Download as pdf or txt
Download as pdf or txt
You are on page 1of 26

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/317568342

The expression of clause negation: from Latin to Early French

Chapter · April 2018


DOI: 10.1515/9783110551716-014

CITATIONS READS
7 452

1 author:

Maj-Britt Mosegaard Hansen


The University of Manchester
92 PUBLICATIONS   1,164 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

semantic change View project

The evolution of negation (with special reference to French) View project

All content following this page was uploaded by Maj-Britt Mosegaard Hansen on 13 June 2017.

The user has requested enhancement of the downloaded file.


[to appear in Anne Carlier & Céline Guillot-Barbance, eds., Latin tardif/français ancien :
continuïtés et ruptures. Berlin: De Gruyter Mouton.
https://www.degruyter.com/view/product/490123?format=G]

The expression of clause negation: from Latin to early French


Maj-Britt Mosegaard Hansen
The University of Manchester

1 Introduction
This paper will look at the evolution of clause negation from Latin to the earliest stages of
French, with a principal focus on the latter. While the main part of the chapter centers on
forms of standard clause negation (cf. Payne 1985: 198), i.e. the most basic and productive
type of clause negation, which is used in declarative main clauses and which does not involve
quantifiers, some consideration (principally in sect. 5.1.2.5 below) will be given to negative
clauses containing quantifiers. Constituent negation (cf. Klima 1964), on the other hand, will
not be discussed.
The French data analyzed have been sourced from the Base de Français Médiéval
(BFM2013), and includes a choice of sixteen langue d’oïl texts (totalling 337,536 words)
composed on the Continent before 1230. Five texts were composed around or prior to 1100,
nine in the course of the 12th c., and two in or after 1200. Three of the texts are translations
of earlier works originally written in Latin. Together they represent the range of genres and
dialects included in the BFM2013 for this period.1
All the texts in the data base belong to relatively formal domains, being legal, literary,
religious, or historiographic in nature. It is thus an open question how closely the apparent
functions and relative frequencies of forms of negation instantiated in these texts can be
assumed to reflect those that might have been found in a corpus of Early French colloquial
speech, did such a corpus exist. Moreover, ten of the texts are in verse, while only six are
prose texts. It is worth keeping in mind that in the composition of the former considerations
of meter and rhyme will evidently have played a role. On the other hand, Old French verse
texts were meant to be recited orally rather than read, and more generally speaking,
Fleischman (1990: 21) argues that “from the standpoint of its grammar and discourse
structure, Old French is very much a spoken language”.

1
The chosen texts are listed below, followed by an indication of whether they are prose (p) or verse (v) texts, or
translated from Latin (t), the abbreviated title used in examples cited in the main text of this paper, and by their
(approximate) date of composition. The title marked by * has been removed from the BFM data base,
presumably for copyright reasons, after this study was carried out:

Serments de Strasbourg (p, Strasb, 842), Séquence de Sainte Eulalie (v, Eulali, after 881), Sermon sur Jonas (p,
Jonas, 938/952), *Vie de Saint Alexis (v, Alexis, c. 1050), Chanson de Roland (v, Roland, c. 1100), Gormont et
Isembert (v, Gormont, c. 1130), Le couronnement de Louis (v, Louis, c. 1130), Li vers del Juise (v, Juise, 1125-
1150), Enéas 1+2 (v, eneas1/2, c. 1155), Chrétien de Troyes, Erec et Enide (v, ErecKu, c. 1170), Benoît de
Sainte Maure, Chronique des Ducs de Normandie (p., BenDuc1b, c. 1174), Li dialoge Grégoire lo Pape (p, t,
DialGreg1, end 12th c.), Vie de Saint Benoît (p, t, DialGreg2, end 12th c.), Dialogue de l’âme de Saint Isidore
(p, t, DialAme, c. 1200), La queste del Saint Graal (p, Qgraal_cm, 1225/1230).

1
The structure of this chapter is as follows: Sect. 2 introduces some central theoretical notions.
In sect. 3, a brief account is given of negation marking and its evolution in Latin. Sect. 4
moves on to unmarked clause negation in Old French. Sect. 5 discusses marked forms of
negation in that language, and sect. 6 is a conclusion.

2 Negation, markedness, and cyclical developments


The expression of standard clause negation in French is usually cited as an example of the
cross-linguistic phenomenon known as the Jespersen Cycle (Jespersen 1917: 4, Dahl 1979:
88), whereby an original negative marker is reinforced by an additional marker, which
subsequently becomes reanalysed as the principal exponent of negation. The original marker,
having thus been made redundant, may eventually disappear, and the new marker may move
into its syntactic slot. While French has evidently not completed this cycle (and indeed, may
never do so, for reasons discussed in Hansen 2013: 66f), it can be illustrated as in Table 1
below if we take into account certain dialects and French-based creoles:
Table 1: The Jespersen Cycle

1. je ne dis The negator is preverbal

2. je ne dis (pas) The preverbal negator is


optionally complemented by
a postverbal element

3. je ne dis pas The postverbal element


grammaticalizes as part of a
discontinuous negator
embracing the verb

4. je (ne) dis pas The original preverbal


negator becomes optional

5. [Colloquial Québécois]2 je dis pas The negator is postverbal

6. [Lousiana French Creole] mo pa di The previously postverbal


negator migrates to preverbal
position

In French, postverbal pas originates in the polarity-neutral noun pas ‘step’ (< Latin passu(m)),
which underwent grammaticalization as a negative marker in Old French. The data
considered in the present paper represent only Stages 1-2 of the Jespersen Cycle, but see

2
This is not meant to suggest that the construction with only a postverbal negator is not common in colloquial
French as spoken in France. On the contrary, the latter is clearly at Stage 4, and ne-deletion is frequently
observed. Relatively recent studies (Ashby 2001, Hansen & Malderez 2004) show, however, that ne-retention
remains at 10-20% in spontaneous speech in France, whereas it amounted to a mere 1.5% in Québécois already
in the 1970s (Sankoff & Vincent 1977).

2
Hansen (2009, 2013) for discussion of the evolution of standard clause negation from Stage 2
onwards.
The key question with respect to the initial stages of the Jespersen Cycle is why reinforcing
markers are introduced in the first place. Sect. 5.1 below will argue (contra Jespersen’s
[1917:4] original hypothesis) that this was not principally due to the phonetic weakening of
the preverbal marker from non to ne, but that it happened rather for discourse pragmatic
reasons.
As all but the very earliest Old French texts feature variation in the expression of standard
clause negation, the discussion of standard clause negation in sects. 4-5 will make use of the
concept of markedness. Where two different expressions can express similar meanings in a
language, the case may often be made that one of them is “unmarked” with respect to the
other, which is then the “marked” form. As pointed out by Haspelmath (2006), the notion of
markedness is used with a wide range of meanings in the literature, and is not always clearly
defined. It is therefore important to point out that the model used in the present paper is that
of Givón (1990: 945).
This author defines three types of markedness in language: textual, structural, and cognitive.
Textual markedness has to do with the token frequency of different items across text types,
the more frequently used expression being the unmarked one.3 Structural markedness has to
do with form, marked expressions being larger and/or more formally complex than unmarked
ones. Finally, cognitive markedness has to do with semantic/pragmatic function, such that
the marked item will cover only a subset of the possible uses of its unmarked counterpart: for
instance, the English noun bitch is marked with respect to the noun dog, such that, depending
on the context, referents of the latter may be either male or female, whereas bitch can only
refer to a female dog. While the three may go together, such that an expression may be
marked or unmarked on all three parameters, they are in principle independent of one another.
In modern French (corresponding to Stage 4 in Table 1) clause negation does not have to take
the form of standard negation, as defined in sect. 1 above, i.e. using (ne)…pas. It may also be
expressed by one of a small group of quantifiers (so-called n-words, cf. Laka Muzarga 1990)
including personne, rien, aucun, nul, jamais, plus, and nulle part, with or without preverbal
ne, as in (1):
(1) Il (n’)y avait personne dans la salle.
‘There was no-one in the room.’
In this type of case, which we will refer to as quantifier negation, the proposition as a whole
is negated through the quantification of one of its constituents as zero.
Most of the Modern French n-words have originally polarity-neutral etymons: thus, for
instance, the superficial formal identity between the n-word personne (‘no-one/anyone’) and
the noun personne (‘person’ < Latin persona ‘mask/character’) is due to layering (Hopper
1991), the former originating from the latter. Similarly, the n-word rien originates in the now
largely defunct Old French noun riens (< Latin re(m) ‘thing’).

3
I emphasize that it is frequency across text types that is at stake. Clearly, an otherwise marked item may be the
statistically preferred one in certain genres.

3
When looking at the evolution of these two n-words in particular, it is tempting to see the
development of quantifier negation in French as having been subject to a type of cyclic
development not unlike the Jespersen Cycle, represented in Table 2 below.4
Table 2: A possible quantifier cycle

Stage 1. Je ne dis (rien) ‘I do not A polarity-neutral bare NP


say a thing’ optionally accompanies
preverbal ne to make the
scope of the negation
explicit

Stage 2. Je ne dis rien ‘I don’t say ne + Negative Polarity


anything’ Item

Stage 3. Je (ne) dis rien ‘I don’t say N-word optionally


anything/I say nothing’ accompanied by preverbal
ne

Stage 4. [Future French?] Je dis rien ‘I say nothing’ Negative quantifier

I will argue in sect. 5.1.2.5 below, however, that while Table 2 does provide an accurate
account of the evolution of personne and rien, there is increasing evidence against a
generalized quantifier cycle being operative in French.

3 Negation in Latin: a brief overview


The preverbal negative adverb non was the standard marker of clause negation in Latin, as
illustrated in (2), although the markers ne (presumably inherited from Proto-Indo-European)
and haud were also used in certain restricted kinds of context (see Fruyt 2011: 716ff):
(2) Non poteram Cn Pompeium, praestantissima virtute virum, timidum suspicari;…
(Cicero, Pro T. Annio Milone 24,66, 1st c. BC)5
‘I could not suspect Gnaeus Pompeius, a man of the most outstanding valor, of being
timid’
For pragmatic purposes, non could be reinforced by minimizers of various types, such as the
noun mica in (3), which denotes a minimal quantity of bread. In some cases, e.g. gutta
(‘drop’), these nouns were fully conventionalized minimizers, which could be used with
semantically non-harmonious expressions, as in (4) (from Fruyt 2011: 839):

4
The notion of a diachronic cycle involving indefinites was, to my knowledge, first introduced by Ladusaw
(1993).
5
Where nothing else is indicated, Latin examples are sourced from the electronic Perseus data base.

4
(3) quinque dies aquam in os suum non coniecit, non micam panis (Petronius, Satyricon,
1st c. AD)
‘for five days he didn’t put water in his mouth, not a crumb of bread’
(4) quoi neque paratast gutta certi consilii (Plautus, Pseud. 397, 3rd-2nd c. BC)
‘you have prepared no definite plan whatsoever’ (lit.: ‘…not a drop of a definite
plan…’)
Non is generally taken to be a univerbation of ne and the minimizer oinom ‘one’, lit. ‘not one’
(e.g. Fruyt 2011: 709). Etymologically, it is thus itself the result of a cyclic development of
the kind described by the Jespersen Cycle. The source form being a reconstructed one, we
are, however, not in a position to say anything about what may have triggered the cycle in
this case, nor about the nature of the variation that presumably characterized its initial stages.
In addition, Latin had an inventory of quantifying indefinites, both polarity-neutral and
negative, which could appear in negative clauses with nominal or adverbials functions. The
negative indefinites nemo (‘nobody’), nullus (‘no(ne)’), nihil (‘nothing’), and numquam
(‘never’) would negate a clause on their own, cf. (5) below, while polarity-neutral items such
as ullus (‘some’), quis (‘somebody’), or umquam (‘ever’), or the NPI quisquam (‘anybody’)
could be used with similar meaning when in conjunction with the standard negator, as in
(6)(cf. Bertocchi et al. 2010):
(5) …ut intelligatis nihil esse homini tam timendum quam invidiam… (Cicero, Pro A.
Cluentio 3, 7, 1st c. BC)
‘that you understand that there is nothing so much to be feared by a man as envy’
(6) Non fuit illud igitur iudicium iudice simile, iudices, non fuit, in quo non modus ullus
est adhibitur… (Cicero, Pro A. Cluentio 35, 96, 1st c. BC)
‘that court of justice was thus not like a court of justice, o judges, it was not, for in it
there is not any moderation preserved’
In the manner of non, all of the negative indefinites mentioned above are the diachronic
results of univerbation of the old Indo-European negative marker ne with another element.
Thus, we have the following etymologies: nemo < ne+homo ‘man’, nullus < ne+ullus, nihil
< ne+hilum ‘tiny thing’, and numquam < ne+umquam (Fruyt 2011: 710ff). The evolution of
the Latin negative quantifiers is thus not unlike that of Modern French n-words such as
personne and rien (briefly mentioned in sect. 2 above), both being descended from originally
polarity-neutral items that were frequently used in negative clauses. A salient difference,
however, is that, as shown above, the Latin negative indefinites invariably incorporate the
standard clause negator, whereas most of the Modern French n-words do not.
In its use of negative indefinites, Classical Latin is generally acknowledged to have been a
so-called “double negation” (DN) language (similarly to Modern Standard English, for
instance), i.e. one in which two negative markers occurring within one and the same clause
will cancel one another out, resulting in a positive interpretation (e.g. Willis et al. 2013: 30).
Generally speaking, word order would determine the relative scope of the markers, cf. (7)-(8)
(Molinelli 1988: 14):

5
(7) Nemo non videt (M. Cicero Lae 99, 1st c. BC – from Molinelli 1988: 13)

‘No-one does not see’ (→ “Everyone sees”, i.e. ~∃~ or ∀~~)


(8) Non nemo videt

‘It is not the case that nobody sees’ (→ ‘Somebody sees’, i.e. ~~∃ or ~∀~)
There is, however, evidence that these observations were only true of the literary language.
Indeed, outside the literary register of the Classical period, so-called negative concord (NC)
constructions, i.e. structures where two or more negative markers occurring within one and
the same clause give rise to a single-negation interpretation (e.g. Willis et al. 2013: 30), are
attested since the pre-Classical period, cf. (9):
(9) Lapideo sunt corde multi quos non miseret neminis (Ennius, Scen. 139, 3rd - 2nd c. BC
– from Molinelli 1988: 34)
‘Stony of heart are many who do not pity anyone’
While NC initially seems to have characterized the speech of the common people in
particular, observations by grammarians and other commentators who condemned it suggest
that by the 5th c. AD, its usage was widespread (Molinelli 1988: 33-40). Indeed, even highly
literate Classical writers such as Cicero occasionally use NC, cf. (10) below.
(10) debebat Epicrates nummum nullum nemimi (Cicero In Ver. II, 11, 60, 1st c.
BC – from Molinelli 1988: 35)
‘Epicrates owed no money to anybody’
Notice that NC registers of Latin accepted not just combinations of two or more negative
indefinites, as in (10), but also co-occurrence of the standard clause negator non with one or
more negative indefinites, as in (9). While Modern Standard French is likewise an NC
language, combinations of the standard negative marker pas with n-words are excluded (or if
used, result in a DN interpretation), cf. (11)-(12):6

(11) Il (ne)7 devait rien à personne.


‘He owed nothing to anybody.’
(12) Il (n’)a (*pas) pitié de personne.’
6
Indeed, it appears that combinations of the two are generally avoided across languages (Haspelmath 1997:
203f), possibly, as Haspelmath suggests, because indefinites in negative clauses and minimizers like pas, mie etc.
have largely similar extreme-scalar reinforcing functions. In addition, Willis et al. (2013: 42) point out that
minimizers like pas etc. are typically reanalyzed direct objects, which would account for their initial syntactic
incompatibility with indefinite direct objects.
7
I consider ne-deletion in colloquial speech to be part of Standard French, in as much as it has been shown to
not only occur in the production of speakers of all ages and from all social classes, but to actually be
considerably more frequent than ne-retention in this register (e.g. Ashby 2001, Hansen & Malderez 2004). The
term Standard French is used here to distinguish the relevant variety from certain non-standard varieties where
pas and the n-words may co-occur within the same clause.

6
Ungrammatical with the intended NC interpretation ‘He doesn’t pity anyone.’
Possible with a metalinguistic DN interpretation ‘He doesn’t pity no-one’ >
‘He pities someone’.
We shall now proceed to consider the situation in pre-13th c. French, first discussing
unmarked forms, and subsequently the various marked forms of clause negation.

4 Unmarked clause negation in early French


In Old French several different forms of basic clause negation were in competition. From a
perspective of pure textual markedness, one of these forms was at any given time the
uncontroversially unmarked one, i.e. one particular form was the most frequent one, while
any competitors were marked in this sense, i.e. less frequent. In this section, I will be
concerned with the former, while sect. 5 looks at the latter.

4.1 Non
As might be expected, use of simple preverbal non, inherited from Latin, appears to have
been the standard form of clause negation in very old French. Thus, the two earliest texts in
the data base, Les Serments de Strasbourg and La séquence de Sainte Eulalie (both composed
in the 9th c.), use this marker exclusively, without any obvious constraints on the types of
verb that it can co-occur with, cf. (13):
(13) La domnizelle celle koze non contredist (Eulali, v. 23)
‘The maiden did not contradict this’
These texts also contain what appears to be NC constructions involving non and forms of the
indefinite nul, as in (14)-(15) below. However, my use of the hedge “appears to be” reflects
the fact that the status of Old French nul is different from that of its Latin ancestor nullus (see
sect. 5.1.2.5 below).
(14) …in nulla aiudha contra Lodhuuuig nun li iu er (Strasb)
‘of no help against Louis will I (not) be’
(15) Niule cose non la pouret omque pleier (Eulali, v. 9)
‘Nothing could (not) ever make her give in’
It must be noted that both these early texts are very short (115 and 188 words, respectively),
so any conclusions regarding marked and unmarked forms at this stage can only be tentative.
The third oldest text in the data base, the mid-10th c. Sermon sur Jonas, already contains
several instances of the phonetically weakened ne (cf. (16)), while non is found only in the
passages written in Latin:
(16) Et jo ne dolreie de tanta milia hominum si perdut erent dixit. (Jonas, p. 44)
‘And I should not mourn so many thousands of people if they were lost he
said’
7
Indeed, from the 10th c. onwards, all the texts in the data base prefer to use forms of ne rather
than non for basic clause negation. Non continues to be found, but as a both textually and
cognitively marked form, discussed in sect. 5.1.1 below.

4.2 Ne
Compared to non, from the tenth century onwards, there can be no doubt that ne is the
unmarked form of clause negation in terms of textual frequency alone: a search for variants
of this marker in the data base yielded a total of 7,117 tokens with negative adverbial value.8
By comparison, only 25 post-9th c. tokens of non as a finite clause negator were found.
Similarly, in terms of textual frequency alone, preverbal ne appearing on its own appears to
be unmarked in comparison to various forms of discontinuous negation, i.e. forms that
involve ne reinforced by one of the items pas, mie, point, goutte and guère: in the data base,
the latter number 1,350 tokens in total, or 19% of the total number of clauses negated by ne.
That said, performing an automatic search for co-occurrences of ne with various indefinite
zero-quantifiers was not possible due to the very large variety of such indefinites used in
negative clauses in Medieval French, and it was not deemed feasible for the purposes of this
paper to examine all of the ne-marked clauses manually. This means that the total number of
clauses in the corpus where the scope of ne includes an indefinite zero-quantifier is unknown.
While discontinuous negation does occasionally co-occur with indefinites in Old French,
these two forms of negation are on the whole in complementary distribution. In principle,
thus, it might be the case that a majority of the 5,767 clauses which do not contain a
reinforcing element instead contain an indefinite which at least arguably contributes to the
negative marking of those clauses (see sect. 5.1.2.5 below for justification).
In order to support the notion that ne appearing on its own was indeed the textually unmarked
form of clause negation at this stage of the language, a random sample of 300 ne-marked
clauses was therefore examined manually, yielding a total of 88 co-occurrences of ne with an
indefinite within the same clause (29.3%).
Assuming this pattern is generalizable to the corpus as a whole, we can conclude that
between the 10th and the early 13th c., simple preverbal ne was used to negate finite clauses in
roughly half of all cases, while in the remaining half ne was used in combination with either
one of the reinforcing elements listed or with a quantifying indefinite. If we then subtract
30% from the total number of ne-marked clauses, on the assumption that these clauses are
likely to include an indefinite, we are left with a total of 4982 clauses of which 27% (1350)
include a reinforcing element. On that basis, I consider it highly plausible that at least 2/3 of
all instances of “standard” (i.e. quantifier-less, cf. sect. 1 above) clause negation in this early
stage of French would have been marked by preverbal ne alone, and that in frequency terms,
the unmarked status of the simple ne can thus be confidently asserted.

8
The variants searched for were the following: ne, n’, nen, nel, and nes. The latter two forms, and sometimes
nen as well, fuse ne with a pronominal clitic, while n’ and nen are used before a vowel. All these forms, except
n’ , have uses where they do not function as negative adverbials. Such uses were eliminated from the data.

8
In structural terms, the simple ne is evidently also unmarked: compared to non, it is
phonologically reduced, while in addition being morphosyntactically simpler compared to
forms of clause negation which include either a reinforcing element like pas or an indefinite.
From the perspective of cognitive markedness, there does not seem to have been any finite-
clause contexts where simple preverbal ne could not substitute for non or for any of the
discontinuous forms of clause negation, which means that, compared to the latter, ne may at
this stage have been unmarked at the cognitive level as well. Sect. 5 below will argue that the
appearance of alternative forms was indeed governed by specific pragmatic constraints, and
that these forms were therefore marked also in the cognitive sense.
On the basis of data from four Old/Middle French texts, Hansen (2009) shows that, as we
move into Middle French, where bipartite negation gradually becomes generalized, the
simple preverbal ne becomes increasingly specialized for certain syntactic and semantic
contexts, which are highly reminiscent of those in which the form can still occur in formal
registers of Modern French.
As for negation involving indefinites, to the extent that it negates the proposition as a whole
by focusing on the non-existence of one specific essential element of it (be it an argument or
a temporal or spatial adjunct), it is semantically marked in as much as propositions negated in
this way are typically more informative, in the sense of excluding more possible worlds, than
closely related propositions negated through “standard” clause negation. Compare, for
example, the Modern French examples in (17) and (18):
(17) Pierre (n’)aime pas Marie.
‘Pierre does not like/love Marie.’
(Compatible with a range of worlds in which is true that Pierre likes/loves
Françoise/Jeanne/Jacqueline/…)
(18) Pierre (n’)aime personne.
‘Pierre likes/loves no-one.’
(Incompatible with all worlds in which any utterance of the form ‘Pierre
likes/loves N’, where N refers to a human being, is true.)

5 Marked clause negation in texts composed after ca. 900

5.1 Standard clause negation and discourse status


Hansen (2009) and Hansen & Visconti (2009) argue that reinforcement of the negative
marker ne by pas or mie in Medieval French was subject to a discourse-functional constraint,
whereby clauses negated by ne-pas/mie had to express what these authors call “discourse-
old” propositions. Clauses negated by ne alone, on the other hand, were free to express any
type of propositions, whether discourse-new or discourse-old. In terms of the model of

9
markedness used in the present paper, ne-pas/mie were thus cognitively marked forms of
negation in Old French, while the simple preverbal ne was cognitively unmarked.9

The discourse status of propositions is defined following Birner (2006), in terms of inferential
links to prior discourse. According to Birner, inferential links in discourse may work in
either a forwards or a backwards manner. A forwards, or “elaborating”, inference is
automatically invited by a given trigger, where such triggers include explicit prior mention of
the inferable concept or proposition. In Birner’s model, forwards inferences are therefore
both discourse-old and hearer-old. This is illustrated in (19) below, where the notion of a
wedding can be assumed to be triggered automatically upon mention of the word bride:

(19) Sally was the most beautiful bride! Her wedding was certainly the high point
of the summer.
A backwards, or “bridging”, inference, on the other hand, is not automatically triggered upon
mention of a given concept or state of affairs, but is a connection which is made
retrospectively in order to establish discourse coherence. Backwards inferences are thus
discourse-old (to the extent that they are based on prior discourse), but hearer-new (to the
extent that they must be generated actively). This is exemplified in (20), where hearers will

9
In recent work, Schøsler & Völker (2014) claim to falsify the model of Hansen (2009) and Hansen & Visconti
(2009) by adducing an example where different manuscripts show variation between ne-mie and simple
preverbal ne, stating that “variation between the zero form and reinforcement should be excluded, because
discourse-new and discourse-old will be the same in all manuscripts” (Schøsler & Völker 2014: 143). It should,
however, be clear that on a proper understanding of what is meant by saying that ne-mie/pas are cognitively
marked, while ne is unmarked in this sense, variation is in fact expected in discourse-old environments (as
pointed out in Hansen 2009: 244), and thus in no way undermines the hypothesis.

The lone example adduced by Schøsler & Völker (2014: 131), reproduced in (i) below, does not include
sufficient context for readers to be in a position to assess the discourse status of the proposition directly. That
said, one cannot help but notice that version D is substantially different from the other three. This in itself
suggests that discourse-old and discourse-new may, in fact, not be the same in all manuscripts, contra Schøsler
& Völker’s claim. Moreover, in at least one edited version of the text (the 12th c. Charroi de Nîmes,
http://www.russianplanet.ru/filolog/epos/roland/frenchepos/garin/nimes04.htm), the preceding context is as in (ii)
below. Assuming version C of the manuscript to be similar, the mie-marked of Otrant’s utterance can be read as
directly denying the underlined imperative, making it a discourse-old proposition by the criteria set out in
Hansen (2009) and Hansen & Visconti (2009):

(i) A1-A4 Et di(s)t Otran: “De ce ne sai que die” (‘And Otrant said, “Of this I know not what to say”)

B1-B2 Et dist Otran(s): “De ce ne sai que dire” (idem)

C Et dist Otran: “Iche ne sai je mie” (‘And Otrant said, “This I do not know”)

D Et dist Otran: “Grant folie me dites” (‘And Otrant said, “Great madness do you speak”)

(ii) Ce dit Guillelmes a la chiere hardie :/« Otran, fel rois, Damedex te maldie !/Se tu creoies le filz sainte
Marie,/Saches de voir, t’ame sera garie ; /Et se nel fes, ce te jur et afie, /De cele teste n’en porteras
tu mie, /Tot por Mahon, qui ne valt une alie ! » (emphasis added)

‘This said brave-faced William,/ “Otrant, traitorous king, may God curse you!/ If you believed in the
son of holy Mary,/ Know that, in truth, your soul will be healed;/ and if you do not, this I swear to you
and promise,/You will not keep your head despite Mohammed, who is not worth a penny!”’

10
tend to infer that the band in question must have been playing at the wedding, even though
weddings manifestly do not have to involve live music:

(20) I so enjoyed that wedding. The band was fantastic!


Hansen (2009) and Hansen & Visconti (2009) analyze a total of 243 clauses negated by ne-
pas/mie in a corpus consisting of four Medieval French texts, and show that they all express
propositions that are discourse-old in Birner’s (2006) sense. Specifically, the contexts in
which ne-pas/mie are found are of four different types:

1. The negated utterance expresses a direct denial or rejection of part of the preceding
text, as illustrated in (21):

(21) Fols est li reis ki vos laissat as porz. […] “Ultre, culvert ! Carles n’est mie
fol,…” (Roland, vv. 1193, 1207)
‘Mad is the king who left you in these passes. […] “Out of my sight, villain!
Charles is not mad, …”’

2. The negated utterance is a repetition or a paraphrase of part of the preceding text, as


in (22):

(22) segur soiez, ne dotez pas (Eneas1, v.611)


‘be certain, do not doubt’

3. The negated utterance expresses or denies a (pragmatic) presupposition of part of the


preceding text, as in (23), where the act of calling upon someone pragmatically
presupposes that that someone is in a position to help the individual calling upon them:

(23) Lasse, por koi l’apeles ? de sorcurs n’avras mie ! (Juise, v.77)
‘Wretch, why do you call upon him/her? You won’t get any help!’

4. The negated utterance expresses or denies an inference warranted by the preceding


text, as in (24), where the fact that someone’s shield and armour have been badly
damaged in battle standardly suggests that the person may also have suffered bodily
harm:

(24) sil fiert sur sun escu bendé/ k’il la li ad freit e quassé,/ le hauberc rumpu et
desafré;/ mes nen a pas sun cors dampné (Gormont, v. 125)
‘thus he strikes on his banded shield so that he cracked and broke it, fractured and
tarnished his hauberk, but did not harm his body’
Sect. 5.1.2.1 of this paper will adduce further empirical support for this hypothesis. In
addition, the subsequent sects. 5.1.1-5.1.2.4 will show that discourse pragmatics is similarly
relevant to the use of other marked forms of basic clause negation in Old French besides ne-
pas/mie, viz. post-9th c. non, ne-point, ne-goutte, and ne-guère.

11
That this should be so need not surprise us, in as much as the existence of discourse-
pragmatic constraints on marked forms of clause negation is increasingly attested in a number
of modern Romance vernaculars, such as Italian (Visconti 2009), Catalan (Espinal 1993),
Brazilian Portuguese (Schwegler 1988, Schwenter 2006), and arguably also Swiss French
(Fonseca Greber 2007). Moreover, Wallage (2013) shows that Hansen’s (2009) and Hansen
& Visconti’s (2009) model can account for the similar variation between preverbal ne and
bipartite ne-not in Medieval English.10

5.1.1 Non
As noted in sect. 4.2, instances of non scoping a finite verb are rare in the post-9th c. data, a
total of only 25 such examples having been found. In these examples, the verb is normally
one of small subset consisting of être (‘be’), avoir (‘have’), and faire (‘do’), cf. (25) (Reid
1939: 306f, Foulet 1965: 235f):
(25) Sire, por coi m’avez traïe ? – Ge non ai, voir, la moie amie. (Eneas1, v. 1750)
‘My Lord, why have you betrayed me? – I have not, in truth, my friend.’
Attested examples are normally elliptical clauses such as the one in (25), where the lexical
verb and/or one or more complements must be recovered from the preceding context. There
is usually no overt subject, but when there is, non is typically fronted and followed by verb-
subject inversion, as in (26): 11
(26) Ja Dé ne place qu’il m’amor ait ! Non avra il. (Eneas2, v.8489)
‘Never may it please God that he have my love! He will not have it.’

10
Wallage’s (2013) data do not, however, appear to support the assumption that the eventual generalization of
bipartite negation in Middle English can be explained in the way Hansen (2009) and Hansen & Visconti (2009)
propose to explain its generalization in Middle French.
11
Although the data base is too small to allow firm conclusions to be drawn, it is likely that both the normal
word order pattern, whereby non is fronted, while the overt subject – if any – is inverted, and the exception in
(25) can be attributed to information-structural factors, such that non normally appears as a contrastive focus. In
(25), however, the subject arguably constitutes an additional contrastive focus, the wider context of the example,
given in (i) below, being one where Dido has indeed been betrayed, in as much as Aeneas is about to leave her,
but where the Gods, and not he, are ultimately responsible for that betrayal:

(i) Bien sui seüre de morir,/quant ge vos voi de moi partir./Sire, por coi m'avez traïe ?/- Ge non ai, voir, la
moie amie./- Mesfis vos ge onques de rien ?/- Moi n'avez vos fait el que bien./- Destruis ge Troie ?-
Nenil, Greus./- Fu ce par moi ?- Mes par les deus./- Ai ge vos vostre pere ocis ?/- Nenil, dame, gel vos
plevis./- Sire, por coi me fuiez donc ?/- Ce n'est par moi.- Et par cui donc ?/- C'est par les deus, quil
m'ont mandé,/qui ont sorti et destiné,/an Lonbardie an doi aler,/iluec doi Troie restorer. (Eneas1, p. 54)

‘I am quite certain to die/ when I see you leaving me./ My Lord, why have you betrayed me?/ -I have not,
in truth, my friend./ - Did I ever do you any wrong?/ - Towards me you have done nothing but good./ -
Did I destroy Troy?/ - No, the Greeks did./ - Was it at my behest?/ - At that of the Gods, rather./ - Did I
kill your father?/ - No Lady, I assure you./ - My Lord, then why do you flee me?/ - It is not my choice. –
And whose is it then?/ - It is that of the Gods, who have commanded me,/who have drawn lots and
destined/ that I must go to Lombardy/ and must restore Troy there.’

12
The use of non as a finite clause negator is highly pragmatically constrained from the 10th c.
onwards. As Reid (1939: 306) puts it,
[t]he construction […] is used (i) to contradict a preceding affirmative enunciation, or
(ii) to extend the application of, or (iii) to confirm a preceding negative enunciation,
made (or more rarely implied) either by the speaker or by another person.
In other words, as pointed out by Larrivée (2011), non in finite clauses is marked for use in
contexts where the underlying proposition can be assumed to be very recently activated in the
mind of the hearer, in the sense defined by Dryer (1996). The activation of a given
proposition requires that proposition to be explicitly represented in the hearer’s mind, but it is
independent of whether its truth is also endorsed by him. 12 In terms of the framework set out
in sect. 5.1 above, such contexts can be described as both discourse-old and hearer-old.
Ex. (25) above is a textbook example: the utterance produced by the first speaker activates
the proposition “the addressee of the present utterance has betrayed the speaker of the present
utterance at some point prior to the moment of speech”. This proposition is not asserted, but
presupposed by the WH-interrogative format of the utterance. The second speaker’s response,
which is marked by non, explicitly denies the truth of that presupposition.13
A few exceptions to this pattern can be found in the three texts (all composed around or
shortly before 1200) that have been translated from Latin originals, and where isolated
instances of non occur in pragmatically unmarked contexts with the present tense of savoir
(‘know’) and pouvoir (‘be able to’), and with passive forms of trouver (‘find’) and
pourpenser (‘reflect upon’), with être as the auxiliary, cf. (27)-(28):
(27) …car ge lasseiz de la uoie hui cest ior non puis pas eissir. (DialGreg1, p. 23)
‘…for I, being tired from my journey, cannot go out today.’
(28) …si droiture est disjetée et non est trovée… (DialAme, p. 277)
‘…if righteousness is thrown away and is not found…’
These examples presumably reflect an archaic usage, probably facilitated by analogy: as for
(27) and its equivalents, Reid (1939: 307) notes that non as a marked form of clause negation
is very occasionally used with modal verbs, “for which faire is not felt to be a satisfactory

12
As a matter of convention, I use the feminine pronoun for speakers and the masculine pronoun for hearers,
except if authentic examples are discussed where the actual role distribution would make that choice appear odd.
13
Although Godard & Marandin (2006: 192) predict that denials should occur only in root clauses, non may, as
already observed by Larrivée (2011: 1990), also be found in subordinate clauses such as (ii) below. The
rationale for Godard & Marandin’s prediction being that only root clauses are supposed to have speech act value,
it is, however, of interest that in my data non is found only in subordinate clauses governed by a speech act verb:

(ii) …te requist ele que por amende de ce que tu li avoies si meffet devenisses ses hons, et tu deïs que non
feroies… (Qgraal_cm, p.103)

‘…she requested of you that to make up for the harm that you had done to her you should become her vassal,
and you said that you would not…’

13
substitute”. In the case of the two passives, the use of non may have been triggered by the
presence of the auxiliary. In addition, influence from Latin may of course be a factor. It
must be noted, however, that in two cases, the Latin original does not actually use non, but
rather a different, and in one case only implicitly negative, expression (viz. ignoro ‘I don’t
know’ > ge non sai, DialGreg1, p. 7). Together with the fact that, in two other cases ((27)
above being one), these translations employ non in conjunction with postverbal pas (a use
also noted by Reid 1939: 309), this suggests that while non as a productive marker of basic
clause negation may have been archaic by the 12th c., its use was not yet perceived as
distinctly odd by native speakers.
Overall, then, the behavior of non in post-9th c. texts seems to provide incontrovertible
evidence that discourse status could be criterial in the choice between alternative forms of
clause negation in Medieval French.

5.1.2 Reinforced clause negation


Those forms of marked standard clause negation that see preverbal ne reinforced by a
(usually) postverbal marker of nominal origin are, as shown in sect. 4.2 above, considerably
more frequent in the data overall than the use of non. There are, however, significant
discrepancies in the frequency of occurrence of different postverbal markers, pas being by far
the most frequent item used followed by mie, and much further behind, point and goutte.

5.1.2.1 Ne-pas and ne-mie


The data contain 723 tokens of pas and 515 tokens of mie used as negative reinforcers, or in
two cases, NPIs. An example of the latter is reproduced in (29) below:
(29) Tut seie fel, se jo mie l’otrei! (Roland, v. 3897)
‘May I be a complete traitor if I consent to that at all!’
Ne-mie is overwhelmingly favored over ne-pas (by 44 instances to 4) in La chanson de
Roland (ca. 1100). In the remaining texts, the two forms are either roughly equally frequent
(as is the case in Gormont et Isembart and Le couronnement de Louis (both from ca. 1130),
or ne-pas is favored. The latter is the only reinforced form used in the early Vie de Saint
Alexis (ca. 1050), where it is, however, represented by only a single occurrence.
Nine tokens of mie (1.7%) and 20 tokens of pas (2.8%) occur preverbally, as exemplified in
(30) below. The remaining tokens are postverbal. It is worth noting that preposing in my
data base is found only in verse texts, where it may be attributable principally to
considerations of rhyme.
(30) …qui bien aime, pas ne repose ; (Eneas2, v. 9954)
‘…he who is much in love does not rest;’
Five tokens of mie (1%) and one of pas (0.1%) occur in partitive constructions where they
govern a prepositional phrase headed by de, as in (31). All the partitive mie-examples are

14
found in verse texts where the extra syllable provided by the preposition may have been
wanted principally for metric reasons.
(31) De ma fille n’avra il mie, (Eneas1, v. 3376)
‘He’ll not have my daughter’ (lit.: ‘Of my daughter he’ll not have a bit’)
These observations suggest that pas and mie are quite highly grammaticalized already in this
early stage of French: their syntactic position is largely fixed and decategorialization from
noun to adverb has all but been completed. Moreover, the fact that pas and mie sometimes
co-occur with NPIs, as in (32)-(33) below, makes it clear that they are not themselves NPIs,
but an integral part of the negation:
(32) Trestuz les altres ne pris jo mie un guant. (Roland, v. 3189)
‘All the others I don’t consider to be worth a glove.’
(33) Tuit vos Franceis ne valent pas meaille. (Louis, v. 2433)
‘All your Frenchmen are not worth a penny.’
Very sporadically examples are found in which ne-pas or ne-mie enter into what may be an
NC relation with a quantifier (cf. (34) below). As already noted, however, the polarity status
of the quantifiers used in negative clauses is unclear at this diachronic stage (see sect. 5.1.2.5
below for further discussion). The total number of such instances is five in the case of ne-pas,
and four in the case of ne-mie. The data base also includes two examples of pas combining
with the old preverbal form non rather than ne (cf. ex. (27) in sect. 5.1.1 above).
(34) Bien fu gardez et bien serviz, car ne fu pas faite a enviz rien nule qui lui fust
mestiers, mais lieemant et volantiers (ErecKu, vv. 5173-5176)
‘He was well kept and well served, for no thing that he needed was (lit. ‘was
not’) done unwillingly, but gladly and willingly.’
As already discussed in sect. 5.1 above, Hansen (2009) and Hansen & Visconti (2009)
showed that, as far as the pragmatic status of ne-mie/pas is concerned, both markers
systematically occurred in discourse-old contexts in La chanson de Roland and Le
coronemenz de Looïs. According to the analysis of those authors, a sample consisting of the
first fifty tokens of mie and pas, respectively, found in La queste del Saint Graal (ca. 1225-
1230) showed the same pattern.14 For the purposes of the present paper, pragmatic analysis
was in addition carried out on all the examples of ne-pas/mie (487 in total) in La Vie de Saint
Alexis, Gormont et Isambart, Li ver del juïse (ca. 1125-1150), Eneas 1 & 2 (ca. 1155), Erec
et Enide (ca. 1170), and Chroniques des Ducs de Normandie (ca. 1174).
The analysis revealed that, of the 487 clauses in question, 9 (1.8%) went against Hansen’s
(2009) and Hansen & Visconti’s (2009) hypothesis, in as much as they expressed
propositions that were clearly discourse-new. A further 45 clauses (9.2%) expressed
propositions whose discourse status was unclear, in the sense that they could not be classified
as discourse-old on the basis of the immediately preceding text. It is possible, however, that
14
The same was true of the early 14th c. hagiography La vie de Saint Louis, by Sire Jean de Joinville. That text
falls outside the time period considered in the present study, however.

15
such a classification might, in at least some cases, be warranted by consideration of longer
stretches of text.
Conversely, 433 clauses (89%) supported the hypothesis, the propositions expressed being
classifiable as discourse-old. On this basis, it seems that the hypothesis according to which
the use of ne-pas/mie in Old French was subject to discourse-functional constraints, can be
upheld.
Those examples that clearly or apparently went against the hypothesis may be accounted for
by two factors: One is the fact that, as mentioned above, most of the texts in the data base are
verse texts, where the addition of pas/mie may in some cases have provided the extra syllable
needed to complete a verse or constitute a rhyme. The other factor is the gradual
obligatorification of pas which we know took place in the course of medieval French, and
which seems to presuppose a gradual loosening of the original pragmatic constraint on these
markers. 15 In any case, at the stage of French considered in this paper, change in the
expression of standard clause negation is incipient, and wherever a change is in progress,
variation is to be expected, in this as in other domains of language usage.

5.1.2.2 Ne-point
The earliest attestation of negative point in the data is in Le coronemenz Looïs (1130), and the
marker is found in only four texts in total. Moreover, only 39 finite clauses containing point
in negative polarity contexts are attested in the data base. The use of point in bipartite
negation is thus a later development than the use of mie/pas, and it is clearly disfavored by
comparison to the latter two.
One of the host contexts is a weak negative (or affective) polarity context, in which the
marker functions as an NPI rather than a negative marker:
(35) …mes au porter hors fet dongier ;/car qui point en volsist porter/ne s’an seüst
ja mes raler, (ErecKu, vv. 5700-5702)
‘…but taking it out poses difficulty; for whoever would take any of it away
would never be able to leave,’
In a further two instances point occurs in NC constructions with adverbial n-words, either
mes (‘any/no longer’), as in (36), or onques (‘(n)ever’):16
(36) Maboagrins sui apelez,/mes ne sui mes point coneüz, an leu ou j’aie esté veüz,
par remanbrance de cest non, (ErecKu, vv. 6082-6085)
‘Maboagrin is my name, but I’m no longer known at all, in any place that I’ve
been seen, by recollection of that name,’

15
See, however, Wallage’s (2013) analysis of the competition between ne and ne-not in Middle English for the
view that pragmatic “unmarking” was not a factor in the generalization of ne-not in that language. Wallage’s
analysis fails, however, to explain how ne-not could become generalized to discourse-new contexts without the
pragmatic constraints on its use being relaxed.
16
For in-depth studies of these n-words, see Hansen (2012, 2014).

16
The remaining 36 tokens express “standard” clause negation. There are no examples of point
as a minimizer occurring in polarity-neutral contexts. Together these data suggest that it was
very rapidly grammaticalized as a marker of negation. Point does, however, appear to be
somewhat less grammaticalized than pas/mie at this stage, in as much as 22 of the 39 tokens
(56%) are found in partitive constructions either with a prepositional phrase introduced by de
(‘of’), the pronominal adverb en (‘thereof’), or the relative pronoun dont (‘of which/whose’)
as in (37)-(38) below. In such cases, point functions syntactically as a direct object noun
rather than as a negative adverb. Other examples like (39), however, show the marker co-
occurring with a direct object noun, which it does not govern syntactically. Thus, point
appears at this stage to have an intermediate part-of-speech status, not having achieved quite
the same degree of decategorialization as pas/mie.
(37) Al menacier n’a point de hardement. (Louis, v. 866)
‘In threatening there is no courage [lit. not a bit of courage].’
(38) …et quant il veoit qu’il n’en pooit point avoir si s’en retornoit la donc il ert
venuz, (QGraal, p. 150)
‘…and when he saw that he couldn’t have it (lit.: any of it) he went back
where he came from,’
(39) La vostre gent ne puet il point amer. (Louis, v. 830)
‘Your people he cannot love.’
Pragmatically, all of the 39 tokens found in negative contexts occur in propositions that are
discourse-old.
While ne-point thus resembles ne-pas/mie in that respect, it seems to distinguish itself from
the latter markers by being favored with predicates that admit of degrees, although this is far
from being an absolute constraint. Thus, 20 of the tokens of ne-point, incl. those in (35) and
(37)-(39) co-occur with this type of predicate, while a further three co-occur with predicates
that, on at least one plausible interpretation, can be seen as admitting degrees, such as that in
(36), where point could be interpreted as indicating the number of people (namely zero) to
whom the speaker is now known by the name Maboagrin. Together these uses account for
59% of the examples. In comparison, a random sample of 50 clauses negated by ne-pas and
50 clauses negated by ne-mie yielded a total of 27 clauses with predicates that could plausibly
be understood as admitting of degrees. 15 of the clauses negated with ne-mie (30%), and 12
(24%) of those negated with ne-pas were of this type. These contrasting patterns might
explain why later Renaissance commentators described ne-point as being specialized for what
they termed “quantitative” negation, whereas ne-pas was seen by them as preferentially
expressing “qualitative” negation (Catalani 2001: 163).

5.1.2.3 Ne-goutte
My data base contains only eight tokens of ne-goutte. This negative marker is even more
recent than ne-point, being first attested in the data in the mid-12th c. romance, Enéas (1155),
and is found in only two additional texts. Moreover, its distribution is highly restricted at this
stage, all the tokens negating forms of the verb voir (‘see’) used in its basic perceptual sense:

17
(40) La suors lor troble les ialz / et li sans qui avoec degote, / si que par po ne
voient gote, (ErecKu, v. 5932-4)
‘Sweat clouds their eyes / and (so does) the blood that runs down with it, /
such that they almost can’t see,’
While the expression is attested in combination with other verbs (mainly, but not exclusively,
verbs of perception and cognition) in Old French more generally (cf. Price 1990: 204f), ne
voir goutte is generally recognized as being by far the most frequent collocation. None of
the host clauses in my data base contain any complements, and except for six examples of ne-
goutte negating the verbs aimer (‘love’), all the Old French examples cited by Price (1990:
205) similarly have either zero-complements or partitive complements. These observations
suggest that, rather than having been grammaticalized as an actual negative adverb, goutte
may at this early stage of the language have been perceived by many speakers as a
minimizing noun, analogously to Latin gutta in the use illustrated by example (4) in sect. 3
above.
Pragmatically, ne-goutte appears similar to the other bipartite forms discussed above, in as
much as all the examples in the data base are found in contexts where the proposition is
discourse-old. Thus, in (40) above, for instance, the fact that the subjects’ eyes are clouded
by sweat and blood makes it highly inferable that they will be unable to see very well.

5.1.2.4 An intermediate type of negation: ne-guère


The data base contains a total of 65 tokens of the bipartite marker ne-guère. Guère is
assumed to have its etymological origin in the Frankish *waigaro (‘much’, cf. Le trésor de la
langue française informatisée). In Modern French, guère can fulfil both adverbial and
nominal functions in the clause, and my data show that this was also the case in Old French.
In (41) below, guère is thus used adverbially, as a degree marker, in a clause which also
contains a direct object noun (cure ‘concern’), while in (42), it functions as the direct object
of the verb voir (‘see’):
(41) De sa vie n’ot gaires cure, (Eneas1, v. 961)
‘For his life he didn’t care much,’
(42) Ne puis ne fu chevaliers qui gueres en veïst se ce ne fu ausi come en sonjant.
(Qgraal_cm, p. 134)
‘Nor afterwards was there ever a knight who saw nearly anything of them
unless it was as if in a dream.’
The status of guère was and is thus intermediate between that of negative reinforcers like pas,
mie, point, and goutte, and that of indefinites like nul, rien, personne etc. In my data, guère
occurs exclusively in negative polarity environments, but as (42) shows, that includes a weak
(or ‘affective’) negative-polarity context in one instance. Notice, however, that the relative
clause in which guère appears in this example is governed by a negated main clause, together
with which it forms a cleft sentence. This may suggest that, like the former items, but
possibly unlike goutte, guère swiftly grammaticalized as a marker of negation.

18
The data base contains four possible instances of NC constructions, i.e. contexts where guère
co-occurs with one or more indefinites, as in (43):
(43) …nule autre rien ne panse gaire, (Eneas1, v. 1575)
‘she hardly thinks of any other thing,’
Like pas, mie, and point, adverbial guère very occasionally precedes the preverbal negative
marker ne in the linear structure of the clause, as seen in (44) below.
(44) …bien sot que Eneas ventroit et que gaires ne demorroit que il avroit la cite
prise : (Eneas2, v. 9646)
‘…he knew well that Aeneas would win and that it wouldn’t be long before
he’d have taken the city:’
The use of guère does not appear to be constrained to discourse-old contexts in my data.
However, this marked form of negation is pragmatically constrained in a different way,
namely by functioning as a downtoner, which mitigates rather than strengthens the force of
the negation, thereby weakening the speaker’s commitment to the negative state-of-affairs.

5.1.2.5 Ne + indefinite quantifier


If, as shown in sect. 3 above, NC was already common in Late Latin, it is hardly surprising
that it appears to be similarly attested in all the early Romance vernaculars, incl. French. As
already suggested several times, however, the early French data are not as clear as they might
seem at first blush.
For the expression of quantifier negation, Modern Standard French has at its disposal a closed
set of n-words which differ from one another in respect of a small set of syntactic and
semantic features (i.e. whether they function as arguments, determiners, or adverbials, and
whether they refer to human beings or inanimate entities, space or time), viz. personne
(‘nobody/anybody’), rien (‘nothing/anything’), aucun (‘no(ne)/any’ < Latin aliquis unus, lit.
‘some one’), nul (‘no(ne)/any’), nulle part (‘nowhere/anywhere’ < Latin nulla parte
‘nowhere’), jamais (‘(n)ever’ < Latin iam magis, lit. ‘from now on more’), and plus (‘no/any
more/longer’ < Latin plus ‘more’).
In Old French, however, this closed functional paradigm did not yet exist. Instead, several
forms with very different origins were in a number of cases in competition for very similar
functional slots. Thus, personne competed with (among others) ame (‘soul’) and homme
(‘man’), rien with nient (< Latin ne gentem ‘not people’) and chose (‘thing’), jamais with
onques (< Latin umquam ‘ever’), and plus with mais (< Latin magis ‘more’).
Moreover, whereas the n-word status of the Modern French quantifiers is not in doubt, it is
unclear whether their Old French equivalents are polarity neutral, NPIs, or inherently
negative items (n-words or negative quantifiers). As a consequence, it is also unclear when
French can be said to have become the NC language it is today.
As mentioned in sect. 2 above, it seems pretty clear that the Modern French n-words
personne, rien, and aucun have gone through the cyclical development illustrated in Table 2,
starting out as polarity-neutral, and subsequently becoming grammaticalized, first as NPIs
19
and then as n-words (e.g. Déprez & Martineau 2004). It has traditionally been assumed that
this sort of development characterized the n-word paradigm as a whole, with the exception of
nul, which as mentioned above is the direct descendant of a Latin negative indefinite.
Furthermore, it has been claimed (Haspelmath 1997: 230ff) that the evolutionary cline from
positive > negative polarity sensitive > negative is unidirectional as far as indefinites are
concerned.
Recent work has shown, however, that, rather than developing as a group, individual French
n-words have experienced different diachronic trajectories starting at different times, and for
different reasons. Thus, Déprez & Martineau (2004) convincingly argue that changes
elsewhere in the nominal system are ultimately responsible the grammaticalization of
personne, rien, and aucun. For obvious reasons, however, that explanation cannot work for
adverbial n-words like jamais and plus.
Buridant (2000: 614) and Ingham (2011: 445) note that Old French nul is found in weak
negative-polarity contexts such as conditionals, cf (45) below (from Buridant 2000: 614),
where the inherently negative Latin nullus would have been excluded. In other words,
between Latin and Old French, nul evolved from a negative indefinite into an NPI, which is
counter to Haspelmath’s (1997: 230ff) hypothesis, but between Old French and Modern
French, it has moved back in the opposite direction, behaving very clearly as an n-word in the
modern language.
(45) Se tu sez nul deduit fere/ Par quoi tu puisses a gent plere (Rose, 2177-78)
‘If you know how to engage in any pleasant pursuit by which you can please
people’
Labelle & Espinal (2014) show that a number of now defunct Old French indefinites such as
nient and nesun (‘no one’ < neis un ‘not even one’) could similarly be used in weak negative-
polarity contexts despite the fact that they morphologically incorporate the negative marker
ne.
It is for these reasons that it cannot be stated with certainty that exx. (14), (15), (34) and (43)
above do indeed represent NC structures: The first two contain the negative marker non
alongside nul, but the latter’s status as either a negative indefinite, and NPI or an n-word is
indeterminate. The last two examples mentioned contain nul alongside rien, whose status as
a polarity-neutral item, an NPI or an n-word is similarly indeterminate at this stage of
development.
Hansen (2012, 2014) adduces evidence to suggest that the temporal(-aspectual) adverbial
markers jamais and mais/plus were initially grammaticalized with negative meaning in
Medieval French, and only subsequently developed NPI uses, most of which disappeared
again in the case of plus17, whereas Modern French jamais has retained them in more formal
registers at least. Assuming this is a correct account, then examples like (36) above, where
mais co-occurs with point support the assumption that Old French was an NC language of the
same type as colloquial Latin (cf. ex. (9) in sect. 3 above).
Based on these various types of evidence, there can be little doubt that the evolution of
French negative indefinites falsify Haspelmath’s (1997) strong unidirectionality hypothesis.
17
Ne…mais, of course, did not survive beyond Middle French.

20
We therefore need to consider two alternative hypotheses which have been formulated, by
Hoeksema (1998: 104) and Jäger (2010), respectively. The former is a weaker version of the
unidirectionality hypothesis, according to which items can move into, but not out of, the
negative-polarity domain. Hoeksema’s (1998) hypothesis makes no claims about possible
directions of movement inside that domain, however. It thus allows for the movement from
weak to strong negative-polarity sensitivity instantiated by items like rien and personne, as
well as for the movement from strong to weak negative-polarity sensitivity which seems to
characterize items like nul, nient, jamais, and plus. According to Jäger’s (2010) “random
walk” hypothesis, on the other hand, there is no directionality whatsoever to the diachronic
evolution of indefinites, which can move freely in and out of the negative-polarity domain.
In so far as the data adduced in this paper have yielded no examples of negative indefinites,
n-words or NPIs becoming polarity-neutral, the evolution of negative expressions in French
appears to be compatible with Hoeksema’s account, but fails to support the “random walk”
hypothesis. Jäger (2010: 50) does adduce examples of items that have moved out of the
negative polarity domain, but while a handful of these appear to be uncontested (viz. German
jemand ‘someone/anyone’ and immer ‘always’, Dutch ooit ‘once’, and English anymore in
some dialects), Willis (2011: 308) proposes an alternative account of the Slavonic nekto-
series, as well as of a few other purported instances. It thus remains a matter for further
research whether incontrovertible examples of “backwards” developments are sufficiently
numerous to support the assumption that indefinites develop in a genuinely random manner,
or whether there is at least a strong statistical tendency for them to move towards the
affective/negative, rather than the neutral/positive, end of the spectrum.

6 Conclusion
In this paper, I have traced the evolution of the most common forms of clause negation, my
focus being predominantly on so-called standard clause negation, which does not involve
quantifiers. I have argued that Givón’s (1990: 945) theory of markedness is useful in
accounting for the variation that we find among different markers of standard clause negation
in Early French. More specifically, I have shown that the marked forms of standard clause
negation in that language were all subject to discourse-pragmatic constraints on their use,
constraints which in all cases except ne-guère can be described using Birner’s (2006)
classification of the discourse status of propositions.
With respect to quantifier negation, I have suggested that Early French may have been an NC
language of the same type as colloquial Latin, i.e. one that allows combinations not only of
several quantifiers, but also of the standard clause negator + one or more quantifiers to yield
only a single-negation reading. More (ideally carefully quantified) research is needed to
support that assumption, however.
I have also shown that the evolution of French polarity indefinites fails to support the strong
unidirectionality hypothesis put forward in Haspelmath (1997: 230ff), while being
compatible with the weaker hypothesis formulated in Hoeksema (1998: 104). Although other
languages offer limited evidence against even that weaker hypothesis, there does not at
present appear to be sufficient grounds for assuming, along with Jäger (2010), that the
diachronic evolution of indefinites is essentially random, however. Instead, Hoeksema’s
hypothesis might be upheld if reformulated as a strong statistical tendency rather than an
absolute rule.
21
Acknowledgements
I would like to thank the two anonymous referees (one of whom – Johan van der Auwera –
has subsequently chosen to reveal his identity) for their comments and suggestions, and my
former colleague Daron Burrows (Oxford) for helpful discussion of some of my examples.
Needless to say, none of them can be held responsible for any remaining errors in the
presentation. I would also like to thank the volume editors, Anne Carlier and Céline Guillot,
for their time and efforts on this project.

Data bases
Base de français médiéval: http://bfm.ens-lyon.fr/
Perseus Digital Library: http://www.perseus.tufts.edu/hopper/
Trésor de la langue française informatisée: http://atilf.atilf.fr/

References
Ashby, William J. 2001. Un nouveau regard sur la chute du ne en français parlé tourangeau :
s’agit-il d’un changement en cours ? Journal of French Language Studies 11: 1-22.
Bertocchi, Alessandra, Mirka Miraldi & Anna Orlandini. 2010. Quantification. In Philip
Baldi & Pierluigi Cuzzolin, eds., New Perspectives on Historical Latin Syntax. Vol. 3:
Constituent Syntax: Quantification, Numerals, Possession, Anaphora. Berlin: De
Gruyter Mouton, 19-173.
Birner, Betty J. 2006. Semantic and pragmatic contributions to information status. Acta
lingvistica hafniensia 38: 14-32.
Buridant, Claude. 2000. Grammaire nouvelle de l’ancien français. Paris: SEDES.
Catalani, Luigi. 2001. Die Negation im Mittelfranzösischen. Frankfurt-am-Main: Peter
Lang.
Dahl, Östen. 1979. Typology of sentence negation. Linguistics 17: 79-106.
Déprez, Viviane & France Martineau. 2004. Micro-parametric variation and negative concord.
In Julie Auger, Clancy Clements & Barbara Vance, eds. Contemporary Approaches to
Romance Linguistics. Amsterdam: John Benjamins, 139-158.
Dryer, Matthew. 1996. Focus, pragmatic presupposition, and activated propositions.
Journal of Pragmatics 26:475-523.
Espinal, Maria-Teresa. 1993. The interpretation of no-pas in Catalan. Journal of Pragmatics
19: 353-369.
22
Fleischman, Suzanne. 1990. Philology, linguistics, and the discourse of the Medieval text.
Speculum 65: 19-37.
Fonseca-Greber, Bonnibeth Beale. 2007. The emergence of emphatic ne in conversational
Swiss French. Journal of French Language Studies 17(3): 249-275.
Foulet, Lucien. 19653. Petite syntaxe de l’ancien français. Paris: Honoré Champion.
Fruyt, Michèle. 2011. Grammaticalization in Latin. In Philip Baldi & Pierluigi Cuzzolin,
eds., New Perspectives on Historical Latin Syntax. Vol. 4: Complex Sentences,
Grammaticalization, Typology. Berlin: De Gruyter Mouton, 661-864.
Givón, Talmy. 1990. Syntax. A Functional-Typological Introduction. Vol. II. Amsterdam:
John Benjamins.
Godard, Danièle & Jean-Marie Marandin. 2006. Reinforcing negation: the case of Italian.
In Stefan Müller, ed., Proceedings of the HPSG06 Conference. Stanford, CA: CSLI
Publications, 174-194.
Hansen, Anita Berit & Isabelle Malderez. 2004. Le ne de négation en région parisienne : une
étude en temps réel. Langage et société 107: 5-30.
Hansen, Maj-Britt Mosegaard. 2009. The grammaticalization of negative reinforcers in Old
and Middle French: a discourse-functional approach. In Maj-Britt Mosegaard Hansen &
Jacqueline Visconti, eds. Current Trends in Diachronic Semantics and Pragmatics.
(Studies in Pragmatics 7.) Leiden: Brill, 227-251.
Hansen, Maj-Britt Mosegaard. 2012. On the evolution of temporal n-words in Medieval
French. Language Sciences 34(1): 76-91.
Hansen, Maj-Britt Mosegaard. 2013. Negation in the history of French. In David Willis,
Christopher Lucas & Anne Breitbarth, eds. The History of Negation in the Languages of
Europe and the Mediterranean. Vol. I: Case Studies. Oxford: Oxford University Press,
51-76.
Hansen, Maj-Britt Mosegaard. 2014. The grammaticalization of negative indefinites: the
case of the temporal/aspectual n-words plus and mais in Medieval French. In Maj-Britt
Mosegaard Hansen & Jacqueline Visconti, eds. The Diachrony of Negation. (Studies in
Language Companion Series.) Amsterdam: John Benjamins, 187-214.
Hansen, Maj-Britt Mosegaard & Jacqueline Visconti. 2009. On the diachrony of
“reinforced” negation in French and Italian. In Corinne Rossari, Claudia Ricci &
Adriana Spiridon, eds. Grammaticalization and Pragmatics. Facts, Approaches,
Theoretical Issues. (Studies in Pragmatics 5.) Leiden: Brill, 137-171.
Haspelmath, Martin. 1997. Indefinite Pronouns. Oxford: Clarendon.
Haspelmath, Martin. 2006. Against markedness (and what to replace it with). Journal of
Linguistics 42: 25-70.
Hoeksema, Jack. 1998. On the (non)loss of polarity sensitivity. Dutch ooit. In Richard M.
Hogg & Linda van Bergen, eds. Historical Linguistics 1995, vol. 2. Amsterdam: John
Benjamins, 101-114.
23
Hopper, Paul J. 1991. On some principles of grammaticization. In Elizabeth Closs Traugott &
Bernd Heine, eds. Approaches to Grammaticalization, vol. I. Amsterdam: John
Benjamins, 17-35.
Ingham, Richard. 2011. Grammar change in Anglo-Norman and Continental French.
Diachronica 28(4): 441-467.
Jespersen, Otto. 1917. Negation in English and Other Languages. Copenhagen: Høst & Søn.
Jäger, Agnes. 2010. Anything is nothing is something. On the diachrony of polarity types of
indefinites. Natural Language and Linguistic Theory 28: 787-822.
Klima, Edward S. 1964. Negation in English. In Jerry A. Fodor & Jerrold A. Katz, eds. The
Structure of Language. Readings in the Philosophy of Language. Prentice-Hall:
Englewood Cliffs, NJ, 246-323.
Labelle, Marie & Maria-Teresa Espinal. 2014. Diachronic changes in negative expressions:
the case of French. Lingua 145: 194-225.
Ladusaw, William A. 1993. Negation, indefinites, and the Jespersen cycle. In Joshua S.
Guenter, Barbara A. Kaiser & Cheryl C. Zoll, eds. Proceedings of the 19th Annual
Meeting of the Berkeley Linguistics Society. General Session and Parasession on
Semantic Typology and Semantic Universals, 437-446. Berkeley, CA: Berkeley
Linguistics Society.
Laka Muzarga, Itziar, 1990. Negation in Syntax: On the Nature of Functional Categories
and Projections. Doctoral dissertation, Massachusetts Institute of Technology.
Larrivée, Pierre. 2011. The role of pragmatics in grammatical change: the case of French
preverbal non. Journal of Pragmatics 43: 1987-1996.
Molinelli, Piera. 1988. Fenomeni della negazione dal latino all’italiano. Florence: La
nuova Italia.
Payne, John R. 1985. Negation. In Timothy Shopen, ed., Language Typology and Syntactic
Description. Vol. 1: Clause Structure. Cambridge: Cambridge University Press, 197-
242.
Price, Glanville. 1990. The origins and syntax of ne…goutte. In John N. Green & Wendy
Ayres-Bennett, eds., Variation and Change in French. Essays Presented to Rebecca
Posner on the Occasion of her Sixtieth Birthday. London: Routledge, 201-209.
Reid, T.B.W. 1939. Non, nen and ne with finite verbs in French. In Studies in French
Language and Mediaeval Literature: Presented to Professor Mildred K. Pope by Pupils,
Colleagues and Friends. Manchester: Manchester University Press, 305-313.
Sankoff, Gillian & Diane Vincent. 1977. L’emploi productif de ne dans le français parlé à
Montréal. Le français moderne 45: 243-256.
Schwegler, Armin. 1988. Word-order changes in predicate negation strategies in Romance
languages. Diachronica V(1): 21-58.

24
Schwenter, Scott A. 2006. Fine-tuning Jespersen’s Cycle. In Betty J. Birner & Gregory Ward,
eds. Drawing the Boundaries of Meaning. Neo-Gricean Studies in Pragmatics and
Semantics in Honor of Laurence R. Horn. Amsterdam: John Benjamins, 327-344.
Schøsler, Lene & Harald Völker. 2014. Intralinguistic and extralinguistic variation factors in
Old French negation with ne-Ø, ne-mie, ne-pas and ne-point across different text types.
Journal of French Language Studies 24(1): 127-153.
Visconti, Jacqueline. 2009. From “textual” to “interpersonal”: on the diachrony of the Italian
particle mica. Journal of Pragmatics 41: 937-950.
Wallage, Phillip. 2013. Functional differentiation and grammatical competition in the English
Jespersen Cycle. Journal of Historical Syntax 2(1): 1-25.
Willis, David. 2011. Negative polarity and the quantifier cycle: comparative diachronic
perspectives from European languages. In Pierre Larrivée & Richard P. Ingham, eds. The
Evolution of Negation. Beyond the Jespersen Cycle. Berlin: De Gruyter Mouton, 285-323.
Willis, David, Christopher Lucas & Anne Breitbarth. 2013. Comparing diachronies of
negation. In David Willis, Christopher Lucas & Anne Breitbarth, eds. The History of
Negation in the Languages of Europe and the Mediterranean. Vol. I: Case Studies.
Oxford: Oxford University Press, 1-50.

25

View publication stats

You might also like