The study also highlights the constructs of current linguistic theory, arguing for distinctive features and the notion 'onset' and against some of the claims of Optimality Theory and Usage-based accounts.
The importance of Henk Zeevat's new monograph cannot be overstated. [...] I recommend it to anyone who combines interests in language, logic, and computation [...]. David Beaver, University of Texas at Austin
INTRODUCTION It is true that, as suggested by the title of this book, the study of negative polarity items (NPIs) has tended to focus on selected expressions in a handful of languages. English “any” and to a degree “ever” have often been taken as the measure of negative polarity, which relate to the series of items whose distribution or interpretation depends on particular environments.
(1) I didn’t ever buy any tickets in the end. / *I ever bought any tickets in the end.
(2) If I ever bought any tickets, / *Although I ever bought any tickets,
(3) I’m glad we ever got any tickets. / ?*I know we ever got any tickets.
Polarity raises three issues referred to by Israel as the licensing, the sensitivity and the diversity problems (2011: 31). What defines the various licensing contexts polarity items such as negation, conditionals and emotive factives illustrated above? A first type of definition -- called “monolithic” by the editors in their introduction -- relies on one factor. This comprises the Affective feature proposed by Klima (1964), Downward Entailment (DE) by Ladusaw (1979), and Non-Veridicality by Giannakidou (1998). Ladusaw claims that DE licenses NPIs as in (1), a context that validates inferences from sets to subsets to support the inference from “I didn't buy a car'' to “I didn't buy a BMW”; but since non-DE environments such as (2) and (3) also license NPIs, revisions are proposed, notably by von Fintel (1999), whose notion of Strawson Downward Entailment integrates utterance presuppositions. Giannakidou purports that non-veridical contexts license NPIs because they do not assert the existence of an event (as is found with negation in (1) and the conditional in (2), neither of which communicating that a purchase took place). While such monolithic approaches have the elegance of simplicity, they may not account for all licensing cases -- conditionals as in (2) not being a DE environment, emotive factives not being non-veridical -- and may lack the finesse to account for granular variation between particular items.
The problems of sensitivity and diversity relate to why some items should be sensitive to polarity context at all, and why this sensitivity should vary between items (proposals to distinguish sub-classes of negative polarity on the basis of sub-types of entailment having been made by Zwarts 1993). One way to account for the intricacy of negative polarity is what the editors identify as the second type of definitions of negative polarity, “conspiracy theories”. Relating to scalar views of polarity (e.g. Duffley and Larrivée 2010), such theories envisage negative polarity as the convergence between the meaning of items and the meaning of contexts. This is a prominent assumption in the formal scalar perspectives developed by many German colleagues, notably Manfred Krifka (1995). That perspective shines through in the selection of papers, collected from a 2011 Göttingen conference. Wider empirical coverage is advocated, as is the conspiracy theory of polarity. The 13 chapters are summarised and commented below, before an overall evaluation is provided.
SUMMARY Following the introduction by Regine Eckardt and Manfred Sailer, the first chapter is on “Mapping the West Germanic ‘Any’” by Johan van der Auwera and Lauren Van Alsenoy (pp. 21-44). Their objective is to account for the historical development of the equivalents of “any” in West Germanic -- the English version of course, and also Dutch “enig” and German “einig”, with further reference to Frisian, Afrikaans, Pennsylvania Dutch and Yiddish. They do so through a revised version of the semantic maps devised by Haspelmath (1997) for synchronic typological explorations. Because they focus on contexts rather than readings, Haspelmath’s maps do not distinguish between NPI and Free Choice status in contexts such as conditionals in (2). The revised maps taking explicit inspiration from the Aristotelian Square of Opposition have a central spine of non-specific direct negation, non-specific negative polarity and non-specific free-choice function, to which are adjoined qualitative interpretations such as Dutch “Dit vind ik enig” ‘this find I any’ “I find this wonderful”, comparable to qualitative uses of “some” (“That was some party”). Although the revised maps have fewer cells and therefore make weaker proposals in terms of possible routes of evolution, the paper illustrates how Haspelmath’s semantic maps can be redrawn to make sense of item evolution in diachrony.
Jack Hoeksema contributes the second chapter on “Polarity items in Strawsonian contexts: a comparison” (pp. 47-77). He critically assesses characterisations of negative polarity contexts as (Strawson) Downward Entailment and non-veridicality. An impressive amount of quantitative data is presented on the contextual distribution of “any”, “ever”, “anymore” (in US and UK English, compared to the Dutch and German equivalents), adverbial “any” (as in “any longer”), “remotely” in English and Dutch, minimizers, post-nominal “alive”, “yet” and “as yet”, temporal NPIs such as “in weeks” in English and Dutch, “need” in English, Dutch and German. Focus is on particular licensing contexts of superlative and exceptive types (“only”, “the only”, “the first”, “the last”). The intricacies of the data seem to support a conspiracy analysis where the semantics of each item and context should account for distributional differences. However, whether the unexpected variation can be explained without stipulation remains to be seen: American English “anymore” is licensed by “the only” but not by “only”, Dutch “meer” is licensed by “only” but not by “the only”.
The evolution of one use of the little studied polarity item “either” is offered by Volker Gast in “From ‘aegewaeder’ to ‘either’: The distribution of a negative polarity item in historical perspective” (pp. 79-101). “Either” as a determiner in present-day English is analysed as an existential quantifier, that lives on discourse-old information -- “I bought two Austrian novels, but I haven’t read either book” is felicitous because the notion of book is introduced in the antecedent context, which isn’t the case of the notion of sexes in the infelicitous “There didn’t exist either sex before the world was created”. The evolution of “either” into an existential quantifier is due to the decline of existential quantifier “outher” that pulls “either” in existential functions. Its restriction to non-veridical contexts is explained by the emergence of “both” that pushes “either” into negative polarity. The convergence of meaning and context is thus illustrated by diachronic evolution.
A new general proposal on NPI licensing is put forward by Johan Brandtler in “Evaluability: an alternative approach to polarity sensitivity” (pp. 103-125). He observes that contemporary approaches to polarity licensing do not provide an account of the distribution of Positive Polarity Items in Negative Polarity Contexts such as conditionals (“If you’re still / ever going to visit your aunt, you can ask her”) and of Negative Polarity Items in factive environments (as in (3) above). Factive complement clauses in Swedish are identified by overt marking on the edge of C that can be characterised by the notion of evaluability, the ability to be accepted or challenged. While it is claimed that the notion of (non)evaluability is superior to DE, non-veridicality and Progovac-style syntactic binding, I found it difficult to assess how this approach -- that seems very close to the non-veridicality analysis -- fares better in accounting for problematic emotive factive contexts.
Interaction between negation and focus particles is discussed by Luka Crnic in “How to get even with desires and imperatives” (pp. 127-153). “Even” normally indicates the most unlikely occurrence in a series (“John made even a FILM (on top of a book and an exhibition)”). This indication yields odd results in ?*“John made even ONE film” due to the conflicting knowledge that one occurrence is the most likely quantity to be realised if a film is to made at all. The oddity disappears in some factive subordinates and imperatives (“John hopes to make even ONE film of this quality”), where “even” highlights the least unlikely occurrence, a case highlighted by the existence of German focus particle “auch nur”. Moving “even” over negation at LF is not going to be available with hope predicates and positive imperatives (“Inspire even ONE child and your job as a teacher is done”). An exhaustivity operator is difficult to restrict to the relevant contexts. The upward-entailing value of imperatives and factives makes them unlikely candidates to a modal environment analysis that would license this strange NPI. The proper adjustments to either of these treatments is left to future research, which it is suggested would help in solving the licensing of NPIs in factives where a covert “even” might be involved.
I-Ta Chris Hsieh provides a theoretical contribution “On NPI licensing in possibility conditionals” (pp. 155-182). Conditionals of the type “If you’re ever going to visit your aunt, she can tell you about the Kosovo war” are generally assimilated to a possibility modal (“You might be going to visit your aunt”). This assimilation however leads to inaccurate predictions: it doesn’t follow from the cited example that if I visit my Canadian aunt as opposed to say my Bosnian aunt that she can tell me about the bombing of Kosovo, and the lack of downward monotonicity should bar NPI licensing, contrary to observations. So rather than treating possibility modals as existential quantifiers over worlds, they could be seen as universal quantification over a subset of worlds, with which Strawson Downward Entailment goes through to license NPIs.
“An analogy between a connected exceptive phrase and polarity item” is proposed by Jon Gajewski (pp. 183-211). The Connected Exceptive Marker “but” in “Nobody but Bill has visited their aunt this year” (as opposed to free exceptives such as the “besides” phrase in “Besides Bill, nobody has visited their aunt this year”) is comparable to a Negative Polarity Item. It demands a polarity context, and introduces alternatives. These alternatives are deemed to be of a second-order type, and exploited by an exhaustification operator. The recursive application of the operator imposes syntactic minimality on the connected exceptive phrase, and uniqueness at the second level.
The issue of what defines licensing contexts is revisited by Elena Herburger and Simon Mauck in “The chance of being an NPI” (pp. 213-239). The remind the reader that not all licensing contexts are Downward-Entailing (notably conditionals, factive complements, as in (2), (3) above). The conventional dimensions of NPI status should be captured by lexical properties of items, and the view that NPIs include a covert “even” is considered but found wanting because not all NPIs relate to the bottom of a scale. It is proposed in a Minimalist approach that negative polarity be defined by an abstract feature applying to NPIs and negative polarity licensing contexts. Such a view accounts for historical changes in NPI status. It however raises the issue of whether items such as “few” and “most” as sometimes licensors should be analysed as feature-bearing. The strength of NPIs is also a question that the authors refer to future research. It is surprising that recent proposals such as Giannakidou (2010) and well-known feature-based analyses such as Martins (2000) are ignored in this context.
Robert Levine considers “The modal need VP gap (non)anomaly” (pp. 241-265). “Need” is a NPI as an auxiliary (“I need not know”), and is not as a “normal” verb (“I need to know”). As an auxiliary, “need” should allow VP ellipsis (“Should you worry about this? I think you needn’t”), in which case it is licensed only with local negation (*“Should you worry about this? I don’t think you need”) we are told, whereas other contexts license “need” outside ellipsis cases (“OK I don’t think you need worry about this yet”). The HPSG treatment suggests that “need” with a gap requires a local superstrong licenser.
Regine Eckardt and Eva Csipak discuss “Minimizers -- Towards pragmatic licensing” (pp. 267-298). They further a conspiracy analysis by looking at the behaviour of minimisers such “lift a finger”. These are not always readily acceptable in the scope of “few”, “only” and “rarely”. The reason why “Few students even lifted a finger” is infelicitous is that it evokes a minimal action that does not exist in the real world. This raises the question of why both “few” and “a few” license “a shred of” according to Google, in real-life examples such as “few have a shred of real evidence to back up the claim”. More empirical work is required on such expressions.
High degree polarity items are explored by Ai Matsui in the chapter “Revisiting the licensing problem through understating NPIs -- The case of Japanese ammari '(not) very much’” (pp. 299-322). The comparison between Japanese ‘a(n)mari’ and near synonyms “sonnami” and “all that” is conducted to validate the view that polarity emerges from the interaction between items and context. The unexpected distribution of the Japanese items is found in “because” clauses. In the equivalent of “I turned on the air conditioner because the room was so hot”, the occurrence of the NPI is analysed in terms of relevance, lack of relevance of a question such as “Is it all that hot?” being blamed for the apparent infelicity of the sequence. How to establish relevance is not spelt-out in a clearly transferable way.
Edgar Onea and Manfred Sailer deal with degree expression “all that” in “Really all that clear?” (pp. 323-350). They use three methods to test their semantic analysis according to which the presupposition of a high degree held typically by someone other than the speaker is denied typically by the speaker: in “The task isn’t all that hard”, the speaker is rejecting the previously entertained notion that the task is hard to a high degree. Corpus profile, informant judgement (of 5 native speakers) and an informant judgement experiment (by 23 native speakers) suggest that “all that” is a weak NPI and document some complexities, as the convergent role of “realise” and “only” for the acceptability of “John realised that only very happy people are all that satisfied with their lives”, as compared to ??“Only very happy people are all that satisfied with their lives” which seems less felicitous. The notion that lexical semantics accounts for polarity distribution without explicit reference to a negative polarity category supports the conspiracy theory evoked in the introduction.
Mingya Liu, Regine Eckardt and Janina Rado present the results of three experimental studies in “Polarity in context” (pp. 351-368). They are concerned with the scalar analysis of polarity, such that polarity items give rise to a range of alternatives and represent the strongest alternative in the contexts that license it. This implies that scalar values uncontroversially evoked by a minimiser should favour the licensing of polarity items. This assumption is verified in three acceptability judgement tests, one concerning “auch nur” with and without a minimiser NPI, one relating to “auch nur” without a minimiser but in a scalar context, and the last integrating the three variables of “auch nur” and a minimiser in a scalar context. The results confirm expectations in that “auch nur” is rated as more acceptable when used with an NPI. However, “auch nur” actually received lower ratings in a scalar context than out-of-context. In this light, the claim made in the conclusion that “auch nur” is sensitive to discourse-old information, whereas “even” would accommodate such information, seems unsupported.
EVALUATION The volume has the material quality and the price tag associated with de Gruyter volumes. It contains few typographical errors and only a brief notional index. Unsurprisingly, the book is marked by the German tradition in more ways than one. Many issues that arise in Germanic languages find their places in the volume, notably the little discussed “weak even” case represented by German “auch nur”. Formal semantic frameworks are well represented, often integrating pragmatic dimensions.
Interesting contributions from leading scholars on the question of polarity are brought together with no less interesting chapters from younger colleagues. The objectives of dealing with unexplored questions and of furthering a conspiracy analysis of polarity where felicity is explained by convergence between meaning of items and contexts are partially met. Obviously, however, a conspiracy analysis is bound to be ill-equipped to explain arbitrary conventions imposed by particular languages. As noted by Larrivée (2007), “at all” and “du tout” are indistinguishable at the interpretation level, but have very different distributions, such that the English version is a weak NPI (“If you need anything at all” is fine) whereas the French is a superstrong one (bad not only in conditionals and questions, but also with most n-words, and only good with clause-mate propositional negator).
Inevitably for a volume emerging from a conference, many chapters explore similar concerns, while other issues in the field are not touched on. Maybe a thematic grouping of chapters would have furthered homogeneity. It is regrettable that more cross-referencing is not found between the contributions, and that there is some amount of “dialogue de sourds” involved -- claims shown not to hold in some parts of the book being used as explanations in other parts. Apart from a couple of exceptions, the bibliography of individual chapters tends to be reasonably inclusive.
The audience will mostly be specialists working on negative polarity.
REFERENCES Duffley, Patrick and Pierre Larrivee. 2010. Anyone for non-scalarity? English Language and Linguistics 14,1, 1-17.
Giannakidou, Anastasia. 2010. The dynamics of change in Dutch enig: From nonveridicality to strong negative polarity. Natural Language and Linguistic Theory 28,4, 861-875.
Giannakidou, Anastasia. 1998. Polarity sensitivity as (non)veridical dependency. Amsterdam: Benjamins.
Haspelmath, Martin. 1997. Indefinite pronouns. Oxford: Oxford University Press.
Israel, Michael. 2011. The Grammar of Polarity: Pragmatics, Sensitivity, and the Logic of Scales. Cambridge: Cambridge University Press.
Klima, Edward. 1964. Negation in English. J. A. Fodor & J. J. Katz (eds.). The Structure of Language. Englewood Cliffs: Prentice-Hall. 246-323.
Krifka, Manfred. 1995. The semantics and pragmatics of polarity items. Linguistic Analysis 25, 1-49.
Ladusaw, William. 1979. Negative Polarity as Inherent Scope Relations. PhD dissertation, U. of Texas. Published New York: Garland, 1980.
Larrivee, Pierre. 2007. Du tout au rien: libre-choix et polarité négative. Paris: Champion.
Martins, Ana Maria. 2000. Polarity Items in Romance: Underspecification and Lexical Change. Susan Pintzuk, George Tsoulas and Anthony Warner (eds). Diachronic Syntax: Models and Mechanisms. Oxford and New York: Oxford University Press. 191-219.
von Fintel, Kai. 1999. NPI-licensing, Strawson-entailment, and context-dependency. Journal of Semantics 16, 97-148.
Zwarts, Frans. 1993. Three types of polarity. MS. Published in 1998 in F. Hamn and E. Hinrichs (eds). Plural Quantification. Dordrecht: Kluwer.
ABOUT THE REVIEWER:
Pierre Larrivee is Professor in French Linguistics at Universite de Caen (France). He currently works on the interface between pragmatics and grammatical variation and change in French and English. His recent papers on the evolution of negation have been published in journals such as Lingua, Linguistics, the Journal of Pragmatics and English Language and Linguistics. He heads the research centre CRISCO, which primarily develops work on the relationship between syntax and interpretation.