The use of polarity items, especially negative polarity items (NPI), has received a lot of attention, and it has been investigated in many languages. As far as the Scandinavian languages are concerned, polarity items have been studied in Danish (Jensen Reference Jensen1999) and Norwegian (Lindstad Reference Lindstad1999; Johannesen Reference Johannesen2003), but until now no extensive investigation of Swedish polarity items has been published. Johan Brandtler's dissertation is thus a welcome contribution to the field. Examples of polarity items are given in the comprehensive Swedish Academy reference grammar (Teleman, Hellberg & Andersson Reference Teleman, Hellberg and Andersson1999, vol. 4:187ff.), where they are considered to be ‘expressions dependent on negation’ (Swed. negationsberoende uttryck). Brandtler argues that this conception of dependency is too narrow. Instead he aims for a characterization of polarity item licensing that works for both affirmative and negative items, in declarative clauses as well as in non-affirmative contexts. See the following examples with the NPI någonsin ‘ever’ and the positive polarity item (PPI) fortfarande ‘still’ (1f.):
(1)
(2)
(3)
(4)
Instead of starting from the almost complementary distribution of NPIs and PPIs in declarative sentences, illustrated in (1) and (2) above, as most other researchers have done, Brandtler asks ‘if negative polarity items are sensitive to negation, what is “negative” about the yes/no-question and the conditional in (3)? And if positive polarity items are sensitive to affirmative contexts, what is “affirmative” about the yes/no-question and the conditional in (4)?” (p. 2). His strategy is to take the seemingly non-licensed occurrences of NPIs and PPIs in (3) and (4) above as the starting point and to argue that the contexts in (1) and (2) are restrictive environments where polarity items in addition need to be formally licensed. He proposes that polarity items are sensitive to evaluability, a pragmatic notion “that refers to the possibility of accepting or rejecting a sentence as true in a communicative exchange” (p. 3). Although evaluability resembles semantic notions like truth and veridicality (see Giannakidou Reference Giannakidou1998), it is distinct in that what is at issue in a conversational exchange is not so much the logical or actual truth of a given utterance; rather it is whether or not the participants accept an utterance as true. Furthermore, only sentences which are used by the speaker to assert, presuppose or entail the truth are evaluable in this technical sense; other sentences are non-evaluable (82f.).
The gist of Brandtler's proposal is that NPIs and PPIs may occur freely in non-evaluable clauses like yes/no-questions and if-clauses. When they occur in evaluable clauses like declaratives, they must be licensed. He then proceeds to show that this distinction between evaluable and non-evaluable sentences is mirrored syntactically in Swedish in terms of clause structure. Evaluable sentences, he argues, are characterized by the fact that they have an initial Spec position, [Spec, CP] (Swed. fundament), whereas non-evaluable clauses typically lack such a position. (1) and (2) above are examples of clauses with overt [Spec, CP]; hence they are evaluable and polarity items need to be licensed. Supporting evidence for the importance of syntactic clause type comes from the contrast between (3a) and (5).
(5)
Whereas the NPI någonsin is acceptable in a non-evaluable clause like the verb-initial yes/no-question in (3a), it is not acceptable in the declarative question with a filled Spec position in (5). One consequence of this syntactic characterization is that wh-questions, which arguably involve overt Spec positions, are evaluable and consequently should not allow polarity items unless these are licensed. Questions like in (6) at first glance seem to go against the prediction.
(6)
However, Brandtler argues that (6) is not a genuine information question but a rhetorical question, to which the expected answer is ‘Nothing’. He then demonstrates how the interpretation of the wh-word in rhetorical questions can be seen as downward entailing and thus function as the licensing element for the NPI (cf. Ladusaw Reference Ladusaw1979).
In addition to the distinction between evaluable and non-evaluable clauses, Brandtler makes a further distinction among the evaluable clauses between those that are subjected to evaluation and those that are not (86f.). Clauses which express asserted information may be accepted or challenged in a discussion; they are hence subjected to evaluation. Clauses that express presupposed information are not normally challenged, although they may be. Consequently they are not subjected to evaluation. This distinction, Brandtler argues, corresponds to whether the Spec position is overt or covert. Clauses that are subjected to evaluation always have overt Spec positions whereas clauses with presupposed content do not, as exemplified by embedded (non-V2) declarative and relative clauses. This distinction allows the author to capture a subtle distinction between att-exclamatives, where NPIs may occur, as in (7a), and wh-exclamatives, where NPIs are restricted (7b).
(7)
The presence or absence of a Spec position plays a crucial role in Brandtler's analysis, which is carried out within the Minimalist Program. Following Chomsky (Reference Chomsky, Freidin, Otero and Zubizarreta2008) and Platzack (Reference Platzack2008), Brandtler assumes that the availability of [Spec, CP] is dependent on an edge feature in C. He furthermore suggests that the instantiation of this edge feature in C in Swedish is a syntactic reflex of evaluability (113). However, this reflex is not to be taken as a universal connection. In Swedish, the edge feature has come to be associated with evaluability, but this association is arbitrary and languages are thus expected to vary in whether and how they make use of this feature.
Throughout the book, Brandtler displays a thorough familiarity and understanding of the prolific literature on the topic of polarity licensing. He discusses previous approaches to polarity licensing in an insightful way, addressing both the mainly semantic analyses proposed by Ladusaw (Reference Ladusaw1979) and Giannakidou (Reference Giannakidou1998), and the syntactic accounts put forward by e.g. Progovac (Reference Progovac1994). As the book's title indicates, Brandtler is mainly concerned with the general principles for polarity licensing and he has discovered an interesting connection between clause type and the distribution of polarity sensitive expressions in Swedish. By taking a pragmatic approach, he is able to take into account how speakers and hearers may treat asserted versus presupposed information. The advantages of this approach become particularly clear in his discussion of long-distance NPI licensing (Chapter 9), where speaker attitude and the nature of embedding predicates turn out to be important. Brandtler writes and argues well and for the most part his reasoning is easy to follow. The alphabetic index at the end of the volume is very useful
I find the ideas underlying Brandtler's evaluability hypothesis interesting and well worth pursuing. However, there are some unclarities in his presentation, which need to be resolved. For instance, what does evaluability apply to? In the definition quoted above, it is sentences which are either evaluable or non-evaluable. But in other places Brandtler talks about clauses and utterances as being evaluable. This is actually not a terminological quibble. For Brandtler it is important that an if-clause, as in the antecedent of a conditional (3b) above, counts as non-evaluable and lacking [Spec, CP]. The entire conditional statement, on the other hand, is a declarative clause, with [Spec, CP], which is evaluable. Given Brandtler's pragmatic approach, one would think that a definition based on the notion utterance would be most natural, but it is difficult to argue that the antecedent of a conditional is an utterance separate from the rest of the sentence.
As already mentioned, Brandtler aims to account for the general principles underlying polarity licensing in Swedish. However, in this book, he does not discuss individual items in any detail. Applying his hypothesis to a range of strong, weak and intermediate polarity items would be a natural next step to take, and also a way of testing the validity of the predictions and the usefulness of the notion evaluability. I will end by commenting on the subtitle of the dissertation, which is actually a bit misleading. Given the emphasis on pragmatic notions, it would have been more appropriate to let the subtitle be The Syntax, Semantics and Pragmatics of Polarity Item Licensing in Swedish.