shigeru miyagawa

Pre-copyedited version, Shigeru Miyagawa, April 2016
Agreement Beyond Phi
Linguistic Inquiry Monograph (to appear), MIT Press
Chapter 1 Introduction: Strong Uniformity

1. Introduction
A linguistic theory should minimally tell us the following:
• How are natural languages the same?
• In what ways can they be different?
GB theory (Chomsky 1981) had a straightforward answer to these questions. All languages contain the same set of principles such as subjacency and ECP; where the languages differ is in the setting of the parameter built into many of the principles, head parameter being one such example. This vision allowed the theory to attain not only descriptive adequacy, but also explanatory adequacy as well — in the ideal, of course — because this framework gave what appeared at the time to be a compelling picture of the initial state of UG. However, as we learned more about the nature of these principles, it became evident that many, if not all, of these principles are a description of the problem they are supposed to solve. Why, for example, should a movement that crosses two nodes of a particular kind lead to ungrammaticality? Subjacency simply builds this observation into a condition on movement, failing to tell us anything beyond what we already know to be the problem. Chomsky (1986) begins to address this issue, but it is in the minimalist program (MP) (Chomsky 1993, 1995) that the problem inherent to GB comes fully into light, and effort is made to rid
1
the theory of anything that does not have independent and intuitive motivation. As Chomsky states (1995: 233), the assumptions of earlier theories were often "of roughly the order of complexity of what is to be explained." While this is progress, it leaves us without an answer to either of the questions posed at the outset. Without universal principles, it is not obvious how we state the uniform nature of human language, and without principles, there can be no parameters that can be built into them to capture the potential for variation that languages exhibit. Recognizing the vacuum left by ridding the theory of universal principles, Chomsky suggests the Uniformity Principle in their place.
(1) Uniformity Principle (Chomsky 2001: 2)
In the absence of compelling evidence to the contrary, assume languages to be uniform, with variety restricted to easily detectable properties of utterances.
To understand the UP, we need to have more specificity to both parts of the statement. In assuming languages to be uniform, precisely what are the elements that are shared by all languages? In what ways can the languages vary within this uniform profile? It surely is not the case that the detectable properties of utterances are random in nature, just as the parameters in GB are not random in their formulation. I will attempt to provide a concrete instantiation of both portions of the Uniformity Principle (UP) by extending the proposal in Miyagawa (2010), in order to understand both the content of the universal statement and the precise nature of the variation being described in the UP. As we will see, the result is not radically different from the way that principles-and-parameters in GB is conceptualized, and it is also consistent with
2
recent discussion of “micro parameters” by Baker (2008), Kayne (2005), and many others.
2. Strong Uniformity: An instantiation of the Uniformity Principle
In Miyagawa (2010), I focus on elements in linguistic theory that are responsible for triggering the operation of movement. Unlike in GB, in which movement is viewed as entirely optional, where Move α moves anything, anywhere, at any time (Chomsky 1981), in MP, virtually every instance of movement is considered as last resort (Chomsky 1995). What triggers it are grammatical features that must somehow be checked off. These grammatical features vary from language to language, the most common of them being φ-feature agreement.1 Given the central role that the grammatical features have come to play in linguistic theory, it is only natural to ask which grammatical features are found in which languages, and what accounts for the variation. To answer this question, I proposed the Strong Uniformity.
(2) Strong Uniformity (Miyagawa 2010)
Every language shares the same set of grammatical features, and every language overtly manifests these features.
What Strong Uniformity states, in the sprit of UP, is that the same stock of grammatical features is found in every language. The idea that these features overtly appear in some fashion provides the basis for delineating the possible variations in how the grammatical features manifest themselves. Right away, a whole host of questions arise. How does one account for the variety of φ-feature agreements across languages, from an impoverished set like in English to the rich agreement of
3
Romance? What about languages that do not evidence any agreement, such as Chinese, Japanese, and Korean?
There are a variety of φ-feature agreements, but in this monograph I will mostly focus on person agreement, since it is person agreement that is operative in the kinds of phenomena I will look at, such as pro-drop and allocutive agreement. For the second question, I argue that there are two types of grammatical features, φ-feature agreement and what Kiss (1995) calls “discourse configurational” features, which are topic and focus. In some languages, topic/focus plays the same role as agreement in triggering movement to positions such as Spec,TP. By Strong Uniformity, every language has both φ-feature agreement and topic/focus, making all languages uniform. In this monograph, I will often use δ to stand for the discourse- configurational feature without distinguishing between topic and focus.
These grammatical features have a similar status as the universal principles in GB: they are shared by all languages. What differentiates the grammatical features from the universal principles is that the grammatical features actually occur in the language as detectable entities, and they are closely associated with linguistic operations (Chomsky 2005, 2008; Miyagawa 2010), hence they have an independent and intuitive motivation to be included in the theory. What remains is how languages can vary within the framework of Strong Uniformity.
2.1. Examples of typology based on Strong Uniformity
I begin with the assumption that all grammatical features start out on a phase head; I will focus in this monograph on the phase head C (Chomsky 2005, 2007, Miyagawa 2010, Richards 2007). These grammatical features may be inherited by T in certain circumstances. As I noted in Miyagawa (2010), the patterns of inheritance
4
can capture variations across languages. In that work, I only dealt with two such patterns.
(3) Agreement-based languages CP
TP
C’
CP
C’
TP
C δ-feature φ-feature
C φ-feature δ-feature
inheritance
T
(4) Discourse-configurational languages
inheritance
T
In (4), the φ-feature agreement is inherited by T, leading to what I termed an agreement-based language such as English. In (5), the δ-feature is inherited by T, resulting in a discourse-configurational language such as Japanese. In this monograph, I will explore all the basic variations predicted by this approach.
Let us consider the two types of grammatical features, φ-feature and δ-feature. Without making further distinctions, such as dividing the δ-feature into topic and focus, we predict four different types of languages: (I) φ-feature occurs in C, δ-feature on T; (II) δ-feature on C, φ-feature on T; (III) both φ-feature and δ-feature occur on T; and (IV) both φ-feature and δ-feature occur on C. These are given below with representative languages.
5
(5) Some predicted languages
Category I: Category II: Category III:
Category IV:
Cφ, T𝛿 C𝛿, Tφ C, T φ/𝛿
Cφ/𝛿, T
Japanese English
Spanish Dinka
In this monograph, I will look at each of these possibilities, and in fact more. As we will see, there are cases where the δ-feature must be distinguished between topic and focus, as noted originally by Kiss (1995). For example, Spanish has been argued to be a Category III language, with the δ-feature occurring on T (Jiménez-Fernández 2010, Jiménez-Fernández & Miyagawa 2014). However, this turns out to hold only for the δ-feature of topic; the other δ-feature, focus, occurs on C in Spanish. We will see this in Chapter 4 when we look at the various forms of why questions across languages.
Also, the δ-feature of topic is not a unitary feature, but comes in at least three versions: Aboutness, Contrastive, and Familiar/Given (Frascarelli and Hinterhölzl 2007). As we will see in Chapter 2, while Contrastive and Familiar/Given topics are subject to the parametric variation of occurring either on C or being inherited by T, the Aboutness topic uniformly occurs on C across all languages. It is simply the nature of the Aboutness topic that it must have the entire clause as its domain.
Below, I will discuss two instances of the typology given above: the occurrence of the δ-feature topic at T, which we find in Categories I and III; and the occurrence of both types of grammatical feature, φ- and δ-feature, at C, which we find in Category IV. In the chapters to follow, I will motivate other parts of the typology.
2.2.1. δ-feature at T
6
Categories I and III have the δ-feature at T. I will show this with Japanese (I) and Spanish (III). In both languages, I will take up topic movement, which applies within the TP domain in these languages. The points I will demonstrate are that the movement is indeed for topicalization, which is well-known in Spanish, and that it is A-movement, hence a movement that occurs within the TP domain, a point already extensively argued for by Saito (1985, 1992).
The type of movement in Japanese that I take up in Miyagawa (2010) for Japanese that is triggered by δ-feature on T is scrambling. As already established in the literature, clause-bound scrambling has properties of A-movement. To show this, we can turn to some typical properties of A-movement: A-movement can overcome Weak Crossover violation, and it is able to create a new binder (Mahajan 1989).
(6) Whoi ti seems to hisi mother ti to be smart?
(7) Johni seems to himselfi to be ti smart.
In (6) the wh-phrase who undergoes A-movement from the subordinate subject position to the matrix Spec,TP, crossing the pronoun his. Despite this, the sentence is grammatical because a Weak Crossover (WCO) violation is only invoked if there is a variable and there is a pronoun coreferential with the variable that the variable fails to c-command. A-movement does not create a variable because it is not an operator movement, so in (8) the trace and the pronoun in the subordinate subject position are not subject to WCO. In (9), John A-moves to Spec,TP and is able to bind himself from this new position. Presumably, such binding only takes place from A-positions. We can see below that A’-movement is incapable of suppressing a WCO violation; it also cannot create a new binder.
7
(8) ?*Whoi does hisi mother love ti?
(9) *To whomi did each otheri’s friends introduce Mary ti?
The following are examples of clause-bound scrambling in Japanese that demonstrate that it is A-movement; these examples are modeled after Mahajan’s work and similar examples are discussed by Hoji (1985), Saito (1992), Tada (1993), and Yoshimura (1989, 1992). As shown in the (b) example, “A” scrambling can suppress a WCO violation.
(10) a. *[Kinoo proi proj atta hitoi]-ga dare-oj hihansita no? yesterday met person-NOM who-ACCi criticized Q
Lit. ‘The person who met (him) yesterday criticized whom?’
b. Dare-oj [kinoo proi proj atta hitoi]-ga tj hihansita no?
who-ACCi yesterday met person-NOM criticized Q Lit. ‘Who, the person who met (him) yesterday criticized?’
A-movement can also create a new binder (Mahajan 1990, Saito 1992).
(11) a. *Otagaii-no sensei-ga [Taroo-to Hanako]i-o each other-GEN teacher-NOM Taro-and Hanako-ACC
suisensita. recommended
‘Each other’s teachers recommended Taro and Hanako.’
b. Taroo-to Hanako-oi otagai-no sensei-ga ti
8
Taro-and Hanako-ACCi each other-GEN teacher-NOM suisensita.
recommended
‘Taro and Hanako, each other’s teachers recommended.’
Unlike this kind of local scrambling, long-distance scrambling solely has A’ properties, so that long-distance scrambling is unable to suppress a WCO violation and it cannot create a new binder (Mahajan 1990; Saito 1992; Tada 1993).
(12) *Dare-oj who-ACCi [Taroo-ga
Taro-NOM
Lit. ‘Who, the person who met (him) yesterday said that Taro knows
(him)?’
(13) ?*Taroo-to Hanako-oi
Taro-and Hanako-ACCi [koutyou-ga ti sikaru to]
[kinoo proi proj atta
hitoi]-ga
person-NOM no?
yesterday
tj sitteiru to]
met itta
know
C saidQ
otagai-no sensei-ga each other-GEN teacher-NOM omotta.
thought
Having established that clause-bound scrambling may be A-movement, let us move on to the evidence that its function may be topicalization. One piece of evidence comes from acquisition (Miyagawa 2010). Hayashibe (1975) noted that there appears
principal scold C
Lit. ‘Taro and Hanako, each other’s teachers thought that the
principal will scold.’
9
to be a period, sometime up to 5 years of age, in which children tend to interpret scrambled sentences like (14b) as if they were nonscrambled sentences like (14a) in word order, completely ignoring the case marking on the arguments.
(14) a. SOV: Kamesan-ga ahirusan-o turtle-NOM duck-ACC
‘A turtle pushed a duck.’ b. OSV: Ahirusan-o kamesan-ga duck-ACC turtle-NOM
osimasita. pushed
osimasita. pushed
Hayashibe concludes from this that scrambling is acquired late in language development. However, Otsu (1994) shows that children before or around the age of three years of age have no problem with scrambling when they are presented with a discourse context that makes the scrambled sentence sound natural.
c. Kooen-ni ahirusan-ga imasita.
park-in duck-NOM was
Sono ahirusan-o kamesan-ga osimasita.
the duck-ACC turtle-NOM pushed
‘There was a duck in the park. A turtle pushed the duck.’
What Otsu has shown is that scrambling of the object, ‘the duck-ACC’, is possible if there is prior context that establishes it as the discourse topic.2
Spanish is a typical agreement language in that the φ-feature agreement occurs on T; given the rich nature of agreement, it is able to license pro-drop.
10
(15) ___ baila bien. (Jaeggli 1982) dance-3SG well
‘She dances well.’
At the same time, the δ-feature of topic apparently lowers to T as well (Jiménez- Fernández 2010, Jiménez-Fernández and Miyagawa 2014). This topic construction, which Jiménez-Fernández calls topic dislocation, is also called Clitic Left Dislocation (CLLD) in the literature.
(16) a. Estos libros, Juan los leyó ayer. these books Juan them read yesterday
‘These books, Juan read yesterday.’ b. Algunos libros, Juan los leyó ayer.
some books Juan them read yesterday
‘Some books, Juan read yesterday.’ (Arregi 2003)
As we saw with Japanese, if this topic dislocation in Spanish is applying within the TP domain, it is an instance of A-movement. Two pieces of evidence that it is indeed A-movement are Floating Quantifiers (FQ) and binding. On the basis of Catalan data, López (2009) concludes that FQs are allowed only in A-movement, not in A’-movement (Lasnik, 2003).3 We see the same for Spanish, where A-movement such as raising and passive constructions are compatible with FQs.
(17) a. Los padres parecen haber asistido todos a la reunión. the parents seem-PRES.3PL to.have attended all to the meeting
11
‘Parents seem to have all attended the meeting.’
b. Los exámenes han sido corregidos todos.
the exams have-PERF.3PL been corrected all ‘The exams have all been graded.’
If topic displacement involves A-movement, it should allow FQs. We see this in in (18) (Jiménez-Fernández 2010).
(18) Los exámenes los ha corregido todos este profesor. the exams CL have-PERF.3SG corrected all this teacher
‘This teacher has corrected all the exams.’
The second piece of evidence that topic displacement, or CLLD, in Spanish applies within TP relates to the fact that it exhibits A-properties (Jiménez-Fernández 2010). Specifically, it is able to create a new binder, which is clearly an indication of A-movement.
(19) a. *Sui enfermera llamó al pacientei. self’s nurse call-PAST.3SG to.the patient
‘His nurse called the patient.’
b. Al pacientei lo llamó sui enfermera.
to.the patient CL called-PAST.3SG self’s nurse ‘The patient was called by his/her nurse.’
12
While the anaphor fails to be bound in (19a), topicalization of the antecedent to its left makes it possible to create a binder for the anaphor, thus making (19b) grammatical.4
3. Agreement at C: Dinka
Two types of languages are predicted to have agreement at C: Category I (Cφ, T𝛿) and Category IV (Cφ/𝛿, T). Reversing the order, I will first discuss Category IV. According to the typology, a Category IV language has both types of grammatical features, φ and 𝛿, at C. This means that such a language would have, for example, topicalization to
Spec,CP and the φ-feature on C agrees with this topic.5
A language that evidences these properties of Category IV is Dinka, a Nilo-
Saharan language spoken in southern Sudan. The analysis here is drawn from van Ark (2015). Dinka is a V2 language, with the verbal element, either the main verb or an auxiliary element, occurring at C. What occurs as the first element preceding the V2 verbal item is topic. In the first example below, the topic is the subject, and C agrees with it in number (singular). In the second example, the topic is the object, and C agrees with this singular object. In the third example, the topic is a plural subject, and the agreement at C inflects for plurality.
(20)a.
b.
c .
Àyén à-càm cuɪ̤̂n nè̤ pàl.
Ayen 3-eat.SV food P knife. 'Ayen is eating food with a knife.' Cuɪ̤̂n à-cɛ́ɛm Áyén nè̤ pàl. food 3S-eat.OV Ayen.NOM P knife 'Food, Ayen is eating with a knife.'
K ɔ́ c á a - c é̤ r ɔ̀ t h t ɪ̤ ̂ ŋ .
13
people 3P-PRF.SV self.PL see 'The people have seen themselves.'
What occurs in Spec,CP need not always be a topic. Dinka is a Wh-movement language, and the Spec,CP may host a wh-phrase. In that case, C agrees with the wh- phrase moved into Spec,CP.6
(21) Agreement with wh-phrases
a. Ye kɔɔ̀c-kó è̤-kè-thɛ̤̀t?
Q people-CS1-which.PL PST-3P-cook.SV
'Which people were cooking?'
b. Ye kɔɔ̀c-kó è̤-kè-cɪ̤́i Áyén kè gàam gàlàm?
Q people-CS1.which.PL PST-3P-PRF.OV Ayen-NOM PL
'Which people had Ayen given a pen to?'
c. Ye kɔɔ̀c-kó è̤-kè-yè kètàak,
Q people.CS1-which.PL PST-3P-HAB.2SG PL think.NF Áyén kè gàam gàlàm]?
Ayen-NOM PL give.NF pen
'Which did (s)he think that Ayen had given a pen to?'
give.NF pen
[CP è̤-kè-cɪ̤́i PST-3P-PRF.OV
Wh-phrases carry focus, hence, in these cases, presumably C has the 𝛿-feature of
focus. Whether the 𝛿-feature is topic or focus, it occurs at C along with φ-feature
agreement, making Dinka a Category IV language.
Based on the discussion above, we have the following examples for each of the
categories.
14
(22) Some predicted languages
Category I: Category II: Category III:
Category IV:
Cφ, T𝛿 C𝛿, Tφ C, T φ/𝛿
Cφ/𝛿, T
– Japanese
4. Outline of the monograph
– –
–
English Spanish Dinka
In Chapter 2, “Allocutive agreement and the root,” I will look at agreement at C in Japanese. Japanese is traditionally considered as a language without any agreement, yet, Strong Uniformity predicts that it has φ-feature agreement that occurs at C. I will argue that the politeness marker –mas- (-des- for nominal) is this agreement at C. To make this argument, I will draw on the study of some dialects of Basque that exhibit a type of agreement called allocutive agreement, which agrees with one of the discourse participants, the speaker or the hearer. The allocutive agreement agrees with the hearer, hence it is 2nd person. Drawing on the work of Oyharçabal (1993), who gives arguments that the Basque allocutive agreement is standard agreement, and not some other phenomenon, we will see that the Basque allocutive agreement mirrors the politeness marking in Japanese both in its function (politeness) and the position (at C). As agreement, the allocutive agreement requires a “goal,” a second person entity that can furnish the features for person, gender, number, and politeness level. I will argue that such an entity is part of the “performative analysis,” originally proposed by Ross (1969) and recently updated by Speas and Tenny (2003). Speas and Tenny call
15
the structure above the expression that contains the speaker and hearer representations “speech act projection” (saP). I will show that the distribution of saP matches exactly the original conception of the root by Emonds (1969). Thus, the “allocutive agreement” in Japanese, the politeness marker –mas-, occurs in the three environments Emonds specified as root.
(23) Root
A root will mean either the highest S in a tree, an S immediately dominated by the highest S or the reported S in direct discourse. (Emonds 1969: 6)
In Chapter 3, “Pro-drop, E-type Pronouns, and Agreement,” I take up a topic that was extensively studied in the 1980s — the phenomena of pro-drop. Kuroda (1965) suggested that the empty slots in Japanese sentences are pronominal in nature, an analysis that foreshadowed later works of Taraldsen (1978), Huang (1984), and Rizzi (1986), among many others. In the 1990s, starting with Huang (1991) and Otani and Whitman (1991), a new breed of what in the past would have been called pro-drop began to be discussed. These are cases in which the empty element in an argument position has an indefinite interpretation that allows sloppy interpretation. Oku (1998), in a work that opened the door to the so-called argument ellipsis analysis, notes that in Japanese, the subject empty element may get either the strict or the sloppy interpretation.
(24) a. Mariko-wa [zibun-no kodomo-ga furansugo-o benkyoosuru to] omotteiru. Mariko-TOP self-GEN child-NOM French-ACC study that think ‘Lit. Mariko thinks that self’s child will study French.’
16
b. Haruna-wa [e surobeniago-o benkyoosuru to] omotteiru. Haruna-TOP Slovenian-ACC study that think ‘Lit. Haruna thinks that e will study Slovenian.’
Strict/Sloppy
The e in the subordinate subject position in (b) may be interpreted as “he/she,” which would be the standard “pro” referring to Mariko’s child, but it can also have the sloppy interpretation of “Haruna’s child.” Following Otani and Whitman (1991), Oku assumes that the sloppy interpretation, which arises from an indefinite expression, cannot be pro. He then proposes that the sloppy interpretation is made possible by a fully specified argument (“self’s child”) that has undergone ellipsis. He calls this argument ellipsis. Oku makes one additional observation that has led to an important body of work on the relationship between the possibility of sloppy interpretation and agreement. He notes that in the Spanish example below, only the strict reading is possible.
(25) a. María cree que su propuesta será aceptada. Maria believes that her proposal will.be accepted ‘Maria believes that her proposal will be accepted.’
b. Juan también cree que e será aceptada.
Juan also believes that will.be accepted
‘lit. Juan also believes that e will be accepted.’ (Oku (1998))
Strict/*Sloppy
17
Oku suggests that the difference here is that the subject has agreement, and this blocks the subject position from undergoing argument ellipsis. This is consistent with the observation by Taraldsen (1978) and Rizzi (1986) that rich agreement licenses pro- drop. Thus, if there is agreement, and the target of agreement — subject — is empty, the agreement is sufficiently rich enough to license the pro. We would not expect anything else in that position, such as a covert fully specified argument. The idea that agreement blocks argument ellipsis has been reinforced and extended by a series of important works by Saito (2007) and Takahashi (2008, 2010 with Şener, 2013).
One problem with Oku’s observation is that there are languages such as Chinese and Malayalam that do not evidence any overt subject agreement, yet, Takahashi reports that the subject position does not allow sloppy interpretation. I will draw on the work of Liu (2014) for Chinese and various works on Malayalam, including Swenson and Marty (2014), to show that these languages indeed have agreement that targets the subject even though the agreement is not pronounced. The evidence for it comes from the so-called blocking effect of anaphor binding. Having defended Oku’s original observation even with languages that do not have overt agreement, I will then turn the table around and argue, following Oikonomou (to appear), that the empty element is a pro even when it allows a sloppy interpretation as originally suggested by Kuroda (1965). It is not argument ellipsis that gives rise to the sloppy interpretation. Cases have been reported of sloppy interpretation even with overt pronouns and they are presumed to have the E-type pronoun reading (Karttunen 1969; I have changed the example slightly to make it less provocative).
(26) The man who gave his paycheck to his wife was wiser than the man who gave it to his child.
18
The impersonal pronoun it allows an E-type pronoun interpretation which leads to it being reinterpreted as a fully specified noun phrase, his paycheck, where his stands for a variable. The idea that the sloppy interpretation is related to E-type pronoun is similar to Tomioka's (2003) proposal that the element that gets this interpretation is type <e, t> (so a predicate); it must have Existential Closure; and it is type shifted from predicate to individual. It is also related to the "indefinite pronoun" idea of Hoji (1998), which I will discuss in some detail. I will argue along the lines of Oikonomou that the sloppy interpretation is due to an E-type pronoun interpretation of the pro. I will also correlate the difficulty of E-type pronoun interpretation when agreement is operative, thereby incorporating Oku’s original idea without having to assume argument ellipsis.
Chinese is a Category II language, in which the δ-feature remains on C and φ- feature is inherited by T. Adopting an idea of Sato (2015a/b), I will argue that the difficulty of sloppy interpretation has to do with pro being topicalized to Spec,CP in the relevant constructions. I will further argue that this notion of topicalization is also what is operative with the agreement languages, such as Spanish, to make sloppy interpretation difficult, although not impossible, as we will see. To show that the sloppy interpretation of the subject pro is readily available in Japanese but not so easily detectable in Chinese, I present two large-scale surveys, one on Japanese, the other on Chinese. We will see that while the Japanese speakers readily perceive the sloppy interpretation without the help of any additional context, the Chinese speakers rarely get the sloppy interpretation. However, if an appropriate context is provided to induce the sloppy interpretation, as many as 50% of the speakers reported that they get the sloppy reading. I will suggest that these variations are due to factors –
19
topicalization is the factor we take up – that render the example as easy or difficult for interpreting the pro as an E-type pronoun.
In Chapter 4, “On the distribution and structure of 'why',” I take up two approaches to ‘why’, the movement analysis and the externally merge analysis, the latter originally due to Bromberger (1987, 1992) and Rizzi (1990), later extended by Ko (2005), Stephanov and Tsai (2008), among others. A well-known externally merged ‘why’ is how come (Collins 1991), which is merged directly into the Spec,CP where it takes scope. The lack of movement is indicated by the absence of Aux inversion (How come you left the party early?). While many languages have the EM option, a language such as Japanese apparently does not, leaving a gap in the paradigm for ‘why’. It is not the case that this gap exists because Japanese is a wh-in- situ language. Chinese, another wh-in-situ language, has the EM ‘why’, zenme ‘how come’ (Tsai 2008), that behaves similarly to how come in English. I argue that the gap in the paradigm for a language such as Japanese (and presumably Korean) is due to the fact that Japanese is a Category I language in which the δ-feature is inherited by T. This means that FOCUS, a discourse feature, never occurs at C. I show that the EM option for ‘why’ requires FOCUS at C, something that Chinese allows since it is a Category II language, in which the δ-feature of FOCUS remains at C. In the literature, one ostensible piece of evidence for the EM nature of naze ‘why’ in Japanese is that it is unique among wh-phrases in being able to escape the intervention effect (Miyagawa 1997). The intervention effect (Takahashi 1990, Rizzi 1992, Beck 1996a; the effect studied also in Hoji 1985) is a phenomenon in which the covert movement of wh-phrase is blocked when it is c-commanded by certain types of expressions such as a quantifier or something with focus.7 Takahashi (1990) noted that the NPI focus marker –sika triggers an intervention effect.
20
(27) *Hanako-sika dare-ni erab-are-nakat-ta no? Hanako-only who-by choose-PASS-NEG-PST Q ‘By whom was only Hanako chosen?’
The occurrence of “only Hanako” in the subject position blocks the wh-phrase ‘by who’ from taking scope. All wh-phrases are subject to this intervention effect save one: ‘why’ may circumvent the effect of the intervenor and be able to take proper scope (Miyagawa 1997).
(28) Hanako-sika naze erab-are-nakat-ta no? Hanako-only why choose-PASS-NEG-PST Q
‘Why was only Hanako chosen?’
Ko (2005) shows that the same anti-intervention effect shows up in Korean.
(29) a. *Amwuto/*John-pakkey mwues-ul ilk-ci-anh-ass-ni? (Beck and Kim 1997) Anyone / John-only what-Acc read-CI-not-Past-Q
‘What did no one/only John read?’
b. Amwuto/?John-pakkey way ku chayk-ul ilk-ci-anh-ass-ni? (Ko 2005)
Anyone / John-only why that book-Acc read-CI-not-Past-Q ‘Why did no one/only John read that book?’
In (29a), the argument wh-phrase ‘what’ cannot take scope because of the c- commanding intervenor, ‘anyone/only John’ in the subject position. As in Japanese,
21
‘why’ is able to escape this intervention effect, as we see in (29b). In contrast, in Chinese ‘why’ cannot escape the effect of intervention (Yang 2012).
(30) *Zhiyou Zhangsan weishenme cizhi?
only Zhangsan whyadv resign ‘Why did only Zhangsan resign?’
I will argue that it is no accident that Japanese and Korean, but not Chinese, have the anti-intervention effect. Drawing on the work of Beck (1995) and Shlonsky and Soare (2011), I will propose a structure of ‘why’ that, looked at from Strong Uniformity, predicts that only languages such as Japanese and Korean have the anti-intervention effect with ‘why’. These are wh-in-situ languages and Category I languages in which the δ-feature is inherited by T. In addition to naze 'why', I will look at another expression that has a similar meaning, the use of 'what' for 'why'.
(31) Taroo-wa nani-o awatete-iru no? Taro-TOP what-ACC panick-ing Q ‘Why (in the hell) is Taro panicking?’
A similar use of 'what' for 'why' is found in other languages; the following is a German example from Ochi (2014).
(32) Was tadeln Sie Hans denn? what blame you Hans
‘Why (the hell) are you blaming Hans?’
22
I will show that this usage of 'what' for 'why' in Japanese differs from the regular 'why' word naze in having a structure that is causative in nature. As part of the argument for this, I show that this construction evidences the "deep" Double-o constraint proposed by Harada (1973, 1975) based on the causative construction.
In Chapter 5, “Ga/No conversion, Strong Uniformity, and focus,” I look at the well-known phenomenon of genitive subject in Japanese, a phenomenon found across many Altaic languages, though there are many variations across these languages. The basic fact is that in Japanese, the subject of a relative clause or a complex NP may be marked with the nominative –ga or the genitive –no.
(33) Hanako-ga/-no katta hon Hanako-Nom/-Gen bought book
‘the book that Hanako bought’
Two competing theories of how to account for the occurrence of the genitive –no appear in the literature, the D-licensing and the C-licensing approaches. In the D- licensing approach, the linguists key in on the fact that there must be a nominal head to license the genitive, and this is consistent with the nominal phrase in Japanese, where everything must be marked with the genitive.
(34) [DP Hanako-no gakkai-de-no Taroo-no hihan] Hanako-Gen conference-at-Gen Taro-Gen criticism ‘Hanako’s criticism of Taro at the conference’
The C-licensing approach focuses on the morphology of the predicate: the prenominal predicate is in an adnominal form, although the actual distinction between adnominal
23
and finite had been lost hundreds of years ago except for the copula (na vs. da). The idea is that despite the loss of morphological difference, the actual difference still exists between adnominal and finite inflections. I will show that the ga/no construction provides further evidence for Strong Uniformity. The key observation, by Akaso and Haraguchi (2010, 2011), is that focus deprives the possibility of the genitive marking on the subject.
(35) Taroo-dake-ga/*-no nonda kusuri Taro-only-Nom/-Gen took medicine ‘medicine that only Taro took’
This is predicted on the D-licensing analysis viewed within Strong Uniformity. D- licensing states that for the genitive to be licensed, the RC must be a TP, and not a CP. It is the same as the ECM construction in English, where the lack of a CP layer allows a higher head to assign Case to it. Because focus requires C to furnish the pertinent feature, the occurrence of focus naturally leads to the occurrence of C, and that in principle should block the possibility of the genitive subject under D-licensing, which assumes that only a TP RC allows a genitive subject. However, Ochi (in press) observes that focus is possible with genitive subjects if the focus occurs on an adjunct.
(36) kinoo/sukosi-dake Taroo-no nonda kusuri yesterday/little-only Taro-GEN took medicine
‘the medicine that Taro took only yesterday/only a little’
Here, the focus marking is on the adjunct time adverbial ‘yesterday’ or the quantity adverbial ‘just a little’.
24
What I will show is that this argument/adjunct asymmetry for focus marking in the ga/no conversion construction reflects a fundamental point about Strong Uniformity. Strong Uniformity holds that φ-feature and δ-feature are computationally equivalent as far as narrow syntax operations are concerned. I will argue that the argument/adjunct asymmetry in focus marking of the ga/no conversion construction shows that Activation must trigger the agreement involving a δ-feature, in particular, focus agreement. The Activation we see in the ga/no conversion is Case, just as we find Case for Activation for φ-feature agreement (e.g., Chomsky 2001). Although the actual case system for δ-feature agreement is slightly different — what Rackowski (2002) calls Case Agreement — it nevertheless operates on the familiar case system of nominative, accusative, dative, and so forth. This analysis provides evidence for the Strong Uniformity notion that two types of grammatical features are computationally equivalent. Together with what we will see in Chapter 3 about the Chinese subject pro, which depends on either the φ-feature of person agreement or the δ-feature of topic for participating in coreference, the picture emerges of human language that is uniform in containing the same set of grammatical features, with variation arising from where these grammatical features may occur in the structure and how they are used by language to implement the various operations that make human language the dynamic and expressive system that it is. Sections 1-5 of this chapter reproduces Miyagwa (2013). I am grateful to the Society of English Linguistics in Japan for allowing me to use the article from their fine journal, English Linguistics.
1In the literature, there is a debate as to what triggers last-resort movement. For “EPP” type A-movement, some linguis have argued that Case is responsible (Bošković 1997,
25
2002 and Martin 1999). There are other studies that assume that movement correlates with agreement (e.g., Chomsky 2000, 2005, 2007, 2008, Kuroda 1988, Miyagawa 2010, Pesetsky and Torrego 2001).
2See Sano (2005) for related discussion on the acquisition of topicalization in Japanese.
3See McCloskey (2000) for an analysis of wh-movement (A’-movement) being compatible with FQs in West Ulster English.
4An anonymous reviewer notes that in (19b), if the resumptive pronoun does not occur, one gets a contrastive focus interpretation instead of a topic interpretation. If what I say in Chapter 4 is correct that the focus feature in Spanish stays at C, this may mean that the contrastive focus is due to A'-movement instead of A-movement.
5 The most well-known kind of agreement at C is what we find in languages such as West Flemish.
(i) a. Kpeinzen dan-k (ik) morgen goan. I-think that-I (I) tomorrow go
‘I think that I’ll go tomorrow.’
b. Kpeinzen da-j (gie) morgen goat. I-think that-you (you) tomorrow go
‘I think that you’ll go tomorrow.’
(Hageman 1992, Haegeman and van Koppen 2012)
The complementizer agrees with the subject, thus indicating agreement at C. At the same time, there is subject-verb agreement. I will not deal with this type of agreement at C. See, for example, Miyagawa 2010 and Haegeman and van Koppen 2012 for possible analyses.
6In the wh-movement construction, the agreement is impoverished; there is no person agreement. In certain tenses the agreement disappears all together. See van Ark (2015).
7More recently, Kim (2002, 2006) argues that the intervention effect is due to the focus feature on the wh-phrase being blocked from being associated with the focus feature on C by another focus-bearing item. Beck (2006) presents a formal semantic analysis based on this assumption that intervention is an instance of the failure of focus-agreement. See also Miyagawa (2010).
26