Reconstructing the syntax of focus operators *

This paper presents novel evidence that the exclusive operator alleen in Dutch (and nur in German) can directly attach to the focus constituent it associates with, and against an analysis like the one in Jacobs 1983 and Büring & Hartmann 2001 which analyzes all instances of alleen/nur as sentential adverbs that take a single syntactic argument that denotes a proposition. Instead, we argue that alleen/nur takes two syntactic arguments, which combine to denote a proposition. The evidence comes from novel data showing scope reconstruction of [ alleen/nur + DP] sequences from the prefield in Dutch (and German), adding to earlier arguments in Reis 2005 and Meyer & Sauerland 2009.


Introduction
In English, focus-sensitive operators like only can occur both in adnominal and in adverbial positions, as shown in (1a) and (1b) respectively.In (1a), only occurs adjacent to the DP it associates with, and arguably adjoins directly to the DP.In the structure in (1b), only attaches in adverbial position to the VP, which contains the semantic focus.Only differs from adverbs like always in that it can adjoin to both VPs and DPs, while adverbs are more restricted, as shown in the comparison between (1a) and (2a).In other words, cases in which only and its focus are apparently placed in first position and the verb in second position are analyzed as involving a verb-third word order instead, which is usually considered impossible in German and Dutch.This analysis also seems in conflict with the fact that this word order necessarily involves association with the subject, and is incompatible with sentence-wide focus: (5) A: 'Did anyone dance?' B: #Nee. No.
Alleen Only

[Anna Anna heeft has
Maria Maria gekust] F .kissed Intended:'The only thing that happened is that Anna kissed Mary.' The Dutch response in (5) cannot convey that the only thing that happened is that Anna kissed Maria, which is what the context in (5) would require.Only cannot associate with the entire sentence in (5), it has to associate with the subject, conveying that nobody other than Anna kissed Maria.The sentence would be compatible as a response to the question 'Did Hans kiss Mary?'.In general, only can only associate with its sister or with constituents within its sister constituent (i.e., constituents it c-commands).The association pattern in (5) is then as expected if only must attach to the subject in this configuration, but remains unexplained under the Adverbial Analysis.
The strength of the Adverbial Analysis is that it can explain certain syntactic restrictions in Dutch and German that seem surprising under a Mixed Analysis.One such restriction is that nur/alleen cannot attach inside Prepositional Phrases (*mit nur Hans/ *met alleen Hans 'with only Hans'),2 .This is expected if nur/alleen can attach only in adverbial positions (Jacobs 1983, Büring & Hartmann 2001).Jacobs (1983) and Büring & Hartmann (2001) also note that the Adverbial Analysis predicts that in sentences in which nur associates with the verb, it will only be able to occur adjacent to the verb (its focus) if material that would otherwise intervene can scramble 'out of the way'.This prediction seems to be borne out.As the comparison between (6a) and (6b) shows, nur cannot surface adjacent to the verb in (6b), because directional PP complements cannot scramble.[fahren] F drive darf].may 'because one can only drive the car into the garage.'Adapted from Büring & Hartmann (2001: ex.20a and 22a) Interestingly, it seems that nur actually has to be placed as close to its focus as syntactically possible, or at least there is a strong preference for positioning it in this way.Jacobs (1983) and Büring & Hartmann (2001) capture this by positing a constraint that directly enforces this word order preference for being placed as close to the focus as possible, and use this constraint to explain why it often appears as if only attaches to the constituent it associates with-whenever possible, it will occur adjacent to it.
In addition to arguments in terms of surface syntactic distribution, Büring & Hartmann (2001) add a novel argument in favor of the Adverbial Analysis and more specifically for the claim that nur, when sentence-initial, does not attach to the constituent immediately preceding the verb, but to a node containing the entire sentence.The argument rests on the claim that although DP reconstruction from first position is generally possible, nur can never reconstruct along with the DP, as would be expected if [ONLY + DP] formed a constituent.
In this paper, we will present evidence that reconstruction of [ONLY + focus constituent] is possible after all, strengthening earlier evidence against Büring & Hartmann's (2001) claim presented in Reis (2005) and Meyer & Sauerland (2009).In the following, we will use the term ONLY when referring to the exclusive operator when we mean to refer to both alleen and nur, and use the language-specific operators (alleen, and nur) otherwise.
Before embarking on this argument, we will outline an analysis of ONLY that is compatible with the syntactic flexibility assumed in the 'mixed' approach, and yet can account for some of the syntactic constraints that motivated Jacobs (1983) and Büring & Hartmann (2001) to adopt the Adverbial Analysis.We will then use scope reconstruction as a tool to compare the two accounts.

early access
that nur can attach to extended projections of the verb, but not to argument DPs.In both analyses, in order to associate with an argument DP, nur has to attach to an XP node containing it (e.g., VP or CP), similar to sentence adverbials like always.We therefore call this type of analysis the 'Adverbial Analysis'.
We will compare this analysis to the analysis of ONLY proposed in Wagner 2006, under which ONLY takes two syntactic arguments, a constituent that acts as the focus and a constituent that combines with it to form a proposition as shown in (7) (see Rooth 1985, Drubig 1994, Krifka 1996, Bayer 1996: for earlier analyses of only with similar assumptions).( 7) The two analyses of only considered here-the two-place analysis according to which both adnominal and adverbial attachment are possible, and the Adverbial Analysis, according to which only adverbial attachment is possible-derive the same overall truth conditions, assuming that we restrict the alternatives appropriately. 5oth analyses assume that focus association is constrained such that ONLY can only associate with constituents within the constituent it attaches to.Both employ Alternative Semantics to explain why the apparent semantic focus does not necessarily have to be the entire constituent it attaches to, but can be further contextually restricted.
The two analyses differ regarding their claims about the syntax of ONLY.However, they are both compatible with some of the facts that led Jacobs (1983) and Büring & Hartmann (2001) to propose the Adverbial Analysis.
For example, the above-mentioned unavailability of attachment inside PP's and VP's seems like a straightforward argument in favour of the Adverbial Analysis.However, the particular version of the Mixed Analysis assumed here is compatible with this observation.Following Bayer (1996), we argue that the unavailability of attachment inside PP's can be related to the fact that preposition stranding is not allowed in Dutch and German.The LF required to interpret only, illustrated in (10), cannot be derived because the preposition is left behind:  (11b) shows that the VP containing the PP may be fronted, (11c) shows that movement of the verb by itself is unnatural.This is as expected if fronting a predicate always involves fronting the entire VP node containing it, and if directional PPs cannot scramble out of the VP 7 .The source of the infelicity of placing alleen next to the verb in (11a) is the same in both theories: alleen actually has to attach to a VP node containing the focus in both theories.In the Adverbial Analysis, this is generally true due to semantic and selectional properties of ONLY; in the two-place theory this is a result of how ONLY interacts with syntax: By attaching ONLY to a constituent that cannot move, the right configuration to interpret ONLY cannot be derived at LF.In sum, at least two of the syntactic arguments from Jacobs (1983) and Büring & Hartmann (2001) do not distinguish between the Adverbial Analysis and our particular version of a Mixed Analysis.Sudhoff (2010) discusses several other related data points, and shows corpus data that illustrates that the distribution of nur is in fact not as restricted as Jacobs (1983) claimed.A discussion of the full range of syntactic arguments is beyond the scope of this paper.See Bayer (2016)  Hirsch (2017) for relevant recent discussions.In the following, we will focus on how the two analyses compare when it comes to the facts about scope reconstruction.
3 The reconstruction argument Büring & Hartmann (2001) ' Meyer & Sauerland (2009: Ex.11) In this example with the negative quantifier 'keiner', both readings are clearly available in (13).One natural way to pronounce this sentence under the reconstructed scope reading is to use a contrastive topic intonation (Büring 1997), andSudhoff (2010: p.168) reports that under this intonation only the reconstructed reading is available.When using the same intonation in ( 12), a reconstructed reading seems unavailable, just as Büring & Hartmann (2001) observe.But as Sudhoff (2010) points out, this is not surprising, given that the reconstructed reading would be incompatible with the pragmatic import of this intonation: According to Büring (1997) the intonation implies that there has to be some alternative that remains disputable after uttering the sentence, but the reconstructed reading would arguably resolve all relevant alternatives. 8The important point about ( 13) is that if it is true that reconstruction is possible here, then this shows that nur can at least sometimes attach to DPs, and the Adverbial Analysis can therefore not be correct in its strongest form.Meyer & Sauerland (2009) assume with Büring & Hartmann (2001) that the reconstructed reading in ( 12) is indeed at least undetectable, but they present an alternative explanation to Büring & Hartmann's for this.Meyer & Sauerland (2009) argue that the availability of ( 12b) is masked by a reading that can be judged true in a larger set of circumstances, namely (12a).Their argument is based on the Truth Dominance principle, which claims that if a reading is ambiguous and it is true under its most accessible reading, it will be judged as true (see also Reinhart 1976, Abusch 1994).In ( 12), the reconstructed scope reading is undetectable as its truth entails that of the surface scope reading.In other words, we cannot find a context where the reconstructed scope reading is true and the surface scope reading is false; if everyone loves only Mary, it also has to be the case that Mary is the only one who is loved by everyone.In (13), by contrast, this explanation correctly predicts that the reconstructed reading should be detectable.Given that the reconstructed scope reading does not entail the surface scope reading, this sentence has a logically independent reading that can be tested for.In structures with a negative quantifier in subject position, as is the case for (13), the reconstructed reading is available in (13b) because the entailment relations are reversed.
In the prior literature, the claim that the original example in (12) does not allow reconstruction has not been contested.We first note, however, that the intuition that the reconstructed reading is absent in (12) is intuitively not obvious at all when pronouncing the sentence with main prominence on the fronted constituent (Bayer 2016 independently made this observation).In the following, we will discuss several arguments that show that reconstruction in fact is available and detectable for [ONLY+universal] after all.In addition, we look at the scopal interactions between ONLY and various adverbs, which provide clear evidence for the possibility of reconstructing [ONLY + DP].Our arguments show that the parse of ( 12) that treats [ONLY + DP] as a constituent is available after all, and we will in fact see evidence that it might be the only parse available. 9early access Liz Smeets, Michael Wagner

Argument 1: Revealing a logically stronger reading
In ambiguous structures in which the two readings stand in an entailment relationship, the presence of the stronger reading is notoriously hard to detect.This makes it difficult to probe our intuitions about (12).In any situation in which everyone likes only Mary it will also be true that only Mary is such that everyone likes her.A similar issue often arises with evaluating the scope of negation.Consider the following example, in which the existential operator within the negative quantifier might take different scopes with respect to negation (Potts 2000): The company needs to fire no employee.a.There are no employees x such that the company is obligated to fire x. b.It is not the case that the company is obligated to fire employees.
It is clear that reading (a) can be true while (b) is false, for example, maybe the company decided that they need to fire one employee but does not know which one.
Then there is no particular employee that the company must fire, as claimed in (a).It thus seems clear that the sentence has reading (a).But how can we be sure that it also has reading (b)?To show proof of an ambiguity, we would want to construct a scenario in which (b) is true and (a) is false (cf.Gillon 1987).But this is impossible here, since any situation that makes (b) true will also make (a) true, since (b) entails (a).However, Potts (2000) (see also Gajewski 2005: 43-44) illustrates a strategy to show that ( 14) is indeed ambiguous, by adding the following context: (15) Suppose Mike, nervous employee of much-hyped.comwhose stock has plummeted, says to his fellow employee Greg, "I hear the company's going to fire someone.We're all equally likely to get the boot; they just need to make a cut."

early access
We can test for the reconstructed reading by creating a situation in which only that reading would be a coherent way to deny something that was previously said.

(17)
Context: There's been a bank robbery, and the inspector asks one of his assistants about the investigation.16) only had the surface scope reading, the dialogue in ( 17) should be incoherent, but it is perfectly fine-at least when the response is pronounced with main prominence on the subject.Rather than looking at which propositions our target sentence involving alleen can be used to contradict, we can also test what responses are possible contradictions of that sentence.Consider the following dialogue: York York zullen will komen.come 'that two people from New York will again come.' (modified from Neeleman & Koot (2007) as cited in Szendroi ( 2010)) We can use reconstruction under adverbs from first position as a further test for the constituency of [ONLY+DP].To do so, we need a context that is incompatible with the surface scope reading.In order to test whether our context is successful in making surface scope infelicitous, we can use the configuration in which only precedes alweer in the middle field, where we know independently that reconstruction should be impossible.And indeed, our context is successful in making wide scope over the adverb infelicitous: 11 As shown in the appendix, the same generalization holds in German.We note that this conflicts with the generalization about scope and word order in Frey (1993).According to Frey, word orders which deviate from the canonical word order lead to scope ambiguities, since they always allow reconstruction of the moved argument into the lower trace position, including reconstruction below an adjunct (Frey 1993: 193ff).This, however, appears not to be true for the adverb alweer in Dutch and wieder in German, since neither word order in (20) is ambiguous.One possible explanation why movement across an adverb cannot reconstruct in the middle field is that the traces left by movement in the middle field are necessarily all of type e.Another possibility is to assume that arguments can be base-generation above the adverb (see Fanselow 2003, 2001, Neeleman & Koot 2008: for discussion), and to further assume that base-generating an argument low and then moving above an adverb is impossible.For example, Neeleman & Reinhart (1998) assume a cross-derivational constraint that prefers base-generation over deriving a word order by movement.Maybe movement to first position, for some reason, can leave high type traces, or maybe syntactic reconstruction is possible.Important here is that the two configurations differ in this regard.'John was again the only one who didn't do his homework.' The reason for the infelicity of B 's response is that it conveys that nobody other than Jan did not do their homework for a second time, but since last week everyone other than Jan did do their homework, none of them could possibly have not done it again this week anyway.By contrast, the response in B is perfectly felicitous.We can now use this judgment as the baseline for the case in which [ONLY+DP] occur in first position as shown in ( 22).The sentence is felicitous, at least when pronounced with nuclear stress on Jan:

Argument 3: Question-answer congruence
Our third argument is closely related to the second.It relies on the observation that it is infelicitous to introduce certain adverbs into an answer if they do not already form part of the question.The source of such effects is that the alternatives that the answer makes salient have to be congruent with the question (Hamblin 1973, Stechow 1986/1989, Rooth 1992).We will illustrate this effect based on an English example involving again.We should note, however, that the word order that we have to use in English in order to be sure of the scope of again is not the most natural.It would be most natural to place again in final position, but then both low and wide scope attachments are possible.Instead, we have to 'sandwich' again between the auxiliary and the VP.Note that we are only using English here to convey the basic intuition about the Dutch example, where this issue does not arise.Consider the following dialogue (cf.discussion in McKillen 2016: 117-119): (23) Context: The TAs discuss who they think failed the last exam.
in the English illustrates a difference between English and V 2 languages like Dutch (and German), where reconstruction of the first constituent is generally more freely available.

Extending the observations to other adverbs
To ensure that the reconstruction effects are not tied to specific properties of We can also apply the semantic requirement on congruent sentential answers to this case.The question in (31) generates a set of propositions of the type x passed, not x definitely passed.The answer in (31a) is felicitous.The response in (31b) conveys who definitely passed, but leaves open the possibility that others past as well, and hence seems a bit odd: To summarize, by now we have seen several arguments showing that reconstruction of [ONLY+DP] is possible, which shows that the two can form a constituent.This means that the adnominal parse in (4b) is available.We now turn to an argument that shows that the adverbial parse in (4a), according to which only attaches to the entire clause, and the DP occurs in second and verb in third position, is in fact unavailable.

Evidence from long-distance movement
When the constituent surfacing in the first position in V 2 originates in an embedded clause, it has a special property: It is obligatorily interpreted within the embedded clause, and cannot take scope over the material that forms part of the matrix clause.Wagner (2004) observes that in German, when a constituent is fronted from an embedded clause, the fronted constituent cannot bind a variable in the subject of the matrix clause, while a variable in the fronted constituent can happily be bound by the following matrix subject (32).The same is true for Dutch, where long-distance movement also appears to obligatorily reconstruct: ( The availability of the reconstructed reading in (34) shows that ONLY can form a constituent with the following focus constituent.However, the examples involving long-distance movement support a stronger claim: The fact that reconstruction is obligatory shows that ONLY apparently must form a constituent with the focus constituent.Otherwise, we would expect that only should be able to attach to the left periphery of the matrix clause, and therefore be able to take scope over adverbs in the matrix clause.If only could attach high as in the adverbial parse in (4a), then the wide scope reading over the matrix adverb should be available.Unless there is an independent reason that blocks attaching only to the entire sentence exactly when the constituent in first position originates in an embedded sentence, the scope facts for long-distance movement are unexpected if the adverbial parse of first-position only was possible.The facts are as expected if ONLY+DP in first position in fact must form a constituent.
And yet, there is an important difference between the long-distance case and the examples we looked at earlier: We saw that our earlier examples were always ambiguous between surface scope and inverse scope.One way to think about this is that fronting to first position obligatorily reconstructs when the fronted constituent originates in an embedded clause but optionally reconstructs when it originates in the same clause.However, our observations are also compatible with the view that reconstruction of [ONLY+DP] is in fact always obligatory.Looking again at the example in ( 21), note that it is not apparent from which position alleen Jan has moved to first position.It could have moved from a position above the adverb (see (35a)), or from within the VP underneath the adverb (see (35b)).The ambiguous sentence may thus have two different possible derivations, in each [only Jan] will be interpreted in its position before dislocation to first position: ( We could thus derive both readings even if reconstruction of [only Jan] was in fact obligatory even here, just as in the case of long-distance movement, so we might reconstructs and the reconstruction site is realized with reduced prominence.We do not offer an explanation for these effects, but merely note that prosody appears to correlate with reconstruction.Exploring this apparently systematic prosodic effect further would warrant a separate investigation.

early access
Liz Smeets, Michael Wagner not need to find a reason why reconstruction from first position would be obligatory only when it involves long-distance movement.14

Conclusion
In this paper we presented novel evidence for the availability of scope reconstruction of [ONLY + DP] sequences from the prefield in Dutch (and German), including cases where this reading was previously thought to be unavailable or at least undetectable.
used to pass [ DP only [ DP [syntax] F exams]] Adnominal b. Mary i used to [ vP only [ vP t i pass [Syntax] F exams]] Adverbial (2) a. *Mary used to pass [ DP always [ DP [syntax] F exams]] Adnominal b. Mary i used to [ vP always [ vP t i pass [syntax] F exams]] Adverbial The syntax of Dutch makes it harder to identify the left edge of the VP, and hence the attachment site of alleen is often not obvious.Two potential parses of the Dutch b.Adnominal attachment: [ Alleen Anna ] [heeft Maria gekust].Only Anna has Maria kissed 'Only Anna kissed Mary.' s not true.A little boy also saw the perpetrator.'Thesetests based on the coherence of contradictions show that reconstruction is possible, even in cases where it is hard to detect for the reasons outlined inMeyer & Sauerland 2009.The possibility of reconstruction shows that [ONLY + DP] can form a constituent after all.3.2Argument 2: Scope reconstruction relative to adverbialsAnother test can be devised based on the relative scope of a constituent in first position and adverbs in the middle field.10An argument placed in first position can scope below an adverb: again > two people from New York / two people from New York > again A standard assumption is that low scope requires reconstruction into the trace position.Interestingly, when an argument precedes an adverb in the middle field, the argument cannot take scope below the adverb: <s,t>> .Presupposes: ∃ x.P(x) in w 4 .
(Bayer 1996rtmann 2001: [y = x → P(y) is false in w]Similar to the analysis inBüring & Hartmann 2001, ONLY requires a whole family of lexical entries, one for each type σ of constituent that it can attach to.While inBüring & Hartmann 2001only always attaches to a node in the extended projection of a verb or a non-argument, under the current account ONLY can combine with a constituent of any type, as long as an LF can be derived in which [ONLY + focus constituent] are in a sister relation to a constituent that it can compose with to form a proposition.In this analysis, any syntactic constraints on the distribution of ONLY must be due to the impossibility of deriving an appropriate LF to interpret ONLY(Bayer 1996).Let's look at an example: Meyer & Sauerland (2009)nerally cannot reconstruct because nur and the following DP do not form a constituent under the Adverbial Analysis.The sequence in first position in (12), for example, is argued to be unable to reconstruct under the universal quantifier every.The inverse scope reading in (12b) should then be unavailable, which is unexpected if nur can attach to a DP: Surface Scope: 'Mary is the only one such that everyone loves her.' b.Reconstructed Scope: * 'Everyone loves only Mary.'If the reconstructed reading was available, then this would constitute evidence for [nur + DP] to form a constituent.Reis (2005)andMeyer & Sauerland (2009)countered this argument with the observation that reconstruction of the [nur + DP] sequence in the prefield is possible with certain other quantifiers: i a. Surface Scope: 'The only one nobody loves is Mary.' b.Reconstructed Scope: 'Nobody loves Mary and no other person.
The felicity of using the sentence to contradict Mike therefore shows that reading (b) must be available.The trick is to create a situation in which only the logically stronger reading would lead to a contradiction.We can apply this test to our case (thanks to Junko Shimoyama for the suggestion).Consider the sentence in (16).If reconstruction is possible, it should have two readings: 'Only the escape car is such that everyone saw that.'ReconstructedScope: 'Everyone only saw the escape car.' Potts observes that Greg could use (14) to contradict Mike's statement.But if (14) only had reading (a), this contradiction would not make any sense.
AGAIN, we extend our observations to two other adverbs, POSSIBLY and DEFINITELY.The reconstructed scope is easily detectable in sentences with existentially quantified adverbs such as POSSIBLY because here, the reading obtained under surface scope entails the reconstructed reading.Surface Scope: Only Mary is such that she possibly passed this time To be sure that it is really reconstruction that makes the Dutch example felicitous we can compare the example with placing [ONLY+DP] before the adverb within the middle field.Here, as expected, only the surface scope reading is available: Surface Scope, infelicitous: #'Only Mary is such that she possibly passed this time, but it can be that Jan passed too.' b.Reconstructed Scope, unavailable: 'It is possible that only Mary passed this time, but it can be that Jan passed too.'When using a universally quantified adverb, such as DEFINITELY, the reconstructed scope reading is harder to detect because it entails the surface-scope reading.Surface Scope: Only Mary is such that she passed for sure this time ∀ p' ∈ Alt, ¬ p' unless p' = p 'Mary passed, and for people other than her, it is not necessarily the case that they passed.'b.Reconstructed Scope: For sure, only Mary passed this time ∀ p' ∈ Alt, ¬ p' unless p' = p.'Mary passed, and it is definitely the case that no person other than Mary passed.'Again,we can reveal a logically stronger reading by asking whether the sentence is felicitous in a context where the surface scope reading is denied.Speaker A thinks that there is a chance that Jan passed.Speaker B objects to A's utterance and utters B. The objection is only felicitous under the reconstructed reading, because the surface scope reading does not say anything about the possibility of whether Jan passed or not.No. Definitely only Mary passed, John didn't even study, he can't have passed the exam.' ∀ p' ∈ Alt, ¬ p' unless p' = p 'Mary possibly passed, and for the others, it is not possible that they passed.'b.Reconstructed Scope: It is possible that Mary is the only one who passed this time ∀ p' ∈ Alt, ¬ p' unless p' = p.'Mary possibly passed, and for the others, it is possible that they didn't pass.'In a context in which people other than Mary have possibly passed, the reconstructed reading leads to a true statement and the surface scope reading to a false one.The sentence in (27) illustrates that the reconstructed reading is available.The unambiguous paraphrases of the two readings in English show that the surface scope reading would be infelicitous here but the reconstructed scope reading is felicitous: ' 35) a. LF: [ CP [Only Jan] i [ C has [ T P t i again [ vP passed]]]]] For no other person x than [John] it holds that [x passed again] b.LF: [ CP [Only Jan] i [ C has [ T P again [ vP t i passed]]]]] It is again the case that for no other person x than [John] it holds that [x passed] Reis 2005 andMeyer &Sauerland 200983 a constituent, contrary to what is predicted by the adverbial analyses of ONLY inBüring &Hartmann 2001 andJacobs 1983, adding to earlier arguments inReis 2005 andMeyer &Sauerland 2009.We also showed evidence suggesting that the adverbial parse in which only attaches to the entire sentence is in fact unavailable.The data is compatible with an alternative two-place syntactic analysis of ONLY, which also provides an alternative way of thinking about some of the syntactic restrictions of ONLY that formed the original motivation for the Adverbial Analysis.Surface Scope: Only Mary is such that she possibly passed this time ∀ p' ∈ Alt, ¬ p' unless p' = p 'Mary possibly passed, and for the others, it is not possible that they passed.'b.Reconstructed scope: It is possible that Mary is the only one who passed this time ∀ p' ∈ Alt, ¬ p' unless p' = p.'Mary possibly passed, and for the others, it is possible that they didn't pass.' Surface Scope: #No other person than Mary has possibly passed this time, but it can be that Jan passed too.b.Reconstructed Scope: 'It's possible that only Mary passed this time, but it might be that Jan passed too.' Surface Scope: #'It's possible that only Mary passed this time, but it can be that Jan passed too.' b.Reconstructed Scope: Unavailable No other person than Mary has possibly passed this time, but it can be that Jan passed too.It is definitely the case that only Mary passed./OnlyMary has definitely passed.' a. Surface Scope: Only Mary is such that she passed for sure this time ∀ p' ∈ Alt, ¬ p' unless p' = p 'Only has definitely passed.' early access Liz Smeets, Michael Wagner b.Reconstructed Scope: For sure, only Mary passed this time ∀ p' ∈ Alt, ¬ p' unless p' = p.'It is definitely the case that only Mary passed.'Surface Scope, unavailable: 'It is only the case that Mary claimed that Jan failed.'Reconstructed Scope, available: 'Maria claimed again that only Jan failed.'