CHARACTERIZING CONCEALED-QUESTION DPS By Zachary Wayne Schmidt A Thesis Submitted to the Honors College In Partial Fulfillment of the Bachelor s Degree With Honors in Linguistics THE UNIVERSITY OF ARIZONA M A Y 2 0 1 7 Approved by: Dr. Heidi Harley Department of Linguistics
Abstract Concealed questions (CQs) are DPs that function as embedded questions. Although various attempts have been made to establish which DPs can and cannot be CQs, none of these attempts has been completely successful. I argue that the constraints on CQ DPs are the same as the constraints on subjects of embedded specificational wh- questions. I show that one apparent exception, the requirement that a specificational subject be topical, is a matter of felicity rather than grammaticality, but the existence of nontopical CQs does offer insight into the structural difference between CQs and embedded questions. I conclude by proposing that a CQ is a small clause whose predicative DP has risen into the matrix clause to produce a specificational structure. Acknowledgements I d like to thank my thesis adviser, Dr. Heidi Harley, for her wise advice, keen judgment, and good sense of humor.
2 Contents 1 Introduction 1.1 Concealed Questions... 3 1.2 Nathan s Approach... 5 2 Constraints on Concealed Questions 2.1 Specificational Sentences... 8 2.2 CQs Behave like Specificational Subjects... 10 3 Concealed Questions as Small Clauses 3.1 The Problem of Bare Indefinites... 15 3.2 The Small-Clause Analysis of CQs... 17 4 Conclusions... 19
3 1. Introduction 1.1. Concealed Questions A concealed question (CQ) is a determiner phrase that shares its meaning, as well as certain distributional properties, with an embedded question. Many questionembedding verbs, like figure out in (1) and tell in (2), can take CQ complements. As (1) demonstrates, the DP the winner from (1a) is semantically and syntactically interchangeable with the embedded question who the winner was from (1b) as the complement of figure out. Therefore, the winner is a CQ. Similarly, (2) shows the CQ the president of Russia in (2a) alternating with the embedded question who the president of Russia is in (2b) as the complement of tell. (1) a. Mary figured out the winner. b. Mary figured out who the winner was. (2) a. Tell me the president of Russia. b. Tell me who the president of Russia is. There are, of course, multiple ways to paraphrase any given CQ using an embedded question. In (1b), for example, who the winner was could be replaced by who won without significantly changing the meaning of the sentence. But I argue here that embedded questions of the form who the winner was and who the president of Russia is
4 show an especially close relationship to CQs, in that, with one class of exceptions, every CQ X has a semantically equivalent corresponding embedded question of the form who/what X is/was. Where no embedded question of the appropriate type exists, a CQ is impossible, so the ungrammaticality of *Tell me the weather follows from the ungrammaticality of *Tell me what the weather is, even though the similar Tell me what the weather is like is grammatical. The only exceptions to this generalization are indefinite unmodified CQs. For instance, a saint is an acceptable CQ in Tell me a saint, but who a saint is is strongly infelicitous in #Tell me who a saint is. This contrast is a problem for any analysis of concealed questions that relies on their similarity to wh- questions. In 3, I will use a suggestion in Mikkelsen (2005) about indefinite topics to propose a solution. Finally, although almost every CQ X can be replaced with an embedded question of the form who/what X is/was, there are many embedded questions of the form who/what X is/was that cannot be replaced by a concealed question X. In 2, I point out that questions of the form what X is are the surface realizations of at least two different underlying structures and show that CQs are closely related to just one of those structures, namely, the specificational one. This fact, combined with Mikkelsen s (2005) work on specificational subjects, makes it possible to thoroughly characterize the kinds of DPs that can and cannot be CQs. In 3 I will offer a syntactic analysis of CQs as small clauses. First, though, I will examine one of the most familiar descriptions of the constraints on concealed-question DPs.
5 1.2. Nathan s Approach Nathan (2005) divides concealed-question DPs into two classes: those that are headed by relational nouns and those that are not 1. He points out that relational nouns seem to be naturally capable of heading CQ noun phrases. See (1) and (2) for examples. By contrast, non-relational nouns like city seem to require particular kinds of modification in order to receive CQ interpretations (pp. 292, 293). Nathan uses the following examples 2 to illustrate this point: (3) a. #Tell me the large city (of Vermont). b. Tell me the largest city in Vermont c. Tell me a city you visited last month. By itself or with an adjective as in (3a), city is awkward as a CQ. Nathan argues that the CQs in (3b) and (3c) are licensed, respectively, by the superlative head -est and by the relative clause you visited last month. On the other hand, a relational noun, say capital, can appear in the same context without modification: Tell me the capital of Vermont. But this generalization is not perfect. Nathan never claims that all relational nouns can head CQ noun phrases, but it is still worth mentioning that some relational nouns, including destruction and life, cannot. 1 Nathan uses the term functional instead of relational. I use relational because it better describes nouns like capital, which denotes a relation that is not a function from one set of individuals to another (Bolivia, for instance, has more than one capital). 2 (3a), (b), and (c) are Nathan s (7a), (9a), and (9b), respectively. Nathan marks his (7a) with a star.
6 (4) Not all relational nouns can head CQ NPs: a. *Mary learned the destruction of the city. b. *Mary learned Evan s life. Furthermore, many non-relational nouns can head CQ NPs without the type of modification that Nathan claims that such nouns require. In fact, I and other Englishspeakers I have consulted disagree with several of Nathan s judgments about the grammaticality of non-relational CQs. He marks both Tell me a city in Vermont and Tell me a good restaurant (p. 292) ungrammatical, but they are acceptable to me and my consultants, as are their completely unmodified counterparts Tell me a city and Tell me a restaurant. (5) Non-relational nouns can head CQ NPs unmodified: a. Tell me a city (in Vermont). b. Tell me a (good) restaurant. I might say Tell me a city (in Vermont) if I needed help with a crossword puzzle, and I might say Tell me a (good) restaurant if I had just arrived in my interlocutor s hometown and I wanted to eat somewhere. In light of the acceptability of the non-relational CQ NPs in (5a) and (b), the unacceptability of the large city in (3a) seems anomalous. The key difference here is that the large city is definite whereas both a city (in Vermont) and a (good) restaurant are indefinite. The definite determiner requires an NP that, together with the discourse context, picks out a unique individual. In the absence of a discourse context, an NP headed
7 by a non-relational noun almost never satisfies this requirement. For example, #Mary likes the large city is infelicitous because large city, a non-relational NP, might refer to any large city in the world. This is not to say that #Mary likes the large city is impossible, though; it is fine, for instance, in a contrastive-focus-inducing context: Mary commutes between two cities, one small and one large. Mary likes the LARGE city. If (3a) really is infelicitous for the same reason that #Mary likes the large city is, it should be possible to license (3a) using contrastive focus. It turns out that it is. For instance: Vermont has exactly two cities, one large and one small. Tell me the LARGE city. Nathan s (10a) (p. 293), reproduced below as (6), presents a similar case. (6) #Sam wanted to know the responsible person. Like (3a), (6) can be salvaged by contrastive focus: There are two people, one responsible and one irresponsible. Sam wanted to know the RESPONSIBLE person. The fact that contrastive focus licenses definite non-relational CQs like those in (3a) and (6) is more evidence that definite CQs behave exactly like regular definite DPs in terms of the relational/non-relational distinction. Furthermore, the large city and the responsible person can be added to the growing list of examples of grammatical non-relational CQs. Unlike non-relational nouns, relational nouns can often pick out unique individuals with little or no discourse context. For instance, the NP capital of Vermont can only refer to Montpelier, so the capital of Vermont is perfectly felicitous with or without context. Modifiers like superlatives and relative clauses act similarly; (3b) s largest city in Vermont, for example, is unlikely to have more than one possible referent, so it is acceptable in an out-of-the-blue definite DP. What the special modifier licenses is
8 not a CQ reading but the presence of the definite determiner. Notice that, since the CQ a city you visited last month in (3c) is indefinite, it would be acceptable even without the relative clause (see (5a)). Replacing a with the produces the equally acceptable sentence Tell me the city you visited last month, in which the relative clause makes it easy to reconstruct a definite-determiner-licensing context. To summarize: (1), (2), and (5) show that both relational and non-relational nouns can head CQ NPs. The CQs in (3a) and (6) are infelicitous not because they are nonrelational CQs but because they are non-relational definite DPs. And (4) shows that even some relational nouns do not make good CQs. It is safe to conclude that relationality does not reliably distinguish those nouns that can head CQ NPs from those that cannot. I will propose a different distinguishing characteristic in the next section. 2. Constraints on Concealed Questions 2.1. Specificational Sentences In the next subsection, I will show that the embedded wh- question to which a CQ corresponds must be specificational. Here I will introduce the concept of specificational sentences and develop some methods of testing sentences for specificationality. A specificational sentence is a sentence of the form DP is DP in which the order of the two DPs is intuitively backwards, with the first DP denoting a property and the second denoting an individual. For instance, The winner is Evan is specificational because the property-denoting the winner precedes the individual-denoting Evan. Reversing the DPs produces Evan is the winner, in which the DPs appear in the standard order. Sentences like Evan is the winner are called predicational.
9 According to Mikkelsen s (2005) analysis of copular sentences (see pp. 164-186 in particular), predicational Evan is the winner and specificational The winner is Evan are both constructed around the same small-clause core: [PredP Evan the winner]. PredP requires its DPs to be in the standard order, so *[PredP the winner Evan] is impossible. In unmarked cases, the individual-denoting DP (Evan) moves to Spec-TP to produce a predicational sentence: Evan is [PredP Evan the winner]. However, in a case where the winner is topical, Evan is not topical, and T bears an uninterpretable topic feature, the winner must move to Spec-TP instead, yielding a specificational sentence: The winner is [PredP Evan the winner]. Figure 1. On the left, the structure of a predicational sentence; on the right, the structure of the corresponding specificational sentence. These structures come from Mikkelsen (2005, pp. 170, 172). I have left out the vp between PredP and TP. I will call a DP predicative if it is generated as the complement of Pred and referential if it is generated as the specifier of Pred. For instance, in I consider [PredP Evan the winner], Evan is referential and the winner is predicative. Echo questions can be used to test whether a wh- word is capable of functioning as a predicative or referential DP. For example, the grammaticality of You consider [PredP WHO the winner]? shows that who can serve as the subject of a small clause, which means
10 that who can be referential. On the other hand, the ungrammaticality of *You consider [PredP Evan WHO]? shows that who cannot be predicative. Unlike who, what can be either predicative or referential, as demonstrated by You consider [PredP WHAT the capital of Russia]? and You consider [PredP Moscow WHAT]? But, when used referentially, what is inanimate. The inanimateness of referential what is responsible for the infelicity of sentences like #You consider [PredP WHAT the winner]?, in which the subject of the small clause is inanimate and the predicate is expected to be animate. No such restrictions apply to predicative what, which appears freely in both inanimate small clauses like You consider [PredP Moscow WHAT] and animate ones like You consider [PredP Evan WHAT]?. 2.2. CQs Behave like Specificational Subjects In this subsection I will show that the same constraints that determine which DPs make acceptable CQs also determine which DPs make acceptable subjects of specificational wh- questions 3. If we assume that the embedded question to which a CQ corresponds must be specificational, then its subject must be predicative and its wh- word must be referential. Then, for each of the following examples, the properties of the first sentence explain the ungrammaticality of the second: (7) Subject of embedded question is not predicative: a. Mary learned who Evan is. b. *Mary learned Evan. 3 For an alternative analysis that is also based on the similarity between CQs and specificational subjects, see Romero (2005).
11 (8) Wh- word is not referential: a. Mary learned what the winner is, namely, an American. b. *Mary learned the winner, namely, an American. In (7a), since the embedded question s subject, Evan, is a proper noun, we can be sure that it is non-predicative. (Notice that it is ungrammatical as the predicate of a small clause: *I consider [PredP him Evan].) As a result, who Evan is is not specificational 4 and (7b) s Evan cannot be a CQ. In (8a), the embedded question s wh- word is what. Furthermore, the winner is animate, and, as established in 2.1, whenever what appears in an animate context, it must be non-referential. So (8a) is not specificational and (8b) s the winner cannot be a CQ. I have included namely, an American in (8a) and (b) to indicate that what Mary learned was not the winner s identity but the winner s nationality. (8b) becomes grammatical when the sentence-final DP is, for instance, a proper name. In this case, the wh- word of the relevant embedded question must be referential in order to match the provided answer. (9) Wh- word is referential: a. Mary learned who the winner is, namely, Evan. b. Mary learned the winner, namely, Evan. 4 In fact, since who and Evan are both non-predicative, (6a) is neither specificational nor predicational.
12 This is exactly what we would expect; (9a) s who the winner is has a predicative subject and a referential wh- word, so (9a) is specificational. Therefore, (9b) s the winner is a good CQ. So far, I have only shown that every CQ corresponds to the predicative DP in an embedded copular who/what question. Conceivably, the embedded question could be predicational, as in (10a). (10) a. Predicational: I told Mary who s a good lawyer. b. Specificational: I told Mary who a good lawyer is. c. CQ: I told Mary a good lawyer. But (10a) is exhaustive; it means something like For every relevant person X, I told Mary whether X is a good lawyer. On the other hand, (10b) and (c) both mean There is a good lawyer X such that I told Mary that X is a good lawyer. This difference in meaning produces contrasts like the one between #I told Mary who s a good lawyer, and she still doesn t know Evan s a good lawyer and I told Mary {a good lawyer / who a good lawyer is}, and she still doesn t know Evan s a good lawyer. At least with respect to exhaustivity, specificational wh- questions and CQs pattern together to the exclusion of predicational wh- questions. Evidence from quantified CQs 5 also supports an analysis under which CQs correspond to specificational rather than predicational structures. Unlike their 5 For an in-depth study of quantified CQs, see Frana (2013).
13 specificational counterparts 6, predicational who/what questions with quantified predicative DPs are generally unacceptable, as illustrated by (11a). (11) a. Predicational: #Tell me who s every good lawyer. b. Specificational: Tell me who every good lawyer is. c. CQ: Tell me every good lawyer. Clearly, the CQ in (11c) can only be paraphrased by the specificational construction in (11b). A final line of evidence suggesting an unusually deep connection between CQs and specificational questions comes from the idea of a conceptual cover. Aloni (2001) informally defines a conceptual cover as a method of identification (p. x). More formally, a conceptual cover is a set of individual concepts such that, in each world, each individual constitutes the instantiation of one and only one concept (p. 16). Specificational sentences require their predicative DPs to describe sets for which conceptual covers exist. For instance, there is no conventional way of uniquely identifying a rock, so a DP headed by rock cannot be the subject of a specificational sentence. (12) a. Specificational: *I figured out what the rock that Evan threw was. 6 Mikkelsen (2005, pp. 112, 113) shows that strongly quantified DPs cannot be specificational subjects in declarative sentences (e.g. *Every good lawyer is Mary or Evan). This result can be extended to nonembedded wh- questions (e.g. *Who s every good lawyer?), so it is unclear why who every good lawyer is is an acceptable embedded question.
14 b. CQ: *I figured out the rock that Evan threw. Again, the ungrammaticality of the CQ seems to follow from the ungrammaticality of the specificational sentence. The observation that CQs require conceptual covers is due to Schwager (2008, in particular pp. 6, 14). Many relational nouns, too, are incapable of heading CQs for lack of a conceptual cover. Destruction and life from (4b) and (c) fall into this category, as does basement in (13b). (13) a. Specificational: *Tell me what the basement of the hotel is. b. CQ: *Tell me the basement of the hotel. Incidentally, Nathan s (2005) theory would require elaboration in order to explain the ungrammaticality of (12b) and (13b), since (12b) s the rock that Evan threw contains a relative clause and (13b) s the basement of the hotel is a relational DP. Both types of DP make good concealed questions according to Nathan s account. Having examined the close relationship between specificational who/what questions and CQs, I will consolidate my observations. Since every concealed-question DP is capable of serving as the subject of a specificational sentence, a DP must meet the following two requirements in order to be a viable CQ: Constraint A: The DP must be capable of being predicative.
15 Constraint B: There must be a conceptual cover available for the identification of individuals that meet the the description of the DP. Together, Constraints A and B can explain all of the impossible CQs that I have mentioned so far. Constraint A accounts for (7b) s Evan. Constraint B accounts for (4a) s destruction, (4b) s life, (12b) s rock and (13b) s basement. (8b) s the winner is not actually impossible as a CQ, so it does not need to be ruled out by either requirement. Furthermore, A and B do not rule out anything they shouldn t in particular, the wide range of non-relational nouns like restaurant, city, and lawyer that can head CQs unmodified. In 3, I will try to give a syntactic explanation of the relationship that gives rise to Constraints A and B. 3. Concealed Questions as Small Clauses 3.1. The Problem of Bare Indefinites The close correspondence between concealed questions and embedded specificational wh- questions suggests that there might be some syntactic connection between the two. Indeed, the earliest account of CQs, found in Baker (1968), treated them as wh- questions in which everything but the subject DP is unpronounced. Under this analysis, Mary learned the president of Russia would be the phonetic realization of Mary learned who the president of Russia is. Even if we assume that only specificational embedded wh- questions can undergo this partial deletion, the deletion account has two major problems: it incorrectly predicts
16 that any question-embedding predicate should also be able to embed CQs, as shown in (14), and it fails to predict the acceptability of indefinite unmodified CQs like Tell me a lawyer, as shown in (15). (14) Grammatical embedded question, ungrammatical CQ: a. Embedded question: Mary wonders who the president of Russia is. b. CQ: *Mary wonders the president of Russia. (15) Infelicitous embedded question, felicitous CQ: a. Embedded question: #Tell me who a saint is. b. CQ: Tell me a saint. (14) could, of course, be explained by saying that some predicates idiosyncratically allow the questions they embed to undergo partial deletion. This is not a satisfying explanation, since it involves an extremely specific deletion rule attested nowhere else in English. I prefer the explanation defended in Nathan (2005), according to which a predicate embeds CQs if and only if it embeds both questions and propositions, but for the remainder of this paper I will be more concerned with explaining the data in (15). The first thing to notice about (15a) is that its unacceptability vanishes when an adjective is included in the subject of the embedded question: Tell me who a famous saint is. In fact, Mikkelsen (2005, p. 157) observes that all indefinite specificational subjects require their head nouns to be modified in some way. For instance, #A saint is Joan of Arc is infelicitous but A famous saint is Joan of Arc is much better. The more information an indefinite specificational subject contains, the better it sounds.
17 Mikkelsen (2005) concludes that bare indefinites cannot be specificational subjects because they do not contain enough discourse-old material (p. 158) to be felicitous topics. In her analysis, a specificational sentence is the result of a predicative DP moving to subject position because of its topic feature. Therefore, a non-topical DP can never be a specificational subject. Since bare indefinites cannot be topics, they cannot be specificational subjects. But bare indefinites can be concealed questions, as in (15b). Under Mikkelsen s analysis, the only possible explanation of (15b) is that the DP in a CQ is not required to be a topic. Therefore, whatever the structure of a CQ is, it must not include a topic phrase. 3.2. The Small-Clause Analysis of CQs With the conclusion of 3.1 in mind, I propose the following structure for concealed questions, using the winner as an example: Figure 2. The small clause that gives rise to a concealed question. Here I follow Mikkelsen (2005) in treating the small clause as a PredP. Under Mikkelsen s analysis, both the referential DP and the predicative DP in PredP have unvalued case features (p. 170). Since PRO appears only in caseless environments, I will assume that it does not have an unvalued case feature.
18 The structure in Figure 2 lacks a topic phrase, so it correctly predicts the availability of bare indefinite CQs like the one in (15b). However, in two other respects, the small-clause analysis of CQs appears to be deficient: first, it does not have a specificational structure, and, second, it is unclear what semantic interpretation to assign to PRO. Both problems disappear if the predicative DP moves out of PredP. Hong and Lasnik (2010) show that the subject of a small clause in object position obligatorily raises to Spec-AgrOP. Assuming that this movement is motivated by an uninterpretable accusative case feature on AgrO, PRO would not move since it cannot receive case. The only other DP that can move is the predicative one. As a result, the predicative DP ends up c-commanding the referential DP, as it would in a specificational sentence. The case feature on AgrO plays the same role here that the topic feature on T plays in a real specificational sentence. Thus, we end up with a specificational structure in which the predicative DP need not be a topic. Figure 3. A portion of the structure of know the winner, where the winner is a CQ. The predicative DP the winner has moved from Comp-PredP to Spec-AgrOP, where it c- commands the referential DP PRO.
19 Furthermore, since the predicative DP now c-commands PRO, it can bind PRO. PRO s being bound by the predicative DP explains why a sentence like Mary told Evan the winner means Mary told Evan that the winner is X, where X must refer to the actual winner. In a CQ, the equivalent of X is PRO, which is guaranteed to refer to the actual winner because its referent is established by the DP the winner. This is why tell behaves like a factive verb when it takes a CQ complement. In addition to explaining why CQs act like specificational subjects, the small-clause analysis of CQs, if correct, can be used to determine which syntactic contexts will and will not license CQs. At the very least, it predicts that all CQ-embedding verbs assign case and embed propositions. More work will have to be done to determine what other predictions this analysis makes and whether those predictions are borne out by the data. 4. Conclusions I have argued that a DP can serve as a CQ if and only if it meets two requirements, here called Constraints A and B, which also apply to subjects of embedded specificational wh- questions (see 2.2). In fact, I believe that Constraints A and B completely determine the set of grammatical specificational subjects 7. If this is the case, the result of 2 can be restated: a DP can serve as a CQ if and only if it can serve as a specificational subject. In an attempt to explain the relationship between CQs and specificational subjects, I have proposed that a CQ is actually a small clause whose predicative DP has 7 Although the set of felicitous specificational subjects is smaller, thanks to the topicality requirement mentioned in 3.
20 risen into the matrix clause to produce a specificational structure. If this analysis proves to be correct, it will have important consequences for the study of concealed questions, case, and small clauses.
21 References Aloni, Maria D. 2001. Quantification under conceptual covers. Amsterdam: Institute for Logic, Language and Computation. http://maloni.humanities.uva.nl/index.html. Baker, Carl L. 1968. Indirect questions in English. Urbana, Illinois: University of Illinois. Frana, Ilaria. 2013. Quantified concealed questions. Natural Language Semantics 21(2). 179 218. doi:10.1007/s11050-012-9089-y. Hong, Sungshim & Howard Lasnik. 2010. A note on Raising to Object in small clauses and full clauses. Journal of East Asian Linguistics 19(3). 275 289. doi:10.1007/s10831-010-9062-z. Mikkelsen, Line. 2005. Copular clauses: Specification, predication and equation. Amsterdam: John Benjamins Publishing Company. Nathan, Lance. 2005. The Interpretation of Concealed Questions. http://www.lingref.com/cpp/wccfl/24/paper1234.pdf. Romero, Maribel. 2005. Concealed Questions and Specificational Subjects*. Linguistics and Philosophy 28(6). 687 737. doi:10.1007/s10988-005-2654-9. Schwager, Magdalena. 2008. Keeping prices low: an answer to a concealed question. http://magdalena-kaufmann.uconn.edu/papers/schwagersub08.pdf.