Fine-tuning Jespersen’s Cycle

1

Fine-Tuning Jespersen’s Cycle*

Scott A. SchwenterThe Ohio State University

1. Jespersen’s Cycle and its parts

“Jespersen’s Cycle”, a term apparently first used by Östen Dahl in a 1979 paperpublished in the journal Linguistics, refers to the cyclical process by which sentencenegators increase and reduce their formal complexity in different historical steps orstages. In the original formulation of this process, Otto Jespersen described what wouldlater become his “Cycle” as follows:

The history of negative expressions in various languages makes us witnessthe following curious fluctuation: the original negative adverb is firstweakened, then found insufficient and therefore strengthened, generallythrough some additional word, and this in its turn may be felt as thenegative proper and may then in course of time be subject to the samedevelopment as the original word (Jespersen 1917: 4; see also Jespersen1924: 335).

Jespersen exemplified the workings of the Cycle with data from French, English, andDanish, three languages in which the cyclical process of weakening, strengthening, andreplacement could be found (albeit not exactly to the same degree). However, it is, Ibelieve, fair to say that the paradigm case of Jespersen’s Cycle, and definitely the onethat has been repeated most often in the literature, is that of French. Although the level ofdetail at which the French case is presented varies widely (cf. Hopper and Traugott2003), the overall characterization of the change is typically quite uniform amongresearchers. In one frequent way of presenting the diachrony of the change, a generalschema of four basic structural stages for the French negative cycle is offered, much as in(1) (cf. Schwegler 1988, 1990; Geurts 2000; Ladusaw 1993; Posner 1985):

(1) Stage 1. NEG + VERB Je ne sais. ‘I don’t know’Stage 2. NEG + VERB + EMPHATIC NEG Je ne sais (pas).Stage 3. NEG + VERB + OBLIGATORY NEG Je ne sais pas.Stage 4. VERB + NEG Je sais pas.

*It is a pleasure and an honor to contribute to this volume in honor of Larry Horn. It is fair to say that Larryhas been as supportive as anyone of my research—he’s really had no reason to be so supportive, but (#so)supportive he has been! For comments on issues in this paper, I am grateful to Larry, Elizabeth Traugott,Patrícia Matos Amaral, Chad Howe, Michael Israel, Craige Roberts, the audience at the 2004 LSA AnnualMeeting in Boston, where it was originally presented, and to the editors of this volume. Any and all errorsof fact or interpretation are solely the fault of the author.

2

It should be noted that there is a well-known intermediate stage, which we might callStage 3/4, wherein the preverbal negative (e.g. French ne) is variably realized. Indeed, inthe colloquial spoken language of the present-day, French negation is found at thisintermediate Stage 3/4 (Armstrong 2002; Ashby 1981, 2001; Coveney 1996; Martineauand Mougeon 2003; Sankoff and Vincent 1980).1

Jespersen’s Cycle (usually as exemplified by French), along with the development ofthe be going to future in English, is without doubt one of the standard textbook examplesof grammaticalization (inter alia: Dahl 2001a, 2001b, 2004: 137; Detges and Waltereit2002; Geurts 2000; Hopper 1991; Hopper and Traugott 2003). From asemantic/pragmatic point of view, the rise of obligatory post-verbal negatives—depictedin Stage 3 above—is widely considered to be due to a gradual loss of the “emphatic”value that the post-verbal element conveys in Stage 2: this element (e.g. pas), over timeand through frequent use (leading to “overuse”), loses its emphatic value and is thereforereanalyzed as an obligatory element within the canonical sentential negation construction.

The principal goal of this paper is to show that this position is problematic from thepresent-day perspective of other Romance languages which have post-verbal negativesthat cannot be strictly considered either “emphatic” (Stage 2), unless this term is definedin an extremely restricted fashion, or “obligatory” (Stage 3). Instead, the argument to beadvanced is that there is a clearly identifiable stage wherein the post-verbal negativeelement is heavily regulated by information-structural factors, and specifically by thediscourse-old status of the denied proposition.

Although the data and analysis to follow are strictly present-day synchronic in nature,an important assumption of this paper is the correctness of the uniformitarian principle(Labov 1975), which maintains that “the linguistic forces that are evidenced today are inprinciple the same as those that operated in the past” (Hopper and Traugott 2003: 50). Inbasic terms, what I intend to show is that there is a good deal of commonality among thesynchronic states of negation in several Romance varieties, and that what ties these statestogether is their reliance on information structure, not emphasis (or presupposition; seebelow). This kind of investigation comparing the present with the (presumed) pastalready has precedent in the realm of negation and Jespersen’s Cycle, but from a strictlysyntactic perspective. Indeed, Zanuttini has already pointed out that, “since thesynchronic stages [of the Cycle] correspond so closely to the diachronic ones, aninvestigation of the former will help us better understand the latter” (1997: 14). To mymind, what’s good for syntax is also good for pragmatics!

The remainder of this paper is structured as follows. In Section 2 I discuss the notionof emphasis and argue that it is inadequate for a complete analysis of noncanonicalnegative constructions in Catalan and Italian. Instead, I demonstrate that theseconstructions are sensitive to information-structural properties of the proposition beingdenied. In Section 3 I examine a related, yet more complex, example of noncanonical 1 There are, however, broad differences between different dialects of spoken French with respect toNEG2/3. Thus, for example, Canadian French negation is overwhelmingly (99%) NEG3 (Sankoff andVincent 1980), while this percentage is much lower in European French, which conserves NEG2 to a muchgreater extent (approximately 20% to 40%) in the spoken language (Ashby 1981; Coveney 1996).

3

negative constructions from Brazilian Portuguese, and show how these forms are likewisesensitive to information structure and specifically how they differ with respect to thediscourse accessibility of the denied proposition. Section 4 provides concluding remarksabout the analysis and its relation to Jespersen’s Cycle. Finally, Section 5 is a briefpostscript in which I discuss a larger issue concerning form-function pairings in the realmof sentential negation. Throughout the discussion, I will use the labels NEG1, NEG2, andNEG3 as abbreviations for three types of sentential negative constructions: NEG 1 refersto strictly preverbal negation (i.e. canonical negation in the languages of interest here);NEG 2 refers to negation with co-occurring pre- and postverbal negative elements, suchas ne V pas in French; and NEG3 refers to strictly postverbal negation, e.g. V pas inFrench. Although these formal options have traditionally been linked to distinct stages ofJespersen’s Cycle, in this paper I am primarily interested in noncanonical negatives atStage 2 of the Cycle, i.e. the stage where the postverbal negative element is optional.

2. NEG2 in other Romance languages: beyond “emphasis”

The term most often associated with noncanonical negation in the early stages ofJespersen’s Cycle is EMPHASIS: the post-verbal negative element adds emphasis to thecanonical pre-verbal sentential negative. To the best of my knowledge, however, thenotion of emphasis associated with post-verbal elements in the Cycle—corresponding toStage 2 in (1) above—has never been defined in explicit terms. Indeed, althoughemphasis is a label that abounds in linguistic research, it is typically undefined no matterwhat its domain of application. As regards Jespersen’s Cycle specifically, Schwegler(1990: 158) has noted that “there seems to be a constant and universal psycholinguisticneed for negative emphasizers”. Schwegler’s footnote to this same statement, however,exemplifies the aura of mystery that surrounds these so-called emphasizers:“‘Psycholinguistic proclivity’ is not an explanation. It is really an unknown for a causalfactor whose existence must, however, be acknowledged for an understanding of theprocess involved” (Schwegler 1990: 239).

It is not clear exactly what stage(s) of the Cycle Schwegler is referring to in thepreceding quotes. What must be distinguished, however, are the possibilities for emphasis(in intuitive terms) that constitute the precursors or input to the Cycle proper, and theemphatic post-verbal elements that become incorporated as GRAMMATICAL elements intothe Cycle, such as in Stage 2 in (1). The first kind of emphasis and their correspondingemphasizers are a seemingly universal property of language(s): the possibility of usingpost-verbal elements, most typically nouns denoting minimal quantities, in emphaticnegation contexts as negative polarity items (NPIs). As pointed out by Detges andWaltereit (2002), it is elements such as minimizing NPIs like a crumb or idiomaticminimizers like lift a finger that are widespread across languages. However, as noted bymany authors (e.g. Hopper [1991] for French), typically only one of these minimizersbecomes generalized for sentential negation, in the sense of extending its domain ofapplication to all kinds of predicates and of bleaching its meaning so that the erstwhilenominal meaning is no longer tangible. Thus, French pas was semantically extended

4

beyond its nominal meaning (‘step’), ultimately to the point where it could combine withany verb, not just verbs of motion, which are the assumed starting point for this particularemphasizer. On the other hand, English a crumb is still heavily constrained by themeaning of its source noun, so that it is permissible to talk about not eating a crumb, but*not walking a crumb remains impossible. The French and English forms illustrateopposite ends of the semantic spectrum for post-verbal negative elements: while pas inthe negative construction has lost its nominal meaning, a crumb retains its nominalmeaning in the negative construction. In between these poles, however, lie many otherpossibilities. Thus, the English NPI jack(shit) (cf. Postal 2004) is fairly general withcertain verbs, e.g. He doesn’t know jack, My son won’t eat jack anymore. However, itdoes not combine well with all verbs, e.g. ?I don’t sleep jack, and likewise still seems toconserve the “minimal quantity” meaning. The upshot of all this is that there is a fine-grained semantic/pragmatic continuum of post-verbal elements, and defining them asbeing emphatic at Stage 2 does not help resolve the issue.

On the bright side, one of the few attempts, and perhaps the only attempt, to defineemphasis in the realm of semantics and pragmatics more generally has been made byMichael Israel in his dissertation (1998) and a series of papers since then. Given the lackof alternative explanations for the concept in question, I assume, following Israel (1998,2001), that emphasis describes the high informativity of a proposition relative to a scalar“norm”—“if a proposition entails the norm, its assertion is informative because it exceedswhat one would normally expect to be asserted” (1998: 47). The “normally expectedassertion” of Israel’s model is the canonical, unmarked sentential negative. Thus, theinformative value of a marked negative (plus NPI) like “I didn’t move a step/at all”entails that of the unmarked negative “I didn’t move”, but not vice versa (cf. also Detgesand Waltereit 2002). Therefore, the marked negative is emphatic relative to the unmarkedform.

Most crucial to the point of this paper, however, is the fact that emphatic negatives inIsrael’s sense are not necessarily linked to information structure, or, to put it morecolloquially, they do not necessarily “respond” to anything or anyone. While it may betrue that emphatic NPI minimizers are most often employed in contexts where there is anassumption that the denied proposition is true (see Detges and Waltereit 2002),2 there isno restriction such that these NPIs MUST occur in such a context. Thus, an example with aminimizer NPI like I didn’t sleep a wink last night could be uttered felicitously in Englishwith no necessary prior assumptions on the part of any of the interlocutors about thequality of the speaker’s sleep the night before. Upon uttering this sentence, the speakerdoes create (or invoke) a context of counterexpectation, contrasting what she just assertedwith shared norms (e.g. having slept well, having slept some, not sleeping well but morethan a wink, etc.). The crucial part of Israel’s analysis is simply that this utterance has tobe understood as expressing an informative proposition that entails other propositions inthe relevant scalar model, and, crucially, as denying the smallest possible quantity of

2 I know of no studies examining the use of NPIs in naturally-occurring discourse that could either supportor refute this prediction.

5

sleep. These aspects of interpretation are what justify the classification of sleep a wink asan emphatic NPI.

However, there are present-day Romance languages whose NEG2s are neitheremphatic in the sense defined by Israel nor obligatory elements. We’ll first consider twocases that are rather similar: Catalan and Italian (see Schwenter 2002). By grouping thesetwo languages together, I am not being innovative: NEG2 in Catalan and Italian havealready been considered similar by other authors, e.g. Bernini and Ramat (1996: 17), whostate, “In Catalan post-verbal pas is used to give the negative construction … adversativemeaning on the pragmatic level; its use, comparable to that of mica in Italian, implies thatthe speaker presupposes that whatever he is denying is on the contrary considered true orunderstood as realizable by his interlocutor”. While I agree with Bernini and Ramat intheir assessment of the comparability of NEG2 across the two languages, my analysis ofwhat is “implied” by their use is distinct.

2.1 Catalan

The case of Catalan would appear to be of utmost importance for comparative Romancepurposes, since both the NEG2 post-verbal element (pas < ‘step’) and its syntax (no Vpas) are essentially identical to that of the more well-known French negative. However,to my knowledge, none of the authors who have analyzed negation and the negative cyclein French—from either a diachronic or a synchronic perspective—have drawncomparisons with the neighboring Catalan negatives. Catalan could be especiallyimportant for hypothesizing about the details of the path of change taken by Frenchnegation, assuming (as one would be inclined to do, given the linguistic and geographicalcloseness of the two languages) that Catalan is following or at least has followed asimilar path of development to this point.

The main difference between the French and Catalan NEG2 constructions is that thepost-verbal negative pas is not obligatory in Catalan, or at least not in those Catalanvarieties (i.e. the vast majority) spoken in Spain, which are very obviously not in asituation of bilingualism with French but rather with Spanish. Since Spanish is alanguage with no “dominant” NEG2 construction, one would assume that Spanish cannotbe providing a model for transfer into the Catalan construction. Moreover, as far as Iknow no one to date has claimed that the Catalan negative system is actively changing,i.e. that it is on its way to an obligatory NEG2 stage (e.g. Stage 3 in [1] above), althoughsuch a stage is certainly a future possibility.

In existing descriptions of the Catalan negative system, no V pas has been consideredan emphatic construction by some authors (Hualde 1992; Yates 1993[1975]), but not byothers (Espinal 1993; Wheeler, Yates, and Dols 1999). Clearly, as in the case of itsFrench counterpart, the erstwhile direct-object noun pas in the NEG2 construction hascompletely lost its former meaning of physical movement (meaning still preservedoutside this construction in the noun pas). As a result of this meaning change, the particlehas been generalized to all classes of verbs, including intransitives and copulas (Jordi noés pas intel.ligent ‘Jordi is not intelligent’).

6

At the same time, however, this semantic and syntactic generalization of pas is notunrestricted, and the NEG2 no V pas is clearly not acceptable everywhere the canonicalNEG1 no V can be used. Rather, the distribution of no V pas is strictly controlled byinformation-structural factors, as the following examples illustrate:

(2a) [Stepping outside on a fairly warm day, after several days of unusually coldweather]Avui no fa (#pas) fred!today not makes cold‘Today it’s not cold!’[Add morpheme-by-morpheme glosses]

(2b) [speaker B to A; B believes that A believes the cold weather will continue]Avui no fa (#pas) fred!‘Today it’s not cold!’

(3a) [Same weather scenario as in (2a)]A: Avui fa fred també?

today makes cold also‘Will be it cold today as well?’

B: No, avui no fa (pas) fred.no, today not makes (NEG) cold‘No, today it’s not cold.’

(3b) [speaker B sees interlocutor A putting on a heavy coat]Avui no fa (pas) fred.‘Today it’s not cold’

Notice first that post-verbal pas is infelicitous in both (2a) and (2b), while thecanonical NEG1 is normal. By contrast, NEG2 is felicitous, but not required, in both (3a)and (3b). In general terms, the difference between the examples in (2) and those in (3) isthat in (2) there is no “trigger” element in the discourse to license pas. The speaker-internal counterexpectation context in (2a) is not sufficient to license pas, and not even astrong expectation about an interlocutor’s belief-state, as in (2b), is sufficient. If thebehavior of pas were in fact licensed by speaker emphasis (per, e.g., Israel’s definition),then it would be difficult to explain why the utterances in (2a, b), situated as they are inemphatic contexts that run counter to speaker or interlocutor expectations, areinfelicitous. What the examples illustrate is that there must be a salient proposition,evoked either linguistically (3a) or situationally (3b), for pas to be felicitous (cf. Prince1981). Following Prince’s (1992) model, we can say that NEG2 is sensitive to thediscourse-status (3), not the hearer-status (2) of the denied proposition. Also importantis the salience of the denied proposition, i.e. not only must this proposition be discourse-old, it must also be salient in the discourse at the time that the speaker utters NEG2. Inspoken discourse, NEG2 is uttered as a denial of a proposition that can be derived from

7

an immediately preceding utterance, usually that of another speaker, as in (3a), or as aresponse to a salient non-linguistic action from which a particular proposition can beinferred, as in (3b).

The distinction drawn by Dryer (1996) between activated and believed propositionsis also applicable to the Catalan data. According to Dryer, the activation of a propositionis independent of one’s belief in the truth of that same proposition: there are manypropositions in whose truth we believe that are not at all activated in the discourse, andothers may be activated but not believed. Such is the case, for instance, in (3a) above,where speaker A’s question activates the proposition “It will be cold today”, but does notexpress A’s belief in that same proposition. As B’s reply to A in the same exampleillustrates, the Catalan NEG2 is felicitous in denials of propositions that are activated butnot necessarily believed. (See also Caton [1981] and Fretheim [1984] on illocuted vs.propounded propositions.)

In some descriptions of NEG2 in Catalan (e.g. Yates 1993[1975]), the constructionhas also been described as a marker of counterexpectation. But an interesting property ofno V pas is that it can actually be used to agree with a prior negative statement (Espinal1993):

(4) A: La Maria ja no vindrà a aquestes hores.the Maria already not will.come at these hours‘Maria won’t be coming at this hour.’

B: Efectivament, la Maria no vindrà pas tan tard.indeed the Maria not will.come NEG so late‘True, Maria won’t come so late.’

For Espinal (1993), this particular use of NEG2 for confirmation is separate from thedenial use seen in the examples in (3) above. The present analysis suggests, however, thatinstead of considering this use a separate function of NEG2 (pace Espinal), it can be seenas another instance of the denial of an activated discourse-old proposition. Thus, in (4),the affirmative proposition denied by B (“Maria will come at this late hour”) is activatedindirectly by A’s prior (negated) utterance. That is, from A’s utterance it is possible toinfer a point of view that holds that Maria will be coming at the late hour in question. Thefunction of no V pas in B’s reply is to index the negation as replying to this affirmativepoint of view, i.e. the SAME function it displays in (3). The confirmatory interpretation ofno V pas in (4) is merely an indirect feature of the normal meaning/function of no V pasin this particular kind of discourse context.

2.2 Italian

Let us now turn to a very similar case that strengthens the claim that informationstructure is crucial for a correct characterization of the meaning/use of NEG2, at leastwhen it corresponds to Stage 2 in (1) above. The form of NEG2 in (standard) Italian (Iwill not deal here with the wide variety of dialectal variants) is non V mica. The post-verbal element derives from an erstwhile noun meaning ‘crumb’ (cf. Fr. mie, Sp. miga).

8

However, in present-day Italian it has generalized its domain of use so much that nowfunctionally it resembles Catalan NEG2 to a great extent. Molinelli (1987: 170) alsonotes that there is substantial NEG3 use of mica in popular spoken Northern Italian.

In her groundbreaking monograph on the syntax of negation in Romance (mainlylimited to Italian and other Romance varieties spoken in Italy), Zanuttini presents thefollowing pair of examples, contrasting the standard NEG1 form (non V) with thenoncanonical NEG2 form (non V mica):

(5a) Gianni non ha la macchina.Gianni not have the car

(5b) Gianni non ha mica la macchina.‘Gianni doesn’t have the car.’

Zanuttini describes the pragmatic difference between these two examples in the followingmanner:

As was first discussed in Cinque (1976), the occurrence of mica ispragmatically restricted to those contexts in which the non-negativecounterpart of the proposition expressed by the sentence is assumed inthe discourse. For example, in order for mica to be uttered felicitously in[(5a)], it is necessary that the proposition that Gianni has a car be entailedby the common ground. If such a proposition is not part of the commonground, the presence of mica renders the sentence infelicitous and itscounterpart without mica must be used…(1997: 61; emphasis added)

This explanation is obviously much more explicit than an intuitive one that statessimply that non V mica is more emphatic than its NEG1 counterpart. However, it alsohinges on the definition of “common ground”, and unfortunately Zanuttini does notclarify how she is using this term. To see why such a clear understanding of commonground is needed, compare (6a) and (6b), which provide greater (constructed)contextualization for the NEG2 example in (5b) above:

(6a) A: Chi viene a prenderti?who comes to grab-you‘Who’s coming to pick you up?’

B: Non so. Ma Gianni non ha (#mica) la macchina.not I.know but Gianni not have the car‘I don’t know. But Gianni doesn’t have the car.’

(6b) A: Chi viene a prenderti, Gianni?‘Who’s coming to pick you up, Gianni?’

B: Non so. Ma Gianni non ha (mica) la macchina.‘I don’t know. But Gianni doesn’t have the car.’

9

Most notable in this pair of examples is that mica is infelicitous in (6a) even if theinterlocutors share the common ground—which following Stalnaker (1978) we couldstandardly characterize as the set of propositions that the interlocutors hold in common tobe true (see also Kadmon [2001: 9])—that Gianni has a car, that he usually picks B upwith that car, etc. Thus, Zanuttini’s “entailed by the common ground” (again, on theStalnakerian view of common ground) is actually too weak a characterization of thefelicitous use of mica, insofar as the postverbal NEG is infelicitous in (6a) despite the factthat Zanuttini’s constraint is met. By contrast, in (6b) mica is felicitous (though notobligatory) precisely because the proposition “Gianni is coming to pick up B” is activatedand salient in the discourse context. Thus, mica can be employed felicitously when thenegated proposition is accessible or entailed in the common ground of the discourse, butthe manner in which that proposition becomes part of the common ground is crucial tothe felicity of mica; it is NOT the case that any proposition entailed by the commonground can be negated using mica. Rather, as was the case in Catalan, the propositionbeing denied by NEG2 must be discourse-old (whether introduced linguistically or not)and salient in the discourse context. In addition, as (6b) shows, prior belief in thecorresponding affirmative proposition (“Gianni has the car”) is not a necessaryrequirement on the use of NEG2; rather, activation of that proposition (even viaimplicature) is sufficient. Finally, it is also notable that, like the Catalan NEG2construction, Italian non V mica is possible in a “confirmatory” context like the one in (4)above.

In summary, the NEG2 constructions in Catalan/Italian are licensed only when thereis a denial of a salient discourse-old proposition. This discourse-old proposition does nothave to be believed by anyone; all that is required is its activation (Dryer 1996) in thediscourse context. While it may indeed be the case that these NEG2s are used primarilyin emphatic discourse contexts (though this is an empirical question in need of ananswer), the notion of emphasis itself—whether on an intuitive level or explicitly definedà la Israel—cannot account for the information-structural constraints that regulate theseconstructions. Indeed, it is these constraints that actually make the negative constructionsin question unavailable to many more types of intuitively emphatic contexts, such asthose illustrated in (2) above.

As regards Jespersen’s Cycle, the NEG2 constructions in both languages are clearlyat Stage 2 with respect to their generality. The post-verbal negative elements have lost itsnominal meaning and have generalized to all kinds of verbal predicates, but they are notyet an obligatory concomitant of the pre-verbal negative, as they would be at Stage 3.The important point to take away from the discussion and analysis is that themeaning/function of NEG2 (or, more narrowly, the post-verbal negative element) atStage 2 of the Cycle is not done justice by labels such as emphasis. Instead, there is acrucial relationship between the NEG2 construction and information structure, such thatNEG2 in these languages is felicitously employed only when the negated proposition isdiscourse-old and salient.

10

3. NEG2/3 in Brazilian Portuguese

Let us turn now to a similar, but considerably more complex, case that I have analyzed inmuch greater detail elsewhere (Schwenter 2005). In spoken Brazilian Portuguese (BP),all three negative patterns in the model of Jespersen’s Cycle given in (1) above arepresent, but the post-verbal negative occurs sentence-finally instead of immediately afterthe verb (cf. French, Catalan, Italian):

(7a) A Cláudia não veio à festa. (NEG1)(7b) A Cláudia não veio à festa não. (NEG2)(7c) A Cláudia veio à festa não. (NEG3)

All mean: ‘Cláudia didn’t come to the party.’

Another important difference between BP and the other languages mentioned so far isthat the post-verbal negative form is a repetition of the pre-verbal negative morpheme(não). That is, instead of the originally “minimizing” nominal element common to theaforementioned languages, NEG2 in BP constitutes what is commonly known as“embracing” negation. This pattern is found in other Romance varieties as well(Schwegler 1990, 1996), and arises via the incorporation of a post-sentential,afterthought-like “resumptive negation” (Jespersen 1917: 72) into sentence-final position,with loss of the intonation break between the sentence and the negative morpheme(Alkmim 2001; Bernini and Ramat 1996; Schwegler 1990). These formal differencesnotwithstanding, the synchronic distribution and diachronic development of thenoncanonical BP negatives in (7) have been widely considered an instance of Jespersen’sCycle (Bernini and Ramat 1996; Schwegler 1990; inter alia). As I intend to show in therest of this section, there are clear discourse-functional similarities shared with theCatalan and Italian NEG2 constructions which indeed justify situating both NEG2 andNEG3 in BP at Stage 2 of Jespersen’s Cycle.

Although negation in BP has been considered a “change in progress” (NEG1 >NEG2/3) by some researchers (Schwegler 1991; Bernini and Ramat 1996), the overalluse of NEG2/3 in conversational BP is actually quite low. Grouping the results of severaldifferent studies, we find that NEG2 never exceeds 20% of all sentential negatives acrossdialects, while NEG3 never exceeds 5% (Alkmim 2001; Furtado da Cunha 1996, 2001;Roncarati 1996). In addition, cross-generational patterns of NEG1 vs. NEG2/3 use do notclearly evince the kinds of patterns one would expect to find in a case of change inprogress (Alkmim 2001). As a result, I do not make the assumption, contra the views ofSchwegler (1991) and Bernini and Ramat (1996), that a change towards obligatory NEG2(or NEG3, for that matter) is currently in progress in BP.

Very much like the description of noncanonical negatives in other languages, themeaning or function of NEG2 in BP has been variously characterized as “emphatic”(Barme 2000), “reinforcing” (Uppendahl, 1979), “contrary to expectation” (Furtado daCunha, 2001), or “presuppositional” (Schwegler, 1991, 1996; Roncarati, 1996). However,upon surveying the data from BP one finds that none of these characterizations canaccount for the distribution of NEG2, which, again, turns out to look very similar to the

11

function of the corresponding NEG2s in Catalan and Italian. Compare the followingpatterns of (in)felicity of BP NEG2:

(8a) [Speaker realizes that she missed a TV program she wanted to see]Eu não vi esse programa (#não)!I not saw that program‘I didn’t watch that program!’

(8b) [Sister realizes that HER BROTHER missed a TV program he wanted to see]Você não viu esse programa (#não)!you not saw that program‘You didn’t watch that program!’

(9) A: Você viu esse programa?‘Did you watch that program?’

B: Não vi não.not I.saw NEG‘I didn’t watch it.’

Discourse-oldness of the negated proposition, as in (9), is again the relevant property forthe use of NEG2. And, as already noted for the examples from Catalan and Italian, it ismerely activation, not belief, of a proposition p that licenses the use of NEG2 for thesubsequent denial of p. Thus, the question posed by A in (9) does not have to beinterpreted as “biased” towards the truth of the corresponding proposition (i.e. towards“B watched the program”) in order for B to employ NEG2 in the reply. On the contrary,seemingly “emphatic” contexts like (8a) or hearer-old/counter-expectation contexts like(8b) do not license NEG2; indeed, only NEG1 is possible in these examples.

Further evidence for the necessity of distinguishing between believed and activatedpropositions can be discerned in examples where NEG2 does not occur in a sentenceproviding an answer to a yes/no-question. In (10), from the PEUL corpus of spoken BPfrom Rio de Janeiro, speaker F offers up possible eventualities that could keep the soccerteam under discussion from winning a particular game:

(10) E- (est.) Quer dizer que tem possibilidade de ganhar?‘You mean that there’s a chance of winning?’

F- Tem possibilidade. A não ser se acontecer, no campo mesmo, um desastre:alguém quebrar uma perna, do outro ser expulso, daí, pode até perder, masisso aí, se deus quiser, não acontece não.‘It’s possible. Unless there were to occur, on the field itself, a disaster:someone breaking a leg, another one being ejected, in that case, they couldeven lose, but that there, if God wishes, won’t happen.

[E19]

12

It is clear in (10) that F does not BELIEVE that someone will break a leg or be ejected fromthe game in question. Instead, it is understood that he is creating possible scenarios thatcould damage the team’s chances of winning. His use of NEG2—in the apodosis of aconditional sentence—is to deny the possible occurrence of these hypothetical disastrousevents, events that exist only in the mental model of the speaker. The occurrence of theseevents is therefore activated within the speaker’s mental model, and in the commonground of the interlocutors, but not believed by any participants in the conversation.

As is the case in Catalan and Italian, NEG2 in BP can also occur in agreement with aprior negative assertion. Again, this is completely in line with the present analysis,insofar as NEG2 is sensitive to a salient, discourse-old proposition; here, that propositionis the affirmative counterpart underlying A’s negative utterance:

(11) A: O João não foi à festa.the João not went to.the party‘João didn’t go to the party.’

B: Não foi não.‘He didn’t go.’

The necessity of taking information structure into account for the analysis ofnoncanonical negation becomes even clearer when comparing BP NEG2 and NEG3: theaccessibility of the denied proposition also plays an important role in constraining whichnoncanonical negative construction can be selected in a given context (Schwenter 2005).The salience and discourse-oldness of the denied proposition are also requirements forNEG3, which would likewise be infelicitous in the examples in (8), but felicitous in (9).The main difference between NEG3 and NEG2 is that only NEG2 can be used to denypropositions which are inferrable on the basis of other propositions, while NEG3 mustdeny a proposition that is activated explicitly in the discourse context. Compare (12)where both NEG2 and NEG3 is felicitous and (13), where only NEG2 is felicitous:3

(12) A: Você gostou da palestra da Maria?you liked of.the talk of.the Maria‘Did you like Maria’s talk?’

B: Gostei não. (Or: Não gostei não)I.liked NEG‘I didn’t.’

3 The judgments here reflect those of speakers from Rio de Janeiro. At least some speakers from thenortheast region of Brazil (e.g. Bahia) appear to accept NEG3 in (13). I am grateful to Tjerk Hagemeijer forpointing this out to me.

13

(13) A: Você gostou da palestra da Maria?‘Did you like Maria’s talk?’ (+> B went to Maria’s talk)

B: Eu #(não) fui não.I not went NEG‘I didn’t go.’

The question asked by A in (12) activates directly the proposition “you (=B) like Maria’stalk”, thereby permitting both NEG2 and NEG3 in B’s response. In (13), however, thesame question does not directly activate the proposition “you (=B) went to Maria’s talk”,but rather invites the inference that A believes that B went to Maria’s talk. In this case,NEG3 is infelicitous since its domain of use is restricted to discourse-old propositionsthat are directly activated in the discourse. By contrast, NEG2 is not restricted in thisway, and is therefore felicitous in B’s reply.

The canonical NEG1 form (não VP) can be used in all the contexts where NEG2 andNEG3 are felicitous, and of course many more where the noncanonical forms are notpossible. In short, negative statements that are discourse-new, and therefore primarily“informative” (i.e. not denials of an already accessible proposition), are encoded byNEG1 (Armin Schwegler, p.c.). The differences between the three BP negatives can besummarized as in Table 1, where only NEG1 is permitted with discourse-newpropositions, NEG1 and NEG2 with inferrable propositions, and all three forms withdirectly-activated propositions:4

Table 1: BP negatives, by information status of the denied proposition

Form Discourse-new

Inferrable DirectlyActivated

NEG1 √ √ √NEG2 # √ √NEG3 # # √

4. Conclusion

The data presented in this paper from three Romance varieties point to a clear functionalgap between Stages 2 and 3 of Jespersen’s Cycle in its usual formulation (see [1] above).The varieties surveyed have NEG2 constructions that are neither unambiguouslyemphatic (in whatever sense of that term is taken) nor obligatory for the expression ofsentential negation. The common denominator among these NEG2s is that they arerestricted to denials of activated, salient discourse-old propositions. It is in this sense thatfine-tuning of Jespersen’s Cycle is needed: Much of the diachronic “action” in the Cycle

4 Another issue of interest to Jespersen’s Cycle that I cannot deal with here due to space restrictions is theprobable diachronic connection between NEG2 and NEG3. The data from present-day BP suggest that theNEG2 > NEG3 change, as purportedly reflected by the transition from Stage 3 to Stage 4 of the Cycle (e.g.from ne V pas to V pas in French), is considerably more complex than previously recognized.

14

takes place between Stages 2 and 3, between the generalization of a post-verbal elementas a GRAMMATICAL (i.e. no longer lexical) exponent of negation, and its continuedgeneralization to become an obligatory part of the canonical sentential negationconstruction (e.g. (ne) V pas in present-day French). The synchronic data provided in thispaper should cause us to re-think the way in which “emphatic” post-verbal negativewords lose their purported emphatic value and gradually become pragmatically unmarkedelements. It now seems clear that this diachronic process must take information-structuralconsiderations into account, and likewise that the NEG2 constructions in Catalan, Italian,and BP will need to loosen their restrictions on the negated proposition before evermoving into Stage 3 of the Cycle.5

Now, before I take too much credit for the linking of negation and informationstructure, I should point out that this link was prefigured by—who else?—Larry Horn inhis 1978 CLS contribution, which was later updated in Appendix 2 of A Natural Historyof Negation (1989). Commenting on the distinction between verbs with incorporatednegation as in “a discouraged b from Xing” and verbs lacking this negation as in “aencouraged b not to X”, Horn noted that the use of the incorporated form “is limited tocontexts where the contained affirmative proposition is already understood. Thus, adenied (doubts) that p is appropriate only when p is a proposition evoked in (i.e.,appearing in or directly inferable from) the earlier discourse” (1989: 523). While I thinkHorn’s analysis might itself need a bit of fine-tuning—it doesn’t seem, for instance, thatthe proposition p necessarily has to be discourse-old, but simply presupposed by theinterlocutors—it nevertheless illustrates clearly the intricate connection betweeninformation structure and the choice of negative FORM, a result also issuing from theanalysis I’ve presented here.

5. Postscript: A form-function paradox?

To conclude this study, I would like to point out an apparent paradox between negativeform and negative function, and then suggest that such a paradox does not, in fact, exist.As a starting point, note that it is a widespread assumption among linguists (specialists innegation or otherwise) that negative sentences are in some sense replies to thecorresponding affirmatives. Consider, for instance, Horn’s position that, “There should bea reason to utter a sentence, and for a negative sentence, that reason … is generally theearlier consideration of its contained affirmative counterpart” (1978: 203). Likewise, inwhat is certainly one of the most quoted citations of all time on negation, Givón notedthat “negatives are uttered in a context where corresponding affirmatives have alreadybeen discussed, or else where the speaker assumes the hearer’s belief in—and thusfamiliarity with—the corresponding affirmatives” (1978: 109).

The paradox arises upon confronting the positions of these (and other) esteemedscholars with the data offered in this paper. What my analysis has revealed is that thediscourse function of the noncanonical forms of sentence negation (i.e. NEG2/3) in the

5 In Schwenter (2005) I analyze two examples from BP that appear to indicate that the information-structural restrictions imposed by NEG2 are actually being loosened in this variety.

15

three languages surveyed above is very similar—nearly identical, actually—to thefunction that Horn, Givón and many, many others (see the copious citations to this effectin Horn [1989, ch. 3]) have argued to be the most canonical function for sentencenegation in general!

So, the question to be asked is this: Why are clearly marked forms for sentencenegation being used to carry out the discourse function considered the most UNMARKED

one by most researchers? But, more importantly, what is the evidence that the function ofdenying “an earlier-considered affirmative counterpart” is the unmarked or prototypicalone for negation? This function may seem to be prototypical in isolation, where anegative sentence is most easily interpreted in this way. However, in stark contrast to thestandard position, researchers studying negation in conversational discourse—which Itake to be the locus of language change—have found that denials of explicit propositionsare actually quite infrequent. Tottie (1991) found that only 14.7% of all sentencenegations in a corpus of British English have the function of denying a prior affirmative.Even more strikingly, Thompson (1998) found that only 5% of the sentential negatives ina corpus of American English have this function. As Thompson states, the “data … showclearly that there is typically NO sense in which a negative clause denies anything, eitherexplicit or implicit, in the conversation” (1998: 323; emphasis in original).

Taking these findings on the use of negative sentences in conversation into account,the functional differentiation of paradigms of sentence negators is more readilycomprehensible. Noncanonical forms like NEG2 and NEG3 in the three languagessurveyed here serve to indicate in explicit fashion, via their distinct coding vis-à-visNEG1, what is actually a highly infrequent/marked function in naturally-occurringdiscourse, that of denying a proposition that is already activated in the current discoursecontext (see Fretheim 1984 on “denials”). It is the pairing of noncanonical NEG2 formswith a noncanonical discourse function that can set the wheels of Jespersen’s Cycle inmotion.

References

Alkmim, Mônica. 2001. As negativas sentenciais no dialeto Mineiro: uma abordagemvariacionista. PhD dissertation, Universidade Federal de Minas Gerais.

Armstrong, Nigel. 2002. Variable deletion of French ne: a cross-stylistic perspective.Language Sciences 24.153-73.

Ashby, William. 1981. The loss of the negative particle ne in French: a syntactic changein progress. Language 57.674-87.

Ashby, William. 2001. Un nouveau regard sur la chute du ne en français parlétourangeau: s’agit-il d’un changement en cours? French Language Studies 11.1-22.

Barme, S., 2000. Negação (e afirmação) no português brasileiro falado informal. Ms.,University of Mainz.

Bernini, Giuliano, and Paolo Ramat. 1996. Negative sentences in the languages ofEurope. Berlin: Mouton de Gruyter.

Caton, Charles E. 1981. Stalnaker on pragmatic presupposition. Radical pragmatics, ed.by Peter Cole, 83-100. New York: Academic Press.

16

Cinque, Guglielmo. 1991 [1976]. Mica: note di sintassi e pragmatica. Teoria linguistica esintassi italiana (by Guglielmo Cinque), 311-23. Bologna: Il Mulino.

Coveney, Aidan. 1996. Variability in spoken French: a sociolinguistic study ofinterrogation and negation. Exeter, UK: Elm Bank.

Dahl, Östen. 1979. Typology of sentence negation. Linguistics 17.79-106.Dahl, Östen. 2001a. Grammaticalization and the life cycles of constructions. RASK

14.91-133.Dahl Östen. 2001b. Inflationary effects in language and elsewhere. Frequency and the

emergence of linguistic structure, ed. by Joan Bybee and Paul Hopper, 471-80.Amsterdam: John Benjamins.

Dahl, Östen. 2004. The growth and maintenance of linguistic complexity. Amsterdam:John Benjamins.

Detges, Ulrich, and Richard Waltereit. 2002. Grammaticalization vs. reanalysis: asemantic-pragmatic account of functional change in grammar. Zeitschrift fürSprachwissenschaft 21.151-95.

Dryer, Matthew S. 1996. Focus, pragmatic presupposition, and activated propositions.Journal of Pragmatics 26:475-523.

Espinal, M. Teresa. 1993. The interpretation of no-pas in Catalan. Journal of Pragmatics19.353-69.

Fretheim, Thorstein. 1984. Denials and other negatives. Riepmocála, essays in honour ofKnut Bergslund: presented on the occasion of his seventieth birthday, ed. by KnutBergsland et al., 49-65. Oslo: Novus Forlag.

Furtado da Cunha, Maria Angélica. 1996. Gramaticalização dos mecanismos de negaçãoem Natal. Gramaticalização no português do Brasil, ed. by Mário EduardoMartelotta, Sebastião Josué Votre, Maria Maura Cezario, 167-89. Rio de Janeiro:Tempo Brasileiro.

Furtado da Cunha, Maria Angélica. 2001. O modelo das motivações competidoras nodomínio funcional da negação. D.E.L.T.A. 17.1-30.

Geurts, Bart. 2000. Explaining grammaticalization (the standard way). Linguistics38.781-8.

Givón, Talmy. 1978. Negation in language: pragmatics, function, ontology. Syntax andsemantics 9: Pragmatics, ed. by Peter Cole, 69-112. New York: Academic Press.

Hopper, Paul J. 1991. On some principles of grammaticization. Approaches togrammaticalization, vol. I, ed. by Elizabeth Closs Traugott and Bernd Heine, 17-35.Amsterdam: John Benjamins.

Hopper, Paul J., and Elizabeth Closs Traugott. 2003. Grammaticalization. 2nd ed.Cambridge: Cambridge University Press.

Horn, Laurence R. 1978. Lexical incorporation, implicature, and the least efforthypothesis. Papers from the parasession on the lexicon, ed. by Donka Farkas et al.,196-209. University of Chicago Press, Chicago.

Horn, Laurence R. 1989. A natural history of negation. Chicago: University of ChicagoPress.

Hualde, José Ignacio. 1992. Catalan. London: Routledge.

17

Israel, Michael. 1998. The rhetoric of grammar: scalar reasoning and polarity sensitivity.PhD dissertation, UCSD.

Israel, Michael. 2001. Minimizers, maximizers and the rhetoric of scalar reasoning.Journal of Semantics 18.297-331.

Jespersen, Otto. 1917. Negation in English and other languages. Copenhagen: A. F.Høst.

Jespersen, Otto. 1924. The philosophy of grammar. London: George Allen & Unwin.Kadmon, Nirit. 2001. Formal pragmatics. Oxford: Blackwell.Labov, William. 1975. On the use of the present to explain the past. Proceedings of the

11th International Congress of Linguists, ed. by L. Heilmann, 825-51. Bologna: IlMulino.

Ladusaw, William A. 1993. Negation, indefinites, and the Jespersen Cycle. BLS 19.437-46.

Martineau, France, and Raymond Mougeon. 2003. A sociolinguistic study of the originsof ne deletion in European and Quebec French. Language 79.118-52.

Molinelli, Piera. 1987. The current situation as regards discontinuous negation in theRomance languages. Linguistic typology, ed. by Paolo Ramat, 165-72. Berlin:Mouton de Gruyter.

Posner, Rebecca. 1985. Post-verbal negation in non-standard French: a historical andcomparative view. Romance Philology 39.170-97.

Postal, Paul. 2004. Skeptical linguistic essays. Oxford: Oxford University Press.Prince, Ellen F. 1981. Toward a taxonomy of given-new information. Radical

Pragmatics, ed. by Peter Cole, 223-255. New York: Academic Press.Prince, Ellen F. 1992. The ZPG letter: subjects, definiteness, and information-status. In

Discourse description: diverse analyses of a fundraising text, eds. Sandra Thompsonand William Mann, 295-325. Amsterdam: John Benjamins.

Roncarati, Cláudia. 1996. A negação no português falado. Variação e discurso, ed. byAlzira Tavares de Macedo, Cláudia Roncarati, and Maria Cecília Mollica, 97-112.Rio de Janeiro: Tempo Brasileiro.

Sankoff, Gillian, and Diane Vincent. 1980. The productive use of ne in spoken MontréalFrench. The social life of language (by Gillian Sankoff), 295-310. Philadelphia:University of Pennsylvania Press.

Schwegler, Armin. 1988. Word-order changes in predicate negation strategies inRomance languages. Diachronica 5.21-58.

Schwegler, Armin. 1990. Analyticity and syntheticity: a diachronic perspective withspecial reference to Romance languages. Berlin: Mouton de Gruyter.

Schwegler, Armin. 1991[1985-87]. Predicate negation in contemporary BrazilianPortuguese: a change in progress. Orbis 34.187-214.

Schwegler, Armin. 1996. La doble negación dominicana y la génesis del españolcaribeño. Hispanic Linguistics 8.247-315.

Schwenter, Scott A. 2002. Pragmatic variation between negatives: evidence fromRomance. University of Pennsylvania Working Papers in Linguistics, vol. 8.3: Papersfrom NWAV 30, ed. by Daniel Ezra Johnson and Tara Sánchez, 249-63.

18

Schwenter, Scott A. 2005. The pragmatics of negation in Brazilian Portuguese. Lingua115.1427-56.

Stalnaker, Robert. 1978. Assertion. Syntax and semantics 9: pragmatics, ed. by PeterCole, 315-32. New York: Academic Press.

Thompson, Sandra A. 1998. A discourse explanation for the cross-linguistic differencesin the grammar of incorporation and negation. Case, typology, and grammar: inhonor of Barry Blake, ed. by Anna Sierwiska and Jae Jung Song, 307-40.Amsterdam: John Benjamins.

Tottie, Gunnel. 1991. Negation in speech and writing. New York: Academic Press.Wheeler, Max W.; Alan Yates; and Nicolau Dols. 1999. Catalan: a comprehensive

grammar. London: Routledge.Yates, Alan. 1993[1975]. Teach yourself Catalan. Lincolnwood, IL: NTC Publishing.Zanuttini, Raffaella. 1997. Negation and clausal structure: a comparative study of

Romance languages. Oxford: Oxford University Press.