Abstract

We present novel data showing that indexicals, first and second person pronouns in particular, occurring in a certain kind of attitude report in Uyghur are interpreted with respect to the reported context (indexical shifting). While previous authors report similar shifted interpretations of indexicals in languages such as Amharic and Zazaki, we observe a unique feature of Uyghur indexical shifting: it is sensitive to structural positions of the indexical item, and as a consequence can be partial. We account for the structural sensitivity of Uyghur indexical shifting with a context-shifting operator (or monster) that is syntactically independent from the embedding attitude predicate.

1 Introduction

Every natural language has words and phrases whose meanings refer to certain aspects of the context of utterance (e.g., I, you, here, now). These items are called indexicals. Kaplan (1977) made the important observation that the interpretation of indexicals is insensitive to modals, unlike that of definite descriptions. For example, while the definite description in (1a) has an interpretation under which its referent is not the actual department head, the indexical I in (1b) is obligatorily interpreted relative to the context of the current utterance, even though they are in the same modal contexts.

(1)

  • a.

    John thinks that the department head is a phonologist.

  • b.

    John thinks that I am a phonologist.

This rigidity of reference exhibited by indexicals led Kaplan to conjecture that indexicals are obligatorily dependent on the actual context of utterance. In other words, he claimed there is no operator in natural language that shifts the context. He called such supposedly nonexisting context-shifting operators monsters. To put it differently, if a monster did exist, indexicals under its scope would be interpreted relative to nonactual contexts and hence I could refer to somebody other than the speaker of the sentence, for example. We refer to such a phenomenon as indexical shifting.

Contrary to Kaplan’s (1977) surmise, however, recent studies demonstrate that many languages, including Amharic (Schlenker 1999, 2003, Anand 2006), Navajo (Schlenker 1999, Speas 2000), Zazaki (Anand and Nevins 2004, Anand 2006), Slave (Rice 1986, Anand and Nevins 2004, Anand 2006, Bittner 2012), Catalan Sign Language (Quer 2005), Nez Perce (Deal 2008, 2011), Matses (Ludwig et al. 2010, Munro et al. 2012), and Turkish (Gültekin Şener and Şener 2011), exhibit indexical shifting. In light of this, several authors including Schlenker (1999, 2003), Anand and Nevins (2004), and Anand (2006) have proposed that monsters do in fact exist.

In this article, we present novel data showing that a monster exists in attitude report constructions in (Modern) Uyghur (Turkic; North China and Kazakhstan), which exhibit peculiar properties that have hitherto been unobserved in other languages. In particular, indexical shifting in Uyghur is sensitive to structural positions of the indexical items, and as a consequence can be partial. We account for this with a monster operator that is syntactically independent from the embedding attitude verb.

The organization of the article is as follows. In section 2, we introduce the basic properties of Uyghur indexical shifting. We present our main observation about Uyghur in section 3. Section 4 contains the bulk of our analysis and the supporting evidence. Section 5 is devoted to validating the predictions of our theory. We conclude in section 6.

2 Uyghur Embedding and Indexical Shifting

Just as in other languages with indexical shifting, indexical shifting in Uyghur is confined to attitude report constructions. Attitude reports in this language can appear in one of two syntactic forms: nominalized complement clause (2a) and finite complement clause (2b).

(2)

  • a.

    Ahmet [profesor-ning kit-ken-lik-i-ni] di-di.

    Ahmet [professor-gen leave-rel-nmlz-3-acc] say-past.3

    ‘Ahmet said that the professor left.’

  • b.

    Ahmet [profesor ket-ti] di-di.

    Ahmet [professor.nom leave-past.3] say-past.3

    ‘Ahmet said that the professor left.’

In a nominalized complement clause, the main verb in the embedded clause (ket- ‘leave’ in (2)) does not bear a tense suffix. Instead, such a verb takes a relative clause suffix -ken, the nominalizer -lik, as well as the possessive agreement, which cross-references the person features of the embedded genitive subject. In such constructions, the entire embedded clause is morphologically casemarked; for instance, in (2a) the nominalized embedded clause bears the accusative marker -ni. The case marking of the embedded nominalized clause is determined by the embedding attitude verb; for example, de- ‘say’ selects for accusative complement clauses, qoshul- ‘agree’ selects for dative complement clauses, guman qik- ‘suspect’ selects for ablative complement clauses, and so on.

In contrast to the verb in nominalized complements, the embedded verb in finite complement clauses is fully tensed. Finite embedded clauses are grammatical as independent sentences, in contrast to nominalized complements, as shown in (3).

(3)

  • a.

    *Profesor-ning kit-ken-lik-i-ni.

    professor-gen leave-rel-nmlz-3-acc

    ‘The professor left.’

  • b.

    Profesor ket-ti.

    professor.nom leave-past.3

    ‘The professor left.’

Although nominalized and finite complement clauses can be used to convey similar meanings (e.g., (2a) and (2b) are synonymous), they exhibit different characteristics with respect to the interpretation of indexicals: in nominalized complement clauses, nominative subjects and verbal agreement are interpreted relative to the context of the matrix utterance (the nonshifted reading), whereas in finite complement clauses, indexical subjects and verbal agreement are interpreted relative to the reported context (the shifted reading). For instance, the following sentences cannot be used to report the same event, in contrast to the synonymous pair in (2):

(4)

  • a.

    Ahmet [mening kit-ken-lik-im-ni] di-di.

    Ahmet [1sg.gen leave-rel-nmlz-1sg-acc] say-past.3

    ✓(nonshifted) ‘Ahmet said that Ispeaker left.’

    ✗(shifted) ‘Ahmeti said that hei left.’

  • b.

    Ahmet [men ket-tim] di-di.

    Ahmet [1sg leave-past.1sg] say-past.3

    ✓(nonshifted) ‘Ahmet said that Ispeaker left.’

    ✗(shifted) ‘Ahmeti said that hei left.’

The same contrast obtains with second person indexicals, as shown in (5).

(5)

  • a.

    Tursun Muhemmet-ke [xet jaz-ghan-lik-ing-ni] di-di.

    Tursun Muhemmet-dat [letter write-rel-nmlz-2sg-acc] say-past.3

    ✓(nonshifted) ‘Tursun told Muhemmet that youhearer wrote a letter.’

    ✗(shifted) ‘Tursun told Muhemmeti that hei wrote a letter.’

  • b.

    Tursun Muhemmet-ke [xet jaz-ding] di-di.

    Tursun Muhemmet-dat [letter write-past.2sg] say-past.3

    ✓(nonshifted) ‘Tursun told Muhemmet that youhearer wrote a letter.’

    ✗(shifted) ‘Tursun told Muhemmeti that hei wrote a letter.’

Indexical shifting in finite embedded clauses is a general phenomenon in Uyghur and is not confined to speech reports under the verb de-mek ‘to say’, unlike what happens in other languages where shifted indexicals have been reported: verbs such as bil-mek ‘to think, believe, know’, sözle-mek ‘to say’, and ojla-maq ‘to believe’ all exhibit similar shifting properties. For expository purposes, we will present examples of indexical shifting mostly in the context of the verb demek; however, our conclusions apply to all attitude verbs that take complement clauses.1

Given the data above, one might wonder whether the finite complement clauses are simply obligatorily quotational: this would make the shifting facts above not at all surprising. Under this analysis, expressions enclosed in quotation marks are mentioned rather than used, and hence indexicals in quotations superficially look shifted (see Kaplan 1977). However, there is evidence demonstrating that finite clausal complements in Uyghur need not be quotational, while the indexical shifting properties remain constant. The evidence we will present in favor of a nonquotational analysis comes from (i) long-distance wh-questions, (ii) long-distance licensing of negative items, and (iii) nonverbatim reports (see Schlenker 1999, Anand 2006, Oshima 2006).

The first piece of evidence that finite clauses exhibiting indexical shifting need not be quotations comes from wh-questions. One hallmark of quotations is that they resist quantifying-in (Quine 1960, Cappelen and Lepore 2012). One example of this restriction is the unavailability of quotation-internal wh-phrases that take scope outside the quotation, as in (6).

(6) *What did John say ‘‘I saw t’’?

Uyghur finite embedded clauses can contain a wh-phrase taking scope in the matrix clause. Indexicals in such finite embedded clauses still receive a shifted interpretation, as in (7).

(7)

  • Tursun [men kim-ni kör-dim] di-di?

  • Tursun [1sg who-acc see-past.1sg] say-past.3

  • ‘Who did Tursuni say hei saw?’

Even though the embedded clause in (7) cannot be a quotation, the first person subject still refers to the attitude holder.

Further evidence against an obligatory quotation interpretation comes from negative items. In Uyghur, negative items require a licenser, such as negation, in the same clause or a higher clause. In finite complement clauses, negative items can be licensed long-distance by negation in the matrix clause while embedded indexicals remain shifted, as in (8).

(8)

  • Tursun [men hichkim-ni kör-dim] di-mi-di.

  • Tursun [1sg nobody-acc see-past.1sg] say-neg-past.3

  • ‘Tursuni didn’t say that hei saw anyone.’

This demonstrates that the embedded clause in (8) could not be a quotation because, on its own, the embedded sentence is ungrammatical, as demonstrated by (9).2

(9)

  • *Men hichkim-ni kör-dim.

  • 1sg nobody-acc see-past.1sg

  • ‘I saw no one.’

The fact that embedded clauses can be nonverbatim also demonstrates that finite embedded clauses need not be quotations, as shown in (10).

(10)

  • Context: Ahmet and Muhemmet are two students who took a test in class today. After the test, I met Ahmet. He said, ‘‘(I) passed the test’’ (men) imtihandin öttim. A while later, I met Muhemmet, who said exactly the same thing. I can report on this situation to Aygül as follows:

  • Ahmet we Muhemmet [(biz) imtihan-din öt-tuq] di-di.

  • Ahmet and Muhemmet [(we) test-abl pass-past.1pl] say-past.3

  • ‘Ahmet and Muhemmet said that they passed the test.’

In (10), the embedded speech report differs from the original speech in having plural embedded subject and agreement; while the original claims of test-passing are made individually, in (10) these claims are reported collectively. The fact that the speech report in (10) differs from the original speech demonstrates that the embedded clause is not a quotation. Nonetheless, the pronominal indexicals remain shifted.

Note that a partial quotation analysis is not available for the situation described in (10): the original utterance, made separately by Ahmet and Muhemmet, featured a singular pronoun and singular agreement. In the speech report, both the pronominal subject and the verbal agreement are plural. In partial quotations, we would expect some phrases, but not others, to receive a shifted interpretation. We would not expect different φ-features of the same phrase to vary with respect to whether they are interpreted relative to the matrix or embedded context.

The data presented in this section demonstrate that indexical shifting in Uyghur is observed outside of quotations. In the next section, we introduce the main syntactic puzzle that Uyghur indexical shifting presents. The solution to this puzzle will provide a unique means of interrogating the syntax of indexical shifting, as we will show in later sections.

3 Puzzle: Partial Indexical Shifting

We begin with the observation that in finite embedded clauses, but not in matrix clauses, subjects can bear either nominative or accusative case.

(11)

  • a.

    Ahmet [profesor-{∅/ni} ket-ti] di-di.

    Ahmet [professor-{nom/acc} leave-past.3] say-past.3

    ‘Ahmet said that the professor left.’

  • b.

    Profesor-{∅/*ni} ket-ti.

    professor-{nom/*acc} leave-PAST.3

    ‘The professor left.’

Recall that as we showed in (4), nominative subjects of finite complement clauses are interpreted obligatorily with respect to the embedding context, that is, exhibit a shifted interpretation. In this respect, nominative embedded subjects differ from accusative embedded subjects, which cannot shift. This is illustrated in (12a–b).

(12)

  • a.

    Ahmet [men ket-tim] di-di.

    Ahmet [1sg.nom leave-past.1sg] say-past.3

    ✗(nonshifted) ‘Ahmet said that Ispeaker left.’

    ✓(shifted) ‘Ahmeti said that hei left.’

  • b.

    Ahmet [meni ket-ti] di-di.

    Ahmet [1sg.acc leave-past.3] say-past.3

    ✗(nonshifted) ‘Ahmet said that Ispeaker left.’

    ✓(shifted) ‘Ahmeti said that hei left.’

These examples feature a first person embedded subject: shifted when nominative, nonshifted when accusative. This contrast between embedded subjects bearing nominative case and embedded subjects bearing accusative case is not limited to first person: accusative second person subjects cannot shift either. This also contrasts with nominative second person subjects, which must shift.

(13)

  • a.

    Ahmet [sen ket-ting] di-di.

    Ahmet [2sg.nom leave-past.2sg] say-past.3

    ✗(nonshifted) ‘Ahmet said that youcurrent hearer left.’

    ✓(shifted) ‘Ahmet said that heoriginal hearer left.’

  • b.

    Ahmet [seni ket-ti] di-di.

    Ahmet [2sg.acc leave-past.3] say-past.3

    ✗(nonshifted) ‘Ahmet said that youcurrent hearer left.’

    ✓(shifted) ‘Ahmet said that heoriginal hearer left.’

To our knowledge, the relevance of syntactic factors like morphological case to indexical shifting has not been observed elsewhere. As we show in the sections that follow, this feature of Uyghur provides a unique window into the syntax of indexical shifting. First, we spell out our analysis of the clausal structure of finite complements. This is the task of the next section.

4 Proposal: Bipartite Structure of Finite Complements

Following Schlenker (1999, 2003), Anand and Nevins (2004), and Anand (2006), among others, we assume that a monster is syntactically present in Uyghur finite attitude report constructions, and that this operator is responsible for shifted interpretation of indexicals.

First, in order to account for the difference between finite and nominalized complement clauses, we assume that the Uyghur monster appears only and always in finite complement clauses. With Anand and Nevins, but contra Schlenker, we assume that the monster is independent from the attitude verb itself. An attitude construction is assumed to be licit with or without the monster. Thus, when and only when the monster is present, indexical shifting takes place. See Anand and Nevins 2004, Anand 2006, and Sudo 2012 for semantic details and for empirical motivation for this analysis. In this setting, the difference between nominalized and finite attitude complement clauses in Uyghur is simply analyzed as the presence versus absence of the monster.

Second, we claim that the monster partitions the embedded clause into two parts: the part where indexicals shift and the part where indexicals do not shift. Since the monster is the operator that shifts the context, everything in its scope must shift. Certain constituents, however, can be located outside the scope of the monster (structurally higher than the monster) and therefore do not receive shifted interpretation. Thus, the basic clause structure would look like (14) (with the monster symbolized by ).

(14)

graphic

Recall that in section 3, we showed that embedded accusative subjects never undergo indexical shifting, while embedded nominative subjects always shift. To account for this, we claim that accusative embedded subjects are always structurally higher than the monster, whereas nominative embedded subjects always remain in the scope of the monster operator.

In the remainder of this section, we present evidence to support this analysis. In section 4.1, we show that accusative indexical subjects are base-generated in the lower clause, just like nominative subjects. In section 4.2, we present evidence that accusative subjects are structurally higher than nominative subjects. In section 4.3, we show that accusative subjects are able to be interpreted within the scope of the embedding attitude verb.

4.1 Accusative Subjects Are Embedded Subjects

Given the nonshiftability of accusative subjects, one might be tempted to analyze them as proleptic arguments of the attitude verbs, just like the of-phrase in John said of Mary that she left. In pro-drop languages like Uyghur, this analysis seems tenable at first sight, and in fact, such an analysis has been proposed for the accusative subjects of some constructions in syntactically similar languages (see Bruening 2001 on Japanese). However, we will present five pieces of evidence suggesting that accusative embedded subjects in Uyghur are generated in the lower clause: sentential idioms, licensing of negative items, adverb placement, the double accusative constraint, and accusative case assignment. All of these data suggest that accusative subjects can be generated in the lower clause, and the argument about the double accusative constraint furthermore suggests that they must be generated in the lower clause.

The first piece of evidence against the proleptic analysis comes from idiomatic interpretation: sentential idioms retain their idiomatic meaning even when the subject is marked accusative.3

(15)

  • a.

    Toqquz qiz-ning tolghaq teng kel-di.

    nine girl-gen labor together arrive-past.3

    ‘Times are hard.’ (literally: ‘Nine girls’ labor pains came all at once’)

  • b.

    Tursun [toqquz qiz-ning tolghaq-ni teng kel-di] di-di.

    Tursun [nine girl-gen labor-acc together arrive-past.3] say-past.3

    ‘Tursun said that times are hard.’

If the accusative embedded subject were thematically related to the matrix verb (or not generated in the lower clause), we would expect (15b) to be ungrammatical.

The second piece of evidence against the proleptic analysis comes from negative item licensing. We showed earlier that negative elements in Uyghur require a licenser in the same or a higher clause, as demonstrated again in (16).

(16)

  • Men hichkim-ni kör-*(mi)-dim.

  • 1sg.nom nobody-acc see-*(neg)-past.1sg

  • ‘I didn’t see anybody.’

(17) shows that negative accusative subjects can be licensed by embedded negation, demonstrating that embedded accusative subjects are located in the lower clause at least at some point in the derivation.

(17)

  • Ahmet [hichkim-ni ket-mi-di] di-di.

  • Ahmet [nobody-acc leave-neg-past] say-past.3

  • ‘Ahmet said that nobody left.’

Third, adverbial material belonging to the embedded clause can appear to the left of the accusative embedded subject, as in (18).

(18)

  • Tünügün Ahmet manga [ete Aygül-ni ket-idu] di-di.

  • yesterday Ahmet 1sg.dat [tomorrow Aygül-acc leave-impf.3] say-past.3

  • ‘Yesterday Ahmet said that Aygül would leave tomorrow.’

In (18), the adverb ete ‘tomorrow’ is located to the left of the embedded accusative subject. This adverb cannot be a part of the matrix clause in this case, as the matrix clause already contains a temporal adverb. This entails that ete is part of the embedded clause, and so is the embedded accusative subject.4

Fourth, there is a language-specific constraint banning two accusative NPs in the same clause (Halpert 2009).5 The effects of this constraint can be seen with causativized ditransitives that disallow accusative objects.

(19)

  • a.

    Muhemmet Aygül-ge gül-(ni) ber-di.

    Muhemmet Aygül-dat flower-(acc) give-past.3

    ‘Muhemmet gave Aygül a flower.’

  • b.

    Men Muhemmet-ni Aygül-ge gül-(*ni) ber-güz-dim.

    1sg Muhemmet-acc Aygül-dat flower-(*acc) give-caus-past.1

    ‘I made Muhemmet give Aygül a flower.’

For the evaluation of this constraint, the accusative subject is treated as belonging to the embedded clause. Specifically, the contrast between (20a) and (20b) shows that accusative subjects are licit in the presence of unmarked objects in the lower clause; however, if the lower clause contains an accusative direct object, an accusative embedded subject is ungrammatical. (20c) demonstrates that when the embedded object bears nonaccusative case, an accusative subject is once again available.

(20)

  • a.

    Tursun [meni nan yaq-ti] di-di.

    Tursun [1sg.acc bread bake-past.3] say-past.3

    ‘Tursun said that Ispeaker made bread.’

  • b.

    *Tursun [meni nan-ni yaq-ti] di-di.

    Tursun [1sg.acc bread-acc bake-past.3] say-past.3

    ‘Tursun said that Ispeaker made bread.’

  • c.

    Tursun [meni imtihan-din öt-ti] di-di.

    Tursun [1sg.acc test-abl pass-past.3] say-past.3

    ‘Tursun said that Ispeaker passed the test.’

The double accusative case constraint demonstrates that accusative subjects are treated as part of the embedded and not the matrix clause, providing another argument against the prolepsis analysis of Uyghur embedded accusative subjects.

Fifth, it can be shown that the embedded subjects receive accusative case within the embedded clause, and hence that the accusative case assignment does not depend on an assigner in the matrix clause, as it would in a prolepsis analysis. We begin by demonstrating that Uyghur passive verbs do not assign accusative case.6

(21)

  • a.

    Doxtur Ahmet-ni kör-di.

    doctor Ahmet-acc saw-past.3

    ‘A doctor saw Ahmet.’

  • b.

    Doxtur teripidin Ahmet-(*ni) kör-el-di.

    doctor by Ahmet-(*acc) saw-pass-past.3

    ‘Ahmet was seen by a doctor.’

This fact also holds for nominalized complements.

(22)

  • a.

    Ahmet manga [Aygül-ning kit-ken-lik-i-ni] di-di.

    Ahmet 1sg.dat [Aygül-gen leave-rel-nmlz-3-acc] say-past.3

    ‘Ahmet told me that Aygül left.’

  • b.

    Manga [Aygül-ning kit-ken-lik-i-(*ni)] di-el-di.

    1sg.dat [Aygül-gen leave-rel-nmlz-3-(*acc)] say-pass-past.3

    ‘I was told that Aygül left.’

Nonetheless, as (23) shows, an embedded subject can receive accusative case even when the embedding verb is passivized. This demonstrates that the embedding verb is not the locus of accusative case assignment.

(23)

  • Manga [Aygül(-ni) ket-ti] di-el-di.

  • 1sg.dat [Aygül(-acc) leave-past.3] say-pass-past.3

  • ‘I was told that Aygül left.’

These data suggest that a nonproleptic parse is always available for the accusative embedded subject. However, as an anonymous reviewer points out, this does not necessarily rule out a dual parse, whereby some accusative embedded subjects are proleptic while others are not. This is not at all implausible, given proposals such as Bruening’s (2001): namely, that Japanese and Passamaquoddy raising-to-object constructions alternate between raising and prolepsis. As the reviewer correctly notices, some of the evidence presented in the next section (adduced in support of the hypothesis that the accusative embedded subject is not part of the matrix clause) rules out a proleptic analysis. There is additional evidence against the availability of a proleptic parse from embedded-pronoun data. English copy-raising constructions (Rogers 1971, Potsdam and Runner 2001) require a pronoun or some kind of resumptive element (Heycock 1994) in the embedded clause (24a). The same is also true of more ‘‘traditional’’ proleptic constructions of the believe of type (24b).

(24)

  • a.

    *Richard seems like is in trouble.

  • b.

    *I believe of Mary that is a genius.

Other types of analysis (see Bruening 2001; though see Tanaka 2002 for objections) show that a possible distinguishing feature of prolepsis (in contrast to movement) is the availability of a resumptive pronoun in the embedded clause. Therefore, according to this analysis, the possibility that a resumptive element can occur in the lower clause can be used as a diagnostic of prolepsis. With this in mind, we observe that subject pronouns are not licit in combination with accusative embedded subjects in Uyghur.

(25)

  • *Ahmet meni u ket-ti di-di.

  • Ahmet 1sg.acc 3sg.nom leave-past.3 say-past.3

  • ‘Ahmet said that Ispeaker left.’

We conclude, then, that embedded accusative subjects are base-generated in the embedded clause regardless of case; in other words, embedded subjects are never proleptic.

4.2 Accusative Subjects Are Structurally Higher Than Nominative Subjects

Having demonstrated that accusative embedded subjects originate in the embedded clause, we proceed to show that subjects bearing accusative case are structurally higher than nominative subjects. The first piece of evidence comes from binding facts. The examples in (26) show that an embedded reflexive subject can be coreferential with the matrix subject only when it bears accusative case. In these examples, we enclose the embedded subjects in ‘only’-phrases in order to force the use of the pronoun, which otherwise tends to be omitted.

(26)

  • a. Meni [peqet özi-em-ni-la nan ye-men] di-dim.

  • 1sg [only refl-1sg-acc-only bread eat-impf.1sg] say-past.1sg

  • ‘I said that only I eat bread.’

  • b. *?Meni [peqet özi-em-∅-la nan ye-men] di-dim.

  • 1sg [only refl-1sg-nom-only bread eat-impf.1sg] say-past.1sg

  • ‘I said that only I eat bread.’

These examples demonstrate that accusative subjects are closer to a binder in the matrix clause than their nominative counterparts. (26a) is licit because the reflexive is in the same binding domain as the c-commanding antecedent, satisfying Principle A of binding theory (Chomsky 1981). In contrast, (26b) is ungrammatical because the anaphor is not bound within its binding domain, violating Principle A. Examples (27a–b) show similar facts from Principle B: an embedded pronominal subject cannot corefer with the matrix subject when it bears accusative case, while a nominative pronominal embedded subject can corefer with a matrix subject.

(27)

  • a.

    Meni [peqet meni-la nan ye-men] di-dim.

    1sg [only 1sg.nom-only bread eat-impf.1sg] say-past.1sg

    ‘I said that only I eat bread.’

  • b.

    *Meni [peqet menii-la nan ye-men] di-dim.

    1sg [only 1sg.acc-only bread eat-impf.1sg] say-past.1sg

    (Intended) ‘I said that only I eat bread.’

  • c.

    Ahmet [peqet meni-la nan ye-du] di-dim.

    Ahmet [only 1sg.acc-only bread eat-impf.3] say-past.1sg

    ‘Ahmet said that only I eat bread.’

According to Principle B, a pronoun must not be in the same binding domain as a c-commanding coreferring expression. If accusative embedded subjects are higher in the structure than nominative embedded subjects, this would explain the ungrammaticality of (27b): in this example, the accusative embedded pronominal subject is in the same binding domain as the coreferring pronoun in the matrix clause. The nominative embedded subject in (27a), on the other hand, is not in the same binding domain, and therefore avoids violating Principle B. Example (27c) shows that it is not the case that first person accusative embedded subjects are banned altogether.

The second piece of evidence comes from the fact that the embedded subject can raise into the matrix clause; however, crucially, only an accusative subject can do so.7

(28)

  • a.

    graphic

    ‘Ahmet said the cup broke.’

  • b.

    graphic

    ‘Ahmet said the cup broke.’

Having demonstrated that accusative embedded subjects are structurally higher than nominative embedded subjects, in the next section we show that accusative embedded subjects need not raise into the matrix clause and can stay in the scope of the attitude verb.

4.3 Accusative Embedded Subjects Can Be below the Embedding Verb

In this section, we show that accusative subjects can be in the scope of the embedding attitude verb. Our evidence comes from three sources. First, we observe that de dicto readings of accusative subjects are available.

(29)

  • graphic

  • ‘Tursun said that a winged horse arrived, but winged horses do not exist.’

Here, the accusative subject tulpar-ni ‘winged horse’ must be interpreted in the scope of the attitude verb de- ‘say’, since its existence is explicitly denied in the continuation without giving rise to an inconsistency. On the assumption that the de dicto reading requires the NP part of the accusative subject to be below the attitude verb at the level of semantic interpretation (Fodor 1970, Partee 1974, Cresswell and von Stechow 1982, Keshet 2008, 2011), example (29) demonstrates that the accusative subject can be structurally below the matrix attitude verb at LF.

Second, embedded clauses with accusative subjects can be coordinated under the same verb.

(30)

  • graphic

  • ‘Each student believes that Aygül left and that Ahmet arrived.’

This example suggests that the embedded accusative subject can form a constituent with the rest of the embedded material to the exclusion of the matrix attitude verb, which in turn indicates that the accusative subject can be located below the matrix verb, that is, in the finite embedded clause.

Third, recall that embedded accusative subjects trigger double accusative constraint violations in the embedded clause, as shown in (20). We now further observe that embedded accusative subjects do not trigger violations of this constraint with respect to matrix material. This is illustrated with a causativized verb, de-mek ‘to say’, in (31). Causativized de-mek assigns accusative case to the causee argument in the matrix clause, and the fact that an accusative embedded subject is still possible demonstrates that for the purposes of double accusative constraint evaluation, the embedded accusative subject is not in the same domain as the causee argument in the matrix clause.

(31)

  • Ahmet-ni Aygül-ge [herbir oqughuchi(-ni) ket-ti] di-güz-dim.

  • Ahmet-acc Aygül-dat [each student(-acc) leave-past.3] say-caus-past.1sg

  • ‘I made Ahmet say to Aygül that every student left.’

Additional evidence comes from the fact that a quantificational accusative subject can take scope below the attitude verb, as shown in (32).

(32)

  • Context: Ahmet heard from Aslan that one of my friends is from Urumchi, but Ahmet has no idea who that friend is. He thought it could be John, it could be Bill, or it could be Sue.

  • Ahmet [bir dostu-m-ni Ürümchilik dep] oyla-idu.

  • Ahmet [one friend-1sg-acc Urumchian c] think-impf.3

  • ‘Ahmet thinks a friend of mine is from Urumchi.’

In this example, the indefinite accusative subject bir dostum-ni ‘a friend of mine’ takes scope below oyla- ‘think’, since there is no particular friend of mine who Ahmet thinks is from Urumchi. This again suggests that the accusative subject can be interpreted in a position that is structurally lower than the attitude verb.

From the data above, we propose that the basic clausal architecture of Uyghur finite complement clauses is as shown in (33).

(33)

graphic

Importantly, given that accusative embedded subjects can be below the embedding verb but never receive shifted interpretation, it follows that attitude verbs themselves are not monsters. This claim is attractive, since many Uyghur verbs take either nominalized or finite complement clauses, but indexical shifting happens only with finite complement clauses, and never with nominalized complements. We submit, therefore, that the monster is its own lexical item, different from the verb. This in turn enables us to entertain the idea that languages do not differ in the semantics of indexicals. Rather, the difference between languages lies in whether or not they have a monstrous lexical item. This is in line with proposals made in Anand and Nevins 2004 and Anand 2006, and contra those in Schlenker 1999, 2003.

Notice that our analysis predicts that case marking on the embedded pronominal subject can be used as a tool for diagnosing the scope of the monster operator. That is, it is predicted that when some particular phrase is structurally higher than (to the left of ) the accusative subject, it cannot shift, whereas all phrases lower than (to the right of ) the nominative subject must shift, as depicted schematically in (34).

(34)

graphic

In the next section, we show that these predictions are borne out.

5 Embedded Subjects as Diagnostics

In this section, we demonstrate that case morphology on the embedded subject in a finite clause can be used to localize the position of the monster: an indexical pronoun to the left of an accusative subject receives an unshifted interpretation, as it is positioned outside the scope of the monster operator, while an indexical pronoun to the right of a nominative embedded subject is interpreted in the scope of the monster operator (receives a shifted interpretation). We begin by observing that in Uyghur, accusative and dative NPs can scramble freely within the clause.

(35)

  • a.

    (Xet-ni) men (xet-ni) yaz-dim.

    (letter-acc) 1sg.nom (letter-acc) write-past.1sg

    ‘I wrote a letter.’

  • b.

    (Aygül-ge) Ahmet (Aygül-ge) bowaq-ni (Aygül-ge) körset-ti.

    (Aygül-dat) Ahmet (Aygül-dat) baby-acc (Aygül-dat) show-past.3

    ‘Ahmet showed the baby to Aygül.’

What example (36) shows is that a dative argument can shift when it linearly follows the nominative subject.

(36)

  • Context: Ahmet told me, ‘‘I sent you a letter the other day.’’ Now I am telling you

  • what he said.

  • Ahmet manga [men sanga xet ewet-tim] di-di.

  • Ahmet 1sg.dat [1sg.nom 2sg.dat letter send-past.1sg] say-past.3

  • ‘Ahmeti told me that hei sent a letter to me.’

Crucially, (37) demonstrates that with the same word order, the nonshifted interpretation is unavailable, which is what we predict. The logic of (37) is that the context only supports the nonshifted interpretation, and the infelicity of the sentence indicates the lack of this reading.

(37)

  • Context: Muhemmet told me, ‘‘I sent a letter to Aygül.’’ I am talking to Aygül.

  • #Muhemmet manga [men sanga xet ewet-tim] di-di.

  • Muhemmet 1sg.dat [1sg.nom 2sg.dat letter send-past.1sg] say-past.3

  • Unavailable interpretation: ‘Muhemmeti told me that hei sent a letter to youAygül.’

The same facts obtain with accusative objects.

(38)

  • a.

    Context: Ahmet said to Aygül, ‘‘I like you.’’ Now I tell Tursun what he said.

    Ahmet Aygül-ge [men seni yaxshi kör-ymen] di-di.

    Ahmet Aygül-dat [1sg.nom 2sg.acc well see-impf.1sg] say-past.3

    ‘Ahmet told Aygül that he likes her.’

  • b.

    Context: Ahmet told me, ‘‘I like Aygül.’’ I tell Aygül what he said.

    #Ahmet [men seni yaxshi kör-ymen] di-di.

    Ahmet [1sg.nom 2sg.acc well see-impf.1sg] say-past.3

    Unavailable interpretation: ‘Ahmet said that he likes youAygül.’

Our second prediction is that a dative argument preceding an accusative embedded subject must receive a nonshifted reading. First, observe that in (39), a nonshifted reading is available for such dative arguments.

(39)

  • Context: I am John. Ahmet said to Muhemmet about me, ‘‘John sent a letter to Aygül.’’ Muhemmet told me what Ahmet said, and I tell Aygül about this.

  • Ahmet [sanga meni xet ewet-ti] di-di.

  • Ahmet [2sg.dat 1sg.acc letter send-past.3] say-past.3

  • ‘Ahmet said that I sent a letter to you.’

Moreover, (40) shows that the nonshifted interpretation is the only interpretation available in such constructions.

(40)

  • Context: I am John. Ahmet said to Aygül, ‘‘John sent a letter to you.’’ Aygül told me what Ahmet said. Now I tell Muhemmet what I heard from Aygül.

  • #Ahmet Aygül-ge [sanga meni xet ewet-ti] di-di.

  • Ahmet Aygül-dat [2sg.dat 1sg.acc letter send-past.3] say-past.3

  • ‘Ahmet said to Aygül that I sent a letter to Muhemmet.’

We predict that in principle, embedded objects should behave the same way; however, this cannot be tested because embedded accusative subjects cannot be preceded by accusative objects originating in the embedded clause owing to the double accusative constraint.

We have just shown that our basic predictions are correct and that the case of the embedded subject can be used to localize the position of the monster. Now let us consider examples where datives and objects are ambiguous between shifted and nonshifted interpretations. One such situation arises because Uyghur allows subject pro-drop, as demonstrated in (41).

(41)

  • (Men) ket-tim.

  • (1sg.nom) left-past.1sg

  • ‘I left.’

In the absence of an overt subject, we cannot fix the location of the monster, and hence either interpretation should be available. More concretely, the string [CP NPnonsubject V] could have the two parses shown in (42).

(42)

graphic

That this prediction is correct is shown by (43), with dative phrases. The context in (43a) supports only a shifted interpretation of the dative indexical and that in (43b) supports only its nonshifted interpretation; the sentence allowing for shifting ambiguity is felicitous in both contexts.

(43)

  • a.

    Context: Ahmet told me, ‘‘I sent you a letter the other day.’’ Now I am telling you what he said.

    Ahmet manga [sanga xet ewet-tim] di-di.

    Ahmet 1sg.dat [2sg.dat letter send-past.1sg] say-past.3

    ‘Ahmet told me that he sent a letter to me.’

  • b.

    Context: Muhemmet told me, ‘‘I sent a letter to Aygül.’’ Now I am telling Aygül what he said.

    Muhemmet manga [sanga xet ewet-tim] di-di.

    Muhemmet 1sg.dat [2sg.dat letter send-past.1sg] say-past.3

    ‘Muhemmet told me that he sent a letter to you.’

Moreover, as (44) demonstrates, accusative objects behave the same way. The context in (44a) supports a shifted interpretation, while the context in (44b) supports a nonshifted interpretation. The same string is licit in both contexts.

(44)

  • a.

    Context: I heard Ahmet say to Aygül, ‘‘I like you.’’ I am telling Tursun what he said.

    Ahmet Aygül-ge [seni yaxshi kör-ymen] di-di.

    Ahmet Aygül-dat [2sg.acc well see-impf.1sg] say-past.3

    ‘Ahmet said to Aygül that he likes her.’

  • b.

    Context: Ahmet told me, ‘‘I like Aygül.’’ I tell Aygül what he said.

    Ahmet [seni yaxshi kör-ymen] di-di.

    Ahmet [2sg.acc well see-impf.1sg] say-past.3

    ‘Ahmet said that he likes you.’

Before closing this section, we present two more pieces of evidence corroborating our analysis. The first observation concerns obligatorily nonshifting expressions. Some Uyghur NPs are not felicitous in shifted contexts, and phrases headed by qaysi ‘which’ are one such type of expression. Indexicals inside qaysi-phrases can only receive nonshifted interpretations, as demonstrated in (45).

(45)

  • Ahmet [qaysi oqughuchi-m-ni kör-dim dep] bil-idu.

  • Ahmet [which student-1sg-acc see-past.1sgc] believe-impf.3

  • ✓(nonshifted possessor) ‘Which of myspeaker students does Ahmet think he saw?’

  • ✗(shifted possessor) ‘Which of hisi students does Ahmeti think he saw?’

We predict that nominative embedded subjects cannot precede qaysi objects, since this word order would force the qaysi-phrase to be under the scope of the monster operator. Example (46) shows that this prediction is borne out.

(46)

  • *Ahmet [men qaysi oqughuchi-m-ni kör-dim dep] bil-idu.

  • Ahmet [1sg.nom which student-1sg-acc see-past.1sgc] believe-impf.3

  • ‘Which of myspeaker/hisi students does Ahmeti think he saw?’

As expected, when the subject is accusative, the sentence is perfectly grammatical, as demonstrated by (47).8

(47)

  • Ahmet [meni qaysi imtihan-din öt-ti] di-di?

  • Ahmet [1sg.nom which test-from pass-past.3] say-past.3

  • ‘Which test did Ahmet say that I passed?’

Second, our theory predicts that indexicals occurring within the same NP must either shift together or not shift at all, depending on the position where the entire NP appears (cf. Anand and Nevins’s (2004) Shift-Together constraint). That is, all indexicals within an accusative subject must not shift, while all indexicals within a nominative subject must shift. That this prediction is correct is demonstrated in (48) and (49).

(48)

  • Ahmet Aygül-ge [[sen yaxshi kör-idi-ghan] oqughuchi-m-ni imtihan-din öt-ti] di-di.

  • Ahmet Aygül-dat [[2sg well see-impf-rel] student-1sg-acc test-from pass-past.3] say-past.3

  • ‘Ahmet told Aygül that the student of mine that you like passed the test.’

  • *‘Ahmet told Aygül that the student of mine that Aygül likes passed the test.’

  • *‘Ahmet told Aygül that the student of his that you like passed the test.’

  • *‘Ahmet told Aygül that the student of his that Aygül likes passed the test.’

(49)

  • Ahmet Aygül-ge [[sen yaxshi kör-idi-ghan] oqughuchi-m imtihan-din öt-ti] di-di.

  • Ahmet Aygül-dat [[2sg well see-impf-rel] student-1sg test-from pass-past.3] say-past.3

  • *‘Ahmet told Aygül that the student of mine that you like passed the test.’

  • *‘Ahmet told Aygül that the student of mine that Aygül likes passed the test.’

  • *‘Ahmet told Aygül that the student of his that you like passed the test.’

  • ‘Ahmet told Aygül that the student of his that Aygül likes passed the test.’

Thus, we have shown that by using case marking on embedded subjects and linear order between embedded subjects and other DPs containing indexicals, we can determine whether a DP takes scope above or below the monster operator. All indexicals within the same DP behave identically with respect to the monster operator, as expected, and phrases that disallow indexical shift are also not permitted in syntactic configurations that would require them to exhibit shifting behavior.

6 Conclusions

Our analysis of Uyghur indexical shifting crucially relies on the syntactic position of the monster operator, which partitions the embedded clause into a shifted domain and a nonshifted domain. The novelty of our account lies mainly in the correlation of clausal syntax and shifting properties, and we have demonstrated that embedded objects and datives can appear within the upper (nonshifted) or the lower (shifted) domain, with the predicted consequences. To our knowledge, such a correlation between the syntax and semantics of indexical shifting has not been documented elsewhere.

Before concluding the article, we would like to mention one remaining issue that arises as a consequence of our analysis of indexical shifting. As shown earlier, the accusative subject of a finite complement clause never undergoes indexical shifting, but the verbal agreement on the embedded verb always does. This can create a mismatch between person features expressed via verbal agreement and the person feature of the pronominal argument that ostensibly controls the agreement, as shown in (50).

(50)

  • a.

    Ahmet [meni ket-ti] di-di.

    Ahmet [1sg.acc leave-past.3] say-past.3

    *‘Ahmet said that he himself left.’

    ‘Ahmet said that I left.’

  • b.

    *Ahmet [meni ket-tim] di-di.

    Ahmet [1sg.acc leave-past.1sg] say-past.3

    ‘Ahmet said that he/I left.’

In (50a), the embedded agreement is third person, but the pronoun that it agrees with is not shifted and is first person. Descriptively speaking, then, third person agreement and a first person pronoun refer to the same entity, but belong to shifted and nonshifted domains. Note that the verbal agreement with an accusative subject is not always third person, as shown by (51a–c), and hence it cannot be said that the agreement in (50a) is default agreement.

(51)

  • a.

    Ahmet [Aygül-ni nan ye-isen] di-di.

    Ahmet [Aygül-acc bread eat-impf.2] say-past.3

    ‘Ahmet said to Aygül, ‘‘You eat bread.’’ ’

  • b.

    Ahmet [meni nan ye-isen] di-di.

    Ahmet [1sg.acc bread eat-impf.2] say-past.3

    ‘Ahmet said to me, ‘‘You eat bread.’’ ’

  • c.

    Men [peqet öz-em-ni-la nan ye-imen] di-dim.

    1sg.nom [only self-1sg-acc-foc bread eat-impf.1sg] say-past.1sg

    ‘I said that only I eat bread.’

Together with our claim that accusative subjects originate in the embedded clause and are not proleptic arguments, this implies that pronoun/agreement mismatch is a regular phenomenon in embedded finite clauses in Uyghur. This poses an interesting puzzle for the theory of agreement.

One widely held account of agreement is roughly as follows: person features (first person, second person, etc.) are manipulated in syntax. Agreeing heads either enter into the derivation with person features and check them against nominals or copy person features from nominals via Agree relations. Either way, it is the nominal that is agreed with that licenses person features on agreeing heads. The person features on agreeing heads are spelled out as morphology descriptively called ‘‘agreement morphology.’’

The present case is problematic for this type of account, as there is no matching feature on the embedded subject nominal to license the agreement on the embedded predicate. In (51b), for example, the embedded subject pronoun has first person features, which on the standard account cannot license second person morphology on the embedded predicate.

Therefore, an account of this phenomenon requires a theory of agreement that differs from the prevailing view. In particular, this phenomenon suggests that the mechanism of agreement must be able to refer to the semantic property of the accusative subject with respect to the reported context. Developing such a system is far beyond the scope of this article, and we leave it for future research.9

Notes

We would like to thank the two anonymous LI reviewers for helpful comments that greatly improved the quality of this article. Our thanks also go to Pranav Anand, Alya Asarina, Amy Rose Deal, Rose-Marie Déchaine, Danny Fox, Martin Hackl, Claire Halpert, Irene Heim, Sabine Iatridou, Makoto Kanazawa, Yusuke Kubota, Andrew Nevins, David Oshima, Orin Percus, David Pesetsky, Norvin Richards, Philippe Schlenker, Chris Tancredi, Wataru Uegaki, Martina Wiltschko, and the audiences of NELS 40, MIT Syntax-Semantics Workshop 2009, and the MIT LF Reading Group for many helpful discussions and insightful suggestions. We also gratefully acknowledge the help and patience of Mettursun Bedulla, without whom this project would not have been possible. The second author was partially supported by a Euryi grant from the European Science Foundation (Schlenker – Presupposition: A Formal Pragmatic Approach); the ESF is not responsible for the claims made here. All errors are the authors’ own.

1 Sudo (2010, 2012) gives detailed data and a semantic analysis of indexical shifting under a variety of attitude predicates in Uyghur (and other languages). As the main interest of this article is the syntax of indexical shifting, we will mostly use de-mek as the main verb to avoid potential complications.

2 Strictly speaking, the quotation reading is not impossible for (9), since ungrammatical expressions can be quoted. However, such a reading is pragmatically highly marked and generally infelicitous.

3 That the idiom in (15) is a true sentential idiom, and not an NP idiom, is shown by the fact that only with the given verb (kel-mek ‘to come’) does the idiom retain the idiomatic meaning.

4 In (18), ete ‘tomorrow’ receives a nonshifted interpretation, which complies with our claim that an indexical appearing above the accusative subject does not shift. This point will be further clarified below with person indexicals. We should mention here, however, that the shiftability of temporal adverbials including ete is not very clear to us at this moment, calling for further investigation. In any event, for present purposes, the interpretation of ete in (18) is irrelevant to the point being made here, and any other adverbial, indexical or not, can be used to make the same point. We thank an anonymous reviewer for discussion on this.

5 A similar constraint is known for Japanese (e.g., Kuroda 1965, Harada 1973, Hiraiwa 2002, 2010, Poser 2002).

6Öztürk (2013) reports different facts for the variety of Uyghur she works with. Our consultant consistently rejects passive sentences with accusative case. We have no explanation for this difference.

7 There is a complicating factor: with a passivized matrix clause, the embedded nominative subject can appear within the matrix clause if it is agreed with in the matrix clause. We set aside this issue here.

8 We thank an anonymous reviewer for suggesting that we clarify this prediction.

9 An anonymous reviewer generously suggested to us the following analysis: Assume, following Sauerland (2003), that the locus of φ-features is φP, a functional projection above the DP, and that it is φP that controls the verbal agreement. When the subject is accusative, what gets moved above the monster is just the DP part, and φP is left behind. A crucial assumption here is that a DP like Aygül can have φP with a second person feature, for example, as far as the interpretation is coherent. However, it is not clear to us how this analysis would account for examples like (51b) where the accusative subject is a first person pronoun and the embedded agreement is second person, as arguably another φP is necessary in the moved position for the first person pronoun to be realized as first person. Whether this line of analysis can be modified to explain such cases is left open here.

References

References
Anand
Pranav
.
2006
.
De de se
.
Doctoral dissertation, MIT, Cambridge, MA
.
Anand
Pranav
, and
Andrew
Nevins
.
2004
.
Shifty operators in changing context
. In
Proceedings of SALT 14
, ed. by
Robert B.
Young
,
20
37
. .
Bittner
Maria
.
2012
.
Perspectival discourse referents for indexicals
.
Paper presented at SULA 7, Cornell University. To appear in the proceedings. Available at http://www.rci.rutgers.edu/_mbittner/bitt12_sula7.pdf
.
Bruening
Benjamin
.
2001
.
Raising to object and proper movement
.
Ms., University of Delaware, Newark
.
Cappelen
Herman
, and
Ernest
Lepore
.
2012
.
Quotation
. In
The Stanford encyclopedia of philosophy
, ed. by
Edward N.
Zalta
. .
Chomsky
Noam
.
1981
.
Lectures on government and binding
.
Dordrecht
:
Foris
.
Cresswell
Max J
., and
Arnim
von Stechow
.
1982
.
De re belief generalized
.
Linguistics and Philosophy
5
:
503
535
.
Deal
Amy Rose
.
2008
.
Events in space
. In
Proceedings of SALT 18
, ed. by
Tova
Friedman
and
Satoshi
Ito
,
230
247
. .
Deal
Amy Rose
.
2011
.
Nez Perce embedded indexicals
.
Handout, Ling Lunch, MIT, Cambridge, MA
.
Fodor
Janet Dean
.
1970
.
The linguistic description of opaque contexts
.
Doctoral dissertation, MIT, Cambridge, MA
.
Gültekin Șener
Nilüfer
, and
Serkan
Șener
.
2011
.
Null subject, indexicality in Turkish and Uyghur
. In
Proceedings of the 7th Workshop on Altaic Formal Linguistics (WAFL7)
, ed. by
Andrew
Simpson
.
MIT Working Papers in Linguistics 62
.
Cambridge, MA
:
MIT, MIT Working Papers in Linguistics
.
Halpert
Claire
.
2009
.
Light verb constructions and case patterns in Uyghur
.
Ms., MIT, Cambridge, MA
.
Harada
Sin-iti
.
1973
.
Counter equi NP deletion
.
Annual Bulletin, Research Institute of Logopaedics and Phoniatrics, Tokyo University
7
:
113
147
.
Heycock
Caroline
.
1994
.
Layers of predication
.
New York
:
Garland
.
Hiraiwa
Ken
.
2002
.
Facets of case: On the nature of the Double-o Constraint
. In
Proceedings of the Third Tokyo Conference on Psycholinguistics
, ed. by
Yukio
Otsu
,
139
163
.
Tokyo
:
Hituzi Publishers
.
Hiraiwa
Ken
.
2010
.
Spelling out the Double-o Constraint
.
Natural Language and Linguistic Theory
28
:
723
770
.
Kaplan
David
.
1977
.
Demonstratives
.
Ms., UCLA, Los Angeles, CA
.
Keshet
Ezra
.
2008
.
Good intensions: Paving two roads to a theory of the de re/de dicto distinction
.
Doctoral dissertation, MIT, Cambridge, MA
.
Keshet
Ezra
.
2011
.
Split intensionality: A new scope theory of de re and de dicto
.
Linguistics and Philosophy
33
:
251
283
.
Kuroda
S.-Y
.
1965
.
Generative grammatical studies in the Japanese language
.
Doctoral dissertation, MIT, Cambridge, MA
.
Ludwig
Rainer
,
Robert
Munro
,
David W.
Fleck
, and
Uli
Sauerland
.
2010
.
Reported speech in Matses: Obligatory perspective shift with syntactic transparency
. In
Proceedings of SULA 5: Semantics of Under-Represented Languages in the Americas
, ed. by
Suzi
Lima
,
33
48
.
University of Massachusetts Occasional Papers 41
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Organization
.
Munro
Robert
,
Rainer
Ludwig
,
Uli
Sauerland
, and
David W.
Fleck
.
2012
.
Reported speech in Matses: Perspective persistence and evidential narratives
.
International Journal of American Linguistics
78
:
41
75
.
Oshima
David Y
.
2006
.
Perspectives in reported discourse
.
Doctoral dissertation, Stanford University, Stanford, CA
.
Öztürk
Balkiz
.
2013
.
Postverbal constituents in SOV languages
. In
Theoretical approaches to disharmonic word orders
, ed. by
Michelle
Sheehan
and
Theresa
Biberauer
,
270
305
.
Oxford
:
Oxford University Press
.
Partee
Barbara H
.
1974
.
Opacity and scope
. In
Semantics and philosophy
, ed. by
Milton K.
Munitz
and
Peter K.
Unger
,
81
101
.
New York
:
New York University Press
.
Poser
William
.
2002
.
The Double-O Constraints in Japanese
.
Ms., University of Pennsylvania, Philadelphia
.
Potsdam
Eric
, and
Jeffrey
Runner
.
2001
.
Richard returns: Copy raising and its implications
. In
Proceedings of Chicago Linguistic Society
, ed. by
Mary
Andronis
,
Christopher
Ball
,
Heidi
Elston
, and
Sylvain
Neuvel
,
1
:
453
468
.
Chicago
:
University of Chicago, Chicago Linguistic Society
.
Quer
Josep
.
2005
.
Context shift and indexical variables in sign languages
. In
Proceedings of SALT 15
, ed. by
Effi
Georgala
and
Jonathan
Howell
,
152
168
. .
Quine
Willard Van Orman
.
1960
.
Word and object
.
Cambridge, MA
:
MIT Press
.
Rice
Keren
.
1986
.
Some remarks on direct and indirect speech in Slave (Northern Athapaskan)
. In
Direct and indirect speech
, ed. by
Florian
Coulmas
,
161
178
.
Berlin
:
Mouton de Gruyter
.
Rogers
Andy
.
1971
.
Three kinds of physical perception verbs
. In
Papers from the Seventh Regional Meeting, Chicago Linguistic Society
,
206
222
.
Chicago
:
University of Chicago, Chicago Linguistic Society
.
Sauerland
Uli
.
2003
.
A new semantics for number
. In
Proceedings of SALT 13
, ed. by
Robert B.
Young
and
Yuping
Zhou
,
258
275
. .
Schlenker
Philippe
.
1999
.
Propositional attitudes and indexicality
.
Doctoral dissertation, MIT, Cambridge, MA
.
Schlenker
Philippe
.
2003
.
A plea for monsters
.
Linguistics and Philosophy
26
:
29
120
.
Speas
Margaret
.
2000
.
Person and point of view in Navajo
. In
Papers in honor of Ken Hale
, ed. by
Andrew
Carnie
,
Eloise
Jelinek
, and
Mary Ann
Willie
.
MIT Working Papers on Endangered and Less Familiar Languages 1
.
Cambridge, MA
:
MIT, MIT Working Papers in Linguistics
.
Sudo
Yasutada
.
2010
.
Person indexicals in Uyghur indexical shifting
.
Paper presented at Berkeley Linguistics Society 36, Berkeley, CA. To appear in the proceedings. Available at http://web.mit.edu/ysudo/www/pdf/bls36.pdf
.
Sudo
Yasutada
.
2012
.
On the semantics of phi features on pronouns
.
Doctoral dissertation, MIT, Cambridge, MA
.
Tanaka
Hidekazu
.
2002
.
Raising to object out of CP
.
Linguistic Inquiry
33
:
637
652
.