Navigation – Plan du site

The Foundations Of Discourse: The Case Of British Stereotypes Of The French

Pierre Larrivée et Julien Longhi

Texte intégral


1It is widely accepted that individuals use language to describe the world around them. This perspective would lead us to expect that what people say about the world reflects both diversity in terms of individual perspective, and the multiplicity of the world itself. However, there are regularities in discourse in the sense that what people say about the world is often a repetition of what has been said before (Bakhtin 1981), as opposed to a reflection of a given individual’s immediate experience of the world. French people talking about the British seem to focus more frequently on their phlegmatic character, their reserve and their hypocrisy than on other alleged character traits (Crouzet 2006). These predicates may or may not be applied to Britons in other communities, and the American people I have consulted do not seem to agree that phlegm would be the first characteristic of British nationals. However, a community may change its views through time, and it is plausible that the stereotypes of the British in the US differ today from those of 1776.

2The purpose of this paper is to advocate the use of discovery procedures in the study of conventional discourses. In order to distinguish discourses that are the result of socially ratified knowledge, and those that are not, it is necessary to establish some criteria. The investigative work documented in this paper represents an important contribution to the understanding of meaning in language: discourse analysis in its various guises seldom spells out the methodological basis for the interpretations it puts forward, and therefore little substantiation is available to establish the cognitive reality of discourses at all: a sad state of affairs given the prominent role of discourse in shaping individual and collective experience. Our innovative approach is achieved through looking at stereotypes as an epitome of socially shared knowledge. We conduct three preliminary experiments inspired by corpus linguistics, social psychology and pragmatics in order to assess the degree to which bringing together different methods can illuminate the existence of socially ratified discourses. The focus is therefore less on the depth and breath of these experiments than on their joint potential to reveal what is a conventional discourse, and what is not.

3The first section presents research traditions for which some discourses are regarded as socially ratified knowledge rather than merely the expression of personal belief. This is followed by an analysis of British stereotypes of the French in a press corpus, sentence completion questionnaires and inference judgments. The degree of convergence of predicative structures applied to all the relevant members of the target group in revealing its corresponding conventional discourses is emphasised in the conclusion.

1. The discourse matrix assumption

4The view that socially stable representations attach to a topic to inform discourse productions has played an important role in the French research community. The impetus given by Michel Pêcheux to discourse analysis relies on the notions of inter-discourse and discursive formation. These two separate notions come to merge in later work in the field: inter-discourse corresponds to background knowledge about the conventional contents of some discourses, finding an abstract representation in discursive formations (Guillaumou 2006). The merger is pursued by Foucault (1972), for whom a linguistic sequence is an utterance only if it is situated in an enunciative field – i.e. if it refers to the conventional discourses about the topic – whereas the set of these conventional discourses constitutes a discursive formation. Foucault (1972) proposes that discourse is a social practice underpinned by areas of conventional knowledge called discursive matrixes (formations discursives). A discursive matrix on social stereotypes of the British would, in France, comprise predicates concerning their phlegm, reserve and hypocrisy. The existence of such matrixes would explain why these predicates would seem to come up with notable frequency in discourse productions. They would further explain factors such as the lag between expectations and reality, as documented in a large literature on irrationality (Marcus 2008, Morel 2002, Poundstone 1988, Reason 2000); how experience can be reinterpreted following social interpretative frameworks (Spanos, Burgess and Burgess 1994); and the documented influence of stereotypes on subject decisions (Bodenhausen and Wyer 1985). Comparable to the idea that action is informed by social forms of knowledge (Weber’s frame of perception, Panofsky’s habit, Bourdieu’s habitus, Goffman’s frames, Schank’s scripts, Dawkins’ memes and Levinson’s semtem), the notion of discursive matrixes has led to an ever-growing body of descriptive work in the different strands of discourse analysis. However, the empirical work has not always provided immediate answers to the crucial questions raised by the notion of socially shared forms of knowledge underlying discourse productions. Amongst these questions lies the issue of how a discursive matrix can be demonstrated.

5What kind of evidence would confirm the hypothesis that the French harbour the belief that the British are phlegmatic, reserved and hypocritical? If such ideas emanate from social knowledge, unlike the propositions that the British are energetic, open-minded and generous, the distinction between the two requires a demonstration. The ability to establish the proposed conventional representations appears pressing for a number of reasons. Sociologically, discourse analysis, practices and assumptions vary considerably according to communities, as evidenced by the scant cross-references between, for instance, the French School of Discourse Analysis (Pêcheux 1969, Foucault 1972) and the British Critical Discourse Analysis (Fairclough and Holes 1995, Wodak and Meyer 2001). One undesirable scientific consequence is that any results achieved in one community might be difficult to replicate in another, due to the fact that the analytical tools might rely on intuition or culture-specific notions (which is eloquently demonstrated for cross-cultural pragmatics by Wierzbicka 1991). Whilst considerably complicating the testing of particular hypotheses, this also obscures the approach of general questions raised by discursive matrixes, which to select a few major ones, comprise their acquisition, sharing, use, variation and change, and causes. Let’s mention here some of the key questions which arise when approaching research on discourse matrixes:

  1. How are matrixes learned? The acquisition of (self) stereotypes is evidenced to occur around 10 years of age (Rutland 1999). However, what input this process is based on remains to be elucidated, although an answer may be formed through a comparative study involving results from the acquisition of lexical meaning (Bloom 2000).

  2. How can representations that are learned by individuals come to converge in a community (Zlatev, Racine, Sinha and Itkone 2008)? Social convergence does seem to apply to stereotypes (Karasawa 2007). The paradox of socially convergent knowledge on the part of cognitively independent individuals, and the degree of that convergence, tends to remain, and does so for linguistic and other aspects of social meaning.

  3. By what processes are different dimensions of matrixes highlighted? That the British are hypocritical is not likely to be evoked by the French in every context, and indeed the type of context that might bring about this predicate is questionable (Blair 2002). The contextual dependency of discursive meaning is comparable to the complex linguistic question of polysemy, where an under-specified item like ‘go’ or ‘any’ has different readings in characterised contexts (future and movement in ‘I am going to go to the beach’, negation and universal quantification in ‘He can’t do anything right’ vs. ‘He can do anything he likes’).

  4. Why do different discourse matrixes have a varying degree of efficacy? Some propositions are more performative than others, and a course of action is more readily commended as furthering competitiveness than as being virtuous.

  5. How do discursive matrixes vary and change? Discourse productions change through time, as, presumably, do discursive matrixes: no contemporary equivalent is found to the medieval French belief that the sub-human English people had a tail, and their presumed phlegm is not shared by other communities. A diachronic version of discourse analysis remains to be provided to match three hundred years of inquiry into shifts in linguistic meaning.

  6. One question relating to variation and change is that raised by Fairclough (2003) of hegemony. How does a discourse matrix become dominant in a community? This particular issue concerns discourse matrixes fulfilling ideological functions (van Dijk 1998), and could be compared to linguistic forms becoming markers of normative registers.

  7. What are the causes of discursive matrixes? Why should such knowledge exist? What purpose is served by the belief that Britons are reserved? In this sense, the study may benefit from a comparative approach involving other conventional knowledge, such as highly routinised grammatical meanings that allow for quick communication, according to Givón (2002). Routinised discourses allow rapid communication (Moscovici 2000), and may constitute a marker of identity and exclusion (Wodak 2008), furthermore, the knowledge that underlies them helps make sense of the world (Taijfel 1981: 33).

6The resolution of the issues documented above rests on the ability to identify and characterise discursive matrixes. Proposals as to how discursive matrixes can be identified in a testable way constitute the purpose of this work. In this paper, three methods are reviewed from the fields of Corpus Linguistics, Pragmatics and Social Psychology. Each method is applied to the particular issue of British stereotypes about the French in order to establish their general applicability and degree of convergence.

7The analysis has considerable relevance for a better understanding of the workings of cognition. If discourse productions conventionalise in a form of knowledge that sees particular predicates related to a given topic, then that knowledge should have an impact on the perception, recognition, categorisation and memory of the empirical experience relating to that topic. In other words, the existence of socially shared knowledge that underlies discourse production has a considerable importance for the notions of agency and models of cognition. The model of the free individual endowed with reason is certainly circumscribed if their discursive production refers to conventional knowledge that may or may not reflect experience. An individual’s knowledge is often presented as a result of his or her relationship with the world, and Aquinas’ empiricist maxim that what is in the mind has first been in the senses is often taken for granted by different models of cognition. If, however, a proportion of what people know is partially derived from social conventions, as illustrated not only by fictitious beings that populate discourse, but also by presumably real beings of which experience might be mainly discursive (such as the dark side of the moon and the bottom of oceans), then sensory experience is not the unique source of knowledge: in effect, it might well be possible to know that the British are reserved without ever having had an encounter with a member of that community. If experience were the basis of knowledge of national character, considerably more variation might be expected than if its origins were socially ratified. This is illustrated by the following:

(1) Heaven is: French cook, British police, German mechanic, Italian lovers, and the Swiss run everything
Hell is: British cook, German police, French mechanic, Swiss lovers, and the Italians run everything

8This slightly tired joke does make the point that not all predicates are conventionally associated to all topics, and that, irrespective of individual experience, there is a social belief that Italians are better lovers than they are administrators. The ways of establishing that one discourse is conventional and the other is not, is the topic of the next section

2. How do we know there are Discursive Matrixes?

9It has been claimed that certain existing discursive matrixes constitute a conventional relation between a topic and different predicates. There are three interrelated approaches to the issue of how this relation can be established, each promising transferable methods. Primarily, the frequency of the relationship between the predicate and topic should be observed with conventional discourse representation as proposed by Corpus Linguistics. Secondly, this relationship should facilitate inferences about the topic, as sometimes noted in Pragmatics. Lastly, Social Psychology observes regularities in judgments revealed by discourse completion tasks. In this section, each factor is considered in turn before being applied to the stereotypes of the French in section 3.

10It has been observed for some time that the particular frequency of certain words and expressions in a language may reveal the cultural assumptions of the community that speaks it. Predated by psychoanalytical techniques of word association, Georges Matoré and Algirdas Greimas’ notion of the key word was rediscovered by Wierzbicka (1991) and Corpus Linguistics (Stubbs 2001, 1998); it plays a formidable role in the design of search engines. Yet it is not only applicable to isolated words, Corpus Linguistics has shown that collocations between words and whole expressions can be revealing (as shown by the Google endeavour Culturomics that aims to diagnose cultural change on the basis of collocation changes in giant web corpora). This is demonstrated by Wolfgang Teubert, for instance, whose exploration of the British media discourses about the European Union shows that it is characterised with high frequency as a bureaucratic federal super-state that obscures the global transatlantic aspirations of Britain (Teubert 2007; see also on another topic the work by Gianoni 2010). The notion that conventional discourses are established by the frequency of attributes with predicates suggests that this could be a way to establish discourse matrixes. A Google query shows that the sequence ‘The French are arrogant’ gets 1 090 hits in July 2009, whereas ‘The French are hypocritical’ gets 10 and ‘The Belgians are arrogant’ gets none. Clearly this communicates something about the social perceptions of the French in English-speaking communities, although the internet does not in itself allow an immediate definition of which community is involved, and the considered sequence is merely one way in which the expected discursive matrix surfaces.

11A second avenue of enquiry is provided by pragmatic inferences. What is explicitly said constitutes but a part of the communication that "takes place against a background of beliefs or assumptions which are shared by the speaker and his audience, and are recognized as such" (Stalnaker 1974: 199). The common ground that speakers assume makes exchanges manageable, as these assumptions do not have to be spelled-out in every exchange. Background knowledge does surface in certain contexts, especially where implicit conclusions are concerned, as pointed out independently by both Grice and Ducrot through the respective notions of implicatures and sous-entendu. Implicit knowledge is necessary for the rationality of the argument made in the following exchange:

(2) A: Is John really such a bad cook?
B: Well, he’s English
A: That’s outrageous / a fair point. (Wedgewood 2007)

12Without the generally accepted idea in England that the English are hopeless at cooking, the relation between B’s answer to A’s question would seem anomalous. This anomaly is illustrated by a fictitious exchange such as –Is John a good cook? –Well, he’s a discourse analyst, which is definitely curious not because the intended slur cannot be read, but because there are no generally accepted ideas about the cooking abilities of members of the aforementioned profession. The discourse matrixes about the French are exploited by attested exchanges, one particular example occurs in the subtitles of the Italian film Novecento:

(3) – Is she chic?
– Well, she’s French!

13A further example can be seen in the English subtitles of the French film Les Visiteurs, where the rudeness of a character is justified by the assertion that he is French (cited in Humbert 2003; see also for the representation of the French in cinema, Durham 2008, Knox 2003, Verdaguer 2004). Another way to identify whether such ideas are part of the common ground is to consider the connectors that are used. Since but generally coordinates notions that proceed in different argumentative directions, unlike and which should bring together notions that lead to the same conclusion, the conjunction of John is English but a good cook and that of John is English and a bad cook should and does yield felicitous sequences, whereas John is a discourse analyst but a good cook and John is a discourse analyst and a bad cook that can only be a description of accidentally related properties, and much more so that John is English and a good cook and John is English but a bad cook. Exploiting the argumentative direction of connectors allows for the diagnostic of the relation between predicates and a topic to be elicited rapidly (Anscombre 2001, Larrivée 2009: 191). A consequence of this is to point to the argumentative dimension of discourse matrixes (Galatanu 2002, Longhi 2008, after Ducrot, see Anscombre and Ducrot 1983), which provide a conclusion that is part of the common ground.  

14A third way to validate the existence of a conventional relation between a predicate and a topic is through the controlled methods of social psychology. The study of socially conventional perceptions is carried out through a variety of empirical methods that generally involve subjects making judgements along pre-defined dimensions. The resulting data is often rated in accordance with Likert’s scales of socially meaningful occurrences, and sentence completion tasks. Sentence completion tasks typically involve subjects completing specified sentences in order to identify their latent beliefs on a topic, which in turn leads to the identification of collective opinions. Such work is pursued by Suárez-Orozco (2000) in her exploration of how children of immigrants think that their communities are viewed by the American society. These subjects were asked to complete the sentence ‘Most Americans think [people from my country] are...’. Overall, the responses in Suárez-Orozco’s study reflected a negative evaluation (‘bad’), which was related in some cases to poverty, lack of intelligence or knowledge, impoliteness, laziness and criminality. Less depressing and equally informative results are provided by the long tradition of experiments eliciting judgements on accents. The British subjects asked to rank the prestige and social attractiveness of the speakers of audio clips labelled as representing different English accents in Coupland and Bishop (2007; see Garrett 2010 for an overview of work on language attitudes) provide remarkably consistent judgements: the production labelled ‘standard English’ and the one closest to the subjects’ own accent were judged most favourably on both dimensions, whereas less favour was granted to immigrant accents (German and French doing much better than Asian English), which nonetheless fared better than some urban vernaculars such as Birmingham’s. It is not just the accent that is at stake, but the groups with which that accent might be associated; consequently, collective social perceptions of those groups can be elicited through such experiments.

15These three methods represent an interdisciplinary approach that offers the potential to reveal the collective knowledge of members of a community, beyond what can be extracted through analysis of their immediate, personal experience of the world. The interdisciplinary nature of this approach allows a robust demonstration of the predicates that are conventionally related to a topic in a community through convergence, and by the elimination of biases introduced by each method. Corpora generally include a large proportion of occurrences that have nothing to do with the relevant discourse matrixes. The strictly controlled inference structures and judgement elicitations may provide answers that are not of direct interest, and relevant answers are amenable to the observer’s paradox as subjects provide the response that they think that the observer wants. Undoubtedly, a more accurate picture can be obtained by bringing together the three methods of corpus linguistics, inference judgement and sentence completion.

16Before we turn to our analytical contribution on the stereotypes of the French in Britain, it would be appropriate to mention a further experimental method that has been carried out using recent technology. Event-related potentials (ERP) are employed by Robert (2009) for the study of syntactic transitivity. The study involves submitting syntactically anomalous sequences such as La concierge bavasse un ragot ‘The concierge dribbled a gossip’ and Marie bavasse un ragot ‘Marie dribbled a gossip’. The anomaly is due to the fact that the verb bavasser ‘to dribble’ is normally intransitive, and generally relates to a bodily fluid as opposed to speech-content. However, these two sequences are received differently by subjects; while the electrophysical indicator N400 relating to semantic difficulty is activated in both cases, the indicator P600 relating to syntactic difficulties peaks with the Marie version, but not with the concierge version. The reason for this appears to be that in France, gossiping is typical of a concierge, an association that a proper noun like Marie does not in itself invite. In other words, the presence of a received idea is enough for the speaker to produce a stable interpretation and disregard the syntactic difficulty. If this reasoning is correct, discursive formation could be tested with a view to revealing that morphosyntactic difficulties can be seen to be overridden in such situations during experimental observations. However, the elaborate nature of such experiments forces us to leave their use to future work.

3. Stereotypes of the French in Britain

3.1 Corpus analysis

17The existence of discourse matrixes is indicated by the frequency of the relevant predicates for a given topic. This frequency can be studied in any extant body of language production where the topic is likely to come up. Studies of the British press have established its reference to stereotypes of the French: military ineptitude, cowardice, a lack of personal hygiene and immorality are predicates used by the populist daily The Sun (Drake 1998). Press articles are readily available in hard copies and increasingly in electronic format, and can be brought together and searched using everyday text processing facilities. They are dated, generally have a named author and target a well-defined national community, unlike on-line resources such as blogs. Furthermore, newspaper articles are believed to largely reflect conventional social discourses, and contribute to reinforce them. They have the supplementary advantage of escaping observer effects, since observation cannot modify already published texts. They should therefore allow stereotypes and other axiological matrixes in a given community at a given time to be documented. However, such stereotypes may not always be as prevalent in the press as one might expect, given that public commercial publications often shy away from axiological characterisations of groups (Fowler 1991) – except possibly in certain article genres such as reader letters and columns. That stereotypes should be found at all in the press would therefore be an indication of their inscription in the larger community they represent. The objectives pursued in this section are to apportion a) whether Britain has stereotypes of the French, b) which ones can be evidenced in the press, and c) whether corpus analysis support the idea of discourse matrix.

18Having established these objectives, on July 12th 2009 we carried out a search of two sequences relating to the ethnonym French in the UK National Newspapers category of Lexi-Nexis UK. The first sequence was ‘all French people are’; it was chosen to see what predicates are applied to the French in general, and all available dates were selected. This search yielded 165 texts, with 111 occurrences of the exact sequence ‘all French people’ (excluding 4 repeated strings). The second sequence ‘is French but’ was chosen so as to elicit concessive predicates about the French. The reasoning was that concessive relations should be found with well-established stereotypes. The necessity for fine-grained analysis led us to limiting the query to the previous six months. This yielded 89 occurrences ‘French but’ (omitting 3 repeated strings). The number of occurrences was significant, while allowing a fine-grained analysis, which is the reason why these particular strings were chosen among all those that could be relevant. The analysis was expected to provide a number of recurring predicates across the two chosen structures if stereotypes on the topic existed in the society where and for whom the newspapers were published, within the limits of the norm and discouraging the expression of negative axiology. In the event that there were no such stereotypes, the predicates were expected to present as divergent and provide for few recognisable patterns.

19Classification was the first step in the analysis of the occurrences found. The data was classified on the basis of the sentence containing the ethnonym only. In most cases, the sentence was a description of particular individuals, things or events that did not present a general property of French nationals. The clear application of an axiological predicate to French people outside ofa particular situation was of primary interest. Of the 111 occurrences of ‘all French people’, 46 concerned the views of French people on a particular thing or event (the euro, the tour de France or serial killer Jacques Mesrine), 48 were associated with political issues (very often French politicians “calling” or “appealing” to the French), and 17 reporting judgements applying to the French in general. These related to behaviour:

(4) "What about le televiseur or le telephone?", I asked. "Well, they’re objects," he replied. "So is la table, yet that’s feminine", I countered. At that point, he did what all French people do in such a situation. "Mais, c’est comme ca," he shrugged.


(5) "He didn’t seem like a very warm man," said one man. "But perhaps all French people are like that."


(6) You and I might mistakenly believe that all French people take lengthy afternoon breaks, but look further south, say the Asterix books.


(7) a. It’s time the Brits stopped being so rude about the French, stopped assuming all French people wear woolly jumpers and berets, carry onions over one shoulder and reek of garlic.

b. And there is a word related to prejudice. Stereotypes. Like all French people have strings of onions around their necks or that all Scottish people play the bagpipes and wear kilts.


(8) a. A national survey into personal hygiene has found less than half of all French people wash themselves all over in the bath or shower every day.


c. But you can’t say that about everybody. It’s the same as saying all French people stink.

24wine and food consumption:

(9) a. But if modest consumption of wine during pregnancy was as bad as they claim, then all French people would be brain-damaged.

b. Van de Velde - a man who applies the preparatory skills to commentary that top-class caddies apply to yardage charts - tries to say something insightful, then Alliss, or his partner in misnomers Alex Hay, interrupts by suggesting that all French people eat insects or describing a "beautiful" shot by "David" Love III.

c. Like all French people, I am a great meat eater, and since I spend most of my time in England, I eat British beef.

d. In fact there are very few places outside France that can carry off a meal of more than half-a-dozen courses. The old school, French-style, full-works, slap-up extravaganza, done real proper, is an unforgettable experience - and one enjoyed by all French people not just the rich.

25To this can be added for illustration the following sequence:

(9’) It is a nation of wine-bibbers, cries the cliché. The statistics tell us half of all French people don’t drink wine at all.

26sexual penchant:

(10) It [lingerie] has naughty connotations in England because of the widely held (and quite correct) assumption that all French people are perverts.

27cultural sophistication:

(11) Because all French people have style and class and we British can bea bit naff.

28or overall evaluation:

(12) Distressed French expatriates point to an increasingly inflammatory media: one headline on the cover of an Australian magazine asked: ’’Are all French people bastards?’’

29Some of the stereotypes are listed and challenged in the following passage:

(13) They go through the pictures together then. No, neither has eaten frogs’ legs, nor would they. The raw meat is disgusting and the fruit looks artificial. The baguette is still very French, although wrapped, square loaves are becoming more popular in the towns. Models of the Eiffel Tower are hateful to all French people, of course, and even the real thing is dismissed in some quarters (though not by these ladies) as ’just metal - how do you say, scrap metal?’

30Notably, example (9) containing the ethnonym also represents an instance where stereotypes are challenged. These predicates intuitively correspond to the British stereotypes about the French, which can be demonstrated by immediate collocations attested on Google (7 520 ‘French shrug’, 21 700 ‘smelly French’, 30 700 ‘wine-drinking French’, 35 100 ‘French sexual’, 8 190 ‘French sophistication’, 7 700 ‘French bastard’, well above combinations with other ethonyms – accessed in May 2010). An exception to this is the predicate warm in (5): this stereotype emerges neither through intuition nor through collocations (only 3 ‘warm French man’). This occurrence seems to report the judgment of a particular writer, as indicated by the edging perhaps, rather than the community as a whole. The collective responsibility for stereotypes (Anscombre 2001) is recognised explicitly in a majority of sequences, through speech verbs (8c), or notions such as assumption ((7a), (10)), prejudices (7b), stereotypes (7b), belief (6), or reported speech ((9b), (9c), (12)), and these collective views are dissociated from the speakers’ in (6), (10).

31The 89 sequences ‘French but’ comprised homonyms – brand mark BUT (1 case), the family names of Dawn French (1), Katy French (1), Philip French (3) – and the language was concerned in 34 cases. The 49 remaining examples related to specific (groups of) individuals (24), things and event (16, out of which 9 related to sporting competitions). The concessive argument focused on something other than the French in 6 further cases, the last 3 providing undisputed evidence of collective axiological judgment.

(14) a. I don’t want to stereotype the French, but their discipline can go if things are not going their way.

b. "Britain is by far - and I mean by far - the best place in Europe to live if you are not white, " he opined.
Well, most of us had guessed as much, but it made a change to hear someone like Mr Phillips say so.
Since devolution was ushered in by an unsuspecting Scottish public, the administrations - mostly Labour - have spent millions of pounds telling us the exact opposite. They may not have said we were actually more intolerant than, say, the French, but their advertising budget - which rose inexorably year by year - was spent trying to make us feel deeply guilty about our inhumanity to our fellow man.

c.  Stockings stuffed with literary references, puppet theatre, manic bickering and emotional frankness, it’s an understatement to call such a movie French. But boy, is it French.

32The above examples refer to collective perceptions of the disposition of French people, invoking character traits such as indiscipline and intolerance; argumentativeness and assertiveness; as well as cultural sophistication, with the primary utterance (a) identifiable as a stereotype. These recur in indirect ways in other sequences; the following examples refer to underhand behaviour:

(15) An e-mail? A blinking e-mail? A furtive little electronic note sent from an anonymous Hotmail account is no way to bring about the end of the Prime Minister of Britain’s career.

It sounds like the kind of thing the Italians might do, or the French. But not the Brits.  That’s just not sporting and shows a complete lack of, well, kugelsack. Plotters in Gordon Brown’s own Labour party are using Internet messages to try to bring about his resignation.

(16) It’s amazing to sell soufflés to the Frenchbut they really appreciate our type of food.

33Both searches established that the British press does evidence stereotypes of the French. The view of the French as either a smelly breed who sport garlands of onions or garlic around their necks, or as culturally sophisticated people primarily interested in sex, wine and food can be reconstructed through 15.5% out of all axiological occurrences of ‘all French people’ in UK Lexis-Nexis. They constitute 6% of the sequences ‘French but’ in a six month period. They are often mentioned rather than used, by being identified through reported speech or knowledge of the community. There is disparity between the percentages that is explained by the apparently more open-ended nature of concessive relation, which can be established on a topic other than what immediately precedes the connector for the string ‘French but’. However, it remains that the presence of stereotypes in a type of publication averse to direct axiological judgments constitutes remarkable evidence of their existence in the society under study.

34Even though the universally quantified ethnonym seems a good way to diagnose stereotypes, the case of example (5) has shown that axiological predicates applied to the national group are not all endorsed by a community. While the proportion of stereotype occurrences is significant, the small number of occurrences for each calls for further enquiry. The question of whether the stereotypes of the French in Britain can be validated by other methods is explored in the next two sections.

3.2 Sentence completion tests

35Conventional predicates about a topic can be documented through public discourses found in newspapers. They raise the question of the extent to which they are an entirely reliable witness of knowledge shared by social actors, given their communication functions and norms. These may therefore usefully be supplemented with other types of data. Supplementary material can be obtained through the protocols devised by social psychology. One such protocol is the sentence completion test (SCT) (Sherry, Dahlen and Holadey 2004; Holadey, Smith and Sherry 2000; Rogers, Bishop and Lane 2004). Extending the word association techniques used in psychoanalysis, SCTs involve presenting subjects with the task of filling in the gaps in incomplete sentences. The gap may be considerable in terms of its semantic potential, as in the Rotter Incomplete Sentences Blank ‘I dislike…’, or very much narrowed down as in Suarez-Orozco’s ‘Most Americans think that [people from my country] are...’. The answers should be provided by subjects as quickly as possible, in order to give access to latent beliefs and escape the observer paradox. Used extensively in personality assessments, the tests have also been employed for the assessment of cultural values and stereotypes. An early example is Golde and Kogan (1959) in their study of the way the young view older adults; attitudes to smoking among college students is explored through the completion of the sentence ‘Smoking makes one...’ by Hendricks and Brandon (2004); the perceived differences between the notions of gender and sex is considered by Pryzgoda and Chrisler (2000).

36The robust and simple method of sentence completion tests confirms the expected stable recurrent predicates of discursive matrixes on a variety of topics. One such topic might relate to national stereotypes. In order to test the existence of stereotypes of the French amongst British nationals, we devised a questionnaire comprising two sentences containing the strings ‘all French people’ and ‘French but’ extracted from the press corpus, to ensure comparability. The sentences invited group judgment (‘Most people think that all French people are...’) and individual evaluation (‘She is French but...’). The same format was adopted for the four preceding filler sentences (‘Most people think that philosophy is’, ‘She studies philosophy but...’, ‘It’s widely believed that fruits and vegetables are...’, ‘He likes fruits and vegetables, although...’). The SCT questionnaires were administered to a group of 21 British Sociology undergraduate second year university students at the beginning of one of their first classes in November 2010. Sociology students were selected because it was felt that Modern Languages students, to whom one author has direct access, might have a reason not to revealwith the hypothesised stereotypes. The subjects were instructed to  “Complete the following sentences. Try to do this fairly quickly.” The results were expected to evidence the stable and recurrent predicates that were found in the press, and the sequence all French people would yield a greater proportion of stereotypes than French but if the press results indicated non-accidental tendencies. These expectations were borne out by the responses. The sequence ‘Most people think that all French people are...’ was completed with eleven different predicates; 38% of answers were represented by rude; followed by romantic at 14% and arrogant at 10%; the other 8 predicates distributed between negative (weak, racist, stuck up) and positive axiology (sexy, fancy, nice, posh); one was entirely neutral (from France), maybe in an attempt to avoid axiology altogether.

‘French people are...’

37As in the press corpus, the predicates in the sequence ‘She is French but...’ are much more diverse. The usual suspect (not) rude came up in only 10% of the answers. They had to do with a constellation of behaviour typically associated to the French – their disposition (19%: she’s not that bad; she’s ok; she is quiet and reserve [sic], all right), language (14%: she doesn’t speak French, she can’t speak French, she does not speak with an accent), food (14%: she doesn’t like eating frogs legs and snails, she doesn’t like cheese, she doesn’t like French food), appearance (10%: she’s fashionable, she is lovely), sexual ethos (10%: she can’t kiss, she doesn’t (French kiss presumably)) and three indications of national domicile (14%: lives in England (2), lives in UK). One answer was leading away from axiology (heisn’t) and one wasn’t provided at all, reinforcing the point that the sequence was not the most revealing in terms of the existence of discourse matrixes.

38The limitations of the concessive individual sequence She is French but ... with respect to the generic attributive Most people think that all French people are ... can be related to their formulation, the characterisation of an individual rather than a group, and to being asked to disclose one’s own values as opposed to one’s group. The concessive sequence does ask for a whole clause to be provided, with no overt requirement that it be axiological, whereas axiology is more likely to occur in an attributive structure in which only the predicate has to be provided. An individual (she) can be engaged in a larger number of contingent situations that a whole group (all French people). Unlike the generic which was related to the values of one’s group (Most people think), the concessive may be read as inviting personal judgment, which subjects might be reluctant to disclose for fear that these might in turn lead to them being judged.

39The social correlates of judgments were difficult to investigate because despite the gender imbalance (4 males and 17 females), the group was relatively homogeneous (all British, all around 20 years of age). It is interesting to note that there was no material difference between the answers of subjects who declared that they “have a direct personal experience of the French” (n = 10) and those who do not  (n = 11). If anything, subjects declaring experience of the French had a blunter use of stereotypes, claiming 7 of the 8uses of rude in the completion of the generic sentence. This may be because the respective subject’s personal experience of the French happened to reinforce the socially shared stereotypes.

40 Diagnosis of discourse matrixes is provided for by sentence completion tests. This method of gathering data provides a higher rate of expected axiology with group judgment than with individual judgement, and a higher rate with generic attributive than with the concessive structures, as in the press corpus. Whether or not these results are replicated by the judgment of inferences is evaluated in the next section.

3.3 Inference judgment task

41Stereotypes have been related to the facilitation of information management and communication.  This is the case in examples elicited in pragmatics studies, in occasional attestations, and in the following press occurrence, whose semantic felicity relies on the stereotype that all French people consume large amounts of wine:

(9) a. But if modest consumption of wine during pregnancy was as bad as they claim, then all French people would be brain-damaged.

42Felicity would be questionable if other topic or predicates were involved (Anscombre 2001, Larrivée 2009: 191), as they go against the socially sanctioned stereotypes.

(9”) a. ?? If modest consumption of wine during pregnancy was bad, then all tea-totals would be brain-damaged.

b. If modest consumption of wine during pregnancy was bad, then all French people would be healthy.

43Stereotypes can thus be brought to light by normative and anti-normative inferential relations. Whilst this has never been tested to our knowledge (however, see Geurts and Pouscoulous 2009), we consequently sought to establish the status of such relations by submitting them to subjects from the target culture, in order to mitigate the impact of individual variation in judgment and establish their expected convergence. We designed the following four sentences with normative predicates rude and arrogant and anti-normative polite and modest, based on the findings of the press corpora.

44It is true that she is rude, but then she’s French.

45The French were very polite, as they always are.

46Since he’s French, he’s very modest.

47The French were arrogant, as they always are.

48The sequences followed the format of individual evaluation and generic attributive structure employed in the preceding judgment test. To keep the questionnaire short, no fillers regarding other topics were provided. Four possible answers were offered – Yes, Maybe, No, Not sure. This task followed the sentence completion test to avoid the predicates provided in the judgment task colouring the predicates chosen by subjects during the SCT. The question put to subjects was “Would you say that the link between the following pairs of sentences is what you would expect? If not, briefly say why.” The expected answers were categorical agreement with the first and last statement, and categorical disagreement with the second and third. Categorical answers were not found in the subject answers, although they followed the previously identified trends. These can be seen from the following graph:

49The normative rude and arrogant received the greatest proportion of outright agreement (43% and 24%, against 5% for polite and 10% for modest). Rude had the lowest disagreement (14%), which was at 19% for modest and arrogant; the highest level of disagreement rested with the anti-normative polite (38%). The anti-normative predicate modest associated to a lesser extent with outright rejection (19%) or agreement (19%) than it did with uncertainty (71% for the combination of maybe and unsure). Remarkably, it is the only term that did not provoke explicit comments about stereotyping. Even though provision was not explicitly made for comments in the available answers, subjects registered that a stereotype was involved for rude (10%), polite (14%) and arrogant (10%), which in turn stopped them providing a judgment on the inference.

50This suggests that the expected stereotypes about French rudeness and arrogance can be validated by inference judgment tests, although not in a categorical way, but at a threshold of around 35%: this threshold is met by the agreement level for rude, and by arrogant if agreement at 24% is aggregated to 10% of notes about stereotyping. The lower than expected threshold can be explained by the attitudes of the subjects. Some explicitly recognised stereotyping and expressed disagreement with it. During a debriefing discussion, the four British adult subjects who took part in a pilot experiment involving the questionnaire raised their uncertainty as to whether the answers were supposed to reflect their own personal beliefs, or those of that of their national group. The subjects’ identification of the presence of stereotyping, and their subsequent negative perception of it, might have meant that they were reluctant to share social judgments for fear of being negatively judged themselves. This observer’s paradox effect might have been heightened by the absence of fillers, the invitation to reflect on the reasons of disagreements (“If not, briefly say why”), and the focus on individual answers (‘Would you say...’, ‘you would expect’). Ways in which this might be avoided would be to focus on the group’s views rather than those of the individual (“Would most people agree that the link between the following pairs of sentences is expected?”); An invitation to answer as quickly as possible, and use of fillers might have led to more categorical answers. The answers to the inference judgment task attest to stereotypes of French rudeness and arrogance in meeting a threshold of about 35% agreement, which is comparable to levels found in the SCT and very significantly higher than even all the stereotypes taken together in the press corpora. A further indication that a discursive matrix is present is provided by the uninvited subject comments that stereotypes are involved. Little or no such comment bore on the anti-normative attributes modest and polite.


51The purpose of this paper has been to propose a discovery procedure of discursive matrixes. If some discourses are conventional and therefore ratified by a group, then the questions that arise are how do we know which ones are thus, and which ones are not? The expected frequency and saliency of predicates for a topic can be assessed by methods of Corpus Linguistics and psycho-social procedures of theSentence Completion Test and the novel Inference Judgment Task. During our research, these were applied to the stereotypes of the French in Britain. The results validate the socially shared representation of the French as rude, arrogant and sophisticated. Such validation is provided by the Corpus Linguistics approach, although the journalistic corpus may not be the most revealing media available as it shies away from certain stereotypes, which are now frowned upon in the public sphere. Confirmation is provided through answers supplied by British subjects to questionnaires verifying the saliency of particular predicates, which are ratified at a rate of about 35%, indicating the presence of an observer effect. This may be why sequences involving judgments on an individual are less revealing than those about groups, where subject responsibility is more obviously defused. Concessive sequences were also less informative than attributive sequences both in the press corpora and in the Sentence Completion Tests (the Inference Judgment Task does not allow to assess this).

52Directions for future research go beyond the validation of the procedure to other topics with more effectively tuned protocols. The consequences for a theory of cognition are significant in showing that knowledge is not entirely accrued by individual experience, and does indeed originate from socially shared views. This may or may not appear obvious to some commentator or other, and the point is that the proposed protocol allows us to demonstrate shared knowledge as a cognitive fact. Socially shared knowledge is significant for language, as it demonstrates how semantic reflexes shaping discourse relate to social knowledge, and illuminate how linguistic meaning can refer to a common experience of the (cultural and material) world. The conventional nature of these views has far-reaching consequences for the everyday behaviour of subjects, notably in terms of rapid production and recognition of messages, and persuasion (see Mercier and Sperber in press):

There are also real advantages in conformity beyond the rewards and privileges that it yields. If one chooses to denounce Qaddafi, or the Sandinistas, or the PLO, or the Soviet Union, no credible evidence is required. The same is true if one repeats conventional doctrines about our own society and its behavior [...]. But a critical analysis of American institutions, the way they function domestically and their internal operations, must meet far higher standards; in fact, standards are imposed that can be barely be met in the natural sciences. One has to work hard, to produce evidence that is credible, to construct serious arguments, to present extensive documentation – all tasks that are superfluous as long as one remains within the presuppositional framework of the doctrinal consensus. It is small wonder that few are willing to undertake the effort, quite apart from the rewards that accrue to conformity and the costs of honest dissidence. (Herman and Chomsky 1988: 305)

Haut de page


Anscombre, J-C. (2001). Dénomination, sens et référence dans une théorie des stéréotypes nominaux. Cahiers de praxématique 36, 43-72.

Anscombre, J-C. and Ducrot, O. (1983). L’argumentation dans la langue. Brussels: Mardaga.

Bakhtin, M. (1981). The Dialogic Imagination: Four Essays. Austin and London: University of Texas Press.

Bar-Tal, D. (2000). Shared beliefs in a society: Social psychological analysis. London: Sage.

Blair, I. V. (2002). The Malleability of Automatic Stereotypes and Prejudice. Personality and Social Psychology Review 6, 246-261.

Bloom, P. (2000). How children learn the meaning of words. Cambridge: MIT Press.

Bodenhausen, G. V. and Wyer, R. S. (1985). Effects of stereotypes in decision making and information-processing strategies. Journal of Personality and Social Psychology 48,2, 267-282.

Coupland, N. and Bishop, H. (2007). Ideologised values for British accents. Journal of Sociolinguistics 11,1, 74-93

Crouzet, F. (2006). Images d’Outre-Manche: la France vue par les Britanniques, la Grande-Bretagne vue par les Français 1904-2004. Histoire, économie et société 1, 131-141.

Drake, D. (1998). Hop off you Frogs: Francophobia and the Sun. Europa 2,2, 1998

Durham, C. A. (2008). Finding France on Film: Chocolat, Amelie and Le Divorce. French Cultural Studies.

Foucault, M. (1972). The Archaeology of Knowledge. London: Tavistock.

Fairclough, N. (2003). Analysing Discourse: Textual Analysis for Social Research. London: Routledge.

Fairclough, N. and Holes C. (1995). Critical Discourse Analysis: The Critical Study of Language. Longman.

Fowler, R. (1991). Language in the News: Discourse and Ideology in the Press. London and New York: Routledge.

Galatanu, O. (2002). La sémantique des possibles argumentatifs et ses enjeux pour l’Analyse du Discours. M. Jesus Salinero Cascante and Inarea I. Las Heras (eds). Actes du congrès International d’Etudes Françaises, vol. 2. Logroño: Universitad de La Rioja. 203-215.

Garrett, P. (2010). Attitudes to Language. Cambridge: Cambridge University Press.

Geurts,  B.  and  Pouscoulous N. (2009). Embedded  implicatures ? Semantics and Pragmatics 2, 4, 1–34.

Gianoni, D. S. (2010). Mapping Academic Values in the Disciplines Subtitle: A Corpus-Based Approach. Berne: Peter Lang.

Givón, T. (2002). Bio-linguistics: The Santa Barbara Lectures. Amsterdam and Philadelphia: Benjamins.

Golde, P. & N. Kogan (1958). A sentence completion procedure for assessing attitudes toward old people. Journal of Gerontology 14, 355-363.

Guillaumou, J. (2006). Discours et événement. L’histoire langagière des concepts. Besançon: Presses Universitaires de Franche-Comté.

Hendricks, P. S. and Brandon, T. H. (2004). Smoking expectancy among college smokers. Addictive Behaviors 30,2, 235-245.

Herman, E. S. and Chomsky, N. (1988). Manufacturing consent. The Political Economy of the Mass Media. Vintage.

Holaday, M., Smith, D. A. and Sherry, A. (2000). Sentence Completion Tests: A Review of the Literature and Results of a Survey of Members of the Society for Personality Assessment.  Journal of Personality Assessment 74,3, 371–383.

Humbert, B. (2003). Screening France. French Politics, Culture and Society 21, 81-94.

Karasawa, M. (2007). Stereotypes as Shared Beliefs: Effects of Group Identity on Dyadic Conversations. Group Processes and Intergroup Relations 10,4, 515-532.

Knox, E. C. (2003). Deja Views: How Americans Look at France. French Politics, Culture and Society 21, 24-32.

Larrivée, P. (2009). Les Français, les Québécois, et la langue de l’autre. Paris: L’Harmattan.

Longhi, J. (2008). Objets discursifs et doxa. Essai de sémantique discursive. Paris: L’Harmattan.

Marcus, G. (2008). Kluge: The haphazard construction of the human mind. London: Faber.

Mercier, H. and Sperber, D. (In press). Why do humans reason? Arguments for an argumentative theory. Behavioral and Brain Sciences.

Morel, C. (2002). Les décisions absurdes. Sociologie des erreurs radicales et persistantes. Paris: Gallimard.

Moscovici, S. (2000). Social Representations: Explorations in Social Psychology. Polity Press.

Paveau, M.-A. (2006). Les Prédiscours. Sens, mémoire, cognition. Paris: Presses Sorbonne nouvelle.

Pêcheux, M. (1969). Analyse automatique du discours. Paris: Dunod.

Poundstone, W. (1988). Labyrinths of Reason. Puzzles, Paradoxes and the Frailty of Knowledge. New York: Anchor Press/Doubleday.

Pryzgoda, J. and Chrisler, J. C. (2000). Definitions of Gender and Sex: The Subtleties of Meaning. Sex Roles 43, 7-8.

Reason, J. (2000). Human error: models and management. British Medical Journal 320, 768-770.

Robert, S. (2009). Construction grammars and brain imagery: an ongoing story. Paper read at the Third International AFLiCo Conference, University Paris Ouest – Nanterre, May 28, 2009.

Rogers, K. E., Bishop, J. and Lane, R. C. (2004). Considerations for the Use of Sentence Completion Tests. Journal of Contemporary Psychotherapy 33,3, 235-242.

Rutland, A. (1999). The Development of national prejudice, ingroup favoritism and self stereotypes in British children. BritishJournal of Social Psychology 38, 55-70.

Sherry, A.,  Dahlen, E. and Holaday, M. (2004). The Use of Sentence Completion Tests with Adults. Michel Hersen, Mark J. Hilsenroth, Daniel L. Segal (Eds). Comprehensive Handbook of Psychological Assessment: Personality and Psychopathology. Hoboken: Wiley, 372-386.

Spanos, N. P.,  Burgess, C. A. and Burgess, M. F. (1994). Past-Life Identities, Ufo Abductions, and Satanic Ritual Abuse: The Social Construction of Memories. International Journal of Clinical and Experimental Hypnosis 42,4, 433-446.

Stalnaker, R. C. (1974). Pragmatic presuppositions. M. K. Munitz and P. K.  Unger (Eds). Semantics and Philosophy. New York University Press. 197-214.

Stubbs, M. (2001). Words and Phrases: Corpus Studies of Lexical Semantics. Oxford: Blackwell.

Stubbs, M. (1998). German loanwords and cultural stereotypes. English Today 14, 19-26.

Suárez-Orozco, C. (2000). Identities under siege: immigration stress and social mirroring  among the children of immigrants. A. C. G. M. Robben and M. N. Suarez-Orozco. 2000. Culture under siege. Collective violence and trauma. New York:  Cambridge University Press, 194-226.

Taijfel, H. (1981). Human Groups and Social Categories. Cambridge: Cambridge University Press.

Teubert, W. (2007). From argumentation to formulaic expressions: a corpus linguistics perspective. Pdf document, 25 slides.

van Dijk, T. A. (1998). Ideology: A Multidisciplinary Approach. London: Sage.

Verdaguer, P. (2004). Hollywood’s Frenchness: representations of the French in American Films. Contemporary French and Francophone Studies 8,4, 441-451.

Wedgewood, D. (2007). Shared assumptions: semantic minimalism and Relevance Theory. Journal of Linguistics 43, 647-681.

Wierzbicka, A. (1991). Cross-cultural pragmatics: The semantics of human interaction. Berlin and New York: de Gruyter.

Wodak, R. (2008). ’The Contribution of Critical Linguistics to the Analysis of Discriminatory Prejudices and Stereotypes in the Language of Politics’, in R. Wodak and V. Koller. Handbook of Applied Linguistics ’The Public Sphere’ 4 Berlin: de Gruyter, 291-316.

Wodak, R. and Meyer M. (2001). Methods of Critical Discourse Analysis: Methods of Critical. London: Sage.

Zlatev, J., Racine, T. P., Sinha Ch. and Itkone E. (2008). The Shared Mind. Perspectives on intersubjectivity. Amsterdam: Benjamins.

Haut de page

Table des illustrations

Légende ‘French people are...’
Fichier image/png, 23k
Légende Liker test
Fichier image/png, 23k
Haut de page

Pour citer cet article

Référence électronique

Pierre Larrivée et Julien Longhi, « The Foundations Of Discourse: The Case Of British Stereotypes Of The French », Corela [En ligne], 10-1 | 2012, mis en ligne le 19 juin 2012, consulté le 16 août 2017. URL : ; DOI : 10.4000/corela.2676

Haut de page

Droits d’auteur

Licence Creative Commons
Corela – cognition, représentation, langage est mis à disposition selon les termes de la licence Creative Commons Attribution - Pas d’Utilisation Commerciale - Partage dans les Mêmes Conditions 4.0 International.

Haut de page
  • Logo Université de Poitiers
  • Logo MSHS de Poitiers
  • Logo CerLiCO