Linguistics

introduce what we call variable evidentiality. In (1), Korean -te-introduces the implication that the speaker perceived the event denoted by the prejacent (direct evidentiality) or the implication that the speaker infers the prejacent based on perceived evidence (inferential evidentiality), depending on the absence/presence of a tense. Similarly, (2) shows Bulgarian -l can be used when the speaker has inferential or hearsay evidence (inferential vs. hearsay evidentiality: Smirnova 2013, a.o.). (1) a. Pi-ka o-te-la. 'It is raining.' Rain-Nom come-te-Decl (~> I saw that it rained) b. Pi-ka o-ass-te-la. 'It rained.' (~> I infer that it rained from Rain-Nom come-Past-te-Decl perceived evidence such as the wet ground) (2) Maria svirela na piano. 'Maria plays the piano.' Maria Plays.Imperf.Pres.Ple on piano (~> I heard/I inferred) Relative tense? proposes that a tense under -te-should be evaluated not against the utterance time (UT) but against the time when the speaker learned the evidence (Learning Time, LT): when LT follows the time when the event of the prejacent occurred (Event Time, ET), the past tense is used, whereas when LT overlaps with ET, no overt tense is used. That is, the variability is due to the relation between ET and LT. Similarly, Koev (2017) argues that the indirectness is because -l requires the prejacent be spatiotemporally distant from the perceived evidence, and unlike -l, -te-allows the temporal and spatial overlap between the prejacent and the perceived evidence. However, with respect to the temporal relation, both Lee and Koev assume that a tense under an evidential expresses the relation between LT and ET, not UT and ET. Puzzles. However, there are several issues which cannot be explained by relative tense analyses. (i) Both -te-and -l require the direct perception-based inference: the speaker cannot obtain evidence based on pure reasoning. This is why Smirnova claims that -l requires externally experienced evidence, and we may be tempted to apply this analysis to -te-. (ii) However, while -l can introduce variable evidentiality between hearsay and inference, -te-cannot. For example, neither (1a) nor ( ) is felicitous under a scenario where the speaker only heard the raining event from someone else, without having any direct evidence. Finally, (iii) a more detailed examination shows that (1a) also involves inference: in (1a), the evidence should be directly perceived before UT; (1a) can be uttered only if the speaker cannot see the raining event anymore. However, in (1a) the speaker still makes commitment to the prejacent which holds in the present (see the translation). Without any overt past temporal adverb, if the speaker does not believe at UT that it is raining, (s)he cannot utter (1a), which means that the tense of the prajecent is evaluated with respect to UT, not with respect to LT (Hoe et al. 2020). Now, the following issues arise. In (1a), the speaker commits the prejacent just relying on an inference based on the past perceived evidence. Then, what is the nature of such inference? And how do two inferences in (1a) and (1b) differ from each other? Tense and the not-at-issue meaning of -te-. To account for these, let us assume the followings. First, we lexically specify that in -te-, LT should precede UT (which explains why sometimes -te-is classified as past: Chung 2007, a.o.). Notice that it is not a mere stipulation, since (1b) cannot be uttered when the speaker is watching the wet ground at UT even if the inference from the perception of the wet ground to the past event of raining arises properly: there should be a temporal restriction between LT and UT in using -te-(cf. Lee 2013; Koev 2017). Second,

This paper concerns a puzzle about the interpretation of Korean di- rect evidential -te- in interrogatives which I call evidential perspective shift in questions ('origo shift of evidentials' in quesitons in Garrett 2001). This puzzle consists in the evidential being interpreted as regarding a source of information of the answerer rather than the questioner. Here I introduce a novel semantic ap- proach to this phenomenon which derives it as a consequence of the interaction of the meaning of questions and of evidentials. I will also illustrate the concep- tual and empirical advantages of this view over previous accounts.

Korean direct perceptive evidential -te- is said 10 introduce two puzzles. First, when it is used without any tense marker, the sentence hosting it is interpreted as past. Second, when it is used with so-called tense markers, such as past -ess- and future -keyss-, -te- introduces the implication similar to what is introduced by inferential evidentials in other languages. In this paper I first show that this apparent puzzle is misleading, by presenting some counterexamples which do not fit to what the ``puzzle`` says. Then`` propose a new lexical entry for -te-, according to which -Ie- introduces the presupposition that the speaker acquired some direct perceptive evidence before the utterance time, and then infers the prejacent from this direct perceptive evidence r also show that this analysis accounts for the data in the previous puzzle as well as other data presented in this paper. Finally I point out several problems of previous analyses which arc based on this ``misguided`` puzzle.

Korean verbal ending -teintroduces different presuppositions depending on whether it appears with a tense marker (such as the past tense marker -ass/essor the future tense marker -keyss-) or not. When it appears without any tense marker, it introduces the presupposition that the speaker has direct perceptive evidence relative to the ‘prejacent’ (hereafter direct evidential presupposition), as in (1):

Korean adjectives in general cannot combine with measure phrases (MP), but MPs are compatible with adjectives when they appear with the inchoative morpheme -(e)ci. In this case, MPs can only denote the difference between two states along the dimension denoted by the root adjective. To account for this, this paper proposes that i) -(e)ci is a spell-out of V in the directed motion construction which takes an abstract path argument, like become, and ii) this path argument contains a comparative morpheme. By assuming this we can explain why MPs appear with -(e)ci, as well as other interesting phenomena such as variable telicity in deadjectival verbs with -(e)ci.

for their comments and questions. I am also indebted to the discussion with Semoon Hoe and Barry Schein. Special thanks go to Elena Guerzoni for her comments and suggestions to the earlier version of this paper. All remaining errors are mine, of course. This work is partly supported by the project Estructura eventiva y 'construcción' léxico-sintáctica de las oraciones. Teoría y experimentación (FF2009-07114).

In this paper we identify two types of the evidential perspective shifts (the evidence holder indicated by an evidential marker is different from the utterance of the sentence containing that evidential marker) in Korean, the shift in interrogatives and the shift in declaratives, and explain these two shifts by proposing specific lexical entries for Korean evidential markers without any significant revision of the existing theory of interrogatives or indexicals. We also discuss an implication of our proposal, especially with respect to the interaction between our lexical entry for the Korean direct evidential marker and the Korean long-distance anaphor caki.

This paper concerns a puzzle about the interpretation of Korean direct evidential -tein interrogatives which I call evidential perspective shift in questions ('origo shift of evidentials' in quesitons in Garrett 2001). This puzzle consists in the evidential being interpreted as regarding a source of information of the answerer rather than the questioner. Here I introduce a novel semantic approach to this phenomenon which derives it as a consequence of the interaction of the meaning of questions and of evidentials. I will also illustrate the conceptual and empirical advantages of this view over previous accounts. 2.1 Semantics of Indexicals: Kaplan (1989) theory of indexicals is intended to explain the difference between (2a) and (2b): 2 Higginbotham (2009) also pointed out this characteristics under the term First-person authority.

In Colloquial Dari -particularly the variety spoken in and around Kabul, Afghanistancomparatives obligatorily include a marker, kada, the past participle of kadan 'to do'. Tajik, a closely related language, also has a counterpart to kada -dida, the past participle of 'see'. Its distribution is similar, except that dida is optional:

Illusory "Escher" sentences (More people have been to Russia than I have) are a unique puzzle for theories assuming full and thorough grammatical analysis and semantic composition during processing, since people generally appear to accept them before noticing that the meaning is incoherent. Prior work has shown, however, that meaning is extracted from the illusion: in particular, reactions to the illusion are sensitive to whether a comparison of events is possible , suggesting that the perception of grammaticality is related to the fact that determiners may sometimes count events . The series of self-paced reading experiments reported here suggest the illusion is ambiguous between an individual and a (less salient) event quantification interpretation, since reading times are generally slower for comparatives with plural predicates. Reading times indicate that the meaning of the illusion is plausibly shifted to a comparison of events at the point of anomaly -a costly process that causes observable processing difficulty. The success of this operation determines how felicitous an interpretation comprehenders can obtain for an essentially ungrammatical sentence.

Semantic analyses often take degree words like more, fewer, too, enough and most to be quantifiers over degrees , drawing important parallels with quantification in the domain of individuals. Scope ambiguities -or lack thereof -constitute an important diagnostic for the quantificational properties of the degree word, including the presence of quantifier raising . Importantly, however, lack of scope ambiguity may suggest a quantificational treatment of the degree word is unjustified (Kennedy 1999), or it may implicate independent grammatical restrictions on the relevant scope configurations .

Alternative paths to linguistic annotation, such as those utilizing games or exploiting the web users, are becoming popular in recent times owing to their very high benefit-to-cost ratios. In this paper, however, we report a case study on POS annotation for Bangla and Hindi, where we observe that reliable linguistic annotation requires not only expert annotators, but also a great deal of supervision. For our hierarchical POS annotation scheme, we find that close supervision and training is necessary at every level of the hierarchy, or equivalently, complexity of the tagset. Nevertheless, an intelligent annotation tool can significantly accelerate the annotation process and increase the inter-annotator agreement for both expert and non-expert annotators. These findings lead us to believe that reliable annotation requiring deep linguistic knowledge (e.g., POS, chunking, Treebank, semantic role labeling) requires expertise and supervision. The focus, therefore, should be on design and development of appropriate annotation tools equipped with machine learning based predictive modules that can significantly boost the productivity of the annotators. 1