2.1 Alternatives and exhaustification

Chierchia (to appear a) extensively argues that the properties of EIs are best understood within a more general theory of polarity and free choice phenomena, within an alternative-based semantics. The core hypothesis underlying this strongly unitary program is that all kinds of polarity sensitive items (e.g. NPIs and FCIs like those in (3) above, or the EIs under consideration here) are (weak) indefinites (i.e., existentially quantified elements) that activate sets of alternatives. Once they are active, alternatives must be factored into meaning (just like alternatives activated through focus). One way to do this is via the insertion of covert exhaustification operators, whose semantics is roughly akin to only and even. Exhaustification through a silent only has been argued to be at work when calculating scalar implicatures (cf. Chierchia, Fox, and SpectorReference Chierchia, Fox, Spector, Maienborn, von Heusinger and Portner2012; Chierchia to appear b), but there is evidence that such covert operators can be found in a wide variety of instances:

In (4a), we normally interpret B’s answer as conveying that the speaker only saw Paul and Sue, i.e., he did not see any other individual(s) that might have also been contextually relevant. The positive answer in (4b) indicates that such an exhaustive meaning can also come about via an even-like operator, whereby we get the interpretation that B’s ex came to the party, and she was the least likely person to do that (plus an additional inference that someone else was there too, an additive component which is not relevant here). The semantics of these alternative-sensitive operators is as in (5) below:

O and E are binary operators that combine with a set of propositional alternatives (here represented by the subscript C) and a proposition p (the prejacent). According to the definitions in (5), O says that that p (and its entailments) is the only true member of the set of alternatives C, and the application of E conveys that p is the least likely among the relevant set of alternatives C. This mechanism is assumed to enrich meaning whenever alternatives are activated, regardless of whether they come about through focus, contextual factors, or are lexically determined. To understand the alternative-based treatment of polarity sensitive items, let us briefly go through an example:

The assertion in (6a) is identical to what we would have with a plain indefinitea class, i.e., an existential statement, made with respect to a contextually relevant domain of quantification D. For simplicity, assume D is a set consisting of only two elements, a semantics class and a pragmatics class. The presence of the polarity sensitive elementany leads to the consideration of subdomain alternatives, i.e., subsets of the relevant contextual domain. Active alternatives compose via pointwise functional application, so at the sentential level we have alternatives that look as in (7b):

The alternatives introduced by the NPI must be exhaustified, i.e., require the presence of an appropriate alternative-sensitive operator in the structure, O in (7c). As defined in (5a) above, the application of the operator O says that the assertion is true and all non-entailed alternatives (in C) are false. Therefore, once we apply O to the assertion in (7a), we get a contradictory meaning: I teach a class chosen from the set {semantics, pragmatics}, but I do not teach a semantics class and I do not teach a pragmatics class. Whence its deviance.² Things are different in a negative statement like (6b), where the presence of negation (or any other downward-entailing operator) renders the exhaustified meaning non-contradictory. Since all alternatives are entailed by the assertion, there are no stronger alternatives, and therefore nothing gets eliminated via exhaustification (7d). Put another way, exhaustification in this configuration is vacuous, meaning it simply amounts to the initial assertion I do not teach a class in D.

Let us now turn to EIs, which in this framework are characterized as alternative bearing existentials, on a par with the other members of the polarity system. To understand how their interpretation comes about, it is best to consider the following set of modal sentences:

The disjunction in (8) and the EI un quelconque in (9) – both existential elements under the scope of an existential modal – end up having a conjunctive interpretation, where each of the relevant alternatives (guitar or violin in (8) or any musical instrument in (9)) constitutes a legitimate choice. On the present alternative-based account, the observed free choice effects uniformly arise via recursive application of O, the exhaustification operator defined in (5a).³ More specifically, it is assumed that the alternatives activated in both cases are identical, and come in two varieties: scalar alternatives (sc-alt), which are stronger alternatives on a numerical scale, two, three, etc. (10b), and subdomain alternatives (d-alt), which consist of subdomains of the relevant set of musical instruments (10c).

To emphasize the parallel with disjunction, let us assume that the EI in (9) existentially quantifies over a domain with only two elements, i.e. guitar and violin. If D = {guitar, violin}, then the assertion and the alternatives can be represented as follows:⁴

Exhaustification with respect to this set of alternatives leads to a contradiction (e.g., SauerlandReference Sauerland2004). The solution, extensively discussed in Fox (Reference Fox, Sauerland and Stateva2007a) and Chierchia (to appear b) to whom I refer for details, is to allow for recursive exhaustification, i.e., the consideration of ‘pre-exhaustified’ domain alternatives.⁵ Roughly speaking, exhaustified domain alternatives (exh d-alt) are obtained by applying O to each one of the alternatives in (11c), as shown in (12a,b). For an alternative like “you may study the guitar,” the exhaustified meaning is equivalent to “you may only study the guitar,” i.e., “you may study the guitar and you may not study the violin” (12a), and similarly for the other active alternative (12b).

Exhaustification relative to this set leads to the enriched meaning in (13):

Once we put together the assertion with the derived implicatures, the enriched meaning of the sentence in (9) expresses that every musical instrument in D (i.e. both the guitar and the violin) is a possible option. The assumption that EIs activate scalar and (pre-exhaustified) domain alternatives, just like disjunction or other scalar terms, thus derives the free choice inference observed in (9).

These considerations suffice to illustrate the alternative-based analysis of EIs adopted in this chapter. The account just sketched relies on obligatory activation of alternatives, a property EIs share with polarity sensitive elements. The idea that polarity or free choice indefinites activate alternatives of some kind is not new, and various aspects of the theory defended by Chierchia have been made in previous literature (cf. Kadmon and LandmanReference Kadmon and Landman1993; KrifkaReference Krifka1995; LahiriReference Lahiri1998; Kratzer and ShimoyamaReference Kratzer, Shimoyama and Otsu2002; Aloni and van RooijReference Aloni, van Rooij, Bouma, Krämer and Zwarts2004, among others). The novelty lies in the execution of this proposal in a framework which deals uniformly with ordinary scalar implicatures (cf. Chierchia, Fox, and SpectorReference Chierchia, Fox, Spector, Maienborn, von Heusinger and Portner2012 and references therein) and a wide variety of polarity sensitive phenomena. The result is a conceptually unitary system, with a broad empirical coverage, where the interplay between alternatives and exhaustification generates the various patterns of polarity attested across languages. On this view, the main difference between scalar terms (e.g. disjunction, simple indefinites) and polarity sensitive indefinites is that alternatives associated with the former are only optionally active, whereas those associated with the latter are obligatorily active. As briefly illustrated for NPIs and EIs above, the narrower distribution of polarity sensitive items falls out from their conjectured lexical semantics (active alternatives and constraints on exhaustification) and the way it interacts with the operators in their context of occurrence (e.g. downward-entailing or modal operators). Variation in the polarity system is captured through a restricted number of parametrical choices, along two interconnected dimensions: (i) types of active alternatives and (ii) modes of exhaustification. A closer look at variation among EIs offers the opportunity to illustrate both.

2.3 Variation among EIs: NPI use

A second parameter of variation among EIs, which will turn out to be relevant for the possibility to associate with focus, concerns the use as a negative polarity item. Some EIs can be freely used in downward-entailing contexts, where they behave as regular NPIs and have narrow scope existential readings:

The disappearance of the scalar and the free choice implicature in downward-entailing contexts is fully expected on the present approach. To see this in slightly more detail, assume that the domain relevant for the interpretation of irgendein in (22a) contains three books {a,b,c}. The assertion to which exhaustification applies looks as in (23a) (assuming for simplicity that niemand translates as sentential negation), and the alternatives as in (23b):

It is clear that both the scalar alternatives and the negation of the exhaustified domain alternatives are entailed by the assertion.⁷ As a result, no scalar or free choice inference arises and the EI is interpreted as an existential, i.e., it acquires an NPI-reading.

In contrast to this, some other EIs do not have NPI uses: if they occur under a downward-entailing operator, they can only acquire a “not just any” reading (in the presence of focus or contextual bias), illustrated in (24b) and (25b):

On the present account, this difference among EIs is captured by assuming a restriction on exhaustification. More precisely, certain EIs (and more generally free choice indefinites) require that the result of exhaustification be properly stronger than the assertion. This requirement cannot be satisfied in downward-entailing contexts, where the assertion entails all alternatives, and hence, exhaustification simply returns the assertion. Formally, this is cashed out by defining O_PS, a presuppositional operator selected by certain EIs:

O_PS functions like the usual exhaustification operator O (i.e., eliminates stronger alternatives), but is defined only if the result of exhaustification properly entails the assertion. In modal contexts, the derived free choice interpretation is clearly stronger than the assertion, and thus EIs that select for O_PS, e.g., un NP qualunque or oarecare, can be felicitously used. On the other hand, in downward-entailing contexts, the presuppositional exhaustification does not work, and the EIs that select for it are therefore ruled out.

This brief overview demonstrates how differences among EIs are captured on an alternative-based framework. Just like everywhere else in the polarity system, variation stems from constraints on lexically determined sets of alternatives (e.g., minimal vs. non-minimal subdomains lead to partial and total variation inferences) or on exhaustification (the proper strengthening requirement precludes an NPI use). With this background in mind, we now turn to a further parameter of variation, namely the interaction with focus.

3.1 Total variation and focus in downward-entailing contexts

It has often been observed that certain EIs can acquire a “not just any” reading in downward-entailing (DE) contexts, if they are used in a denial context or if focused. The following Italian examples (from Chierchia to appear a) and Romanian sentences (due to Săvescu-Ciucivara Reference Săvescu-Ciucivara, Alboiu, Avram, Avram and Isac2007) illustrate this behavior:⁸

Recall that these EIs do not have NPI-uses, and hence are typically ruled out under downward-entailing operators. Here however, we see that emphatic stress, which I assume to signal focus, enables their occurrence in these environments, and (28a) for instance conveys that the speaker doubts that Gianni saw any old student, and thinks Gianni saw some special student. A similar reading arises if a modal is present in the structure, as in (29b), where the free choice effect is denied. The main issue to be dealt with in this connection concerns the way the presence of focus affects the meaning of the EI.

Given the parallel treatment of EIs and disjunction, it seems reasonable to look for an answer by considering what happens with focus on scalar items outside the polarity system. Implicatures typically disappear under downward-entailing operators, but it is well-known that focus on the implicature trigger signals that the implicature is retained (e.g., HornReference Horn1989, Chierchia, Fox and SpectorReference Chierchia, Fox, Spector, Maienborn, von Heusinger and Portner2012 among others), as illustrated in the following example repeated from above:

How is this effect of focus derived on the alternative-based approach we are pursuing? Recall that disjunction optionally activates scalar and possibly domain alternatives. When active, scalar alternatives lead to the scalar implicature, i.e., the exclusive interpretation, whereas (recursive) exhaustification over domain alternatives is responsible for the free choice effects arising in modal contexts (cf. (8)–(9) above). Let us assume in the sentence in (30) only scalar alternatives are active and see how the exclusive construal of disjunction comes about. If exhaustification applies above negation, the assertion entails all its alternatives and hence the enriched meaning would be identical to the assertion (cf. (23) above). To derive the correct interpretation of (30), we must assume that the exhaustification operator applies below negation, as in the following LF:

Focus thus simply signals that the (scalar) alternatives are active and exhaustification takes place within the scope of negation. Could we envisage a similar solution for EIs? The answer is negative: EIs obligatorily activate both scalar and domain alternatives, and consequently, if exhaustification were to take place in the scope of negation, the associated LF of (28a) would be as follows:

This however does not seem a viable option: exhaustification in (32b) yields ungrammaticality (something which can be easily checked) – in the absence of a modal, the scalar and the free choice implicature contradict each other. On the basis of this, we must conclude that exhaustification applies above negation, at least in the absence of modal operators. In other words, the assumption that exhaustification takes place below negation can only be exploited in modal contexts, leaving us with no (or a different) explanation for the “not just any” reading arising in non-modal contexts. But the examples above show that the “not just any” reading arises regardless of whether a modal “intervenes” between the downward-entailing operator and the EI, if the EI is focused. So we are back to our original question: how does focus lead to the “not just any” interpretation of a total variation EI like un NP qualunque/oarecare?

In principle, there are two options we could pursue. Focus can either affect the way in which alternatives get factored into meaning (i.e., mode of exhaustification) or can have an effect on the alternatives considered for exhaustification. The first line of analysis is pursued in Crnič (Reference Crnič2011a), who convincingly argues (building mainly on KrifkaReference Krifka1995; Chierchia to appear a) that stressed any behaves on a par with emphatic NPIs and selects for exhaustification via silent even (as defined in (5b)). The assumption that the presence of focus involves even-like exhaustification is exploited to derive the licensing of stressed any in non-downward-entailing contexts (e.g., non-monotone quantifiers or desire predicates). It remains to be seen to what extent this solution is applicable to EIs, a matter I will leave for future research. In the remainder of this chapter, I would like to explore the latter option, namely the hypothesis that focus affects the alternatives to which the exhaustification operator O applies. More precisely, I will defend the claim in (33):

This hypothesis is directly inspired by the treatment of Chinese wh-indefinites in Liao (Reference Liao2011).⁹ Let me briefly review the relevant aspects of her proposal, and then show how it extends to EIs.

In a detailed study of Chinese wh-indefinites, within the alternative-based framework adopted here, Liao discusses focused wh-phrases in negative sentences, illustrated in the following example:¹⁰

The wh-indefinite shenme can acquire either an NPI-reading (34a), or the interpretation in (34b), whereby Zhangsan did buy some (ordinary) book. Wh-indefinites are similar to EIs in that they activate both scalar and domain alternatives (of any size). The NPI-reading is obtained along the lines of (23) above. More relevant for our present purposes is the second construal of the wh-indefinite in (34). According to Liao, this can be derived by conjecturing that (34) involves focus and the presence of focus leads to the consideration of a domain wider than the one initially associated with the assertion. The desired interpretation then comes about in the following way. Assuming Rooth’s (Reference Rooth1992) focus theory, and putting the entire proposition in (34) in focus, its LF representation is as in (35a):

Focus adds alternatives that range over the wider domain D′, i.e. the second alternative in the set in (35b). Since this proposition is stronger than the assertion, it will get excluded through exhaustification. Once we put together the assertion and the derived implicatures, the resulting interpretation is that Zhangsan didn’t buy any book in D, but he did buy some book in D’.¹¹

Liao’s proposal provides a straightforward account for the “not just any” reading of the focused EIs in (28) and (29). To see this in more detail, consider once again the following sentence:

Recall that a total variation EI like un qualunque lexically activates scalar and subdomain alternatives, given in (37)–(38). As we have already seen in (23), exhaustification applied above negation is vacuous (all alternatives are entailed by the assertion), i.e., it returns the assertion: Gianni didn’t read any book in D, where D is {a,b,c}:

Since the NPI reading is disallowed by “proper strengtheningm,” this configuration is normally ruled out for un qualunque. In the absence of focus, this sentence is ungrammatical. But here the EI is focused and consequently, the LF for (36) is as in (39a). Following Liao’s proposal reviewed above (formulated as in (33)), we assume that the presence of focus adds domain alternatives, i.e., leads to the consideration of a superset of D, namely D’= {a,b,c,d}:

Exhaustification applies to this enlarged set and seeks to eliminate stronger alternatives. The alternatives in (38) are all entailed by the assertion (i.e. ¬(a ∨ b ∨ c)), but the alternative based on D’ (¬(a ∨ b ∨ c ∨ d)) is not. So through exhaustification, it will be excluded. The resulting meaning can be represented as in (40):

According to (40), the sentence in (36), with un libro qualunque focused, is interpreted as saying that Gianni did not read a book in the initially considered domain of quantification, but he did read a book in some other, special domain. This does not explain how the book in the “special” domain is singled out. However, there is little doubt that it is a contextual matter, and a pragmatic effect of having changed the initial domain. Importantly, it is not something that comes from how exhaustification proceeds. For this proposal to work and derive the intended reading, it is crucial that the set considered at the final level of exhaustification (i.e., exhaustification applying to the lexically activated alternatives plus those brought in by focus) includes alternatives that are not entailed by the assertion. This ensures that we get a coherent meaning after applying O. But the choice of D and D’ is regulated by pragmatic factors. As we have seen in the previous example, a reasonable way to get the right result (and thus explain why focus rescues sentences like (40)) is by adding alternatives to D. But there are cases involving a different strategy. For example, imagine the initial D is set to the books on a given reading list. Clearly one can use (40) in this situation and add a continuation like I read the most difficult one (on the list).¹² What this means is that the difference between the books that have not been read (D) and the one(s) that have been read (D’) is established by bringing in additional factors (e.g., difficulty). I do not have a full story on how the various contextual factors may affect the observed readings. But I hope the general line of reasoning is clear even in the absence of more precise assumptions on the choice of domains. What matters for our present purposes is that focus on total variation EIs adds domain alternatives, thus offering a way to avoid the ungrammaticality of configurations like (40).¹³ In DE-contexts, the addition of non-entailed alternatives amounts to the consideration of super-domains (something we represented by adding alternative d to the initial domain of books).

In the following, I will extend this proposal and exploit the hypothesis in (33) to account for the restrictions governing focus on partial variation EIs likevreun or un qualche.¹⁴

3.2 Partial variation and focus

In the previous section, we have seen that EIs like un NP qualunque or oarecare can be focused, acquiring a “not just any” reading in DE-contexts. In order to make sense of this behavior, I have adopted the proposal in Liao (Reference Liao2011) whereby focus adds domain alternatives to the set of lexically activated alternatives.

I now turn to what I think is a previously unnoticed restriction, namely the impossibility of Romanian vreun to associate with focus, illustrated below: as soon as vreun is focused in the sentences in (41), they become unacceptable:¹⁵

The fact that vreun cannot be stressed is surprising, on at least two grounds. First, we have just seen that EIs like un qualunque/oarecare can be focused. Second, vreun also has an NPI use, and NPIs are well-known to be able to carry emphasis (cf. (27)). So no matter how we look at the data in (41), we are faced with a puzzle. Moreover, as illustrated in (42), this restriction seems to be shared by other partial variation EIs – focus on algún or un qualche is systematically rejected:

The conclusion that can be drawn from these facts is that partial variation EIs cannot be focused, unlike total variation EIs, NPIs, or simple indefinites, to which they are closely related. How can an alternatives-and-exhaustification approach accommodate this behavior? Ideally, from the interaction between lexically activated alternatives and focus activated alternatives. Recall that the sole difference between un oarecare and vreun (and more generally between total and partial variation EIs) lies in the set of subdomain alternatives they activate, which must consist of all subdomains of the relevant quantificational domain for the former, and singleton subdomains only for the latter. Accordingly, if we pursue the same line of reasoning as before, we expect the (im)possibility to associate with focus to be somehow related to this difference. In other words, the fact that vreun activates singleton domain alternatives has to be incompatible with the posited effect of focus on domain alternatives. In the following, I will argue that this route can be fruitfully pursued to capture the restriction in (41b), and show that this explanation can be extended to the partial variation EIs exemplified in (42). I briefly return to the negative polarity uses of vreun in Section 4.

In the previous section, we accounted for the interpretation of the focused occurrences of EIs like un oarecare by assuming that focus adds domain alternatives to the set of lexically activated alternatives. Recall that on the present approach, the choice of the relevant quantificational domain D is a matter of pragmatics, but the subdomains are semantically determined and must consist of all subdomains of D (a property that is responsible for the total variation inference). Consequently, the only way to add domain alternatives (in the presence of focus) is by extending the initial domain, thus making exhaustification apply over both sub- and super-domain alternatives. This delivers the right interpretation.

Things are different for partial variation EIs. In particular, their lexically determined alternatives are different, in that they are restricted to singleton domain alternatives. As shown in Section 2.2, this assumption derives the fact that EIs like vreun or un qualche always prompt partial variation. I will now argue that it is this very same restriction to singleton alternatives that can explain the ban on focus illustrated in (41b). To see this, consider the set of domain alternatives lexically activated by vreun (in a context with three elements in D):

In the absence of focus, the sentence is grammatical and triggers a partial variation inference. Once we bring in focus, domain alternatives must be added to the set in (43). Here however, we can include domain alternatives that look as in (44a), i.e., non-minimal subdomains, and as a result, exhaustification will apply to the set in (44b). Once we allow for pre-exhaustification of these alternatives, and put together the assertion with the derived (scalar and free choice) implicatures, through steps familiar from Section 2.2 above, we wind up with the total variation meaning in (44c).

This state of affairs is not tolerated by vreun, which lexically encodes a ban on total variation, as extensively argued in Fălăuş (Reference Fălăuş2012). The gist of the proposal I am referring to can be summarized as follows. Romanian vreun is only felicitous in contexts where one of the alternatives in the quantificational domain can be excluded (although the speaker does not necessarily know which one). In other words, instead of triggering the partial variation inference (i.e., some but not necessarily all), it imposes a stronger – anti-total variation – restriction (i.e., it is necessary that not all alternatives constitute viable options). This requirement is built in the semantics of vreun, an idea that is formally implemented by assuming that vreun includes among its alternatives the total variation element un oarecare (and its alternatives). Essentially, the total variation inference normally associated with un oarecare (i.e. ◊ a ∧ ◊ b ∧ ◊ c) gets added to the set of alternatives in (43) and excluded via exhaustification, as sketched in (45):

This derives the correct “partial variation only” associated with vreun and accounts for its restricted distribution (most prominently its exclusion from deontic contexts), a property that is not directly relevant here. The crucial aspect of the analysis is that the lexical semantics of vreun encodes a ban on total variation. Focus on the other hand derives total variation, as illustrated in (44). This correctly accounts for the clash observed in (41b): the alternatives brought in by focus conflict with the standard meaning of vreun. On the theory outlined here, the non-association of vreun with focus in modal contexts receives a principled explanation, based on two independently motivated assumptions, namely the constraint on singleton subdomains and the role of focus.

Crucial to the analysis proposed above is the hypothesis that focus affects the set of domain alternatives considered for exhaustification. Importantly, the shape of the domain alternatives that get added through focus depends on the shape of the lexically activated subdomains. In case the initial domain alternatives are singletons, focus brings in larger (i.e., non-singleton) subdomains. If focus were allowed to add further singleton domain alternatives (like for example in the case of total variation EIs), we would expect to get a partial variation inference that holds for a wider quantificational domain. The fact that this never happens indicates that the effect of focus on the domain alternatives is constrained by the lexical alternatives.¹⁶

This brings us to un qualche and algún, which are similar to vreun in that they never trigger total variation inferences. However, they are not so strictly incompatible with total variation set-ups: the sentence in (42c) for example, without focus on un qualche, is perfectly acceptable in a context where the speaker would be happy to play any musical instrument, without excluding any relevant alternative. In other words, unlike in the case of vreun, nothing in the meaning of un qualche (or algún) enforces partial variation. Focus is nevertheless infelicitous, as indicated in (42). On the present line of analysis, this means that these EIs disallow total variation (although this requirement is not as strong as in the case of Romanian vreun). A plausible way of thinking about this is in terms of a blocking effect: the fact that both Spanish and Italian have total variation elements (un NP qualsiasi/cualquiera) arguably blocks the possibility of the partial variation item to sustain the total variation inference triggered by focus. This leaves open the option that if for some reason (or for some speakers) this blocking effect can be overridden, focus becomes acceptable. Further empirical investigations are needed to see if this expectation is borne out in Spanish and Italian.

A related prediction made by the account outlined here is that an EI that can convey both partial and total variation inferences should be able to associate with focus, and the presence of focus should correlate with total variation. As shown in the next section, there is interesting preliminary evidence that German irgendein verifies this prediction.

3.3 Irgendein and focus

The previous section proposed an explanation for the fact that vreun does not associate with focus in modal contexts. Building on the conjecture that focus adds domain alternatives, we have tied the behavior of vreun to the activation of singleton domain alternatives and the ban on total variation which is part of its core meaning. Furthermore, we have seen evidence indicating that this restriction is shared by other partial variation EIs. What all these EIs have in common is the fact that they can never trigger total variation. This option however is available for the German EI irgendein, as illustrated in the following examples repeated from Section 2.2:

As discussed in detail in Aloni and Port (Reference Aloni, Port, Fainleib, LaCara and Park2010), the sentence in (46a) receives a total variation interpretation, whereby any doctor could satisfy the modal requirement. The epistemic statement in (46b) on the other hand is most readily used in partial variation scenarios (e.g., the hide-and-seek context in (15)). The precise distribution of the partial vs. total variation inferences is not directly relevant for our present purposes. What is crucial to the current discussion is that the configurations where irgendein associates with total variation seem to require focus. According to Haspelmath (Reference Haspelmath1997: 249), “in the free choice function, the irgend-series may occur, but it must be stressed”:

Similar observations can be found in Aloni and Port (Reference Aloni and Port2011), who furthermore note that irgend- cannot be stressed in contexts where it triggers partial variation (i.e., episodic and epistemic modal contexts):

I think the data require closer scrutiny, but if these facts turn out to be solid, we have a further argument in favor of the hypothesis that focus leads to the addition of domain alternatives. As mentioned in Section 2.2 above, Chierchia (to appear a) suggests that irgendein can activate both singleton and non-singleton subdomains, but the larger alternatives are optional. By default, exhaustification applies to the active singleton alternatives (together with the scalar alternatives), yielding partial variation. If on the other hand irgendein occurs in a context where total variation is enforced (by the operator in the context of occurrence, in particular by deontic modals, cf. Fălăuş Reference Fălăuş2012 and Aloni and Franke, this volume), irgendein is focused.¹⁷ As soon as focus is present (i.e., non-minimal domains are considered for exhaustification), irgendein winds up sustaining the total variation inference. Put another way, focus is required to avoid drawing the weaker, partial variation inference.¹⁸ A more careful empirical investigation is clearly needed before drawing any conclusions on the interaction between the “licensing” operators, the interpretation of irgendein and focus. But the evidence available at this point provides interesting support in favor of the correlation between focus and total variation observed for the Romance EIs discussed in the previous sections.