I. Information Structure and Scalar Implicatures This part of my talk shows how the information structure categories of Contrastive Topic (CT) and Contrastive Focus (CF) are correlated with PA (pero/aber) (haciman(K)/ga(J)). vs. SN (sondern/sino) (anira(K)/naku(J) conjunction respectively and also with DN (descriptive negation) vs. MN (metalinguistic negation) respectively when negative utterances are involved. More importantly it claims that the CT -- PA pattern underlies the phenomenon of scalar implicatures, whereas the CF -- SN pattern, which is MN or correction, typically `block' them or is irrelevant to them. Scalar implicatures are conventionalized by CT or conversationalized by covert CT and are thus argued that they cannot be adequately treated by `only' or its equivalent Exh(austivity) operator, as proposed by Fox (2006), by a hint from Sauerland (2005).

2. Negative Poalrity and Free Choice (in Korean and Japanese) Unlike in scalar implicatures generated by CT, -nun or -wa-marked, NPIs and FCIs are crucially dependent on the Concessive markers -to and -mo. These markers show that their complement clauses (to(mo)(p)) have the existential presupposition of alternatives and the conventional implicature of [likelihood (p) ≥ likelihood (q), where q is an alternative to p]. In other words, they are inherently scalar. (The same markers are also interpreted as additives with the same existential presupposition of alternatives.)

Monotone-decreasingness (Ladusaw 1979) and non-veridicality (Zwarts 1995) are nice function types to characterize the licensing contexts of NPIs and Free Choice Items (FCIs) but the former fails to account for weak NPIs and the latter for weakly negative predicates (turn off, remove, etc., see Joe and C. Lee 2001 J/K) and emotive factive predicates (lucky, etc.)(C. Lee 1999 UCLA Working Papers). Here we propose a unified solution in terms of concession. The majority of languages of the world such as Japanese, Korean, Chinese, Mongolian, Hindi, Zapotec and Basque form NPIs and FCIs by combining wh-based (otherwise, [any]-like) indefinites and concessives that denote the notion of concession, mostly equivalent to "even" in English (see Haspelmath 1993 [57 out of 100 languages are wh-based] and others for facts). In all languages, the lowest indefinite natural number "one" or a minimizer accompanied by a concessive also forms an NPI. This type is quantitative and can be explained by Horn (1972) and Fauconnier's (1975) scales, which, I claim, are triggered by concession. Going down to the lower bound for the easiest (or likeliest) on a contextually relevant scale of graded alternative quantities is making concession. If it is not the case with the relevant proposition even in that concessive minimized situation, which is the least likely, then, a higher or the maximized quantity does not hold either and the consequent emphatic total negation is what the speaker means to convey.

For the former wh-based type, which is qualitative, concession is made by arbitrary choice. However arbitrarily, property-wise, you may choose a member, up to maximization, from the wh-domain (the most arbitrary way is the easiest), if it is not the case with the relevant proposition, it is an NPI, and if it is the case in uncertain but modally maximally possible contexts, it is an FCI, but if it is the case in uncertain but modally existentially possible contexts, it is a weak NPI. A wh-question is a set of alternative answers as (true) propositions (Hamblin 1973 and Karttunen 1974) and an indefinite from it can stand for any (arbitrary) non-specific member of the same set (as in a choice function). I call the set of individuals etc. that correspond to the wh-information focus a wh-domain. The wh-based NPIs, however, are indefinite wh-forms with Concessives (CNC), not interrogative wh-words, contrary to claims made in the literature. This talk will show in Korean and Japanese how the notion of concession is central to understand polarity and how compositionally as well as intensionally polarity-related phenomena can be resolved.