François RASTIER

(Pubished in Andreas Blank and Peter Koch (eds.), Historical Semantics and Cognition,  
Coll. Cognitive Linguistics Research, Berlin: Mouton de Gruyter, 1999, p. 109-144)

For the centennial of the publication
of Bréal's Essai de sémantique

Besides from the fact that structural semantics and cognitive semantics are not unified movements, they remain difficult to compare because their objects and objectives — none of which have been reached yet — are somehow different. In the field of diachrony, the contributions of structural semantics have been, to tell the truth, notorious for a long time (see Coseriu 1964) and if the cognitive issue managed to stimulate it (see the works of Koch 1995 and Blank 1997), it remains that diachronic cognitive semantics (Sweetser 1990, for instance) has not or not yet shown any visible theoretic or practical progress: most of the time, it does not do much more than retrace well known problems (see Nyckees 1997).

1. The panchronic problem of qualitative inequalities

At the beginning of the thirties, Hjelmslev set the problem of qualitative inequalities among linguistic categories by distinguishing intense terms and extense terms [1]. The main merit of the cognitive theory of prototypes will have been to restate this problem among lexical classes. Yet, the concept of prototype does not by itself account for these inequalities. Indeed, it introduces two types of fuzziness: one through its own definition which varies from one author to the other, and the other from its use to describe matters of graduality. Moreover, it must be founded [2]. Why does a given term become prototypical or ceases to be such? This question is never asked and nothing in the theory of prototypes gives us a way of knowing how prototypes are born, grow and disappear.

Hjelmslev's problem of qualitative inequality of lexical units inside the taxeme must be considered in synchrony as well as in diachrony in a panchronic perspective taking into account the structure of lexical classes. Whereas in a synchronic description, discrete representations are preferred, continuous representations with thresholds are necessary in diachrony. The panchronic perspective thus requires to articulate two kinds of representations: discrete and continuous. If in synchrony the relations inside lexical classes can be characterised by discrete semic oppositions, the gradual nature of diachronic evolution can be represented by dynamic models which, without contradicting the semic analysis, identify sememes as characteristic zones within dynamic evolutions (see infra, 2. 2).

1.1. Some weak points of cognitive semasiology

1.1.1. The problem of semantic classes

The main shortcoming of cognitive semantics is the weakness of its theory of semantic or conceptual classes. This is firstly due to the vagueness of Rosch's concept of category, to the theorization of this vagueness (see the cue validity) and to the absence of a defined point of view on its constitution: from which point of view, according to which system of relevance can a specimen be considered deviant? Rosch's take-over consists in centering the inequalities of the category around the prototype, furthermore postulating that there is one and only prototype per category [3].

Correlatively, one must recall the weakness of the concept of domain in cognitive semantics. Langacker for instance sustains that “semantic structures […] are relative to "cognitive domains", and that a domain “can be any sort of conceptualization: a perceptual experience, a concept, a conceptual complex, an elaborate knowledge system, and so forth” (1986: 4). How can one solve this vagueness?

The use of the theory of typicality in diachronic lexical semantics would therefore request some re-elaborations, especially of the concept of category, an undefined concept which certainly is only an experimental artefact of cognitive psychology. The concepts of taxeme (Pottier 1974) and domain (Coseriu 1968) taken from structural semantics seem to us to be better defined and more operative. Especially, they are not founded on an ontology like Rosch's concept of category (Rosch 1978) nor on the illusion of a perceptive naturality [4], as for Berlin and Kay (1969) ; one can hence conceive their evolution inside a culture and a history.

The question of lexical classes cannot be solved, nor even asked, in a semasiological perspective. Indeed, the inventory of all acceptions of a lexeme or grammatical morpheme is not a semantic class, since it has no other common principle than the identity of its signifiers (contingent criteria based on expression and not on content).

1.1.2. The semasiological method and onomasiology

Cognitive psychology and cognitive semantics diverge precisely on a crucial point of methodology. Whereas for Rosch and the psychologists who came after her categories are classes of concepts (and their putative corresponding objects), the study of which should be submitted to an onomasiological method [5], for cognitive semantics they are classes of acceptions which are studied through a semasiological method. Taking a signifier as invariant, it tries to cross-define its various acceptions, as formerly Katz and Fodor (1963) did with the word bachelor, cross-defining daringly male sea lions, knights, unmarried men and students. It thus neglects that these acceptions do not have the same history, are not generally found in the same discourses, nor the same genres, nor the same contexts and that they therefore are distinct linguistic units.

This seems paradoxical: the onomasiological method of structural semantics (and, with due allowances, of Rosch's theory of concept), in so far that she starts from the signifier to discriminate the synchronic varieties and diachronic varieties of the signifier, is closer to a “conceptual” approach than cognitive semantics which nevertheless claims such an approach.

But the semasiological method, the most traditional one can find, is only descriptive and not explicative. It can only rest upon the prelinguistic prejudice, born from the philosophy of language, that to a word corresponds a signified; and, as this is obviously not the case, one must find for it a preferential signified, or more precisely a basic conceptualization (counterpart to the literal sense in vericonditional semantics). Cognitive semantics therefore postulates an identity or a partial semantic equivalence between the different meanings of the word and tries to reduce polysemy by organising the acceptions around one acception considered as central, i.e. the prototype – without even thinking that polysemy is an artefact of a sign-based linguistics of. It is, as always, a way to weaken semantics in order to reinforce ontology [6].

Nevertheless, the semasiological approach, as it is usually applied in cognitive semantics, has three characteristics which make it hardly compatible with a diachronic approach: (i) the list of acceptions of a lexeme or grammatical morpheme is considered achronic. (ii) The structurally central acception or prototype is defined independently of any diachronic consideration (in Lakoff 1987 : app.2 for over or Langacker 1986 : 3 for ring). (iii) The only temporality retained is the one, internal and ideal, of cognitive operations: it is abstract and anhistorical, as it is represented in a transcendental space [7].

By declaring that the history of language is a series of distributions, Bréal had introduced a diachronic structural principle [8], which Saussure transposes in synchrony in the theory of value. This differential principle will precisely enable the constitution of semic analysis and its use to account for the different phases of evolution. The semic analysis is then considered in context, since the new meanings are described from the new contexts (a semic analysis which would reduce itself to analysing words out of context would be illegitimate because normative). Yet, if the polysemy of acceptions, contingent fact in synchrony, can be enlightened by diachronic studies which show how one managed to pass from one acception to another, this does not entail that it should be the preferential object of diachronic lexicology.

1.1.3. The factors of diachronic changes

Factors are not causes, and the description generally limits itself, and wisely so, to the how of evolution; but we shall see further that one cannot avoid the question of why.

a) The four operations. — Since the beginnings of lexical semantics, four operations have persistently qualified changes in meaning: extension and restriction, metaphor and metonymy. These very operations which are used to describe the diachronic relations (for instance in Darmesteter) are also used to describe the synchronic relations between acceptions from Reisig (1839) to Clédat (1895) up to Robert Martin (1992). They are furthermore linked to a semasiological perspective, because they link two states (synchronic and diachronic) of a signified and they take the signifier as invariant.

They can be grouped by two: extension and restriction are logical concepts which belong to the classical theory of ideas [9]; metaphor and metonymy belong to the restricted rhetoric, theorised by the grammarians of the 18th century (Dumarsais 1730 in particular), precisely to save the literal sense and with it the own identity of concepts.

(i) Logico-referential operations. — Extension and restriction belong to referential semantics [10] which is applied to objects according to an unverifiable principle of quantitative variation very well-known under the name of law of Port-Royal. These operations rest upon (i) a discrete ontology, in such a way that the designated objects can be counted and (ii) a logic of classes, in such a way that relations of inclusion can be described.

(ii) Tropes — It is known that the use of tropes in diachronic semantics, even when it claims to have its roots in Lakoff (Sweetser), is the most traditional thing one can find: it pre-exists even to Bréal's semantics, since it takes an important place in Darmesteter (1877), and already earlier in Reisig (1839). The description through tropes supposes a deviation, or at least a variation between the literal sense and the figurative sense. Yet, except if we repeat Saint Paul, one cannot pretend that the ancient meaning is literal and the new one figurative, nor a fortiori that the evolution of languages departs from nature, as Dumarsais regretted in a famous page about catachresis [11].

The use of rhetorical concepts in diachronic semantics is not self-evident because tropes are used there to name relations that must still be described and explained. Moreover, if one accepts that the trope is a textual form, the relations which are in this way decontextualized are not tropes, strictly speaking. We consider metaphor and metonymy as critical points of semantic forms, but it is not enough to name them to consider them described [12]. What is the difference between the units they unite? In other words, what is the metaphorical or metonymical orientation? What motivates these paths? To answer these questions, we shall study hereunder the relative evaluations of the terms matched, especially their respective position in relation to the evaluative thresholds inside the taxeme.

b) Two evolutions. — Studying the relative valuation of the units matched seems to be for us a means to set the problem of evolution as a whole without resorting to logics or rhetorics. In a study on paragons (1991: 198-202), we have extracted a law of panchronic valuation which accounts for two complementary evolutions:

(i) The evolution through extension goes from the higher valued term to the lower ones : pecunia(Latin for cattle) has extended to mean 'richness'.

(ii) In the same way, restriction goes from the less valued to the most one: frumentum, which in Latin meant 'cereal', became froment (wheat) in French (the most valued cereal); a form of this valuation remains in the familiar acception of blé (dough). Viande(with the generic meaning of food in Old and Middle French) “becomes” viande (meat) to designate food par excellence [13]. This applies also to synchrony (diatopical): in Marseille, for instance, you can hear J'ai un enfant et deux filles (I have a child and two girls), enfant being restricted to boy, eminently valorized in this Mediterranean city.

Qualitative inequalities between marked terms (or paragons) and unmarked terms are thus linked to two principles of distribution and panchronic summation. Extension is hence a distribution of the positive evaluation from the paragon, and restriction the summation towards the paragon. The law of panchronic valuation expresses in this way the relations between intense and extense zones, the restriction towards the highly valued consisting in a passage from the extense zone to the intense zone, and the extension from the highly valued bringing about the opposite movement [14].

Figure 1  Law of panchronical valuation

Examples : synchronic restriction, see enfant ; diachronic restriction, see viande ; synchronic extension, see gagner son bifteck “to own one’s living”; diachronic extension, see egeria, hoover.

Let's return to our four diachronic operations.

(i) Extension and restriction. — We can now reverse in terms of valuation the law of Port-Royal (which nevertheless is only a law of quantitative and not qualitative variation): the more an intension is valued, the more extensions it has.

(ii) Tropes. — If we agree on this point, metonymy (and synecdoche which it is often confused) does not obey to other rules : it is from this point of view only a particular case of extension. It is the most priced part or associated unit which extends its name to the whole or the set. The transformation of proper names into "common nouns", extremely widespread and witnessed everywhere, goes in the same direction: after egerias and Hercules, Hoovers and PCs are everyday examples. Admittedly, these valued terms become unmarked and get neutralized by their very use, and other words come and replace them in a never ending process without any direction — but not without showing some regularities.

The usage of the concept of metaphor answers a totally different problem, the one of changes in semantic domains. But, since taxemes are included in domains, any change of domain leads to an alteration of the taxeme. There was a time when, in old French ouailles (from the Latin ovis) meant 'ewes' in the domain //agriculture// and 'faithful companions' in the domain //religion//. In modern French, only the last acception survives. The evolution process can be described in the following way: the evangelical metaphor of the good shepherd introduces, among others, a picture of the domain //agriculture// into the domain //religion//. As every metaphor, it shows two effects: in the classeme, it virtualizes the generic seme and actualizes a new generic and afferent seme; in the semanteme, it reshuffles the hierarchy of attributes (see Rastier 1987: ch. VII).

2. The taxeme as semantic form

We have used the concept of semantic form according to our hypothesis that semantic perception [15] pertains to pattern recognition and not to computation.

2.1. Qualitative inequalities inside the taxeme

That qualitative inequalities inside lexical classes can be described through intensity and extensity (Hjelmslev) is a fact that we shall relate with the hypothesis of fundamental aesthetics : languages do not articulate descriptions (as the objectivist tradition would like it to do) but evaluations. In particular, qualitative inequalities inside lexical classes seemed to be linked to social evaluations which evolve in history. This is not surprising if we recall that in spite of their name natural languages are indeed cultural productions [16].

Structural semantics, or at least the impoverished image given of it by most manuals, presents nevertheless one deficiency: the absence of qualitative inequalities. Here are two ways to overcome it:

a) To distinguish evaluative zonesinside a taxeme (by drawing one's inspiration from descriptions given formerly by Coseriu 1968). It is a means to break with denotation as no metrics can distinguish the big from the huge or the cold from the icy.

b) To take account of diachronic variationsof the signified: (i) by variations of thresholds inside the taxeme, (ii) by changes of taxeme (iii) and finally by changes of domain.

2.2. Elements of a morphodynamic model

We shall use for this purpose a model taken from the theory of differential varieties and dynamic systems. From Thom's and Zeeman's works in particular, it has been applied to the domain of speech and case-frames by Petitot (1983, 1985), to semasiological semantics (on the polysemy of encore by Victorri and Fuchs 1996), to the structural analysis by Piotrowski (1997) [17].

Continuist modelization presents one important advantage: the gradual and the discrete can be described as particular cases of the continuous and not the reverse. Hence, without suggesting any strong hypothesis on the continuous character of the ‘semantic space’, neither on the spatial nature of cognitive schemes, we admit that the semantic discretization consists in isolating outstanding points on dynamics. We make the assumption that semantic evolution can be represented on a gradual process with thresholds: the discrete elements (like semes and phemes) result from the capture of discontinuities (see the phenomenon of categorial perception discovered by Liberman).

A morphodynamic model is characterized by the functional coupling of an external space (or substrate space or also space of control) with the internal states of a system S. The singularities of the internal space are projected as discontinuities on the external space.

Figure 2: Dynamic system S

As we have chosen an onomasiological perspective, we consider as semantic space the set of the sememes of a taxeme [18], for instance the one of levels of temperature: 'icy', 'cold', 'cool', 'lukewarm', 'hot', 'piping hot' [19].

Figure 3 : Dynamic representation of a taxeme

Regrettably, this representation ignores the context, because determining the main attractor and the position of qualitative and acceptability thresholds depends on the context (cool is neutral for a beer, not for a bath [20]).

The taxeme displays three main basins of attraction separated by two absolute maxima: we call doxal zonethe big basin of attraction delimited by two absolute maxima and located below the acceptability thresholds; and paradoxal zonesthe two zones located beyond these thresholds.

Each sememe corresponds then to one local attractor. All the various contexts attested for it define its basin of attraction (in synchrony). The bottom of the basin of a sememe corresponds to the zone of semantic stability, in other words to its meaning (prevailing acception, or in other words set of semes inherent to the sememe). The slopes of the basin, zones of instability, correspond to usages. The basin broadens or narrows as the number of contexts grows or diminishes. The marked or intense terms correspond to narrow basined attractors and the unmarked or extense terms to wide basined attractors.

The slope of the basins vary whether the sememe has more or less usages in the corpus. Two neighbouring sememes are separated by passes which altitude is also variable: they are low when the sememes count many similar contexts and high when they only count a few or none at all. Parasynonyms are separated by low passes which do not cross any differential threshold; the distinction, which is always possible, consists in lowering the closest differential threshold.

Figure 4: Discretization and marking

The differential principle of structural semantics expresses itself by a covariance of the basins of attraction of sememes. Any local distortion may have consequences on the global form and neighbouring sections of local forms [21]. Yet, as the global determines the local, even if an important local perturbation may modify the global organization, a taxeme shows a certain resistance to the distortions brought by the occurrence of new contexts.

The local evolution of the sememes inside a taxeme can be described by the modification of their basin of attraction and of the passes which separate this basin from its neighbours. The global evolution of a taxeme can be described as a merging or separation of basins of attraction in a general dynamic.

The proximity of sememes is expressed by the presence of their basins inside larger semantic zones (which we shall call evaluative zones: the main ones are the doxal and paradoxal zones). The semantic distance between sememes is expressed by high passes between their basins. When the basins are shallow and the passes low, the equivalence prevails on the opposition or, in other words, coactivation prevails on reciprocal inhibition; when the basins are deep and the passes high, inhibition prevails on coactivation.

When the pass which separates two sememes is low, they share common contexts (as is the case for vis and face, which in old French are present in the same contexts, as is still shown today by the partial equivalence of the phrases vis-à-vis and face-à-face). The high passes (local maxima) indicate qualitative thresholds: their slopes are in the same evaluative zone. The highest passes (absolute maxima) indicate acceptability thresholds: their slopes are located in different evaluative zones.

The thresholds are located at points of singularity: local minima and absolute maxima (which are transition points between opposed evaluative zones) [22].

As far as diachronic or synchronical evolutions are concerned, two opposed movements can be considered: to the extension corresponds a broadening of the attraction basin with a lowering of the point of stability, whereas to the restriction corresponds the downcoming of the highest qualitative threshold.

When a semantic zone contains several attractors, two outstanding attractors may represent the class, because every form is recognized by its singularities:

— The attractor closest to the highest relative maximum or to the superior qualitative threshold (corresponding to the paragon, ex. bifteck)

— The attractor furthest to the absolute maximum, which is also the one whose basin of attraction is the most widespread and/or the deepest because it accepts the highest number of contexts (we call it generic neutral, for instance rue).

To these two attractors correspond the two main and contradictory definitions of the prototype as defined by Rosch's theory.

Figure 5: Thresholds

The relationships between part-whole inside the global basin of the taxeme can also be described as relationships between these two outstanding points, in a way which does not have recourse to the concept of inclusion, but to the one of relationships between minima and maxima.

A "logical" representation can indeed norm and codify a taxeme but not represent the dynamic it has created nor the ones that make it evolve. According to the modelization we have put forward (1987), forces are inhibitions, activations and propagations. These operations suppose "potential differences" which are valuations.

2.3. Evolutive paths: endoxal and paradoxal

We distinguish two kinds of differential thresholds, qualitative thresholds and acceptability thresholds. They divide as we have seen the taxeme into two types of zones (doxal and paradoxal). Whether the evolution crosses or not the acceptability thresholds, it will be called doxal or paradoxal.

The paradoxal evolution is obviously complex, because it supposes the referral of two contradictory dynamics (which correspond to two opposed doxas): to cross an acceptability threshold in a dynamic A, the litigious content must be below the acceptability threshold in a dynamic B (See Chamfort [1795] 1968: 341 : "a little boy asked his mother for jam: 'Give me too much', he said to her").

One can further distinguish two kinds of doxal evolution, whether it crosses or not a qualitative threshold. Among the evolutions which cross a threshold, we shall finally distinguish between neutralizing paths and valuating paths.

We can then deal with a third type of evolution, which the law of panchronic valuation does not account for: the extension from the neutral term. This endoxal evolution starts from the middle or neutral terms with respect to evaluation (which are often the most frequent terms), to designate the whole taxeme. In this way, rue (street) can express in French all types of traffic lanes, avenues, walks (in which streets are included) [23]. It can be considered that the neutral term has the largest number of contexts (simply because it contains less specific attributes: it is therefore compatible even isotopic with a large number of sememes). Thus, the sememe which in the taxeme has the widest and most stable basin (coinciding with the minimum) may designate the whole taxeme.

Figure 6. Extension from the neutral term

The same endoxal evolution can take place several times in history. For instance passer in Latin (sparrow) becomes pajaro (bird) in Spanish, passereau in French (afr. passe(re) > passerel > passereau, passerine, which designates all kinds of small birds); in a renewed way, piaf, which in the first half of this century meant 'sparrow' has extended its meaning to 'bird' in familiar French.

Whereas the law of panchronic valuation expressed the relationships between extense and intense zones, this form of evolution expresses the relationships inside the extense zone: being established inside a same zone, they characterise themselves by the conservation of evaluation.

2.4. Force and form

The description of taxemes as sets of static relations has indeed to its credit simplicity and economy: these relations can be characterised as logical relations, and this brought quite a success to descriptions through semantic features (actually used by defenders as well as by opponents of structural semantics).

But the description of forms has no relevance if one cannot account for their evolution. Cognitive semantics acknowledges abstract forces, especially of course in its analyses of actancy when, following Talmy, it enacts a kind of transcendental mechanics which owes a lot to Aristotelian physics (see the concept of impetus in Talmy 1988). Forces and semantic forms must nevertheless be articulated. The morphodynamic description may account for forms in terms of forces. To say the least, these two aspects, form and force, are complementary: a force is felt and measured by the distortions it entails; a stabilized form results from an always momentary balance of forces.

The effect of forces can be understood in two ways: the movement of critical points; the concomitant distortion of the "normal" sections of the form. The sections that exceed a threshold gain a potential of disturbance. From this point of view, the edges of the evaluative zones are privileged spots of disturbance. Beyond a certain level, peripheral disturbances modify the main basin of the taxeme.

2.5. The conditions of form evolution

In morphodynamics, regular points and singular points are distinguished. As a form is better recognized by its singular points rather than its regular points, some of the relationships which are characterized, by analogy with perception, as relationships form / content can be described or reformulated as relationships between the regular sections of the form and its singular sections. For instance, at the textual level, we have described isotopies as products of the Gestaltist law of good continuation: they are as such regular portions of textual forms, and appear then as semantic backgrouds. On the contrary, allotopics are singular points and certain tropes introduce qualitative discontinuities by a disruption of isotopy [24].

At the lexical level, the isotopic contexts of a sememe keep a regularity, whereas allotopic contexts create singularities which may sustainably disturb the basin of attraction of the sememe. This point must be qualified according to the typology of sememes: sparse with low associated values sememes are compatible with a large number of contexts. For instance the one of grammatical morphemes, which, as a general rule, does not present any generic feature of domain, can show a great deal of isotopic contexts — and this is the reason why the same grammatical morphemes are found in all discourses : hence the diachronic evolution of grammatical morphemes is slower that the one of lexemes.

2.6. Value and values

Let's admit that forms (here semantic forms) are inhibited movements. As a normed set of evaluations, a doxa is precisely composed of prescriptions and inhibitions which ensure a synchronic or diachronic stability to semantic configurations. We retrieve here Barthes's intuition that the lexicon is a frozen doxa (1984 : 129) but we pluralize it: frozen doxas. For instance in French there is no taxeme of levels of temperature: the sememes expressing levels of temperature are organized in different taxemes according to contexts. The same goes for sizes: 5 ft 10 is considered normal for a man, tall for a woman, and giant for a child. This size is hence respectively below and beyond the qualitative threshold, then beyond the acceptability threshold.

Whereas cognitive semantics, after having reinvented the Kantian oversimplicity, is looking for the descriptive categories around transcendental aesthetics as a prior frame to any perception, we have used the term aestheticsin a more restricted meaning [25]. Therefore the historic project of diachronic semantics leads us rather to look amongst social evaluations to find the forces which shape and distort the lexicon. Thus only the structural principles of semantic organization pertain to language; relevant categories pertain to the specific language considered; but the particular organization of its configurations pertains to evaluative norms which are subject to variations. The changes in evaluation resulting from a crossing of a qualitative threshold or mainly an acceptability threshold introduce modifications that can be both fast and long lasting.

We clearly understand that everything depends upon the position of the evaluative thresholds (whether they are qualitative or acceptability thresholds). According to the position of these thresholds, the structure of the taxeme changes; and this position does not depend on language but on doxa. Therefore, every stabilized semic formula results from an undefined series of contextual relations, i.e. from an interpretative tradition which varies with discourses and social practices.

One cannot talk about doxas, as systems of valuation, without mentioning the social values which they reify. Without trying to play with words, one can wonder if the value (relevant linguistic difference) is not ultimately founded upon the values, in the social sense of the term, including the ethic and aesthetic judgements [26]. The law of distribution and the absence of perfect synonyms enable us to outline a positive answer and to second it with examples: an icterus is more posh than a jaundice, ethylism (especially if it is social) than alcoholism, etc. In this case, interpreting an occurrence is not or is not anymore a simple operation of relating it to a type, but of positioning it in a taxeme, and of locating it in one of the evaluative zones of this taxeme.

3. Implementation: the denominations of FACE in French

For our proposition of illustration, a study of the joint evolution of the words face and visage (figure) in French, we rely on Vaugelas (1647), F. Brunot (1905) and especially Renson (1962) [27].

3.1. The different steps of this evolution

According to Renson (1962: I, 227), face is first attested in the 12th century with the meaning of 'visage' (figure). In fact, it means essentially the surface of the face and in particular the cheeks, as it appears in contexts like cent foiz li baise de randon / Les ueuz la face et le menton “one hundred times he kisses at random / eyes, face and chin” (Roman de Thèbes, 6378). In fact, at the time, the plural faces with the meaning of 'cheeks' is attested [28]. This acception justifies the abundance of colour adjectives (vermaille, palie, etc.).

Yet, from the 13th to the 15th century, the number of contexts of face grows substantially, face meaning then, apart from the physical aspect of the figure, its expression (/moral/ feature): for instance in Greban doulce et tant prudente, benigne, or in Molinet ayreuse et furibonde, humble et fort accountable (Renson 1962 : I, 229). This evolution is general and also noticed for visage. It is evidence of the psychologization of literature (which in fact constitutes the best part of our documents).

But, as soon as the 12th century, as has been stressed by J. Trenel, a new class of contexts appears with the use of face in the translations of the Bible (with the meaning of presence: la face de Dieu, (which translates the Hebrew plural p’nim Elohim) and also of surface : la face de la terre (the face of the earth). The face of God is without any precise features (see Renson: I, 233). This new acception, translated literally from Hebrew, shows perhaps a relation with the spiritualization of the acception which describes the human figure: thus we find in Arnoul Greban adjectives like dampnee, digne and saintissime, or in Molinet angélique.

On the opposite, visage is more rarely found in religious contexts, and in the assumption of a reciprocal reinforcement of neighbouring semes, it can be said then that 'face' shows the afferent feature /religious/, brought to the forefront in certain contexts.

It remains that in the 16th century face presents two acceptions, for God and for the human beings, the first being ameliorative and the second neutral, liable to physical as well as moral uses, and sharing with visage many common contexts. The configuration is then as follows: 'face 2' (/human/) and 'visage' share a common general basin of attraction and are only separated by a low pass, whereas 'face 1' (/divine/) lies beyond.

Figure 7. End 15th.

Yet, at the end of the century appears the expression face du Grand Turc, which designates a totally different body part; thus Sieur Tabourot des Accords wrote : “son mari qui était tout nud sur le lict, avait la face du grand Turc tournée de ce côté là” (Escraignes dijonnaises, 42 v°) [her husband who was stark naked on the bed had the face of the Grand Turk turned that way]. Must we recall the establishment of diplomatic relations by François 1er with the Sublime Porte and suppose that the wars between Christendom and Islam could well make of this expression the ludicrous and discreetly blasphematory reverse of the divine face? Whatever the case may be, the word face itself, according to Ferdinand Brunot and in spite of Vendryes's reluctance to admit that this could be the reason of it, finds itself marked with pejoration. In 1627, Mlle de Gournay feared that the "new critics" might decide suddenly to refuse to “escrire face, [...] généralement refusée du nouveau jargon, parce qu’on dit la face du Grand Turc”(L’Ombre, 1627: 958, in Livet, 1895: 303) ["write face (...) generally rejected from the new lingo, because one says the face of the Grand Turk]. Twenty years later, Vaugelas already emphasized that one "would not any longer dare say face for visage except for certain expressions" (1647 : Préface, IX). Hence its sudden rarefaction according to Renson's counting: from 23% of the designations of figure in the 16th century, it drops to 3% in the 17th century [29].

Figure 8. End 16th.

Thanks to the afference of the /pejorative/ feature, the acception ‘face 3’ is now separated from ‘visage’ by an acceptability threshold. Now ‘visage’ covers the doxal zone, surrounded by two antipathetic and paradoxal acceptions of face, one religious and the other one infamous. This configuration is unstable, not by its form, but most certainly because 'face 1' and 'face 3' share the same signifier.

Figure 9. End 17th.

The reduction of the basin of face benefits then to figure, which, as soon as the middle of the 16th century, by a specialization of the meaning of 'exterior shape', was used for 'shape of the human face', and later, in the 17th century, was also used for the facial expression or look (mine, 1662), and came to replace visage and face in the current usage (see Brunot 1905 s.v.).

In today's French, figure has become a neutral and generic term, liable to the most diverse uses; face keeps its /pejorative/ feature, as shown by the insulting phrase face de... Finally, visage, reserved to the written language or to an elevated spoken style, has gained a /ameliorative/ feature (widely used in advertisements for cosmetic products). Visage keeps the feature /human/ but acquires a new acceptation (although already very common in Montaigne) as it is suitable for the aspect of very diverse objects (socialisme à visage humain, 1968), which anticipates perhaps, as formerly for 'face 1', its disappearance from the class of the denominations for human face.

Figure 10. 20th century

3.2. Discussion

3.2.1. About the unity of the taxeme

It can be objected that ‘face’ in face du Grand Turc belongs to another taxeme, the taxeme of the designations of the posterior. Yes, but 'face 2' remains a pejorative designation for visage: “Caesar believed that long and lean figures (visages) were real faces (faces) of conspirators” (Voltaire, letter to d'Argental, February 11th 1764) [30].

Furthermore, it could be objected that, strictly speaking, ‘face 1’ belongs to the religious domain. Fine, but the human visage is also designated there by face : the expression face-à-face (face to face) comes thus from religious texts describing the encounter of man and God.

Moreover, expressions like face d'abbé (face of a reverend) 'red and illuminated face' (Cotgrave 1611 ; Oudin c. 1640), face de carême 'pale and pallid face' (Panckoucke 1749), are already attested in the 17th century in Bois-Robert (Epîtres) or Racine (Les Plaideurs). These two antithetic acceptions of face persist in religious and moral contexts. Littré places side by side face de réprouvé (face of outcast), sinister and frightening physiognomy, and avoir une face de prédestiné (having the face of a predestined), ‘having a full, rosy and serene face’ [31].

To account for these opposed acceptions of face, the 17th century lexicographers used the concept of style. Richelet (1680), the first to relate the condemnation of face, said: "This word is still used in solemn and majestic poetry but not in gallant, lively poetry". Furetière (1690) claims on the contrary that "the word face is only used for figure in mockery for a face that is too thick or too large" (“le mot face pour visage ne se dit plus guère en ce sens qu’en raillerie d’un visage qui est trop gros ou trop large”). The Dictionnaire de l'Académie, in its second edition, will suggest a synthesis by distinguishing the "serious style when talking of God" and the "familiar style: une face réjouie, enluminée (a joyful, illuminated face)".

The theoretic problem raised by the notion of style should not be underestimated; in fact, we admit that a taxeme belongs to one and only one semantic domain. In so far as taxemes are classes of sememes which constitute a basis for term selection within a given practice, and where domains reflect different practices, the sememes of a given taxeme pertain to the same domain.

In today's French, 'face 1' remains in religious texts, since, in spite of the aggiornamenti, the diachrony of religious discourses does not follow the same evolution than the other discourses grouped under the enigmatic name of langue générale (general language) [32]. This diachronic autonomy of a discourse is not surprising: in medicine, face still means 'visage', or more precisely 'the front part of the head'. More generally, it can be pointed out that the different discourses and the different social practices that they reflect are moving around differentiated diachronies: the evolution of a language, and in particular of a lexicon, obeys to very various types of temporality.

The faces d’abbé and the faces de prédestinés, which looked very much alike by their florid complexion, have disappeared from today's French: no ameliorative contexts are found where face would mean 'visage'. Thus, in an expression like face d'ange (angel face), the pejorative feature of 'face' is propagated to ‘angel’ and not the ameliorative feature of ‘angel’ which contradicts the one of 'face': see e.g. Télérama about the trumpeter Chet Baker who died of an overdose: “La mort donna des ailes à une face d’ange” (death gave wings to an angel face). The disappearance of the ameliorative acception of face, which appeared in the 12th century with the translations of the Bible, is perhaps an indication of a secularization of society, where the terms taken from the Scriptures are not any longer a pledge of what the Academy called a "serious style". It indicates also that /pejorative/ has become an inherent feature of 'face 2'.

3.2.2. From syntagms to phrases

We must point out here a deficiency of the lexicographical approach: traditionally linked to an ontology of the concept and the isolated word, it deals only with simple lexical items and neglects complex lexical items. Yet, in the religious discourse, naturally prone to formulas through its links with the ritual, complex lexical items like face de Dieu are particularly important.

In fact, for textual linguistics (and I believe that lexicology should be founded on such linguistics), the minimal semantic unit is the syntagm. Words are units deprived of their preferential contexts, decontexualized artefacts both of ontology and lexicography. In this perspective, simple lexical items are defined as strongly integrated syntagms (fact actually confirmed by linguistic evolution: each word is a fragment of a myth). Yet, the diachronic status of syntagms and complex lexical items are different:

— A non integrated syntagm like face du Grand Turc can be an initiator: semantic innovation can create a local disruption, which, amplified by other factors, will lead to a revision of the taxeme (creation or deletion of an attractor, or simple distortion of the global basin).

— Integrated to a complex lexical item, like face-à-face, a simple lexical item does not retain the properties of its original taxeme: its sememe is redefined in a new taxeme and is not revised by the disruptions of the original taxeme. Thus, as Vaugelas pointed out: “Pour les personnes, on dit encore regarder en face, reprocher en face, soustenir en face, résister en face mais toujours sans l’article la” (1647 : 60 [for people, we still say regarder en face, reprocher en face, soustenir en face, résister en facebut always without the definite article la ]). And indeed, face-à-face has survived without any ameliorative feature (although it was drawn from translations of the Bible), nor subsequent pejorative feature; just like vis-à-vis, although vis owes probably its obsolescence to some form of ostracism if not some taboo.

3.2.3. The two networks of content and expression

If the norms of the doxa may account for the evolution of the signifieds, they also participate in the evaluation of the signifiers, and the signs find themselves caught into two networks: (i) an onomasiological network which accounts for the evolution of the signs according to the evaluative variations of the signifieds; (ii) a semasiological network which takes into account the evaluative variations of the signifiers. The evolution of a sign obeys to these two networks of constraints.

Yet, and this is worth being noticed, the same evaluative forces are at work in the onomasiological and semasiological networks: in the first case, we have seen that the proscription concerning the face of God has extended to the human face. For the second, let us consider the case of a prohibition concerning the signifier and let's take the example of vis which belonged in old French to the designations of the visage. The dropping of the last consonant has transformed it into a homonym of vit. (prick).

The homophony with vitwas obviously not lost on facetious minds [33]. The opposition between noble and evil parts of the body was sufficient to inhibit vis in favour of visage [34]. Thus the same kind of prohibition may contribute to inhibit a signifier (vis) or a signified ('face 2'). And in both cases, the controversial sign is eliminated from the taxeme.

4. Conclusion

4.1. The historic contingence and the ideological conditions

Let us imagine that the hilarious Sieur Tabourot, being or not the inventor of it, had never published this face du Grand Turc, then the designations of visage might perhaps have been changed. Without returning to Plekhanov's conclusions on the role of the individual in history, it is enough to extend to languages the well known fact in the dialectology of small communities that a lexical innovation can always be attributed to one person. In the written languages, it is not rare that literature innovates and that its innovations are taken up in the oral language (see in Chinese the four character expressions).

Nevertheless, for an innovation to be taken up, it must use salient semantic categories: it is useless to recall, at the end of the 16th century, the disturbance caused by the wars of religion, the expansion of scepticism and philosophical dissoluteness. During the following century, the censorship of the social, pious or academic good taste, contributed, as we know, to normalize the language and to put an end to semantic ambiguities: hence face was sacrificed on the altar of decency.

4.2. The anthropological foundation of diachronic evolutions 

We have seen that the evolution of taxemes can be described as a succession of distortions. Among the forces that are instrumental in it, ameliorations and pejorations play an antithetic role. The more a content is highly valued, the narrower its basin and the more it can be distorted. The zones neighbouring the acceptability thresholds are unstable, the pejoration of a very valorized term (as was the case for face) is enough to disturb a complete arrangement. Thus, the dogmatic value that was attached to face enabled to transform a new context into a blasphemy (as if faith was closer to blasphemy than indifference).

More generally, as doxal contradictions correspond to social contradictions, a conflict between groups can express itself by a conflict between doxas, and a paradiastole is always possible (example: your heroes are murderers).

We can nevertheless retain antithetic evolutions. Valuation is expressed through: an activation, the broadening of the attraction basin (and therefore of the contexts), the acquisition of a typical value, the use in an elevated style. On the opposite, taboo, prohibition are expressed by: the active inhibition of the singularity and of the related points, the narrowing of the attraction basin, the quantitative rarefaction of form, and often a euphemistical shift and the use of a sloppy style [35].

One wonders then, in the continuation of what we said about value and values, if the founding of semantic oppositions does not lie in the doxal opposition between the valorized and the devalorized, and, ultimately, between the prohibited and the prescribed. If this is the case, semantics, be it cognitive or not, can only have an anthropological foundation, articulated upon ethnology and history.


[1] Cf. e.g. 1971: 164.

[2] Unless you believe, like Rosch following Berlin and Kay, that prototypes are founded by nature. Surprisingly, for these authors, nature is achronic (and yet, for instance, the mammoth is not any longer a prototype of big game animal).

[3] One can indeed wonder if her insistence on fuzzy membership does not come from an insufficient definition of the categories (which do not correspond to any lexical classes listed by linguists).

[4] In fact, nature is only a naturalization, i.e. a naturalization of doxa or the ideology of common sense.

[5] One knows that semasiology takes the signifier as invariant and considers the problem of polysemy as fundamental, whereas onomasiology starts from the signified and considers synonymy as prevalent. From the thesis that synonyms exist in any language, advanced by Abbé Girard (after Prodicos) and pursued by the synonymists of the 18th century, was born today's differential semantics (see n. 9).

[6] This ontological conception of meaning can be found in Guillaume with the theory of the potential signified, as in the theory of prototypes. The main discordance between Guillaume's approach and the cognitive approach lies in the fact that the potential signified is an abstract form, whereas for cognitivist semanticists (furthermore as opposed to Rosch), it is a better specimen.

[7] One knows that the transcendental theories of meaning, like Guillaume's, Culioli's which was born from it and, lastly, cognitive semantics, describe a time that it not the time of history, whereas diachronic semantics is a historical discipline.

[8] The concept of value explains as well the law of distribution, which Bréal, developing in a historical perspective the synonymists' findings defines as follows: “We call distribution the intentional order as a result of which words which should be synonymous and which were such in fact, have nevertheless taken different meanings and cannot any longer be used one for the other ” (1897: 22). He concludes: “ The history of language is a series of distributions ”

[9] With which the semic analysis bears incidentally some relation. See Bréal about “our fathers from the Condillac School” (1897: 277).

[10] And for us the reference is an effect, not a starting point (see the author 1991: ch. VII, 1994: ch. II).

[11] Anxious to found in nature the linguistic evolution, Sweetser makes her own the legend, long ago cleared up by Meillet and Benveniste, that the concrete meaning comes first in comparison to the abstract meanings.

[12] Metaphor and metonymy, emblems of the restricted rhetoric since Jakobson has strangely enough coupled them, are neither symmetric nor converse; and their choice is furthermore problematic: why exclude the other tropes?

[13] To the contrary and complementarily, as we have just seen, bifteck came to mean ‘food’ by extension from the valorized form.

[14] This law has perhaps anthropological foundations; at least according to Louis Dumont, the hierarchy subordinates the includer to the included, the extense term to the intense term, the unmarked to the marked (see 1992: 11-24).

[15] We consider language not as the reflection or the transposition of perceptive forms, but as an object of perception (hence our proposals for a theory of semantic perception): the link between language and perception lies here. We leave open the question of the incidence of semantic perception on the other forms of perception.

[16] The diachronic structural semantics and the cognitive semantics, in so far as they take into account general conditions of the capture of meaning and find probably a common foundation in Husserl's phenomenology (especially the concept of qualitative discontinuity presented in the third Recherches logiques: 1901), will perhaps give us a way to articulate the cognitive (as transcendental) and the cultural by taking into account cultural and historical conditions of cognition as languages force and articulate it.

[17] Due to lack of space, we refer you to his excellent presentation, 1997, pp. 167-210.

[18] And not the whole series of contexts associated to one expression (as Victorri 1996 did for instance in a semasiological perspective).

[19] Representational theories of meaning do not succeed in accounting for qualitative inequalities. Yet, distinguishing evaluative zones inside the taxeme gives us a means to break with the representational theory of meaning, since no metric enables us to distinguish the big from the huge and the cold from the icy.

[20] See the author 1996. Among the levels of acidity of a wine (weak, soft, cool, vivid, nervous, sour, green), the same level, measured in pH, and expressible in abstracto by nervous, will be categorised as vivid (non pejorative) for a young wine, and by sour (pejorative) for a wine that aged.

[21] In traditional semic analysis, for instance, the introduction of a sememe in a taxeme modifies the semic composition, i.e. the relational structure of the other sememes.

[22] The problem raised by the sememes which would mean one thing and its opposite (as the famousaddad of the Arabic tradition) is in our opinion only raised in cases of antiphrasis or syllepsis, and therefore in context.

[23] Thus the Annuaire des rues de Paris(Directory of Parisian streets) contains not only streets but also boulevards, avenues, courts, etc.

[24] In this way, we reformulate, without linking it to any zero level, the problem of the deviation which preoccupies stylistics.

[25] The human universe is not made of knowledge on the one hand and emotions on the other hand. This omnipresent distinction, up to nowadays cognitive sciences, reiterates without any foundation the archaic separation of heart and reason. Undoubtedly, the neutrality of information is only a modern artefact which agrees with the persistent prejudice that language is a simple ideographic instrument for the use of rational thought. Let's agree that the human universe is made of social and individual appreciations, which are the object of fundamental aesthetics.
Fundamental aesthetics pertains to linguistics when it takes as object the linguistic material itself. At the morphological level, all languages contain appraising morphemes (see for instance the affix -acci- in Italian). At the immediately superior level, the lexicon of languages swarms with evaluations, and acceptability thresholds structure the elementary lexical classes. All the more for phraseological units, very numerous in any text, which reflect and propagate a social doxa. At the level of the sentence, it can be considered that any predication is an evaluation. At the textual level at last, narrative analysis for instance has underlined many times the importance of modalities qualified as thymic. To put it shortly, fundamental aesthetics defines the semiotic substratum on which arts of language are built, and remains well short of philosophical aesthetics. It has nothing in common with any aesthetic, poetic or stylistic function.

[26] This question may take Saussure as an authority when he defines two indissoluble aspects of value: the internal value, which draws its principle from differential semantics, and the external value, for which he gives the example of the coin. This counter value remains metaphorical, but the articulation of the two types of value raises a problem (see Piotrowski, 1997): it is in our opinion the correlation between linguistic valuations and social values (of which the economical exchange values are only an isolated case, exemplary because normed).

[27] My thanks to Evelyne Bourion for having passed to me the elements of this document.

[28] Example : “D’andeus ses oiz ses faces moille” [Both his eyes wet his face] (Benoit, Ducs de Normandie, 5114 ; six similar examples are found in this piece) to be compared with “Plure de ses oils, si li moille sa face” [his eyes weep and soak his face] (Chanson de Guillaume, 478).

[29] One example : Corneille still uses it in the sense of 'visage' in Médée (1635), but uses it only from then on in religious contexts.
Vaugelas may find that "usage put it out of use" (1647: Préface, IX) for ridiculous, extravagant and very insulting reasons. He recognizes that “qu’en même temps que je condamne la raison pour laquelle on nous a osté ce mot dans cette signification, je ne laisse pas de m’en abstenir” (whereas I condemn the reason why this word has been taken away from us with this meaning, I can only abstain from using it).
The first edition of the Dictionnaire de l'Académie (1694), already behind usage, simply defines face by visage; but, if the fifth edition (1798) confirms this, it adds that “Dans le sérieux, il ne se dit en ce sens qu'en parlant de Dieu” (seriously, it is only said with this meaning when referring to God).

[30] Even if the comparison with the posterior remains possible, in Zola for instance where the eaters in L’Assommoir “avaient des faces pareilles à des derrières” (ch. VII [had faces like bottoms]).
We make sure to distinguish the metaphor from its expression: it would have been preferable to associate the word visage to a comparison of the face and the bottom. Renson quotes l’autre visage in Voiture and in various authors luckily forgotten visage sans nez, gros visage, visage à rendre un lavement (1962 : I, 210 [face without nose, big face, face to bring up a lavement]). But these occurrences remain isolated and were not adopted by usage.

[31] The opposition of these two expressions can already be found in the Dictionnaire de l'Académie (1835). Victor Hugo unites brilliantly the two acceptions: “Le voyage qu’ils [mes parents morts] font est profond et sans bornes, / On le fait à pas lents, parmi des faces mornes, / Et nous le ferons tous” (Hugo, Feuilles d’automne, 6). In this context, 'face 2' appears in the religious domain (a funeral).

[32] The evaluative norms belonging to this kind of discourse have saved face from infamy] Using an understatement, the Jesuit authors of the Dictionnaire de Trévoux(1721) notice about expressions containing the word face that "these phrases (face-à-face, etc]) are imitated from the Scriptures (]]]). This makes the use of certain terms more acceptable". The discourse of Law is probably not subject to taboo. At least, the taxeme of designations of the face does not know the same diachrony in the religious discourse than in others.

[33] One of Molinet's ballads, by and large obscene in view of its rhymes obligingly itemised in the manuscript for absent-minded readers: “ Madame, j’ai sentu les façons / Du feu d’amour, puisque je vis / Les yeux plus aspres que faucons / De vostre gent et plaisant vis ” (Dupire, N., ed., Les Faictz et dictz de Jean Molinet, Paris, Picard, 3 vol., 1936, pp. 866-867).

[34] The competition of visage, suggested by Renson, is all the more unconvincing since visage is derived from vis ; the monosyllabic character of vis has also been put forward, but we can find monosyllabic words which are quite durable.
Like face, vitfinds itself banished because visage is traditionally opposed to parts of the body considered as evil. This theme, as we know, has been developed by Freud, who takes up Schopenhauer's polarization ("the head and the genitals are, so to speak, the opposed poles of the individual", excerpt from Le monde, in Insultes, Monaco, Le Rocher, 1988, p. 29).

[35] Euphemization has certainly an anthropological generality: for instance the name of the living body is extended to the dead body (the word cadavre (corpse) is not used during a funeral). The use of a neutral term like corps (body) (ex. levée du corps (= funeral) is not limited to an extension: we thus avoid the pejoration of the dead body.


Barthes, Roland 1984 . Le bruissement de la langue. (Essais critiques 4) Paris, Seuil.

Bischofsberger, Marco 1996. Sémantique historique et cognition, SCOLIA, Strasbourg, 9, pp. 46-59.

Blank, Andreas 1997 . Prinzipien des lexicalischen Bedeutungswandels am Beispiel der romanischen Sprachen. (Beihefte zur Zeitschrift für romanischen Philologie 285) Tübingen, Niemeyer.

Berlin, Brent and Key, Paul 1969 . Basic Color Terms. Berkeley, University of California Press

Bréal, Michel 1897 . Essai de sémantique, Paris, Hachette [new edition Brionne, Montfort, 1983].

Brunot, Ferdinand 1905-1937. Histoire de la langue française des origines à 1900, 12 volumes. Paris, Armand Colin.

Chamfort, Sébastien-Roch Nicolas [1795] 1968 . Maximes et pensées, caractères et paradoxes. Edited by Jean Dagen, Paris, Flammarion

Clédat, Louis 1895 . Les lois de la dérivation des sens appliquées au français, Revue de Philologie, 9, 49-55.

Coseriu, Eugenio 1964. Pour une sémantique diachronique structurale. Travaux de linguistique et de littérature II/1: 139-186.

----------------------1968. Les structures lexématiques. In : W. Theodor Elwert (ed.), Probleme der Semantik, 3-16 (Beihefte zur Zeitschrift für französische Sprache und Literatur, N.F.1), Wiesbaden, Steiner.

Darmesteter, Emile 1877. De la création actuelle de mots nouveaux dans la langue française et des lois qui la régissent, Paris, Vieweg.

----------------------1887. La vie des mots étudiée dans leurs significations, Paris, Delagrave [new edition Paris, Champ Libre, 1979].

Dumont, René 1992. Anthropologie, totalité et hiérarchie, in Anthropologie et philosophie, Paris, Centre Pompidou, 11-24.

Dumarsais, César C. [1730] 1988 . Traité des tropes. Edited by Françoise Douay-Soublin. Paris, Flammarion.

Dupire, N. (ed) 1936. Les Faictz et dictz de Jean Molinet, 3 volumes, Paris, Picard.

Geeraerts, Dirk 1988 . Cognitive grammar and the History of Lexical Semantics, in Rudzka-Ostyn, B. (ed.) Topics in Cognitive Linguistics, Amsterdam, Benjamins, 647-677.

Hjelmslev, Louis 1971. Essais linguistiques, Paris, PUF.

Husserl, Edmond [1901] 1963. Recherches logiques. Volume 3. Eléments d’une élucidation phénoménologique de la connaissance. PUF.

Katz, Jerrold J. and Fodor Jerry A. The Structure of a semantic Theory. Language 38, 170-210.

Koch, Peter 1995 . Der Beitrag der Prototypentheorie zur Historischen Semantik: eine kritische Bestandsaufnahme. Romanitischen Jahrbuch 46, 27-46.

Jackendoff, Ray 1983. Semantics and Cognition, Cambridge (Mass.), MIT Press.

Kleiber, Georges 1991. La sémantique du prototype, Paris, PUF.

Langacker, Ronald 1986. An Introduction to Cognitive Grammar, Cognitive Science, X, 1, 1-40.

------------------------1987. Foundations of Cognitive Grammar, I, Stanford, Stanford University Press.

Lakoff, George 1987. Women, Fire, and Dangerous Things. What Categories Reveal about the Mind, Chicago/London, The University of Chicago Press.

Livet, Charles-Louis 1895-1897. Lexique de la langue de Molière comparée à celle des écrivains de son temps, 3 vol., Paris, Imprimerie Nationale.

Luhmann, Niklaus 1980 . Gesellschaftsstruktur und Semantik. Studien zur Wissenssoziologie der modernen Gesellschaft, Band 1, Frankfurt a. M., Suhrkamp.

Nyckees Vincent 1997. Pour une archéologie du sens figuré, Langue française, 113, pp. 49-65.

Petitot, Jean 1983. Les catastrophes de la parole, Paris, Maloine.

-----------------1985 . Morphogenèse du sens, Paris, PUF.

-----------------1996. Les modèles morphodynamiques en perception visuelle, Visio, I, 65-73.

Piotrowski, David 1997 . Dynamiques et structures en langue, Paris, Editions du CNRS.

Rastier, François 1987. Sémantique interprétative, Paris, PUF.

----------------------1991. Sémantique et recherches cognitives, Paris, PUF.

----------------------1996. Chamfort : le sens du paradoxe, in Ronald Landheer & Paul J. Smith (eds), Le paradoxe en linguistique et en littérature, Genève, Droz, 119-143.

Reisig, Karl 1839 . Vorlesungen über lateinisches Sprachwissenschaft, Leizig, Lenhold.

Renson, Jean 1962 . Les dénominations du visage en français et dans les autres langues romanes, Paris, Les Belles Lettres.

Rosch, Eleanor 1978. Principles of Categorization. In : Rosch Eleanor and Lloyd Barbara (eds) Categorization and Cognition, 27-48, Hillsdale, Erlbaum.

Sweetser, Eve 1990. From Etymology to Pragmatics, Cambridge, Cambridge University Press.

Tabourot des Accords, Jehan 1922 . Escraignes dijonnaises (after Cioranesco XVI), Dijon

Talmy, Leonard 1988 . Force Dynamics in Language and Cognition. Cognitive Science 12, 49-100.

Vaugelas, Claude Faivre de [1647] 1934. Remarques sur la langue française. Edited by Jeanne Streicher. Paris, Droz.

Victorri, Bernard & Fuchs, Catherine 1996. La polysémie, Paris, Hermès.

Vous pouvez adresser vos commentaires et suggestions à :

©  décembre 2005 pour l'édition électronique.

Référence bibliographique : RASTIER, François. Cognitive semantics and diachronic semantics : the values and evolution of classes. Texto! [en ligne], décembre 2005, vol. X, n°4. Disponible sur : <>. (Consultée le ...).