] And thus we can pass in review many other groups of games and see analogies appear and disappear.

And such will be the result of this consideration: we can see a complex network of analogies that overlap and envelop one another. General analogies or analogies in detail.
I could not define these analogies better than with the word “family resemblance”, for it is in this way that the various resemblances that exist between the members of a family overlap; height, traits of the face, the colour of the eyes, gait, disposition… – And I said: ‘jeux’ form a family.” (adapted from Wittgenstein, 1961, 147-148)

And this is not only true at the level of the different specific meanings that may take the so-called primary meaning of the word “jeu”, but even at the level of the five major meanings that dictionaries such as the DFC and Le Nouveau Petit Robert admit concerning this word: the DFC considers that we are faced with five cases of homonymy and Le Nouveau Petit Robert five cases of polysemy. We can indeed note a certain connection between these meanings when taken two by two, but we do not find any semantic feature common to the five meanings expressed by the DFC as follows:

1. “Activity undertaken for pleasure, self entertainment or entertainment of the others, e.g. le jeu d’échec; chess, le jeu de boule; bowls…

2. “Amusement or activity of gambling money”, e.g. le jeu de la roulette; roulette.

3. “way of acting that is fickle, easy-going, gratuitous, devoid of value or seriousness (pejorative or not, according to cases)”, e.g. Cest un jeu, ce n’est pas sérieux; it’s a game, it isn’t serious.

4. “way of playing a role”, e.g. le jeu d’un acteur; the play of an actor, from which “manner of using a musical instrument”, e.g. le jeu brillant d’un violoniste, the brilliant technique of a violin player.

5. “regular movement, easy functioning of an organ or of an organism”, e.g. le jeu d’un ressort; the play of a spring.

In fact, the second meaning manifestly contains the semantic feature “entertainment” in common with the first meaning, but adds that of “gambling money”. The third one may be considered as a broadening of the fourth meaning and of the first one as well. The fifth one is, in a way, a passive version of the fourth meaning corresponding to “the way something is used or functions” and no longer to “the way someone uses or operates something”. But contrary to all expectation a priori, there is no such semantic feature as “entertainment” common to all the primary meanings of the word “jeu”. There is just between them what Wittgenstein metaphorically but rightly called a “family resemblance”, which doesn’t give a similar appearance to all of them but makes them resemble one another gradually, and possibly cross with one another some of the similarities.

T.Givon clearly schematised this notion of family resemblance by a sequence of circles, each pair of which presents a non-empty intersection (cf. Givon, 1986, 78), which means that there is actually a link between all the members thus gathered, but also that there is no common intersection between them all.

The two major meanings of the Latin verb sentire might also illustrate that sort of chain of resemblance between the meanings of a single lexical unit. Apparently, there is nothing common between one meaning of a verb of perception and one meaning of a verb of opinion as:

“1. to have a sensory impression”

“2. to become aware of something”

except when we mention the links that permit to switch from the first meaning to the second one, and that Paul Morillon words in the following way :

“1b. to become aware – through a sensory impression – of a phenomenon of a material nature” (Morillon, 1974, 533)

where the feature “to become aware” is only a semantic impact produced by the contextual particularity according to which the object/argument of the verb sentire is then the semantic content of the clause and not that of the noun phrase,

“2a. to become aware – of something difficult to discover try to hide – or that other people try to hide – thanks to a perspicacity effort” (Morillon, 1974, 533)

where the feature “to become aware” no longer needs the context to be added to the notion of perception.

“2b. to become aware – through reasoning or attention, – of something that first escaped one’s notice” (Morillon, 1974, 15)

If standard prototype theory seemed to question the notion of sememe, while allowing to keep a notion of seme which is not very useful after all, the extended theory, for its part, even seems to ruin the notion of seme, since it shows that there may be no common semantic feature between the separate meanings of a single lexical unit. For all that it does not mean, as Georges Kleiber rightly noted, that prototype semantics challenges the principle of the ‘componentiality’ of the meaning of a word […and] the validity of a componential analysis” (adapted from Kleiber, 1990, 67); because, he explains, “the typical approach is an alternative to necessary and sufficient features, but not to semantic features themselves (adapted from Kleiber, 1990, 68). But in spite of appearances, it may be possible to keep more than the sole principle of componential analysis, and even to find a kind of synthesis between prototype semantics and structural semantics and this, it seems to us, under various conditions. It would be advisable to redefine the concepts of semantic analysis, first by taking the linguistic sign theory into account, then by making sure not to transpose too much the phonological pattern, finally by bearing in mind that prototype theory is above all a psychological or a cognitive theory.

If, on the basis of extended prototype theory, we define the content (or signified) of a morpheme as a set of semantic definitions connected to one another at least two by two, content that is associated with a given single expression (or signifier) or rather with a single class of allomorphs in complementary distribution, then we establish a certain imbalance within the linguistic sign while admitting that the expression plays a major and decisive role. It is indeed almost the one that allows to determine the morpheme, its range of meanings having no foundation stone, it begins nowhere and, having no reason to stop, it may go on endlessly. But then, to what extent do we still have the right to say that morphemes have only one content, if the set of its separate semantic definitions this content corresponds to, is virtually infinite and finally has nothing common? From this point of view, the stance of standard prototype theory would be more satisfying, insofar as it admits a semantic nucleus common to all the separate semantic definitions, which allows to delimit the morpheme, at the meaning level, and would really give us the right to use the term content opposite to the expression of this morpheme.

In other respects, if it is unquestionable that the two meanings of the French lexical unit femme (“1. woman”, “2. wife”) or the two meanings of the Latin lexical unit sentire or the five meanings of the lexical unit jeu (“1. game”, “2. play”) correspond to different logical categories and to different concepts, for all that it does not imply that linguistically they are different, and that even if the content of morphemes, must be conceived, as we think, not as a concept, but as something conceptual. Nothing indeed prevents one from thinking that a single linguistic content might allow the listener to build up and the speaker to suggest different concepts, which will be, of course, inevitably related to one another. Because of that, we must not conceive this content as a kind of semantic invariant which would be found in all the actual meanings of the lexical unit, as would wish both standard prototype theory and structural semantics. The content must be considered as a common semantic nucleus from which all the actual meanings of the lexical unit are built up and this, thanks to precise rules of elaboration of meaning such as semantic narrowing, semantic broadening, metonymy and metaphor. Such a connection between the specific meanings of a lexical unit with a basic semantic nucleus may be direct or indirect. This means that from a single content we can draw a range of primary meanings directly derived from the said content, and that from these primary meanings we can draw sub-meanings derived from primary meanings and not from the basic content. And why not say that sub-sub-meanings are derived from sub-meanings of lexical units? There again, we are faced with Wittgenstein’s family resemblance theory, but not with a common basis, but with an identical basis, and which is in a way the core of the network branched out into related meanings that may take the linguistic element. This being established, what prevents one from claiming that this semantic basis may and must be defined in terms of necessary and sufficient conditions, as is implicitly admitted by structural semantics, when it speaks about sememe, and not just in typical terms? Because a definition, even if it is a semantic definition, cannot be fuzzy and ill determined, or it is not a definition. On the contrary, it is certain that there are, at the psychological and linguistic levels, a great number of separate ways to use a single semantic nucleus, which may even be more or less rigorous, or even more or less fuzzy. And in this view, the various prototype theories, whether they are cognitive theories or their transpositions into the linguistic field, present the great interest of showing well that in psychology and semantics, things do not necessarily happen as in logic.

For that effort of synthesis, which proposed a less static and more subtle conception of sememe, not to seem too programmatic, it would be interesting to take up some of the examples about which it was sometimes spoken concerning the disappearance of a seme in some particular meaning of a lexical unit, and to show that this particular meaning, in spite of everything, rests on the allegedly deleted seme. Let us take the example of the word minute. It is certain that in the phrases:

Je serai de retour dans une minute (I’ll be back in a minute)

Cet automobiliste a eu une minute d’inattention (This motorist had a moment of inattention)

the lexical unit minute refers, in the first case, to a space of time that could last longer than and, in the second, less than “one sixtieth of a hour ” and that in this case it simply means “a short period of time”. But this semantic feature “short” is not there by chance. It is there because it is adapted to the extralinguistic situations it is applied to and to the semantic contexts where it occurs. It is not a new seme, namely a distinctive semantic feature, it is just a semantic feature built up from the seme “one sixtieth of an hour”, in view of the situations designate. Therefore this semantic feature is only a metamorphosis of the seme “one sixtieth of an hour” itself. Where Robert Martin said: the seme “s¹₁ implies the seme ‘short’ ” (Martin, 1983, 66), it would be better to say: the seme “one sixtieth of an hour” implies and develops the semantic feature “short”, which amounts to saying that the broadening of meaning of minute is built up from the semantic definition corresponding to the so-called primary meaning of this word, which can be considered as its sememe. What Robert Martin terms sememe S¹ is thus purely and simply the sememe of the lexeme minute nothing. As for the alleged sememe S², it is only a semantic impact of this sememe, which gives a particular realization (traditionally called broadening of meaning) to one of the semes of this sememe.

We could support the same view concerning the alleged metaphorical meaning of a lexical unit. Admittedly in:

Achille est un lion. (“Achilles is a lion”)

the meaning of the lexical unit lion does not correspond to its sememe, since Achilles is in no way a “large wild carnivorous mammal, having a tawning yellow coat with a shaggy mane”. But this sememe is not purely and simply deleted for all that, since it is from its semes - and not among its semes - that it is possible to discover or imagine the particular semantic features it would be proper, in this context, to associate with the use of this word. The listener must indeed try to find, among all the characteristics that possesses the animal the word lion refers to, the one that is attributed to Achilles by the speaker. And this agrees with our reformulation of the definition proposed by Jean Molino, Françoise Douay-Soublin and Joëlle Gardes-Tamine, according to which metaphor would essentially be defined as the possibility to find new common semantic features. But we will however note that in this metaphorical use, the lexical unit lion does not only receive the meaning of “courageous person”. The sentence indeed does not mean that Achilles has courage, but that he has the courage that a lion may have, which is more suggestive and revealing than if it was simply said:

Achille est un homme courageux. (“Achilles is a courageous man”)

We can see to what extent the metaphorical meaning of a lexical unit, far from invalidating the concept of sememe, is finally not really explainable without the notion of sememe. Indeed if this word did not have a content which allowed to identify a certain type of referent, it would be impossible to give it a meaning when its context of use manifestly prevents it from designating this expected referent. That is why the definition of metaphor proposed by Michel Le Guern seems so interesting to us. Indeed, according to him the metaphorical meaning does not correspond to the deletion of a part or of all of the semes of a lexical unit, but rather to their “parenthesising”. This expression is even not yet positive enough, for the semes apparently parenthesised in this way do not stop for all that to play a role in the interpretation of the lexical unit or to be present in the very result of the interpretation. Perhaps could we speak about a backgrounding or better still about a surpassing of the sememe?

To conclude is it possible to clarify the sememe of “jeu” and to give a fairly basic explanation of the main lines of polysemy? We would be tempted to propose as sememe the essential elements dictionaries usually consider as the primary meaning of this word, namely:

“Activity aiming at the distraction of anybody playing a part in it”.

This wording allows developments concerning pleasure (cf. un élève qui ne pense qu’au jeu; a schoolboy who only thinks of playing), gratuitousness (cf. les jeux de l’imagination, games of imagination), the absence of seriousness (cf. ce n’est qu’un jeu; it’s just a game), self interest or profit (cf. les jeux d’argent; gambling), namely in fact the first two major meanings of the DFC. From that sememe we might also derive, drawing our inspiration from the organisation of the meanings of Le Nouveau Petit Robert rather than from that of the DFC, the following three meanings:

“system of rules allowing this play activity” (cf. tricher au jeu; to cheat at a game)

“equipment needed for this play activity” (cf. jeu de ballon; ball set, jeu d’échecs; chess set)

“way of practising this play activity” and metaphorically “way of practising an activity implicitly assimilated with an art” (cf. le jeu d’un acteur; the play of an actor, être pris à son propre jeu; to be caught out at one’s game)

And from this last meaning, when it is applied to an object and not yet to a person, we come to the fifth meaning of the DFC

“easy and regular functioning”(cf. le jeu des pistons; the play of the pistons)

where we obviously recognize the notion of activity, and that of activity which conforms to precise rules, as when a game normally unfolds. It is then justified to speak about family resemblance concerning the major meanings of the word jeu. However we can see that it is also possible to associate them, as dictionaries do, with a basic meaning we would like to consider as the sememe of this lexical unit, namely the semantic nucleus from which all the other meanings of the word are built up. There is no doubt that the major meanings just mentioned do not cover all the possible meanings of the word jeu, but it seems to us that these other meanings, including those of such phrases as:

un jeu de mots (a play on words); d’entrée de jeu (from the outset); jeux de physionomie, (facial expressions); jeu de clefs (bunch of keys); entrer dans le jeu de quelqu’un (to play somebody’s game)…

may all, as dictionaries try to show, be associated with one of the major meanings by a new broadening of metaphor at a second level. Thus while even more wandering from the primary meaning, which consists, for its part, of all the meanings of the sememe, and making appear new semantic features that are often completely different from the semes of this basic meaning, the meanings derived from meanings themselves derived are not less indirectly linked with the sememe of the primary meaning by a chain of semantic associations or resemblances. And it is not unthinkable to say, as a conclusion, that the sememe is the distinctive semantic definition from which are built up directly or indirectly or on which are based all the particular meanings which, in view of the speech situation, the linguistic context and the situation involved, the linguistic sign may have in its various meanings.