3.1 Categorial gradience of English key

Gradience is defined as ‘the phenomenon of blurred boundaries between two categories of form classes’ (Aarts, Reference Aarts2007: 34). It implies that some members of a category are more prototypical than others (‘subsective gradience’) and that boundaries in between categories are not clear-cut (‘intersective gradience’). Subsective gradience can be illustrated by contrasting the English adjectives happy and alive (Aarts, Reference Aarts2007: 105–107). Whereas happy is a prototypical adjective as it displays five typical adjectival properties (attributive position (epithet), predicative position, intensification, gradedness and un- prefixation) (8), alive is considered less prototypical since it only allows the second and the third criteria (9):

1. (8)

   1. a. a happy woman

   2. b. she is happy

   3. c. very happy

   4. d. happy / happier / happiest

   5. e. unhappy

2. (9)

   1. a. *an alive hamster

   2. b. the hamster is alive

   3. c. very (much) alive

   4. d. ?alive / more alive / most alive

   5. e. *unalive

Our working hypothesis is that French clé manifests synchronic categorial variation or ‘intersective gradience’, thus occurring in typical nominal constructions, in typical adjectival constructions, and in ambiguous constructions that can be occupied by nouns and adjectives.

The categorial gradience of the English counterpart of French clé, viz key, has already been studied by Denison (Reference Denison and Brinton2001), Aarts (Reference Aarts2007), and De Smet (Reference De Smet2012). Denison (Reference Denison and Brinton2001: 128–129), for instance, observes a noun-adjective cline for English key, with examples as the following illustrating the different positions on this cline:Footnote ⁴

1. (10) Fear and ambition are the respective keys to their characters.

2. (11) Occupants of key offices such as the Presidency of the Attorney-Generalship. (OED, 1926)

3. (12) More emotional weight is carried in the key domestic scenes (. . .) (FLOB, C01 103)

4. (13) There are a number of reasons why people lose their hair, stress is a very key factor. (BNC, HVE 174)

5. (14) Meirion Rowlands, one of the Ashley's most key appointments of this time, (. . .) (BNC, GU9 7)

6. (15) Noting that such incidents are not marginal but key to Edgeworth's plots, (. . .) (Butler, 1992)

In examples (10) and (11), Denison (Reference Denison and Brinton2001) analyses key as a noun, used as a predicate (with determiner) (10) or as a premodifier (11). Instead, the use of key with adverbial modification in (13), its superlative use in (14), and its predicative use (without determiner) in (15) are considered to be adjectival uses. It should be noted that in the latter example, key is also coordinated with an adjective, marginal, which could be taken as another argument in favour of the adjectival analysis. However, this criterion is not decisive since adjectives can be coordinated with nouns too, as Denison (Reference Denison and Brinton2001: 127) himself illustrates with the example It's lovely but a mess. Example (12) could be seen as an intermediate type: key is still used attributively, but has scope over a complete noun phrase (domestic scenes). Denison concludes that the cline suggests ‘a graded series of transitions’ (Denison, Reference Denison and Brinton2001: 129) all attested in current language use, providing evidence for synchronic gradience.

Aarts (Reference Aarts2007: 132–133), on the other hand, is more sceptical and even more radical about the status of key. He argues that even in example (11), the compound-like type, key could be assigned adjectival status, given the synchronic co-existence with unambiguously adjectival uses (13–15) and the semantic discrepancy between attributive key and its use as an independent noun (compare (10) in which key refers metaphorically to a way of access with example (11) in which key means ‘principal’). More generally, Aarts (Reference Aarts2007) questions many examples claimed as cases of intersective gradience in this way, assuming that they belong either to category A or B, but not to both.

However, (at least) two counterarguments can be advanced against this view. First, the semantic change from (10) to (11) is not as abrupt as suggested: just like in the case of clé (see Section 2), the evaluative meaning of English key is very likely to be derived from its metaphorical reading outside the ‘key + N’ pattern, as in Experience is the key to success. Second, synchronic language use naturally reflects the different stages of diachronic evolution, including the transitional and possibly ambiguous bridging contexts (cf. Hopper's idea of layering (Hopper, Reference Hopper, Traugott and Heine1991)). Or to put it in another way, synchronic gradience is the result of diachronic gradualness, as claimed by Traugott and Trousdale (Reference Traugott, Trousdale, Traugott and Trousdale2010). In correlation with this diachronic point of view, De Smet (Reference De Smet2012)'s meticulous analysis of the evolution of English key provides strong evidence for a noun-adjective cline, the (ambiguous) attributive uses of key, considered to be nominal by Denison (Reference Denison and Brinton2001) and adjectival by Aarts (Reference Aarts2007), being attested before the (non-ambiguously adjectival) predicative ones.

In the next sections, the English data on key will be compared with French clé. More specifically, Section 3.2 will investigate by an in-depth corpus study the extent to which the typical adjectival properties (attributive construction, predicative construction, intensification and gradedness) are attested for clé in its synchronic use and Section 3.3 will examine to which degree the innovative constructions with clé are accepted by young speakers of modern French in Belgium, France and Quebec.

3.2. Categorial gradience of French clé

For want of sufficient data found in the ‘traditional’ written corpora of French, such as Frantext, I have made use of the search engines GlossaNet (3.2.1) and WebCorp (3.2.2) designed to explore the web as a corpus and offering the advantage of including very recent and even informal language use (blogs, forums, etc.) in the corpus material. By means of these corpus tools, instances of the following syntactical properties, typically associated with adjectives (Denison, Reference Denison and Brinton2001; Aarts, Reference Aarts2007), were looked for:

1. (a) coordination of clé with an adjective;

2. (b) attributive use of clé over a (lexicalised) polylexical unit;

3. (c) intensification by degree adverbs / modification by other adverbs (in attributive construction);

4. (d) gradedness (comparative and superlative constructions);

5. (e) predicative use.

It should be noted that, very marginally, morphological adjectival properties are also attested for French clé, which very rarely allows gender inflection (e.g. position clée vs position clé). These morphological properties have been analysed in Van Goethem (Reference Van Goethem2012), but will not be taken into account in the present study.

3.3 Degrees of acceptability and geographic variation

3.3.1 Acceptability survey Louvain-la-Neuve (Belgium)

In order to check the validity and the representativeness of the corpus results, I have carried out an acceptability survey with 66 French-speaking Language and Literature students of the university of Louvain-la-Neuve in 2011, investigating the synchronic categorial gradience of clé in terms of degrees of well-formedness (cf. Traugott and Trousdale, Reference Traugott, Trousdale, Traugott and Trousdale2010: 20). All selected students were between 19 and 26 years old, had Belgian nationality and French as their mother tongue. The test presented 14 different strings with clé, all attested on the Internet, and had to be evaluated on a scale from 1 (not acceptable) to 5 (fully acceptable).

The results are summarised in Table 1 and Figure 1. In Table 1, the first column presents the different strings in the expected order of acceptability, ranging from the ‘nominal uses’ to more and more typical ‘adjectival uses’ (in the actual test these constructions were obviously ordered at random). The next three columns indicate the average score, the median score, and the standard deviation for each string. The diagrams in Figure 1 represent the mean and median scores for each string.

Table 1: Results survey LLN

Figure 1: Results survey Louvain-la-Neuve (Belgium)

The following conclusions can be drawn from Table 1 and Figure 1.

First, the metaphorical use of la clé meaning ‘the crucial factor’ (S1) and its use as N2 in the compound-like construction (S2) are fully accepted (median scores of 5/5).

Second, the median value of the acceptability is still 4/5 for the coordination of clé with an adjective (S3), its attributive use with scope over a (lexicalised) noun phrase (S4), and its adverbial modification within the attributive construction (S5), although these strings are inconsistent with the compound analysis and suggest that clé acts as an adjective rather than as a noun.

Third, the comparative (S6) and superlative (S7) constructions are very poorly accepted (median values of 2/5), even if clé is still used attributively in the proposed examples.

Fourth, in sentences (S8) until (S14), clé occurs in a predicative construction; in (S8) and (S9), the subject is a noun phrase; the other examples have a non-nominal, impersonal subject (ce, cela, il). The results clearly indicate that the examples with a nominal subject are much better accepted (median scores 3–4/5) than the ones with an impersonal subject (median scores 1–2/5). However, it is striking that when clé is preceded by the adverb vraiment ‘really’, the mean and median acceptability scores dramatically rise. As the comparison of the scores for the examples (S8-S9) and (S13-S14) indicates, this fact holds for both the predicative constructions with a nominal and an impersonal subject.Footnote ⁵ It can be assumed that the relatively high degree of acceptability of the collocation ‘vraiment clé’ in the predicative constructions results from the fact that this sequence was already well established and well accepted in the attributive use, since S5 obtained a median score of 4/5.

3.3.2 Acceptability surveys Lille (France) and Sherbrooke (Quebec)

In order to check the possible geographic variation of these results, I have submitted the same survey to native French-speaking students at the universities of Lille 3 (France) and Sherbrooke (Quebec). Table 2 presents the results from Lille 3 (20 students) and Table 3 from Sherbrooke (34 students).

Table 2: Results survey Lille (France)

Table 3: Results survey Sherbrooke (Quebec)

The diagrams in Figure 2 compare the means found for each string in the three varieties of French.

Figure 2: Comparison of survey results Louvain-la-Neuve - Lille - Sherbrooke

As for Lille, the results are generally comparable to the ones found in Louvain-la-Neuve, even if a slightly more positive attitude can be noticed for the well-established constructions (S1-S4) and a less favourable position for the more ‘innovative’ constructions (S5-S14).

A more conservative tendency is in general observable in the Sherbrooke results. As Figure 2 clearly indicates, this finding is most striking in strings 9 and 14 where the positive effect of the insertion of the degree adverb vraiment is almost non-existent. Although I expected the results from Quebec to be more influenced by the more entrenched English adjectival constructions with key, according to the survey results Quebec presents inversely a more conservative (and possibly protective) attitude towards the innovative constructions with French clé.

This conservative attitude can also be detected in the Canadian media. As an illustration, I provide in (50) an Internet extract originating from a Canadian website related to the radio programme ‘Français au micro’, in which a language specialist comments on the predicative use of clé in sentences such as Cette mesure est-elle vraiment clé? (lit. ‘This measure is it really key?’).

1. (50) Cette mesure est-elle vraiment clé?

   Le mot clé peut être employé en apposition, avec ou sans trait d'union, pour parler d'une chose très importante. On peut dire, par exemple: Un poste-clé, des mots-clés, un témoin-clé, etc. En revanche, le mot clé ne peut pas être utilisé comme épithète ni comme attribut. Ici, il aurait mieux valu dire: Cette mesure est-elle vraiment essentielle, vitale, indispensable, déterminante, capitale, etc. Et bien sûr, on pouvait dire: Cette mesure est-elle vraiment une mesure-clé. Footnote ⁶

   http://www.radiocanada.ca/radio/francaisaumicro/index.asp?dateachercher=1/12/2009, 01.12.2009)

Obviously, the language specialist rejects the use of clé in the predicative position, providing alternative formulations based on synonymous adjectives such as essentiel(le) ‘essential’ and indispensable ‘indispensable’, besides the use of the compound-like construction mesure-clé ‘key measure’. The mere discussion of this construction however indicates that this language use is actually observed.

4.1 Diachronic data

In what precedes, it has been shown that French clé participates in a wide range of adjectival constructions, positioned on a cline ranging from ambiguous attributive constructions to more and more typically adjectival constructions, increasingly distanced from the original compound-like constructions. Moreover, this cline is largely confirmed by the scores of acceptability ascribed to the different uses of clé by native speakers of French in three distinct geographic regions. In line with the idea that synchronic gradience is the result of gradual diachronic change (Traugott and Trousdale, Reference Traugott, Trousdale, Traugott and Trousdale2010), these findings provide strong evidence for a gradual N > A recategorialisation of clé.

Obviously, in order to be validated, this hypothesis needs to be corroborated by diachronic corpus data, even if the emergence of the adjectival uses of clé appears to be a very recent phenomenon. The historical and literary Frantext corpus is obviously not the most appropriate tool for its diachronic study: no single adjectival use of French clé is attested in Frantext, which however contains more than 5000 occurrences of clé, used as a free noun or in the [[N][clé]] construction. The oldest unambiguously adjectival use of clé retrieved by WebCorp only dates from 2001 and, as the results of the acceptability surveys show, many of the uses are not (yet) fully or even very poorly accepted by young native speakers of French, even if they are attested on the Internet.

Since the emergence of the adjectival constructions of French clé seems to be a very recent and still ongoing process, we need a tool that gives access to exhaustive dated and innovative language material. In this respect, the archive search on the http://news.google.com websiteFootnote ⁷ turned out to be more helpful. It allowed me to explore an extensive database of historical and modern newspaper archives and automatically browses timelines showing the frequencies of the searched items over time.Footnote ⁸

Although the number of results is still limited, this web archive provides us with some interesting indications about the diachronic development of the different constructions with clé. It can for instance be observed that the compound-like use (cf. secteur clé ‘key sector’) is attested since the eighties in the collected documents (51), whereas the attributive use with scope over a noun phrase (cf. secteur économique clé ‘key economic sector’) only occurs from 2002 on (52).

1. (51) L'histoire de l'industrie est intimement liée à celle des machines : les grandes mutations du secteur industriel sont en général commandées par l'apparition ou la diffusion de nouvelles machines, l'autre secteur clé étant celui de l'énergie.

   ‘(. . .) the other key sector being that of energy.’

   http://www.ina.fr/fresques/jalons/fiche-media/InaEdu01452/la-modernisation-de-l-appareil-de-production-francais-la-robotisation-des-usines-renault.html (1981)

2. (52) A cette tension s’ajoute la concentration de cette main-d’œuvre dans le secteur économique clé du pays: le cacao.

   ‘(. . .) the economic key sector of the country: cacao.’

   http://www.aujourdhui.ma/imprimer/?rub=actualite&ref=23525 (2002)

Comparative and superlative uses containing the sequence ‘plus clé’ (‘more key’) are observed from 2003 on, first in attributive constructions (53), and only from 2011 onwards in the predicative ones (54).

1. (53) Dans le nouveau cabinet, le poste le plus clé est confié à un proche du Président Arafat, le général Nasser Youssef.

   ‘(. . .) the most key position has been assigned to a close relative of President Arafat, general Nasser Youssef.’

   http://www.aujourdhui.ma/imprimer/?rub=actualite&ref=1865 (2003)

2. (54) Toutefois, David Horne, directeur marketing de Blaze, a tenu à préciser que les deux types de mobiles chargeaient les sites rapidement, un point qui devenait de plus en plus clé dans le choix de telle ou telle solution parmi les consommateurs.

   ‘(. . .) an issue that has become more and more key (. . .)’

   http://www.linformaticien.com/actualites/id/10580/android-plus-rapide-de-50-que-l-iphone-dans-le-chargement-des-pages-web.aspx (2011)

The first attested predicative use of clé goes back to 1997 (55), but isolated cases left aside, this construction only becomes more common from 2006 onwards (56). The first occurrence of an adverbial modification also dates from 2006 (57); the collected material only provides examples of adverbial modification in predicative constructions.

1. (55) Le rôle des prospectus est clé.

   ‘The role of flyers is key.’

   http://www.lsa-conso.fr/la-valeur-s-installe-pas-la-serenite,52359 (1997)

2. (56) Nous sommes impatients d'avoir l'opportunité de développer notre partenariat avec ILOG et considérons que cette alliance est clé pour nous aider à poursuivre notre croissance (. . .).

   ‘(. . .) that this alliance is key for helping us to continue our development (. . .)’

   http://www.communique-de-presse.com/200610255929/Entreprises/ILOG,entre,au,capital,d,un,editeur,de,logiciel,chinois.html (2006)

3. (57) Aux Etats-Unis, ce que les analystes regardent en premier lieu est le bénéfice par action publié par les sociétés, par rapport aux prévisions. Le poids du consensus est vraiment clé, là-bas.

   ‘(. . .). The importance of a consensus is really key, over there.’

   http://www.journaldunet.com/0604/060427-bourse.shtml (2006)

Finally, the construction c'est clé is attested since 2007 (58) and the impersonal construction il est clé is attested from 2009 onwards (59).

1. (58) C'est le début, mais le mobile est l'outil social par définition, donc pour un réseau social, c'est clé.

   ‘(. . .) but mobile telephone is the social tool by definition, so for a social network, it is key.’

   http://www.journaldunet.com/chat/retrans/070116-mayor.shtml (2007)

2. (59) S’il est cléd’obtenir la sympathie et l'adhésion du peuple, soyez gentil avec les autochtones.

   ‘If it is key to obtain sympathy and adhesion of the people, be kind to autochthons.’

   http://www.lematin.ma/Actualite/Express/ArticlePrint.asp?id=123133 (2009)

To conclude, even if the material found thanks to the Google web archives should be interpreted with caution, it proves at least to be in line with our assumptions derived from the synchronic data: in these data the attributive uses precede the predicative ones and within the predicative construction, the ones with a nominal subject precede the impersonal constructions.

4.2 Degrammaticalisation or lexicalisation?

Whereas Traugott and Trousdale (Reference Traugott, Trousdale, Traugott and Trousdale2010) essentially investigate the interface between (synchronic) gradience and (diachronic) grammaticalisation, the gradual emergence of free adjectival uses out of more or less bounded nouns, participating in compound-like constructions, shares important characteristics with a specific type of ‘degrammaticalisation’, that is so-called ‘debonding’. Debonding is ‘a composite change whereby a bound morpheme in a specific linguistic context becomes a free morpheme’ (Norde, Reference Norde2009: 186). As a matter of fact, the typical parameters of debonding play a crucial role in the evolution of clé, that is severance (i.e. decrease in bondedness), flexibilisation (i.e. increase in syntactic freedom), scope expansion and recategorialisation.

Norde (Reference Norde2009: 232) argues that debonding minimally requires a decrease in bondedness, called severance. Our data suggest that the adjectival uses of clé emerged in the [[N][clé]] construction: originally used as a bound noun, clé gradually develops more and more typical adjectival uses, not only in the original attributive setting (cf. secteur économique clé ‘key economic sector’) but also in predicative contexts (cf. ce secteur est vraiment clé pour l'économie ‘this sector is really key to the economy’).

Recategorialisation, scope expansion and flexibilisation also apply to clé: clé gradually develops adjectival properties (recategorialisation), at word level (e.g. secteur clé ‘key sector’), phrase level (e.g. secteur économique clé ‘key economic sector’), and sentence level (e.g. ce secteur est vraiment clé pour l'économie ‘this sector is really key to the economy’) (scope expansion), while simultaneously increasing its syntactic freedom (flexibilisation).

However, even if the typical parameters of debonding seem to apply to the evolution of French clé, true degrammaticalisation is not involved. As the term itself clearly indicates, degrammaticalisation implies loss of grammatical content and/or function (or at least gain in lexical content and/or function), and this does not hold for the category change from noun to adjective, entirely situated within the lexical domain. Moreover, the use of nouns within compounds cannot be equated to bound uses of clitics and (inflectional and derivational) affixes, dealt with by Norde in her chapter on debonding (Norde, Reference Norde2009: 186–227). If Norde and Van Goethem (Reference Norde and Van Goethem2013 (ms)) argue that debonding of lexical affixes and affixoids should be seen as a process sui generis, to be distinguished from the debonding of clitics and grammatical affixes, this will all the more hold for the debonding of compound members with co-existing free counterparts. Finally, Van Goethem and De Smet (forthcoming) demonstrate that debonding is subject to language-specific constraints, such as the degree of cohesion of compounds. While analogous processes in Dutch can therefore be seen as more proper instances of debonding, the morphological stage being characterised by a true bound use in the form of affixoids, this analysis is much less evident in French.

An alternative theoretical approach consists in considering this N > Adj change as an instance of lexicalisation (Brinton and Traugott, 2005). However, Brinton and Traugott (2005: 96) adopt a narrow definition of lexicalisation, qualified as a gradual, diachronic process of fusion, from which conversion, considered to be an ordinary process of word-formation because of its instantaneous and predictable character, is excluded (Brinton and Traugott, 2005: 37–40; 97–98). Norde (Reference Norde2009: 11), by contrast, adopts a broader definition of lexicalisation, including processes such as clipping and conversion, because these changes ‘result in new lexemes, the meaning of which is not fully predictable from the (part of the) word from which they evolved, nor from the nature of the word formation process that formed them’ (Norde, Reference Norde2009: 11).

We end up with a theoretical impasse in which the gradual emergence of adjectival uses of the noun clé combines properties of both degrammaticalisation (viz, debonding) and lexicalisation (viz, conversion), but neither framework seems to be able to describe the process accurately.

From a constructional point of view, we have described the ‘N + clé’ sequence in Section 2 as a semi-schematic morphological construction, which has been shown to give rise to new, syntactic constructions. In the next paragraph, it will be investigated if the notion of ‘constructionalisation’ can provide a more insightful account for our data.

4.3. Constructionalisation

The emergence of new constructions has recently been labeled ‘constructionalisation’ (among others Bergs and Diewald, Reference Bergs and Diewald2008; Trousdale and Norde, Reference Trousdale and Norde2013; Traugott and Trousdale, forthcoming). This process involves ‘a sequence of changes in the form and meaning poles of a construction, whereby new formal configurations come to serve particular functions, and to encode new meanings’ (Trousdale and Norde, Reference Trousdale and Norde2013: 36). We believe that this process exactly describes what is occurring to French clé. In what follows, the different stages of this constructionalisation process will be described.

In a first stage, the noun clé is used in a morphological semi-schematic construction, represented in (60), in which it is typically associated with the meaning ‘very important, crucial’. This is a first step in the constructionalisation process: clé is embedded in a conventional yet productive form-meaning pair.

1. (60)

   

Hüning and Booij (Reference Hüning and Booij2013 (ms)) present a similar analysis of Germanic affixoids being integrated in Germanic compounds, such as the German intensifying element stock ‘lit. stick’, which productively forms part of adjectival compounds such as stockbesoffen ‘very drunk’ and stockconservativ ‘conservative to the core’. The authors suggest that this kind of processes should not be qualified as instances of grammaticalisation or lexicalisation, because the dichotomy between what is grammatical and what is lexical is not adequate for a correct approach of word formation phenomena, mostly combining both aspects. In turn, the formation of productive form-meaning mappings at word-level can better be accounted for by constructionalisation, since this approach does not focus on the grammatical or lexical in- and output of the process, but on the course of the process itself.

Throughout this article, it has, however, been shown that the constructionalisation of clé goes beyond this morphological stage since clé occurs in syntactic constructions with the same semantics. As already suggested by Amiot and Van Goethem (Reference Amiot and Van Goethem2012), the adjectival uses of clé, and other French N2s occurring in similar constructions (see Section 1), could be seen as the result of an interaction or blending between the, closely related, morphological [[N][N]]_N and syntactic [[N][AP]]_NP constructions. The fact that clé developed an evaluative meaning in the former construction, typical of adjectives, may have favoured this constructional ambiguity. In constructional terms, this blending can be translated as an instance of ‘multiple inheritance’. As described in Section 2, within the constructional approach to language, our linguistic knowledge is organised as an inheritance network in which each construction inherits its properties from a more schematic construction. Constructions are, however, often linked to more than one parent construction, as stated by Trousdale and Norde (Reference Trousdale and Norde2013: 35):

Schematically, this multiple inheritance can be represented as in (61). The [[N][clé]] sequence inherits its properties from two distinct parent constructions, the morphological compound-like ‘N+N’ pattern and the syntactic ‘N + AP’ pattern. As a consequence, it is ambiguous between a morphological and a syntactic construction: in line with N+N compounds, clé does for instance not (or only very rarely) display gender inflection (cf. Van Goethem, Reference Van Goethem2012), while it has been proven to occur in typical adjectival constructions, for instance being modified by degree adverbs, a property inherited from its syntactic parent.

1. (61)

   

The fact that clé still resists to a large extent to be used in adjectival constructions such as the predicative one, can also be understood in this account: it can logically be assumed that clé prefers to be used in constructions that reconcile properties of both of its parent constructions, such as the attributive construction, the predicative one being incompatible with the compound analysis.

A further constructionalisation step could possibly imply that [[N][clé]] strengthens its inheritance relation to its syntactic parent construction, which would entail that clé entrenches and extends its syntactic uses and further develops the morphological properties typical of adjectives, such as gender inflection.