Accounting for the data According to the semantics above, exhibition-ka in (1a)

corresponds to the unique x which is an exhibition in context C at context time, while exhibition-

kii (1b) is the unique x which was an exhibition at a time prior to c(time). The use of -kii in (1b)

implies that the stronger alternative with -ka is false. Hence, in (1b) a cessation implicature is generated and the exhibition is understood to be closed at c(time), in this case utterance time. Similarly, our semantics predicts the readings of kii in (2), (3) and (4). ‘Bike’ is a permanent property, consequently a cessation implicature is not available and bike-kii is discourse anaphoric. The ‘sun’ is permanent and unique. The use of sun-ka in (4) would imply the existence of new sun. Since this is not a plausible alternative, -kii is used for locating the sun, regardless of its visibility. Invisibility is a derivable implicature in (6), where the use of sun-kii refers to an eclipsed sun.

(6) a. Shalay waxa=an firinayay cadcee-da.

yesterday FOC=1S look.PAST sun-ka

‘Yesterday I looked at the sun (visible).’ [2016-01-30]

b. Shalay waxa=an firinayay cadcee-dii. yesterday FOC=1S look.PAST sun-kii

‘Yesterday I looked at the sun (eclipsed).’

According to (5), cadcedii in (6b) would be the x which was a sun BEFORE the time of observing the sun. cadceeda in (6a) corresponds to the x which was a sun AT the time of looking at x. As with (4), there is a competition between the two meanings, but the sun cannot stop being a sun so the only possible meaning of (6b) is that the sun is invisible. The remaining puzzle is

why the invisibility implicature arises in (6b), but not in (4), where the use of –da is even

ungrammatical.

Salience and (not-)at-issue status of subordinate clauses Katja Jasinskaja, University of Cologne

Not-at-issue content (presuppositions, conventional implicatures) is often diagnosed by its in- ability to be targetted by direct rejection like No, she didn’t or That’s not true. However, An- derBois et al. (2011) have observed and Syrett and Koev (2015) have confirmed experimentally that appositive relative clauses (ARCs), which according to Potts (2005) contribute conventional implicatures, are easier to reject if they stand in the sentence-final position:

(1) a. A: Chloe, who decided to dress in a classical ballet style, has been chosen to audi- tion for the All Stars Dance Company. B: ?? No, she didn’t.

b. A: The All Stars Dance Company has chosen to audition Chloe, who decided to dress in a classical ballet style. B: No, she didn’t.

Developing Koev’s (2013) idea, the present paper argues that the apparent variable rejectability of ARCs is related to their salience in discourse, which in turn is affected by recency of process- ing. However, subordinate clauses of other kinds, in particular adverbial clauses, have a similar effect on salience, which raises the question of whether this is due to the same underlying prin- ciples, and whether adverbial clauses, like ARCs, contribute not-at-issue content.

Anaphoricity of direct rejection: Phrases that fall under Farkas and Bruce’s (2010) no- tion of direct rejection are composed of anaphoric expressions whose interpretation is strongly dependent on resolution to a salient antecedent in immediate context: answer particles like yes and no, VP ellipsis in she didn’t, demonstrative pronoun that in That’s not true. In contrast, “indirect” rejection, which is able to target all kinds of content, is usually more explicit, which makes its target more easily identifiable even if it is not the most salient one:

(2) A: Chloe, who decided to dress in a classical ballet style, has been chosen to audition for the All Stars Dance Company. B: Chloe didn’t dress in a classical ballet style! Main clause preference: But if direct rejection is really sensitive to salience, and since subor- dinate clauses are quite generally known to reduce the salience of the proposition they express, then not only ARCs, but also restrictive relatives (3-a), as well as adverbial clauses, (3-b)–(3-c), should be difficult to directly reject, or the rejection prefers to target the main clause:

(3) a. A: The dancer who decided to dress in a classical ballet style has been chosen to audition for the ‘All Stars’ Dance Company. B: ?? No, s/he didn’t.

b. A: After Mary laughed, she made a joke about the supervisor. B: No, she didn’t. (; Mary didn’t make a joke; 6; Mary didn’t laugh) adapted from Frazier & Clifton

c. A: Because/Although Chloe danced like an amateur, she was (not) chosen to audi- tion for ‘All Stars’. B: ?? No, she didn’t.

Clause order: Although experimental evidence is fragmentary on this point (Frazier and Clifton, 2005; Cooreman and Sanford, 1996), it apears that causal and concessive adverbial clauses give rise to the same clause order effect as ARCs, cf. (4) vs. (3-c), whereas temporal adverbial clauses and restrictive relatives show a main clause preference regardless of clause order.

(4) A: Chloe was (not) chosen to audition for ‘All Stars’, because/although she danced like an amateur. B: No, she didn’t.

Explaining rejection of subordinate clauses: We assume that the unit of discourse structure is a speech act. Koev argues that ARCs constitute independent speech acts, since they can contain

speech act adverbials like frankly in (5), which is not the case for restrictive relatives. Because- and although-clauses (5-b) also allow frankly, whereas temporal adverbial clauses do not (5-c), which shows that causal and concessive adverbial clauses can constitute speech acts of their own, while this is not the case for temporal clauses.

(5) a. Chloe, who frankly danced like an amateur, has been chosen to audition for the All Stars Dance Company.

b. Chloe has (not) been chosen, because/although frankly she danced like an amateur. c. Chloe cried after (#frankly) she danced like an amateur.

The standard machinery of e.g. Segmented Discourse Representation Theory (SDRT, Asher and Lascarides, 2003) can be used to explain the rest: Speech acts in discourse are connected by co- ordinating (Narration, Parallel, Contrast) and subordinating coherence relations (Elaboration, Explanation, Background). Syntactically subordinate clauses normally express subordinating relations (although there are exceptions). The Right Frontier Constraint (RFC) says that only the last processed speech act and the speech acts it is discourse-structurally subordinated to are accessible for attachment of new discourse material. If an ARC, a causal or a concessive clause is in sentence-final position, as in (4), it is the last processed speech act, but it is (normally) discourse subordinate to the speech act of the main clause, therefore RFC correctly predicts that a rejection can be attached to the speech act of either the main or the subordinate clause.

However, if the ARC is sentence-medial (1-a) or if the causal or concessive clause is sentence- initial (3-c), then the main clause is the last processed speech act and therefore the rejection can only attach to the main clause. Further assuming that anaphoric devices like no and VP ellip- sis have a strong preference to be resolved to an antecedent in the discourse unit to which the present sentence is directly attached, we predict that direct rejection can always target the main clause, but it can only target the subordinate clause if it is sentence-final.

An indirect rejection (2) may attach in the discourse structure to the whole preceding sen- tence, but because it provides more explicit cues that help identify its target, it is less restricted in finding its antecedent. In this respect the distinction between direct and indirect rejections is parallel to that between anaphoric pronouns like he vs. definite descriptions like the man.

Finally, the main clause preference in subordinate clauses that do not constitute indepen- dent speech acts, e.g. temporal after-clauses, can be explained by a principle like Frazier and Clifton’s (2005) main assertion hypothesis: “Other things equal, comprehenders prefer to relate material in a new sentence to the main assertion of the preceding sentence.” The main assertion of a complex sentence is normally found in the main clause (but again there are exceptions).

(Not-)at-issue status of subordinate clauses: In sum, ARCs and other subordinate clauses that can function as speech acts independent of the main clause show parallel behaviour with respect to direct rejection which can be explained by the same underlying principles. What does this mean for semantic theory? One possibility is that contra Potts (2005) ARCs are difficult to directly reject not because they contribute some special kind of content (conventional implicature), but because they are subordinate clauses that have reduced salience in discourse. This would also mean that the rejection test is not reliable in identifying not-at-issue content. However, in this paper we explore another theoretical possibility: that discourse salience and at-issueness are closely related notions, that adverbial clauses can be not-at-issue like ARCs, which however makes us rethink the relationship between at-issueness and projection.

2

References

AnderBois, S., Brasoveanu, A., and Henderson, R. (2011). Crossing the appositive/at-issue meaning boundary. In Proceedings of Semantics and Linguistic Theory, SALT’20, pages 328–346.

Asher, N. and Lascarides, A. (2003). Logics of Conversation. Studies in Natural Language Processing. Cambridge University Press.

Cooreman, A. and Sanford, A. J. (1996). Focus and syntactic subordination in discourse. Re- search paper 79, HCRC, Edinburgh, Scotland, U.K.

Farkas, D. F. and Bruce, K. B. (2010). On reacting to assertions and polar questions. Journal of Semantics, 27(1):81–118.

Frazier, L. and Clifton, C. (2005). The syntax-discourse divide: processing ellipsis. Syntax, 8(2):121–174.

Koev, T. K. (2013). Apposition and the Structure of Discourse. PhD thesis, Rutgers University- Graduate School-New Brunswick.

Potts, C. (2005). The Logic of Conventional Implicatures. Oxford University Press.

Syrett, K. and Koev, T. (2015). Experimental evidence for the truth conditional contribution and shifting information status of appositives. Journal of Semantics, 32:525–577.

Tonhauser, J. (2012). Diagnosing (not-) at-issue content. Proceedings of Semantics of Under- represented Languages in the Americas (SULA), pages 239–254.

Nominal property concept words in Basaá

P. Jenks (UC Berkeley), A. Koontz-Garboden (U Manchester), & E.M. Makasso (ZAS, Berlin) Theories of gradability and comparison (e.g., Kamp 1975; Klein 1980; von Stechow 1984; Kennedy 2007) have been developed on the basis of adjectival constructions in familiar languages, mostly English. In many languages, however, the main predicate in translationally equivalent con- structions, which we call, following Dixon 1982, the property concept (PC), is not an adjective. Francez and Koontz-Garboden (2015; FKG) examine constructions in several languages with nom- inal PC words, a key finding being that their morphosyntax is fundamentally different, invoking the morphosyntax of possession, as in Hausa, cp. (1) and (2).

(1) Mun_Źa we.CONT dŹa with Îarf¯ı. strength

‘We are strong.’ (Newman 2000:224)

(2) Y¯arinyŹa girl tan_Źa she.CONT dŹa with z¯obŹe. ring

‘The girl has a ring.’ (Newman 2000:222) FKG argue that possessive morphosyntax in these constructions is conditioned by the denotation of the PC noun. Whereas adjectives denote (contextually sensitive) sets of ordinary individuals, PC nouns denote substances, i.e., mass-type denotations (Link 2002), so that e.g.,Îarf¯ı in (1) denotes a mereologically ordered set of portions of strength. Such a meaning can be related to an entity by the possessive relation, but not by predication, at least with the intended meaning, as can be seen with the limited number of nominal PCs in English:

(3) _{Kim has strength6=Kim is strength (cf. Kim is strong)}

In the sample of languages FKG examine, PC nouns always trigger possessive morphosyntax in predication, suggesting a connection between nominality of PCs and mass-type denotation. Such a finding would offer intriguing insight into the semantic nature of lexical categoryhood, which is quite poorly understood, notwithstanding a great deal of study (see discussion in e.g., Baker 2003). In this paper we show, based on a detailed case study of Basaá (Bantu; Cameroon) that PC nouns do not uniformly have mass-type denotations; what Hyman et al. (2012) call nominal adjectives in Basaá are PC words that are nominal yet characterize sets of individuals rather than portions of substances. We begin by giving three arguments for the nominality of the relevant class. We then show that these PC words are individual, rather than substance characterizing. We conclude with discussion of consequences for the lexical semantics of PC words generally.

Three classes of PC words in Basaá PC words in Basaá fall into three groups, easily distin- guished by their behavior in modification. Group 1 subordinate the noun they modify via an agree- ing connective particle (4). These are the nominal adjectives of Hyman et al. (2012). Group 2 are subordinated to the noun they modify by the same connective particle (5). Group 3 directly modify nouns in a post-nominal attributive position, agreeing with the noun in noun class and number (6). We concur with Hyman et al. (2012) that the post-nominal PCs in (6) are true adjectives, given their ability to be adnominal modifiers with no connective particle. The question is about groups 1 and 2, which we argue are nouns.

(4) lì-múg´E 5-quiet !_lí 5.PRT hí-nùní 19-bird ‘quiet bird’ (5) hí-nùní 19-bird hí 19.PRT Ngùy 9.strength ‘strong bird’ (6) hì-nùní 19-bird hì-k´EN´ı 19-big ‘big bird’ Group 1 and 2 PCs are nominal We give three arguments for the nominal status of group 1 words, largely following Hyman et al. (2012:3ff.). Both group 1 and group 2 words have inherent noun class, something generally true only of nouns (Hyman 2003); different words in the classes are lexically specified with the full range of Basaa noun classes. Additionally, group 1 PCs control agreement on higher modifiers (Hyman et al. 2012:7d), like ordinary nouns:

(7) lì-múg´E 5-quiet !_lí 5.PRT dí-nùní 19-bird líní 5.this / *tíní 19.this ‘this quiet bird’

Finally, neither group 1 nor 2 words attributively modify nouns like the group 3 PCs in (6). Instead, both require a linker particle, as in (4) and (5). This all suggests that group 1 and 2 words are nouns. Group 1 are individual characterizing while group 2 are substance characterizing Our cen- tral claim is that group 1 words are a novel category—PC nominals that characterize sets of in- dividuals rather than sets of portions of substance (as with group 2). Our first argument relies on FKG’s observation that substance-denoting PCs trigger the use of possessive morphosyntax. While group 2 PCs in Basaa do trigger possessive morphosyntax, group 1 PCs do not. Instead, like adjectives and nominal predicates, they occur after the copula ye (8).

(8) à 1.AGR yè be lì-m´ug´E 5-quiet ‘(S)he is quiet.’ (9) à 1.AGR gw`E´E have mà-s´Od´a 6-strength ‘(S)he is strong.’

This indicates that like common nouns and adjectives, group 1 words are individual characterizing. On the other hand, the distribution of group 2 after have (9) indicates that these words have the substance-characterizing meaning described by FKG.

A second set of arguments illustrates a count/mass contrast between group 1 (count) and group 2 (mass), expected if the latter are substance characterizing and the former are not. While group 1 inflect for number (10), group 2 are invariant with respect to number (11):

(10) a. à 1(SG).AGR yè be lì-m´ug´E 5(SG)-quiet b. á´a 2(PL).AGR yè be mà-m´ug´E 6(PL)-quiet

‘(S)he is quiet.’ ‘They are quiet.’

(11) a. à 1.AGR gw`E´E have mà-s´Od´a 6(PL)-luck / *lì-s´Od´a 5(SG)-luck b. á´a 2.AGR gw`E´E have mà-s´Od´a 6(PL)-luck

‘(S)he is lucky.’ ‘They are lucky.’

The contrast reflects the behavior of count vs. mass nouns in Basaá: generally, only count nouns inflect for number. A second mass-count test involves numerals: while numerals can combine with group 1 words (ma-múg´E!_{má dí-nuní mátân’five quiet birds’), group 2 words cannot be modified}

by numerals (e.g., *ma-s´Od´a mátân ‘*five lucks’).

Lastly, the particle w`E´E can serve as an anaphor for group 2 (12) but not group 1 or group 3: (12) líhàt, w`E´E Paul

Rich, that’s Paul.

(13) #`nlám, w`E´E Paul Pretty, that’s Paul.

(14) #Nk´ENí, w`E´E Paul Important, that’s Paul.

We propose that w`E´E is a pro-substance anaphoric to mass denotations (group 2) but not individual- characterizing ones (groups 1 and 3), akin to the ability of demonstrative that but not pronouns to be anaphoric to substances in English (Partee 1986 a.o.).

To summarize, previous work on PC nouns suggests that nominality might entail a substance- type denotation. In this paper we show that Basaá has a class of PC words which are at once nominal and which denote in the domain of ordinary individuals. This shows that within the class of PC nouns there is a greater lexical semantic diversity than has been appreciated. It remains an open question whether there are similar kinds of variation in denotation of PC words when these are lexicalized as adjectives and verbs or whether these categories impose constraints on the kinds of denotations that such words can have, and why. We expect exploration of such questions to provide insight into the semantic nature of lexical categoryhood.

Stativity and progressive: The case of Japanese 'tokoro da' Magdalena and Stefan Kaufmann

The Japanese noun 'tokoro' (lit. ‘place’) has a grammaticalized variant with a temporal or modal inter- pretation (Takubo 2011). Syntactically, this variant is a “formal noun”: it requires modification by a tensed clause 'S' (or a suitable proform), and 'S-tokoro' behaves outwardly like a noun phrase, roughly meaning ‘time at which S’. Followed by the copula 'da' /'datta' (Non-Past/Past) it can be used as a ma- trix sentence; 'S-tokoro' can also be embedded as a temporal adjunct clause. In all this, it is similar to 'S-mae' (‘time before S’), 'S-ato' (‘time after S), and 'S-toki' (also ‘time when S’). But acceptability and interpretation of 'S-tokoro' display puzzling interactions with the temporal and aspectual properties of 'S' , which among other challenge the widespread conception of Progressives as a derived statives (Katz 2003). We focus on matrix uses (embedded ones being similar) and develop a unified analysis that treats Progressives as combining properties of eventives and statives and draws on similarities between temporal and modal orderings.

Data: First, if S is non-stative, then 'S tokoro' means ‘(just) before S’ or ‘(just) after S’, depending on the tense of S, cf. (1). What 'tokoro' in (1c,d) adds to the respective plain counterparts (1a,b) is the restriction to immediate future or past.

(1) a. Kaigi-ga meeting-NOM hazima-ru begin-NONPAST b. Kaigi-ga meeting-NOM hazimat-ta begin-PAST

‘The meeting will begin.’ ’The meeting began.’

c. (1a) tokoro tokoro {da COP-NONPAST / datta COP-PAST } d. (1b) tokoro tokoro {da COP-NONPAST / datta COP-PAST } ‘The meeting {is/was} about to begin.’ ’The meeting {has/had} just begun.’ Second, 'tokoro' is degraded with statives. The only interpretation available is a counterfactual one (Takubo 2011): (2) requires the accommodation of a counterfactual antecedent (e.g., ‘if I had grown up in the US’); as in full-fledged counterfactuals (Ogihara 2014), past tense is prefered.

(2) Eigo-ga English-NOM dekiru can-NONPAST tokoro TOKORO {???_da COP-NONPAST /?datta}. COP-PAST

‘I would be able to speak English.’

Third, the aspectual marker 'te-i-' allows for Progressive, Result Perfect, and Experiential Perfect readings (Igarashi & Gunji 1998), cf. (3a). With 'tokoro' (cf. (3b)), the Progressive interpretation is retained, whereas Perfect readings are marginal and involve counterfactuality (similar to lexical statives as in (2); Result vs. Experiental become hard to distinguish, the difference is ignored in the following).

(3) a. Zyon-ga John-NOM akai red zyaketto-wo jacket-ACC {ki-teiru

wear-TEI-NONPAST

/ ki-teita}. wear-TEI-PAST

Progressive: ‘John {is/was} putting on a red jacket.’ Result Perfect: ‘John {is/was} wearing a red jacket.’

Experiential Perfect: ‘John {has/had} once {put on/worn} a red jacket.’ b. Zyon-ga John-NOM akai red zyaketto-wo jacket-ACC ki-teiru

wear-TEI-NONPAST

tokoro TOKORO {da COP-NONPAST / datta} COP-PAST

Progressive: ‘John {is/was} putting on a red jacket.’ Perfect (marginal): ‘John would have worn a red jacket.’

Thus with 'tokoro' , the Perfect patterns with lexical statives (and habituals, not exemplified): no temporal reading is available.

The Progressive, in contrast, displays a hybrid pattern: [i.Stative-Like] (3b) expresses that the event is on-going, rather than just over or about to begin as obtained for non-statives with 'tokoro' ((1) above). In fact, the Progressive displays the same behavior as with other formal nouns, resulting in compatibility

1

only with 'toki' ‘when’. In this, it patterns with all lexical and other derived statives, cf. (4), and contrasts with non-statives (see Kaufmann & Miyachi 2011, henceforth KM):

(4) Zyon-ga John-NOM { ofuisu-ni office-LOC iru be-NONPAST / / hasit-tei-ru run-TEI-NONPAST

} { *mae *before / / *ato *after / / toki while } ‘while John {is/was} {in the office/running}’

[ii.Non-stative like] The Progressive reading of (3b) does not force a counterfactual interpretation, in contrast to the Perfect readings of the same sentence and to lexical statives (see (2)) or other derived statives.

Main idea: To capture the specific temporal readings with non-statives, the difference between 'te-i-' and (other) statives, and the counterfactuality effect with statives (other than progressives), we draw on an ordering relation and a notion of change in the immediate surroundings of the reference index. Intuitively, we argue that 'S-tokoro' requires that, right next to the reference time, there is a change in a contextually salient measure reflecting how much of the event described in 'S' is instantiated.

Analysis: We adopt a world-time (W × T ) model and let hI, <i be the set of non-empty intervals in T ordered by complete precedence (Landman 1991). Eventive predicates denote properties of events (which have intervals as temporal traces), stative predicates denote properties of intervals (Katz 1995).

Japanese tenses are relative, denoting binary relations between intervals (Ogihara 1996, KM). The relata i, i0intuitively correspond to the reference times of embedding and embedded clauses, or to speech and reference time in matrix contexts. In copular ' S-tokoro {da/datta}' sentences, i, i0are the reference