Scalar Additive Operators in the Languages of Europe

54
* This article is a result of a research visit by Volker Gast at the Center for Grammar, Cognition and Typology at the University of Antwerp in summer 2007. Financial support from the Alexander von Humboldt Foundation, the Research Council of the University of Antwerp, and the Belgian Federal Government (IAP- grant P6/44) is gratefully acknowledged. We are indebted to several native speakers and specialists for first- hand information: Cleomar Alves da Silva (Portuguese), Olga Barsukova (Russian, Ukrainian), Alexis Dimitriadis (Modern Greek), Andreea Dumitrescu (Romanian), Viktor Elšik (Czech), Petar Kehayov (Bulgarian and Estonian), Elma Kerz (Croatian), Mikhail Kissine (French), Lene Lam (Danish), Benita Lašinytė (Lithuanian), Vladimir Matveev (Russian), Matti Miestamo (Finnish), Brian Nolan (Irish Gaelic), Olga Pavlovskaya (Russian, Belarusan, Ukrainian), Peter Schrijver (Breton), Katerina Stathi (Modern Greek), LetiziaVezzosi (Italian), and Mónica Zepeda (Spanish). Moreover, we would like to thank Holger Diessel, Nick Evans, Florian Haas, Martin Haspelmath, Daniel Hole, Ekkehard König, former editor Brian Joseph, former associate editor Laura Michaelis, present editor Greg Carlson, and three anonymous referees for valuable comments and suggestions. Any inaccuracies are our own. An early outline of the general framework is given in Gast & van der Auwera 2010. All maps in this article were generated with the open-source software ‘R: A language and environment for statistical computing’, version 2.9.2 (R development core team, Vienna, 2010). We thank Balthasar Bickel for providing instructions for map making on his website. 1 As is seen in the course of this article, the expressions illustrated in 1–4 are not fully equivalent. While some aspects of variation are captured in the typology presented below, others had to be left out of considera- tion for reasons of space (see e.g. König 1981a,b, 1991 for additional parameters of variation). As always in crosslinguistic studies, we have to work on the hypothesis of ‘crosslinguistic comparability’. Near equiva- lence of expressions from different languages is assumed to be reflected in translational correspondence. The definition of scalar additive operators provided in §2.3 is independent of any particular language and can thus be regarded as a tertium comparationis of the study. 1 SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE VOLKER GAST JOHAN VAN DER AUWERA Friedrich Schiller University Jena University of Antwerp Scalar additive operators, such as Engl. even, Fr. même, Germ. sogar, Sp. aun, and so forth, vary crosslinguistically in terms of their distributional behavior, in particular with respect to se- mantic and pragmatic properties of the sentential environment (scale-reversing vs. scale-preserv- ing, negative vs. nonnegative). This article proposes a semantic framework for the crosslinguistic analysis of scalar additive operators and a typology based on that framework. Five major types of operators are distinguished and the distribution of these types in forty European languages is sur- veyed. The synchronic patterns found in the languages of the sample are interpreted in the light of historical developments in the domain of investigation, and implications for the division of labor between lexical meaning and sentential context are discussed.* Keywords: pragmatic strength, scalarity, focus, scope, polarity, typology 1. INTRODUCTION. 1.1. THE OBJECT OF STUDY. The term ‘scalar additive operator’ (SAO, for short) is here used for elements like Engl. even, Span. aun, Germ. sogar, and French même, as illus- trated in 1–4 with an example from the Bible (Luke 8:25). A definition of these expres- sions is provided below (§2.3). 1 (1) English Who is this? He commands even the winds and the water, and they obey him. (Luke 8:25; New International Version, from http://www.bibleserver.com)

Transcript of Scalar Additive Operators in the Languages of Europe

* This article is a result of a research visit by Volker Gast at the Center for Grammar, Cognition andTypology at the University of Antwerp in summer 2007. Financial support from the Alexander von HumboldtFoundation, the Research Council of the University of Antwerp, and the Belgian Federal Government (IAP-grant P6/44) is gratefully acknowledged. We are indebted to several native speakers and specialists for first-hand information: Cleomar Alves da Silva (Portuguese), Olga Barsukova (Russian, Ukrainian), AlexisDimitriadis (Modern Greek), Andreea Dumitrescu (Romanian), Viktor Elšik (Czech), Petar Kehayov(Bulgarian and Estonian), Elma Kerz (Croatian), Mikhail Kissine (French), Lene Lam (Danish), BenitaLašinytė (Lithuanian), Vladimir Matveev (Russian), Matti Miestamo (Finnish), Brian Nolan (Irish Gaelic),Olga Pavlovskaya (Russian, Belarusan, Ukrainian), Peter Schrijver (Breton), Katerina Stathi (Modern Greek),Letizia Vezzosi (Italian), and Mónica Zepeda (Spanish). Moreover, we would like to thank Holger Diessel,Nick Evans, Florian Haas, Martin Haspelmath, Daniel Hole, Ekkehard König, former editor Brian Joseph,former associate editor Laura Michaelis, present editor Greg Carlson, and three anonymous referees forvaluable comments and suggestions. Any inaccuracies are our own. An early outline of the general frameworkis given in Gast & van derAuwera 2010.All maps in this article were generated with the open-source software‘R: A language and environment for statistical computing’, version 2.9.2 (R development core team, Vienna,2010). We thank Balthasar Bickel for providing instructions for map making on his website.

1 As is seen in the course of this article, the expressions illustrated in 1–4 are not fully equivalent. Whilesome aspects of variation are captured in the typology presented below, others had to be left out of considera-tion for reasons of space (see e.g. König 1981a,b, 1991 for additional parameters of variation). As always incrosslinguistic studies, we have to work on the hypothesis of ‘crosslinguistic comparability’. Near equiva-lence of expressions from different languages is assumed to be reflected in translational correspondence. Thedefinition of scalar additive operators provided in §2.3 is independent of any particular language and can thusbe regarded as a tertium comparationis of the study.

1

SCALARADDITIVE OPERATORS INTHE LANGUAGES OF EUROPE

VOLKER GAST JOHAN VAN DERAUWERA

Friedrich Schiller University Jena University of AntwerpScalar additive operators, such as Engl. even, Fr. même, Germ. sogar, Sp. aun, and so forth,

vary crosslinguistically in terms of their distributional behavior, in particular with respect to se-mantic and pragmatic properties of the sentential environment (scale-reversing vs. scale-preserv-ing, negative vs. nonnegative). This article proposes a semantic framework for the crosslinguisticanalysis of scalar additive operators and a typology based on that framework. Five major types ofoperators are distinguished and the distribution of these types in forty European languages is sur-veyed. The synchronic patterns found in the languages of the sample are interpreted in the light ofhistorical developments in the domain of investigation, and implications for the division of laborbetween lexical meaning and sentential context are discussed.*

Keywords: pragmatic strength, scalarity, focus, scope, polarity, typology

11.. INTRODUCTION.11..11.. THE OBJECT OF STUDY. The term ‘scalar additive operator’ (SAO, for short) is here

used for elements like Engl. even, Span. aun, Germ. sogar, and French même, as illus-trated in 1–4 with an example from the Bible (Luke 8:25). A definition of these expres-sions is provided below (§2.3).1

(1) EnglishWho is this? He commands eevveenn the winds and the water, and they obey him.

(Luke 8:25; New International Version, from http://www.bibleserver.com)

2 We use the following abbreviations: ABL: ablative, ACC: accusative, ART: article, AUX: auxiliary, CLSF: clas-sifier, COMP: complementizer, CONJ: conjunction, COP: copula, DAT: dative, DEM: demonstrative, DET: deter-miner, EX: existential, EXPL: expletive, FNOM: factive nominalizer, FUT: future, GEN: genitive, IMPF:imperfect(ive), LOC: locative, NOM: nominative, MID: middle voice, NEG: negation, PART: partitive, PL: plural,POSS: possessive, PREP: preposition, PRON: pronoun, PST: past, PSV: passive, PTCL: particle, REFL: reflexive, REL:relative, SG: singular, SUBJ: subjunctive, TOP: topic.

3 See for instance Fillmore 1965, Horn 1969, Fraser 1971, Anderson 1972, Fauconnier 1975a,b, Altmann1976, Ducrot 1980, König 1981a,b, 1991, Bennett 1982, Anscombre & Ducrot 1983, Jacobs 1983, Kay 1990,Barker 1991, Lycan 1991, Berckmans 1993, Francescotti 1995, Kalerikos 1995, Wilkinson 1996, Rullmann1997, 2003, Guerzoni 2003, 2004, Schwarz 2005, Sudhoff 2010, to name just a selection of pertinent publi-cations.

2 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

(2) Spanish2

¿Quién es éste, que manda aauunn a los vientos y al agua,who is this PTCL commands even PREP the winds and PREP.ART water

y le obedecen?and him they.obey

(Luke 8:25; La Biblia al día, from http://www.bibleserver.com)(3) German

Was ist das für ein Mensch, dass ssooggaarr die Winde und das Wasser seinemwhat is that for a man that even the winds and the water his

Befehl gehorchen?order obey

(Luke 8:25; Einheitsübersetzung, from http://www.bibleserver.com)(4) French

Voyez: il commande mmêêmmee aux vents et aux vagues, et look he commands even PREP.ART.PL winds and PREP.ART.PL waves and

il s’-en fait obéir!PRON REFL-PRON makes obey

(Luke 8:25; Bible du Semeur, from http://www.bibleserver.com)

While specific scalar additive operators—most notably Engl. even and Germ. sogar(as well as its near equivalent selbst)—are rather well studied in terms of their semanticand distributional behavior,3 our knowledge of the patterns and limits of crosslinguisticvariation in this domain is still rather limited, the most important publications probablybeing König 1991, Rullmann 1997, Guerzoni 2003, 2004, and Giannakidou 2007. Inthis article, we pursue four major goals: First, we propose a semantic framework for theanalysis of scalar additive operators that is intended to support a crosslinguistic study.Second, on the basis of this framework we devise a typology of scalar additive opera-tors. Third, we examine the distributional behavior of relevant operators from a sampleof forty European languages. Finally, we interpret our findings in a historical perspec-tive and consider their implications for an important theoretical question, that is, howbest to account for distributional variation in the domain of investigation (differences inscope behavior vs. differences in lexical content; see §3).We first summarize earlier approaches to the lexical-semantic analysis of scalar addi-

tive operators and introduce our own analysis, and then discuss the relationship betweenthe lexical meaning of scalar additive operators and the sentential environment, with afocus on matters of scope. Having established the most important parameters of varia-tion concerning the distribution of scalar additive operators in Europe and having dis-tinguished seven major types of operators we describe the distribution of the types ofexpressions under study in European languages and introduce a way of representingpatterns of polysemy and systems of scalar additive operators. The relevant systems of

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 3

some languages, including two non-European languages, are surveyed. Finally, we con-sider our findings in a diachronic perspective by outlining typical historical develop-ments in the domain of investigation and discussing some implications for a synchronictypology of scalar additive operators.The remainder of this introduction contains a note on terminology (§1.2) and some

remarks on the empirical basis of the investigation, that is, European languages (§1.3).

11..22.. ‘SCALAR ADDITIVE OPERATORS’ AS A SEMANTIC COVER TERM. Scalar additive opera-tors are often dealt with under the rubric of ‘focus particles’ in comparative work (e.g.König 1991, Giannakidou 2007). In individual grammars, these elements are typicallyfound in a section on ‘(focus) adverbs’ (e.g. Press 1986 on Modern Breton) or simply‘particles’ (e.g. Buchholz & Fiedler 1987 on Albanian), if they are mentioned at all. Weprefer the more general term ‘operator’ because many of the relevant items do not ex-hibit properties typical of adverbs, and even the term ‘particle’ is too specific, as it en-tails that the items in questions are (uninflected) words. In many languages, scalaradditive operators can be found that are not single independent words, and which there-fore cannot be classified as ‘particles’, since they are really ‘particle groups’. For in-stance, in contexts such as 1–4 above even translates as till och med (lit. ‘until and with’,i.e. ‘up to and including’) into Swedish.

(5) SwedishVem är han, eftersom till och med vinden och vågorna lyder honom?who is he CONJ even the.winds and the.waves obey him‘Who is this? Even the winds and the waves obey him!’

(Luke 8:25; En levande bok, from http://www.bibleserver.com)

Hungarian combines the scalar particle még (also ‘still’) with a nonscalar additiveoperator, either is (‘also’, in positive contexts; 6a) or sem (‘(not) either, neither/nor’, innegative contexts; 6b). In other words, még contributes the scalar meaning, while is andsem (< is nem ‘also not’) trigger a (negative) existential presupposition. Such operatorscould be classified as ‘discontinuous particle groups’.

(6) Hungarian (Abrusán 2007:3)

a. Péter még Marit is üdvözölte.Peter even Mary also greeted‘Peter greeted even Mary.’

b. Péter nem üdvözölte még Marit sem.Peter not greeted even Mary (not.)either‘Peter did not even greet Mary.’

Finally, and most importantly perhaps, there are scalar additive operators that are af-fixes, not particles, even though this situation is only marginally found in European lan-guages and seems not to be particularly widespread in other parts of the world either.For instance, Finnish and Estonian have (cognate) suffixes -kin (Finnish) and -gi/-ki(Estonian), which have a function similar to that of Engl. also. Estonian -gi, illustratedin 7, is used in both scalar and nonscalar contexts, that is, as an equivalent of both alsoand even.

(7) EstonianTuuled-gi ja meri-gi kuulevad Tema sõna. wind-even and sea-even obey his word‘Even the wind and the sea obey him.’ (Petar Kehayov, p.c.)

To conclude these terminological remarks, the term ‘scalar additive operator’ (SAO)has the advantage of being applicable to any type of exponent, irrespective of its mor-

4 Note that in calling such multiword expressions ‘operators’ we do not wish to imply that they cannot beanalyzed in a compositional way.

4 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

phological and distributional properties. It is more general than both ‘adverb’ and‘(focus) particle’ and can also be used for markers that are ‘distributed’ over the sen-tence, as for instance the Hungarian ‘discontinuous operators’ még ... is and még ... semillustrated in 6 above.4

11..33.. REMARKS ON THE EMPIRICAL BASIS OF THIS STUDY. Restricting this survey to Euro-pean languages may seem arbitrary or simply convenient, but is in fact well motivated.First, the existence of specialized (grammatical, lexical) SAOs (as well as entire sys-tems of such operators) is certainly not a universal feature of language. In some parts ofthe world SAOs are not lexicalized at all, or have a marginal status at best. For example,they are rarely found in Australian languages (see McConvell 1983), and many nativeAmerican languages have borrowed relevant items from Spanish or other contact lan-guages (see Gast & van der Auwera 2010:306 for some examples). This is not of courseto say that SAOs are nonexistent outside of Europe. For instance, (South) East Asianlanguages are known to have elaborate systems of focus operators, including SAOs(see, for instance, Hole 2004 and references cited there for Mandarin Chinese, andNakanishi 2006 on Japanese). Before considering SAOs in a global perspective, how-ever, it seems reasonable to focus on a specific and well-studied language area such asEurope, since this allows us to take into consideration aspects of (external and internal)language history, including matters of language contact (see Kortmann 1997:§3.2 forsimilar arguments about comparative studies of European languages). Still, we venturean outlook on two non-European languages also (§8.2; see also Gast & van der Auwera2010).Our sample comprises forty major European languages. The data have been taken

from sources such as dictionaries, grammars, articles, native speakers, or experts (seethe acknowledgments) and the Bible. The languages of the sample are shown in Map 1(Bosnian, Croatian, and Serbian are subsumed under Bosnian-Croatian-Serbian or‘BCS’ for short, since we have not found any noticeable differences among the opera-tors of these languages).

22.. THE LEXICAL SEMANTICS OF SCALAR ADDITIVE OPERATORS. In this section we intro-duce a framework for the analysis of scalar additive operators on the basis of dynamicmodels of discourse (see e.g. Roberts 2004 for an overview) and common conceptsfrom research into focus-sensitive operators and focus semantics more generally (e.g.Rooth 1985, 1992, 1996). Much of what is said relates to Engl. even, but the underlyingconcepts can be applied to corresponding operators from other languages as well. Be-fore presenting our own analysis in §§2.2 and 2.3, we summarize a prominent, basicallysemantic (i.e. context-invariant), early approach to the analysis of even in §2.1 that isstill influential, pointing out three problems that such an analysis has to face.

22..11.. SEMANTIC ANALYSES OF SCALAR ADDITIVE OPERATORS. In spite of a large number ofrelevant publications (see n. 3), there is no generally accepted analysis of Engl. even orother SAOs. A number of early analyses (e.g. Karttunen & Karttunen 1977, Karttunen& Peters 1979, Rooth 1985) regard even as an operator that adds to the additive infer-ence of also a scalar presupposition or implicature to the effect that the proposition inquestion is the least likely one from a set of alternative propositions. Consider 8.

(8) Even [the Pope]F congratulated Fred.

1

MAP 1. The languages of the sample. MAP 1. The languages of the sample.

[AQ1]

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 5

According to the authors mentioned above, 8 says that (i) ‘the Pope congratulated Fred’(propositional content), (ii) ‘there is someone other than the Pope who congratulatedFred’ (existential presupposition or conventional implicature), and (iii) ‘ “the Pope con-gratulated Fred” is the least likely element from a set of alternative propositions thathave the form “x congratulated Fred”, with some contextually salient value replacing(the focus) x’ (scalar presupposition or conventional implicature). The set of sentencesunder consideration can be ordered on a scale as shown in 9, with an (assumed) increasein unlikelihood from the bottom to the top, and with the proposition expressed in theeven-sentence occupying the top position on that scale.

(9) FOCUS

the Pope less likely (propositional content)the governor

Even his students congratulated Fred.his best friendhis wife more likely

This analysis faces (at least) three problems, which have been extensively discussedelsewhere (see for instance König 1991:68–76, Rullmann 1997:44–48, and Guerzoni2003:106–15), so we restrict ourselves to a brief summary here: (i) even does not al-ways interact with scales of unlikelihood, (ii) it does not always require that the propo-sition in question occupy a top position on the relevant scale, and (iii) the exact status ofthe additive inference associated with even (e.g. existential presupposition, conven-tional implicature) is not entirely clear.

5 It is worth mentioning that the problem of how to define scales has also played an important role in theanalysis of restrictive focus operators like only, which is sometimes regarded as an antonym of (scalar) even(e.g. by Horn 1969, but see van der Auwera 1985). Horn (1969) analyzes only (and, accordingly, even) as in-teracting with ‘scales of degree’ that are given by the context. Jacobs (1983) likewise assumes that scales arebasically given by the context. Löbner (1990) takes it that only interacts with scales of informativeness, andthat only-sentences make the weakest statement out of a set of possible alternatives (see also von Stechow1991); but again, the notion of ‘informativeness’ is defined relative to the context, not in terms of semanticentailment relations (cf. also §2.3). See also Beaver & Clark 2008:70–72 for recent discussion.

6 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

The assumption that even always interacts with scales of unlikelihood is challengedby specific instances of this operator where probability does not seem to play a role.5

Two pertinent (attested) examples are given in 10 (see also Kay 1990:80–81, n. 21,König 1991:71, and Rullmann 1997:45 for similar English examples; see Jacobs1983:128–31 on German sogar/selbst). In both sentences, the alternative values are ex-plicitly mentioned in the sentential environment (indifferent vs. hostile, dark gray vs.black).

(10) a. However, dealing with Native individuals, who were often indifferent oreven hostile to government intervention, presented challenges to dedi-cated bureau employees.

(Corpus of Contemporary American English, http://corpus.byu.edu/coca)

b. It is more for aesthetic reasons that leather seats in automobiles are mainlycoloured dark grey, indeed mostly even black.

(http://www.freepatentsonline.com/7521118.html, accessed March16, 2010)

In 10a, there is no indication that a hostile attitude of native individuals toward govern-ment intervention is less likely than an indifferent one. In 10b, it is even stated thatleather seats in automobiles are MOSTLY black, and there is no reason to assume thatblack is, in any way, a less likely color for a car seat than dark gray.In view of examples like those in 10, it has been argued that the scales even interacts

with are more pragmatic than semantic in nature—that is, they are sensitive to informa-tion from the context (see e.g. Fauconnier 1975a,b, Anscombre & Ducrot 1983, Kay1990, Francescotti 1995, Schwenter 1999a, among others). This line of reasoning ispursued in §2.2.The second problem of a semantic analysis as sketched above concerns the assump-

tion that the focus of an even-sentence always occupies an extreme position on the rele-vant scale. That this cannot be right from a strictly semantic point of view can beillustrated with examples like 11. Obviously, the quantifier most is not the highest ele-ment on a Horn scale of the form <all, most, some>, but it nevertheless qualifies as afocus of even (see Kay 1990:89–90, Rullmann 1997:45 for similar arguments and examples).

(11) If some or even most of them do not respond, how can we know whether theanswers that we have got are representative? (British National Corpus)

This problem can also be solved by taking a pragmatic (context-sensitive) approach.With respect to 11, this means that the value all is not even under consideration.Finally, the assumption that even triggers an existential presupposition (Horn 1969,

Fauconnier 1975a,b) or conventional implicature (Karttunen & Peters 1979, Rooth1985) is also disputed, for several reasons. Fraser (1971:152–53) has pointed out thatthe truth-conditional behavior of the additive inference associated with even-sentencesdiffers from that of ‘canonical’ types of presuppositions (see also Horn 1989:148–52 fordiscussion). Krifka (1991) and von Stechow (1991) have drawn attention to exampleslike 12, which seem to be incompatible with the postulation of an existential presuppo-

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 7

sition, since only excludes values contrasting with the focus Sue (see also König 1991:70, Rullmann 1997:45–46, Guerzoni 2003:111 for discussion).

(12) John even danced only with [Sue]F.While the exact status of the existential inference associated with even is a matter of

debate, there is no disagreement that even is in some sense ‘additive’. One way of ex-plaining the somewhat special status of this inference is to regard it as being derivativeof the scalar inference (cf. Rullmann 1997:58; see also Schwarz 2005 on German auchnur and einmal ). We argue that the assumption of an independent existential presuppo-sition is in fact not needed, since it can be derived as a by-effect of other appropriatenessconditions to which SAOs are subject.

22..22.. A FRAMEWORK OF ANALYSIS. The analysis proposed in this study is ‘pragmatic’ in-sofar as it regards the meaning of even (and other SAOs) as being sensitive to, and sys-tematically interacting with, information from the context. A (simplified version of a)dynamic model of discourse along the lines of Roberts 1996, 2004 is outlined in thissection, integrating concepts from RELEVANCE THEORY (e.g. Sperber & Wilson 1986,Wilson & Sperber 2004; cf. also Iten 2002 for an analysis of even in terms of relevancetheory). Our analysis of even and other SAOs is presented in §2.3.We use a ‘content-based’ notion of context in the tradition of Stalnaker (1974, 1978,

1998) (cf. Glanzberg 2002 for an overview of theories of context). One of the most im-portant aspects of contextual knowledge is thus the ‘common ground’. Unlike Stalnaker(1974, 1978), we assume that the common ground comprises not only factual informa-tion, but also sets of discourse referents (in the sense of Karttunen 1969, 1976) andmore abstract meanings such as (unsaturated) predicates and nonfactual propositions(see also Stalnaker 1998:84 on nonfactual context information). The common ground isthus conceived of as a set of entities (of all semantic types) that interlocutors have ‘con-sciousness’ or ‘awareness’ of, that is, entities that are ‘given’ or at least (mentally) ‘ac-cessible’ (cf. Chafe 1974, 1976, Lambrecht 1994, Schwarzschild 1999, Krifka 2008,among many others).Moreover, the context of an utterance encompasses what Grice (1975:45) called an

‘accepted purpose or direction of the talk exchange’. Following Roberts (1996, 2004)(and those inspiring her game-theoretic approach, e.g. Lewis 1969, Hintikka 1973, andCarlson 1983), we assume that conversation is a joint enterprise whose purpose consistsin finding answers to specific ‘questions under discussion’ (cf. also Ginzburg 1996,Büring 2003). The participants of a ‘language game’devise specific ‘strategies’and make‘moves’ to achieve their goals. In the following discussion, we use the term ‘move’ for‘contribution made by some interlocutor to a conversation’. Note that ‘[m]oves … are notspeech acts, but rather the semantic objects expressed in speech acts: A speech act is theact of proffering a move’ (Roberts 2004:209). Following Carlson (1983), Roberts (1996,2004) distinguishes two types of moves, questions (Carlson’s ‘setup moves’), and asser-tions (Carlson’s ‘payoff moves’), which provide (at least partial) answers to some ques-tion under discussion. We use the terms ‘interrogative move’ and ‘declarative move’interchangeably with ‘question’ and ‘assertion’, respectively.In the model outlined above, each assertion is regarded as an answer to some (more

or less specific) question under discussion. The question to which a given declarativemove/assertion is intended to provide an answer—Roberts’s (1996) ‘immediate ques-tion under discussion’—we call the ‘quaestio’ (cf. Klein & von Stutterheim 1987,1992). This term is defined in 13.

[AQ3]

8 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

(13) The QUAESTIO of an assertion A is the (potentially implicit) question underdiscussion to which A is intended to provide a (partial or full) answer.

According to our analysis, even and other SAOs do not (necessarily) interact with thepropositional content of an assertion, but may also make reference to entailments thatcan be derived from that content in interaction with contextual information. We adoptthe notion of ‘contextual implication’ from relevance theory (Sperber & Wilson 1986,Wilson & Sperber 2004). Contextual implications are ‘conclusion[s] deducible frominput and context together, but neither from input or context alone’ (Wilson & Sperber2004:608). We are only interested in ‘relevant’ contextual implications. Contexual im-plications are ‘relevant’ to a quaestio Q to the extent that they provide information aboutQ (or specify Q in the case of an interrogative move; cf. Roberts 2004:216). Let us re-consider one of the examples used above to illustrate the role of ‘relevant contextual im-plications’ in the interpretation of SAOs. The sentence in 14 is the relevant part(complement clause) of 10b.

(14) Leather seats in automobiles are mainly coloured dark grey, indeed mostlyeven black.

Example 14 has been taken from a patent application for a ‘[f ]lat element having adark surface exhibiting a reduced solar absorption’ (US Patent 7521118). In the appli-cation it is pointed out that cars tend to exhibit dark surfaces for various reasons, amongwhich are the aesthetic ones mentioned in 14, and that dark surfaces absorb heat to agreater extent than lighter surfaces do. Leather seats provide one type of surface forwhich this problem is relevant. The quaestio corresponding to 14 is thus How dark areleather seats in automobiles?. Against the background of this quaestio, we can derivethe two ‘relevant contextual implications’ in 15.

(15) Quaestio: How dark are automobile seats?PROPOSITIONAL CONTENT RELEVANT CONTEXTUAL IMPLICATION

They are black. → They are completely dark. ↓ ENTAILMENTThey are dark gray. → They are very dark.

Crucially, the two propositions literally expressed—‘Leather seats are black’ and‘Leather seats are dark gray’—do not stand in a relationship of entailment to each other;they are incompatible. The contextual implications on the right-hand side in 15 do ex-hibit a systematic asymmetry in informativeness, however, since ‘They are completelydark’ entails ‘They are very dark’. As we argue, the interpretation of SAOs makes refer-ence to such contextual implications and their relative informativeness or ‘strength’. Weuse the concept of ‘pragmatic strength’, defined in 16 to capture differences in infor-mativeness between contextual implications (see Fauconnier 1975b, Anscombre &Ducrot 1983, Kay 1990, Löbner 1990, Francescotti 1995, and Iten 2002 for similar no-tions of informativeness/strength that are relativized to the context).

(16) A proposition π is PRAGMATICALLY STRONGER (relative to a given quaestio Q)than a proposition ρ iff the RELEVANT CONTEXTUAL IMPLICATIONS of π (withrespect to Q) entail the RELEVANT CONTEXTUAL IMPLICATIONS of ρ (with re-spect to Q).

Relative pragmatic strength is symbolized by ‘>s’ and ‘<s’, that is,‘π >s ρ’ stands for‘(proposition) π is pragmatically stronger than (proposition) ρ’ and, accordingly, ‘π <sρ’ for ‘(proposition) π is pragmatically weaker than (proposition) ρ’.

22..33.. THE INTERPRETATION OF SCALAR ADDITIVE OPERATORS. Scalar additive operatorsare focus-sensitive elements that interact with propositional meanings. We assume that

}

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 9

they take a proposition as their argument, that is, a unit of meaning that is capable ofbeing true or false (cf. Frege’s 1918 ‘thought’ (Gedanke) and Wittgenstein’s 1922 (prop.4.024) ‘sentence meaning’ (Satzbedeutung)).8 Propositions are thus minimally consti-tuted by a predicate and an appropriate set of arguments. To distinguish propositionsfrom the constituents denoting them, we enclose the former in double brackets symbol-izing the interpretation function; that is, [[John snores]] is the proposition denoted bythe English sentence John snores. Note that, unlike in some semantic traditions (e.g.Heim & Kratzer 1998), [[John snores]] is thus an intensional meaning (proposition),not an extensional one (truth value) (see e.g. von Stechow 2007 for this practice).A proposition taken as an argument by a given SAO is said to be in the ‘scope’ of that

operator. The syntactic structure in 17a thus corresponds to the semantic representationin 17b (semantic entities are written in small caps, so that EVEN is the semantic correlateof the English word even; in other words, EVEN = [[even]]).

(17) a. Even the winds obey him.b. EVEN([[the winds obey him]])

SCOPE

Following the tradition of ALTERNATIVE SEMANTICS (Rooth 1985, 1992, 1996), we as-sume that the effect of focusing consists in relating the denotation of a given constituentto a set of alternative (contrasting) denotations from the common ground. Since SAOsare propositional operators, we are always dealing with sets of alternative propositions.The propositional content of a sentence containing an SAO is called the ‘text proposi-tion’; the alternative propositions are called ‘context propositions’ (see Kay 1990 forthese terms). For text propositions we use the variable ‘τ’, and for context propositionsthe variable ‘κ’ (as there is no Greek letter corresponding to <c> as in context). Capital‘K’ stands for sets of context propositions (i.e. K = {κ1, κ2 ... κn}). The set of contextpropositions associated with a given sentence can be defined as in 18 (it is approxi-mately equivalent to Rooth’s 1985 ‘focus semantic value’, except that the latter com-prises the ‘ordinary semantic value’—our ‘text proposition’—as one of its elements).

(18) The set of context propositions K associated with a sentence S denoting a(text) proposition τ is the set of all propositions κ (from the common ground)that differ from τ only with respect to the denotation of the focus of S.

Let us reconsider an example in order to illustrate the notions introduced above. Thesentence Even [the winds]F obey him is associated with the text proposition [[the windsobey him]] and context propositions of the type [[x obey(s) him]], x ≠ [[the winds]]. Wewrite ‘τS’ for the text proposition τ of a sentence S, ‘κS’ for any context proposition, and‘KS’ for the set of context propositions associated with S. This is illustrated in 19.

(19) a. Even [the winds]F obey him. (= S )b. τS = [[The winds obey him]] (= text proposition of S )c. KS = {[[His children obey him]], [[His dogs obey him]], … } (= set of context propositions of S)

According to our analysis, even (as well as other SAOs) indicates that the text propo-sition of the sentence in question is pragmatically stronger than all context propositions.We regard this inference as a pragmatic presupposition in the sense of Stalnaker(1974:200): ‘P pragmatically presupposes Q iff whenever the utterance of P is conver-sationally acceptable, the speaker of P assumes Q and believes his audience to assumeQ as well’. SAOs can now be defined as in 20.

10 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

(20) SCALAR ADDITIVE OPERATORS are propositonal operators that trigger the prag-matic presupposition that their TEXT PROPOSITION is PRAGMATICALLY STRONGERthan all corresponding CONTEXT PROPOSITIONS.

We use the term ‘pragmatic presupposition’ rather than ‘conventional implicature’(Karttunen & Peters 1979) because it seems to us that the (partial) ordering of thepropositions under consideration (the text proposition and the context propositions) is aprecondition or compatibility requirement, rather than a consequence, of a move inwhich an SAO is used. In terms of the dynamic model of discourse presently assumed,this implies that SAOs are felicitous iff there is a ‘scalar quaestio’ in the commonground, that is, a quaestio that asks for the extent to which a given proposition is true(e.g. How dark are automobile seats?).To summarize, the example considered in 19 above can now be interpreted as shown

in 21. The quaestio of that assertion can be assumed to be How powerful is that man?.The propositional content is simply [[the winds obey him]] (21b), and there is a prag-matic presupposition saying that [[the winds obey him]] is pragmatically stronger thanany context proposition of the form [[x obey(s) him]], x ≠ [[the winds]] (21c).

(21) a. Even [the winds]F obey him. (= S )b. Propositional content of S: [[The winds obey him]] (= τS)c. Pragmatic presupposition of S: τS >s all κ ∈ KS

Our analysis of even does not make reference to an existential presupposition. Aspointed out in §2.1, we take it that the additive inference of even can be derived from thescalar presupposition. Note first that it is a prerequisite of focusing a given expressionthat alternative expressions (of the appropriate semantic type) be in the common ground.If even is taken to interact with scales of ‘pragmatic strength’, and if even-sentences make(pragmatically) stronger statements than all other propositions in the common ground,these alternative propositions will, under normal circumstances of conversation, be takenfor granted. For instance, with respect to our example in 1, lower-ranking alternativepropositions like those in 22 will by default be assumed to be true if the highest-rankingone is true. That this need not be so can be shown with examples like 23. It seems to usthat the concept of a ‘default assumption’ proposed by Schwarz (2005) is quite appropri-ate to capture the type of additive inference triggered by even.

(22)the winds and the water

Even his children obey(s) him. PRAGMATIC STRENGTH

his dog.(23) Can it be possible that his children do not obey him, given that even the

winds and the water do?

33.. THE LEXICAL CONTENT AND SCOPE OF SCALAR ADDITIVE OPERATORS. So far, we haveanalyzed SAOs as operators that take a proposition—the ‘text proposition’—as an argu-ment. This view is complicated by specific instances of even where the operator occursin ‘scale-reversing’ contexts. This problem is introduced in §3.1 by considering system-atic ambiguities associated with even. These ambiguities can be regarded either as amatter of scope (the ‘scope theory’), or as resulting from differences in lexical content(the ‘lexicalist theory’; cf. §3.2). After considering arguments in both directions, wemake some heuristic assumptions in §3.3, intending to establish comparability in ourcrosslinguistic study without forestalling any theoretical decisions.

33..11.. AMBIGUOUS INSTANCES OF even AND SCALE REVERSAL. Under circumstances to bemade more precise below, even gives rise to systematic ambiguities, as in 24, for exam-

6 For example, Giannakidou (1997, 1998) has pointed out that some quantifiers like few and at most aredownward entailing, but not nonveridical. Similarly, possibility operators are nonveridical but not downwardentailing.

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 11

ple (for similar examples, see Rooth 1985:Ch. 4, Wilkinson 1996:195ff., Rullmann1997:40, Guerzoni 2003:§2.6).

(24) I refuse to believe that Bill even [slapped]F that man.In its first (more prominent) reading, 24 is compatible with a context like 25.

(25) Bill is accused of murder, but I’m sure he’s innocent. In fact, I refuse to be-lieve that he even [slapped]F that man.

Let us call this context the ‘murder context’. Slapping is regarded as a relatively minoroffense in this case. A context for the second reading is provided in 26. This type of con-text, where ‘slapping’ is regarded as a major offense, we call the ‘insult context’.

(26) I refuse to believe that Bill (not only [insulted]F but) even [slapped]F thatman.

The behavior of even in the two contexts differs in systematic ways. First, in the murdercontext, but not in the insult context, even can be replaced with the scalar operator somuch as (cf. 27a vs. 27b; see also König 1981a,b, 1991).

(27) a. *I’m sure that Bill is innocent. I refuse to believe that he so much as[slapped]F that man.

b. *I refuse to believe that Bill (not only [insulted]F but) so much as[slapped]F that man.

Second, the interpretation of even in the ‘murder context’ is more or less equivalent tothe reading that results when even forms part of the higher clause I refuse to believe …(28a). In the insult context, this type of substitution is not possible (28b).

(28) a. #I’m sure that Bill did not kill that man. I even refuse to believe that he[slapped]F him.

b. #I even refuse to believe that Bill (not only [insulted]F but also) [slapped]Fthat man.

The two ‘tests’ differentiating the alternative readings of 24 can be combined. The mur-der reading allows both the ‘raising’ of even to the higher predicate and substitutionwith so much as in the lower clause (29). The resulting sentence may sound somewhatredundant, but relevant attested examples are not hard to find, as in (30).

(29) I even refuse to believe that Bill so much as [slapped]F that man.(30) Some docs even refuse to so much as scribble notes with the flashy pens em-

blazoned with drug and equipment brand names.(http://postscripts.typepad.com/the_daily_meds/, accessed March 31, 2009)

Ambiguities like the one illustrated in 24 above arise in contexts that have been called‘affective’ (Klima 1964), ‘downward entailing’ (Ladusaw 1979), ‘nonveridical’ (Zwarts1995, Giannakidou 1997, 1998, 2007), and ‘scale-reversing’ (Haspelmath 1997),among other terms (note that these terms are not fully equivalent).6 In GENERALIZED

QUANTIFIER theory, the attributes ‘monotone decreasing’ (e.g. Barwise & Cooper 1981)or, alternatively, ‘downward monotonic’ (e.g. Gamut 1991) are commonly used for thequantifiers in question. The term ‘negative polarity context’ is also widespread, eventhough the relevant set of contexts comprises not only environments that are negative ina narrow (lexico-grammatical) sense, as in 31, but also ‘nonnegative’ contexts like spe-cific types of relative clauses (most notably those whose head noun is specified by a

[AQ2]

12 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

universal quantifier; see 32) and the protasis of a conditional clause (33). The term‘negative polarity context’ is motivated by the fact that the relevant contexts license‘negative polarity items’ such as any, ever, and so forth (cf. Ladusaw 1979, Krifka 1995as well as references cited there). As can be seen in 31–33, it is in such contexts thateven may combine with a focus value that is typically associated with (pragmatically)weak propositions (e.g. slapped as opposed to killed).

(31) I didn’t (ever) even [slap]F that man. (clausemate negation)

(32) Every person who (ever) even [slaps]F a fellow citizen will go to jail.(relative clause modifying a universally quantified head noun)

(33) If you (ever) even [slap]F your fellow citizen you will go to jail.(protasis of conditional clause)

We use the term ‘scale-reversing’ for the type of context in which ambiguities of evenarise, since scale reversal is their most important property in the present context (cf.Haspelmath 1997:114; for matters relating to scale reversal more generally, see Faucon-nier 1975b, 1978, König 1991, and Krifka 1995, among others). Contexts in which noscale reversal applies are called ‘scale-preserving’.Let us consider 24 for illustration once again, starting with the embedded sentence

Bill [slapped]F that man. The set of alternative predicates contrasting with the focusslapped contains elements such as killed and insulted, that is, other instances of themore general predicate harm. The set of context propositions may thus contain elementssuch as [[Bill killed that man]] and [[Bill insulted that man]]. When used to answer thequaestio To what extent did Bill harm that man?, the sentence with killed yields thestrongest contextual implications, and the one with insulted the weakest ones. This is il-lustrated in 34.

(34) Quaestio: To what extent did Bill harm that man?killed MAXIMAL

Bill slapped that man. degree of harm (contextual implications)insulted MINIMAL

If we consider the entire sentence in 24, the correlation between focus values andpropositional strength is inverted. Let us assume that the quaestio in this case is To whatextent are you convinced of Bill’s innocence?. The proposition [[I refuse to believe thatBill slapped that man]] triggers the contextual implication ‘I am absolutely convincedof his innocence’ (maximal certainty about the falsity of a minimal degree of [physical]violence). The proposition [[I refuse to believe that Bill killed that man]] merely li-censes the weaker inference that (the speaker believes that) Bill did not apply a maximaldegree of violence (i.e. commit murder). The ordering of the propositions under consid-eration is thus as shown in 35: insulted yields the pragmatically strongest propositionand killed the weakest one.

(35) Quaestio: To what extent are you convinced of Bill’s innocence?insulted MAXIMAL

I refuse to believe that Bill slapped that man. degree of conviction of innocencekilled MINIMAL

Scale-reversing contexts are thus characterized by the fact that focus values leadingto weaker degrees of strength in some subpart of a proposition lead to higher degrees ofstrength at the propositional level, and vice versa. This applies—among many othercontexts—to all types of environments illustrated in 31–33 above, that is, to clausematenegation, conditional clauses, and relative clauses modifying a universally quantifiedhead noun. What exactly triggers scale reversal and how such contexts can be defined

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 13

independently is an intricate question to which we would not like to contribute at thispoint. What matters to us is the observable fact that scales are reversed (cf. also Haspel-math 1997:114 for this position). As seen below, scale reversal is an important factor inthe distribution of SAOs, and a central parameter of variation in their typology.

33..22.. EXPLAINING THE AMBIGUITY OF even—SCOPE OR LEXICAL CONTENT? There are twoways of explaining the ambiguity of 24: (i) we can assume that the meaning of even isinvariant, and that the interpretive variability results from a difference in scope (the‘scope theory’); or (ii) we can assume that the two instances of even have different(though related) functions (the ‘lexicalist theory’). Both theories are briefly discussed inthis section.

THE SCOPE THEORY. One way of capturing the ambiguity of sentences like 24 (I refuseto believe that Bill even [slapped]F that man) is to assume that the two readings differwith respect to the scope of even. In the ‘insult reading’, even can be interpreted as tak-ing scope over the embedded proposition only ([[Bill slapped that man]]). In otherwords, only a subpart of the entire proposition functions as an argument of even. Ac-cordingly, it is this part of the proposition that is compared in terms of strength to thecorresponding context propositions (e.g. [[Bill insulted that man]]). The higher-levelpredication ([[I refuse to believe that …]]) is simply not relevant in this case. It merelyprovides an epistemic evaluation that is not directly related to the quaestio (To what ex-tent did Bill harm that man?). The interpretation of the entire sentence according to thisanalysis can thus be represented as shown in 36. Note that we assume that a sentencelike Bill even [slapped ]F that man is appropriate in a context in which only insulted (orsome other predicate leading to weaker contextual implications than slapped ) is avail-able in the common ground (cf. the definition in 18), while a value such as killed is notunder discussion.

(36) a. [[I refuse to believe S]],[[S]] = EVEN([[Bill [slapped]F that man]])

b. Text proposition: τS = [[Bill slapped that man]]c. Strongest context proposition: κS = [[Bill insulted that man]]d. Pragmatic presupposition: τS >s all κ ∈ KS

If even is (syntactically) attached to the higher predicate (I even refuse to believe thatBill slapped that man, cf. 28a above), it is the entire sentence that functions as a basis ofthe partial ordering in terms of pragmatic strength. In other words, even has sententialscope. This is illustrated in 37. The text proposition [[I refuse to believe that Billslapped that man]] is pragmatically stronger than the strongest context proposition [[Irefuse to believe that Bill killed that man]] (in this case, the focus value insulted is notunder discussion/in the common ground).

(37) a. EVEN([[I refuse to believe that Bill [slapped]F that man]]) (=S )b. Text proposition: τS = [[I refuse to believe that Bill slapped

that man]]c. Strongest context proposition: κS = [[I refuse to believe that Bill killed

that man]]d. Pragmatic presupposition: τS >s all κ ∈ KS

According to the scope theory, I refuse to believe [that Bill even [slapped]F that man]is interpreted just like 28a, that is, as represented in 37b–d. Even though even formspart of the embedded clause, it is regarded as taking scope over the entire sentence. Thescope theory thus allows us to assign a single function or meaning to even. The ambi-guity of 24 can be regarded as resulting from the fact that even may take variable scope.

7 These locality principles prevent S-scope from transcending certain bounding nodes, the most importantof which [are] … the Sn-nodes of [X-bar] theory (0 ≤ n)’ (Jacobs 1984:187). ‘The scope of a focus particlecannot only be determined by another operator co-occurring with the particle in the same clause, but also bycertain “bounding nodes” (NP, S)’ (König 1991:50).

8 Taglicht (1984) has pointed out (apparent?) counterexamples to this generalization (cf. also König 1991).The minimal pair They were advised to learn only Spanish (wide scope of only) vs. They were advised to onlylearn Spanish (narrow scope) has been discussed under the label of ‘scope-fixing effect’ (cf. also Rooth1985:145–54).

14 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

Among the most prominent advocates of this analysis are Horn (1971), Kay (1990),Wilkinson (1996), and Guerzoni (2003). In terms of the scope theory, the distribution ofeven can invariably be characterized as shown in 38. The proposition in its scope ispragmatically stronger than all context propositions.

(38) EVEN([[S]])Propositional content: τS (= [[S]])Pragmatic presupposition: τS >s all κ ∈ KS

While allowing for a monosemous analysis of even, the main problem of the scopetheory is that even is regarded as exhibiting a scope behavior that is in some way ‘non-canonical’. It is often assumed that the scope of focus operators is syntactically re-stricted, in particular insofar as ‘sentential focus adverbs … do not take scope outsidetheir clause’ (Hoeksema & Zwarts 1991:58; cf. also Jacobs 1983, 1984, König 1991,von Stechow 1991:812).7 For instance, the scope of only is generally clause-bound. Thiscan be illustrated with the pair of examples in 39.8 Unlike in a corresponding sentenceswith even (cf. 24), there is no ambiguity, since 39a cannot be equivalent to 39b; that is,only can only be interpreted within its ‘minimal clause’ (i.e. the smallest clausal con-stituent containing it).

(39) a. I refuse to believe that Bill only slapped that man. ( … I’m sure that he also shot at him.)

b. I only refuse to believe that Bill slapped that man.( … but I do not refuse to believe that Bill insulted him.)

With respect to the ambiguity of even in sentences like 24, advocates of the ‘scope the-ory’ are thus forced to assume that even behaves differently from other focus-sensitiveoperators like only: ‘One suspicious feature of the scope theory is that it attempts to solvea semantic problem by assigning wide scope to an element without any independent jus-tification that this sort of exceptional scope assignment is actually possible’ (Rullmann1997:48). This ‘disadvantage’ of the scope theory has even been noticed by proponentsof the latter. For instance, Guerzoni remarks:

Summing up, the scope theory needs to stipulate that the movement of even is less constrained than move-ment operations of a more familiar type. This seems to deprive this theory of much of its initial appeal, acriticism that has often been brought into the debate by defenders of the alternative lexical ambiguity view.(2003:94)

This alternative analysis, to which we now turn, is not without problems either, however.

THE LEXICALIST THEORY. An alternative analysis to the scope theory is to regard theambiguity of 24 as a lexical one. According to this line of reasoning, prominently ar-gued for by Rooth (1985), Rullmann (1997), and Giannakidou (2007), there are twolexical entries for even, one of which is a negative polarity item—say, ‘negative polarityeven’, or Neg-even for short—while the other one is either polarity neutral (cf. von Ste-chow 1991:811) or a positive polarity item (Giannakidou 2007). We call it ‘Pos-even’ inthe following. This analysis is often called the ‘NPI-theory’ (e.g. Rooth 1985, Rullmann

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 15

1997), but we refer to it as the ‘lexicalist theory’, as it attributes the ambiguity of caseslike 24 to variation in the lexical content of even.Pos-even can be analyzed in a straightforward way. It takes ‘canonical scope’, indicat-

ing that the proposition contained in its scope is stronger than all context propositions.Neg-even is interpreted differently. It is assumed to take particularly WEAK propositionsas its arguments. Given that Neg-even is a negative polarity item and, as such, restrictedto scale-reversing contexts, the sentence-level proposition will be particularly strong (butsee below for a complication).The scope relationships corresponding to the murder reading of 24 can thus be repre-

sented as shown in 40. Neg-even indicates that from among the propositions under con-sideration, [[Bill slapped that man]] is the weakest one (40b). In a scale-reversingcontext, the weakest embedded proposition will yield the strongest sentence-levelproposition (40c).

(40) a. S1 = [[I refuse to believe that S2]][[S2]] = Neg-EVEN([[Bill [slapped]F that

man]])b. Text proposition of S2: τS2 = [[Bill slapped that man]]Pragmatic presupposition: τS2 <s all κ ∈ KS2

c. Text proposition of S1: τS1 = [[I refuse to believe that Bill slappedthat man]]

Pragmatic presupposition: τS1 >s all κ ∈ KS1The main advantage of the lexicalist theory vis-à-vis the scope theory is that it is

prima facie not forced to attribute noncanonical scope behavior to even. Its most obvi-ous disadvantage is that it assumes two lexical (sub)entries for even. Moreover, there isa second problem of the lexicalist theory, which has been pointed out by proponents ofthe scope theory (e.g. Guerzoni 2003). It is not sufficient for Neg-even to occur in ascale-reversing context in combination with a particularly weak focus value; it is alsonecessary that the sentence-level proposition be particularly strong, and the latter con-dition does not necessarily follow from the former. The following pair of examples (dueto Irene Heim and discussed in Guerzoni 2003:95) illustrates this.

(41) a. #Every student that even handed in [one]F assignment got an A.b. #Every student that even handed in [one]F assignment was wearing blue

jeans.The relative clause of 41b is pragmatically weaker than all context propositions, but thesentence-level proposition is not pragmatically strong. This is due to the fact that there isno (obvious) implicational relation holding between the two states of affairs expressed,which means that there is no scale reversal. The fact that Neg-even is sensitive to the in-terpretation of the sentence-level proposition seems to speak against regarding it as tak-ing narrow scope. One way of solving this problem in a lexicalist approach would be toassume a licensing condition for Neg-even requiring that this operator be contained in ut-terances making particularly strong statements. Such an analysis could be implementedby assuming that reference to ordered sets of alternatives is an independent property ofutterances, and that scalar operators are merely (lexico-grammatical) reflexes of specifictypes of utterances. This line of reasoning has been pursued by Krifka (1995) (buildingon Fauconnier 1975b) in his analysis of negative polarity items. We do not go into any de-tail here, and refer the reader to Krifka 1995 for more information.

16 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

33..33.. THE DIVISION OF LABOR BETWEEN SCOPE AND LEXICAL CONTENT. Having presentedsome (but not all) pros and cons for and against the scope theory and the lexical theoryof the ambiguity of even, we have to consider the consequences for our crosslinguisticstudy. As has been seen, there are two factors potentially determining the interpretationof SAOs: scope and lexical content. In order to carry out a comparative study, it is thusnecessary to abstract away from one of these factors. In what follows, we regard thescope properties of SAOs as a given and treat their lexical content as a variable. As isseen later (§9.3), however, a (historically and crosslinguistically) adequate theory ofscalar additive operators should take both factors into account.Regarding scope as a given means to assume that the scope of an operator is a func-

tion of its syntactic position. We make two assumptions about the relationship betweensyntax and scope. First, we assume that the scope of an SAO cannot extend beyond the‘minimal clause’ containing that operator (cf. Jacobs 1983, 1984, Hoeksema & Zwarts1991:58, König 1991, von Stechow 1991). Our second assumption is that there are lan-guage-particular rules of scope assignment within the minimal clause. For example, inGerman relative scope is reflected in left-to-right sequence in the middle field (cf.König 1991:47; see also Jacobs 1982, Lerner & Sternefeld 1984). While other Euro-pean languages basically behave like German (see Progovac 1994:31 on Bosnian-Croatian-Serbian, Junghanns 2006 on Czech), in English the relationship between syn-tactic structure and scope relations is more complicated, and various, partly competing,factors seem to play a role (linear precedence, c-command, intonation; see König1991:47–48 for discussion; see also Bayer 1996 for an analysis of matters of scope interms of ‘LF-movement’). For reasons of space, we cannot enter into this discussionhere. Suffice it to say that in a comparative study like this, language-specific particular-ities need to be taken into account when scope relations within the minimal clause areconcerned.The syntactic domain within which focus-sensitive operators ‘canonically’ take scope

is called the ‘local scope domain’, or simply ‘local domain’, abbreviated as ‘LD’. Thisis defined in 42.

(42) The LOCAL DOMAIN of a scalar additive operator O is the syntactic domainwithin which members of the same syntactic class as O canonically takescope.

The proposition within the local domain of an SAO we call the ‘local proposition’. Thehighest proposition containing both the SAO and its propositional argument we call the‘host proposition’. The syntactic domain corresponding to the host proposition we callthe ‘host domain’. We can now capture the difference between Pos-even and Neg-evenpointed out above in distributional terms as follows: in the case of Pos-even, the localproposition ([[Bill slapped that man]]) is pragmatically stronger than all context propo-sitions, while in sentences with Neg-even, it is the host proposition ([[I refuse to believethat Bill slapped that man]]) that is pragmatically strong, while the local proposition ispragmatically weak. This distinction provides the basis of the typology presented in §4.

44.. A TYPOLOGY OF SCALAR ADDITIVE OPERATORS. The foregoing discussion has shownthat scalar additive operators should not be regarded as a single type of expression, butrather as a family of items that share one feature—that of occurring in propositions thatare, at some level, ‘pragmatically strong’—but vary among other dimensions. Our ty-pology is based on two parameters: (i) the degree of pragmatic strength associated withthe local proposition (§4.1), and (ii) the presence vs. absence of clausemate negation inthe minimal clause (§4.2). On the basis of these parameters, five types of SAOs can be

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 17

distinguished. In §4.3, an additional type of operator is introduced, which straddles theline between the major types identified in §§4.1 and 4.2. Moreover, the family of scalaradditive operators comprises two further important types, ‘scalar negators’, which arenot merely licensed under negation but rather trigger negation (§4.4), and ‘general ad-ditive operators’, which are used in both scalar and nonscalar contexts (§4.5).

44..11.. BEYOND-OPERATORS VS. BENEATH-OPERATORS. As pointed out in §3.1, the operatorso much as is often (at least approximately) equivalent to even in scale-reversing con-texts. So much as is in fact distributionally restricted to such contexts; that is, it is a neg-ative polarity item. Similar operators exist in many other European languages, forexample, Italian (anche solo), Czech (i jen), German (auch nur), and Greek (kan) (cf.König 1991, Rullmann 1997, Guerzoni 2003, Schwarz 2005, Giannakidou 2007,Nakanishi 2008). Just like Engl. so much as, Germ. auch nur occurs only in combina-tion with weak local propositions. When the local proposition is pragmatically strong,the particle sogar (or, alternatively, selbst) is used as an SAO. Accordingly, the tworeadings of 24 above are (obligatorily) differentiated lexically. This is illustrated in 43.

(43) a. ‘insult context’Ich weigere mich zu glauben, dass Willie den Mann sogar geohrfeigt hat.I refuse me to believe that Bill the man even slapped has‘I refuse to believe that Bill even/*so much as slapped that man.’

b. ‘murder context’Ich weigere mich zu glauben, dass Willie den Mann auch nurI refuse me to believe that Bill the man so much as

geohrfeigt hat.slapped has

‘I refuse to believe that Bill even/so much as slapped that man.’Elements of the type of sogar we call ‘BEYOND-operators’ and elements of the auch

nur-type we call ‘BENEATH-operators’. The attributes ‘BEYOND’ and ‘BENEATH’ refer tothe relation between the local proposition (the proposition denoted by the material inthe local domain) and the relevant context propositions. The local proposition of aBEYOND-operator is located higher than (‘beyond’) the corresponding context proposi-tions on the scale of pragmatic strength, whereas the local proposition of a BENEATH-operator is located lower than (‘beneath’) the relevant context propositions on thatscale. BEYOND-operators are defined in 44 and illustrated in 45 with an example fromItalian (perfino).

(44) BEYOND-operators are SCALAR ADDITIVE OPERATORS whose LOCAL PROPOSI-TION is PRAGMATICALLY STRONGER than all corresponding CONTEXT PROPOSI-TIONS.a. [ … [LD BEYOND … ]]b. τLD >s all κ ∈ KLD

(45) Italian[LD PERFINO [i venti e le onde]F gli ubbidiscono].

even the winds and the waves him obey‘Even the winds and the water obey him.’ (Letizia Vezzosi, p.c.)

Note that BEYOND-operators are found in both scale-preserving and scale-reversingcontexts. An example of the latter case was provided by the (German version of the) ‘in-sult reading’ of 24 (cf. 43a). By contrast, BENEATH-operators are distributionally re-stricted to scale-reversing contexts. They are defined in 46, and a relevant example fromItalian (anche solo) is given in 47.

(i) universal SAOs

(ii) beyond-operator (iii) beneath-operator

FIGURE 1. Major types of SAOs.

44..22.. SUBCLASSIFYING BENEATH-OPERATORS: NEGATIVE AND NONNEGATIVE POLARITY. BE-NEATH-operators are, by definition, restricted to scale-reversing contexts. Subtypes ofBENEATH-operators can be established on the basis of the distinction between negativeand nonnegative polarity. Specific types of operators can only be used in combinationwith ‘clausemate negation’, that is, when their minimal clause is (directly) negated. Inother words, these elements are ‘superstrong NPIs’, in terms of Zwarts 1998 (cf. alsovan der Wouden 1997; similarly, Progovac 1994 identifies a class of NPIs that requireclausemate negation). Such operators are widespread among Germanic languages andare found inter alia in German ((nicht) einmal), Dutch ((niet) eens), and Norwegian((ikke) engang). The restriction of the German operator einmal to directly negatedclauses is illustrated in 48.9

(48) a. *Nicht einmal [seine Hunde]F gehorchen ihm.*not even his dogs obey him.‘Not even his dogs obey him.’

9 Note that the distributional restriction of einmal to a single context (immediately following the clausenegator nicht) has led some authors to analyze nicht einmal as a single lexical item (e.g. König 1991:80; seealso Schwarz 2005:126, n. 1).

18 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

(46) BENEATH-operators are SCALAR ADDITIVE OPERATORS whose LOCAL PROPOSI-TION is PRAGMATICALLY WEAKER than all corresponding CONTEXT PROPOSI-TIONS, while their HOST PROPOSITION is PRAGMATICALLY STRONGER than allcorresponding CONTEXT PROPOSITIONS.a. [HD ... [LD BENEATH ... ]]b. τLD <s all κ ∈ KLD

c. τHD >s all κ ∈ KHD

(47) [HP È una vergogna [LD ANCHE SOLO [parlare]F delle cose che certais a shame also only speak of.the things that certaingente fa di nascosto]].people do in secret

‘It is a shame to even speak about the things that many people do in secret.’(Letizia Vezzosi, p.c.)

Operators like Germ. sogar/auch nur and Ital. perfino/anche solo are thus restrictedto specific types of local propositions, and operators of the auch nur/anche solo-typeare moreover restricted to scale-reversing contexts. By contrast, Engl. even can combinewith both strong and weak propositions within the local domain. On the basis of theconsiderations made so far, we can thus distinguish three types of SAOs: (i) SAOs thatoccur in all types of environments, such as Engl. even; we call these operators ‘univer-sal SAOs’; (ii) BEYOND-operators such as Germ. sogar, whose local proposition is prag-matically relatively strong; and (iii) BENEATH-operators such as auch nur, whose localproposition is pragmatically relatively weak while their host propostion is relativelystrong. This first distinction is shown in Figure 1. Note that Figure 1 does not classifythe operators themselves, but the contexts in which the operators are used.

10 Note that 48b is grammatical in the irrelevant reading in which einmal is interpreted as an event quanti-fier meaning ‘once’. This reading is characterized by a different intonation, however.

11 Note that Giannakidou (2007) takes presuppositions, not clause polarity, as a criterion distinguishingsubtypes of scalar additive operators.

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 19

b. *Es ist undenkbar, dass ihm einmal [seine Hunde]F gehorchen.*it is unthinkable that him even his dogs obeyint.: ‘It is inconceivable that even his dogs obey him.’10

While einmal is thus restricted to directly negated clauses—and in fact to a positionfollowing the clause negator nicht—auch nur is used in both negative and nonnegativecontexts (49a,b). It is, however, not normally used when einmal is possible (50), proba-bly due to a blocking effect (i.e. einmal is preferred because it is more specialized thanauch nur).

(49) a. auch nur in negative sentencesNie haben auch nur seine Hunde ihm gehorcht.never have also only his dogs him obeyed‘Never have even his dogs obeyed him.’

b. auch nur in nonnegative sentencesWenn du sie auch nur ansiehts, bekommst du Ärger.if you her also only look.at get you trouble‘If you even look at her, you’ll get into trouble.’

(50) Nicht einmal/*auch nur seine Hunde gehorchen ihm.not once/also only his dogs obey him‘Not even his dogs obey him.’

The difference between Germ. auch nur and (nicht) einmal provides us with a secondparameter of variation subclassifying BENEATH-operators: both auch nur and einmal areBENEATH-operators, but they differ in their distributional restrictions with respect to thepresence or absence of clausemate negation. Parallel distinctions are found in manyother languages (see Giannakidou 2007 for a similar classification of SAOs in Greek).11

Operators such as auch nur we call ‘neutral BENEATH-operators’, as they are ‘polarity-neutral’; elements of the type of einmal we call ‘negative BENEATH-operators’.The question arises of whether the third possible type—a BENEATH-operator that is re-

stricted to nonnegative scale-reversing contexts—is also attested in the languages ofEurope. In fact, such elements do seem to exist. With respect to the Greek particle esto,Giannakidou (2007:43) remarks that it ‘is a curious PI [polarity item]—bad in both positive and negative sentences’ (see 51a,b). However, it ‘improves in polarity environ-ments that are not negative, but nonveridical … e.g. questions, imperatives, subjunc-tives, protasis of conditionals, and with modal verbs’ (cf. the conditional in 51c).

(51) a. esto cannot be used in scale-preserving (veridical) contexts?I Maria efaje esto to pagoto.?DET Maria ate even DET ice.creamint.: ‘Mary ate even the ice cream.’

b. esto cannot be used with clausemate negation?I Maria dhen efaje esto to pagoto.?DET Maria not ate even DET ice.creamint.: ‘Mary did not even eat the ice cream.’

(i) universal SAOs

(ii) beyond-operator (iii) beneath-operator

(iv) negative (v) nonnegative

FIGURE 2. Five types of SAOs.

44..33.. EXTENDED BEYOND-OPERATORS. The typology displayed in Fig. 2 meets with aproblem when specific East and South East European languages are considered. For ex-ample, the BEYOND-operators of Czech (dokonce), Romanian (chiar), and Bosnian-Croatian-Serbian (čak) may occur in the scope of clausemate negation, thus beingassociated with a weak local proposition. We consider Czech dokonce for illustration.

20 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

c. esto can be used in conditionalsAn diavasis esto ke mia selida ap’ afto to vivlio, katiif you.read even also one page PREP DEM DET book something

tha mathis.you.will.learn‘If you read even one page from this book, you will learn something.’

(Giannakidou 2007:43, 66)

Another element that seems to qualify as a BENEATH-operator with a distributional re-striction to nonnegative contexts has been reported to exist in Hungarian. According toAbrusán, the Hungarian particle akár ‘cannot appear in a minimal sentence with nega-tion … However, it can appear in DE [downward entailing/scale-reversing] contextsother than clausemate negation, and in questions’ (2007:4). The examples in 52 illus-trate this.

(52) Hungarian (Abrusán 2007:4)

a. akár is not possible with clausemate negation*Péter nem üdvözölheti akár Marit sem.*Peter not greet.can even Mari eitherint.: ‘Peter may not even greet Mari.’

b. akár with higher negation*Nem igaz, hogy Péter akár (csak) egy példát is megoldott.*not true that Peter even (only) one exercise too solved‘It is not true that Peter solved even one exercise.’

c. akár in conditionals*Ha akár (csak) egy ember is megszólal, kiüríttetem a termet.*if even (only) one person too speaks empty.1SG the room‘If even a single person says a word, I will empty the room.’

Using the criterion of (negative vs. nonnegative) polarity, we can thus distinguishthree types of BENEATH-operators: (i) neutral BENEATH-operators, such as Germ. auchnur, which are used in both negative and nonnegative contexts; (ii) negative BENEATH-operators, which are only used in the scope of (clausemate) negation (e.g. Germ. (nicht)einmal); and (iii) nonnegative BENEATH-operators, which are only used in the absence ofclausemate negation (e.g. Hung. akár). Our typology of SAOs can now be extended asshown in Figure 2 (once again, what is classified is not the operators themselves but thecontexts licensing them).

12 Junghanns (2006) points out that specific Czech elements take scope beyond their c-command domain,especially sentence adverbs. To the best of our knowledge, focus particles exhibit canonical scope behavior.

13 If we regarded ani as a local negator, we could of course argue that—irrespective of the presence of theclause negator není—the minimal scope domain of dokonce in 55b is basically identical to that in 55a (thereis no x). This assumption, however, would be at variance with most theories of negative concord, which regardthe ‘internal’ negative element (here aní) as a reflex of a higher negator (here the prefix ne-; see Corblin et al.2004 for an overview of theories of negative concord). Under an NPI-analysis of negative concord along thelines of Laka 1990 and Giannakidou 2000, ani is interpreted as an existential quantifier in 55b.

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 21

According to our diagnostics, dokonce is a BEYOND-operator and can thus be used withlocally strong focus values, as in 53.

(53) CzechKdo to jen je, že dokonce odpouští hříchy?who this only is COMP even forgives sins‘Who is this? He even forgives sins?’

(Luke 7:49; Český ekumencký překlad, from http://www.bibleserver.com)

Just like Germ. sogar, dokonce can also be used in (nonnegative) scale-reversing con-texts, but only with a BEYOND-semantics. Consider 54.

(54) O tom, co oni dělají potají, je odporné dokonce mluvit.PREP that REL they do secretly is disgusting even speaklit.: ‘It is shameful to evenBEYOND speak about what they do in secret.’

(V. Elšik, p.c.)

Example 54 suggests that ‘speaking (about these things)’ is stronger than all contextpropositions. This sentence is thus appropriate only if an even weaker focus than[[speak about]] is under discussion (e.g. It is shameful to so much as think about thesethings … ). What 54 shows is that dokonce is a BEYOND-operator, just like Germ. sogar.However, there is a difference between Germ. sogar and Czech dokonce: under spe-

cific circumstances—most notably, when the focus is locally negated by the ‘scalarnegator’ ani (cf. §7.3)—dokonce, unlike sogar, can be used in the scope of clause nega-tion. Consider the two examples in 55.

(55) a. Dokonce tady není ani [voda]F k napití.even there NEG.EX not.even water PREP drink‘There is not even water to drink.’

b. Není tady dokonce ani [voda]F k napití.NEG.EX there even not.even water PREP drink‘There is not even water to drink.’ (V. Elšik, p.c.)

The sentences in 55 are semantically equivalent, even though there is a syntactic differ-ence: in 55a, dokonce precedes the negative existential predicate není. According torules of scope mapping generally assumed for Czech (cf. Junghanns 2006), it is thusoutside the scope of clausemate negation (56a). In 55b, the negative existential predi-cate není precedes the SAO dokonce, which is thus expected to be contained in thescope of the former12 (56b; note that we follow those who regard the internal negator innegative concord—ani, in 55b—as an existential quantifier).13

(56) a. DOKONCE (NEG ([[there is [drinkwater]F]]))b. NEG (DOKONCE ([[there is [drinkwater]F]])).

No problem arises in 55a/56a. The text proposition τLD = [[there is no drinkwater]] ispragmatically stronger than any context proposition, say, κLD = [[there is no wine]] (as-suming the quaestio To what extent did you suffer privations?). The condition τLD >s κLDis thus met. In 55b and 56b, however, dokonce appears to be associated with a weak text

14 For discussion of similar elements in Ibero-Romance languages, see Vallduví 1994 on Catalan and Span-ish and Herburger 2003 on Spanish.

22 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

proposition, that is, [[there is drinkwater]], so τLD >s κLD does not hold, and dokoncedoes not exhibit the distributional behavior expected of a BEYOND-operator.The ‘noncanonical’ scope behavior of operators like Czech dokonce has probably re-

sulted from clause-internal processes of (syntactic) reanalysis (cf. §9.1). We assumethat operators like dokonce are BEYOND-operators with an ‘extended distribution’. Suchoperators are thus called ‘extended BEYOND-operators’, as opposed to ‘basic BEYOND-operators’ like German sogar. The distribution of extended BEYOND-operators can becharacterized with respect to the ‘minimal clause’ (MC) containing them. Even thoughthe local domain of dokonce is weak in 55b ([[there is drinkwater]]), the proposition de-noted by the minimal clause is strong ([[there is no drinkwater]]). The proposition denoted by the minimal clause is called the ‘minimal host proposition’. ExtendedBEYOND-operators can now be characterized as in 57.

(57) Extended BEYOND-operators are SCALAR ADDITIVE OPERATORS whose MINIMALHOST PROPOSITION is PRAGMATICALLY STRONGER than all relevant contextpropositions.a. [MC ... (NEG) ... [LD EXT-BEYOND ... ]]b. τMC >s κMC

44..44.. SCALAR NEGATORS. Many (especially South East European) languages have oper-ators that not only are licensed in negative contexts—like Germ. einmal—but that alsoINCORPORATE an expression of negation. In order to understand the difference, let uscompare Germ. einmal to Greek oute (see also Haspelmath 1997:212–23).14 On theface of it, the two elements are very similar (see also Giannakidou 2007:43). Considerthe Greek example 58 and its German translation in 59.

(58) Modern GreekI Maria dhen efaje oute (kan) to pagoto.DET Maria not ate not.even (even) DET ice.cream‘Maria didn’t even eat the ice cream.’ (Giannakidou 2007:43)

(59) GermanMaria hat nicht einmal das Eis gegessen.Maria has not evenNEG the ice.cream eaten‘Maria didn’t even eat the ice cream.’

There is, however, a crucial difference between Greek oute and Germ. einmal: unlikeeinmal, oute can trigger clause negation by itself if it occupies the right syntactic posi-tion. This is illustrated in 60 (the clause negator dhen is optional). The correspondingGerman sentence does not even make sense if einmal is interpreted as an SAO (it is fineif einmal is regarded as an event quantifier meaning ‘once, on one occasion’).

(60) Oute (kan) ti Maria (dhen) proskalese o pritanis.not.even (even) DET Maria (not) invited the dean‘The dean didn’t even invite Maria.’ (Giannakidou 2007:58)

(61) Einmal hat der Dekan Maria eingeladen.once has the dean Maria invited‘On one occasion, the dean invited Maria.’ (not: ‘The dean did not even in-vite Maria.’)

As is also pointed out by Giannakidou (2007), occurrences of oute in the scope of nega-tion as illustrated in 60 are only possible because Greek has negative concord. Greek

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 23

oute and similar elements thus correspond to the combination of the German negatornicht and the (negative) SAO einmal, rather than just einmal. Unlike plain einmal, theyincorporate a negative morpheme and an SAO. Accordingly, they cannot be regarded asa special type of SAO, but should be analyzed as portmanteau operators containing a(specific type of) SAO. We call such elements ‘scalar negators’.Scalar negators are typically characterized by a type of polysemy involving negative

conjunction (‘nor … ’), contrastive negative coordination (‘neither … nor’; cf. Haspel-math 2007), and scalar negation (‘not even’). This three-way polysemy is here illus-trated with examples from Spanish (see also Vallduví 1994 on Spanish ni).

(62) Spanish (Monica Zepeda, p.c.)

a. negative conjunctionNo lo sé; ni lo sabe él.not it I.know nor it he.knows he‘I don’t know; nor does he (know it).’

b. contrastive negative coordinationNi el uno ni el otro lo sabe.not the one not the other it know‘Neither one nor the other knows it.’

c. scalar negationNi el Papa lo sabe.not.even the Pope it knows‘Not even the Pope knows it.’

44..55.. GENERAL ADDITIVE OPERATORS. The last type of expression to be introduced in ourtypology can be illustrated with examples from Latin (et) and Ancient Greek (kai) (see63 and 64, respectively). These additive operators are used in both scalar and nonscalarcontexts; that is, they neutralize the functions of even and also, perhaps in interactionwith prosodic cues (for instance, scalar readings might have been associated with anemphatic focus accent). We call such operators ‘general additive operators’.

(63) LatinFas est et ab hoste doceri.right is even from enemy learn‘It is rightful to learn even from an enemy.’ (Ovidius, Metamorphoses IV, 428)

(64) Ancient GreekPotapos estin houtos hoti kai hoi anemoi kai hē thalassa autōwho is this that even the winds and the sea him

hypakouousin.obey

‘Who is this that even the winds and the sea obey him?’(Matt. 8:27; from http://www.greekbible.com)

There is one contemporary European language in our sample that does not seem todistinguish at all between scalar and nonscalar additive expressions, namely Basque. Ituses the operator ere in all cases.

(65) Basque (Hualde & Ortiz de Urbina 2003:556)

a. Gure ikasleak ere, joan dira.our students also/even go AUX

‘Our students, too, they went.’b. Ez da matrikulatu ere (egin).not AUX register even do‘He didn’t even register.’

general additive operator

nonscalar AOs universal SAO

beyond-operator beneath-operator

negative nonnegative

FIGURE 3. A typology of additive operators.

55.. UNIVERSAL SAOS IN THE LANGUAGES OF EUROPE. We now proceed to determine thedistribution of the types of operators distinguished above in a sample of forty major Eu-ropean languages, starting with universal SAOs. Remember that these operators can beused in all types of contexts, that is, in scale-preserving and scale-reversing ones, withas well as without clausemate negation. We use here scale-preserving and nonnegativescale-reversing contexts to illustrate the distributional properties of universal SAOs,since our data suggest that there is an implicational relation to the effect that if an SAOcan be used in scale-preserving contexts (as a BEYOND-operator) and in nonnegativescale-reversing contexts (as a nonnegative BENEATH-operator), it can also be used in

24 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

c. Hitz bat ere egiten badu, akabatuko dut.word one even do.IMPF if.AUX kill.FUT AUX

‘If he says even one word, I’ll kill him.’Moreover, many (especially Slavic and (non-Slavic) Balkan) languages tend to use

general additive operators in scale-reversing contexts even though universal SAOs orBENEATH-operators are also available. The following examples from Bulgarian—wherethe general additive operator i is used instead of the more specific choices daže (i) anddori (i)—illustrate this.

(66) Bulgariana. Kakave toya, che i vetrovete i vylnite mu sewho is this COMP also/even winds and waves him REFL

pokoryavat?obey

‘Who is this, that even the winds and the water obey him?’(Luke 8:27; Bulgarian Bible, from http://www.biblegateway.com)

b. Zashoto tova koeto skrishom vŭrshat nepokornite, sramotno e iCONJ that what secretly they.do disobedient shameful is and/even

da se govori.COMP MID speak

‘For what the disobedient do in secret is shameful to even speak about.’(Eph. 5:12; Bulgarian Bible, from http://www.biblegateway.com)

44..66.. SUMMARY. Our typology can be summarized as shown in Figure 3. General addi-tive operators (AOs) cover the whole range of contexts: that is, they have both scalar andnonscalar uses. The sister category of universal SAOs is the one of ‘nonscalar additiveoperator’. It contains elements like English also, that is, additive operators that do nothave scalar uses. In Fig. 3, no distinction is made between negative BENEATH-operatorsand scalar negators, as these types of operators are complementarily distributed in thelanguages of our sample.

15 Zokén was originally polymorphemic but can no longer be so analyzed from a synchronic point of view.It derives from a combination of the third-person singular relativizer zo and the particle kén, which covers avariety of functions. Originally, zokén probably meant ‘what is comparable to’. We owe this information toPeter Schrijver.

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 25

negative scale-reversing contexts (as a negative BENEATH-operator). As seen below, thisgeneralization is ultimately a historical one, as it reflects the pathway of developmentfrom BEYOND-operators to universal SAOs (cf. §9).Universal SAOs are often particles (i.e. single invariant words). Such ‘monomor-

phemic universal SAOs’ are found in many West European languages.Among the Celticlanguages, Modern Breton (zokén),15 Irish Gaelic ( féin), and Manx (hene) exhibitSAOs of this type. English even, French même, and Spanish aun also qualify as univer-sal SAOs. Examples of BEYOND-readings of these operators were provided in 1, 2, and 4above. The use of Spanish aun in a nonnegative BENEATH-context is illustrated in 67.

(67) Spanish… da vergüenza aun mencionar lo que los desobedientes hacen en … it.gives shame even mention PRON REL the disobedient do in

secreto.secret

‘ … it is shameful to even mention what the disobedient do in secret.’(Eph. 5:12; La Biblia al día, from http://www.bibleserver.com)

Welsh uses the particle group hyd yn oed (‘as far as, up or down to (and including),even’, according to the Dictionary of the Welsh Language, p. 1950) as a universal SAO.It is illustrated in 68a for BEYOND-readings and 68b for BENEATH-readings.

(68) Welsha. Mae e wedi gwneud ei wely e, hyd yn oed.

COP he after making his bed he even‘He’s even made his bed.’ (Dictionary of the Welsh Language, s.v. hyd)

b. Mae’n warthus hyd yn oed sôn am y pethau mae pobl ynis shameful even talk about DET things is people PREP

eu gwneud o’r golwg.their do out.of view

‘It is disgraceful to even mention what they do in secret.’(Eph. 5:12; Welsh Bible, translated by Arfon Jones, from http://www.beibl.net)

Universal SAOs can also be found in (geographically) Baltic languages. Estonianuses isegi in both scale-preserving and scale-reversing contexts. This word is composedof the intensifier ise and the suffix -gimentioned in §1.2. Latvian uses pat, and Lithuan-ian net as a universal SAO, as in 69.

(69) Lithuaniana. Kas Jis per vienas, kad net vėjai ir ežeras Jo klauso?what he PREP one COMP even wind and lakes him obey‘What kind of man is this? Even the winds and the waves obey him!’

(Matt. 8:25; Lithuanian Bible, from http://unbound.biola.edu)b. Nes kà jie slapèia daro, gëda net sakyti.

CONJ REL they secretly do shame even speak.‘For it is disgraceful even to mention what they do in secret.’

(Eph. 5:12; Lithuanian Bible, from http://unbound.biola.edu)

16 Poljakova (2000), like Haspelmath (1997:225–26), classifies daže as a positive polarity item. Poljakova(2001), however points out that daže can occur in negative polarity contexts when it combines with i. See also§9.1 for more information.

17 Polish nawet may have originally been a prepositional phrase of the form *[PP na [NP wet]]. According toVasmer (1953:193), Old Russian vet and Old Bulgarian vĕt’ mean ‘council, contract’ (cf. also Engl. bet,Germ. Wette), that is, the endpoint of negotiations. The etymology underlying nawet and so forth could thusbe similar to that of Span. hasta (‘as far as’) and Swed. till och med (‘up to and including’).

18 There seems to be some dialectal variation in Albanian. Buchholz et al. 1992 lists édhe (with an accenton the first syllable) and edhé (accent on last syllable) separately. Édhe is translated as ‘noch; schon, sogar’,and edhé as ‘und, auch, und auch’. Other dictionaries do not make that distinction. Madjé ‘indeed’ can op-tionally be added to édhe.

19 We thank Čeyhan Temürcü for help with the Turkish glosses.

26 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

Most Slavic languages likewise have universal SAOs. Russian and Bulgarian use dažefor this purpose (cf. Poljakova 2000, 2001 on Russian).16 Bulgarian also has dori (cf.also Macedonian duri), which probably has the same lexical source as Russian and Bul-garian daže (cf. §9.1). The universal SAOs of Ukrainian (navit’ ), Belarusan (navat),and Polish (nawet) have most likely derived from the combination of a preposition anda noun.17

Among the major (non-Slavic) South East European languages, universal SAOs canbe found in (Tosk) Albanian ((madjé) édhe)18 and Turkish (bile).

(70) Turkish19

a. Bu adam kim ki, rüzgar-a, su-ya bile buyruk veriyor, onlar dathis man who that wind-DAT water-DAT even order gives they CONJ

söz-ü-nü dinliyor.word-3.POSS-ACC obey

‘Who is this man, who commands even the wind and the water andthey obey him?’

(Luke 8:25; Turkish Bible, from http://www.bibleserver.com)b. Karanlık-ta-ki-ler-in gizlice yap-tık-ları-ndan söz etmek biledark-LOC-REL-PL-GEN secretly do-F.NOM-3PL.POSS-ABL word do even

ayıp-tır.shame-is

‘For it is shameful to even mention what they do in secret.’(Eph. 5:12; Turkish Bible, from http://www.bibleserver.com)

The universal SAOs from the languages of our sample are summarized in Map 2. Thelanguages in which we have not found a relevant operator are represented by ‘X’. Three‘clusters’ of languages with universal SAOs can be identified: (i) a group of West Euro-pean languages spoken on the British Isles and in France, (ii) a group of East Europeanlanguages located between the Baltic Sea and the Black Sea, and (iii) some Balkan lan-guages (plus/including Turkish). Moreover, Spanish has a universal SAO (aun), whilemost languages spoken on the Iberian Peninsula lack one (but recall from §4.5 thatBasque has a general additive operator ere).The areal distribution shown in Map 2 is at least partially a genetic one. Most lan-

guages of group (i) are Celtic and those of group (ii) (except Estonian) are Balto-Slavic.All Germanic languages except English lack universal SAOs. Among the nine Romancelanguages of our sample, only two have universal SAOs, French (même) and Spanish(aun).Map 2 provides some indication that universal SAOs are susceptible to contact-

induced processes of convergence such as ‘polysemy-copying’ (cf. Heine & Kuteva2005:100–103). In situations of language contact, expressions from different languages

2

MAP 2. Universal SAOs in European languages. MAP 2. Universal SAOs in European languages.

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 27

with overlapping (though not fully equivalent) functions are often ‘assimilated’ distrib-utionally (cf. also Gast & van der Auwera 2011). For example, if language L1 has a BE-YOND-operator B, and language L2 has a universal SAO U, the two expressions may be‘interlingually identified’ by bilingual speakers, and B may develop into a universalSAO by analogy to U. As Heine and Kuteva (2005) have shown, such instances of con-vergence are also often the result of parallel (‘contact-induced’) grammaticalization. Itseems to us that the relationship between the SAOs of Celtic languages and English, andperhaps also French, might be worth exploring from this point of view. Interesting in-sights into the role of language contact might also be gained by determining thechronology of the convergence processes apparently reflected in Estonian, Baltic, and(some) Slavic languages.Alternatively, we may consider to what extent ‘language-internal’ explanations for

the areal patterns shown in Map 2 are conceivable, that is, to what extent the presence orabsence of specific types of operators correlates with other linguistic features. Our datasuggest that the existence of a universal SAO correlates positively with the presence ofnegative concord in the relevant languages. This hypothesis is compatible with the di-achronic developments sketched in §9.1, where it is shown that universal SAOs oftendevelop from BEYOND-operators when the latter are used to reinforce scalar negators. Ifthis is right, negative concord provides a fertile ground (though neither a necessary nora sufficient condition) for the emergence of universal SAOs.

66.. BEYOND-OPERATORS IN THE LANGUAGES OF EUROPE. BEYOND-operators are found in avertical stretch of languages in the center of Europe. It ranges from Scandinavian lan-

28 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

guages (Swed. till och med, Norw. til og med, Dan. selv, Icel. jafnvel) and Finnish (jopa)via the Continental Germanic languages (Germ. sogar/selbst, Dutch zelfs) to Sursilvan(perfin) and Italian (perfino, persino; see Visconti 2004, 2005 for historical discussion).These BEYOND-operators are ‘basic’ ones; that is, they are fine in combination with astrong text proposition in the local domain (cf. the (a) examples), but they cannot occurin the scope of negation (cf. the (b) examples). Examples from Finnish are provided in71 (cf. also the Italian examples in 45 above).

(71) Finnisha. *Jopa hänen ystävänsä-kin moittivat häntä.*even PRON3.GEN friends-also criticize PRON3.PART‘Even his friends criticize him.’ (Hirvensalo 1963:1159)

b. *Ei-vät jopa äidin pyynnöt voineet häntä liikuttaa.*NEG-3SG even mother.GEN praying could her moveint.: ‘Not even the mother’s praying could move her.’ (M. Miestamo, p.c.)

Basic BEYOND-operators also exist in Spanish, which in addition to the universal SAOaun has hasta (lit. ‘as far as’) and incluso (‘included’; cf. Schwenter 1999b, 2000,2002). The latter element also exists in Galician. Catalan has a particle group fins (i tot),and Portuguese uses até as a BEYOND-operator, a cognate of Spanish hasta, as in 72.

(72) Portuguesea. *Mas quem é este que até os ventos e as ondas lhe obedecem?*but who is this PTCL even the winds and the waters him obey‘But who is this, who even the winds and the waters obey?’

(Matt. 8:25; O livro, from http://www.bibleserver.com)b. *Mas quem é este que nem até o seu cachorro o obedece?*but who is this that not even DET his dog him obeysint.: ‘But who is this? Not even his dog obeys him!’

(Cleomar Alves da Silva, p.c.)

BEYOND-operators are also relatively widespread among East and South East Euro-pean languages (e.g. Czech dokonce, Slovak dokonca, Hungarian még is, Slovene celo,Bosnian-Croatian-Serbian čak, Romanian chiar (şi)). As has been pointed out (cf. §4.3),many of these operators are ‘extended’ BEYOND-operators—that is, they may occur inthe scope of clausemate negation. Moreover, many of these elements are optionallyused in combination with a general additive operator, for example, Czech dokonce (i),Slovene celo (i), and Greek akomi (ke) (see 73).

(73) I Maria efaje akomi ke to pagoto.DET Maria ate even also the ice.cream‘Maria ate even the ice cream.’ (Giannakidou 2007:43)

The BEYOND-operators of the languages of our sample are surveyed in Map 3, whereno distinction is made between basic and extended BEYOND-operators. As a comparisonof Map 3 with Map 2 shows, the distribution of universal SAOs and BEYOND-operatorsis almost complementary. The only language of our sample that clearly has both a uni-versal SAO and a BEYOND-operator is Spanish (aun/universal SAO, hasta/BEYOND-operator). BEYOND-operators are found on the Iberian Pensinsula and in (longitudinal)central Europe, extending well into Eastern Europe and the Northern Balkans. They arewidespread in both Germanic and Romance languages (with English and French pro-viding the notorious exceptions). Moreover, they are found in those Slavic languagesthat lack universal SAOs.

3

MAP 3. BEYOND-operators in European languages.

MAP 4. Neutral BENEATH-operators in European languages.

MAP 3. BEYOND-operators in European languages.

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 29

The near complementarity of universal SAOs and BEYOND-operators can be explainedin historical terms. As is shown below in §9, universal SAOs typically derive from BE-YOND-operators. ‘Redundancies’ as in Spanish (which has both aun and hasta) mayarise if a BEYOND-operator is newly created after a former BEYOND-operator has beengeneralized to a universal SAO. Such developments seem to be rare, however, perhapsbecause ambiguities of the type discussed in §3.1 arise only under very specific cir-cumstances, so there is no functional pressure to differentiate BEYOND-readings from BENEATH-readings of a universal SAO and, hence, no need to create a new (unambigu-ous) BEYOND-operator.

77.. BENEATH-OPERATORS IN THE LANGUAGES OF EUROPE.77..11.. NEUTRAL BENEATH-OPERATORS. Just like BEYOND-operators, and to an even greater

extent, BENEATH-operators constitute a rather heterogeneous class of expressions, as faras their morphological make-up is concerned. We start with neutral BENEATH-operators.There are a few monomorphemic (nontransparent) BENEATH-operators in our sample,for example, Greek kan (< Ancient Gr. kai ‘and’ plus the (conditional) modal clitic an).Kan cannot be used in scale-preserving contexts (74a). It may either cooccur with oute‘not even’ (74b), or stand by itself in other types of scale-reversing contexts (74c; seeGiannakidou 2007 for more information).

(74) Greeka. *I Maria efaje kan to pagoto*DET Maria ate even DET ice.creamint.: ‘Maria ate even the ice cream.’ (K. Stathi, p.c.)

20 Like many other BENEATH-operators (e.g. Greek esto), Romanian măcar is also used in scale-preservingcontexts with the meaning ‘at least’.

21 See also Herburger 2003 on (ni) siquiera. Note, however, that in the variety of Spanish described by Her-burger, siquiera is always accompanied by the scalar negator ni, even in contexts other than clausemate nega-tion. Accordingly, our siquiera in many cases corresponds to Herburger’s ni siquiera.

30 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

b. *O Janis efije xoris na milisi oute kan me tin Maria.*DET Janis left without SUBJ talk.3SG not.even even with DET.ACC Maria‘John left without even talking to Maria.’ (Giannakidou 2007:54)

c. *Metaniosa pu aniksa kan to vivlio.*I.regret that I.opened even the book‘I regret that I so much as opened the book.’ (Giannakidou 2007:60)

A similar distribution is exhibited by the Finnish particle edes. It is not used in scale-preserving contexts (with the function of a BEYOND-operator), but it is found both in thescope of clausemate negation (75b) and in other scale-reversing contexts (75c).

(75) Finnisha. *Edes hänen ystävänsä-kin moittivat häntä.*even PRON3.GEN friends-also criticize PRON3.PARTint.: ‘Even his friends criticize him.’ (M. Miestamo, p.c.)

b. *Ei-vät edes äidin pyynnöt voineet häntä liikuttaa.*NEG-3SG even mother.GEN praying could her move‘Not even her mother’s praying could move her.’

(Suomi-Saksa-Suursanakirja, Katara & Schnellbach-Kopra 1997, s.v. edes)c. *Kun-han vain saan koskettaa edes hänen vaatteitaan, niin*when-surely just manage touch even his clothes so

tulen terveeksi.recover

‘If I so much as touch his clothes, I will be healed.’(Mark 5:28; Finnish Bible, from http://www.christianisrael.com/finnish/index.html)

Many Balkan languages have elements that have spread as a result of language contact,deriving either from Greek (< makári, orig. ‘blessed’; cf. Diez 1887) or from Persian (< meger, cf. Meyer 1891:255, Miklošich 1890:23). These words have ‘wandered’through the languages of the Balkans, and as far as Hungary and Italy. Bosnian-Croatian-Serbian uses the particle makar and Romanian măcar as a neutral BENEATH-operator.20

(76) Romaniana. Fiecare student care a citit măcar o carte a trecut examenul.every student who has read even a book has passed the.exam‘Every student who even/so much as read a book has passed the exam.’

(Geber 2005:5)

b. [Do you have dry socks?]Nu am nici măcar sosete umede.NEG I.have not.even even socks wet‘I don’t even have wet socks.’ (Geber 2005:6)

Words of the macar-family also exist in Albanian, Bulgarian, Macedonian, Bosnian-Croatian-Serbian, Slovene, Italian, and Greek, among other (minor) languages. Theytypically function either as conjunctions or interjections. In many cases, they are asso-ciated with concessivity and free-choice quantification. These meanings are also recov-erable in particles from Spanish, Portuguese, and Galician, that is, the cognates siquiera(Spanish),21 sequer (Portuguese), and sequera (Galician). These items originally meant

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 31

‘if wanted’. An example of Spanish siquiera is given in 77 (see also Lahiri 1998 on therelationship between SAOs and free-choice elements).

(77) SpanishPues da verguenza siquiera hablar de lo que ellos hacen en secreto. well it.gives shame even talk of what they do in secrecy‘For it is disgraceful even to mention what they do in secret.’

(Eph. 5:12; Biblia al día, from http://www.bibleserver.com)

While the operators mentioned above are morphologically (basically) opaque in thecontemporary languages, there are many BENEATH-operators in the languages of Europethat can be analyzed (almost) compositionally. Germ. auch nur simply combines the ad-ditive presupposition of auch (‘also’) with the ‘low’ scalar value of nur (‘only’; cf.Guerzoni 2003). More variety of operators can be found in Dutch, which has a combi-nation of particles parallel to German auch nur (ook maar ‘also only’), but also allowscombining the BEYOND-operator zelfs with maar (zelfs maar, lit. ‘even only’; cf. Van-deweghe 1981, Rullmann & Hoeksema 1997, Hoeksema & Rullmann 2001). Similarcombinations are used in Italian (anche solo/soltanto), Slovak (i len ‘also only’), andCzech (i jen ‘also only’; see 78).

(78) CzechO tom, co oni dìlají potají, je odporné i jen mluvit.PREP that REL they do secretly is disgraceful also only say‘It is disgraceful even to mention what they do in secret.’

(Eph. 5:12; Český ekumencký překlad, from http://www.bibleserver.com)

A slightly different strategy also involves a restrictive operator, but no indicator of anadditive inference: Spanish and Catalan combine ‘only’-elements with an (emotive) ad-verb of degree (tan ‘so’). An example of Catalan tan sols is given in 79 (the Spanishequivalent is tan solo or tan solamente ‘so only’).

(79) CatalanTot home del poble d’Israel que tan sols pensa en els seusevery man of.the people of.Israel REL so only thinks in DET.PL PRON.POSS

ídols …idols

‘Every Israelite who even thinks of his idols … ’(Ezek. 14:4; Biblia Catalana interconfesional, from http://www.biblija.net)

A third major type of neutral BENEATH-operator is instantiated by English so much asand Swedish så mycket som (cf. 80). This type of operator is also found in Danish andNorwegian/Bokmål (så meget som) as well as in Icelandic (svo mikið sem).

(80) Swedish… den som ger så mycket som bara ett glas friskt vatten till den minsthe REL gives so much as only one glass fresh water to the leastansedda avmina efterföljare … han ska få lön för det.respected of my disciples he will get reward for that

‘And if anyone gives even/so much as a cup of cold water to the least re-spected of my disciples … that person will certainly be rewarded.’

(Matt. 10:42; En levande bok, from http://www.bibleserver.com)

A similar etymology is reflected in Irish Gaelic, whose BENEATH-operator fiú—originallya noun meaning ‘worth’ (and optionally accompanied by the numeral amháin ‘one’, also‘only’)—establishes a comparison, just like Engl. so much as: ‘the (mere) worth of x’ isroughly equivalent to ‘so much as x’. Note that the nominal origin of this word is still re-flected in the fact that it sometimes requires a focus in the genitive case. It is, however,also used as a plain (non-case-assigning) particle.

3

MAP 3. BEYOND-operators in European languages.

MAP 4. Neutral BENEATH-operators in European languages. MAP 4. Neutral BENEATH-operators in European languages.

Map 4 lends support to the hypothesis that was made in §5 with respect to universalSAOs, namely that contact-induced processes of convergence have had an impact on thedistribution of SAOs in European languages. More than any other type of operator, BE-NEATH-operators seem to have resulted from calquing. For example, Dutch ook maar,Germ. auch nur, Czech i jen, and Ital. anche solo are made up of components with (ba-sically) identical literal meanings (‘also only’). The same applies to Engl. so much as andits Scandinavian counterparts. Determining the exact historical conditions under whichthese operators arose would be a challenging, but certainly worthwhile, undertaking.From a language-internal point of view, little can be said about the conditions under

which a given language is expected to have (or lack) a neutral BENEATH-operator. Neu-tral BENEATH-operators do not seem to correlate significantly with the presence or ab-sence of any other type of operator, or any type of grammatical feature. UnlikeBEYOND-operators, they are often found to coexist with universal SAOs. The reason forthis asymmetry might be that BENEATH-operators (unlike BEYOND-operators) do not tend

32 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

(81) Irish GaelicDá mbeadh fiú na leide again air.if EX even DET hint.GEN PREP.us PREP.him‘If we had even/so much as (the benefit of) a hint about him!’

(Brian Nolan, p.c.)

The major BENEATH-operators in the languages of our sample are surveyed in Map 4.The map shows that three major areas with similar etymologies can be distinguished: (i) a North Western area including Scandinavia and the British Isles, with BENEATH-operators based on expressions of comparison; (ii) a central European area, with BENEATH-operators based on restrictive particles; and (iii) a (South) Eastern area withelements from the macar-family. The Iberian peninsula is homogeneous insofar as allIbero-Romance languages represented in the sample have neutral BENEATH-operators.

22 It is of course possible to regard ‘only’ as a simple restrictive operator in these contexts, the interpretationof the entire proposition as ‘strong’ being dependent on other, for example, phonological and contextual, fac-tors. This points in the direction of an analysis in line with that of Krifka (1995), who regards reference to or-dered sets as an independent property of utterances (a specific type of focus). See also §3.2.

23 Ekkehard König (p.c.) has pointed out to us that uses of nur as in 82 are associated with a specific into-nation pattern. Both sentences are pronounced in a single tone group: that is, there is no main accent in thesecond tone group. This can be regarded as a prosodic indicator of scope relationships. See also König1991:48ff. on the relationship between phonological phrasing and scope.

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 33

to develop into universal SAOs and are historically rather stable. For example, Engl. somuch as is considerably older than even and has functioned as a BENEATH-operator formore than 500 years (cf. §9.2). The ‘diachronic stability’ of BENEATH-operators is alsoreflected in elements of the macar-family, which have retained their functions for sev-eral centuries under conditions of intensive language contact without ever showing in-dications of being generalized to a universal SAO.

77..22.. NONNEGATIVE BENEATH-OPERATORS. BENEATH-operators that are restricted to non-negative contexts seem to be rather rare, in Europe and probably also elsewhere in theworld. There seem to be two main reasons why BENEATH-operators are excluded underclausemate negation: (i) a possible, ‘undesired’ interaction between the operator and theclausal negator bars the intended scalar additive interpretation, and (ii) the operator pre-serves distributional restrictions that have resulted from its historical development. Cor-responding to these two reasons for why an operator may be restricted to nonnegativepolarity, we can distinguish two types of operators: (i) plain restrictive operators (e.g.Germ. nur), and (ii) operators based on concessive conditionals (e.g. Spanish aunquesea, probably also Greek esto, originally a third-person singular imperative form of thecopula einai).Let us start with the first type of nonnegative BENEATH-operator, that is, plain restric-

tive operators. Under specific circumstances, Germ. nur can function as a BENEATH-operator.22 A pertinent example is given in 82. Given that the BENEATH-reading emergesonly with heavy stress on the focus, stress is here indicated by small caps (cf. Van-deweghe 1981 for similar examples from Dutch).23

(82) Wenn du sie nur ANSIEHST, bekommst du Ärger.if you her only look.at get you trouble‘If you even look at her, you’ll get into trouble.’

Consider now what happens when the sentence is transformed into a main clause andnegated, so that scale reversal is preserved.

(83) Ich habe sie nicht nur ANGESEHEN.I have her not only looked.at‘I didn’t only look at her.’ (not: ‘I didn’t even look at her.’)

In 83, nur is not interpreted as an SAO, but as a restrictive particle like Engl. only. It isnot entirely clear to us why this happens; it may simply be due to the fact that the re-strictive use of nur is ‘prevalent’ or ‘unmarked’ vis-à-vis its use as an additive operatorin such contexts; that is, there is a ‘blocking effect’ and the ‘unmarked’ interpretation ischosen by default (see also Abrusán 2007 for discussion of Hung. akár under negation).The second type of nonnegative BENEATH-operator pointed out above can be illus-

trated with the Spanish operator aunque sea (lit. ‘even though it were’), which is used asa BENEATH-operator only in nonnegative contexts.

34 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

(84) a. *Si dices aunque sea una palabra, vas a tener problemas.*if you.say even.though.it.were a word you.will have problems‘If you say even one word, you’ll get into trouble.’

b. *No voy a decir aunque sea una palabra.*not I.will say even one wordint.: ‘I will not even say a single word.’ (Monica Zepeda, p.c.)

In all likelihood, this operator developed as a reinforcing (concessive) parenthetical inconditionals of the type illustrated in 85, as a result of deleting the first (pronominal)occurrence of the focus, here algo ‘something/anything’.

(85) Si dices algo, aunque sea una palabra, vas a tener if you.say some/anything although be.it one word you.will have

problemas.problems

‘If you say anything, even though it is (subjunctive) just a word, you’ll getinto trouble.’

The restriction to nonnegative polarity can thus be regarded as a reflex of the historicaldevelopment giving rise to aunque sea, which emerged in the context of (nonnegative)conditional clauses and seems to have preserved a distributional restriction to this effect.

77..33.. NEGATIVE BENEATH-OPERATORS AND SCALAR NEGATORS. In §4.4, we distinguishedthe class of ‘scalar negators’ from that of negative BENEATH-operators. In the languagesof our sample, these expressions are complementarily distributed (but note that thereare languages that have neither type of expression, such as English and French). TheGermanic languages (except English) have negative BENEATH-operators, all of which arederived from the same etymon, a numeral ‘one’, often plus an event quantifier such asGerm. mal ‘times’. Such operators are attested in Dutch (niet eens ‘not once’), Swedish(inte ens ‘not one’), Icelandic (ekki einu sinni ‘not one time’), Danish and Norwegian(ikke en-gang ‘not one-time’). Examples from Dutch and Icelandic are given in 86 and87, respectively.

(86) DutchEr is niet eens drinkwater.EXPL is not once drinkwater‘There is not even water to drink.’

(Num. 20:5; Het boek, from http://www.bibleserver.com)(87) Icelandic

Enginn gat lengur bundið hann, ekki einu sinni með hlekkjum.no.one was.able longer bound him not one time with chains‘No one could bind him, not even with a chain.’

(Mark 5:3; Icelandic Bible, from http://www.biblegateway.com)

Scalar negators are basically found in Southern and Eastern Europe, and they are re-stricted to languages with negative concord. Relevant operators can be found amongBalkan and some (non-Balkan) Slavic languages, for example, Albanian (as), Bosnian-Croatian-Serbian (ni(ti)), Czech (ani), Greek (oute), Macedonian (nitu), Polish (ani),Romanian (nici), and Slovene (niti). As is to be expected from their semantics, scalarnegators can typically be combined with (neutral) BENEATH-operators. For example,Spanish ni often cooccurs with siquiera, Greek oute with kan, and Albanian as withédhe (88).

4

MAP 5. Negative BENEATH-operators and scalar negators in European languages. MAP 5. Negative BENEATH-operators and scalar negators in European languages.

24 This word derives from ne pure ‘not also’; pure < Lat. pūre ‘purely, simply’; pure is also used as a non-scalar additive operator in contemporary Italian.

25 Nemmeno derives from Lat. ne minus ‘NEG less’.

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 35

(88) AlbanianE megjithëatë as edhe një prej tyre nuk harrohetCONJ nonetheless not.even even one PREP PRON.POSS.PL not forgets

përpara Perëndisë.PREP God

‘But God doesn’t forget even one of them.’(Luke 12:6; Albanian Bible, from http://www.biblegateway.com)

Scalar negators are mostly simple (monomorphemic) expressions and considerablyold, that is, etymologically nontransparent. An exception is provided by Italian, whichhas two scalar negators with a (more or less) transparent morphology, that is, neppure24(cf. Mari & Tovena 2006) and nemmeno.25 Examples are given in 89.

(89) Italian (Letizia Vezzosi, p.c.)

a. Non ha mangiato neppure il dolce.NEG has eaten not.even DET dessert‘She did not even eat the dessert.’

b. Non mi ha nemmeno salutato.NEG me has not.even greeted‘S/he didn’t even greet me.’

The negative BENEATH-operators and scalar negators in the languages of our sampleare summarized in Map 5. By definition, the former class of expressions is only found

in languages without negative concord, while the latter class is restricted to languageswith negative concord. A genetic pattern is also discernible, however, insofar as nega-tive BENEATH-operators are attested only in Germanic languages. Given that all of therelevant operators derive from the same etymon (a numeral meaning ‘one’), it seemslikely that negative BENEATH-operators represent a common Germanic heritage.

88.. REPRESENTING SYSTEMS OF SCALAR ADDITIVE OPERATORS. So far, our typology hasbeen represented as a taxonomy of contexts in which the various types of operators(with different degrees of specificity) can be used (cf. Fig. 3 in §4.6). In this section wepropose a way of representing entire systems of SAOs. For this purpose, we transformthe tree version of our typology into a tabulaic form of representation. This is done inTable 1 (‘NSAO’ stands for ‘nonscalar additive operator’).

36 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

GENERAL ADDITIVE OPERATOR

NSAO UNIVERSAL SAO

BEYOND neutral BENEATH

negative nonnegative

TABLE 1. Model for systems of additive operators.

GENERAL ADDITIVE OPERATOR

NSAO UNIVERSAL SAO

BEYOND BENEATH

negative nonnegative

also even

so much as

TABLE 2. The system of English additive operators.

Table 1 is modeled in such a way that neighboring fields are distributionally more sim-ilar than nonadjacent ones. This, in turn, means that if a given expression is used tocover two nonadjacent functions, it will also be used to cover the functions located inbetween. Table 1 thus provides a model of attested and unattested form-to-functionmappings in the languages of Europe. Beyond that, it can be considered a hypothesisabout possible and impossible form-to-function mappings in general. In other words,Table 1 can be regarded as an equivalent of a ‘semantic map’ (cf. Haspelmath 1997:226specifically on SAOs; see van der Auwera & Plungian 1998 and van der Auwera &Temürcü 2006 on semantic maps in general). We consider some examples of Europeanlanguages in order to illustrate how Table 1 is to be interpreted in §8.1 before venturingan outlook at two non-European languages in §8.2.

88..11.. SOME EUROPEAN LANGUAGES. The English system of additive expressions isshown in Table 2. The range of each operator is indicated by a shaded box covering theset of contexts in which the operator in question can be used. There is a nonscalar addi-tive operator also, a universal SAO even, and a neutral BENEATH-operator so much as.Similar systems are found in other West European languages, for example, in French,Irish Gaelic, Welsh, Manx, and Breton.

Spanish differs from English in having BEYOND-operators (hasta and incluso), ascalar negator (ni), and a nonnegative BENEATH-operator (aunque sea). Its comprehen-sive system of additive operators is summarized in Table 3.

The system characteristic of Germanic languages can be represented as shown inTable 4, using the example of German.

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 37

GENERAL ADDITIVE OPERATOR

NSAO UNIVERSAL SAO

BEYOND BENEATH

negative nonnegative

tambien aun

hasta/incluso siquiera

ni aunque sea

TABLE 3. The system of Spanish additive operators.

GENERAL ADDITIVE OPERATOR

NSAO UNIVERSAL SAO

BEYOND BENEATH

negative nonnegative

auch sogar/selbst auch nur

einmal

TABLE 4. The system of German additive operators.

GENERAL ADDITIVE OPERATOR

NSAO UNIVERSAL SAO

BEYOND BENEATH

negative nonnegative

RUSSIAN

i

tože daže (i)

ni xot’

CZECH

i

také dokonce

i jen

ani

TABLE 5. The systems of Russian and Czech.

In Slavic languages, there are basically two types of systems: those with a universalSAO (e.g. Russian, Polish, Bulgarian, Macedonian), and those without one (e.g. Czech,Slovak, Slovene, Bosnian-Croatian-Serbian). Most of the BEYOND-operators found inSlavic languages seem to be extended ones. The Russian and Czech systems are dis-played in Table 5. The Czech system illustrates how our typology deals with extendedBEYOND-operators: they simply cover the domain of both (basic) BEYOND-operators andnegative BENEATH-operators.

GENERAL ADDITIVE OPERATOR

NSAO UNIVERSAL SAO

BEYOND BENEATH

negative nonnegative

súní luxh nikla -tis

TABLE 6. The system of Mitla Zapotec additive operators.

GENERAL ADDITIVE OPERATOR

NSAO UNIVERSAL SAO

BEYOND BENEATH

negative nonnegative

-mo dake-demo

-demo

TABLE 7. The system of Japanese additive operators.

(90) BEYOND-operator: luxhChu-cha dee ni rnibee beh con nis luxh rusoobreni xtiidxni?who-Q DEM REL give.orders wind and water even obey.PL his.word‘Who is this? He orders the wind and the water, and they even obey him?’

(Matt: 8:27; Xdidxcoob dios ni biädnä dad Jesucrist)(91) negative BENEATH-operator/scalar negator: nikla

Luxh rut rahk-di rukuaduuni, nikla kon kaden-giib.and.then no.one could-NEG bind not.even with chain-iron‘And no one could bind him, not even with a chain.’

(Mark 5:3; Xdidxcoob dios ni biädnä dad Jesucrist)(92) nonnegative BENEATH-operator: -tis

Palga sahk kanä-tis roʔ xhahbni, siakä.if can I.tempt-just edge his.clothes I.will.be.healed‘If I even touch his clothes, I will be healed.’

(Matt. 9:21; Xdidxcoob dios ni biädnä dad Jesucrist)

According to Nakanishi (2006), Japanese uses one operator for the range of mean-ings from NSAO to negative BENEATH-operator (-mo), an extended BEYOND-operator (-demo), and a nonnegative BENEATH-operator (-dake-demo). The system is displayed inTable 7 and illustrated in 93–95.

38 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

The systems of additive operators in the forty languages of our sample are summa-rized in the appendix, using the format of the tables given in this section. Altogether,145 additive operators have been classified along the dimensions distinguished in this study.

88..22.. SCALAR ADDITIVE OPERATORS IN NON-EUROPEAN LANGUAGES. While a comprehen-sive discussion of non-European languages is beyond the scope of this article, we take abrief look at two relevant languages: the Mesoamerican language Zapotec (as spoken inMitla, cf. Stubblefield & Miller 1991), and Japanese (as described by Nakanishi 2006).Zapotec is interesting because it represents a maximally differentiating system, withone expression for each cell. Its system of SAOs is shown in Table 6, and relevant ex-amples (from the Bible) are given in 90–92.

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 39

(93) BEYOND-contexts: -(de)moSaru-(de)mo ki-kara otiru.monkey-even tree from falls‘Even a monkey falls from a tree.’ (Nakanishi 2006:141)

(94) nonnegative BENEATH-contexts: -dake-demoJohn-ga hon-o is-satu-dake-demo yonda-to-wa odoroita.John-NOM book-ACC one-CLSF-only-even read-COMP-top was.surprised‘I was surprised that John read even a single book.’ (Nakanishi 2006:147)

(95) negative BENEATH-contexts: -(de)mo, *-dake-demoa. *John-wa hon A-(de)mo yom-ana-katta.*John-TOP book A-even read-NEG-PST‘John did not even read book A.’

b. *John-wa hon A-dake-demo yom-ana-katta.*John-TOP book A-only-even read-PST-PSVint.: ‘John did not even read book A.’ (Nakanishi 2006:142)

Obviously, the application of our typology to non-European languages requires moredetailed investigations, and it is an open question whether our system will stand up tocloser scrutiny. It is of course to be expected that the analysis of more relevant data willlead to revisions and extensions.

99.. THE EMERGENCE OF UNIVERSAL SAOS. The make-up of our typology as representedin Table 1 may be somewhat surprising, as negative BENEATH-operators appear to bemore similar to BEYOND-operators than nonnegative ones. One might have expectedSAOs to be distributed along the dimension ‘veridical – nonveridical – averidical’, interms of Zwarts 1995, 1998. In fact, this type of ‘semantic map’ was also proposed byHaspelmath (1997:226). This is not, however, what we have found in the languages of our sample. There seems to be no operator that is used as a nonnegative BENEATH-operator and as a BEYOND-operator, but not as a negative BENEATH-operator. As we arguein this section, this is due to typical historical developments in the domain of SAOs.Basic BEYOND-operators often develop into extended ones, which in turn may be gener-alized to universal SAOs. The make-up of our typology as represented in Table 1 is thusa consequence of possible (or at least likely) historical processes of change (for the roleof diachrony in the semantic map methodology, see van der Auwera 2008:43: ‘[t]he bestsemantic map is a diachronic one’).Function words like SAOs are notoriously underrepresented in traditional diachronic

work, for example, in Indo-European studies (but see Visconti 2004, 2005 on (Old) Ital-ian). We have therefore focused our attention on some ‘major’ languages, but even forthose languages the evidence is sparse. The tentativeness of the hypotheses made in thissection should thus be obvious. We have found evidence for one major type of develop-ment giving rise to universal SAOs, that is, the ‘reinforcement’ of a negative BENEATH-operator or scalar negator, followed by semantic generalization. This development isillustrated for Slavic languages in §9.1 and for English in §9.2. Some theoretical impli-cations of the observations made in §§9.1 and 9.2 are considered in §9.3.

99..11.. REINFORCEMENT OF A SCALAR NEGATOR IN SLAVIC LANGUAGES. The term ‘rein-forcement’ is well known from research into grammaticalization (cf. Lehmann 1982). Itrefers to the addition of a lexical item to a function word that has lost phonological sub-stance and/or semantic specificity. The most famous case of reinforcement is probablythe addition of the negative polarity item pas (originally a noun meaning ‘step’) to thesimple clause negator ne (< Lat. non), that is, the development of the complex negator

[AQ4]

40 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

ne … pas in French (‘Jespersen’s cycle’; cf. van der Auwera 2009 and references citedthere). A similar process seems to have taken place in many of those languages that haveuniversal SAOs: the scalar negator has been reinforced by a BEYOND-operator. The re-sulting construction leads to a certain ‘indeterminacy’ of scope relations between nega-tion and the SAO in question, thus providing a context where a basic BEYOND-operatorcan develop into an extended one.In §4.3, we used Czech dokonce to demonstrate the distributional properties of ex-

tended BEYOND-operators. We now use this operator to illustrate how the process ofchange from a basic BEYOND-operator to an extended one may have been motivated.Most likely, dokonce started out as a basic BEYOND-operator—that is, it was not usedunder negation. One way of expressing scalar negation in Czech is to use the scalarnegator ani in addition to the clause negator ne- (in ne-ní), as in 96.

(96) CzechNení tady ani voda k napití.NEG.EX there not.even water PREP drink‘There was not even water to drink.’ (V. Elšik, p.c.)

If a speaker wants to add more emphasis, the double negation structure in 96 can be re-inforced by adding dokonce. In other words, ani can be ‘supported by’ a phonologicallymore weighty and semantically more specific element. As was shown in §4.3, dokoncecan occur either in a clause-initial position (as in 97a (= 55a)), or in a position followingthe negated main predicate není (cf. 97b (= 55b)).

(97) a. Dokonce tady není ani [voda]F k napití.even there NEG.EX not.even water PREP drink‘There is not even water to drink.’

b. Není tady dokonce ani [voda]F k napití.NEG.EX there even not.even water PREP drink‘There is not even water to drink.’ (V. Elšik, p.c.)

A structural configuration of the type illustrated in 97 provides a context where a BEYOND-operator can ‘extend’ its distribution, because the scope relations between thenegators ne- and ani on the one hand, and the SAO dokonce on the other, are not entirelyclear. In 97b, dokonce appears to take scope over the internal negator ani, but it is (ex-pected to be) in the scope of clause negation (lit. ‘There was not even no water todrink’). One way of interpreting the scope relations is to regard the local domain asweak ([[There was water]]) and the minimal clause as strong ([[There was no water]]).This is the type of context that characterizes extended BEYOND-operators (cf. §4.3).Extended BEYOND-operators are susceptible to further reanalysis, which eventually

leads to the development of a universal SAO. This is an instance of distributional gener-alization from negative contexts to scale-reversing ones. Some operators from Slaviclanguages seem to have gone this way, for example, Russian and Bulgarian daže. Ac-cording to the relevant etymological dictionaries (e.g. Miklošich 1886, Berneker 1913),da started out as a demonstrative meaning ‘so’ (< Proto IE *do), which first developedinto a consecutive subordinating conjunction (‘so that’), and later into a temporal one(‘until’). At some point, da seems to have merged with the locative preposition do (‘un-til’), which existed independently. Dawas combined with the originally adversative par-ticle že, thus giving rise to (Russian and Bulgarian) daže. Bulgarian dori basically hasthe same history, but has undergone more drastic phonological changes, according toMiklošich (1886) (do-že > doži > dori).

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 41

In Old Church Slavonic, daže (mostly used together with do) was basically a tempo-ral proposition meaning ‘until’ (‘temporal endpoint’), but some dictionaries also pro-vide the meaning ‘even’ (e.g. Stojanov & Janakiev 1983). There seems to be consensusthat the ‘temporal endpoint’ meaning is the older one, and developments of the type‘temporal endpoint’ > SAO are actually widely attested in the languages of Europe (e.g.Spanish aun (< Lat. adhuc ‘until’) and hasta ‘until, even’, Swedish till och med ‘till andwith’, i.e. ‘up to and including’). In Old Church Slavonic texts, daže apparently did not occur in the scope of negation, and the scalar negator ni was used as a negative BENEATH-operator. In other words, in Old Church Slavonic daže seems to have had thedistribution of a basic BEYOND-operator.In the contemporary Slavic languages of our sample that have an item deriving from

OCS daže, the relevant operators function as universal SAOs (though according to Pol-jakova 2001, Russian daže obligatorily cooccurs with i in specific contexts; see also n.16). The development from the basic BEYOND-operator of Old Church Slavonic to theuniversal SAOs of Russian and Bulgarian can be reconstructed on the basis of data fromthe contemporary languages (we do not have solid historical evidence, though). Let usconsider Russian. As in Czech (cf. 96), the scalar negator (ni) can (under specific cir-cumstances) cooccur with a nonscalar clause negator (ne). This is illustrated in 98 (zgi( < IE *ĝigā ‘jot, fragment’) is a negative polarity item that is semantically opaque).

(98) Ne vidno ni zgi.not visible not.even NPI‘It’s pitch-black.’ (Polkjakova 2001:454)

For more emphasis, ni can be reinforced with daže. When the focus is an object, theoperator-focus constituent may either precede (99) or follow the verb (100). This situa-tion is reminiscent of the indeterminacy of scope relations observed in the Czech pair ofexamples given in 96 above.

(99) Ja ej daže ni odnogo pis’ma ne napisal.I to.her even not.even one.GEN letter.GEN NEG wrote‘I did not even write her a single letter.’ (Russian National Corpus)

(100) Maša ne rešila daže ni odnoj zadači.Masha NEG solved even not.even one problem‘Masha did not even solve one problem.’ (Poljakova 2000)

Unlike Czech dokonce, however, Russian daže can also be used (and in fact com-monly is used) in combination with a general additive operator (i) instead of the scalarnegator (ni). This is shown in 101.

(101) Maša ne rešila daže i odnoj zadači.Masha NEG solved even and/even one problem‘Masha did not even solve one problem.’ (Poljakova 2000)

When an SAO occurs in contexts of the type illustrated in 101, it can (presumably) begeneralized further to other types of scale-reversing contexts like, for instance, the com-plement clause of scale-reversing predicates, as in 102.

(102) Stydno daže slušat’.shame even listen‘It’s a shame to even listen.’ (Russian National Corpus)

If the development sketched above is correct, Russian daže started out as a basic BEYOND-operator, extended its distribution in the context of negation (like Czechdokonce), and was finally generalized to a universal SAO (unlike Czech dokonce).More detailed diachronic analyses are required to see whether this scenario is actually

42 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

supported by historical data. As is seen in the next section, a largely parallel develop-ment can be reconstructed for Engl. even.99..22.. REINFORCEMENT OF A BENEATH-OPERATOR IN ENGLISH. The use of even as an SAO

is not attested until Early Modern English. The earliest example provided in the detailedhistorical study by Traugott (2006) dates from 1586 (see 103; note that merely is hereused as a manner adverb meaning ‘without qualification’).

(103) caused me, nether arrogantly nor contemptuouslye, but even merely andfaythfully, to doe hir majesty the best servyce.

(Robert Dudley, l. 3640; 1586; quoted from Traugott 2006:350)

The first instance recorded in the Oxford English Dictionary (OED) is from Shake-speare (see 104). Scalar occurrences of even from around 1600 are rare, however. In thegreat majority of cases, adverbial occurrences of even function as ‘particularizers’, as in 105 (see König 1989, 1991, Traugott 2006 for the development from particularizers to SAOs).

(104) Make sacred even his styrrop. (Shakespeare, Timon I, 2, 82; 1607)(105) Aeneas: ‘Is this great Agamemnon’s tent, I pray you?’

Agamemnon: ‘Even this.’ (Troilus and Criseide I, 3; 1602)

In texts from the early seventeenth century, even still functions as a basic BEYOND-operator (when used as an SAO) and is not used in the scope of negation. In the KingJames Bible (1611), emphatic negation is often expressed with the combination no not,as in 106 (alternatively with neither, e.g. in Mark 14:59).

(106) Curse not the king, no not in thy thought. (Ecclesiastes 10:20; 1611)

The first dictionary that mentions even in the scope of negation is Ainsworth 1736 (cf.OED, s.v. even, 9), where Lat. ne quidem is translated as no, not even. This is also theform commonly found in John Locke’s works, that is, in the late seventeenth century(107).

(107) a. Expansion and duration have this further agreement, that, though theyare both considered by us as having parts, yet their parts are not separa-ble one from another, no not even in thought.

(J. Locke, An essay concerning human understanding, XV, 10; 1690)b. And therefore the magistrate cannot take away these worldly things from

this man or party and give them to that; nor change propriety amongstfellow subjects (no not even by a law), for a cause that has no relation tothe end of civil government, …

(J. Locke, A letter concerning toleration; 1689)

In many cases even is also found accompanying (not) so much as (cf. 108), which is at-tested as a (negative) BENEATH-operator as early as the fifteenth century (see OED, s.v.so, 40b), predating (scalar) even by almost 200 years.

(108) All which abuses, if those acute philosophers did not promote, yet theywere never able to overcome; nay, even not so much as King Oberon and hisinvisible army. (Thomas Sprat, The history of the royal society; 1667)

The data from English suggest that even extended its distribution in the context ofother scalar operators, specifically by reinforcing a negative BENEATH-operator (eitherso much as or no not). Typically, the English examples from the seventeenth century in-volve a parenthetical or otherwise extraclausal, and often elliptical, structure, as in 107and 108. It was probably in these contexts that scope relations were blurred and couldthus be reanalyzed, with the result that even came to be licensed in the scope of nega-tion. In Locke’s works, there are also singular instances where even is clearly containedin the scope of negation, as in 109.

Table 8 illustrates why our typology looks the way it does. While there is a process of his-torical change leading from BEYOND to negative BENEATH, a direct change from BEYONDto nonnegative BENEATH (skipping ‘negative BENEATH’) seems to be unattested.Given that the change from basic to extended BEYOND-operator is (basically) a change

in the scope behavior of the operator, the question arises of whether our typology can be reformulated in terms of a scope-theoretic analysis, and in fact this seems to be pos-sible. Let us assume (like the scope-theorists) that the lexical meaning of SAOs is in-variant, and that the various types of operators differ in their scope properties. Threelevels of scope can be distinguished: (i) narrow scope, which is in accordance with thegeneral mechanisms of scope mapping; that is, the scope of the operator is restricted tothe local domain; (ii) extended scope, where the scope of the SAO is restricted to theminimal clause; that is, the operator may take scope over clausemate negation; and (iii)wide scope, where the operator takes scope over some scale-reversing operator from ahigher clause.The types of operators distinguished in §4 can now be characterized in terms of their

scope behavior as follows: basic BEYOND-operators invariably take narrow scope; ex-tended BEYOND-operators take either narrow or extended scope; the scope of negativeBENEATH-operators is obligatorily extended, though restricted to the minimal clause;neutral BENEATH-operators take either extended or wide scope; and finally, nonnegativeBENEATH-operators obligatorily take wide scope. This is illustrated in Table 9 (nonscalaradditive operators are disregarded, since they cannot be analyzed as differing from theother operators only with respect to scope).

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 43

(109) Not even Americans, subjected unto a Christian prince, are to be punishedeither in body or goods for not embracing our faith and worship.

(J. Locke, A letter concerning toleration; 1689)

Later, even was generalized further to other types of scale-reversing contexts, thus ac-quiring the distribution that it exhibits in the contemporary language. It is hard to tellwhen exactly even came to be used in nonnegative contexts, but our data suggest thatthis happened considerably later than the distributional extension to negative contextsobservable in texts from the seventeenth century like those referred to above. Given thatthe OED (or any other comparable reference work) does not provide any reliable evi-dence in this respect, only a comprehensive diachronic corpus study could elucidate thisdevelopment.

99..33.. SOME IMPLICATIONS AND A SCOPE-THEORETIC TYPOLOGY OF SAOS. In both casesconsidered in §§9.1 and 9.2, the development of a universal SAO seems to have pro-ceeded along the path shown in 110.

(110) reinforcement of scalar negator distributional generalizationBEYOND-operator → extended BEYOND-operator → universal SAO

Table 8 shows the development in our tabulaic format used in §8. The pointed arrowindicates scope reanalysis (and light gray shading the distribution of an extended BEYOND-operator), and the solid arrow stands for semantic generalization (and dark grayshading shows the additional sets of contexts covered by a universal SAO).

UNIVERSAL SAO

BEYOND BENEATH

negative nonnegative

TABLE 8. Combining the synchronic and the diachronic perspective.

44 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

The range of contexts covered by any given operator can be illustrated just as in Ta-bles 1–7 above. This is shown for Spanish and Czech in Table 10 (the scalar negators niand ani are disregarded; due to their negative feature, they cannot be analyzed as lexicalvariants of other SAOs with different scope properties).

NARROW SCOPE EXTENDED SCOPE WIDE SCOPE

BEYOND/basic � * *BEYOND/extended � � *BENEATH/negative * � *BENEATH/neutral * � �BENEATH/nonnegative * * �

TABLE 9. A typology of scalar additive operators in terms of the scope theory.

While a purely scope-theoretic typology of SAOs is thus possible, it seems to us thatentirely abstracting away from variation in lexical content oversimplifies matters in somerespects. As we have tried to show, there is a difference between (i) the development of a basic BEYOND-operator into an extended one, and (ii) the change from an extended BEYOND-operator to a universal SAO. While the former process has been regarded as aninstance of scope reanalysis, that is, a matter concerning the syntax-semantics interface,the latter process has been viewed as an instance of distributional generalization, that is,basically a lexical-semantic one. If this is right, the nature of the change from a basic toan extended BEYOND-operator differs substantially from the development of an (ex-tended) BEYOND-operator into a universal SAO. Accordingly, an adequate theory of SAOsshould make reference to both lexical content and scope. One way of combining a lexi-calist with a scope-theoretic view is to regard the difference between BEYOND-operatorsand BENEATH-operators as a lexical one, and the distinction between basic and extendedBEYOND-operators as resulting from differences in scope relations. This is actually the so-lution that is reflected in our choice of terminology. Other options are conceivable as well,however, and figuring out the ‘division of labor’ between lexical semantics and sentencesemantics (coupled with syntax) certainly remains a major challenge in both language-particular and crosslinguistic studies of SAOs.

1100.. CONCLUSIONS. The present article has pursued four major goals. First, we haveaimed at providing a framework for the semantic analysis of scalar additive operators.Second, we have established a typology that is intended to be general enough to cover aconsiderable amount of crosslinguistic variation without becoming too general, that is,

NARROW SCOPE EXTENDED SCOPE WIDE SCOPE

SPANISH

hasta

aun

siquiera

aunque sea

CZECH

i

dokonce

i jen

TABLE 10. Spanish and Czech in terms of the scope theory.

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 45

meaningless. Building on comparative work like König 1991 and Giannakidou 2007,we have proposed to classify scalar additive operators according to the ‘relative prag-matic strength’ of the proposition contained in the ‘local (scope) domain’ (BEYOND-operators vs. BENEATH-operators). BENEATH-operators have moreover been subclassifiedinto those that are restricted to nonnegative contexts and those that can occur only underclausemate negation. This basic typology gives rise to five major types, as the interme-diate nodes also constitute operators in their own right. Moreover, we have taken threeadditional types of scalar expressions into account: that is, extended BEYOND-operators,general additive operators, and scalar negators.The third objective of this article has been a descriptive one: we have surveyed the

distribution of different types of SAOs in a sample of forty European languages. More-over, we have proposed a model for the representation of systems of scalar additive op-erators, and have used this model to describe the distribution of 145 operators from theforty languages of our sample (see the appendix).Our fourth aim was to interpret the patterns of polysemy found in a historical light,

and to consider the implications of the historical findings for a synchronic analysis ofscalar additive operators. We have argued that a historically adequate typology of SAOsshould take both lexical content and scope into account, though the exact division oflabor remains to be determined in more detailed studies, both language-particular andcrosslinguistic.Finally, it has also been an important goal of this article to instigate more compara-

tive investigations of scalar additive operators, including historical ones. So far, thistopic has for the most part been dealt with by semanticists, often with a focus on ratherspecific questions. By presenting a simple but not simplistic typology, we hope to con-tribute to a wider interest in the domain of scalar additive operators, especially amongnonsemanticists, and to provide a frame of reference for future studies.

Appendix: Systems of additive operators in forty european languages

GENERAL ADDITIVE OPERATOR

NSAO UNIVERSAL SAO

BEYOND BENEATH

negative nonnegative

ALBANIAN

edhé (madjé) édhe

as

BASQUE

ere

BELARUSAN

i

taksáma navat

ni

BOSNIAN

i

takodje čak makar

niti

BRETON

ivez zokén

46 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

GENERAL ADDITIVE OPERATOR

CATALAN

també fins i tot tan sols

ni

BULGARIAN

i

sãšto daže/dori

ni

CROATIAN

i

takodje čak makar

niti

CZECH

i

také dokonce

i jen

ani

DANISH

ogsaa selv så meget som

engang

DUTCH

ook zelfs ook/zelfs maar

eens

ENGLISH

also even

so much as

ESTONIAN

-ki/-gi

isegi

FINNISH

-kin jopa edes

FRENCH

aussi même

ne fût-ce que

GALICIAN

tamén incluso sequera

nin

GERMAN

auch auch nur

sogar einmal

GREEK

ke

akomi ke kan

oute esto

HUNGARIAN

is még is még sem akár

ICELANDIC

einnig jafnvel svo mikið sem

einnu

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 47

GENERAL ADDITIVE OPERATOR

IRISH

freisin féin

fiú (amháin)

ITALIAN

anche perfino anche solo

nemmeno

LITHUANIAN

taip pat net

LATVIAN

ari pat

MACEDONIAN

i

duri

ni

MANX

myrgeddin hene

NORWEGIAN

ogsaa til og med så meget som

engang

POLISH

tezh nawet

ani

PORTUGUESE

tambem até sequer

nem

ROMANIAN

şi chiar (şi) măcar

nici

RUSSIAN

i

tože daže

ni xot’

SERBIAN

i

takodje čak makar

niti

SLOVAK

i

aj dokonca

i len

ani

SLOVENE

i

tudi celo niti

SPANISH

tambien aun

hasta/incluso siquiera

ni aunque sea

48 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

GENERAL ADDITIVE OPERATOR

SURSILVAN

era perfin sulettamain

gnanc

SWEDISH

också till och med så mycket som

ens

TURKISH

de bile

UKRAINIAN

i

takozh navit’

ani

WELSH

hefyd hyd yn oed

REFERENCES

ABRUSÁN, MARTA. 2007. Even and free choice any in Hungarian. Proceedings of Sinn undBedeutung 11, ed. by Louise McNally and Estela Puig-Waldmüller, 1–15. Barcelona:Universitat Pompeu Fabra.

AINSWORTH, ROBERT. 1736. Dictionary, English and Latin. London: Rivington & Woodfall.ALTMANN, HANS. 1976. Die Gradpartikeln im Deutschen: Untersuchungen zu ihrer Syntax,

Semantik und Pragmatik. Tübingen: Niemeyer.ANDERSON, STEPHEN R. 1972. How to get even. Language 48.893–906.ANSCOMBRE, JEAN-CLAUDE, and OSWALD DUCROT. 1983. L’argumentation dans la langue.

Liège: Mardaga.BARKER, STEPHEN. 1991. Even, still and counterfactuals. Linguistics and Philosophy 14.1–38.BARWISE, JON, and ROBIN COOPER. 1981. Generalized quantifieres and natural language. Lin-

guistics and Philosophy 5.403–18.BAYER, JOSEF. 1996. Directionality and logical form: On the scope of focusing particles and

wh-in-situ. Dordrecht: Kluwer.BEAVER, DAVID I., and BRADY Z. CLARK. 2008. Sense and sensitivity: How focus determines

meaning. Oxford: Wiley-Blackwell.BENNETT, JONATHAN. 1982. Even if. Linguistics and Philosophy 5.403–18.BERCKMANS, PAUL. 1993. The quantifier theory of even. Linguistics and Philosophy

16.589–611.BERNEKER, ERICH. 1913. Slawisches Etymologisches Wörterbuch. Heidelberg: Winter.BUCHHOLZ, ODA, and WILFRIED FIEDLER. 1987. Albanische Grammatik. Leipzig: Verlag En-

zyklopädie.BUCHHOLZ, ODA; WILFRIED FIEDLER; and GERDA UHLISCH. 1992. Wörterbuch Alban-

isch–Deutsch. Leipzig: Langenscheidt.BÜRING, DANIEL. 2003. On D-trees, beans, and B-accents. Linguistics and Philosophy

26.511–45.CARLSON, LAURI. 1983. Dialogue games: An approach to discourse analysis. Dordrecht:

Reidel.CHAFE, WALLACE. 1974. Language and consciousness. Language 50.111–33.CHAFE, WALLACE. 1976. Givenness, contrastiveness, definiteness, subjects, topics and point

of view. Subject and topic, ed. by Charles N. Li, 25–55. New York: Academic Press.CORBLIN, FRANCIS; VIVIANE DÉPREZ; HENRIËTTE DE SWART; and LUCIA TOVENA. 2004. Nega-

tive concord. Handbook of French semantics, ed. by Francis Corblin and Henriette deSwart, 417–52. Stanford, CA: CSLI Publications.

DIEZ, FRIEDRICH. 1887. Etymologisches Wörterbuch der romanischen Sprachen. Bonn: Marcus.

DUCROT, OSWALD. 1980. Les échelles argumentatives. Paris: Minuit.FAUCONNIER, GILLES. 1975a. Polarity and the scalar principle. Chicago Linguistic Society

11.188–99.

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 49

FAUCONNIER, GILLES. 1975b. Pragmatic scales and logical structure. Linguistic Inquiry6.353–75.

FAUCONNIER, GILLES. 1978. Implication reversal in a natural language. Formal semanticsand pragmatics for natural languages, ed. by F. Guenthner and Siegfried J. Schmidt,289–301. Dordrecht: Reidel.

FILLMORE, CHARLES. 1965. Entailment rules in a semantic theory. POLA report 10.60–82.Columbus: The Ohio State University.

FRANCESCOTTI, ROBERT M. 1995. Even: The conventional implicature approach reconsid-ered. Linguistics and Philosophy 18.153–73.

FRASER, BRUCE. 1971. An analysis of even in English. Studies in linguistic semantics, ed. byCharles J. Fillmore and D. Terence Langendoen, 141–80. New York: Holt, Rinehart andWinston.

FREGE, GOTTLOB. 1918. Der Gedanke: Eine logische Untersuchung. Beiträge zur Philoso-phie des deutschen Idealismus 2.58–77.

GAMUT, L. T. F. 1991. Logic, language and meaning. Chicago: University of Chicago Press.GAST, VOLKER, and JOHAN VAN DER AUWERA. 2010. Vers une typologie des opérateurs addi-

tifs scalaires. Approches de la scalarité, ed. by Pascale Hadermann and Olga Inkova,285–311. Genève: Droz.

GAST, VOLKER, and JOHAN VAN DER AUWERA. 2011. What is ‘contact-induced grammatical-ization’? Evidence from Mesoamerican languages. Grammaticalization and languagecontact, ed. by Björn Wiemer and B. Hansen. Berlin: Mouton de Gruyter, to appear.

GEBER, DANA. 2005. Telling ‘even’s apart. Proceedings of the 2005 annual conference of the Canadian Linguistic Association, ed. by Claire Gurski. Online: http:// http://westernlinguistics.ca/Publications/CLA-ACL/Geber.pdf.

GIANNAKIDOU, ANASTASIA. 1997. The landscape of polarity items. Groningen: University ofGroningen dissertation.

GIANNAKIDOU, ANASTASIA. 1998. Polarity sensitivity as (non)veridical dependency. Amster-dam: John Benjamins.

GIANNAKIDOU, ANASTASIA. 2000. Negative … concord? Natural Language and LinguisticTheory 18.457–523.

GIANNAKIDOU, ANASTASIA. 2007. The landscape of even. Natural Language and LinguisticTheory 25.39–81.

GINZBURG, JONATHAN. 1996. Interrogatives: Questions, facts and dialogue. The handbook ofcontemporary semantic theory, ed. by Shalom Lappin, 385–422. Oxford: Blackwell.

GLANZBERG, MICHAEL. 2002. Context and discourse. Mind and Language 17.333–75.GRICE, H. PAUL. 1975. Logic and conversation. Syntax and semantics, vol. 3: Speech acts,

ed. by Peter Cole and Jerry L. Morgan, 41–58. New York: Academic Press.GUERZONI, ELENA. 2003. Why even ask? Cambridge, MA: MIT dissertation.GUERZONI, ELENA. 2004. Even-NPIs in yes/no questions. Natural Language Semantics

12.319–43.HASPELMATH, MARTIN. 1997. Indefinites. Oxford: Oxford University Press.HASPELMATH, MARTIN. 2007. Coordination. Language typology and syntactic description,

2nd edn., ed. by Timothy Shopen, 1–51. Cambridge: Cambridge University Press.HEIM, IRENE, and ANGELIKA KRATZER. 1998. Semantics in generative grammar. Oxford:

Blackwell.HEINE, BERND, and TANIA KUTEVA. 2005. Language contact and grammatical change. Cam-

bridge: Cambridge University Press.HERBURGER, ELENA. 2003. A note on Spanish ni siquiera, even, and the analysis of NPIs.

Probus 15.237–56.HINTIKKA, JAAKKO. 1973. Logic, language-games and information: Kantian themes in the

philosophy of logic. Oxford: Clarendon.HIRVENSALO, LAURI. 1963. Saksa-Suomi-Suursanakirja. Helsinki: Werner Söderström.HOEKSEMA, JACK, and HOTZE RULLMANN. 2001. Scalarity and polarity: A study of scalar ad-

verbs as polarity items. Perspectives on negation and polarity items, ed. by Jack Hoek-sema, Hotze Rullmann, Victor Sanchez-Valencia, and Ton van der Wouden, 129–71.Amsterdam: John Benjamins.

HOEKSEMA, JACK, and FRANS ZWARTS. 1991. Some remarks on focus adverbs. Journal of Se-mantics 8.51–70.

50 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

HOLE, DANIEL. 2004. Focus and background marking in Mandarin Chinese: System and the-ory behind cái, jiù, d� u and y� . (Asian linguistics 5.) London: RoutledgeCurzon.

HORN, LAWRENCE. 1969. A presuppositional analysis of only and even. Chicago LinguisticSociety 5.97–108.

HORN, LAWRENCE. 1971. Negative transportation: Unsafe at any speed? Chicago LinguisticSociety 7.120–133.

HORN, LAWRENCE. 1989. A natural history of negation. Chicago: University of ChicagoPress.

HUALDE, JOSÉ IGNACIO, and JON ORTIZ DE URBINA. 2003. A grammar of Basque. Berlin:Mouton de Gruyter.

ITEN, CORINNE. 2002. Even if and even: The case for an inferential scalar account. UCLWorking Papers in Linguistics 14.119–57.

JACOBS, JOACHIM. 1982. Syntax und Semantik der Negation im Deutschen. Munich: Fink.JACOBS, JOACHIM. 1983. Fokus und Skalen: Zur Syntax und Semantik der Gradpartikeln im

Deutschen. Tübingen: Niemeyer.JACOBS, JOACHIM. 1984. The syntax of bound focus in German. Groninger Arbeiten zur Ger-

manistischen Linguistik 25.172–200.JUNGHANNS, UWE. 2006. Scope conflicts involving sentential negation in Czech. Negation

in Slavic, ed. by Sue Brown and Adam Przepiórkowski, 105–33. Bloomington, IN: Slavica.

KALERIKOS, ALEXIS. 1995. Even: How to make theories with a word. Journal of Pragmatics24.77–98.

KARTTUNEN, LAURI. 1969. Discourse referents. Austin: University of Texas at Austin disser-tation.

KARTTUNEN, LAURI. 1976. Discourse referents. Syntax and semantics, vol. 7: Notes from thelinguistic underground, ed. by James D. McCawley, 363–85. New York: AcademicPress.

KARTTUNEN, FRANCES, and LAURI KARTTUNEN. 1977. Even questions. North East LinguisticSociety (NELS) 7.115–34.

KARTTUNEN, LAURI, and STANLEY PETERS. 1979. Conventional implicature in Montaguegrammar. Syntax and semantics, vol. 11: Presuppositions, ed. by Choon-Kyu Oh andDavid A. Dinneen, 1–56. New York: Academic Press.

KATARA, PEKKA, and INGRID SCHNELLBACH-KOPRA. 1997. Suomi-saksa-suursanakirja.[Finnish–German dictionary.] Helsinki: WSOY.

KAY, PAUL. 1990. Even. Linguistics and Philosophy 13.59–111.KLEIN, WOLFGANG, and CHRISTIANE VON STUTTERHEIM. 1987. Quaestio und referentielle Be-

wegung in Erzählungen. Linguistische Berichte 109.163–83.KLEIN, WOLFGANG, and CHRISTIANE VON STUTTERHEIM. 1992. Textstruktur und referentielle

Bewegung. Zeitschrift für Literaturwissenschaft und Linguistik 86.67–92.KLIMA, EDWARD. 1964. Negation in English. The structure of language, ed. by Jerry A.

Fodor and Jerrold J. Katz, 246–323. Englewood Cliffs, NJ: Prentice-Hall.KÖNIG, EKKEHARD. 1981a. The meaning of scalar particles in German. Words, worlds and

contexts: New approaches in word semantics, ed. by Hans J. Eikmeyer and HannesRieser, 107–32. Berlin: Walter de Gruyter.

KÖNIG, EKKEHARD. 1981b. Kontrastive Analyse und zweisprachige Wörterbücher: Die Grad-partikel even und ihre Entsprechungen im Deutschen. Partikeln und Deutschunterricht:Abtönungspartikeln für Lerner des Deutschen, ed. by Harald Weydt, 277–304. Heidel-berg: Groos.

KÖNIG, EKKEHARD. 1989. On the historical development of focus particles. Sprechen mitPartikeln, ed. by Harald Weydt, 318–29. Berlin: de Gruyter.

KÖNIG, EKKEHARD. 1991. The meaning of focus particles: A comparative perspective. Lon-don: Routledge.

KORTMANN, BERND. 1997. Adverbial subordination: A typology and history of adverbialsubordinators based on European languages. Berlin: Mouton de Gruyter.

KRIFKA, MANFRED. 1991. A compositional semantics for multiple focus constructions. Pro-ceedings of Semantics and Linguistic Theory (SALT) 1.128–57.

KRIFKA, MANFRED. 1995. The semantics and pragmatics of polarity items in assertion. Lin-guistic Analysis 15.209–57.

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 51

KRIFKA, MANFRED. 2008. Basic notions of information structure. Acta Linguistic Hungarica55.243–76.

LADUSAW, WILLIAMA. 1979. Polarity sensitivity as inherent scope relations. Austin: Univer-sity of Texas at Austin dissertation.

LAHIRI, UTPAL. 1998. Focus and negative polarity in Hindi. Natural Language Semantics6.57–23.

LAKA, ITZIAR. 1990. Negation in syntax: On the nature of functional categories and projec-tions. Cambridge, MA: MIT Press.

LAMBRECHT, KNUD. 1994. Information structure and sentence form: Topic, focus and themental representation of discourse referents. Cambridge: Cambridge University Press.

LEHMANN, CHRISTIAN. 1982. Thoughts on grammaticalization: A programmatic sketch. (Ar-beiten des Kölner Universalienprojekts 48.) Cologne: Institut für Sprachwissenschaft.

LERNER, JEAN-YVES, and WOLFGANG STERNEFELD. 1984. Zum Skopus der Negation im kom-plexen Satz des Deutschen. Zeitschrift für Sprachwissenschaft 3.159–202.

LEWIS, DAVID. 1969. Convention: A philosophical study. Cambridge, MA: Harvard Univer-sity Press.

LÖBNER, SEBASTIAN. 1990. Wahr neben Falsch: Duale Operatoren als die Quantoren natür-licher Sprache. Tübingen: Niemeyer.

LYCAN, WILLIAM G. 1991. Even and even if. Linguistics and Philosophy 14.115–50.MARI, ALDA, and LUCIA MARIA TOVENA. 2006. A unified account for the additive and the

scalar uses of Italian neppure. New perspectives on Romance linguistics, vol. 1: Mor-phology, syntax, semantics, and pragmatics, ed. by Chiyo Nishida and Jean-PierreMontreuil, 187–99. Amsterdam: John Benjamins.

MCCONVELL, PATRICK. 1983. ‘Only’ and related concepts in Gurindji. Darwin: School ofAustralian Linguistics, MS.

MEYER, GUSTAV. 1891. Etymologisches Wörterbuch der albanesischen Sprache. Strassburg:Trübner.

MIKLOŠICH, FRANZ. 1886. Etymologisches Wörterbuch der slavischen Sprachen. Vienna:Braunmüller.

MIKLOŠICH, FRANZ. 1890. Die türkischen Elemente in den südost- und osteuropäischenSprachen (Griechisch, Albanisch, Rumänisch, Bulgarisch, Serbisch, Kleinrussisch,Großrussisch, Polnisch). Vienna: Tempsky.

NAKANISHI, KIMIKO. 2006. Even, only, and negative polarity in Japanese. Proceedings of Se-mantics and Linguistic Theory (SALT) 16.138–55.

NAKANISHI, KIMIKO. 2008. On compositionality of even: A case study of German auch nur.Proceedings of the 2008 annual conference of the Canadian Linguistic Association, ed.by Susie Jones. Online: http://homes.chass.utoronto.ca/~cla-acl/actes2008/CLA2008_Nakanishi.pdf.

POLJAKOVA, SVETLANA. 2000. Fokusdomäne von Gradpartikeln im Deutschen und Russi -schen. Linguistik Online 6.2. Online: http://www.linguistik-online.com/2_00/poljakov.html.

POLJAKOVA, SVETLANA. 2001. Negative Gradpartikeln im Russischen und Deutschen. Cur-rent issues in formal Slavic linguistics, ed. by Gerhild Zybatow, Uwe Junghanns, GritMehlhorn, and Luka Szucsich, 447–60. Frankfurt: Peter Lang.

PRESS, IAN. 1986. A grammar of Modern Breton. Berlin: Mouton de Gruyter.PROGOVAC, LJILJANA. 1994. Negative and positive polarity: A binding approach. Cambridge:

Cambridge University Press.ROBERTS, CRAIGE. 1996. Information structure in discourse: Towards an integrated formal

theory of pragmatics. OSU Working Papers in Linguistics (Papers in semantics)49.91–136.

ROBERTS, CRAIGE. 2004. Context in dynamic interpretation. Handbook of contemporary prag-matic theory, ed. by Laurence R. Horn and Gregory Ward, 197–220. London: Blackwell.

ROOTH, MATS. 1985. Association with focus. Amherst: University of Massachusetts disserta-tion.

ROOTH, MATS. 1992. A theory of focus interpretation. Natural Language Semantics1.75–116.

ROOTH, MATS. 1996. Focus. Handbook of contemporary semantic theory, ed. by ShalomLappin, 271–97. Oxford: Blackwell.

Volker Gast
Sticky Note
tLondon

52 LANGUAGE, VOLUME 87, NUMBER 1 (2011)

RULLMANN, HOTZE. 1997. Even, polarity, and scope. University of Alberta Papers in Experi-mental and Theoretical Linguistics 4.40–64.

RULLMANN, HOTZE. 2003. Additive particles and polarity. Journal of Semantics 20.329–401.RULLMANN, HOTZE, and JACK HOEKSEMA. 1997. De distributie van ook maar en zelfs maar:

Een corpusstudie. Nederlandse Taalkunde 2.281–317.SCHWARZ, BERNHARD. 2005. Scalar additive particles in negative contexts. Natural Lan-

guage Semantics 13.125–68.SCHWARZSCHILD, ROGER. 1999. GIVENness, AVOIDF, and other constraints on the placement

of accent. Natural Language Semantics 7.141–77.SCHWENTER, SCOTT. 1999a. Pragmatics of conditional marking: Implicature, scalarity and

exclusivity. New York: Garland.SCHWENTER, SCOTT. 1999b. Two types of scalar particles: Evidence from Spanish. Advances

in Hispanic linguistics, ed. by Javier Gutiérrez-Rexach and Fernando Martínez-Gil,546–61. Somerville, MA: Cascadilla.

SCHWENTER, SCOTT. 2000. Lo absoluto y lo relativo de las partículas escalares incluso ehasta. Oralia 3.169–97.

SCHWENTER, SCOTT. 2002. Additive particles and scalar endpoint marking. Journal of Bel-gian Linguistics 16.119–34.

SPERBER, DAN, and DEIRDRE WILSON. 1986. Relevance: Communication and cognition. Ox-ford: Blackwell.

STALNAKER, ROBERT. 1974. Pragmatic presuppositions. Semantics and philosophy, ed. byMilton K. Munitz and Peter K. Unger, 197–214. New York: New York University Press.

STALNAKER, ROBERT. 1978. Assertion. Syntax and semantics, vol. 9: Pragmatics, ed. by PeterCole, 315–32. New York: Academic Press.

STALNAKER, ROBERT. 1998. On the representation of context. Journal of Logic, Languageand Information 7.3–19.

STOJANOV, STOJAN, and MIROSLAV JANAKIEV. 1983. Studienwörterbuch Altbulgarisch–Deutsch. Sofia: Kliment Ochridski.

STUBBLEFIELD, MORRIS, and CAROL MILLER. 1991. Diccionario Zapoteco de Mitla, Oaxaca.Mexico City: Instituto Linguistico de Verano.

SUDHOFF, SEBASTIAN. 2010. Focus particles in German: Syntax, prosody and informationstructure. Frankfurt: Lang.

TAGLICHT, JOSEF. 1984. Message and emphasis: On focus and scope in English. London:Longman.

TRAUGOTT, ELIZABETH. 2006. The semantic development of scalar focus modifiers. Thehandbook of the history of English, ed. by Ans van Kemenade and Bettelou Los,335–59. Oxford: Blackwell.

VALLDUVÍ, ENRIC. 1994. Polarity items, n-words and minimizers in Catalan and Spanish.Probus 6.263–94.

VAN DER AUWERA, JOHAN. 1985. Only if. Logique et Analyse 109.61–74.VAN DER AUWERA, JOHAN. 2008. In defense of classical semantic maps. Theoretical Linguis-

tics 34.39–46.VAN DER AUWERA, JOHAN. 2009. The Jespersen cycles. Cyclical change, ed. by Elly van

Gelderen, 35–71. Amsterdam: John Benjamins.VAN DER AUWERA, JOHAN, and VLADIMIR PLUNGIAN. 1998. Modality’s semantic map. Lin-

guistic Typology 2.49–124.VAN DER AUWERA, JOHAN, and ÇEYHAN TEMÜRCÜ. 2006. Semantic maps. Encyclopedia of

language and linguistics, 2nd edn., ed. by Keith Brown, 131–34. Oxford: Elsevier.VAN DER WOUDEN, TON. 1997. Negative contexts: Collocation, polarity and multiple nega-

tion. London: Routledge.VANDEWEGHE, WILLY. 1981. Ook maar X. Studia Germanica Gandensia 21.15–56.VASMER, MAX. 1953. Russisches Etymologisches Wörterbuch. Heidelberg: Winter.VISCONTI, JACQUELINE. 2004. Sintassi e uso delle particelle perfino, persino e addirittura in

italiano antico. La sintassi dell’italiano antico, ed. by Maurizio Dardano and GianlucaFrenguelli, 445–63. Rome: Editrice L’Aracne.

VISCONTI, JACQUELINE. 2005. On the origins of scalar particles in Italian. Journal of Histori-cal Pragmatics 6.237–61.

SCALAR ADDITIVE OPERATORS IN THE LANGUAGES OF EUROPE 53

VON STECHOW, ARNIM. 1991. Current issues in the theory of focus. Semantik: Ein interna-tionales Handbuch der zeitgenössischen Forschung [Semantics: An international hand-book of contemporary research], ed. by Arnim von Stechow and Dieter Wunderlich,804–25. Berlin: Walter de Gruyter.

VON STECHOW, ARNIM. 2007. Schritte zur Satzsemantik. Tübingen: Universität Tübingen.WILKINSON, KARINA. 1996. The scope of even. Natural Language Semantics 4.193–215.WILSON, DEIRDRE, and DAN SPERBER. 2004. Relevance theory. Handbook of contemporary

pragmatic theory, ed. by Laurence R. Horn and Gregory Ward, 607–32. London: Blackwell.

WITTGENSTEIN, LUDWIG. 1922. Tractatus Logico-Philosophicus. London: Kegan Paul, Trench,Trubner & Co.

ZWARTS, FRANS. 1995. Nonveridical contexts. Linguistic Analysis 25.286–312.ZWARTS, FRANS. 1998. Three types of polarity. Plurality and quantification, ed. by Fritz

Hamm and Erhard Hinrichs, 177–238. Dordrecht: Kluwer.

Gast [Received 14 July 2008;Friedrich Schiller University Jena revision invited 4 January 2009; Department of English and American Studies revision received 12 October 2009; Ernst-Abbe-Platz 8 accepted pending revisions 10 February 2010; 07743 Jena, Germany revision received 24 May 2010; [[email protected]] accepted 24 September 2010][[email protected]]

Author Queries LSA87101

Dear AuthorDuring the preperation of your manuscript for publication, the questions listed below have arisen.Please attend to these matters and return this form with your proof. Many thanks for your assistance.

[AQ1] Text has reduced very slightly (by .25 of a point) to fit in width. Okay?[AQ2] Text has reduced slightly (by 1 point) to fit in width. Okay?[AQ3] Do you really want deletion of ), after (Carlson’s ‘payoff moves’ per comment in word file? There

will be no closing parentheses.[AQ4] Edits are unclear in the word file for this paragraph.