A clausal analysis of free choice demo in Japanese

It is commonly assumed that in Japanese, an indeterminate pronoun followed by demo (indet-demo) corresponds to free choice any in English (FC any). Based on the observation that the two has various different properties, I argue that indet-demo is not a nominal free choice item, but a concealed unconditional adjunct, corroborating the claim made by Nakanishi and Hiraiwa (2019) and Hiraiwa and Nakanishi (2020, to appear). Extending Rawlins’s (2008, 2013) Hamblin analysis of unconditionals in English, I propose a compositional semantics of indet-demo that captures its semantics properties.

and Hiraiwa and Nakanishi (2020, to appear). Furthermore, based on Rawlins's (2008Rawlins's ( , 2013 analysis of unconditionals in English (e.g. Whatever Beth cooks, Al will be pleased), this article provides a compositional semantics of indet-demo where (2a) has the structure in (3) with two pros, one in the subject position in the unconditional clause and the other in the argument position in the main clause where nan-demo appears on the surface (see section 3.2 for details). (3) [pro nan(i) de ar-te mo] Al-wa pro tabe-teii what COP exist-COND MO Al-TOP eat-may 'Whatever (it) is, Al may eat (it).' The structure of the article is as follows. Section 2 presents empirical observations that indet-demo and FC any have various different properties. Section 3 establishes the claim that indetdemo is a concealed unconditional clause. Based on this claim, section 4 provides a compositional analysis of indet-demo. Section 5 shows that the proposed analysis can account for the properties of indet-demo presented in section 2. Section 6 examines the possibility of treating indet-demo as a free relative (e.g. Al will eat whatever Beth cooks). Section 7 concludes the article.

Differences between indet-demo and FC any.
FC any is known to have a restricted distribution (Horn 1972, Ladusaw 1979, Carlson 1981. In particular, it is acceptable in possibility sentences, as in (2b) above, but not in necessity and episodic sentences, as in (4b) and (5b), respectively. In contrast, indet-demo has no distributional restrictions; it is fine in possibility (2a), necessity (4a), and episodic sentences (5a).
Al-TOP what-DEMO eat-must b. *Al must eat anything. However, as exemplified in (6b) and (7b), FC any becomes acceptable in necessity and episodic sentences when an appropriate modifier is present (LeGrand's (1975) subtrigging effects) (Dayal 1998, 2004, 2009, Aloni 2007a. The corresponding examples of indet-demo are also acceptable, as shown in (6a) and (7a), but these sentences are acceptable even without modifiers, as we have seen in (4a) and (5a) above. 4 (6) a. Al-wa Beth-ga tuku-ru mono-o nan-demo tabe-nebanaranai. Al-TOP Beth-NOM make-PRES thing-ACC what-DEMO eat-must b. Al must eat anything that Beth cooks.
a. Al-wa Beth-ga tuku-tta mono-o nan-demo tabe-ta. Al-TOP Beth-NOM make-PAST thing-ACC what-DEMO eat-PAST b. Al ate anything that Beth cooked.
Another difference comes from the availability of non-iterative readings. Dayal (1998Dayal ( , 2004 observes that any favors iterative readings, as in (8a), but iterativity is not a necessary condition, as in (8b) (see also Jayez and Tovena 2005). In contrast, indet-demo requires iterative readings, as in (9), which corresponds to (8).
(8) a. Mary sang for anyone who wanted to hear her. (Dayal 1998: 465) b. Anybody who was there at that time died in the blast. (Dayal 2004: 16) (9) a. Mary-wa uta-o kiki-tagat-ta dare-no tame ni-demo uta-ta. Mary-TOP song-ACC hear-want-PAST who-for-DEMO sing-PAST b.??Sono toki sokoni i-ta hito-wa dare-demo bakuhatu-de sin-da. That time there be-PAST person-TOP who-DEMO blast-by die-PAST There is yet another difference between the two. In imperatives, FC any can be universal, as in (10a), or it can be existential in some context, as in (10b) (Horn 1972, LeGrand 1974, Carlson 1981, Dayal 1998, Giannakidou 2001, Chierchia 2006, Aloni 2007b. In contrast, indet-demo only permits universal readings; while (11a) has the same universal reading as the corresponding English example, (11b) is odd under an intended existential reading. However, (11b) can be used to express a rather unnatural reading where the addressee is requested to press keys indifferently.
*$ which key-DEMO press-IMP The differences presented so far suggest that indet-demo cannot be a simple counterpart of FC any. Then the questions arise as to what it is and what accounts for its properties, which I turn next.
3. Indet-demo as a concealed unconditional clause. In order to address , I establish the claim that indet-demo is a concealed unconditional clause, building on the proposal made by Nakanishi and Hiraiwa (2019) and Hiraiwa and Nakanishi (2020, to Rawlins's (2008Rawlins's ( , 2013 analysis of English unconditionals such as (12a), H&N present an analysis of Japanese unconditionals such as (12b), and further claim that their analysis of rum-of-the-mill unconditionals extends to indetdemo.
(12) a. Whatever Beth cooks, Al will be pleased.
b. Beth-ga nani-o tukut-te mo Al-wa yorokobu-daroo. Beth-NOM what-ACC make-COND MO Al-TOP please-will In particular, H&N propose that indet-demo like nan-demo is an unconditional clause with the structure in (13), which is composed of a copula verb de, an existential verb ar, and the particle mo. The existential ar undergoes ellipsis, yielding a seemingly nominal phrase nan-demo.
As a piece of evidence for treating indet-demo as a clause and not as a nominal element, H&N show that indet-demo in (14b) cannot be case-marked, just like the corresponding fullfledged clause in (14a). I here add another piece of evidence for H&N's claim that indet-demo is an unconditional adjunct. Regarding standard unconditionals such as (12b), H&N demonstrate that there are variations depending on whether the particle -mo is obligatory, prohibited, or optional, as exemplified in (15a). 5 The same variations can be found with indet-demo, as shown in (15b), which is straightforwardly explained if indet-demo is an unconditional adjunct.
3.2. REFINING THE STRUCTURE OF INDET-DEMO. Based on H&N's structure in (13), I propose the structure in (16a) for (16b) ( [ ] corresponds to the unconditional clause nan-demo). 6 There are two pros in the structure, one is in the subject position in the unconditional clause and the other in the argument position in the main clause where nan-demo appears on the surface (i.e., the object position in (16b)). 7 (16) a. [pro nan(i) de ar-te mo] Al-wa pro tabe-teii what COP exist-COND MO Al-TOP eat-may 5 The paradigm in (15a) is not exhaustive. See Hiraiwa and Nakanishi (2020), among others, for a complete list. 6 The surface word order in (16b) differs from the underlying order in (16a). I assume that this order obtains as a result of scrambling the subject of the main clause over the unconditional clause. The following example, which corresponds to (12b), shows that such scrambling is freely available with regular unconditionals.
(i) Al-wai [Beth-ga nani-o tukut-te mo] ti yorokobu-daroo. Al-TOP Beth-NOM what-ACC make-COND MO please-will 'Whatever Beth cooks, Al will be pleased.' 7 Indet-demo does not need to appear in the argument position, as exemplified in (i). In such a case, I remain agnostic as to whether there is a pro in the main clause (because, unlike in (16), its presence is not required by the verb). (i) Al-wa doko-demo nemur-eru.
Al-TOP where-DEMO sleep-can 'Al can sleep anywhere.' (ii) [pro doko de ar-te mo] Al-wa (pro) nemur-eru. where COP exist-COND MO Al-TOP sleep-can 'Wherever (it) is, Al can sleep (there).' In the rest of the article, I only discuss cases like (16b) where indet-demo seemingly serves as an argument of a simplex sentence, and simply assume that the analysis of such cases extends to cases like (i) regardless of whether we posit pro in the main clause.
'Whatever (it) is, Al may eat (it).' Under this analysis, nan-demo in (16b) is not the internal argument of the verb, but it is an unconditional adjunct. That is, although on the surface, (16b) looks like a simplex sentence where nan-demo serves as an argument, it is actually a complex sentence where nan-demo is analyzed as an unconditional clause. The proposed analysis is corroborated by the observation in (17) that indet-demo cannot take any nominal modifiers such as adjectives and relative clauses. Further supporting evidence for the claim that indet-demo is not an argument comes from (18a), which demonstrates that indet-demo can co-occur with an overt argument of the verb in the main clause. (18b) shows that the nominal argument can take the relative clause, unlike (17b).
Under the proposed analysis, nan-demo in (18) is an unconditional clause that is adjoined to the full-fledged sentence Al may eat the dish (that Beth cooks), as in (19). This shows that pro in the main clause in (16a) can be overtly realized.
The current analysis correctly predicts that nan-demo in (18), being an adjunct, is omissible, as illustrated in (20). Put differently, nan-demo in (18) adds to (20) a sense of indifference.
Al-TOP Beth-NOM make-PRES dish-ACC eat-may 'Al may eat the dish (that Beth cooks).' Furthermore, pro in the subject position of the unconditional adjunct in (16a) can also be realized on the surface. This is shown with (21a), which is minimally different from (18a) in that Beth-ga tuku-ru ryoori 'the dish that Beth cooks' takes a nominative marker in (21a), while in (18a) it takes an accusative marker. In the former case, the nominative marked element is considered to be the subject of the unconditional adjunct, as in (21b), yielding the reading 'whatever the dish that Beth cooks is'. Note that in (21a), in the main clause can also be overtly realized (e.g. by the pronoun 'it'). It follows from here that there are indet-sentences without s. 9 Put differently, having s is not a necessary condition for legitimate indet-sentences.
From a semantic perspective, if the unconditional analysis of indet-is on the right track, sentences with indet-such as (16b) and (18a) are expected to have properties that are characteristics of unconditionals. Rawlins (2008Rawlins ( , 2013 observes that unconditionals in English express an implication of relational indifference. For example, (12a) expresses that, relative to whether Al will be pleased, it doesn't matter what Beth cooks. He also claims that the main clause (or the consequent of the unconditional) is entailed by an unconditional adjunct. For instance, (12a) entails that Al will be pleased. I submit that the sentence with indet-have both of these properties. For example, (18b) expresses that, relative to whether Al may eat the dish that Beth cooks, it doesn't matter what she cooks, and also it entails that Al may eat the dish that Beth cooks.
Rawlins argues that these two properties, namely, relational indifference and consequent entailment, are derived compositionally from the interpretation of unconditionals (see section 4 for details). As indet-has the same characteristic properties of unconditionals, in the following, I extend Rawlins's analysis to indet-.
. Rawlins (2008Rawlins ( , 2013) presents a Hamblin (1973) analysis of unconditionals based on the insight that unconditionals are a conjunction of conditionals. A simplified LF structure of (12a) (= ) under his analysis is provided in (22). 9 Examples such as (21a) may be taken as a piece of evidence against the view that indet-is analyzed on a par with correlatives in languages like Hindi. It has been observed that correlatives require the presence of a proform in the main clause and that there is some restriction on what kind of proform is permitted (Dayal 1996, among others). In (16a), I proposed that the adjoined unconditional indet-requires the presence of in the main clause. However, (21a) shows that the proform in the main clause is not restricted to ; it can be realized by a pronoun like 'it'. Note furthermore that can be replaced with other expressions such as 'that food'. The lack of restriction on a proform in the main clause makes indet-look different from correlatives. 10 As noted in footnote 3, in may have either the LH or HL pitch accent pattern, whereas in (21a) must have the HL pattern. Note that when is followed by a copula, as in (21b), the indeterminate necessarily has the HL pattern. Thus, we may take the HL pattern in (21a) as a piece of evidence for the underlying clausal structure in (21b). However, this would wrongly predict that -should always have the HL pattern: in the proposed analysis, is derived from the clausal structure in (16a), which is like (21b) except that the subject in the unconditional clause remains covert. I here hypothesize that the LH pattern reflects a grammaticized lexical entry of indet-. That is, indet-may undergo a grammaticalization process and develop into a (seemingly) nominal lexical item. Presumably, the process is easier to apply to indet-without an overt subject (= (16a)) than to the one with the subject (= (21b)). This explains why indet-is often considered to be a nominal FC item, as discussed in section 1.

(22) [ ∀ [ [ Q [CP whatever Beth cooks ] ] [IP All will be pleased ] ] ]
In a Hamblin semantics, all expressions denote sets. While most lexical items denote singleton sets whose sole member is their usual denotations, as in (23a) and (23b), items denote sets of individual alternatives, as in (23c). 11 The alternatives introduced by indeterminates combine with the denotations of other lexical items via pointwise functional application. 12 In our case at hand, the denotation of is the set of propositions {that Beth cooks meat, that Beth cooks fish, ...}, as in (23d) is an interrogative clause, and that a question operator is syntactically present within the unconditional clause, as in (22). His definition of the question operator is given in (24), where csc stands for the context set provided by the input context of internpreation c (Rawlins 2013: 138). Truth-conditionally, the operator simply lets alternatives through, but it introduces exhaustivity and mutual exclusivity presuppositions.
When (23d) combines with (24), the set of propositions {that Beth cooks meat, that Beth cooks fish, ...} gets augmented with the two presuppositions, which ensures that every world in the context set involves exactly one alternative and that there are no worlds where multiple alternatives hold. Rawlins further argues that each alternative in the unconditional adjunct provides a domain restriction to the modal in the main clause, which yields a set of conditionals whose antecedent and consequent are the alternative in the unconditional adjunct and the main clause, respectively. In our case, we obtain a set of conditionals {if Beth cooks meat Al will be pleased, if Beth cooks fish Al will be pleased, ...}. This cannot be the final denotation of (12a), however, because in Hamblin's system, a declarative sentence must denote a singleton set. Rawlins suggests that a default universal operator in (25) is inserted in the LF of an unconditional, as in (22) (see also Menéndez-Benito 2010). As a result, (12a) denotes a singleton whose sole member is the conjunction of conditionals {if Beth cooks meat Al will be pleased, and if Beth cooks fish Al will be pleased, and ...}.  (Kratzer and Shimoyama 2002) In this analysis, unconditional adjuncts provide a jointly exhaustive set of restrictions to a modal operator in the main clause. This amounts to the relational indifference implication; we try out all domain restrictions for a modal, and find that the modal claim is always true. Consequent entailment also follows from this semantics; under the derived truth conditions, Al will be pleased under any condition.
Turning now to Japanese, H&N provide syntactic arguments that Japanese unconditionals such as (12b) have interrogative adjuncts just like English unconditionals. 13 I submit that indetsuch as (26b), repeated from above, is also an interrogative clause, assuming that it has the same syntactic structure as typical unconditional adjuncts, as proposed in (26a). As for the semantics of indeterminates, it is generally assumed that indeterminates denote a set of individuals just like items in English (Shimoyama 2001(Shimoyama , 2006 (18) and (21)), one as the subject of the unconditional adjunct, and the other as the argument in place of where indet-appears. English unconditionals may have these s in the form of overt pronouns, but not always so. (27a), repeated from earlier, is a typical example of unconditionals without any pronouns, and (27b) is an unconditional with a pronoun in the argument position in the main clause. 14 (27b) can be further paraphrased to (27c), in which case the relative clause is omissible when it is recoverable from the context. (27c) is reminiscent to the indet-sentence in (26a).
(27) a. Whatever Beth cooks, Al will be pleased. (= (12a)) b. Whatever Beth cooks, Al may eat it. c. Whatever it is (that Beth cooks), Al may eat it.
The question then is how to analyze s (or overt pronouns in English). As Hirsch (2016) points out, the pronoun in the main clause in (27b) has the interpretation that co-varies with that of the unconditional adjunct , which leads to the hypothesis that in (27b) is an individual variable bound by . However, Hirsch correctly claims that this hypothesis is incompatible with Rawlins's analysis where the unconditional adjunct is a set of propositions, and thus the adjunct cannot serve as a binder of an individual variable. To resolve this problem, Hirsch argues that in (27b) is an E-type pronoun with the denotation in (28), where the world variable is bound by the modal in the main clause. 15 13 There are two pieces of evidence for the interrogative approach. First, the sentential disjunctive connective 'or', which can disjoin questions but not declaratives, can disjoin unconditional clauses. Second, the indeterminate 'why', which only has a question or an existential construal, is also licit in unconditionals. See Nakanishi and Hiraiwa (2019) for details. 14 As Caponigro and Fălăuş (2018: footnote 7) notes, little attention has been paid to unconditionals like (27b) (except for Hirsch (2016), which will be discussed shortly below). 15 Hirsch's (2016) ultimate goal is to unify the semantics of -free relatives like (i) with that of unconditionals like (27b). See section 6 below for a comparison between unconditionals and free relatives. (i) Al may eat whatever Beth cooks. (29) [[ ]] w,g = ix. P(x)(w), where P is a contextually salient property In (21a), which corresponds to the English unconditional in (27b), anaphorically refers to the dish that Beth cooks, which yields the same denotation as (28).
I further propose that in the subject position of the unconditional adjunct is also an Etype pronoun with the same denotation. (18a) may be viewed as a case of cataphora where in the unconditional adjunct refers to the dish that Beth cooks in the main clause, again yielding the same denotation as (28). When there is no overt cue for the reference of s, as in (26b), in the subject position of the unconditional adjunct may refer to anything that is salient in the context (e.g., the dish that Beth cooks, the food served for dinner, the leftovers in the fridge, etc.). In this case, it is natural to assume that in the main clause anaphorically refers to the same thing as in the unconditional adjunct, just like in (21a). We are now ready to apply Rawlins's Hamblin analysis to indet-sentences. Suppose that in (26b), refers to the thing that Beth cooks. The denotation of the indeterminate is a set of individuals in (30a), and it combines with the denotation of the copula and with that of . As a result, we obtain a set of propositions in (30b). When the set combines with the question operator defined in (24) above, exhaustivity and mutual exclusivity presuppositions are introduced; each proposition in (30c) is interpreted exhaustively (which is indicated by ), and the propositions in this set are mutually exclusive. Each proposition in (30c) provides a domain restriction to the modal in the main clause. Putting the details aside, the denotation of the entire unconditional construction is provided in (30d), where each alternative has a conditional paraphrase. In this analysis, the referent of the E-type pronoun varies between the conditionals. Finally, the universal operator in (25) combines with the set in (30d), which yields a singleton set that contains the conjunction of the propositions in (30d), as in (30e). if the thing that Beth cooks in w' is only fish in w', Al will eat the fish in w', and ... } . Let us now examine whether the proposed analysis is capable of explaining the semantic properties of indet-discussed in section 2. First, we have seen that indet-is not subject to any distributional restrictions. In the current approach, a sentence with indet-is analyzed as an unconditional construction where indetserves as an unconditional adjunct. Then the question is whether the main clause of unconditional constructions in general can be a possibility, a necessity, or an episodic sentence. In English, any of these sentences can be the main clause of unconditionals, as exemplified in (31). 16 The same claim holds for run-of-the-mill unconditionals in Japanese, as illustrated in (32). Thus, the proposed analysis correctly predicts that indet-is not sensitive to the distributional restrictions.
(31) a. Whether he is sick or not, Alfonso can stay home from school. (Rawlins 2013: 126) b. Whatever Alfonso has, he should stay home. ( : 146) c. Last week, whoever Alfonso talked to, he got bad advice. ( (32) Dare-ga ki-te mo Al-wa (sono hito-ni) {at-teii / awa-nebanaranai /at-ta}. who-NOM come-COND MO Al-TOP that person-DAT {meet-may/meet-must/meet-PAST} 'Whoever comes/came, Al {may meet / must meet / met} (that person).' Second, indet-does not require the presence of subtrigging. As discussed in section 3.2, indet-, being a clause, does not take a modifier (see (17) above), and thus there is no genuine example of subtrigging that directly modifies indet-. However, indet-can co-occur with expressions that help us identify what s refer to (i.e., the subject of the adjunct and the argument of the main clause in place of indet-). In (33), is an unconditional adjunct, and 'the thing that Beth cooked' serves as the internal argument of the main clause, the existence of which makes the reference of the E-type pronouns explicit. As we have seen in section 4, the resolution of the E-type pronouns is possible without the presence of such overt expressions (i.e., (33) without the bracketed expression), which account for why indet-is licit without (seeming) subtrigging.
(33) Al-wa (Beth-ga tuku-tta mono-o) tabe-ta. (= (7a)) Al-TOP Beth-NOM make-PAST thing-ACC what-DEMO eat-PAST 'Whatever (it) is, Al ate (the thing that Beth cooked).' The third property is that indet-requires an iterative interpretation, as in (34a) (repeated from earlier). The denotation of this sentence is provided in (34b). This suggests that the blast killed a single individual each time, but such an interpretation is odd, accounting for why (34a) is infelicitous.
(34) a.??Sono toki sokoni i-ta hito-wa bakuhatu-de sin-da. That time there be-PAST person-TOP who-DEMO blast-by die-PAST 'Whoever the person who was there at that time was, he died in the blast.' (= (9b)) b. {if the person who was there was Al in w', Al died in the blast in w', and if the person who was there was Beth in w', Beth died in the blast in w', and ... } Finally, we saw that in imperatives such as (35a), indet-does not have an existential interpretation. However, (35a) has an indifference reading (relative to pressing a key, it doesn't matter which key you choose). A plausible referent of the s in this sentence is the key that you choose, in which case the denotation of (35a) is (35b).
kii-osite-kudasai. *$ (= (11b)) which key-DEMO press-please 'Whichever key (it) is, please press (it).' b. {if the key that you choose is A in w', please press that key in w', and if the key that you choose is B in w', please press that key in w', and ... } As mentioned in section 3.2, one of the indispensable features of unconditionals that Rawlins's analysis successfully derives is relational indifference. Thus, the fact that (35a) has an indifference reading is taken as a piece of evidence for the claim that (35a) is an unconditional construction. I propose here that the intended existential reading obtains when '(lit.) good because' is added after indet-, as in (36a) (cf. Watanabe 2013). In this case, 'good' by itself functions as the main clause of the unconditional, which is interpreted as 'no problem' or 'it doesn't matter'. The subordinating conjunction 'because' lets us form two independent sentences, namely, the unconditional 'whichever key (it) is, no problem' and the imperative 'please press (it)'. By using the unconditional, the speaker reassures the addressee that the choice of a key to press makes no difference. This intuition is consistent with the denotation given in (36b). The unconditional is followed by the imperative which simply requests the addressee to press the key that he chooses, which amounts to an existential interpretation of the internal argument of the verb. 17 (36) a.
kii-ii kara osite-kudasai. which key-DEMO good because press-please 'Whichever key (it) is, no problem, so please press (it).' b. {if the key that you choose is A in w', no problem in w', and if the key that you choose is B in w', no problem in w', and ... } . Before concluding the article, I consider a possibility of analyzing indet-as a free relative with (-FR). It has been proposed that unconditional adjuncts such as (37a) may be analyzed as -FRs such as (37b) (Dayal 1997, Izvorski 2000a. The relevance of the two is apparent at least in English since the same morphology -appears in the two. (37) a. Whatever Beth cooks, Al may eat it. b. Al may eat whatever Beth cooks.
Rawlins presents several syntactic arguments to show that unconditionals have the structure of an interrogative, not the structure of an FR (see Rawlins 2013: section 3.1). He also discusses semantic differences between the two. In particular, he compares the nature of "indifference" in the two constructions. It has been argued that FRs may express indifference (FR-indifference, adopting Rawlins's term) (von Fintel 2000, among others). For example, (38a) means that Zack voted for the person at the top of the ballot indiscriminately. Von Fintel (2000) argues that FR-indifference is counterfactual, as exemplified by the paraphrase in (38b).
(38) a. Zack simply voted for whoever was at the top of the ballot.
(  (Jennings 1994, Horn 2000, 2005, Giannakidou 2001, Dayal 2004, 2009; is interpreted universally, but it is separated from the imperative. (i) Press a key, (it could be) any key. Similarly, we may consider indet + in (36a) as a supplementary unconditional; the universal-like interpretation of the unconditional is conjoined with the imperative. But this makes indet-look similar to FC , which brings us to the question of whether it is possible to unify the two (see section 7 on this point).
As briefly discussed in section 3.2, Rawlins claims that unconditionals obligatorily express an implication of relational indifference. Crucially, he presents four descriptive semantic differences between relational indifference and FR-indifference. One of the differences is that while FR-indifference is counterfactual, as shown in (38), relational indifference is not. This is illustrated with (39), which permits the continuation in (40).
(39) Context: Alfonso talked to Joanna, Henry, and Isabella about his problem. We know this.
üLast week, whoever Alfonso talked to, he got bad advice. (Rawlins 2013: 115) (40) üBut he should have talked to Charles, he would have gotten good advice. ( Turning now to Japanese, it is not obvious whether Japanese has FRs in the first place. The answer may vary depending on the definition of FRs. If FRs require the presence of a item as a head, Japanese does not have relative clauses headed by a item, hence the lack of FRs. However, if FRs are defined as headless relative clauses, Japanese does have such a configuration, which suggests the existence of FRs. Even if we abstract away from this issue and assume that indet-may be analyzed as FRs, there is still a reason to believe that indet-is semantically different from FRs with respect to an implication of indifference. Just like unconditionals in English, indet-expresses an implication of relational indifference that is not counterfactual. The indet-sentence in (41) can be followed by (42), which is unexpected if relational indifference is counterfactual.
(41) Context: Al attended the dinner, where chicken, lamb, and salmon were served. We know this. üYuusyoku-de Al-wa dasa-re-ta mono-o tabe-ta. dinner-at Al-TOP serve-PASS-PAST thing-ACC What-DEMO eat-PAST 'Whatever (it) was, Al ate the thing that was served at the dinner.' (42) üBut if shrimp had been served, he wouldn't have eaten it. He is allergic to shellfish.
Rawlins points out that examples such as (39) are important in another respect. As discussed in section 4, in Rawlins's analysis, the relational indifference implication is derived indirectly from how each component is put together in a Hamblin semantics. This approach makes a prediction that the implication disappears when conditionals act 'non-modally' (in what he calls "material unconditionals"). This is what we observe in (39), where the context limits the relevant worlds to those compatible with the stated facts in the past. It follows that (39) makes no predictions as to what happens outside of the worlds where Alfonso didn't talk to Joanna, Henry, or Isabella. The same claim holds for (41); the indet-sentence does not imply that Al would eat what is served outside of the worlds compatible with the facts. I thus submit that the observation here further corroborates the claim that indet-is an unconditional adjunct.
In this paper, I argued that indet-, which is generally viewed as a nominal FC expression, should be analyzed as an unconditional adjunct. This analysis captures properties of indet-that are different from those of FC expressions such as English . Crucially, I am not suggesting that unconditionals and FC expressions are completely different creatures; it is intuitively obvious that they have some fundamental properties in common. Indeed, Rawlins suggests that both unconditionals and FC effects can be meta-characterized using the notion of "orthogonality" (see Rawlins 2013: section 4.1 for details). Taking up Rawlins's insight, Szabolcsi (2019) examines Hungarian expressions that can serve as unconditional adjuncts, FC items, and negative polarity items, and proposes a unified account of