UNIVERSITY OF CALGARY
Featuring Russian Tense: A Feature-Theoretic Account of the Russian Tense System
by
Ilana Mezhevich
A THESIS SUBMITTED TO THE FACULTY OF GRADUATE STUDIES IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
DEPARTMENT OF LINGUISTICS
CALGARY, ALBERTA AUGUST, 2006
© Ilana Mezhevich
ABSTRACT
This dissertation explores the morphosyntactic expression of the relationship between tense, aspect, and mood in Russian. Russian makes the fundamental distinction between past and non-past tense verb forms. Past tense verbs are marked for both tense and aspect, whereas non-past tense verbs are marked only for aspect. In the non-past, the aspectual morphology performs a double duty of conveying both the aspectual and the temporal meaning. Furthermore, Russian uses past tense morphology to express irrealis mood. The same morphological form conveys the past temporal reference in indicative clauses and remains tenseless in subjunctive clauses. Finally, Russian indicative and subjunctive clauses contrast with respect to long-distance wh-movement. Wh-movement is problematic out of embedded indicative clauses which have independent temporal reference. In contrast, wh-movement is acceptable out of embedded subjunctive clauses, whose temporal reference is determined by the matrix clause. The following questions are addressed: Q1: How can the relationship between tense and aspect in the non-past be represented in the morphosyntax? Q2: How can the interpretation of the past tense morphology be accounted for in terms of the morphosyntactic environment? Q3: Can the independent tense content block long-distance wh-movement?
iii
The answers to Q1 and Q2 are based on the idea that although tense, aspect, and mood are distinct functional categories, they all express a relation that can be characterized as (non-)coincidence. I propose that all three are dyadic predicates that order temporal arguments in syntax. I propose also that aspect, tense, and mood can be represented as formal features [±PAST], [±P], and [±FIN], respectively, that license a particular relation of (non-)coincidence. This relation is represented as the interpretable feature [±COIN] on Asp, T, and C, which serves as the basis for semantic interpretation. The answer to Q3 is negative. I argue that the independent tense content cannot be held responsible for the unavailability of long-distance wh-movement. I propose that the problematic status of wh-movement out of Russian indicative clauses is due to a subjacency violation.
iv
ACKNOWLEDGMENTS
It is well known that as part of a dissertation, acknowledgments enjoy a privileged status. While the introduction, the conclusion, or even the analysis itself can be skipped partially or in its entirety, acknowledgements are always studied with great attention. Therefore, if you intend to send a message to posterity or simply to get the point across to your fellow contemporary reader, make sure to include this point in the acknowledgments. The last occasion on which I exercised my writing skills in this genre was exactly four years ago, following the completion of my masters degree in the same department. Four years is a long time and sure enough, some changes have taken place. But one thing remained the same: this dissertation would not have come into existence were it not for the people who deserve to be included in these pages. So, let’s review. As four years ago, first and foremost I must thank the Linguistics Department at the University of Calgary, which for six years provided me with generous financial support, an extremely stimulating scientific environment, and a warm and friendly atmosphere. I believe very few people are as fortunate as I am in this respect and I want to take this opportunity once again to thank you for all the attention, encouragement, and support, as well as for giving me an opportunity to teach undergraduate courses. Over the past six years I have accumulated incredible work and life experience, which I will continue to benefit from way beyond my graduation date. This dissertation would have never been written, and perhaps even conceived, without my supervisor. Now, in my case the supervisor would not only deserve a thank you in due course of acknowledgments, but would be in need of a prolonged vacation in
v
some mountain resort with plenty of fresh air, where people are not allowed to listen to loud music after 10pm. Let’s put it this way: I am not the easiest person to supervise. I can be extremely stubborn and self-conscious. I can’t decide what I want, and when I eventually do, I can’t decide how I want it. All this considered, I could cover pages and pages describing Betsy Ritter’s merits as a supervisor and as a teacher, praising her patience and wisdom, mentioning all the occasions on which she was right and I was wrong, including her firm belief, from day one, in me as a syntactician despite all the empirical evidence to the contrary. For this and much more I could never thank her enough. It’s true that our relationship throughout these six years has remained strictly professional. Yet, Betsy has had a profound impact on my personality, on the way I perceive the world and my place in it – as a scientist, as a woman, as a human being. I also want to thank my committee members, Martha McGinnis, Amanda Pounder, Elena Bratishenko, Nicole Wyatt, and my external examiner, Karen Zagona, whose comments and suggestions helped to improve this dissertation considerably. No doubt, the PhD defense is one of the most nerve-racking events in academic career, no matter how lightly the survivor seems to take this trial. I am grateful to my committee for making mine as smooth and painless as possible. The external examiner is perhaps everybody’s biggest fear: the unnerving combination of a well-known name and a completely unknown personality and examination strategies. Karen proved this fear to be absolutely groundless in my case. Her clearly articulated questions and friendly attitude contributed in no small part to my success at the defense. Last but not the least, thank you to Robert Murray for his service as a perfectly neutral chair. Speaking of Robert Murray, I am happy to extend my gratitude beyond his
vi
remarkable chair performance. The past two years have been truly amazing. Journal of Germanic Linguistics, for which Robert employed me as an editorial and production assistant, provided me with both a steady all-year-round income and invaluable work experience. The annual GLAC was the hallmark of my year for two years. Thank you for always being there for me as a grad coordinator and as a friend, for sharing my passion for sushi and beer, for wonderfully planned trips, for every time we inevitably got lost, for “breaks” which (or is it “that”?) got me through the final stages of the dissertation, and much more. Whatever you may think, you are the best grad advisor ever! Another person who deserves a tremendous thank you is Amanda Pounder. Thank you for so generously spending hours and hours on my projects, for proofreading and revising endless handouts and abstracts, for extremely stimulating discussions (not only about linguistics) and challenging comments and questions. Not many people I know are capable of going to bed with a Georgian grammar–Tybald, Balthasar and I had a long discussion to that effect the other night. Thank you so much for all your help! My work on aspect and complex predicates was greatly influenced by my conversations with Andrea Wilhelm. Her dissertation has remained permanently on my desk ever since her defense three years ago. Andrea, I am very fortunate to have you as my class- and office-mate and a friend for so many years. I also want to thank other people in the Department of Linguistics in Calgary who have in this way or another contributed to my success over the past six years, especially Darin Howe, John Archibald, Michael Dobrovolsky, and my fellow grad students, Heather Bliss, Karen Jesney, Rachel Klippenstein, Antonio González-Poot, Corey Telfer, Christine Shea, Elizabeth Stacy, and Aili You. Of course, special thank you goes to Linda
vii
Toth for all her help with administrative matters, which is truly invaluable. In 2003, thanks again to the generosity to the Calgary Linguistics Department I spent the winter semester in the Linguistics Department at the University of Ottawa. Thank you to Paul Hirschbühler for the opportunity to indulge myself in lexical semantics, and to Maria Luisa Rivero for the thorough introduction to the minimalist syntax–an asset that can hardly be overestimated in the modern world. Finally, I want to thank my families in Russia and Israel including my cat Philemon, and friends outside of linguistics for bringing so much joy into my life: Annensky, Becka, Golda, Hans, Marc, Natasha, Simona, Travis and Rupi, and Tinu-theMammal. And many more: You know who you are. I especially thank Tinu-the-Mammal for patiently enduring various tortures such as opera, Russian fiction, and life with a PhD student in general, and also for letting me rub his belly for luck before the defense. It helped! Mammal, you did the impossible: you proved to me that Calgary could be a livable and a lovable place! Last but not the least, I thank again Lesley Baker, Peanuts and Patches, who since my first days in Canada have been giving me the feeling of a true home.
viii
To Whom It May Concern…
ix
TABLE OF CONTENTS Approval Page………………………………………………..................................
ii
Abstract…………………………………………………………………………….
iii
Acknowledgements………………………………………………………………..
v
Dedication………………………………………………………………………….
ix
Table of Contents……………………………………………………………..........
x
List of Abbreviations……………………………………………………………….
xiv
Epigraph……………………………………………………………………………
xv
CHAPTER ONE: INTRODUCTION……………………………………………
1
1.0. STATEMENT OF THE PROBLEM……………………..……………………….…… 1 1.1. THE CATEGORIES OF TENSE AND ASPECT……………………………………..
3
1.1.1. “Tense” versus “Temporal Reference”………………………….……..
3
1.1.2. Lexical Aspect and Grammatical (Viewpoint) Aspect…..……………..
7
1.2. DATA AND SPECIFIC RESEARCH QUESTIONS………………….…….………….
11
1.2.1. Aspect and the Past vs. Non-Past Opposition in Russian……………....
11
1.2.2. Past Tense Morphology and Subjunctive Mood………………………..
13
1.2.3. Tense and Long-Distance Wh-Movement……………………………..
14
1.3. PROPOSAL…………………..……………..………………………………..…
15
1.4. ORGANIZATION OF THE THESIS…………………………………..……………
18
CHAPTER TWO: OVERVIEW OF THE RUSSIAN TENSE SYSTEM……
19
2.0. INTRODUCTION……………………………………………………….………..
19
2.1. RUSSIAN ASPECT: THE PERFECTIVE/IMPERFECTIVE OPPOSITION……………..
22
2.1.1. Morphological Characteristics…………………………………………
23
2.1.2. Semantic Characteristics……………………………………………….
26
2.1.2.1. Imperfective Aspect…………………………………………...
27
2.1.2.2. Perfective Aspect………………………………………………
30
x
2.2. PAST TENSE VERB FORMS IN RUSSIAN…………….……………..……….……
33
2.2.1. Temporal and Aspectual Interpretation of Past Tense Verb Forms……
33
2.2.2. Past Tense Verb Forms and Subjunctive Mood………….…………….
36
2.2.3. The Historical Development of Russian Past Tense Verb Forms………
38
2.3. NON-PAST TENSE VERB FORMS IN RUSSIAN…………………………………....
46
2.3.1. Simple Non-Past Forms…………………………………………………
46
2.3.2. Complex Tense: Future Imperfective…………………………………..
51
2.3.3. Future is Tense…………………………………………………………
56
2.4. CONCLUSION……………………………………………………………………
59
CHAPTER THREE: FEATURING RUSSIAN TENSE……………………….
61
3.0. INTRODUCTION…………………………………………………………………
61
3.1. THE COMPONENTS OF TIME…………………………………………………….
64
3.1.1. Klein (1995)……………………………………………………………..
64
3.1.2. One Happy Coincidence: The Feature [±COIN]…………………………
71
3.1.2.1. The Feature [±COIN] on T………………………………………
73
3.1.2.2. The Feature [±COIN] on Asp……………………………………
76
3.1.3. Interpretability, Valuation, and Agreement……………………………..
82
3.2. INTERPRETATION OF PAST TENSE CLAUSES…………………………………….
87
3.3. TEMPORAL INTERPRETATION OF THE NON-PAST……………………………….
93
3.3.1. Aspectual Interpretation of the Non-Past………………………………
94
3.3.2. Recycle, Where Facilities Exist!..............................................................
99
3.3.2.1. Imperfective is Recycled as Present…………………………... 100 3.3.2.2. Perfective is Recycled as Future………………………………
103
3.3.3. The Double Interpretation: Perfective/Future and Imperfective/Present.. 104 3.4. THE PERIPHRASTIC FUTURE TENSE…………………………………………….
108
3.4.1. The Data………………………………………………………………..
108
3.4.2. Featuring the Future Imperfective……………………………………..
111
3.5. CONCLUSION…………………………………………………………………..
115
xi
CHAPTER FOUR: VALUE RECYCLING AND SUBJUNCTIVE MOOD.… 118 4.0. INTRODUCTION……………………………………………………………….... 118 4.1. MOOD AS A DYADIC PREDICATE THAT RELATES TIMES………………………… 120 4.1.1. Mood Relates T-EVL Relative to TU…………………………………… 120 4.1.2. The Featural Makeup of Mood…………………………………………. 126 4.2. THE SUBJUNCTIVE MOOD IN RUSSIAN………………………………………….. 132 4.2.1. Matrix Subjunctive Clauses…………………………………………….
132
4.2.2. Temporal Interpretation of Matrix Subjunctives……………………….
136
4.2.3. Agreement and Value Recycling………………………………………
138
4.3. MOOD IN EMBEDDED CLAUSES………………………………………………..
148
4.3.1. Embedded Subjunctive Clauses……………………………………….
148
4.3.1.1. Temporal Interpretation of Embedded Subjunctives…….…..
148
4.3.1.2. The Subjunctive Complementizer……………………………..
151
4.3.1.3. Featuring Embedded Subjunctives……………………………. 154 4.3.2. Summary……………………………………………………………….
158
4.4. IMPLICATIONS FOR REALIS MOOD….………………………………………….
158
4.5. CONCLUSION…………………………………………………………………..
165
CHAPTER 5: TENSE AND LONG-DISTANCE WH-MOVEMENT……..…
167
5.0. INTRODUCTION………………………………………………………………… 167 5.1. WH-MOVEMENT OUT OF INDICATIVES AND SUBJUNCTIVES IN RUSSIAN……….. 169 5.2. INDEPENDENT TENSE DOES NOT BLOCK WH-MOVEMENT……………………… 171 5.2.1. Wh-Movement and Sequence of Tense: Richards (2001:274–281) …... 172 5.2.2. Wh-Movement and Tense in Russian………………………………….. 176 5.2.3. Further Problems with Richards’ Generalization………………………
179
5.2.3.1. SOT is a Not a Property of Languages………………………… 179 5.2.3.2. The Obviation Phenomenon in Russian Subjunctives………...
182
5.2.4. Summary……………………………………………………………….
187
xii
5.3. RUSSIAN IS A TP-ABSORPTION LANGUAGE……………………..……………..
187
5.3.1. Multiple Wh-Movement and TP-Absorption versus CP-Absorption….
188
5.3.2. Diagnostics for TP-Absorption and CP-Absorption…………………..
194
5.3.2.1. Wh-Islands……………………………………………………
195
5.3.2.2. Weak Crossover and Scrambling…………………………….
200
5.3.2.3. Superiority……………………………………………………
204
5.3.2.4. Weak Crossover and Wh-Movement………………………...
207
5.3.3. Summary………………………………………………………………
211
5.4. RUSSIAN LONG-DISTANCE WH-MOVEMENT VIOLATES SUBJACENCY………..
212
5.4.1. Russian Complementizers: ČTO versus ČTOBY………………………..
212
5.4.2. French and Polish…………………………………………………….
217
5.4.3. Summary……………………………………………………………..
222
5.5. CONCLUSION…………………………………………………………………
223
CHAPTER 6: CONCLUSION AND ISSUES FOR FURTHER RESEARCH.. 224 6.0. INTRODUCTION TO THE CONCLUSION…………………………………………..
224
6.1. RESEARCH QUESTIONS AND THE ANALYSIS: SUMMARY……………………….
224
6.2. THEORETICAL IMPLICATIONS………………………………………………….
227
6.3. FUTURE RESEARCH……………………………………………………………
231
REFERENCES……………………………………………………………………
235
xiii
LIST OF ABBREVIATIONS 1ST 2ND 3RD AOR COIN COND DAT FEM FUT IND INF IMPER IMPF
LF LPART MASC NEUTER P PAST PERF PL PREF PRES REFL SG SOT
Spec SUBJ
TU T-AST T-SIT
first person second person third person aorist coincidence conditional particle dative case feminine future tense indicative complementizer infinitive imperative imperfective Logical Form the l-participle masculine neuter perfective(feature) past tense perfective(aspect) plural prefix present tense reflexive singular sequence of tense specifier subjunctive complementizer utterance time assertion time situation time
xiv
I don’t necessarily agree with everything I say. -- Marshall McLuhan
xv
1 CHAPTER 1 INTRODUCTION
1.0. STATEMENT OF THE PROBLEM In this dissertation, I explore the morphosyntactic expression of the relationship between tense, grammatical (viewpoint) aspect, and mood by examining the Russian tense system. In many languages, a single morphological form expresses a combination of tense and aspect. For example, the English simple past tense expresses not only past tense, but also perfective viewpoint. In the sentences Lisa broke a vase or Phil walked for an hour the past tense verb forms broke and walked not only express that the event denoted by the predicate occurred prior to the moment of speech, but also that the event in question was completed. Such forms are traditionally referred to as “tense”, whereas in fact, they convey both tense and aspect. In Russian, the aspectual morphology can carry both the aspectual and temporal information. Russian finite verbs obligatorily express temporal information, that is, Russian is a tensed language, as defined below. The basic distinction is between past and non-past tense verb forms, where only the former are overtly marked for both tense and grammatical aspect (the imperfective/perfective opposition). The non-past tense verb forms are only marked for aspect. They have no tense morphology, and the temporal reference is determined by the grammatical aspect: imperfective aspect gives rise to the present tense interpretation, whereas perfective aspect gives rise to the future tense interpretation. The intuitive idea is that the aspectual morphology performs a double duty
2 of specifying both the aspectual meaning and the temporal reference. The aspectual value, imperfective or perfective, is “transmitted” to tense, and is then reinterpreted accordingly as present or future.1 Similarly, many languages including Russian do not have a specific morpheme that expresses mood. In particular, the irrealis mood is often expressed by past tense morphology (James 1982; Givón 1994; Dahl 1997; Tynan and Lavín 1997; Iatridou 2000 and references there). For example, in a sentence such as I had a car the past tense verb form had conveys the past temporal reference. In contrast, in a sentence such as If I had a car I would give you a ride the same form had does not convey the past tense reference. Instead, the construction expresses that I don’t have a car at the time of utterance. Following much recent work, tense and aspect each head a maximal projection in syntax, TP and AspP, respectively. Furthermore, tense and aspect are represented as formal features, whose value determines the temporal and aspectual properties of a clause (Giorgi and Pianesi 1997; Demirdache and Uribe-Etxebarria 2000; Fassi Fehri 2004; Pesetsky and Torrego 2001, 2004a,b; Zagona 1990, 1995, 2003). Under this approach, the ability of an aspectual form under certain circumstances to convey both tense and aspect presents an interesting case. This implies that this aspectual form may become associated with both TP and AspP, and the value that it supplies may be shared by the formal features on the two heads. To ensure this state of affairs, TP and AspP, or the formal features on T and Asp must be able to “communicate” somehow.
1
At this point, I present these ideas in an informal way. The formal definitions and details are given in chapter 3.
3 The goal of this dissertation is to develop a formal morphosyntactic mechanism that captures this “communication” between AspP and TP, which would also be applicable to mood. This mechanism relies on the idea that formal features can “communicate” with each other via the syntactic operation of agreement (Pesetsky and Torrego 2004a,b). The analysis developed in this dissertation defines the nature of this syntactic
operation,
and
also
the
specific
circumstances
under
which
the
“communication” between two features can take place. I argue that the same approach can be applied to mood, and in particular to the case where past tense morphology receives an irrealis interpretation.
1.1. THE CATEGORIES OF TENSE AND ASPECT Much recent research focuses on the syntax of tense and aspect, whereas the syntax of mood is a less studied area. One of the goals of this dissertation is to extend the analysis of tense and aspect to mood, and thus develop a unified analysis for all three categories. In this section, I discuss tense and aspect, and leave the discussion of mood till chapter 4.
1.1.1. “Tense” versus “Temporal Reference” For the purposes of this dissertation, it is essential to establish what is meant by tense and temporal reference. Tense is a grammatical category, and like any grammatical category it may or may not be found in a given language. This category expresses the location of the event denoted by a predicate in time relative to the moment of speech (Comrie 1985; Dahl 1985; Chung and Timberlake 1985; Klein 1994). The notions present, past and
4 future designate the location of the event in question relative to the moment of speech. In contrast, temporal reference is a semantic notion. It refers to the information conveyed by tense. In a language where tense is a grammatical category, any finite clause must have temporal reference, that is, such a clause picks out an event which is located in time relative to the moment of speech. The distinction between tense and temporal reference is reflected in the fact that although all languages have means of expressing temporal information, they differ in whether or not the expression of such information is obligatory (Matthewson 2002; Ritter and Wiltschko 2005). Languages such as English or Russian may be referred to as “tensed” languages: In these languages, tense is a grammatical category, and the temporal reference of any finite clause is obligatorily specified by verbal morphology. For example, the propositional content of the English sentences in (1) is the same,
. The sentences differ in the temporal location of the event of Philemon eating fish in time relative to the moment of speech:
(1)
a. Philemon is eating fish.
(present: coincides with/includes the moment of speech)
b. Philemon ate fish.
(past: precedes the moment of speech)
c. Philemon will eat fish.
(future: follows the moment of speech)
In each sentence in (1) the temporal reference is encoded in the verbal morphology: the auxiliary is in (1a), the past tense verb form ate in (1b), and the future auxiliary will in
5 (1c). The choice of a form is obligatory; an English finite sentence may not be unspecified with respect to its temporal reference. One should distinguish between the meaning of the tense morphology on the one hand, and the contexts of its use on the other. For example, it is possible to use a verb with the present tense morphology to refer to an event in the past. This use of the present tense morphology is referred to as historical present, as in (2a). Similarly, the present tense morphology can also be used to refer to an event in the immediate future, as in (2b):
(2)
a. At the urging of his marshals, Napoleon abdicates on April 6, 1814 in favor of his son. b. I am moving to New York next month.
It is important to emphasize that examples such as in (2) do not mean that English finite clauses with the present tense morphology are underspecified with respect to temporal reference. The meaning of the present tense is precisely what allows for such an extended use. The purpose of the historical present in (2a) is to present the historical event as if it was taking place at the moment of speech, thus making it more vivid. In (2b), the future event referred to by the present tense morphology suggests that the event is certain to occur, and may be regarded as happening now. In contrast, languages such as Inuktitut can be referred to as “tenseless”: In Inuktitut, there is no grammatical category tense, and the temporal information is not encoded on the verb. However, this does not mean that Inuktitut lacks means to convey
6 the temporal reference; rather, these means are fundamentally different from those in English. According to Swift (2004), Inuktitut manifests a realis-irrealis split, with the verbs overtly marked for the irrealis mood conveying the future time reference. Verbs that have no overt marking for tense, aspect, or modality convey the non-future time reference, with the further distinction between past and present determined by the verb’s lexical meaning. For example, the expressions in (3) have the same structure: They both contain a verb with the inflectional agreement incorporating person, number, and mood. And yet, they differ in their temporal reference: (3a) refers to an event that precedes the moment of speech, whereas (3b) refers to an event that coincides with, or includes the moment of speech:
(3)
a. Anijuq.
b. Pisuttuq.
ani-juq
pisuk-juq
go.out-MOOD.3RD.SG
walk-MOOD.3RD.SG
‘She went out.’
‘She is walking.’
(Swift 2004:22–23)
The event in (3a) takes place prior to the moment of speech; moreover, it is completed by the moment of speech. The event in (3b) takes place at the moment of speech, and is naturally incomplete. These data suggest that in Inuktitut, the past or the present temporal reference is determined by the inherent temporal structure of the verb stem. Swift (2004) argues that verbs that denote a change of state, such as (3a) are interpreted as perfective/past. Verbs that denote ongoing activities, such as (3b) are interpreted as
7 imperfective/present. Thus, it may be said that in Inuktitut, the temporal reference is determined by lexical aspect. This discussion brings us to the category of aspect, which is closely related to the category of tense. Lexical and grammatical aspects are discussed in the next section.
1.1.2. Lexical Aspect and Grammatical (Viewpoint) Aspect Unlike tense, aspect does not locate an event in time relative to some other time. Instead, aspect is concerned with the internal structure of an event (Comrie 1976; Chung and Timberlake 1985; Dahl 1985; Binnick 1991; Smith 1991). Following Smith (1991), two types of aspect should be distinguished: lexical, or situation type aspect (aktionsart) and grammatical, or viewpoint aspect. In this dissertation, I am concerned exclusively with grammatical aspect. However, since both types of aspect can affect the temporal reference of the clause, in this section I discuss both of them, so that it is clear what properties of the predicate are examined in this dissertation. Lexical aspect is concerned with the inherent temporal structure of the predicate, which arises from the lexical meaning of the verb and its complements. The best-known classification of verb meanings based on lexical aspectual properties is due to Vendler (1967). Vendler classifies predicates into states (know, love), activities (walk, sing), achievements (reach the top, recognize), and accomplishments (draw a circle, eat an apple). Smith (1991) terms these classes situation types; she adds the fifth situation type to this inventory, which she refers to as “semelfactive” (tap, knock). The five classes are defined in terms of stativity, durativity, and telicity as follows:
8 (4)
Temporal-aspectual classification of predicates (Smith 1991) Situation type
Stative
Durative
Telic
State
+
+
–
Activity
–
+
–
Semelfactive
–
–
–
Achievement
–
–
+
Accomplishment
–
+
+
According to (4), the five types of predicates can be divided into two groups based on (a)telicity, that is, the presence or absence of an inherent end point. States, activities and semelfactives are atelic predicates: They denote ongoing situations with no potential end point. Such situations can be terminated, but they cannot be completed, as there is no natural end point which would be associated with the completion of the situation. For example, there is no natural end point associated with loving someone (state), walking (activity) or knocking on the door (semelfactive). In contrast, achievements and accomplishments are telic predicates: They denote situations with natural end points. When such an end point is reached, the situation in question is completed. For example, events such as eating an apple or reaching the top have a natural end point: once the apple is entirely consumed and the top is reached the event in question is completed and cannot continue. Predicates that express a change of state are naturally classified as telic: once the change is completed the event in question is over.
9 Tenseless languages often use (a)telicity as a means of distinguishing between present and past temporal references (see, for example, Swift 2004 for Inuktitut; Lin 2005 for Chinese). Telic events are viewed as completed, and the notion of completedness can be naturally associated with an event in the past. One of these languages is Inuktitut, discussed in section 1.1.1 above. These examples are repeated below: (5a) is a telic predicate; it expresses a change of location. As a result, (5a) conveys the past temporal reference. In contrast, (5b) is an atelic predicate; it denotes an event of walking, which does not have an inherent end point. Thus, (5b) conveys the present temporal reference:
(5)
a. Anijuq.
b. Pisuttuq.
ani-juq
pisuk-juq
go.out-MOOD.3RD.SG
walk-MOOD.3RD.SG
‘She went out.’
‘She is walking.’
I demonstrate in section 1.2.1 that situation type aspect does not play a role in determining the temporal reference of a Russian finite clause. Therefore, for the purposes of this dissertation I put this aspect aside. Henceforth, I will mainly be concerned with the properties of grammatical, or viewpoint, aspect as discussed below. Grammatical, or viewpoint, aspect is concerned with different ways of viewing the internal temporal constituency of a situation (Comrie 1976; Smith 1991). The two major categories of grammatical aspect are perfective and imperfective. Perfective aspect presents a situation as a whole, without reference to its internal structure. Imperfective
10 aspect views the situation from inside, or as an ongoing event. It is often understood as an inflectional property reflected in the morphosyntax. For example, the difference between Philemon ate fish and Philemon was eating fish is that of viewpoint aspect: The former presents a situation as complete (perfective), whereas the latter presents the same situation as ongoing (imperfective). Grammatical aspect also can be used as a means for conveying the temporal reference. For example, as argued in Lin (2005), Chinese does not have tense, and the temporal reference of a clause is determined, among other things, by viewpoint aspect: Perfective aspect conveys past temporal reference, whereas imperfective aspect conveys present temporal reference. To sum up the discussion so far, there are two types of aspect: lexical aspect and grammatical aspect. In languages with no tense morphology, such as Inuktitut and Chinese, aspect can serve as a means of conveying the temporal reference of a clause. In Inuktitut, telic predicates, that is, predicates that denote events with an inherent end point, have past temporal reference, whereas atelic predicates have present temporal reference. In Chinese, the distinction between past and present is conveyed by means of the perfective/imperfective opposition. In the next section, I show that Russian, which is a tensed language, also uses aspectual morphology to convey temporal reference.
11 1.2. DATA AND SPECIFIC RESEARCH QUESTIONS 1.2.1. Aspect and the Past vs. Non-Past Opposition in Russian As far as temporal reference is concerned, Russian makes a three-way distinction between past, present, and future. However, in morphology, Russian makes the fundamental distinction between past and non-past stems (Vinogradov 1947/1971:543ff; Borik 2002:138; Borik, González and Verkuyl 2003; Verkuyl 2005). In the past, verbs are morphologically marked for tense and aspect, as shown in (6). The suffix –l conveys the past tense reference. The morphologically simple verb in (6a) is interpreted as imperfective, whereas the prefixed verb in (6b) is interpreted as perfective:
(6)
a. Philemon je-l
jabloko.
Philemon eat-PAST(IMPF) apple ‘Philemon was eating a/the apple.’ b. Philemon s-je-l
jabloko.
Philemon PREF-eat-PAST(PERF) apple ‘Philemon ate / has eaten an/the apple.’
Thus, temporal reference and aspectual information are expressed by distinct grammatical morphemes and are interpreted independently of one another. In the non-past, verbs contain no tense morphology to mark further distinction between present and future; they only contain aspectual morphology. In the absence of tense morphology, imperfective verbs receive a present tense interpretation, whereas
12 perfective verbs receive a future tense interpretation. The verbs in (7)–(8) have the same agreement morphology (person and number) and differ only in aspect. Note that lexical aspect does not affect the temporal reference: the predicates in (7a) and (8a) are accomplishments, whereas the predicates in (7b) and (8b) are activities. For both situation types, the temporal reference is determined by grammatical aspect:
(7)
a. Philemon je-st
jabloko.
(accomplishment; telic)
Philemon eat-3RD.SG(IMPF) apple ‘Philemon is eating a/the apple.’ b. Philemon guljaj-et.
(activity; atelic)
Philemon walk-3RD.SG(IMPF) ‘Philemon is taking a walk.’
(8)
a. Philemon s-je-st
jabloko.
(accomplishment; telic)
Philemon PREF-eat-3RD.SG(PERF) apple ‘Philemon will eat a/the apple.’ b. Philemon po-guljaj-et. Philemon
(activity; atelic)
PREF-walk-3RD.SG(PERF)
‘Philemon will take a walk.’
Based on the data in (6)–(8), Russian tense verb forms may be divided into two groups based on the role of aspect in determining the temporal reference. In the non-past,
13 aspect determines the temporal reference of a clause (present or future). In the past, aspect does not play a role in determining the temporal reference. This property of the Russian tense system provides the basis for the first research question addressed in this dissertation:
Q1: How can the relationship between tense and aspect in the non-past be represented in the morphosyntax?
The answer to Q1 is provided in chapter 3.
1.2.2. Past Tense Morphology and Subjunctive Mood Russian uses past tense morphology to express the irrealis mood. In particular, subjunctive clauses are formed by the morphological past tense verb form in combination with the conditional particle by in matrix clauses, or the subjunctive complementizer čtoby in embedded clauses, as illustrated in (10a) and (10b), respectively:2
(10) a. Ja by I
COND
uš-l-a
domoj.
leave-PAST-FEM.SG home
‘I would go / would have gone home.’
2
At this point, I am not making any claims as to the syntactic status and the position of by and čtoby. These issues are addressed in chapters 4 and 5.
14 b. Ja xoču, čtoby vy I want
SUBJ
osta-l-is’.
you-PL stay-PAST-PL
‘I want you to stay.’ As discussed in section 1.0, many languages use past tense morphology to express the irrealis mood. Therefore, I assume that this is not an arbitrary choice; rather, past tense morphology carries a feature or features which make it suitable for the expression of both the past temporal reference and the irrealis mood. The second research question addressed in this dissertation is stated below:
Q2: How can the interpretation of the past tense morphology be accounted for in terms of the morphosyntactic environment?
Q2 is the focus of chapter 4.
1.2.3. Tense and Long-Distance Wh-Movement Russian indicative and subjunctive clauses contrast with respect to long-distance whmovement. Wh-movement is problematic out of embedded indicative clauses, which have independent temporal reference. In contrast, wh-movement is acceptable out of embedded subjunctive clauses, whose temporal reference is determined by the matrix clause. This contrast is illustrated in (12):
15 (12) a. ??(*)Kogoi Anna dumajet, [čto Liza ljubit ti ]? who
Anna thinks
IND
(indicative)
Liza loves
‘Whom does Anna think that Liza loves?’ b. Kogoi Anna xočet, [čtoby Liza ljubi-l-a who Anna wants
SUBJ
ti ]?
(subjunctive)
Liza love-PAST
‘Who does Anna want Liza to love?’
The contrast between indicative and subjunctive embedded clauses with respect to longdistance wh-movement is the basis for the third research question:
Q3: Can the independent tense content block long-distance wh-movement?
Q3 is addressed in chapter 5.
1.3. PROPOSAL I adopt a minimalist approach within the Principles and Parameters theory (Chomsky 1995, 2000). I assume that grammatical categories are represented in the syntax in the form of features which receive their value from overt morphology that expresses these categories. Following Pesetsky and Torrego (2001, 2004a,b), at the onset of a derivation features can be valued or unvalued, and interpretable or uninterpretable; however, a completed derivation may not contain unvalued features. I also assume that categories such as tense and aspect express properties of clauses, and not lexical items. In particular,
16 the value and interpretation of features that express these categories is derived compositionally from, or read off a particular syntactic configuration. In this respect, tense and aspect features contrast with features such as number, which express properties of lexical items. The contrast between features of lexical items and features of clauses is reflected in how the two types of features receive their semantic interpretation. In order to be able to identify a noun as plural we do not need to insert it into a clausal structure. Similarly, for an adjective to show plural agreement no clausal structure is required. Further, in languages that morphologically distinguish number, this category is typically associated with a piece of morphological material reserved specifically for these purposes. In contrast, it is only appropriate to talk about tense, aspect, or mood when we have a clause.3 The temporal or the aspectual interpretation should not be confused with the presence of what can potentially be a tense or an aspect morpheme on a verb. For example, English regular verb forms such as played contain the suffix –ed. However, we need a clause (or at least a whole predicate) to establish whether we are dealing with a past tense form, as in (14a), or with a tenseless participle, as in (14b) and (14c). The same is true for irregular forms, such as left:
(14) a. Joan played a sonata. / Philemon left. b. Joan has/had played a sonata. / Philemon has/had left. c. The sonata played by Joan… / A piece of cake left for Philemon… 3
This is not to mention temporal interpretation of DPs (LeCarme 1999, 2004; Musan 1995).
17 Thus, the tense and aspect features are not primitives in the sense that their interpretation cannot be determined by the abstract value of a feature carried by a particular morpheme, such as [±PAST]. Rather, their interpretation arises from the interaction of overt morphology with other components of the structure. I follow Pesetsky and Torrego’s (2004b) proposal in referring to this interaction as syntactic agreement between features. The analysis in this dissertation is based on the idea that tense, aspect, and mood are distinct grammatical categories, which express different relations. However, these relations are similar in some relevant respect, and can be characterized in terms of a single notion. Following much recent work, I analyze tense, aspect, and mood as dyadic predicates, each of which orders, or relates two temporal arguments in syntax (Zagona 1990, 1995, 2003; Klein 1995; Stowell 1995, 1996; Giorgi and Pianesi 1997; Demirdache and Uribe-Etxebarria 2000). I propose that in each case, the relation between the temporal arguments can be characterized as a relation of (non)coincidence. This relation is represented as the interpretable feature [±COIN] on C, T, and Asp. This feature serves as the basis for a semantic interpretation. In addition, C, T, and Asp each contain a purely formal feature, which licenses their temporal arguments in syntax. I argue that the independent tense content cannot be held responsible for the unavailability of longdistance wh-movement. I propose that the problematic status of wh-movement out of Russian indicative clauses is due to a subjacency violation.
18 1.4. ORGANIZATION OF THE THESIS Chapter 2 provides an overview of the Russian tense system. I discuss in detail the tense and aspectual morphology of Russian, and the role of viewpoint aspect (the perfective/ imperfective opposition) in determining the temporal reference of Russian finite clauses. In chapter 3, I propose an analysis of the temporal interpretation of Russian finite clauses. In chapter 4, I extend the analysis developed in chapter 3 to the irrealis mood. In Russian, the irrealis mood is conveyed by past tense morphology in combination with the conditional particle. I propose to account for this in terms of value recycling. In chapter 5, I discuss the relationship between tense and long-distance wh-movement. I argue contra Richards (2001), that the interpretable content on T does not block wh-movement, and that the temporal dependency between the matrix and the embedded clause, often referred to as sequence of tense, does not license it. I propose that in Russian, long-distance whmovement out of indicative clauses violates subjacency, which accounts for its problematic grammaticality status. Finally, chapter 6 summarizes the main points of the dissertation and outlines issues for further research.
19 CHAPTER 2 OVERVIEW OF THE RUSSIAN TENSE SYSTEM
2.0. INTRODUCTION This chapter introduces the Russian tense system. I present the data that provide the basis for my research questions stated in chapter 1, and which are the focus of the analysis developed in the subsequent chapters. The questions are repeated below:
Q1: How can the relationship between tense and aspect in the non-past be represented in the morphosyntax?
Q2: How can the interpretation of the past tense morphology be accounted for in terms of the morphosyntactic environment?
Q3: Can the independent tense content block long-distance wh-movement?
I discuss viewpoint aspect (the perfective/ imperfective opposition) and its role in determining the temporal reference of Russian finite clauses. It is often claimed that in the domain of tense, Russian has one main opposition: past versus non-past (Vinogradov 1947/1971:543ff; Borik 2002:138; Borik, González and Verkuyl 2003; Verkuyl 2005).1
1
Švedova (1980, vol. 1:626) distinguishes between three forms: past, present, and periphrastic future, or future imperfective. The future imperfective is a complex tense formed by the auxiliary byt’ ‘to be’ and an imperfective infinitive, and is discussed separately in section 2.3.2.
20 The past and non-past verb forms have different stems and show different agreement: Past tense verb forms agree with the nominative subject in number and gender. Non-past tense forms agree with the subject in person and number. Past and non-past tense verb forms also differ in how their temporal interpretation arises. The past tense verb forms are marked morphologically for both tense and aspect. In contrast, the non-past tense forms are marked morphologically only for aspect, which determines the further distinction between present and future. This is exemplified in (1) and (2) for the verb čitat’ ‘to read’. The sentences in (1) differ in aspect, but both have the past tense interpretation. The sentences in (2) differ in both aspect and the tense interpretation: The imperfective (2a) and the perfective (2b) have a present and a future interpretation, respectively:2
(1) Past tense stem: čitaa. Liza čita-l-a
knigu.
Liza read-PAST-FEM.SG(IMPF) book ‘Liza was reading a/the book.’ b. Liza pro- čita-l-a Liza
PREF-read-PAST-FEM.SG(PERF)
knigu. book
‘Liza read/has read a/the book.’
2
I do not gloss the prefix in (1b) and (2b) as PERF. I discuss the reasons for that in section 2.1.
21 (2) Non-past tense stem: čitaja. Liza čitaj-et knigu.
(present)
Liza read-3RD.SG(IMPF) book ‘Liza is reading a/the book.’ b. Liza pro-čitaj-et Liza
knigu.
PREF-read-3RD.SG(IMPF)
(future)
book
‘Liza will read a/the book.’
I demonstrate that the temporal interpretation is inherent to the verb form, and does not arise via the pragmatic context. I argue also that future in Russian should be treated as tense, and not as mood. This chapter is organized as follows. In section 2.1, I discuss the morphological and semantic characteristics of the Russian aspectual opposition perfective/imperfective. Section 2.2 deals with the past tense verb forms, their morphology and historical development. I discuss their use in the subjunctive mood and the range of interpretations available for these forms. In section 2.3, I discuss the non-past tense verb forms. I demonstrate that although they contain no tense morphology, the aspectual morphology supplies the present/future tense interpretation. I discuss the periphrastic future tense formed by the auxiliary byt’ ‘to be’ and imperfective infinitive and provide arguments that for the purposes of syntax, future in Russian should be treated as tense and not as mood. In section 2.4, I summarize the main points of this chapter and outline the approach to be taken in chapters 3 and 4.
22 2.1. RUSSIAN ASPECT: THE PERFECTIVE/IMPERFECTIVE OPPOSITION The category of aspect is closely related to the category of tense, and the Russian temporal system is often defined in terms of aspect as follows (Švedova 1980, vol. 1:626–636; Borik 2002). Imperfective aspect allows for the formation of three tense forms: past, present, and future (complex, or periphrastic future tense). Perfective aspect allows for only two tense forms: past and future; there is no present perfective form. The Russian tense system is exemplified below.
(3) Past
Imperfective Vasja čita-l
Perfective knigu.
Vasja pro-čita-l
knigu.
Vasja read-PAST.MASC.SG book
Vasja PREF-read-PAST.MASC.SG book
‘Vasja was reading/read a/the book.’
‘Vasja has read/read a/the book.’
Present Vasja čitaj-et
knigu.
Vasja read-3RD.SG book ‘Vasja is reading a/the book.’ Future
Vasja bud-et
čitat’
knigu.
Vasja pro-čitaj-et
knigu.
Vasja be-3RD.SG read-INF book
Vasja
book
‘Vasja will be reading a/the book.’
‘Vasja will read a/the book.’
PREF-read-3RD.SG
In this section, I discuss Russian aspect in more detail.3 In 2.1.1, I discuss its morphological characteristics. In 2.1.2, I deal with the semantic properties of the perfective and the imperfective aspect.
3
It should be mentioned that the literature on Russian aspect is immense, and here I do not attempt to provide an overview of all the relevant work. The goal of the discussion here is to provide the necessary information about aspect, without loading the reader with too many details. For a detailed treatment of the contemporary theory of Russian aspect see Zaliznjak and Šmelev (1997) and references cited there. Note
23 2.1.1. Morphological Characteristics Aspect is a grammatical category in Russian, and there is no verb form, either finite or non-finite, which is not morphologically marked for aspect. Moreover, nearly all verbs come in aspectual pairs, that is, every verb has the perfective and the imperfective form. The vast majority of morphologically simple verbs are imperfective. This means that such stems contain no overt morpheme that conveys the imperfective aspect. For this reason, the aspect in (4) is given in parenthesis:
(4)
a. pisat’
‘to write(IMPF)’
b. čitat’
‘to read(IMPF)’
c. kričat’
‘to scream(IMPF)’
d. nesti
‘to carry(IMPF)’
There are exceptions to this rule. Certain morphologically simple stems are perfective, as illustrated in (5). And again, there is no morphological exponent of the perfective aspect:
(5)
a. prostit’
‘to forgive(PERF)’
b. pustit’
‘to let in, throw(PERF)’
c. kupit’
‘to buy(PERF)’
that Zaliznjak and Šmelev’s (1997) book is published in Russian, but their references include works published in English and German as well.
24 Each of the verbs in (4)–(5) has a perfective/imperfective counterpart. With respect to verbs such as in (4), the major means of deriving the perfective form is prefixation.4 Russian has about twenty prefixes, which have developed out of directional adverbs and prepositions. Most of these prefixes still retain some of their directional meaning. When combined with a morphologically simple imperfective verb, they often not only perfectivize it, but also affect its lexical properties. Therefore, prefixes may not be regarded as inflectional markers of perfectivity, and this is the reason I do not gloss them as exponents of the perfective aspect.5, 6 Most verbs can combine with more than one prefix, and typically there is only one prefix which only affects the aspectual properties of the verb. The rest would also induce some change in the meaning. This is illustrated in (6):
(6)
pisat’ ‘to write (IMPF)’
a. na-pisat’ ‘to write(PERF)’ b. pod-pisat’ ‘to sign(PERF)’ c. za-pisat’ ‘to jot down(PERF)’ d. vy-pisat’ ‘to write out(PERF)’
4
Other means include suffixation or suppletion, as in (ia) and (ib), respectively:
i. a. kolot’ ‘to stub(IMPF)’ b. brat’ ‘to take(IMPF)’ 5
kol’-nu-t’ stub-PERF-INF ‘to stub( PERF) once (semelfactive)’ vzjat’ ‘to take(PERF)’
See Filip (2000) for an extensive discussion and a number of convincing arguments that prefixes may not be considered inflectional morphemes.
25 In (6a), the prefix na- only affects the aspectual properties of the imperfective verb pisat’ ‘to write’. In (6b–d), the prefixes not only perfectivize the verb, but also change its meaning. The prefixed perfective verbs in (6b–d) are new verbs, and as such require their own imperfective counterpart. Such a counterpart is created by adding the imperfectivizing suffix –(yv)aj. The resulting verb form is referred to as “secondary imperfective”. This suffix may be regarded as a true inflectional aspectual morpheme. This is illustrated in (7):
(7)
a. pod-pisat’ ‘to sign(PERF)’ b. za-pisat’ ‘to jot down(PERF)’ c. vy-pisat’ ‘to write out(PERF)’
pod-pis-yva-t’ PREF-write-IMPF-INF
‘to sign(IMPF)’
za-pis-yva-t’ PREF-write-IMPF-INF
‘to jot down(IMPF)’
vy-pis-yva-t’ PREF-write-IMPF-INF
‘to write out(IMPF)’
The imperfective counterparts of exceptional morphologically simple perfective verbs, such as those in (5) above, are derived by the same imperfectivizing suffix. The shape of the stem may be affected in the process of derivation, as shown in (8):
6
There is nothing “perfective” about the meaning of Russian prefixes. To the best of my knowledge, the question of how the perfective interpretation arises from the meaning of the stem and the prefix has not been explicitly addressed in the literature. See Mezhevich (2006a,b) for my own attempt.
26 (8)
a. prostit’ ‘to forgive(PERF)’ b. pustit’ ‘to let in, throw(PERF)’ c. kupit’ ‘buy(PERF)’
prošč-a-t’ forgive-IMPF-INF ‘to forgive(IMPF)’ pusk-a-t’ let-in-IMPF-INF ‘to let in, throw(IMPF)’ po-kup-a-t’ PREF-buy-IMPF-INF
‘to buy(IMPF)’
In this thesis, I am concerned with the aspectual classification of a given verb form, and a change from imperfective to perfective. The morphological device employed, or any lexical changes accompanying the aspect change, are irrelevant for my purposes. Therefore, throughout the thesis I do not distinguish between different morphological strategies for expressing a given aspect, and gloss all verb forms as either perfective or imperfective, by indicating the aspect in parentheses. Also, I gloss all prefixes as “PREF” without referring to their particular meaning.
2.1.2. Semantic Characteristics What is more important for my purposes is the semantics of the perfective and the imperfective aspect, since in the non-past, this opposition defines the present/future distinction. Grammatical (viewpoint) aspect does not locate an event with respect to the speech time, but is concerned with the internal structure of the event (Comrie 1976; Dahl 1985; Binnick 1991). Perfective aspect presents an event from outside, without distinguishing any of its internal temporal structure; it is often said to indicate a complete
27 action. Imperfective aspect presents an event from inside. It makes explicit reference to the internal temporal structure of the event, and thus denotes an incomplete action. To give the reader an idea about the Russian aspectual system I base the discussion in this section on Smith and Rappaport (1991).7 However, as we will see in chapter 3, their view of the Russian perfective and imperfective aspects does not reflect the way these concepts are formalized in my analysis.
2.1.2.1. Imperfective Aspect Imperfective aspect focuses on a situation in progress, excluding its initial and final endpoints. It is formally available to all situation types, and is therefore the dominant, or unmarked, aspect. Below are examples from Smith and Rappaport (1991:302–303):
(9)
a. Vanja znal
ivrit.
(state)8
Vanja knew(IMPF) Hebrew ‘Vanja knew Hebrew.’ b. Vanja pel
v parke.
(activity)
Vanja sang(IMPF) in park ‘Vanja sang in the park.’
7
According to Klein (1995), there is no generally accepted semantic definition of the Russian aspects. He provides a detailed overview of various attempts to characterize them semantically. 8
For expository purposes, I replaced Smith and Rappaport’s original example in (9a). Their original example involved the verb golodat’ ‘be starving’. Although it is a state, it is a stage level predicate, and as such has potential endpoints. I return to this point immediately below.
28 c. My pisali
pis’mo.
(accomplishment)
we wrote(IMPF) letter ‘We were writing a letter.’ d. On stučal
v okno.
(semelfactive)
he knocked(IMPF) in window ‘He was knocking at the window.’ e. On umiral.
(achievement)
he died(IMPF) ‘He was dying.’
The imperfective verbs in (9) describe various situation types in progress, with no reference to the initial or the final endpoint. Note that Smith and Rappaport’s definition of imperfective is too specific when it comes to stative predicates. Although the stative verb in (9a) is morphologically imperfective, it does not make sense to say that the initial and the final endpoints of the situation it describes are invisible. Individual level predicates such as know Hebrew refer to states that are simply lacking these endpoints. However, this informal characterization does work for other situation types. Smith and Rappaport (1991) demonstrate that neither the initial nor the final point of the situation is visible in the imperfective aspect. They use temporal conjunctions with whenclauses, considering the possible interpretations of the main clause. Since the initial point of the situation is invisible, the situation denoted by the main clause must already be under way at the time of the situation denoted by the adverbial clause. For example, in
29 (10a), the situation of Vanja’s singing in the park must begin before Nina’s appearance. Similarly, in (10b), the situation of our writing a letter must begin before the calling situation happens. An inceptive interpretation of the matrix clauses is impossible:
(10) a. Vanja pel
v parke, kogda Nina pojavilas’.
Vanja sang(IMPF) in park
when Nina appeared
‘Vanja was singing in the park when Nina appeared.’ b. My pisali
pis’mo, kogda on pozvonil.
we wrote(IMPF) letter
when
he called
‘We were writing a letter when he called.’
To show that the final point of the situation is invisible, Smith and Rappaport (1991) give the examples in (11), where imperfective sentences are conjoined with the assertion that the situation continues. If the final point of the situation is visible, the sentences in (11) are expected to be pragmatically odd, contrary to fact:
(11) a. Vanja žil
v Leningrade, i
ešče tam živet.
Vanja lived(IMPF) in Leningrad and still there lives ‘Vanja was living in Leningrad, and still lives there.’ b. My pisali
pis’mo, i
we wrote(IMPF) letter
ešče pišem ego.
and still write it
‘We were writing a letter, and are still writing it.’
30 This property of the imperfective aspect, namely, that it picks out a situation in progress and does not expose either the initial or the final point, makes it highly compatible with the present tense meaning. However, an incomplete event in progress can also be described as taking place in the past or in the future. Thus, the Russian imperfective aspect allows for the formation of all three tense forms, as shown in (3).
2.1.2.2. Perfective Aspect Unlike the imperfective aspect, the perfective aspect is not available for stative verbs. This is illustrated in (12). In (12e), the imperfective form of the stative verb vesit’ ‘to weigh’ is combined with three different perfectivizing prefixes, and the result is ungrammatical:
(12) a. On posidel
v parke.
(activity)
he sat(PERF) in park ‘He sat for a while in the park.’ b. On napisal
pis’mo.
(accomplishment)
he wrote(PERF) letter ‘He wrote a letter.’ c. On stuknul
v okno.
he knocked(PERF) in window ‘He knocked at the window.’
(semelfactive)
31 d. Vanja vyigral
matč.
(achievement)
Vanja played(PERF) match ‘Vanja won the game.’ e. Slon
vesil
(*po-vesil/*na-vesil/*s-vesil) dve tonny.
elephant weighed(IMPF) (wighed(PERF))
(state)
two tons
‘The elephant weighed two tons.’
According to Smith and Rappaport (1991), the perfective aspect presents a situation with both initial and final endpoints. In chapter 3, I show that this is not always true. However, for the purposes of this discussion, this definition will do to make the point. To illustrate the presence of the initial point of the situation, the perfective sentence is used in the context of a when-clause. In (13), only the sequential interpretation is possible; the situation denoted by the main clause must begin after the situation denoted by the when-clause is completed:
(13) a. Kogda on vyšel
iz
doma, on posidel
v parke.
when he went.out(PERF) from house he sat.for.a.while in park ‘When he went out of the house, he sat in the park for a while.’ b. Kogda samolet vzletel, when plane
načalas’ snežnaja burja.
took.off(PERF) began
snow
storm
‘When the plane took off, the snow storm began.’
(Borik 2001)
32 In (13a), the situation of Vanja sitting in the park follows the situation of Vanja leaving the house. In (13b), the snow storm begins after the plane takes off. To show that the perfective aspect also includes the final point of the situation, Smith and Rappaport (1991) show that perfective sentences are incompatible with the assertion that the situation continues. The sentences in (14) are contradictions:
(14) a. #My napisali
pis’mo, i
we wrote(PERF) letter
ešče pišem ego.
and still write it
‘We were writing a letter, and are still writing it.’ b. #On posidel he
v parke, i
ešče tam sidit.
sat.for.a.while(PERF) in park and still there sits
‘He sat for a while in the park, and is still sitting there.’
If the final point were invisible, the sentences in (14) would be semantically well-formed, just like their imperfective counterparts in (11) above. Since the perfective aspect presents a situation with both the initial and the final points visible, it may not be compatible with a present tense interpretation. In particular, if the final point is visible, the event is presented as complete, whereas an event that occurs in the present cannot be complete. Thus, the perfective aspect in Russian allows the formation of the past and the future forms, but not present forms, as shown in (3). In the next section, I turn to the discussion of Russian past tense verb forms, and their temporal and aspectual interpretation.
33 2.2. PAST TENSE VERB FORMS IN RUSSIAN 2.2.1. Temporal and Aspectual Interpretation of Past Tense Verb Forms Past tense verb forms in Modern Russian are morphologically marked for both tense and aspect. They are created by adding the suffix –l, which is the only overt tense marker in the language, to a past tense stem.9 In section 2.2.3, I discuss the historical origins of the past tense verb forms, and show that morphologically, it is a participle and not a verb. Synchronically, however, it has all the verbal characteristics. As discussed in section 2.1.1, morphologically simple verbs are usually imperfective. The perfective form is created most often by adding a prefix. The past tense verb forms agree with the nominative subject in number and gender. The past tense paradigm for the verb čitat’ ‘to read’ is exemplified in (15):
(15) Sing
Plural
9
Imperfective
Perfective
čita-l
pro-čita-l
read-PAST-MASC.SG
PREF-read-PAST-MASC.SG
čita-l-a
pro-čita-l-a
read-PAST-FEM.SG
PREF-read-PAST-FEM.SG
(čita-l-o)
(pro-čita-l-o)
(read-PAST-NEUTER.SG)
(PREF-read-PAST-NEUTER.SG)
čita-l-i
pro- čita-l-i
read-PAST-PL
PREF-read-PAST-PL
There are a few exceptions to this rule. For discussion see Švedova (1980, vol. 1:628).
34 If past tense verb forms are morphologically marked for both tense and aspect, they are expected to have both a temporal and an aspectual interpretation. Moreover, their past tense reference should not depend on their aspectual properties. Below I demonstrate that this is indeed the case. The temporal reference “before the speech time” is independent of whether the event is viewed from the inside or outside. As for the past temporal reference, it is not determined by the pragmatic context or time adverbials, but is inherent to the predicate itself. As illustrated in (16), the use of time adverbials is optional. However, when a time adverbial is used it has to match the temporal reference of the clause. The fact that only yesterday is acceptable in (16) suggests that these clauses are in the past:
(16) a. Liza čita-l-a
knigu (*sejčas/*zavtra/včera).
Liza read-participle-FEM.SG(IMPF) book
now/tomorrow/yesterday
‘Liza was reading a/the book *now/*tomorrow/yesterday.’ b. Liza pro- čita-l-a
knigu (*sejčas/*zavtra/včera).
Liza PREF-read-participle-FEM.SG(PERF) book
now/tomorrow/yesterday
‘Liza has read a/the book *now/*tomorrow/yesterday.’
Thus, both sentences in (16) describe the event of Liza reading a book as taking place before the speech time. However, they differ in how they present this event. In (16a), the imperfective verb describes the event of reading in progress. At some point in the past, Liza was engaged in the activity of reading a book. In contrast, the perfective
35 verb in (16b) presents the same event as complete. At some point in the past, Liza finished reading her book. This contrast gives rise to two different interpretations in the context of a whenclause, as shown in (17). In (17a), the event of Liza reading the book necessarily overlaps with the event of Phil phoning. In (17b), the event of reading is most naturally interpreted as being completed by the time Phil phones:
(17) a. Kogda pozvoni-l
Phil, Liza čita-la
knigu.
when phone-PAST(PERF) Phil Liza read-PAST(IMPF) book ‘When Phil phoned Liza was reading a/the book.’ b. Kogda pozvoni-l
Phil, Liza pro- čita-la
when phoned-PAST(PERF) Phil Liza
PREF-read-PAST(PERF)
knigu. book
‘When Phil phoned Liza had read a/the book.’
The sequential interpretation in (17b) is the most natural interpretation. However, in section 2.2.3, I demonstrate that the event denoted by the perfective verb can also follow the event denoted by the when-clause. In the next section, I demonstrate that the morphological past tense form is also used in subjunctive clauses, where it does not receive a past tense interpretation.
36 2.2.2. Past Tense Verb Forms and Subjunctive Mood The morphological past tense verb form is also used in the subjunctive mood, in combination with the conditional particle by in matrix clauses, or the subjunctive complementizer čtoby in embedded clauses. The examples are given in (18a) and (18b), respectively:
(18) a. Ja by I
COND
uš-l-a
domoj.
leave-PAST-FEM.SG home
‘I would go / would have gone home.’ b. Ja xoču, čtoby vy I want
SUBJ
osta-l-is’.
you-PL stay-PAST-PL
‘I want you to stay.’
In the subjunctive mood, the morphological past tense verb form lacks a temporal interpretation. As I discuss in detail in chapter 4, matrix subjunctive clauses such as in (18a) can refer to a hypothetical event of leaving, which could have occurred either prior to the moment of speech, or at the moment of speech. Embedded subjunctive clauses such as in (18b) typically receive a hypothetical future interpretation relative to the matrix event. The contrast between the temporal interpretations of the morphological past tense forms is shown in (19). The same form has the past tense interpretation in the indicative sentence in (19a) and the hypothetical future interpretation in the subjunctive sentence in
37 (19b). This is illustrated by (in)compatibility of the clauses with different time adverbials. In (19a), only the past tense adverbial včera ‘yesterday’ is acceptable. In contrast, the sentence in (19b) is incompatible with včera ‘yesterday’; however, it is acceptable with either present or future time adverbials, sejčas ‘now’ or zavtra ‘tomorrow’:
(19) a. Ja uš-l-a I
domoj včera/*sejčas /*zavtra.
leave-PAST-FEM.SG home yesterday / now / tomorrow
‘I would go / would have gone home yesterday / *now / *tomorrow.’ b. Ja xoču, čtoby ty I want
SUBJ
uš-l-a
domoj *včera/ sejčas /zavtra.
you leave-PAST-FEM.SG home yesterday / now / tomorrow
‘I would go / would have gone home *yesterday / now / tomorrow.’
It is very common crosslinguistically for irrealis mood to be expressed by the past morphology (James 1982; Givón 1994; Dahl 1997; Tynan and Lavín 1997; Iatridou 2000). However, there exist different views as to whether the subjunctive and the indicative mood in Russian contain the same verbal form and differ only in the presence or absence of the conditional particle and subjunctive complementizer. Traditional grammars, such as (Švedova 1980, vol. 1:625), treat them as distinct forms, which happen to be identical to the past tense form in indicative clauses. In contrast, according to Hacking (1998) and Spencer (2001) among others, this is the same form, whose interpretation differs depending on the syntactic environment. For example, Spencer (2001) argues that the morphological past tense verb form (or l-participle, as discussed
38 immediately below) must be analyzed as a pure, or tenseless form, which, along with the particle by forms the subjunctive mood, and which otherwise receives past as its default tense interpretation. Following Spencer (2001), I argue in chapter 4 that the morphological past tense form is used in both the indicative and the subjunctive mood, and that the precise interpretation is determined by the syntactic environment. In the next section, I discuss the historical origins of the morphological past tense form and show that the historical facts support the unified treatment of the form in the indicative and the subjunctive mood.
2.2.3. The Historical Development of Russian Past Tense Verb Forms Historically, Old Russian distinguished the following four past tenses: the aorist, the imperfect, the perfect, and the pluperfect, all of which were subsequently lost. The aorist and the imperfect were simple (that is, non-periphrastic) tenses formed by means of inflectional suffixes. Both simple tenses could occur in either perfective or imperfective aspect and described a single action in the past, which had no relevance for the present. The aorist, or narrative past, was used to describe sequences of events. The imperfect, or descriptive past, was used to describe durative actions viewed as taking place in more than one step. The following paradigm of the verb byti ‘be’ illustrates this point, with the aorist and the imperfect suffixes in bold (Ivanov 1964:375ff):
39 (20)
Singular
Plural
Dual
Aorist
Imperfect
Aorist
Imperfect
Aorist
Imperfect
1st
byxŭ
bäxŭ
byxomŭ
běäxomŭ
byxově
běaxově
2nd
by
běäše
byste
běaste
bysta
běasta
byša
bäxu(tǐ)
3rd
The perfect and the pluperfect were complex past tenses, formed by the conjugated forms of the auxiliary byti ‘be’ and the l-participle, a deverbal adjective created by adding the suffix –l to the stem. Both complex tenses could occur in either perfective or imperfective aspect. The perfect had a resultative meaning; it expressed a condition at present, which resulted from some action in the past (Ivanov 1964:377). It was essentially parallel to the English present perfect, as in John has left, where neither the auxiliary has, nor the participle left conveys past tense. In the perfect, the verb byti occurred in the present tense (imperfective aspect), the form absent from Modern Russian, and the l-participle appeared in either perfective or imperfective aspect. Below I provide a perfect paradigm for the Old Russian verb nesti ‘carry(IMPF)’ (Ivanov 1964:375ff):
(21)
Singular (m./f./n.)
Plural (m./f./n.)
Dual (m.-n./f.)
1st
jesmǐ
nesŭ-l-ŭ/a/o
jesmŭ nesŭ-l-i/y/a
jesvě nesŭ-l-ě/a
2nd
jesi
nesŭ-l-ŭ/a/o
jeste
nesŭ-l-i/y/a
jesta
3rd
jestǐ
nesŭ-l-ŭ/a/o
sutǐ
nesŭ-l-i/y/a
nesŭ-l-ě/a
40 The pluperfect was a “relative” tense used to describe an action in the past that preceded some other action in the past. It may be compared with the English past perfect, as in John had left. In the pluperfect, the auxiliary byti ‘be’ itself could appear in three different tense forms. It could appear in simple past tenses, aorist and imperfect; it could also appear in perfect, that is, the auxiliary itself could have a complex form. Thus, the resulting tense form would consist of the auxiliary in the present tense (imperfective aspect) of byti followed by two l-participles: that of byti and the main verb.10 Below I give the pluperfect paradigm for the verb nesti ‘carry(IMPF)’, with the auxiliary appearing in the perfect (Ivanov 1964:375ff). For ease of exposition, the paradigm below contains only the masculine form:
(22)
Singular
Plural
Dual
1st
jesmǐ by-l-ŭ nesŭ-l-ŭ
jesmě by-li nesŭ-l-i
jesvě by-l-ě nesě-l-ě
2nd
jesi
jeste
by-li nesŭ-l-i
jesta by-l-ě nesě-l-ě
3rd
jestǐ by-l-ŭ nesŭ-l-ŭ
sutǐ
by-li nesŭ-l-i
by-l-ŭ nesŭ-l-ŭ
Note that the l-participle conveyed no tense, but had perfective or resultative meaning. This claim is supported by the fact that, the subjunctive/conditional mood in Old Russian was also formed with the l-participle, just as in Modern Russian; only in this case the l-participle combined with the aorist form of the auxiliary byti ‘to be’.11 The
10
Old Russian had another way of forming the pluperfect, namely, by the imperfect of byti ‘be’ and the lparticiple. Ivanov (1964:377) states that this version of the pluperfect disappeared from Old Russian much earlier than the one given in (22). 11
The l-participle was also used to express future perfective tense which does not exist in modern Russian.
41 conditional particle by has developed out of the second and third person singular form by. This was the only form in the paradigm which did not contain any inflectional marking, as illustrated by the contrast in (23):12
(23) a. Radi sja happy REFL
[ by-x-omŭ
ja-l-i]
po dan’.
be-AOR-1ST.PL agree-LPART-PL to tribute
‘We would happily agree to tribute.’ b. Poslisja
k bratu… da [by
ti
pomog-l-ŭ].
send-IMPER to brother that be-AOR.2ND/3RD.SG you help-LPART-SG ‘Send to you brother… that he should help you.’
(Willis 2000:339)
The degeneration of the old tense system was triggered by the development of aspect, and by the seventeenth century the system of aspect had already replaced the system of tense (Matthews 1960:209). The aorist and the imperfect vanished without a trace. The complex past tenses have partially survived. What is left of the perfect (or the pluperfect, for that matter) is the l-participle. We have seen that in Modern Russian, it took over the function of the past tense. It is usually analyzed as the simple past tense form, with the suffix –l reanalyzed as a tense morpheme (traditional grammar: Švedova 1980, vol. 1:626–636, 664–675; generative analyses: Franks and Greenberg 1988, 1994). Having taken over the function of specifying the past tense, past tense forms still retain
12
According to Spencer (2001), in Slavic languages other than Russian (Bulgarian, Serbo-Croatian, Czech) this auxiliary is still used to form subjunctives and conditionals.
42 their agreement in number and gender: Singular forms are masculine, feminine, or neuter, whereas plural forms do not show gender agreement. As Borik (2002:149) points out, the development of the Russian past tense is consistent with the general scheme of the diachronic development of past tense crosslinguistically:
(24)
RESULTATIVE > PERFECT > PERFECTIVE > PAST
Thus, as shown in (25a), in Old Russian the interpretation of constructions with the l-participle was determined by the auxiliary: The present and the perfective form gave rise to the past tense interpretation, whereas the aorist form gave rise to the subjunctive interpretation. As shown in (25b), in Modern Russian the auxiliary has disappeared in the indicative mood, and the l-participle took over the function of expressing the past temporal reference. In the subjunctive mood, the auxiliary has been reduced to an uninflected particle.
(25) a. Old Russian:
l-participle
Aux1: the present/perfect form of byti
Aux2: the aorist form of byti
Tense: Perfect, Pluperfect
Mood: Subjunctive
43 b. Modern Russian:
l-participle
Aux1: the present/perfect form of byti
Aux2: by
Tense: Past
Mood: Subjunctive
In addition to changing the morphological form of constructions with the lparticiple, it has been suggested that the loss of the auxiliaries has triggered a change in the range of interpretations available for the past tense form (Borik 2002; Borik, González and Verkuyl 2003). This change is best illustrated by applying Reichenbach’s (1947) tense model to Russian. Reichenbach (1947) proposes a three-way distinction between the Speech time S, the Reference time R, and the Event time E, defined as follows:
(26) a. S:
The time when the sentence is uttered.
b. R:
The time with respect to which an assertion is made.
c. E:
The time at which the event occurs.
Two temporal relations, precedence and coincidence, may be defined between these three elements. Following Klein (1994), the category tense expresses a temporal relation between S and R (Klein’s utterance time TU and topic time TT), and grammatical aspect expresses the relation between R and E (Klein’s situation time T-SIT). In English, the
44 notion R is expressed via distinct forms of the auxiliary and reflects the contrast between simple present, as in (27a), where the auxiliary is absent, present perfect, as in (27b), and past perfect, as in (27c):
(27) a. Philemon saw Lee.
E, R ___ S
b. Philemon has seen Lee.
E ___ R, S
c. (When I returned home) Philemon had seen Lee.
E ___ R ___ S
The three sentences are represented in the same way as far as the ordering of E and S is concerned, namely, E ___ S. They differ only in the Reference time: In (27a), it coincides with E, in (27b) – with S, and in (27c) it coincides with neither. Schoorlemmer (1995:240ff) observes that as far as the Reference time R is concerned, Russian differs from English. Both the perfective sentence in (28a) and the imperfective sentence in (28b) can have any of the three interpretations: (28A) is temporally comparable to the English simple past;13 (28B) is temporally comparable to the English perfect, such as Philemon has seen Lee; (28C) is temporally comparable to the English pluperfect, such as Philemon had seen Lee:
13
Schoorlemmer (1995:240) claims that this interpretation is unavailable. However, a few pages later she contradicts herself saying that the most obvious aorist context for past perfective sentences is the description of a sequence of events (251). Moreover, according to my own judgments, Russian perfective and imperfective past forms both can have an interpretation comparable to the English aorist.
45 (28) a. Philemon u-vide-l
Lee.
Philemon
Lee
PREF-see-PAST-3RD.SG(PERF)
‘Philemon saw (noticed) / has seen / had seen Lee.’ b. Philemon vide-l
A. E, R ___ S
B. E ___ R, S
Lee.
Philemon see-PAST-3RD.SG(IMPF) Lee
C. E ___ R ___ S
‘Philemon saw / has seen / had seen Lee.’
Schoorlemmer (1995:241) claims that Russian sentences such as those in (28) are nonspecific with respect to the interpretation of R. Borik, González and Verkuyl (2003:26) conclude that in the Russian tense system, the loss of the auxiliaries resulted in the loss of the bridge between the moment of speech and the event time. This means that the tense system is free to choose the initial point of reference, without expressing it grammatically. The contrast between English and Russian with respect to the expression of the Reference time R suggests that the grammatical aspects in the two languages are not parallel. For example, the English perfect is not equivalent to the Russian perfective. In chapter 3, I outline Klein’s (1995) theory of temporal relations developed specifically for Russian. It captures the properties of the Russian aspect, by replacing the Reference time R with the Assertion time T-AST. I also modify Demirdache and Uribe-Etxebarria’s (2000) theory, in order to make it suitable for describing the Russian tense and aspectual systems.
46 2.3. NON-PAST TENSE VERB FORMS IN RUSSIAN In Modern Russian, there are two simple non-past tense forms, present and future, and one complex, or periphrastic, non-past tense form, future imperfective. The latter consists of the auxiliary byt’ ‘be’ in the perfective aspect (or future tense, as discussed in section 2.3.2) and an imperfective infinitive. The non-past verb forms agree with the nominative subject in person and number. Unlike the past tense forms, the non-past tense forms contain no tense morphology. In simple non-past tense, the temporal distinction between present and future is conveyed by the perfective/imperfective opposition. In 2.3.1, I discuss the simple non-past tense forms, their morphological and semantic properties. In 2.3.2, I introduce the periphrastic future tense. In 2.3.3, I provide evidence that future in Russian is tense and not mood.
2.3.1. Simple Non-Past Forms In (29), the non-past forms of the verb čitat’ ‘read’ consist of the stem čitaj-, the agreement suffixes and the aspectual morphology. The imperfective forms are unmarked for aspect, whereas the perfective forms contain the perfectivizing prefix pro-. There is no morphology that indicates tense:
47 (29)
Singular Imperfective
1st 2nd 3rd
Perfective
Plural Imperfective
Perfective
čitaj-u
pro-čitaj-u
čitaj-em
pro-čitaj-em
read-1ST.SG
PREF-read-1ST.SG
read-1ST.PL
PREF-read-1ST.PL
čitaj-eš
pro-čitaj-eš
čitaj-ete
pro-čitaj-ete
read-2ND.SG
PREF-read-2ND.SG
read-2ND.PL
PREF-read-2ND.PL
čitaj-et
pro-čitaj-et
čitaj-ut
pro-čitaj-ut
read-3RD.SG
PREF-read-3RD.SG
read-3RD.PL
PREF-read-3RD.PL
Despite the absence of tense morphology, Russian clauses that contain a non-past tense form always have a temporal interpretation, which is determined by aspect: The imperfective and perfective verbs receive present and future interpretation, respectively. Let us first examine the non-past imperfective forms. Just as with the past tense forms discussed above, the optionality of temporal adverbials suggests that the temporal interpretation is inherent to the predicate; it is not supplied by the adverbials and does not arise from the pragmatic context. As illustrated in (30), morphologically imperfective verbs receive the present tense interpretation. Only the adverbial sejčas ‘now’ is acceptable in (30a): the event of Liza reading a book must overlap with the speech time. In (30b), the adverbial zavtra ‘tomorrow’ is acceptable. This is because the event of flying to Paris is most likely planned or scheduled, and the present tense meaning is shifted to refer to an event in the immediate future:
48 (30) IMPERFECTIVE ASPECT --> PRESENT TENSE a. Liza čitaj-et
knigu (sejčas/*včera/*zavtra).
Liza read-3RD.SG(IMPF) book
now/ yesterday/tomorrow
‘Liza is reading a/the book (now/*yesterday/*tomorrow).’ b. Philemon uletaj-et
v Pariž (sejčas/*včera/zavtra).
Philemon fly-3RD.SG(IMPF) to Paris now / yesterday / tomorrow. ‘Philemon is leaving for Paris now / *yesterday / tomorrow.
Now consider the non-past perfective forms. The perfective in the past and nonpast is formed in the same way, that is, by affixation or, rarely, by suppletion. Moreover, the same affix is used to create the perfective form of a given verb in both, past and nonpast. For example, in (31), in both tenses the perfective form of the verb čitat’ ‘read’ is formed by the addition of prefix pro-:
(31) a. Past (Perfective):
Liza pro- čitala
knigu.
Liza
book
PREF-read(PERF)
‘Liza has read a/the book.’ b. Non-past (Future):
Liza pro-čitaet Liza
PREF-
knigu.
read(PERF) book
‘Liza will read a/the book.’
49 However, the perfectivizing prefix in (31a) and (31b) has a different effect on the interpretation of the clause. When the perfectivizing prefix is added to the past tense verb form in (31a) it has no effect on the temporal reference of the clause. It only specifies the perfective aspect. In contrast, when the same prefix is added to the non-past tense verb form in (31b) it affects both the aspectual properties and the temporal reference of the clause. Unlike its imperfective counterpart, the perfective non-past verb form denotes an event that takes place after the speech time. Again, the use of time adverbials is optional, which indicates that the temporal interpretation is inherent to the predicate. When a time adverbial is used it has to match the temporal reference of the sentence:
(32) PERFECTIVE ASPECT --> FUTURE TENSE a. Liza pro-čita-et Liza
PREF-
knigu (*sejčas/*včera/ zavtra).
read-3RD.SG.(PERF) book
now/ yesterday/tomorrow
‘Liza will read a/the book (*now/*yesterday/tomorrow).’ b. Philemon u-let-it
v Pariž (*sejčas/*včera/zavtra).
Philemon PREF-fly-3RD.SG(PERF) to Paris
now / yesterday / tomorrow.
‘Philemon will leave for Paris *now / *yesterday / tomorrow.
According
to
Bybee,
Perkins,
and
Pagliuca
(1994:278),
perfectives
crosslinguistically tend to be restricted to the past tense, and future uses of the perfective aspect are relatively rare. As I argue in chapter 3, since in Russian verbal stems are specified as non-past, the future interpretation does not arise from the perfective aspect
50 alone, but in combination with the non-past interpretation of the stem. This issue is discussed in more detail in chapter 3. It should be mentioned that there are contexts in which the perfective non-past forms receive a present interpretation, such as habitual/iterative contexts, as in (33a), or generic statements of general truth, as in (33b). In the latter, the perfective and the imperfective non-past forms can be used interchangeably:
(33) a. Petja tebe
vsegda pravdu skaž-et.
Petja you-DAT always truth
tell-3RD.SG(PERF)
‘Petja will always tell you the truth.’ b. Eto takaja gazeta,
v kotoroj vse čto
xočeš’ napiš-ut/piš-ut.
this such newspaper in which all what want
write-3RD.PL(PERF)/(IMPF)
‘This is such a newspaper where anything can get published.’
(Borik 2002:141)
Borik (2002:139) claims that the interpretation of the non-past forms does not provide solid grounds for distinguishing between a present and a future in the Russian temporal system. I argue, however, that the existence of examples such as in (33) does not undermine the claim that Russian distinguishes between present and future. First, the use of perfective non-past forms with a present interpretation is restricted to generic and habitual contexts. Sentences such as (33) are lacking a particular temporal interpretation, as they do not pick out a specific event which is ordered relative to the moment of
51 speech. The perfective non-past forms can never receive an actual present interpretation. As Forsyth (1970:148) states, “perfective present in the ‘real’ mode have almost exclusively future meaning”. Thus, examples such as in (33) above can never refer to an event overlapping with the speech time. Second, the sentence in (33a) can quantify over events in past as well as non-past. For example, even if Peter is no longer alive (33a) can still be true. Thus, it is possible that the temporal interpretation of (33a) involves some additional mechanisms unique to the temporal interpretation of such constructions. I conclude therefore that Russian does make a distinction between present and future, and that this distinction is conveyed by aspect.
2.3.2. Complex Tense: Future Imperfective The future imperfective is a complex form that consists of the auxiliary byt’ ‘to be’ and an imperfective infinitive. The auxiliary shows the non-past agreement, that is, it agrees with the nominative subject in person and number. The future imperfective paradigm of the verb čitat’ ‘read’ is exemplified in (34):14
14
i.
The perfective aspect does not allow the periphrastic form: *budet pro-čitat’ be-3RD.SG PREF-read-INF(PERF)
52 (34)
SINGULAR
Ja bud-u
čitat’.
PLURAL
My bud-em
čitat’.
I be-1ST.SG read-INF(IMPF)
we be-1ST.PL read-INF(IMPF)
‘I will be reading.’
‘We will be reading.’
Ty bud-eš
Vy bud-ete
čitat’.
čitat’.
you be-2ND.SG read-INF(IMPF)
you be-2ND.PL read-INF(IMPF)
‘You will be reading.’
‘You will be reading.’
Ona/on bud-et
Oni bud-ut
čitat’.
čitat’.
be-3RD.SG read-INF(IMPF)
they be-3RD.PL read-INF(IMPF)
‘She/he will be reading.’
‘They will be reading.’
The complex future imperfective form is morphologically marked for both tense and aspect. The future interpretation of the constructions in (31) is determined by the auxiliary byt’ ’be’. The agreement in person and number suggests that the auxiliary is in the non-past. As discussed in the previous section, the non-past tense verb forms of lexical verbs receive the future tense interpretation in the perfective. In Modern Russian, the aspectual value of byt’ is hard to establish. According to Jakobson (1957) and Franks and Greenberg (1994), it does not distinguish aspect. However, there is evidence that historically, this form used to be aspectually perfective (van Schooneveld 1951; Jakobson 1957; Junghanns 1997). In Old Russian, the verb byti ‘be’ had two complete paradigms in the imperfective and perfective, just like any other verb. In Modern Russian, only the perfective paradigm has survived:
53 (35)
Non-past paradigm of Old Russian byti ‘be’ Imperfective
Perfective
Singular
Plural
Dual
Singular
Plural
Dual
1st
jesmĭ
jesmĭ
jesvě
budu
budemŭ
budevě
2nd
jesi
jeste
jesta
budeši
budete
budeta
3rd
jestě
sutĭ
jeste
budetŭ
budutŭ
budete
(36)
Non-past paradigm of Modern Russian byt’ ‘be’ Imperfective Singular
Singular
Plural
1st
budu
budem
2nd
budeš’
budete
budet
budut
3rd
jest’
Plural
Perfective
(sut’)
Dual
Dual
As the data in (35)–(36) demonstrate, the imperfective of byti/byt’ did not survive in Modern Russian. The form jest’ is used in Modern Russian as a main verb as well as an auxiliary. The form sut’ is given in parentheses, as it appears mostly in literary language and its use is highly restricted (Ivanov 1964:367–369). Both imperfective forms are morphologically marked for 3rd person singular. However, they both can be used for all three persons, either singular or plural. The form jest’ is used to express possession or location, as in (37a,b), or as an emphatic element, as in (37c):
54 (37) a. V Moskve in Moscow
jest’
tramvai.
be-3RD.SG(IMPF) trams
‘There are trams in Moscow.’ b. U nas jest’
cvetnoj televizor.
at us be-3RD.SG(IMPF) color
TV
‘We have a color TV.’ c. Eto i
jest’
pravda.
this and be-3RD.SG(IMPF) truth ‘This really is the truth.’
(Junghanns 1997:251)
The constructions with sut’ as in (38) are stylistically marked:
(38) a. Povesti sut’
literaturnyje proizvedenija.
stories be-3RD.PL(IMPF) literary
works
‘Stories are (in fact) works of literature.’ b. Slon
sut’
(Junghanns 1997:252)
mlekopitajuščee.
elephant be-3RD.PL(IMPF) mammal ‘Elephant is (in fact) a mammal.’
(Borik 2002:152)
It is clear from the tables in (35)–(38) that the auxiliary budet in the future imperfective in Modern Russian has developed out of the perfective form of byti. In addition to the future imperfective, it is used freely, with no stylistic effect, as a copula
55 verb and as an auxiliary in the participle passive, as shown in (39a) and (39b), respectively:
(39) a. Liza budet
vračom / krasivoj / v sadu.
Liza be-3RD.SG(PERF) doctor / pretty / in garden ‘Liza will be a doctor / pretty / in the garden.’ b. Deti
budut
nakormleny (njanjej).
children be-3RD.PL(PERF) fed
nanny-INSTR
‘The children will be fed (by the nanny).’
Just as in future imperfective constructions, the perfective forms of byt’ trigger the future interpretation of the sentences in (39).15 The tests proposed by Smith and Rappaport (1991) reveal a clear contrast between the perfective and the imperfective future tenses. This contrast suggests that these tense forms also have the aspectual meaning. In the future perfective (40a), only the sequential interpretation of the two events is possible: The snowstorm will begin after the plane 15
Borik (2002:153) points out that it is difficult to establish the aspectual value of byt’ in Modern Russian because the infinitival form patterns with imperfective verbs. For example, unlike other perfective verbs it can appear as a complement to phase verbs, such as prodolžat’ ‘continue’: i. a. Prodolžaj byt’ poslušnym. continue be-INF(PERF) obedient ‘Continue to be obedient!’ b. Prodolžaj čitat’ / *pročitat’ knigu. continue read-INF(IMPF)/ read-INF(PERF) book ‘Continue to read a book!’ Such data suggest that in Modern Russian, byt’ is unmarked for aspect. However, what matters for my purposes is that in the non-past tense, the interpretation of the stem bud- is always future, which betrays its perfective status.
56 takes off. In the future imperfective (40b), the two events overlap: The snowstorm will begin while the plane is taking off:
(40) a Future Perfective (simple tense) Kogda samolet vzletit, when plane
načnetsja
snežnaja burja.
take.off(PERF) begin(PERF) snow
storm
‘When the plane takes off, the snowstorm will begin.’ b. Future Imperfective (complex tense) Kogda samolet budet when plane
vzletat’,
načnetsja
snežnaja burja.
be(PERF) took.off-INF began(PERF) snow
storm
‘While the plane will be taking off, the snowstorm will begin.’
The example in (36b) shows that the aspectual morphology performs a double function: It conveys both aspectual meaning and temporal reference.
2.3.3. Future is Tense Throughout the chapter, I have been treating the Russian tense system as consisting of three tenses, present, past, and future. However, tense is often assumed to be past and present, while future belongs to the domain of mood. Such an approach is characteristic of semantic accounts, such as Palmer (1986), Kamp and Reyle (1993), Enҫ (1996), Iatridou (2000), among others. It is important to emphasize that here I am concerned with
57 the syntactic representation of the tense system. I am going to justify my treatment of future as tense from the syntactic point of view. First, there is no reason to believe that Russian clauses with future tense interpretation differ in their structure from present and past tense clauses. Although the future tense in Russian may be argued to have a modal flavour, this does not seem to be reflected anywhere in the morphosyntax. Following Schoorlemmer (1995:161), Russian finite tensed clauses show a uniform behavior with respect to various syntactic phenomena, such as word order, verb movement, the availability of null subjects, etc. In particular, there are no empirical data to suggest that future clauses contain an additional functional projection above TP. Second, finite embedded clauses in Russian always have an independent temporal interpretation.16 In particular, a future embedded clause can refer to an event in the future relative to the moment of speech. Consider first the matrix clauses in (41). The future perfective (41a) and the future imperfective (41b) both refer to an event in the future following the moment of speech. With respect to (41b), if this sentence is uttered on Monday, the event of working from nine to five is supposed to take place on Tuesday:
(41) a. Philemon pri-nes-et Philemon
PREF-bring-3RD.SG(PERF)
knigi potom. books later
‘Philemon will bring the books later.’
16
Russian does not have sequence of tense. This phenomenon is discussed in chapter 5.
58 b. Zavtra
my budem
rabotat’
s
devjati do pjati.
tomorrow we be-2ND.PL work-INF(IMPF) from nine
to five
‘Tomorrow we will be working from 9 to 5.’
When these sentences are embedded under another verb, as in (42), they still can refer to a event that takes place following the moment of speech, and not the matrix event:
(42) a. Feliks skazal, [ čto Philemon pri-nes-et Felix said
IND
Philemon
knigi potom].
PREF-bring-3RD.SG(PERF)
books later
‘Felix said that Philemon will bring the books later.’ b. Liza dumala, [čto zavtra Liza thought
IND
my budem
rabotat’
s
devjati do pjati.
tomorrow we be-2ND.PL work-INF(IMPF) from nine to five
‘Liza thought that tomorrow we will be working from 9 to 5.’
In both sentences, the future tense of the embedded clauses places the embedded event relative to the moment of speech, and not to the matrix event. If tense, but not mood, can locate an event relative to the moment of speech, the embedded clauses in (42) have independent tense. Moreover, this tense is future. In light of these considerations, I treat future as tense for the purposes of syntax, and the Russian tense system as a three-way distinction between past, present, and future.
59 2.4. CONCLUSION In this chapter, I discussed the properties of Russian tense system. It has been shown that the past and the non-past tense forms contrast with respect to the inflectional morphology and their relationship with aspect. The past tense forms have developed historically from the complex past tenses which were formed by an auxiliary and the l-participle. The lparticiple is the deverbal adjectives formed by adding the suffix –l to the stem. Following the loss of the auxiliaries, the l-participle took over the function of conveying the past tense meaning, and the suffix –l has been reanalyzed as a past tense morpheme. The non-past tense forms are morphologically marked for aspect, but not for tense. The distinction between present and future is conveyed by aspect. Imperfective aspect conveys present tense, whereas perfective aspect conveys future tense. Thus, in the non-past, the aspectual morphology performs a double duty of marking both aspect and tense, whereas in the past, it only marks aspect. In the next chapter, I develop a theory that captures the relationship between tense and aspect in Russian. I assume that tense and aspect each head a projection in syntax, TP and AspP, respectively. I assume also that the temporal and aspectual properties can be represented as formal features. These formal features are carried by lexical items in the form of overt morphology. I treat tense and aspect as properties of clauses, and not of lexical items. Therefore, although verbs are morphologically marked for tense and aspect in the lexicon, these features can only be interpretable once a verb is inserted into a syntactic structure.
60 I propose that the heads T and Asp each contain an interpretable feature that is responsible for the temporal and aspectual interpretation of a clause. At this point, let me call this feature [±X]. This feature receives its value from overt verbal morphology via the syntactic operation of agreement (Pesetsky and Torrego 2004b). In the past, the two features on T and Asp receive their value from the suffix –l and the aspectual morphology, respectively. However, in non-past, there is only aspectual morphology, which supplies the value of [±X] on Asp. Since no features in the derivation may remain unvalued, [±X] on T receives its value via agreement with the valued feature [±X] on Asp.
61 CHAPTER 3 FEATURING RUSSIAN TENSE
3.0. INTRODUCTION This chapter focuses on the temporal interpretation of Russian finite indicative clauses.1 In chapter 2, I demonstrated that the main opposition in Russian is between the past and the non-past, and that the temporal interpretation of these forms arises via different mechanisms. The past tense verb forms are morphologically marked for both tense and aspect, and the aspectual interpretation does not influence the temporal interpretation. The non-past tense verb forms are morphologically marked only for aspect, which performs a double function of determining both the aspectual and the temporal reference. The question that I am going to address here is how this relationship between tense and aspect is represented in syntax. Let me briefly review the main data from chapter 2. The non-past tense forms in Russian have no tense morphology and the temporal interpretation of non-past clauses is supplied by aspect: Imperfective verbs receive present tense interpretation; that is, they denote an event that coincides with the utterance time. Perfective verbs receive future tense interpretation; that is, they denote an event that follows the utterance time. The verbs in (1) differ only in aspect: In (1a), the morphologically simple verb is imperfective, whereas in (1b) the verb contains a
1
I deal exclusively with matrix clauses; embedded indicative clauses are dealt with (not very extensively) in chapter 4.
62 perfectivizing prefix. Accordingly, (1a) and (1b) have a present and a future meaning, respectively:
(1)
a. Philemon čitaj-et
knigu.
Philemon read-3RD.SG(IMPF) book ‘Philemon is reading a book.’ b. Philemon pro-čitaj-et
knigu.
Philemon PREF-read-3RD.SG(PERF) book ‘Philemon will read a book.’
In the past, aspect does not determine the temporal reference of a clause. Past tense forms can be imperfective and perfective, as in (2a) and (2b), respectively. Note that the perfective forms in both the non-past (1b) and the past (2b) are formed by the same perfectivizing prefix pro-. However, in the past, aspect does not affect the temporal interpretation of the clause: In either aspect, the past tense form denotes an event that precedes the utterance time:
(2)
a. Philemon čita-l
knigu.
Philemon read-PAST(IMPF) book ‘Philemon was reading / read a/the book.’
63 b. Philemon pro-čita-l
knigu.
Philemon PREF-read-PAST(PERF) book ‘Philemon read / has read a/the book.’
Thus, Russian tense verb forms may be divided into two groups based on the role of aspect in determining the temporal reference. In the non-past, aspect determines the temporal reference of a clause (present or future). In the past, aspect does not play a role in determining the temporal reference:
(3)
Aspect and Temporal Reference in Russian a. Present and Future: Temporal Reference is determined by Aspect b. Past: Temporal Reference is independent of Aspect
The goal of this chapter is to develop a morphosyntactic account of the Russian tense system that would capture the contrast between the past and the non-past in (3). The analysis proposed here relies on the following assumptions. First, tense and aspect are dyadic predicates that order temporal arguments in syntax (Zagona 1990, 1995, 2003; Stowell 1995, 1996; Demirdache and Uribe-Etxebarria 2000). Second, tense and aspect are represented as formal features on the functional heads T and Asp, respectively. These formal features can be interpretable or uninterpretable and valued or unvalued, where the value of a feature comes from overt morphology (Pesetsky and Torrego
64 2004a,b). Finally, following Pesetsky and Torrego (2004b), I assume that an unvalued feature can receive its value from a valued feature via a syntactic operation of agreement. I propose that, under certain circumstances to be defined below, the same morpheme can be used more than once to supply the value of more than one feature on different syntactic heads. I refer to this as “value recycling”. I propose that the present and future tense interpretation of Russian clauses is due to the fact that the aspectual morphology is used twice: It is first used to supply the value of the “aspect” features on Asp. It is then recycled to supply the value of the “tense” features on T. In contrast, in past the value of the relevant aspect and tense features is supplied by different morphemes, and no value recycling takes place. This chapter is organized as follows. In 3.1, I provide the theoretical framework and outline the main assumptions on which my analysis is based. These assumptions will also be relevant in chapter 4. Section 3.2 deals with the past tense interpretation of finite indicative clauses. In section 3.3, I develop an account of the present and future tense interpretation in terms of the value recycling mechanism. In section 3.4, I introduce the periphrastic future tense, or future imperfective, and show that the analysis proposed in this chapter extends to this tense as well. Finally, section 3.5 is a conclusion.
3.1. THE COMPONENTS OF TIME 3.1.1. Klein (1995) As the basis for my analysis of the Russian tense system I adopt Klein’s (1995) timerelational analysis of tense and aspect in Russian. Klein argues that in Russian, tense and
65 aspect are both time-related notions. His theory makes use of three temporal parameters: the utterance time TU, the situation time T-SIT, and the assertion time T-AST. Let us first consider the first two notions, TU and T-SIT. TU is the speech time, or the time when the utterance is made. T-SIT is the time when the situation referred to by the utterance obtains. Klein (1994:21) points out that “the standard theory of tense” involves the temporal parameters of the utterance time and the situation time, with the tense forms expressing three temporal relations between the two: before, simultaneously, after:2
(4)
a. past tense verb form
T-SIT before TU
b. present tense verb form
T-SIT simultaneous to TU
c. future tense verb form
T-SIT after TU
The third temporal component, the assertion time T-AST, is unique to Klein’s theory of Russian. T-AST mediates the relation between TU and T-SIT.3 The assertion 2
The utterance time TU and the situation time T-SIT may be viewed as equivalent to Reichenbach’s (1947) Speech time S and the Event time E, respectively. The relation between the two notions is defined as precedence and simultaneity. This relation determines three basic temporal relations in natural languages: i. a. Philemon is eating fish. b. Philemon ate fish. c. Philemon will eat fish.
S, E (present) E ___ S (past) S ___ E (future)
3
The assertion time T-AST is not comparable to Reichenbach’s Reference time R. First, contra Demirdache and Uribe-Etxebarria (2000:162), R does not always mediate the relation between the Speech time S and the Event time E. For example, in the future perfect R follows both S and E: i. Mary will have already finished her thesis (by the time she gets the job).
S__E__R
The event of Mary finishing her thesis is located after the speech time (future tense), but before the event of getting the job. Second, the Reference time R is not grammatically expressed in Russian. For example, the past tense sentence in (ii) has three possible interpretations, which differ only in the positioning of R: ii. Philemon u-videl Lee. Philemon PREF-see- PAST(PERF) Lee ‘Philemon saw / has seen / had seen Lee.’
a. E, R ___ S b. E ___ R, S c. E ___ R ___ S
66 time refers to the time for which a particular assertion is made in a given utterance, and captures the perfective/imperfective opposition in Russian. The following considerations motivate the introduction of assertion time. Consider the examples below (Klein 1995:686):
(5)
a. Ivan rabota-l
v Moskve.
(past)
Ivan work-PAST(IMPF) in Moscow ‘Ivan was working/worked in Moscow.’ b. Ivan rabotaj-et
v Moskve.
(present)
Ivan work-3RD.SG(IMPF) in Moscow ‘Ivan is working/works in Moscow.’ c. Ivan bud-et
rabotat’
v Moskve
(future)
Ivan be-3RD.SG work-INF(IMPF) in Moscow ‘Ivan will be working in Moscow.’
All three sentences in (5) have the same propositional content: . Moreover, all verbs are imperfective. The sentences in (5) only differ in their temporal reference, or the time when the situation obtains. (5a) denotes a situation that precedes the utterance time, (5b) denotes a situation that includes the utterance time, and (5c) denotes a situation that follows the utterance time. It is reasonable to assume that the
Klein (1994) proposes the term Topic time TT. My understanding is that this notion is comparable to Reichenbach’s (1947) Reference time R.
67 three distinct verb forms indicate how the situation time T-SIT is related to the utterance time TU, as shown in (4) above. However, Klein points out that there is a problem with the representation in (4). For example, (5a) is appropriate when Ivan is still working in Moscow, that is, if T-SIT includes, or is simultaneous with TU, rather than precedes it. But if it precedes it, it cannot include it or be simultaneous with it. Similarly, (5c) is not false if Ivan is already working in Moscow at the time of utterance, that is, if T-SIT includes TU rather than follows it. Thus, in Russian the following can be uttered without any contradiction:
(6)
Ivan rabota-l,
rabotaj-et
i
bud-et
rabotat’
v Moskve.
Ivan work-PAST(IMPF) work-3RD.SG(IMPF) and be-3 RD.SG work-INF(IMPF) in Moscow
‘Ivan worked, is still working, and will be working in Moscow.’
If, for example, the past tense verb form indeed expressed that the event precedes the moment of speech, that is, T-SIT is before TU, then it cannot be simultaneous with TU, and the sentences in (6) should be pragmatically odd. Klein (1995) argues that “a distinction should be made between WHICH THE SITUATION DESCRIBED BY AN UTTERANCE OBTAINS,
THE TIME AT
on the one hand, and
TIME FOR WHICH A PARTICULAR ASSERTION IS MADE IN THIS VERY UTTERANCE,
THE
on the
other. Aspect is a temporal relation between these two time spans” (p. 680). Crucially, the assertion time is related to the situation time. Under Klein’s approach, Russian tensed verb forms in (6) do not express the relation between the TU and T-SIT directly. Instead,
68 they express that some subinterval of T-SIT is before, simultaneous with, or after TU. The speaker makes the assertion only about this subinterval. For example, the past tense verb form expresses that some subinterval of T-SIT is before TU. Whether or not the rest of T-SIT is before TU is irrelevant. The same is true for the future tense form. The assertion is made only for some subinterval of T-SIT, which is after TU. Thus, the distinction must be made between the time of the situation T-SIT and the time (or the subinterval of the situation) for which an assertion is made T-AST. Klein (1995) defines the relationship between tense and aspect in terms of UT, TSIT, and T-AST as follows:
(7)
a. Tense is a temporal relation between TU and T-AST. b. Aspect is a temporal relation between T-AST and T-SIT.
(7a) states that tense is a temporal relation between the time when the utterance is made and the assertion time, that is, the time for which an assertion is made by the utterance. (7b) states that Russian aspect, that is, the perfective/imperfective opposition, is a temporal relation between the assertion time and the situation denoted by the propositional content. The definitions in (7) give us the correct readings on (5a–c), and they also explain why (7) is not a contradiction in Russian. At the same time, (7) accounts for the intuition that tense relates the situation time to some deictic center; in other words, tense locates
69 the situation in time relative to the time of utterance. This is because TU and T-SIT are temporally related to each other, via T-AST. I propose that the temporal relations in (7) can be represented as the binary opposition of coincidence/non-coincidence. In (7a), TU and T-AST either coincide or don’t. If the utterance time and the assertion time coincide, the result is the present tense. However, if they do not coincide, the result is either the past or the future tense.4 Similarly, in (7b), if the assertion time and the situation time coincide, the result is the imperfective aspect; if they do not coincide, the result is the perfective aspect. I use the notion of coincidence to convey that two time intervals may be identical, but they need not be. For example, in (8a) X and Y coincide. In (8a`), no part of X is such that it is not part of Y, and vice versa. In (8a``), X includes Y. For my purposes here I do not distinguish between these possibilities and refer to the relation between X and Y in both cases as coincidence. In contrast, in (8b) X and Y do not coincide: no part of X is such that it is also a part of Y, and vice versa: (8) a. Coincidence X a`. Y X a``. Y
4
There is no confusion between the past non-coincidence and the future non-coincidence because, as I argue later in the chapter, the past tense and the future tense interpretation arise via different mechanisms.
70 b. Non-Coincidence X Y1
Y2
The diagram in (8a) captures the meaning of imperfective aspect, as well as present tense. Since my definition of coincidence does not make reference to the boundaries or endpoints, this approach is better equipped to capture Russian stative verbs than that of Smith and Rappoport (1991) discussed in chapter 2. The diagram in (8b) captures the meaning of past and future tense. The question is, does (8b) also capture the meaning of perfective aspect? In other words, in what sense does perfective aspect express the relation of non-coincidence? I discuss perfective aspect in more detail in section 3.1.2.2. For the purposes of this section I only propose that the perfective aspect differs from the imperfective aspect (as well as from tense, for that matter) in the following respect. In the perfective aspect, T-AST picks out not an interval, but a point in time. Furthermore, this point is located after the final boundary of the situation time T-SIT. The relation of non-coincidence expressed by the perfective aspect is schematized in (9): 5 5
My treatment of the perfective aspect is similar to Hale’s (1986) idea of non-central coincidence. Hale proposes to define spatial, temporal, and identity relations in terms of “central” versus “non-central” (or “terminal”) coincidence. In the case of non-central coincidence, “the location of the figure corresponds to its trajectory (if moving) or its linear arrangement (if stationary), which can be viewed as ending… or beginning at the place.” (p. 240). Hale’s proposal is schematized by Demirdache and Uribe-Etxebarria as follows (p. 177): i. Central coincidence a. [I//////F] Non-central coincidence b. //////[I F] c. [I F]//////
71 (9)
Non-coincidence: Perfective aspect T-SIT
final boundary
T-AST In the next section, I demonstrate that the binary opposition coincidence/non-coincidence, instead of the three-way distinction between before, simultaneously, and after, is sufficient to derive the temporal interpretation of Russian clauses.
3.1.2. One Happy Coincidence: The Feature [± COIN] In the concluding section of chapter 2, I proposed that the interpretation of tense, aspect, and mood is determined by a semantic feature [X] on the heads T, Asp, respectively.6 I define the semantic content of this semantic feature as [±COIN] for “(non)coincidence”. As the term coincidence suggests, the realization of this feature necessarily involves two elements, which can either coincide or not. Following much recent work, I assume that tense and aspect relate, or order, two times (Zagona 1990, 1995; Klein 1995; Stowell 1995, 1996; Giorgi and Pianesi 1997; Demirdache and Uribe-Etxebarria 2000). Building on Klein’s (1995) theory, Demirdache and Uribe-Etxebarria (2000) propose a uniform structural representation of temporal relations, which captures the semantic parallel between tense and aspect syntactically. Under their approach, tense and aspect are dyadic predicates that take time-denoting arguments.
6
In chapter 4, I extend this analysis to the category mood. I propose that C also contains the feature [±COIN].
72 Following Stowell (1996), Demirdache and Uribe-Etxebarria (2000) use the notions
AFTER, BEFORE,
and
WITHIN
to represent different temporal relations.7 For
example, tense orders the utterance time TU after or within the assertion time T-AST, resulting in the past and present interpretations, respectively. Aspect orders the assertion time T-AST after or within the situation time T-SIT, resulting in the perfect and the progressive interpretation, respectively: (10)
TP T`
TU
T
AspP
AFTER/WITHIN
T-AST
Asp` Asp
VP
AFTER/WITHIN
T-SIT
V`…
I propose to replace the notions WITHIN, BEFORE, and AFTER with the single binary feature [±COIN], as follows (cf. Ritter and Wiltschko 2005):8
(11) a. b.
BEFORE/AFTER WITHIN
= non-coincidence [–COIN]
= coincidence [+COIN]
Thus, the feature [±COIN] on T and Asp determines the temporal relations as follows: 7
Demirdache and Uribe-Etxebarria use the notion BEFORE to convey prospective aspect and future tense. In my discussion, I focus on their notions AFTER and WITHIN. Also, they use the term event time EV-T instead of situation time T-SIT. To be consistent, here I preserve Klein’s (1995) original terminology.
73 (12) a. [±COIN] on T
-->
b. [±COIN] on Asp -->
(Non)Coincidence of TU and T-AST (Non)Coincidence of T-AST and T-SIT
The main advantage of this approach is that the content of the functional projections TP and AspP is described in terms of the same features. In 3.1.2.1–3.1.2.2, I justify the replacement of AFTER, BEFORE and WITHIN on T and Asp by the single feature [±COIN].
3.1.2.1. The Feature [±COIN] on T According to (10), the feature [–COIN] on T corresponds to the notions BEFORE and AFTER in Demirdache and Uribe-Etxebarria’s approach, which in turn give rise to the past and the future temporal reference. In other words, the feature [–COIN] on T can receive either past or future interpretation. To distinguish between the two I propose the following solution. Following Zagona (2003), I assume that tense licensing and tense construal do not require a tense-specific theory of licensing.9 Being syntactic arguments, times are licensed in the structure by the same principles as other arguments. In particular, they are subject to the visibility requirement for interpretation.10 I assume also that the visibility is ensured by formal features. For example, the visibility of a direct object is ensured by the presence of the relevant formal feature [CASE] on V. 8
Although Demirdache and Uribe-Etxebarria (2000) make use of the notion of (non)coincidence they do not employ it as a feature. 9 Zagona’s (2003) analysis contrasts with Enç (1987), who proposes a tense-specific theory of licensing. In particular, according to Enç, in matrix clauses the external argument of T is licensed by Comp, which denotes the speech time. Note that earlier theories of tense by Stowell (1982) and Pesetsky (1982a) also require the presence of a CP projection. This requirement is eliminated under the present approach. This will become relevant in chapter 4.
74 Zagona (2003) proposes that the formal feature on T that licenses its temporal arguments is [±PAST]. Crucially, under her analysis this feature is not a semantic feature; that is, [±PAST] does not have a semantic interpretation of a specific ordering relation. My understanding is that this formal feature is supplied by tense morphology, whose temporal interpretation is not fixed by a tense feature, but is derived compositionally, based on a specific syntactic context. As was discussed in chapter 2, Russian distinguishes between past and non-past stems, which differ morphologically, as shown again in (13):
(13) a. Past:
čita- (past tense suffix -l; agreement: number/gender)
b. Non-Past: čitaj- (agreement: person/number)
Following Zagona (2003), I assume that Russian verb stems are specified in the lexicon for [±PAST], just as transitive verbs are specified in the lexicon for the relevant formal feature of structural case. For example, (13a) is specified in the lexicon as [+PAST]. However, this does not mean that it automatically receives the semantic interpretation past, as, for example, in subjunctive clauses (see chapter 4). A past tense stem (or a past tense morpheme) is one that, if inserted into a specific syntactic environment, receives the semantic interpretation past. Thus, the feature [±PAST] on T is a formal feature whose
10
My understanding is that temporal arguments are “visible” in the structure in the sense that they are present, or available for the interpretation.
75 value does not determine its semantic interpretation. Each value only licenses a certain relation, or a range of relations, which is in turn the basis for the interpretation. Now a distinction should be made between the relations of coincidence and noncoincidence, which are overtly marked by [–PAST] and [+PAST], respectively. I propose that the value [+PAST] always licenses the relation of non-coincidence, which is analyzed as disjoint from the speech time. In contrast, the value [–PAST] can license either the relation of coincidence or non-coincidence. The only relation of non-coincidence with the speech time which is non-past is future:
(14)
Formal feature
[+PAST]
Licensed relation
[–COIN]
[–COIN]
[+COIN]
‘past’
‘future’
‘present’
Semantic interpretation
[–PAST]
To sum up, the feature [±PAST] is a formal feature that licenses certain relations. The feature [±COIN] is a semantic feature that serves as the basis for the specific semantic interpretation. Under this approach, the difference between the past and the future tense interpretation is the value [+PAST] or [–PAST]. The question of how exactly the relation of (non)coincidence arises from the formal feature[±PAST] is addressed in section 3.3. In the next section, I discuss the semantic feature [±COIN] on Asp.
76 3.1.2.2. The Feature [±COIN] on Asp Demirdache and Uribe-Etxebarria transfer Klein’s (1995) analysis of Russian to languages such as English. They define the spatiotemporal relations in terms of the location of the figure with respect to the ground. Thus, in (15) the head Asp locates its external argument T-AST (the figure) relative to its internal argument T-SIT (the ground): (15)
AspP T-AST(the figure)
Asp`
Asp
VP
WITHIN/AFTER
T-SIT(the
ground)
V`…
For example, Demirdache and Uribe-Etxebarria (2000) analyze the perfect as aspect which orders T-AST after T-SIT because the perfect picks out an assertion time TAST (the figure) after the final bound of the situation time T-SIT (the ground). In (16), the assertion time is after the final bound of the leaving event in the sense that the event of leaving must be completed. There is no context in which this event could be interpreted as not completed:
(16) I have left. / I had left.
At first glance, the Russian perfective aspect cannot be analyzed in this way because it does not necessarily pick out T-AST after the final bound of T-SIT. For
77 example, in (17a) the perfective verb expresses the completed event of singing, whereas in (17b), the perfective verb expresses the beginning of a singing event: 11,12
(17) a. Philemon s-pe-l Philemon
PREF-sing-PAST(PERF)
pesnju. song
‘Philemon sang the song (=Philemon finished the song.)’ b. Philemon za-pe-l
pesnju.
Philemon PREF-sing-PAST(PERF) song ‘Philemon started singing the song.’
Under Demirdache and Uribe-Etxebarria’s (2000) approach, in (17a) the perfective aspect locates the assertion time T-AST after the situation time T-SIT: the event of singing is completed. The question is where does the perfective aspect locate T-AST relative to T-SIT in (17b)? Obviously, the answer cannot be “after”. However, it cannot be “before” either, because (17b) explicitly states that the singing event has begun. 11
The semantics of Russian aspect has been a matter of a long debate. Klein (1995) provides an overview of definitions, which include the following: (i) The perfective aspect presents a situation in its totality, whereas the imperfective aspect presents a situation not in its totality, or from within (Forsyth 1970; Comrie 1976). (ii) The perfective aspect presents a situation as complete, whereas the imperfective aspect presents a situation as an incomplete (Miklosich 1883). (iii) The perfective aspect is associated with an inherent boundary, whereas the imperfective aspect is not (Jakobson 1932; Timberlake 1985; Dahl 1985; Zaliznjak and Šmelev 1997). In my view, one of the problems with providing an adequate definition of Russian aspect is that researchers often do not distinguish between the semantics of aspect and the conventions of its use, or pragmatics (see Smith and Rappaport 1991 for a definition that takes into account this distinction). In addition, the perfective/imperfective opposition (grammatical aspect) interacts with lexical aspect (see Paducheva 1995; Smith and Rappaport 1991). For example, as Smith and Rappaport (1991) point out, perfective aspect is not available for states. Stative verbs have perfective forms, but these are classified as achievements. However, since here I am not directly concerned with the semantics of Russian aspect, the examples in (16) are sufficient to make the point.
78 This apparent problem may be solved as follows. Let us suppose that in (17b) the beginning of the singing event is viewed as an event in its own right. In other words, we distinguish between the event of singing and the event of beginning to sing. One consideration that justifies such an approach is the fact that the singing event and the “beginning event” are referred to by different verbs: the perfectivizing prefix za- in (17b) affects the lexical properties of the verb pet’ ‘sing’. The perfective verb in (17b) has its own secondary imperfective counterpart derived by suffixation. As discussed in section 2.1.1 of chapter 2, secondary imperfective is created when a perfectivizing prefix changes the verb’s lexical properties. Thus, the imperfective verbs in (18) have different meaning. In (18a), the singing is in progress, whereas in (18b) the beginning of singing is in progress:
(18) a. Philemon pe-l
pesnju.
Philemon sing-PAST(IMPF) song ‘Philemon was singing a song.’ b. Philemon za-pe-va-l
pesnju.
Philemon PREF-sing-IMPF-PAST song ‘Philemon was starting to sing the song.’
12
This contrast is due to distinct lexical semantic properties of the perfectivizing prefixes s- and za- in (17a) and (17b), respectively. This suggests that the constraints on T-AST picked out by the perfective aspect in Russian are semantic and are not expressed in the morphosyntax.
79 Under this approach, the binary feature [±COIN] on Asp captures the nature of the Russian aspects. The imperfective aspect locates T-AST within T-SIT: it presents a situation in progress and is not concerned with the temporal boundaries of this situation. In this sense, in the imperfective aspect T-AST coincides with T-SIT. In contrast, in the perfective aspect T-AST picks out the point after the final boundary of the T-SIT. The representations of the perfective verbs in (19) contrast with the representation of the imperfective verbs in (18) repeated below as (20):
(19) Perfective Aspect a. Philemon s-pe-l Philemon
pesnju.
PREF-sing-PAST(PERF)
song
‘Philemon sang the song.’ T-SIT (the singing event) T-AST b. Philemon za-pe-l
pesnju.
Philemon PREF-sing-PAST(PERF) song ‘Philemon started singing the song.’ T-SIT (the beginning-of-singing event) T-AST
80 (20) Imperfective Aspect a. Philemon pe-l
pesnju.
Philemon sing-PAST(IMPF) song ‘Philemon was singing a song.’ T-SIT (the singing event)
T-AST b. Philemon za-pe-va-l
pesnju.
Philemon PREF-sing-IMPF-PAST song ‘Philemon was starting to sing the song.’ T-SIT (the beginning-of-singing event)
T-AST The feature [±COIN] on Asp is a semantic feature, which expresses the relation of (non)coincidence, which in turn serves as the basis for the perfective or imperfective interpretation. To maintain the parallel between tense and aspect, we now need to establish a formal feature on Asp, which licenses the relation of (non)coincidence. Following Klein (1995), I assume that whether a given verb is perfective or imperfective is a part of Russian speakers’ lexical knowledge. To describe this knowledge, he posits
81 the feature [±P], which is a property of lexical entries.13 Crucially, this feature should not be confused with the aspect itself. The aspectual differentiation arises only when the lexical verb becomes part of a finite construction, since only finite clauses typically make assertions. Thus, for the aspect to be able to order the assertion time T-AST relative to the situation time T-SIT a finite clause is required. I propose that the feature [±P] is the formal feature which licenses the temporal arguments of Asp, just as the formal feature [±PAST] licenses the temporal arguments of T. The only difference is that with respect to aspect, there are only two possible semantic interpretations, each of which corresponds to the positive or the negative value of [P]:
(21)
Formal feature Licensed relation Semantic interpretation
[+P]
[–P]
[–COIN]
[+COIN]
‘perfective’
‘imperfective’
To sum up, I propose that tense and aspect are realized by means of two kinds of features: purely formal features [±PAST] and [±P] on T and Asp, respectively, and the semantic feature [±COIN]. The purely formal features [±PAST] and [±P] license a particular relation, or a range of relations, that can be expressed by the categories tense and aspect. This relation or range of relations is represented by the semantic feature [±COIN]. This
13
My understanding is that in Klein’s theory, [P] stands for “perfective”. Thus, verbs come from the lexicon marked as either [+perfective] (perfective verbs) or [–perfective] (imperfective verbs).
82 semantic feature is the basis for a particular semantic interpretation, such as ‘past’, ‘present’ or ‘future’ for tense, or ‘perfective’ or ‘imperfective’ for aspect.14 In the next section, I propose, following Pesetsky and Torrego (2004b), that features can be interpretable or uninterpretable, and valued or unvalued. I assume that the purely formal features [±PAST] and [±P] can be valued or unvalued, but they cannot be interpretable or uninterpretable. In contrast, the semantic feature [±COIN] can be valued or unvalued, and interpretable or uninterpretable.
3.1.3. Interpretability, Valuation, and Agreement Following recent work by Pesetsky and Torrego (2004b), I assume that various features interact through agreement and thus influence the shape and the semantic interpretation of the syntactic structure. These features can be interpretable or uninterpretable and valued or unvalued, as shown in (22):15
(22)
14
Types of features (Pesetsky and Torrego 2004b:5) uF val uninterpretable, valued
iF val interpretable, valued
uF [ ] uninterpretable, unvalued
iF [ ] interpretable, unvalued
An alternative approach can be proposed, according to which features [±PAST] and [±P] are semantic interpretable features and not purely formal features. Although it is possible that semantic features license syntactic arguments, such an approach cannot explain why imperfective aspect is always interpreted as present tense, whereas perfective aspect is always interpreted as future tense. Furthermore, if feature [+PAST] is interpretable, it is not clear why it receives past tense interpretation in indicative clauses and irrealis interpretation in subjunctive clauses. In sections 3.3–3.4, I demonstrate that positing one formal and one interpretable feature on T and Asp allows for an elegant account of these facts. 15 The notions of interpretability and valuation were introduced in Chomsky (2000, 2001), with the following biconditional relation between them: A feature F is uninterpretable iff F is unvalued (Chomsky
83 Let us first consider valuation. Pesetsky and Torrego propose that certain features come from the lexicon unvalued, and receive their value from a valued instance of the same feature on another lexical item. For example, in languages such as Latin agreement between adjectives and nouns involves the feature [NUMBER]. The fact that the adjective in (23) bears the feature [NUMBER] (singular in 23a and plural in 23b) is due to the property of the noun that it modifies. The noun puella ‘girl’ but not the adjective Romanus ‘Roman’ is specified in the lexicon as singular or plural:
(23) a. Haec
puella
Romana
ambulat.
this-NOM.FEM.SG girl.NOM.FEM.SG Roman-NOM.FEM.SG walks-3RD.SG ‘This Roman girl walks.’ b. Hae
puellae
Romanae
ambulant.
these-NOM.FEM.PL girl.NOM.FEM.PL Roman-NOM.FEM.PL walk-3RD.PL ‘These Roman girls walk.’
(Pesetsky and Torrego 2004b:1)
Nouns come out of the lexicon with a valued feature [NUMBER]. Pesetsky and Torrego point out that this claim is supported by the existence of pluralia tantum, such as scissors or glasses. However, on adjectives this feature is not valued in the lexicon, but receives its value via agreement with the feature [NUMBER] on the modified noun. This can be represented as follows:
2001:5). Pesetsky and Torrego’s (2004b) analysis differs from Chomsky’s approach in that interpretability
84 agree relation (24) N i [NUM: SG] A u [NUM: ] --> N i [NUM: SG] A u [NUM: SG]
As shown in (24), adjectives start out with the unvalued feature [NUMBER]. If this feature remains unvalued, that is, if the adjective does not agree with the modified noun, the sentence will be ungrammatical. This suggests that the notion of valuation is relevant for syntax. It is assumed by Pesetsky and Torrego that a derivation may not contain unvalued features. Every unvalued feature which enters a derivation must enter into an agree relation with a valued instance of the same feature. The agree relation is defined as follows:
(25) a. Agreement: A syntactic operation that results in valuation of an unvalued feature F by creating a link between an unvalued feature F1 (a goal) and a valued feature F2 (a probe). b. Feature sharing: F2 (a probe) replaces feature F1 (a goal), so that the same feature is present in both syntactic locations.
As a result of the syntactic agreement, the feature [NUMBER] becomes valued on both the noun and the adjective; the adjective and the modified noun share the feature [NUMBER]. I provide a schematic representation of the agree operation in the next section, where I discuss the past tense interpretation of Russian finite clauses.
and valuation are independent of each other.
85 Now let us consider the notion of interpretability. The distinction between interpretable and uninterpretable features is concerned with the question of whether or not a particular feature makes a semantic contribution to the interpretation of a lexical item. For example, although the adjective and the modified noun share the feature [NUMBER] as a result of agreement, the interpretation of this feature differs for the two items. When the feature [NUMBER] appears on a noun it makes a semantic contribution to the interpretation of this noun: chair and chairs have different interpretations because they refer to a single object and to at least two objects, respectively. As discussed above, in Latin the feature [NUMBER] appears on nouns (valued in the lexicon) and adjectives (valued via agreement). However, these features make no semantic contribution to the interpretation of adjectives. We do not think about adjectives as being specified as singular or plural. They are singular or plural by virtue of syntactic agreement with a modified noun, which is specified for these features. The main points of Pesetsky and Torrego’s theory are summarized below:
(26) a. Features can be interpretable or uninterpretable and valued or unvalued. b. A derivation cannot contain unvalued features. c. Agreement between two features on two distinct heads results in the valuation of the unvalued feature F. d. As a result of agreement, the same feature is present on both heads.
86 Following Pesetsky and Torrego, I assume that the purely formal features [PAST] and [P] and the semantic feature [COIN] can be valued or unvalued. However, only the semantic feature [±COIN] can be interpretable or uninterpretable. I assume also that the value of the purely formal features [PAST] and [P] on T and Asp is supplied by verbal morphology via agreement between T and V, and Asp and V. The value of the formal semantic feature [COIN] is typically determined by the value of [PAST] and [P]. I argue in section 3.2 that this is the case in the past tense: [+PAST] gives rise to [–COIN]. However, in the non-past, the value [–PAST] is not specific enough to determine the value of the semantic feature [COIN]: non-past can be either present or future. In section 3.3 I propose that the feature [COIN] in the non-past receives its value via agreement with the valued instance of the feature [COIN] on Asp. The table below summarizes the temporal interpretation of Russian finite clauses in the indicative mood. The tense forms are grouped together based on whether or not the temporal interpretation involves agreement between T and Asp. The representation shows the stage in the derivation before the agreement between T and Asp in the non-past takes place. As the shaded boxes in (27) indicate, the feature [COIN] on T in the non-past is unvalued, because it cannot receive its value from the valued feature [–PAST]:
87 (27)
The distribution of the formal features on T and Asp Agreement between T and Asp Present Future
T
Asp
No Agreement between T and Asp Past Future Imperfective
[–PAST]
[–PAST]
[+PAST]
[–PAST]
i[COIN]
i[COIN]
i[–COIN]
i[–COIN]
[+P]
[–P]
[±P]
[+P]
i[+COIN]
i[–COIN]
i[±COIN]
i[+COIN]
This analysis enables us to maintain that tense and aspect are distinct grammatical categories. This is expressed by different formal features on T and Asp. However, the agreement between these distinct categories is possible because they express relations which are similar in some relevant respect. In particular, both relations can be characterized as a relation of (non)coincidence, which is represented by the interpretable feature [±COIN] on both T and Asp.
3.2. INTERPRETATION OF PAST TENSE CLAUSES I begin with the discussion of past tense clauses because they constitute the unmarked case as far as feature valuation is concerned. Past tense verb forms contain both tense and aspectual morphology, which means that the formal features on T and Asp receive their value from distinct morphemes, and no agreement between T and Asp is necessary. The past tense verbs in Russian are created by adding the suffix –l to a past tense stem, which is the only overt tense marker in Modern Russian. Past tense verbs can be
88 either perfective or imperfective. Thus, the past tense verb forms are marked overtly for both tense and aspect, and the past temporal interpretation does not depend on aspect:
(28) a. Liza čita-l-a
knigu (*sejčas/*zavtra/včera).
Liza read-PAST(IMPF) book
now/tomorrow/yesterday
‘Liza was reading a/the book *now/*tomorrow/yesterday.’ b. Liza pro- čita-l-a Liza
PREF-read-PAST(PERF)
knigu (*sejčas/*zavtra/včera). book
now/tomorrow/yesterday
‘Liza (has) read a/the book *now/*tomorrow/yesterday.’
The fact that the temporal interpretation is not determined by aspect means that there is no agreement between the features on T and Asp, and that tense and aspect are interpreted independently of each other. In the next section, I outline a theoretical mechanism that accounts for the past tense interpretation of Russian finite clauses. I assume that the past tense verb forms come from the lexicon with the uninterpretable valued features [+PAST] and [±P], whose value is supplied by the past tense suffix –l and the aspectual morphology, respectively. However, the verbs themselves do not contain the feature [COIN]. This feature is present on Asp and on T: the tensed verb form cannot express the relationship of (non)coincidence unless it is a part of a syntactic structure. Let us consider the derivation step by step. I begin with the merge of the head Asp:
89 (29) Step 1: Merge of Asp AspP Asp` Asp u[P] i[COIN]
VP V` V u[+PAST] u[±P]
The head Asp contains the uninterpretable feature [P] and the interpretable feature [COIN], both unvalued. The interpretable feature [COIN] cannot receive its value from V because V does not carry an instance of the same feature. The unvalued feature [P] scans its ccommanding domain and finds a valued instance of the same feature on V. The agreement takes place and the formal feature [P] on Asp becomes valued: (30)
Step 2: Agree between Asp and V AspP Asp` Asp u[±P] i[COIN]
VP V` V u[+PAST] u[±P]
90 As a result of the agree relation in (30), a link is established between the two instances of the same feature; the same valued feature [±P] is now present in both syntactic locations. Being valued, the uninterpretable formal feature [±P] becomes interpretable by licensing the relation of (non)coincidence: It supplies the value of the interpretable unvalued feature [COIN] on Asp. As a result, either the perfective or the imperfective interpretation arises, depending on the value of [P]:
(31) Step 3: Licensing of the (non)coincidence relation: perfective/imperfective AspP Asp` Asp u[±P]
VP
i[±COIN: IMPF/PERF]
V`
V u[+PAST] u[±P] At step 4 in the derivation, T merges into the structure. T carries the uninterpretable unvalued feature [PAST] and the interpretable unvalued feature [COIN]:
91 (32)
Step 4: Merge of T TP T` T
AspP
u[PAST] i[COIN]
Asp` Asp u[±P]
VP V`
i[±COIN: IMPF/PERF]
V u[+PAST] u[±P] The uninterpretable unvalued feature [PAST] on T scans its c-commanding domain and finds a valued instance of the same feature on V. The agree operation takes place: (33) Step 5: Agree between T and V TP T` T
AspP
u[+PAST] i[COIN]
Asp` Asp u[±P]
VP V`
i[±COIN: IMPF/PERF]
V u[+PAST] u[±P] As a result of the agree relation in (33), a link is established between the two instances of the same feature; the same valued feature [+PAST] is now present in both syntactic
92 locations. Being valued, the uninterpretable formal feature [+PAST] becomes interpretable by licensing the relation of non-coincidence: It supplies the value of the interpretable unvalued feature [COIN] on T, and the clause receives the past tense interpretation: (34) Step 6: Licensing of the non-coincidence relation: past TP T` T
AspP
u[+PAST] i[–COIN: PAST]
Asp` Asp u[±P]
VP V`
i[±COIN: IMPF/PERF]
V u[+PAST] u[±P] In the past, the value of the feature [COIN] on T is obtained “for free”: Since the value [+PAST] can only result in a non-coincidence relation, where the situation time T-SIT precedes the utterance time TU, no additional operations are required for the interpretation of past tense clauses. In the next section, I discuss the non-past clauses, in which the value of the formal feature [PAST] on T is not specific enough to license the relationship of (non)coincidence.
93 3.3. TEMPORAL INTERPRETATION OF THE NON-PAST As mentioned earlier, the temporal reference of a Russian non-past clause is determined by aspect. The non-past tense verbs in (35) have the same inflectional morphology (person and number) and differ only in aspect. The morphologically simple verb in (35a) is imperfective, whereas the prefixed verb in (35b) is perfective. Accordingly, (35a) and (35b) receive a present and a future tense interpretation, respectively:16
(35) a. Liza čitaj-et
lekciju v universitete (sejčas/*včera/*zavtra).
Liza read-3RD.SG(IMPF) lecture in university
now/ yesterday/tomorrow
‘Liza is reading a lecture at the university (now/*yesterday/*tomorrow).’ b. Liza pro- čitaj-et Liza
PREF-read-3RD.SG(PERF)
lekciju v universitete (*sejčas/ zavtra /*včera). lecture in university
now/tomorrow/yesterday
‘Liza will (have) read a lecture at the university (*now/tomorrow/*yesterday).’
The morphologically imperfective verb in (35a) consists of the stem čitaj- and inflectional agreement morphology. Although there is no tense marking on the verb, the event of Liza giving a lecture is interpreted as coinciding with the utterance time. The morphologically perfective verb in (35b) contains the perfectivizing prefix pro-. In (35b), the event of Liza giving a lecture is interpreted as following the utterance time. The data in (35) can be informally described as follows. The aspectual morphology, that is, the morphologically simple verb in (35a) and the prefix in (35b),
16
See chapter 2 for discussion of temporal interpretations available for perfective and imperfective verbs.
94 performs the double duty of specifying both, tense and aspect. I hypothesize that although the stem is specified as non-past, this is not enough to supply the temporal reference of the clause. The value non-past is too general to distinguish between the present and the future. I propose that in this case the meaning of the aspectual morphology (perfective/imperfective)
can
be
recycled
to
supply
the
temporal
meaning
(present/future). I propose the mechanism of value recycling, which allows the same morpheme to be interpreted in more than one syntactic location within the same structure. In 3.3.2–3.3.3, I formalize this proposal. In the next section, I show that the aspectual interpretation of non-past tense verb forms arises in the same way as the one in the past.
3.3.1. Aspectual Interpretation of the Non-Past The valuation of the formal feature [±P] and [±COIN] on Asp proceeds in the same way as it does in the past tense clauses. The only difference is that non-past tense verb forms come from the lexicon with the uninterpretable valued feature [–PAST]. The unvalued feature [P] on Asp scans its c-commanding domain and finds a valued instance of the same feature on V. As a result of an agree relation, a link is established between Asp and V, and [P] on Asp becomes valued. Now the same feature is present in both syntactic locations. The valued instance of [P] on Asp licenses the relation of (non)coincidence by giving its value to the interpretable feature [COIN]:
95 (36) a. Step 1: Merge of Asp AspP Asp` Asp u[P] i[COIN]
VP V` V u[–PAST] u[±P]
b. Step 2: Agree between Asp and V AspP Asp` Asp u[±P] i[COIN]
VP V` V u[–PAST] u[±P]
c. Step 3: Licensing of the (non)coincidence relation: perfective/imperfective AspP Asp` Asp u[±P]
VP V`
i[±COIN: IMPF/PERF]
V u[–PAST] u[±P]
96 As in past tense clauses, T merges with AspP. It carries an uninterpretable unvalued feature [PAST] and an interpretable unvalued feature [COIN]. The feature [PAST] on T receives its value via the syntactic operation of agreement with the valued instance of the same feature on V: (37) a. Step 4: Merge of T TP T` T
AspP
u[PAST] i[COIN]
Asp` Asp u[±P]
VP V`
i[±COIN: IMPF/PERF]
V u[–PAST] u[±P] b. Step 5: Agree between T and V TP T` T
AspP
u[–PAST] i[COIN]
Asp` Asp u[±P]
VP V`
i[±COIN: IMPF/PERF]
V u[–PAST] u[±P]
97 At this point, the valued feature [–PAST] on T should license the relation of (non)coincidence: It should supply the value of the interpretable feature [COIN]. The problem is that the value ‘–’ is too general in this case, since Russian has two non-past tenses: present and future. Suppose, for the sake of argument, that the derivation proceeds as in past tense clauses, and the valued feature [–PAST] on T licenses the relation of noncoincidence between the utterance time and the assertion time. The semantic feature [– COIN]
licensed by the formal feature [–PAST] gives rise to a future tense interpretation.
This would work for a clause with a perfective verb: (38) Step 6: Licensing of the non-coincidence relation: future TP T` T u[–PAST]
AspP
i[–COIN: FUTURE]
Asp` Asp u[+P]
VP V`
i[–COIN: PERF]
V u[–PAST] u[+P]
This is the structure of a clause which has a future perfective interpretation, as shown in (39):
98 (39) Liza pro- čitaj-et Liza
lekciju v universitete (*sejčas/ zavtra /*včera).
PREF-read-3RD.SG(PERF)
lecture in university
now/tomorrow/yesterday
‘Liza will (have) read a lecture at the university (*now/tomorrow/*yesterday).’
However, such an approach implies that the structure in (40) is also possible, where the non-past tense verb is imperfective:17
(40)
Step 6`: Licensing of the non-coincidence relation: future TP T` T u[–PAST]
AspP
i[–COIN: FUTURE]
Asp` Asp u[–P]
VP V`
i[+COIN: IMPF]
V u[–PAST] u[–P] According to this structure, a clause with a morphologically simple non-past tense verb can receive a future tense interpretation, which is not true. A sentence such as in (41) is not interpreted as referring to an event in progress, which takes place after the time of
17
Although the future imperfective exists in Russian, it is a complex tense, that is, the morphological means of expressing it are very different from the future perfective. I discuss the future imperfective in section 3.4.
99 utterance; it denotes an event in progress, which co-occurs with the time of utterance, that is, it receives the present tense reading:
(41) Liza čitaj-et
gazetu
(sejčas/*včera/*zavtra).
Liza read-3RD.SG(IMPF) newspaper now/ yesterday/tomorrow ‘Liza is reading a newspaper (now/*yesterday/*tomorrow).’
The structure in (40) cannot account for the interpretation in (41). I conclude therefore that although the feature [–PAST] is valued, it is unable to supply the value of [COIN]. In other words, it is unable to license a specific relation of (non)coincidence. Therefore, the interpretable feature [COIN] has to look for its value elsewhere.
3.3.2. Recycle, Where Facilities Exist! I propose that since the feature [COIN] on T cannot receive its value directly from the valued formal feature [–PAST], it has to enter into an agree relation by itself. Otherwise it will remain unvalued and the derivation will crash. I assume that under other circumstances (for instance, in the past tense) an agreement between two instances of a semantic feature is ruled out by usual economy considerations. Purely formal features must enter into an agree relation with the relevant feature on a lexical head because this is the only way they can become valued. In contrast, semantic features express relations that serve as the basis for interpretation; they can be viewed as interpretable instantiations of purely formal features. Where possible, such semantic features receive their value
100 directly from a valued instance of the purely formal feature, for example, [+PAST] --> [– COIN].
I assume that acquiring a value via the agree operation that creates a link between two syntactic locations is more costly than acquiring this value by virtue of being the interpretable instantiation of a purely formal valued feature located on the same functional head. With this in mind, let me introduce the mechanism of value recycling. This mechanism may be employed when a feature F cannot receive its value from a designated morpheme. To put it simply, value recycling is allowed when the number of features that need valuation exceeds the number of morphemes capable of supplying a value. Value recycling results in a situation where the same morpheme is interpreted more than once in different syntactic locations within the same structure.
3.3.2.1. Imperfective is Recycled as Present Now, let us pick up where we left off, namely, the stage in the derivation when the formal feature [PAST] on T receives its value via agreement with the valued instance of the same feature on V:
101 (42) Step 5: Agree between T and V TP T` T u[–PAST] i[COIN]
AspP Asp` Asp u[–P]
VP V`
i[+COIN: IMPF]
V u[–PAST] u[–P] Since the valued feature [–PAST] is too general to license a particular relation of (non)coincidence that could be the basis for a tense interpretation, the semantic feature [COIN] itself has to scan its c-commanding domain looking for a valued instance of the same feature. The appropriate feature is found on Asp: This is an interpretable valued feature [+COIN], which gives rise to the imperfective interpretation:
102 (43) Step 6: Agree between T and Asp (imperfective) TP T` T u[–PAST]
AspP Asp`
i[+COIN: PRES]
Asp u[–P]
VP V`
i[+COIN: IMPF]
V u[–PAST] u[–P] As a result of this agreement, the feature [COIN] on T receives the value of the feature [COIN] on Asp: [+COIN]. However, since these features are located on different heads they give rise to different semantic interpretations. In particular, the feature [+COIN] on T provides the basis for the temporal, and not aspectual interpretation. The combination of [–PAST] and [+COIN] on T results in a present tense meaning. Note that the agreement in (43) takes place between two interpretable instances of the same feature. This is a formal representation of the fact that the same morpheme is interpreted simultaneously in two syntactic locations: on T and on Asp. The prediction is that the non-past tense clause is interpreted not only as present, but also as imperfective. I return to this issue in section 3.3.4 and show that this is indeed the case for both imperfective and perfective verbs. In the next section, I discuss the value recycling in non-past perfective clauses.
103 3.3.2.2. Perfective is Recycled as Future The future tense interpretation of a non-past tense verb form arises when Asp contains an interpretable feature [–COIN]. On Asp, the relation of non-coincidence is interpreted as perfective: (44) Step 6: Agree between T and Asp (perfective) TP T` T
AspP
u[–PAST] Asp`
i[–COIN: FUT]
Asp u[+P]
VP V`
i[–COIN: PERF]
V u[–PAST] u[+P] Note that this analysis captures the fact that in Russian, the perfective aspect in the
absence
of
tense
morphology
receives
a
future
tense
interpretation.
Crosslinguistically, it is much more common for a perfective aspect to receive a past tense reading (Bybee, Perkins, and Pagliuca 1994:278). For example, according to Lin (2005), the perfective aspect in Chinese (a language without tense morphology) always gives rise to a past tense interpretation. Lin (2005) argues that not only does Chinese have no morphological tenses, but there is no need to resort to covert features under a tense node in syntax to arrive at a temporal interpretation of a Chinese clause. Following Lin
104 (2005), Chinese verbs are not specified for tense at all. Consequently, the past tense interpretation arises from the meaning of the perfective aspect alone. The perfective aspect presents an event as completed, and the state of completeness is most naturally associated with an event in the past. In contrast, since Russian verbs are specified as past and non-past, the temporal interpretation arises compositionally from the meaning “perfective” and the meaning “non-past”. In other words, the combination of features on T in (44) cannot give rise to a past tense interpretation due to the presence of [–PAST]; it cannot receive a present tense interpretation due to the presence of [–COIN]. Thus, the only possible interpretation is future.
3.3.3. The Double Interpretation: Perfective/Future and Imperfective/Present In this section, I return to the issue of the double interpretation. Under my analysis, the agreement takes place between two interpretable instances of the same feature. The prediction is that the same morphology should now be interpreted simultaneously as aspect and tense, in two distinct syntactic locations. This prediction is borne out, which can be demonstrated as follows. In Russian, stative verbs such as stoit’ ‘cost’ or vesit’ ‘weigh’ are always imperfective; they do not have perfective counterparts (Smith and Rappaport 1991:301). However, all Russian verbs can have a future tense meaning. Suppose that in the non-past, the feature [–COIN] is interpretable on T but uninterpretable on Asp. This means that the morphologically perfective non-past form has the temporal (future) meaning, but not the perfective meaning. As a result, stative verbs are expected
105 to be compatible with the perfective morphology in the non-past, because they receive the future, and not the perfective interpretation. However, as the examples below demonstrate, this is not the case. In (45a), the morphologically imperfective verb stoit’ ‘cost’ has a present tense meaning. In (45b), it is combined with three different prefixes and the result is ungrammatical:
(45) a. Etot dom
sto-it
sostojanije.
this house cost-3RD.SG(IMPF) fortune ‘This house costs a fortune.’ b. *Čerez god, etot dom po-/na-/so-sto-it in
sostojanije.
year this house PREF1-/PREF2-/PREF3-cost-3RD.SG(PERF) fortune
‘In a year, this house will cost a fortune.’
The ungrammaticality of (45b) is explained if the prefixed form is not only morphologically perfective, but also semantically perfective, in addition to having the future temporal interpretation. Since the perfective aspect is incompatible with the stative meaning of verbs such as stoit’ ‘cost’, the sentence in (45b) is ungrammatical. The only future tense form of stative verbs is the periphrastic future tense, which has the imperfective aspect. The periphrastic future tense (the future imperfective) is discussed in section 3.4. It is formed by the inflected auxiliary byt’ ‘be’ and an imperfective infinitive:
106 (46) Čerez god, etot dom in
bud-et
stoit’
sostojanije.
year this house be-3RD.SG cost-INF(IMPF) fortune
‘In a year, this house will cost a fortune.’
Similar evidence suggests that other verb forms in the simple non-past receive both the aspectual and the temporal interpretation. For example, morphologically perfective non-stative verbs are semantically perfective as well. Following Smith and Rappaport (1991), the perfective aspect presents an event with both initial and final endpoints. This is demonstrated by the following perfective example, where only the sequential reading of the two clauses is available:
(47) Kogda ona po-stroit when she
PREF-build(PERF)
dom, ona kupit
mašinu.
house she buy(PERF) car
‘When she finishes building the house she will buy a car.’
In (47), the event of building a house and the event of buying a car are both in the future with respect to the utterance time. However, the event of building a house must be completed before the event of buying a car occurs. This interpretation contrasts with the future imperfective sentence in (48), where the event of buying a car co-occurs with the event of building a house:
107 (48) Kogda ona budet
stroit’
dom, ona kupit
mašinu.
when she be-3RD.SG build-INF(IMPF) house she buy(PERF) car ‘While she is building a house, she will buy a car.’
In (48), again the event of building a house and the event of buying a car are both in the future relative to the utterance time. However, the event of buying the car is necessarily included in the event of building a house. The interpretation in (48) is due to the fact that unlike the perfective aspect, the imperfective aspect focuses on a part of a situation excluding its initial and final points. The morphologically imperfective non-past tense verb form in (49) is semantically imperfective as well. The event denoted by the when-clause is in progress at the time of the matrix event:
(49) Kogda ona rabotaj-et,
ona odevaj-et
očki.
when she work-3RD.SG(IMPF) she put-on-3RD.SG(IMPF) glasses ‘When she works she puts on her glasses.’
These examples indicate that in the non-past, the aspectual morphology in Russian carries two pieces of information: the perfective/imperfective contrast, and the future/present temporal meaning. This is captured by the proposal that in the non-past, the functional heads T and Asp share a formal feature via the syntactic operation of agree. Moreover, by
108 virtue of the value recycling mechanism, they share an interpretable value of the same feature, since the same morphology is interpreted twice: on Asp and on T.
3.4. THE PERIPHRASTIC FUTURE TENSE 3.4.1. The Data In this section, I discuss the periphrastic future tense, or the future imperfective. This is a complex tense, formed by the inflected form of the perfective auxiliary byt’ ‘be’ and an imperfective infinitive. Note that just like the simple non-past tense main verbs, the periphrastic future contains no tense morphology. The future temporal interpretation is determined by the auxiliary, which shows the same agreement as the non-past tense forms (person and number). The imperfective aspect comes from the infinitive. Note that in (50), the auxiliary is glossed as perfective. Its perfective status is discussed immediately below:
(50) a. My bud-em
čitat’
knigu.
we be-1ST.PL(PERF) read-INF(IMPF) book ‘We will be reading a/the book.’ b. Čerez god, etot dom in
bud-et
stoit’
sostojanije.
year this house be-3RD.SG(PERF) cost-INF(IMPF) fortune
‘In a year, this house will cost a fortune.’
109 The periphrastic future is a “frozen” form. It cannot contain the infinitival form of the auxiliary byt’, as in (51a); the auxiliary must be inflected. Neither can it contain a perfective infinitive, as shown by the ungrammaticality of (51b):
(51) a. *byt’
čitat’
knigu.
be-INF read-INF(IMPF) book ‘to be reading a book.’ b. *My bud-em we
pro-čitat’
knigu.
be-1ST.PL(PERF) PREF-read-INF(PERF) book
‘Philemon will have read a/the book.’
Since the auxiliary in the future imperfective is always inflected, I am going to refer to it by means of the stem bud- to which the inflectional morphology is added, rather than the infinitival form byt’. Before I turn to the analysis of the periphrastic future, let me briefly review the evidence from chapter 2 that the form bud- is perfective. The fact that it agrees with the subject in person and number suggest that it is a non-past tense form. (Recall that past tense verb forms agree with their subject in number and gender). Historically, this form has developed out of the perfective form of the Old Russian auxiliary byti ‘be’, which had complete perfective and imperfective paradigms, just like any other verb (van Schooneveld 1951; Jakobson 1957; Junghanns 1997).
110 (52)
Non-past paradigm of Old Russian byti ‘be’ Perfective
(53)
Imperfective
Singular
Plural
Dual
Singular
Plural
Dual
1st
budu
budemŭ
budevě
jesmĭ
jesmĭ
jesvě
2nd
budeši
budete
budeta
jesi
jeste
jesta
3rd
budetŭ
budutŭ
budete
jestě
sutě
jeste
Non-past paradigm of Modern Russian byt’ ‘be’ Singular
Plural
1st
budu
budem
2nd
budeš’
budete
3rd
budet
budut
It can be seen from the data in (52)–(53) that the Modern Russian auxiliary bud- in the future imperfective has developed out of the perfective form of byti. According to Jakobson (1957) and Franks and Greenberg (1994), in Modern Russian this form does not distinguish aspect and is inherently perfective.18 This accounts for the temporal reference of the future imperfective: Just as with other non-past verbs, the perfective morphology of bud- triggers the future interpretation. This is also true for other constructions in which bud- occurs. As noted in chapter 2, this
18
But see chapter 2 for the discussion of the imperfective interpretation of byt’.
111 form is used as a copula, as in (54a), and as an auxiliary in passive constructions, such as (54b). Such sentences always receive the future tense interpretation:
(54) a. Liza bud-et
vračom / krasivoj / v sadu.
Liza be-3RD.SG(PERF) doctor / pretty / in garden ‘Liza will be a doctor / pretty / in the garden.’ b. Deti
bud-ut
nakormleny (njanjej).
children be-3RD.PL(PERF) fed
nanny-INSTR
‘The children will be fed (by the nanny).’
In the next section, I extend the analysis proposed for the simple tense forms to the future imperfective.
3.4.2. Featuring the Future Imperfective In light of the analysis proposed here, the future imperfective presents an interesting case. It appears to be a combination of the past and the non-past simple tenses. On the one hand, as in the non-past, the future temporal reference is determined by the perfective aspect (of the auxiliary). On the other hand, as in the past, tense and aspect are expressed by different morphological forms. Let us begin from the beginning, that is, at the bottom. I assume that infinitival verbs can be used to make an assertion in the sense of Klein (1995), as long as they appear in a finite structure. Thus, infinitival verbs in future imperfective constructions
112 come from the lexicon with the uninterpretable valued feature [+P]. However, they are not specified as either past or non-past. Asp merges into the structure with the uninterpretable unvalued feature [P] and the interpretable unvalued feature [COIN]. The agree relation is established between the two features, just as in the other non-past and the past tense clauses, discussed above. As a result, the valued feature [+COIN] on the infinitival verb becomes interpretable on Asp as imperfective:
(55) a. Step 1: Merge of Asp AspP Asp` Asp u[P] i[COIN]
VP V` V u[–P]
b. Step 2: Agree between Asp and V AspP Asp` Asp u[–P] i[COIN]
VP V` V u[–P]
113 c. Step 3: Licensing of the coincidence relation: imperfective AspP Asp` Asp u[–P]
VP V`
i[+COIN: IMPF]
V u[–P] So far, the structural representation of the future imperfective is the same as the representation of other tenses. The only difference is that in this case the aspect is always imperfective. The next step in the derivation is the merge of T with the uninterpretable unvalued feature [PAST] and the interpretable unvalued feature [COIN]. I propose that the value of the formal feature [PAST] in this case is supplied by the morphology of the auxiliary. I propose that the auxiliary bud- comes from the lexicon with uninterpretable valued features [–PAST], [+P]:
114 (56) Step 4: Merge of the auxiliary TP T` T budu[–PAST; +P]
AspP Asp`
i[–COIN: FUT]
Asp u[–P]
VP V`
i[+COIN: IMPF]
V u[–P] At first glance, this structure appears to be different from the structures discussed so far in that aspectual features appear on T (and on V and Asp). However, all it means is that the aspectual morphology is interpreted on T, which is exactly the case with the simple non-past tense. The only difference is that in the simple present or future, the lexical item carrying the aspectual morphology merges below the AspP projection, and the aspectual morphology is interpreted on Asp first, and then on T, by virtue of value recycling. In (56), a second lexical item carrying the aspectual morphology (the auxiliary) merges above the AspP projection, in T. As a result, the aspectual morphology is interpreted only on T. I assume that being a syntactic operation, agreement can be driven by valuation, but not by interpretability, since only valuation is relevant for syntax. If the auxiliary budmerges in T it cannot enter into an agree relation with Asp because the feature [+P] (and [–PAST]) on bud- is valued in the lexicon. Since it also carries the valued feature [–PAST]
115 the specification of T is precise enough to license the relation of non-coincidence, which forms the basis for the future tense interpretation. This explains why bud- in the future imperfective does not contribute to the aspectual interpretation, even though it is morphologically perfective: Its formal feature [+P] is never interpreted on Asp. To sum up, in the future imperfective tense and aspect are both interpretable (and interpreted) independently of one another. Moreover, the two instances of the interpretable semantic feature [COIN] have different values: [–COIN] on T and [+COIN] on Asp. It was proposed that the value of [–COIN] on T is supplied by the auxiliary bud-, which comes from the lexicon with the valued uninterpretable features [–PAST], [+P] and merges in T. The value of [–P] on Asp is supplied by the aspectual morphology on the imperfective infinitive in V, via agreement between Asp and V, just as in the past and the non-past simple tenses.
3.5. CONCLUSION In this chapter, I proposed an account of the temporal interpretation of Russian finite indicative clauses. I showed that the temporal interpretation of Russian non-past clauses is supplied by aspect, whereas the temporal interpretation of past clauses arises independently of aspect. I proposed that tense and aspect each are represented in syntax by two features, [PAST] and [P]. These are purely formal features, which license the temporal arguments of tense and aspect. The temporal reference of a clause can be supplied by aspect, as in the non-past, or by an independent morpheme, as in the past (null head) and the future imperfective (auxiliary bud-). The analysis of the Russian tense
116 system proposed in this chapter is based on three elements. First, I adopt Klein’s (1995) time-relational analysis of Russian tense and aspect, according to which both tense and aspect order two times. Second, following Zagona (1990, 1995), Stowell (1996), Demirdache and Uribe-Etxebarria (2000), I assume that tense and aspect are dyadic predicates, which take times as arguments. The account developed in this chapter relies on the presence or absence of the agree relation between instances of the same feature on T and Asp. I proposed that the heads T and Asp contain an interpretable unvalued feature [±COIN]. The value and the interpretation of this feature depends on the agree relation between T and Asp. In the simple non-past, there is agreement between T and Asp, and as a result the temporal reference of simple non-past clauses is determined by aspect. In the past and future imperfective there is no agreement between T and Asp and the temporal and aspectual interpretation of such clauses are determined independently of one another. The analysis developed in this chapter has implications for the issue of interpretability. In particular, it raises a more general question of where exactly any particular morpheme can be interpreted, and what determines the circumstances under which a given morpheme can receive interpretation in more than one place. A particular issue I have in mind concerns Russian prefixation. The grammatical status of Russian perfectivizing prefixes has been the subject of a long-standing debate. Russian prefixes have mixed derivational and inflectional properties: they affect the lexical meaning and the argument structure of a verb, and they affect its aspectual and temporal properties. Most analyses that I am aware of treat Russian prefixes are either lexical morphemes
117 (Spencer and Zaretskaya 1998; Filip 2000), or as inflectional morphemes (Binnick 1991). In both cases, either the inflectional or the lexical properties remain unaccounted for. In chapter 6, I return to this issue and show that my analysis provides a new way of analyzing Russian prefixes. In the next chapter, I extend the time-relational analysis to the third category, mood. I propose that mood can be given the same analysis as tense and aspect: It can be viewed as a dyadic predicate that orders two times. I propose that just like T and Asp, C contains a purely formal feature which licenses the arguments of mood in syntax. This feature licenses a range of relations, which can be characterized as (non)coincidence. The formal feature on C licenses the relation of (non)coincidence by supplying the value of the interpretable feature [±COIN]. The analysis developed in the next chapter allows for a unified syntactic treatment of tense, aspect, and mood. In particular, it sheds light on the fact that crosslinguistically, the irrealis mood is often expressed by the past tense morphology (James 1982; Givón 1994; Dahl 1997; Tynan and Lavín 1997; Iatridou and references there).
118 CHAPTER 4 VALUE RECYCLING AND SUBJUNCTIVE MOOD
4.0. INTRODUCTION In this chapter, I argue that the analysis of tense and aspect developed in chapter 3 can be extended to mood. The focus on this chapter is primarily on the irrealis mood. In particular, I am concerned with the subjunctive interpretation of Russian l-participles, which in the realis mood convey a past temporal reference (see chapter 3). Russian does not have a subjunctive verb form. Matrix subjunctive clauses are formed by the morphological past tense verb form (the l-participle) and the conditional particle by, as in (1a). Embedded subjunctive clauses are formed by the l-participle and the subjunctive complementizer čtoby, as in (1b):
(1)
a. Ja by I
propusti-la etot doklad.
COND
miss-PAST
this talk
‘I would skip / would have skipped this talk.’ b. Ja xoču, čtoby vy I want
SUBJ
ostalis’.
you stay-PAST
‘I want you to stay.’1
1
I gloss the –l suffix as PAST because I treat the l-participle as the same form regardless of its syntactic environment and interpretation.
119 It has been suggested that mood relates (or orders) two worlds, just as tense and aspect order times (Chung and Timberlake 1985; Iatridou 2000). Taking this idea as my point of departure, I assume that mood is a dyadic predicate that takes two arguments. However, I propose that, like tense and aspect, the syntactic category mood relates two times: the evaluation time T-EVL (external argument) and the utterance time TU (internal argument).2 I define the evaluation time as time relative to which the situation described by the utterance is evaluated. As with tense and aspect, the relation expressed by mood can be characterized as the one of (non)coincidence. In realis mood, T-EVL coincides with TU: the evaluation time is always the utterance time, that is, the situation is evaluated relative to the utterance time. In irrealis mood, T-EVL does not coincide with TU: the evaluation time is not the utterance time, that is, the situation is not evaluated relative to the utterance time, but to some other time. My proposal is based on the traditional view of mood as the category that expresses a relation between worlds. I assume that the actual world and all possible alternative worlds have time, which can be represented as a continuum, or a time line. Various events can be associated with points or intervals along the time line in the actual worlds, as well as in alternative worlds. I assume also that the utterance time TU is always located on the actual world time line, whereas the situation time T-SIT can be located either on the actual world time line, or on an alternative world time line. I show that my theory captures the fact that the irrealis mood is expressed by past tense morphology in Russian, a phenomenon widely attested crosslinguistically (James 1982; 2
A similar idea is expressed in Timberlake (1985:163), who suggests that worlds have properties similar to
120 Givón 1994; Dahl 1997; Tynan and Lavín 1997; Iatridou 2000 and references there). At the end of this chapter, I discuss the implications of my analysis for the treatment of the realis mood. This chapter is organized as follows. In section 4.1, I propose a theory of mood as a dyadic predicate that orders times. In section 4.2, I show that this theory enables us to account for matrix subjunctive clauses in Russian. Section 4.3 provides an extension of this analysis to embedded subjunctive clauses. Section 4.4 deals with the implications of the theory for realis mood. Section 4.5 is a conclusion.
4.1. MOOD AS A DYADIC PREDICATE THAT RELATES TIMES In this section, I develop a syntactic analysis of mood as a dyadic predicate that relates two times. In 4.1.1, I argue that mood relates the evaluation time T-EVL and the utterance time TU, retaining the insight that mood characterizes an event by comparing two worlds. Section 4.1.2 deals with the featural makeup of mood: the purely formal feature that licenses its syntactic time denoting arguments and the semantic interpretable feature which serves as the basis for a semantic interpretation.
4.1.1. Mood Relates T-EVL and TU Following Iatridou’s (2000) semantic account, I analyze mood as a dyadic predicate that relates two arguments in syntax. However, I depart from her analysis in that I argue that in syntax, mood relates two times and not two worlds. I propose that mood relates the those of time. However, he does not develop this idea any further.
121 evaluation time T-EVL (external argument) and the utterance time TU (internal argument). Further, the relation expressed by mood can be characterized as the relation of (non)coincidence between T-EVL and TU. As a starting point, let me discuss the more traditional view of mood as a world-ordering category. According to various studies, mood characterizes an event denoted by the propositional content of a clause by comparing the world of this event to the actual world (Chung and Timberlake 1985; Palmer 1986; Bybee, Perkins, and Pagliuca 1994; Iatridou 2000). Let us suppose that the actual world is the world in which the utterance is made. Then, in realis mood, the event world is identical to the actual world. In other words, the event occurs in the same world in which the utterance is made. In contrast, in irrealis mood, the event world is not identical to the actual world. Mood differs from tense in that it is not a deictic category. In other words, mood does not place the situation time relative to the utterance time. Mood also differs from aspect in that it is not concerned with the internal structure of a situation. For example, whether a given situation is presented as ongoing or as completed is unrelated to whether this situation takes place in the actual world or in an alternative world. As Chung and Timberlake (1985) point out, the internal complexity of mood stems from the fact that while there is only one way for an event to be actual, or to take place in the actual world, there are many ways in which an event can be “less than completely actual”. Different types of non-actuality include hypothetical, conditional, subjunctive, etc. In this sense, the categories tense and aspect are relatively simple:
122 typically, we do not consider an event to be “somewhat in the past” or “not quite perfective or progressive”. Despite these differences, mood and tense share a number of properties which suggest that the two categories are closely related. For example, Chung and Timberlake (1985) observe that as far as verbal morphology is concerned, languages commonly make a two-way distinction between realis and irrealis mood. The realis mood is equivalent to indicative, whereas the irrealis mood is equivalent to conditional, hypothetical, subjunctive, etc. In this respect, mood patterns with tense, which is often represented in language as a two-way distinction between Past and Non-Past or Future and Non-Future. Another respect in which mood can be related to tense is the following. To the best of my knowledge, most languages do not have any morphology, verbal or otherwise, to encode realis (indicative) mood.3 In fact, it is typical for tensed languages to “equate” indicative mood with tense: If a clause has a temporal interpretation it is in the indicative mood; and conversely, an indicative clause must receive a temporal interpretation. Furthermore, many languages do not have a particular morpheme for irrealis mood either. Instead, as was mentioned earlier, it is quite common crosslinguistically to convey the irrealis mood by past tense morphology (James 1982; Givón 1994; Dahl 1997; Tynan and Lavín 1997; Iatridou 2000 and references cited therein). I take these observations to suggest that mood and tense may have more in common than was previously thought. In particular, I propose that both categories can be represented in terms of time. Following Comrie (1985) among others, I assume that time 3
This is true at least for matrix clauses. Embedded clauses are discussed below.
123 can be represented as a straight line, with the present moment represented as “0”, and the past and the future placed to the left and to the right of the present moment, respectively.4 Let us call this straight line a time line: (2) PAST
0
FUTURE
Furthermore, I assume that time is a property not only of the actual world, but of worlds in general. This is supported by the fact that we can distinguish between a hypothetical situation in the past and a hypothetical situation in the present or immediate future. The examples from (Iatridou 2000:231) illustrate this point: (3a) conveys that I do not have a car at the time of utterance (now), whereas (3b) conveys that I did not have a car in the past, and says nothing about whether or not I have a car now. The two meanings are expressed by the simple past and the past perfect, respectively:
(3)
a. I wish I had a car. b. I wish I had had a car when I was a student.
Given this, I assume that all worlds have time. Time in any world can be represented with a time line of the type shown in (2), and various events can be associated with various points or intervals along this time line. For example, the time line of the actual world may contain the utterance time and the situation time. The tense
4
Of course, the present is not a moment in time, but rather an interval, whose length depends on a particular context. However, for the purposes of this discussion this representation will do.
124 morphology conveys the relative ordering of the situation time and the utterance time, that is, whether the situation time is located on the time line to the left (past), to the right (future), or coincides with the utterance time (present). For my purposes here, I assume that utterances are typically made in the actual world, and therefore the utterance time TU is always located on the actual world time line.5 In contrast, alternative worlds may contain situations similar to those that occur in the actual world. Therefore, the situation time T-SIT can be located on the actual world time line or on an alternative world time line. This is illustrated in (4): the solid lines in (4a–b) represent the actual world time line and the dotted lines in (4b) represent the time lines for alternative worlds. In realis mood, the utterance time is interpreted as the evaluation time: the situation is evaluated relative to the utterance time. This is only possible if the situation time T-SIT and the utterance time TU are located along the same time line, as shown in (4a). Since, by assumption, the utterance time is always located on the actual world time line, a situation can be evaluated relative to the utterance time only if this situation takes place in the actual world.6 In irrealis mood, the utterance time is interpreted as not being the evaluation time: the situation cannot be evaluated relative to the utterance time. This happens when the situation time T-SIT is not located on the actual world time line, but on an alternative world time line, as shown in (4b). As discussed above, we still can distinguish between non-actual situations located at different distances from the utterance time in the actual 5
Here I am not making any claims regarding whether or not utterances can be made in alternative worlds.
125 world. I assume, therefore, that non-actual situations are evaluated relative to some nonactual evaluation time T-EVL located on the same “longitude” as the utterance time TU. This is captured by the representation of the situations described in (4bi) and (4bii) on the time lines of ALT-W (i) and ALT-W (ii), respectively. However, this is not what we call a temporal interpretation. A clause can have a temporal interpretation only if it describes a situation whose time is located on the same time line as the utterance time:
(4) a. Realis:
I had a car.
ACT-W T-SIT
T-EVL=TU
b. Irrealis: (i) I wish I had a car. (ii) I wish I had had a car (when I was a student). ACT-W TU
ALT-W (i) - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - T-EVL =T-SIT
ACT-W (ii) - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - T-SIT
T-EVL
To sum up the discussion so far, tense locates the situation time to the right or to the left from, or within the same interval as the utterance time on the actual world time 6
I assume that there is an infinite number of alternative worlds. I represent alternative worlds as parallel to the actual world to capture the fact that events in alternative worlds can be ordered relative to the utterance
126 line. Further, we have multiple time lines that represent time in multiple worlds. However, only the actual world time line contains the utterance time. Thus, before tense is able to locate a given situation time relative to the utterance time, it needs to be established whether the situation time in question and the utterance time are located on the same time line to begin with. I propose that this is exactly what we call mood. The situation world is compared with the utterance world via comparing the time lines. This proposal retains the basic idea that mood expresses a relation between worlds. In realis mood, the time line along which the situation time appears is the same as the time line along which the utterance time appears, that is, the actual world time line. As a result, the utterance is evaluated relative to the utterance time, and tense can locate the situation time relative to the utterance time. Only in realis mood can we have a true temporal interpretation. In irrealis mood, the situation time and the utterance time appear along different time lines. As a result, the utterance is not evaluated relative to the utterance time; tense cannot locate the situation relative to the utterance time and we have no temporal interpretation. In the next section, I discuss the features that make up the category mood.
4.1.2. The Featural Makeup of Mood I assume that the information about mood is encoded in the CP layer (Polletto 1995; Rizzi 1997, 2004). There is no empirical evidence to motivate multiple functional projections within the CP layer in Russian. Therefore, I assume that C carries the relevant time in the actual world.
127 mood features. In defining the featural content of C, I follow the line of reasoning in Iatridou’s (2000) analysis of counterfactual constructions, such as in (5) below. One of the issues addressed by Iatridou is the past tense morphology that receives a non-past interpretation. Iatridou refers to this phenomenon as “fake tense” or “fake past morphology”:
(5)
a. If Fred was drunk, he would be louder. b. If Mary knew the answer, she would be the only one.
(Iatridou 2000:244)
The sentences in (5) contain the past tense forms was, knew, and would. However, they convey that the speaker believes that Fred is not drunk now and that Mary does not know the answer now. In other words, the past tense morphology does not have the past tense interpretation. Iatridou (2000) argues that the past tense morpheme always has the same meaning, but the domain it operates on varies according to the environment. The past tense form has a certain meaning σ such that in certain environments E1 it receives the past interpretation and in certain environments E2 it receives a different interpretation. She proposes that what we call the “past tense morpheme” is a feature, which provides a skeletal meaning of the following form:
(6)
T(x) excludes C(x)
128 T(x) stands for “Topic(x)”, that is, “the x that we are talking about”. C(x) stands for “the x that for all we know is the x of the speaker”. The term topic is borrowed from Klein (1994), who introduces the notion Topic Time. Iatridou (2000) uses the term exclusion feature to refer to the variable x and proposes that this variable can range over times and worlds. When it ranges over times we refer to the result as “past tense”, as in (7). When it ranges over worlds, we refer to the result as “irrealis”, as in (8):
(7)
a. T(t): the time interval (set of times) that we are talking about . b. C(t): the time interval (set of times) that for all we know is the time of the speaker (Klein’s 1994 Utterance Time). c. Past tense: The topic time excludes the utterance time.7
(8)
a. T(w): the worlds that we are talking about (topic worlds) b. C(w): the words that for all we know are the worlds of the speaker (actual world) c. Irrealis mood: The topic world excludes the actual world.
I adopt the following components of Iatridou’s (2000) proposal. First, the relationship between times and worlds is determined by the same feature. Second, this feature is defined as exclusion, which can also be represented as a feature with the negative value: [–X]. Finally, the past tense morpheme has a certain meaning σ such that
7
Iatridou (2000) treats tense as only past or present, and the future woll as modal, as do Palmer (1986) and Kamp and Reyle (1993) among many others.
129 in certain environments E1 it receives the past interpretation and in certain environments E2 it receives a different interpretation. With this in mind, I propose that the interpretable feature on C that serves as the basis for the semantic interpretation of the ordering relation between the external argument T-EVL and the internal argument TU is [COIN]. [–COIN] is interpreted as irrealis (T-EVL and TU do not coincide, that is, the utterance time is not the evaluation time). [+COIN] is interpreted as realis (T-EVL and TU coincide, that is, the utterance time is the evaluation time).8 Here I am mainly concerned with [–COIN], which receives either past or irrealis interpretation, depending on the functional head:
(9)
CP T-EVL
C C i[±COIN]
TP T`
TU
T u[±PAST] i [±COIN]
AspP…
What is missing in this structure is the formal feature that licenses the syntactic arguments of mood and supplies the value of the semantic feature [COIN]. Following Rizzi (1997), I propose that this feature on C is finiteness, [FIN].9 Following Binnick
8 9
In 4.4 I propose that unlike irrealis mood, realis mood does not have to be represented as a formal feature.
In Cowper’s analysis, [FIN] is a feature in Infl. Here I do not explore this issue in detail, but I return to this idea briefly in section 4.4, when I discuss the implications of my analysis for realis mood.
130 (1991:69ff), Kayne (1994), Rizzi (1997) and Cowper (2002, 2005) among others, I assume that [+FIN] entails that the clause is finite, that is, it has a nominative subject, and the main verb agrees with this subject in phi-features; accordingly, [–FIN] entails that the clause does not have a nominative subject, and the main verb does not show agreement. 10 I also assume that non-finite clauses cannot pick out a specific situation in the actual world which can be located relative to the utterance time. Therefore, non-finite clauses do not have a temporal reference. As shown in (10) below, [–FIN] on C always licenses [–COIN]: the evaluation time and the utterance time do not coincide.11 In contrast, [+FIN] can license either [–COIN] or [+COIN]: finite clauses may or may not have a temporal reference, as they may or may not pick out a specific situation in the actual world located relative to the utterance time. Thus, in finite clauses, the evaluation time may or may not coincide with the utterance time. In this chapter, I focus on the shaded area in (10), that is, the cases when [+FIN] licenses [–COIN]:
10
Cowper (2002) argues that agreeing infinitives in languages such as Portuguese are pseudo-finite: they acquire the superficial properties of finiteness during the course of the syntactic computation. 11
An example of a non-finite irrealis clause is a matrix infinitival clause with a dative subject, such as in i:
i. Gruzovik-am ne projexa-t’. trucks-DAT NEG ride-through-INF ‘It is not in the cards for the trucks to get through.’
(Moore and Perlmutter 1999)
Such clauses convey that the event denoted by their propositional content is beyond the subject’s control. Although here I do not discuss such constructions, these data are compatible with my analysis. If only finite clauses can pick out a specific event that can be located on the actual world time line relative to the utterance time, non-finite clauses are always interpreted as irrealis. Thus, [–FIN] can only license the relation of non-coincidence by supplying the value of [–COIN] on C.
131 (10)
Formal feature
[–FIN]
[+FIN]
Licensed relation
[–COIN]
[+COIN]
[–COIN]
Semantic interpretation
‘irrealis’
‘realis’
‘irrealis’
The completed structure is now given in (11):
(11)
CP T-EVL
C C u[±FIN] i[±COIN]
TP T`
TU
T u[±PAST] i [±COIN]
AspP…
In the next section, I move to the question of how the value of the feature [FIN] on C is supplied. Under the approach adopted in this dissertation, the value of formal features is supplied by overt morphology. Given that Russian, like many other languages, does not have verbal morphology to indicate mood, this issue is intriguing. I focus on the subjunctive mood in Russian, which is expressed by past tense morphology and demonstrate that the theory of mood developed here allows for an account of Russian subjunctives. I discuss the implications of this analysis for realis mood in section 4.4.
132 4.2. THE SUBJUNCTIVE MOOD IN RUSSIAN 4.2.1. Matrix Subjunctive Clauses Russian does not have a distinct morphological verb form for the subjunctive mood. The subjunctive mood is formed by the l-participle and the conditional particle by, as shown in (12):
(12) a. Ja by I
COND
propusti-la etot doklad. miss-PAST
this talk
‘I would skip / would have skipped this talk.’ b. Ja by I
COND
uš-la
domoj.
leave-PAST home
‘I would go/have gone home.’
According to Hacking (1998:27), the conditional particle has a status of both the marker of the morphological subjunctive mood and the lexical marker of modality. However, if by alone conveys the subjunctive mood, it is not clear why it is incompatible with other verb forms, as illustrated in (13):
(13) a. *Ja by I
COND
propuskaj-u
etot doklad.
miss-1ST.SG(IMPF/PRES) this talk
‘I would skip this talk.’
133 b. *Ja by I
COND
ujd-u
domoj.
leave-1ST.SG(PERF/FUT) home
‘I would go home.’
These examples suggest that it is the combination of the conditional particle by and the lparticiple that has the meaning subjunctive. I return to this issue in section 4.2.3. At this point, I want to address the question of what syntactic position the particle by occupies. At first glance, it may seem that by occupies T. In (12), by immediately follows the nominative subject and precedes the finite verb. Following the traditional assumption that nominative subjects occupy SpecTP, it is reasonable to assume that in (12) by appears in T (Brecht 1977; Progovac 1993). Furthermore, historically, by developed out of the Old Russian auxiliary byti in the aorist tense; more specifically, it developed out of the second/third person singular form of byti in the aorist, by. As the auxiliaries typically occur in T it is possible that by occupies its historical position. However, Willis (2000) provides evidence that in Old Russian, the auxiliary byti in the subjunctive/conditional mood underwent obligatory T-to-C movement in most contexts. He argues that the obligatory conditional verb movement created the conditions for a change whereby C became a basic, rather than a derived position for the conditional auxiliary. According to Willis, the second/third person singular form by underwent a change from being an inflectional element that moves to C to an uninflected particle that originates in C.
134 There is evidence that supports Willis’ (2000) claim that in Modern Russian by appears in C. For example, there is evidence from wh-movement that suggests that in (12) by occupies C and that the nominative subject ja ‘I’ undergoes movement to SpecCP. In (14) below, by again appears in the second position, but this time it immediately follows the wh-phrases čto ‘what’ in (14a) and kuda ‘where’ in (14b):
(14) a. Čto what
by
ty
propusti-la t?
COND
you-NOM miss-PAST
‘What (e.g., which talk) would you skip?’ b. Kuda where
by
ty
uš-la ?
COND
you-NOM go-PAST t
‘Where would you go to?’
Assuming that the wh-phrases occupy SpecCP and the nominative subjects occupy SpecTP, by in (14) appears in C, not in T. Note that although the second position is the most natural for by, it can appear anywhere in the clause. The only position that by cannot appear is clause-initially; thus, it cannot precede the nominative subject in indicative clauses:
(15) a. Ja uš-la I leave-PAST
by
domoj.
COND
home
135 b. Ja uš-la
domoj by.
I leave-PAST home
COND
‘I would leave/have left.’ c. *By COND
ja uš-la
domoj.
I leave-PAST home
‘I would leave/have left.’
I am going to assume that in Modern Russian by has undergone a change from an inflected auxiliary to an uninflected clitic and as such must be attached to another element.12 If by appears in C, some other element must undergo movement to SpecCP or to C, so that by would not be stranded. This would explain the Subject-by-Verb order in (12), and the WH-by-Subject order in (14). The assumption that by is a clitic becomes important in section 4.3, where I discuss mood in embedded clauses. To sum up the discussion so far, Russian matrix subjunctive clauses are formed by the conditional particle by and the morphological past tense verb form (the lparticiple). The subjunctive examples throughout the section suggest that the conditional particle appears in C. Before I present my analysis in section 4.2.3, I demonstrate that past tense morphology in the subjunctive mood does not receive the same past tense interpretation as in the indicative mood.
12
The variation in placement of by can be analyzed in a number of ways. One way is to suggest that this is due to the fact that Russian is a free word order language. However, the fact that by cannot appear sentence initially suggests that this freedom is not due to the free word order. Another possibility is that by always occupies C, and that the elements that precede it undergo movement to adjoin to it. At this point, I leave this question open.
136 4.2.2. Temporal Interpretation of Matrix Subjunctives Unlike indicative clauses, subjunctive clauses in Russian do not have an unambiguous temporal reference. The morphological past tense verb form, or the l-participle, in subjunctive clauses does not receive a fixed past tense interpretation. This can be demonstrated by the compatibility of indicative and subjunctive clauses with time adverbials. The indicative clause in (16a) states that the event of leaving occurred prior to the utterance time. As a result, it is only compatible with a past time adverbial; the use of present or future time adverbials renders the sentence ungrammatical. In contrast, the subjunctive sentence in (16b) is compatible with past, present, and future adverbials:
(16) a. Ja ujexa-la
včera / *sejčas / *zavtra.
I leave-PAST yesterday / now / tomorrow ‘I left yesterday / *now / *tomorrow.’ b. Ja by I
COND
ujexa-la
včera
/sejčas /zavtra.
leave-PAST yesterday / now / tomorrow
‘I would have left yesterday / leave now / leave tomorrow.’
The indicative clause in (16a) and the subjunctive clause in (16b) have the same propositional content . However, they differ in how they present the event of leaving. The indicative (16a) picks out a particular leaving event in the actual world. Thus, the situation time and the utterance time are located along the same actual world
137 time line. Therefore, the past morphology on the verb (the suffix –l) is interpreted as past. As a result, (16a) receives a past tense interpretation. In contrast, the subjunctive (16b) does not pick out any particular leaving event; rather, it expresses the speaker’s attitude towards a hypothetical leaving event. In this particular case, the subjunctive clause expresses the speaker’s wish to leave, or a hypothetical possibility of leaving. Thus, the situation time and the utterance time are located along distinct time lines: some alternative world time line and the actual world time line, respectively. Therefore, the situation time cannot be located relative to the utterance time, and the clause does not receive a past tense interpretation. Note that the English translation of (16b) reflects the temporal contrast between a hypothetical event of leaving that could have occurred at some point prior to the utterance time and the hypothetical event of leaving that could still take place at some point coinciding with the utterance time: I would have left versus I would leave. This suggests that irrealis clauses are not tenseless in the same way as infinitival clauses. Unlike indicative clauses, subjunctive clauses do not denote actual events, but rather present events denoted by their propositional content as hypothetically possible at some point in time relative to the time of utterance. In Russian, the temporal placement of this hypothetical possibility, that is, at what point in time relative to the utterance time this hypothetical possibility exists is not encoded in the structure or in the verb form. Rather, it is determined by external factors, such as the pragmatic context, the preceding discourse, etc.
138 These data suggest that subjunctive mood is in complementary distribution with tense: Whenever the conditional particle by is present the clause cannot receive a temporal interpretation.13 Instead, it is interpreted as irrealis mood. In the next section, I propose that the interpretation irrealis arises when the relevant “tense” feature is interpreted on C.
4.2.3. Agreement and Value Recycling Following Ackerman and Webelhuth (1998), the meaning of morphosyntactic constructions is not necessarily derived compositionally from the meaning of its parts. For example, they claim that in a construction such as John has left the property “Perfect” cannot be identified with either the auxiliary has or the participle left. Rather, it is the construction as a whole that conveys this meaning. Spencer (2001) extends this approach to Russian constructions with the l-participle. He proposes that in the absence of the conditional particle by the l-participle receives the past tense interpretation by default, just as English nouns in the absence of the plural marker receive the singular interpretation by default. Following Spencer (2001), I assume that the subjunctive interpretation does not arise compositionally from the meaning of the conditional particle 13
It should be mentioned that there are cases when the particle by co-occurs with a tensed form, as in the following example: i. Govorjat, budto (by) on pridet zavtra /prixodit každyj den’. say-3RD.PL as-if by he come( PERF) tomorrow/come(IMPF) every day ‘They say as if he will come tomorrow / comes every day.’ Such examples require an explanation under the present analysis. Note, however, that they differ from the examples discussed in this thesis in two respects. First, such clauses are not conditionals/subjunctives. Second, as indicated by parenthesis, the particle by is optional, and its presence or absence has no effect on the semantics of the clause. It is possible that a unified analysis may be developed to include such constructions. I leave this issue for further research.
139 by and the meaning of the l-participle, but rather from the combination [by + l-participle]. This explains why the subjunctive interpretation cannot be obtained if by is combined with any other verb form. The ungrammatical examples are repeated below:
(17) a. *Ja by I
COND
propuskaj-u
etot doklad.
miss-1ST.SG(IMPF/PRES) this talk
‘I would skip this talk.’ b. *Ja by I
COND
ujd-u
domoj.
leave-1ST.SG(PERF/FUT) home
‘I would go home.’
Note, however, that Spencer’s (2001) approach does not explain the connection between the subjunctive/conditional meaning and the past tense meaning. It is not clear why past is the default interpretation of a form that otherwise is a part of the subjunctive form. Below, I propose an analysis that captures this connection. Let us begin from the bottom of the structure, that is, from Merge of Asp with the unvalued formal feature [P]. In (18), I outline the steps 1–3 of the derivation:
140 (18) a. Step 1: Merge of Asp AspP Asp` Asp u[P] i[COIN]
VP V` V u[+PAST] u[±P]
b. Step 2: Agree between Asp and V AspP Asp` Asp u[±P] i[COIN]
VP V` V u[+PAST] u[±P]
c. Step 3: Licensing of the (non)coincidence relation: perfective/imperfective AspP Asp` Asp u[±P]
VP V`
i[±COIN: IMPF/PERF]
V u[+PAST] u[±P]
141 In (18a), the head Asp merges in with the uninterpretable formal feature [P] and the interpretable feature [COIN], both unvalued. The unvalued feature [P] scans its ccommanding domain and finds a valued instance of the same feature on V. The agreement takes place and the formal feature [P] on Asp becomes valued. As a result of the agree relation in (18b), a link is established between the two instances of the same feature. Being valued, the uninterpretable formal feature [±P] becomes interpretable by licensing the relation of (non)coincidence: It supplies the value of the interpretable unvalued feature [COIN] on Asp. As a result, either the perfective or the imperfective interpretation arises, depending on the value of [P]. The examples below demonstrate that although subjunctive clauses do not receive the “indicative” type temporal interpretation, they receive the aspectual interpretation:
(19) a. Ja by I
COND
uš-la
vovremja.
leave-PAST(PERF) on-time
‘I would leave/have left on time (on a particular occasion)’ b. Ja by I
COND
uxodi-la
vovremja.
leave-PAST(IMPF) on-time
‘I would leave/have left on time (usually, as a rule).’
Next, in step 4 in the derivation, T merges into the structure with the uninterpretable unvalued feature [PAST] and the interpretable unvalued feature [COIN]. In step 5, the formal feature on T enters into an agree relation with its valued counterpart on
142 V. In step 6, the feature [–PAST] on T supplies the value of the semantic feature [COIN]. The three steps are illustrated in (20): (20) a. Step 4: Merge of T TP T` T
AspP
u[PAST] i[COIN]
Asp` Asp u[±P]
VP V`
i[±COIN: IMPF/PERF]
V u[+PAST] u[±P] b. Step 5: Agree between T and V TP T` T u[+PAST] i[COIN]
AspP Asp` Asp u[±P]
VP V`
i[±COIN: IMPF/PERF]
V u[+PAST] u[±P]
143 c. Step 6: Licensing of the non-coincidence relation: past TP T` T u[+PAST]
AspP
i[–COIN: PAST]
Asp` Asp u[±P]
VP V`
i[±COIN: IMPF/PERF]
V u[+PAST] u[±P] At this point, the derivation of a subjunctive clause is identical to that of an indicative clause with a past tense interpretation. The difference is that subjunctive clauses contain a CP projection headed by the conditional particle by. Following Willis (2000), I assume that in Modern Russian, the conditional particle by originates in C. I propose that by comes from the lexicon with uninterpretable valued feature [+FIN]. This proposal captures the fact that subjunctive clauses in Russian are always finite: They contain a nominative subject and the main verb shows agreement. Thus, the value of [FIN] on C is supplied by overt morphology. This is illustrated in (21):
144 (21)
Step 7: Merge of C CP C` C by u[+FIN] i[COIN]
TP T` T u[+PAST] i[–COIN: PAST]
AspP Asp` Asp u[±P]
VP…
i[±COIN: IMPF/PERF]
At this point, the uninterpretable feature [+FIN] on C should license the relation of (non)coincidence by supplying the value of the interpretable feature [COIN]. However, here we are dealing with the same problem as in non-past indicative clauses: The value ‘+’ is too general in this case. A finite clause may be either realis or irrealis. If [+FIN] licenses the relation of coincidence by supplying the positive value of [COIN], [+COIN] gives rise to a realis interpretation. As a result, the combination of by + l-participle is expected to receive an indicative interpretation, contrary to fact. Assuming that by itself carries the feature [+IRREALIS] or [+SUBJUNCTIVE] would again predict that verb forms other that the l-participle should be able to co-occur with by. Since [COIN] on C cannot receive its value directly from the valued formal feature [+FIN], it has to enter an agree relation by itself. It scans its c-commanding domain and finds an interpretable valued instance of [COIN] on T. Since the formal feature on T is [+PAST], the interpretable feature on T has the right value, [–COIN]. The value of the
145 formal feature [PAST] on T supplied by the past tense morphology (the suffix –l) is recycled to supply the value of the interpretable feature [COIN] on C, as illustrated in (22):
(22)
Step 8: Agree between C and T CP C` C by u[+FIN] i[–COIN: IRREALIS]
TP T` T
AspP
u[+PAST] i[–COIN: PAST]
Asp` Asp u[±P]
VP…
i[±COIN: IMPF/PERF]
As a result of this agreement, the feature [COIN] on C receives the value of the feature [COIN] on T: [–COIN]. However, since these features are located on different heads they give rise to different semantic interpretations. The feature [–COIN] on C receives an irrealis interpretation. Thus, the mechanism of feature valuation required to supply the value of the formal feature on C is essentially the same as the one that is required to supply the value of the formal features on T and Asp. Recall that according to Iatridou (2000), the domain on which the past tense morpheme operates varies according to the environment (see section 4.1.2). In irrealis mood, the domain on which the past tense morpheme operates is TP, as shown in (22).
146 The interpretable feature [–COIN] on C takes scope over the interpretable feature [–COIN] on T. Both C and T now contain an interpretable instance of the semantic feature [COIN]. This implies that both the irrealis mood and the past tense should be interpretable. I propose that tense in irrealis mood is indeed interpretable, although in this case the past tense morphology appears to be more flexible in its interpretation. This is possibly a result of tense appearing in the scope of mood. The temporal contrast between the two English conditionals repeated here as (23) support this view: (23) a. I wish I had a car. b. I wish I had had a car when I was a student.
My analysis captures the observation that in many languages without overt mood morphology, irrealis mood in finite clauses is conveyed by past tense morphology. Since only the combination of features [+FIN; –COIN] on C gives rise to an irrealis interpretation, the interpretable feature on T must be [–COIN]. Since many languages make the basic distinction between past and non-past, [–COIN] on T is supplied by past tense morphology. However, [–COIN] on T can also mean future, and the prediction is that we should be able to find future morphology in irrealis contexts. This prediction is borne out. For example, Hebrew uses past tense morphology in conditional clauses such as (24a), and future tense morphology in embedded subjunctives, such as (24b). As shown by the
147 ungrammaticality of (24c), the use of future tense morphology is obligatory in Hebrew subjunctives:
(24) a. ilu haiti
Rothschild haiti
gar be-America.
if was.1ST.SG Rothschild was.1ST.SG live in-America ‘If I were Rothschild, I would live in America.’ b. Dina roca
še
Gila taase
uga.
Dina wants that Gila will.make.3RD.SG.FEM cake ‘Dina wants Gila to make a cake.’ c. *Dina roca
še Gila asta
/ osa
uga.
Dina wants that Gila made.3RD.SG.FEM/ is making.SG.FEM cake ‘Dina wants Gila to make a cake.’
In contrast, the interpretable feature [+COIN] on C gives rise to a realis interpretation. Therefore, in irrealis clauses [COIN] cannot receive its value from [+COIN] on T, whose value is supplied by present tense morphology. To the best of my knowledge, no languages use present tense morphology to convey irrealis mood. Thus, the analysis proposed in this chapter is supported by empirical data. In the next section, I extend this analysis to embedded subjunctive clauses in Russian.
148 4.3. MOOD IN EMBEDDED CLAUSES 4.3.1. Embedded Subjunctive Clauses Unlike matrix subjunctive clauses, embedded subjunctive clauses are not introduced by the particle by alone. The conditional particle by is combined with the indicative complementizer čto ‘what’ to form the subjunctive complementizer, as shown in (25):
(25) a. Liza xotela, Liza wanted
čtoby Philemon uše-l. SUBJ
Philemon leave-PAST
‘Liza wanted Philemon to leave.’ b. Ja prošu, čtoby vy I ask
SUBJ
osta-lis’.
you stay-PAST
‘I am asking you to stay.’
I discuss the subjunctive complementizer in detail in section 4.3.1.2. First, however, in 4.3.1.1, I address another property of embedded subjunctives with respect to which they differ from their matrix counterparts. I demonstrate that the temporal interpretation of embedded subjunctives is determined by the matrix verb.
4.3.1.1. Temporal Interpretation of Embedded Subjunctives Embedded subjunctives typically denote a hypothetical situation in the future relative to the matrix event. The verbs that take a subjunctive complement include verbs of volition, prediction, etc. The embedded subjunctive clause in (26) is incompatible with the past
149 tense adverbial v prošlom godu ‘last year’, as in (26a), but is fine with the future tense adverbial v sledujuščem godu ‘next year’, as in (26b):
(26) a. *Liza trebujet, čtoby Philemon v prošlom godu prixodil Liza demands
SUBJ
Philemon in last
vo vremja.
year come-PART in time
‘Liza demands that Philemon show up on time last year.’ b. Liza trebujet, čtoby Philemon v sledujuščem godu prixodi-l Liza demands
SUBJ
Philemon in next
vo vremja.
year come-PART in time
‘Liza demands that Philemon show up on time next year.’
The embedded subjunctive clauses in (26) contrast with the embedded indicative clauses in (27), which contain the past tense verb form. Embedded indicative clauses are grammatical with the past tense adverbial v prošlom godu ‘last year’, but are incompatible with the future time adverbial v sledujuščem godu ‘next year’, as in (27b):
(27) a. Liza govorit, čto Philemon v prošlom godu prixodi-l Liza says
IND
Philemon in last year
vo vremja.
come-PAST in time
‘Liza says that Philemon showed up on time last year.’ b. *Liza govorit, čto Philemon v sledujuščem godu prixodi-l Liza says
IND
Philemon in next
vo vremja.
year come-PAST in time
‘Liza says that Philemon showed up on time next year.’
150 There are contexts in which embedded subjunctives appear to have an interpretation other than the hypothetical future. I argue that such an interpretation is also due to the lexical properties of the matrix verb. For example, the embedded subjunctives in (28) do not refer to a hypothetical event in the future. They make a statement about a permanent or temporary property of an individual, which is relevant to the utterance time:
(28) a. Ja somnevajus’, čtoby on by-l I doubt
SUBJ
na takoje sposoben.
he be-PAST on such capable
‘I doubt that he would be capable of such a thing.’ b. Philemon ljubit, čtoby ego česa-li Philemon loves
SUBJ
za
uxom.
him scratch-PAST behind ear
‘Philemon likes being scratched behind the ear (by anybody).’
In (28a), according to the speaker’s belief somebody is incapable of something in principle, not at any particular point in time. In (28b), the passion for being scratched behind the ear is a property of Philemon. Note that embedded subjunctives can also refer to a hypothetical or nonreferential event in the past. This is often the case in constructions with overt negation. In (29), the speaker denies that a certain event took place prior to the utterance time:
151 (29) Ja ne
pomnju,
čtoby ty
I
remember
SUBJ
NEG
mne govori-l
ob
etom.
you me say-PAST about this
‘I don’t recall you telling me about that.’
(Hacking 1998:28)
In such examples, the past interpretation of the subjunctive clause is not determined by the morphological past tense form, but rather by the semantics of the matrix verb. For example, one can only remember (or not remember) things that happen in the past. Recall that the temporal reference of matrix subjunctive clauses is determined by the pragmatic context or the preceding discourse. The temporal reference of the embedded subjunctive clauses is determined by the matrix clause. The generalization is that the temporal interpretation of subjunctive clauses is determined by the immediate environment, linguistic or otherwise. In embedded contexts, the immediate linguistic environment is the matrix clause.
4.3.1.2. The Subjunctive Complementizer As mentioned above, the subjunctive complementizer is formed by the conditional particle by and the indicative complementizer čto ‘what’. The discussion here is concerned with the following two questions: First, do these two elements form a single lexical item in the lexicon or do they originate separately and subsequently combine via some syntactic operation, such as cliticization? Second, what is the syntactic position of this element?
152 With respect to the first question, I argue, contra Brecht (1977) that čtoby should be analyzed as a single lexical item. Brecht (1977) argues that by originates in T and subsequently cliticizes on to čto. One problem with this view is that under no circumstances can another element ever appear between čto and the particle by when the combination of the two functions as the subjunctive complementizer. As was discussed earlier and as shown again in (30), when by appears alone in a matrix subjunctive clause, it most naturally occurs in the second position. However, it can in principle occur anywhere, as in (30a–c), except clause-initially, as in (30d) (Hacking 1998; Spencer 2001):
(30) a. Ty you b. Ty
by
uš-la
COND
leave-PAST home
uš-la
by
you leave-PAST c. Ty
domoj.
uš-la
COND
domoj. home
domoj by.
you leave-PAST home
COND
‘You would go/have gone home.’ d. *By COND
ty
uš-la
domoj.
you leave-PAST home
‘You would go/have gone home.’
153 Given that in (30) by enjoys such freedom, it is not clear why in embedded subjunctive clauses by must cliticize on to the indicative complementizer; nor is it clear why (30b–c) are grammatical, whereas (31b–c) are not:
(31) a. Ja xoču, čtoby ty I want
uš-la
domoj.
you leave-PAST home
SUBJ
‘I want you to go home.’ b. *Ja xoču, čto ty I want
IND
you
c. *Ja xoču, čto ty I want
IND
by
uš-la
COND
leave-PAST home
uš-la
domoj.
domoj by.
you leave-PART home
COND
Furthermore, in addition to introducing subjunctive clauses, čtoby also introduces adjunct purpose clauses, both finite and infinitival, as shown in (32a) and (32b), respectively:
(32) a. Anna zanjala
deneg, [ čtoby Phil kupi-l
Anna borrowed money
SUBJ
mašinu ].
Phil buy-PAST car
‘Anna borrowed money so that Phil could buy a car.’ b. [ Čtoby sdat’ SUBJ
eksamen], Liza dolžna zanimat’sja.
pass-INF exam
Liza must
‘In order to pass the test, Liza must study.
study
154 Neither čto nor by can introduce adjunct purpose clauses separately. In other words, čtoby imposes selectional restrictions on its complements, which differ from the selectional restrictions imposed by its components in isolation. This property is characteristic of lexical items and is unexpected if by is a clitic which cliticizes onto the indicative complementizer čto in syntax. Based on these properties of čtoby I conclude that it is a single lexical item. 14 More specifically, I propose that it has a compound-like structure, where the conditional particle by is the head, and čto is a semantically empty element, whose only purpose is to indicate that the clause in question is embedded.15 At this point, I do not discuss any further properties of čto. In chapter 5, I argue that even when čto appears alone as the indicative complementizer it is semantically empty as well. In the next section, I address the syntactic location of čtoby and outline the derivation of Russian embedded subjunctives.
4.3.1.3. Featuring Embedded Subjunctives I propose that my analysis of matrix subjunctive clauses can be extended to embedded subjunctives. I propose that the subjunctive complementizer čtoby comes from the lexicon with the uninterpretable valued feature [+FIN], by virtue of being headed by the particle by. The interpretation of embedded subjunctive clauses arises in the same way as
14
Among those who treat čtoby as a single lexical item are Avrutin and Babyonyshev (1997) and Spencer (2001). 15
This compound structure cannot be directly compared with Russian compounds, which have lexical meaning. However, this proposal is consistent with the fact that in “headed” (or “subordinated”) compounds in Russian the head follows its modifier (Švedova 1980, vol. 1:242–254).
155 the interpretation of matrix subjunctives. The difference is that in the embedded subjunctives the complex subjunctive complementizer čtoby merges in C, instead of the particle by alone. The last three steps of the derivation are given below:
(33) a. Step 4: Merge of T TP T` T u[PAST] i[COIN]
AspP Asp` Asp u[±P]
VP V`
i[±COIN: IMPF/PERF]
V u[+PAST] u[±P] b. Step 5: Agree between T and V TP T` T
AspP
u[+PAST] i[COIN]
Asp` Asp u[±P]
VP V`
i[±COIN: IMPF/PERF]
V u[+PAST] u[±P]
156 c. Step 6: Licensing of the non-coincidence relation: past TP T` T u[+PAST]
AspP
i[–COIN: PAST]
Asp` Asp u[±P]
VP V`
i[±COIN: IMPF/PERF]
V u[+PAST] u[±P] At this point, the subjunctive complementizer čtoby merges in C. It supplies the value of the formal feature [FIN]:
(34)
Step 7: Merge of C CP C` C čtoby u[+FIN] i[COIN]
TP T` T u[+PAST] i[–COIN: PAST]
AspP Asp` Asp u[±P] i[±COIN: IMPF/PERF]
VP…
157 The rest of the derivation proceeds as in matrix subjunctives. The uninterpretable feature [+FIN] on C cannot license the relation of (non)coincidence because it is too general to supply the value “–” of the interpretable feature [COIN]. Since [COIN] on C cannot receive its value directly from the valued formal feature [+FIN], it has to enter an agree relation by itself. It scans its c-commanding domain and finds an interpretable valued instance of [COIN] on T. Since the formal feature on T is [+PAST], the interpretable feature on T has the right value, [–COIN], as illustrated in (35):
(35)
Step 8: Agree between C and T CP C` C čtoby u[+FIN] i[–COIN: IRREALIS]
TP T` T u[+PAST] i[–COIN: PAST]
AspP Asp` Asp u[±P]
VP…
i[±COIN: IMPF/PERF]
The value of the formal feature [PAST] on T supplied by the past tense morphology (the suffix –l) is recycled to supply the value of the interpretable feature [COIN] on C.
158 4.3.2. Summary I have argued that the analysis of realis clauses developed in chapter 3 can be extended to account for the interpretation of irrealis clauses (Russian subjunctives). According to this analysis, irrealis mood is a dyadic predicate that relates time denoting arguments in syntax. These arguments are licensed by the formal feature [+FIN], whose value is supplied by the conditional particle by. The value “+” is too general to license a relation of (non)coincidence: finite clauses can be either realis (coincidence) or irrealis (noncoincidence). Therefore, the interpretable feature [COIN] on C receives its value via an agree relation with a valued instance of the same feature on T. Thus, the subjunctive interpretation arises when the value of [COIN] on C is supplied by past tense morphology, via agreement between C and T. This analysis has implications for realis mood. If in irrealis clauses, C contains the formal feature [FIN], whose value is supplied by the conditional particle by, the question is, what is the content of C in realis clauses? This issue is addressed in the next section.
4.4. IMPLICATIONS FOR REALIS MOOD To maintain the analogy between mood and tense, a reasonable suggestion is that in realis clauses, C merges in with the unvalued formal feature [FIN], as shown in (36):
159 (36) Step 7: Merge of C CP C` C u[FIN] i[COIN]
TP T` T u[±PAST] i[–COIN: PAST]
AspP Asp` Asp u[±P]
VP V`
i[±COIN: IMPF/PERF]
V u[±PAST] u[±P] The only combination of features that would give rise to a realis interpretation is [+FIN; +COIN]: a finite clause which expresses a relation of coincidence between the evaluation time and the utterance time. Since there is no verbal morphology that indicates realis mood, the formal feature [FIN] on C would scan its c-commanding domain in search of a valued feature, which could supply its value. Since no head in the structure contains a valued instance of [FIN] this feature on C will remain unvalued and the derivation will crash. This problem can be addressed in a number of ways. One possible solution would be to introduce the formal feature [FIN] on T, in addition to the feature [PAST]. This proposal gives rise to the same problem: there is no morpheme on the verb that indicates that this verb is finite, and as a result, this feature cannot receive its value. Unlike tense or
160 aspect, finiteness as defined here is a collection of properties. A clause is finite if it has all or some of these properties. For example, finite verbs agree with the nominative subject. Russian past tense verb forms agree in number and gender, whereas non-past tense forms agree person and number. This contrast, however, does not affect the status of a given verb as finite. Moreover, while it is reasonable to assume that verbs come from the lexicon with a valued feature that expresses tense or aspect, the assumption that verbs come from the lexicon with a valued feature that expresses finiteness is problematic. Finiteness is not an inherent property of a given verb. The verb becomes finite when it appears in a syntactic structure, where its tense and agreement features are interpreted. Another possibility would be to replace the formal features [FIN] on C and [PAST] on T with a single feature, whose name would be sufficiently neutral to be compatible with both mood and tense. For the sake of argument, let us refer to this feature as [X]. This proposal would solve the problem of agreement. [X] on T would receive its value from the tense morphology on the verb; [X] on C would receive its value from the valued instance of the same feature on T (value recycling). However, this approach does not allow us to distinguish between the categories tense and mood, as both are represented as [±X;±COIN]. Yet another option, which I am going to explore, is to suggest that the indicative mood does not need to be represented as a formal feature at all. Let me begin by briefly discussing Enç’s (1996) treatment of English present tense. Enç suggests that present tense is semantically vacuous: it does not shift the time evaluation either to the past or to
161 the future. Moreover, she argues that English does not have a present tense, and that the morphology on the verb, that is, the suffix –s, is subject agreement:
(37) a. Sarah likes lobster. b. Sarah talks with her mouth full.
(Enç 1996:346)
In contrast, Enç treats the suffix –ed in English as a true past tense morpheme. Under Enç’s approach, in the absence of tense morphology the original evaluation time is automatically understood to be the utterance time, and a clause receives a present tense interpretation, that is, a semantically vacuous interpretation. Building on Enç’s proposal, let us suppose that indicative mood is comparable in the relevant sense to present tense. It does not shift the evaluation time to a world different from the one in which a given situation took place. Unless otherwise indicated, that is, in the absence of any morphological or syntactic phenomena that would suggest a non-indicative mood, the original evaluation time is automatically understood to be the utterance time. In other words, the value of the features [+FIN; +COIN] does not need to be specified. This approach is consistent with the assumption in section 4.1 that in realis mood the evaluation time is always interpreted as the utterance time. Let us look at how this proposal would work. If there is no formal feature that corresponds to indicative mood, indicative clauses do not have to contain a CP projection. The presence versus absence of a CP projection is discussed by Bošković (1997) with respect to the categorial status of zero finite declarative complements such as
162 I believe John likes Mary. He argues that such complements are TPs by virtue of the following economy principle:16
(38) The Minimal Structure Principle Provided that lexical requirements of relevant elements are satisfied, if two representations have the same lexical structure and serve the same function, then the representation that has fewer projections is to be chosen as the syntactic representation serving that function.
(Bošković 1997:25)
Bošković (1997) demonstrates that the analysis of zero finite complement clauses as TPs provides a straightforward explanation for a number of syntactic phenomena. For example, it accounts for the lack of C-trace effects in constructions such as (39):
(39) a. Who do you believe likes Mary? b. *Who do you believe that likes Mary?
(Bošković 1997:30)
If no CP is present in the embedded clause in (39a) the sentence is expected to be grammatical. Russian embedded indicative clauses exhibit a similar contrast. In (40b), the presence of the overt indicative complementizer čto blocks wh-movement:17
16
Bošković’s (1997) theory involves IP.
163 (40) a. Kogo Anna dumaj-et
Liza ljubit ?
who Anna think-3RD.SG(IMPF) Liza love-3RD.SG(IMPF) ‘Who does Anna think that Liza loves?’ b. ??/* Kogo Anna dumaj-et
čto Liza ljubit ?
who Anna think-3RD.SG(IMPF) IND Liza love-3RD.SG(IMPF) ‘Who does Anna think that Liza loves?’
Following Bošković (1997), the grammaticality contrast in (40) suggests that Russian embedded indicative clauses do not always contain a CP projection. In particular, this projection appears to be absent in (40a). I propose that Bošković’s (1997) analysis can be extended to matrix indicative clauses. The fact that the sentences in (41) are grammatical, and that they receive a temporal interpretation suggests that no features in the derivation remain unvalued or uninterpretable:18
(41) a. Philemon jest
rybu.
(present)
Philemon eat-3RD.SG(IMPF) fish ‘Philemon eats/is eating fish.’
17
The phenomenon in (40) is not exactly parallel to English. Long-distance wh-movement in Russian is the topic of the following chapter. For the present purposes it is sufficient to demonstrate that in the absence of the indicative complementizer wh-movement is unproblematic. 18
In these examples I omit future imperfective, since nothing crucially depends here on a particular tense of the construction.
164 b. Philemon s-jest
rybu.
(future)
Philemon PREF-eat-3RDSG(PERF) fish ‘Philemon will eat fish.’ c. Philemon je-l
rybu.
(past)
Philemon eat-PAST(IMPF) fish ‘Philemon was eating fish.’
Under the approach adopted in this dissertation, the presence of tense does not require a CP projection either in matrix or in embedded clauses. At the same time, there is no empirical data to suggest that indicative clauses contain a CP projection.19 I propose that the realis interpretation of a clause does not require it either. Rather, it arises when the evaluation time is interpreted as the utterance time. The realis interpretation arises when tense is able to locate the situation time relative to the utterance time. Since there is only one case when this is possible, namely when the situation time is located along the actual world time line, the realis interpretation may be considered the default reading. No additional formal features or morphemes are required to convey this meaning.20 This view is consistent with the intuition that in tensed languages tense and the indicative mood are “equated” in the sense that a temporal reference suggests indicative mood. 19
Russian has scrambling into a position above the nominative subject. I discuss this property of Russian in chapter 5 and argue that scrambled phrases occupy multiple specifiers of TP. 20 Note, however, that realis mood can in principle be represented as a feature on C. For example, in some languages, embedded indicative clauses can have the evaluation time in SpecCP. The result would be the sequence of tense. A sequence of tense reading arises when the evaluation time is controlled by the matrix event.
165 In contrast, the irrealis interpretation arises when the evaluation time is not interpreted as the utterance time, and tense cannot locate a situation time relative to the utterance time on the actual world time line. There are many possible ways in which the event denoted by the propositional content of a given clause can be “non-actual” or “less than actual”. For example, irrealis clauses can be subjunctive, conditional, imperative, etc. Therefore, additional devices are required to convey a particular meaning. This meaning is conveyed by elements that occupy either SpecCP or C, depending on a particular construction.
4.5. CONCLUSION In this chapter, I argued that the analysis of tense and aspect developed in chapter 3 can be extended to the category mood. Following Iatridou’s (2000) semantic account, I proposed that just like tense and aspect, mood is a dyadic predicate that relates two arguments in syntax. However, I departed from Iatridou (2000) in that I argued that mood should be analyzed as a category that related two times: the evaluation time T-EVL (its external argument) and the utterance time TU (it internal argument). Under this approach, mood indeed expresses a relation between two worlds, by virtue of comparing the actual world time line to the time line of the situation time. In realis, the two time lines coincide: the evaluation time, the situation time, and the utterance time are all located along the actual world time line. In irrealis, the situation time is located on the time line in an alternative world.
166 I proposed that the time denoting arguments of mood are licensed by the formal feature [FIN] that licenses the relation of non-coincidence by supplying the value of the interpretable feature [COIN]. This feature is in turn provides the basis for a semantic interpretation. In particular, [–COIN]: past on T and irrealis on C. I demonstrated that this analysis, together with the paradigm-based approach (Ackerman and Webelhuth 1998), accounts for the interpretation of Russian subjunctive clauses. It captures the connection between the past tense morphology and the subjunctive interpretation. In particular, it accounts for the fact that Russian l-participles convey the past tense meaning and the subjunctive meaning in realis and irrealis clauses, respectively. I also discussed implications of this analysis for the syntactic representation of the realis mood. I explored the possibility that Russian does not have a formal feature associated with realis mood, and that the indicative interpretation arises by default, by virtue of T having interpretable content. This proposal gained support from Bošković’s (1997) Minimal Structure Principle, according to which a particular meaning must be expressed by the smallest number of functional projections possible. In the next chapter, I turn to a different, although related topic. I discuss the relation between tense and wh-movement. I argue that the tense of an embedded clause may not be directly responsible for the availability of long-distance wh-movement. In particular, I argue that the temporal dependency between the matrix and the embedded clauses, which is usually referred to as sequence of tense, does not license long-distance wh-movement. I propose an account of long-distance wh-movement in Russian that relies on the notion of subjacency and does not involve the notion of tense.
167 And Now for Something Completely Different (Monty Python, 1971)
CHAPTER 5 TENSE AND LONG-DISTANCE WH-MOVEMENT
5.0. INTRODUCTION In chapters 3 and 4 I have argued that indicative and subjunctive clauses in Russian contrast with respect to tense: Indicative clauses have independent tense, whereas subjunctive clauses do not. Instead, their temporal reference is determined either by the matrix verb for embedded subjunctives, or by the preceding discourse or pragmatics for matrix subjunctives. Indicatives and subjunctives also contrast with respect to another phenomenon: It has long been observed that in Russian wh-movement out of indicatives is problematic, whereas wh-movement out of subjunctives is acceptable (Perlmutter 1971; Comrie 1980; Pesetsky 1982b; Bailyn 1995; Bošcović 2004). The goal of this chapter is to explore the connection between tense and wh-movement, as summarized in (1):
(1)
Clause Type
Independent Tense
Long-Distance WH-Movement
Indicative
YES
NO
Subjunctive
NO
YES
168 (1) suggests a connection between independent tense and unavailability of whmovement. In particular, it may be hypothesized that independent tense in Russian blocks wh-movement. This hypothesis is supported by the fact that wh-movement out of tensed clauses is problematic in a number of other languages (Richards 2001). However, after careful examination of the relevant data, I conclude that at least in Russian, the connection between tense and wh-movement is not as straightforward as may seem at first, and in particular independent tense cannot be held responsible, at least directly, for the unavailability of long-distance wh-movement. I propose that in Russian, longdistance wh-movement out of indicative clauses violates a strong form of subjacency (Chomsky 2000). My analysis relies on two components: (i) the inability of the Russian CP to have more than one specifier, and (ii) the wh-status of the Russian indicative complementizer. Before I outline my analysis, I examine in detail the interaction between tense and wh-movement in light of an observation by Richards (2001). Richards notes that languages that have sequence of tense (SOT) tend to allow long-distance wh-movement out of indicative clauses (for example, English). In contrast, languages that do not have SOT do not allow long-distance wh-movement out of indicative clauses (for example, Polish). Although Richards does not develop a complete theory that would account for this phenomenon, he sketches a possible solution in terms of a head-movement operation that results in SOT and also licenses wh-movement. First, I show that such a solution does not work for Russian. Second, I demonstrate that even in languages that have SOT such as English, the connection between SOT and wh-movement is not straightforward.
169 However, I do not exclude the possibility that there may be an indirect connection between tense and wh-movement. For example, a particular property of tense may determine whether or not a given language can have multiple CP specifiers. I leave this issue for future research. This chapter is organized as follows. In section 5.1, I introduce the relevant Russian data. In section 5.2, I present Richards proposal and discuss some problems with it. In sections 5.3, I provide evidence that Russian TP can have multiple specifiers, whereas CP can only have one. In section 5.4, I propose that long-distance wh-movement out of indicative clauses violates subjacency. Section 5.5 is a conclusion.
5.1. WH-MOVEMENT OUT OF INDICATIVES AND SUBJUNCTIVES IN RUSSIAN It has long been observed that Russian embedded indicative and subjunctive clauses contrast with respect to wh-movement (Perlmutter 1971; Comrie 1980; Pesetsky 1982a; Bailyn 1995; Stepanov 1998; Bošcović 2004). Wh-movement is problematic out of indicative clauses, across an overt indicative complementizer. In contrast, wh-movement is acceptable out of embedded subjunctive clauses, across an overt subjunctive complementizer. The examples from chapter 2 are repeated below:1
(2) a. Anna dumajet, [ čto Liza ljubit Feliksa ]. Anna thinks
IND
(indicative)
Liza loves Felix
‘Anna thinks that Liza loves Felix.’ 1
In this chapter, I do not gloss the verbal agreement morphology since it is irrelevant for the discussion.
170 b. ??(*)Kogo Anna dumajet, [čto Liza ljubit t ]? who
Anna thinks
IND
Liza loves
‘Whom does Anna think that Liza loves?’
(3) a. Anna xočet, [čtoby Liza ljubi-l-a Anna wants
SUBJ
Feliksa].
(subjunctive)
Liza love-PAST Felix
‘Anna wants Liza to love Felix.’ b. Kogo Anna xočet, [čtoby Liza ljubi-l-a who Anna wants
SUBJ
t ]?
Liza love-PAST
‘Who does Anna want Liza to love?’
It should be mentioned that native speakers’ judgments vary regarding the grammaticality status of (2b). According to some speakers, wh-movement across the indicative complementizer is completely ungrammatical, but others may judge it marginally acceptable. Pesetsky (1982b:298) observes that extraction from embedded tensed clauses is possible, but has a marked status. Comrie (1980:102) points out that such extraction is unacceptable in written language, and that many speakers are horrified to learn that they use such constructions in speech. For this reason, I refer to longdistance wh-movement out of tensed clauses as problematic. What is crucial for my purposes here is that there is a grammaticality contrast between (2b) and (3b), even though the degree to which the former is unacceptable may vary from speaker to speaker.
171 Although this property of Russian wh-movement is well known, to the best of my knowledge no comprehensive analysis has been proposed so far. Following the analysis of Russian finite clauses in chapters 3 and 4, indicative and subjunctive clauses contrast with respect to their temporal properties. Indicative clauses have independent temporal reference, whereas the temporal reference of subjunctive clauses is determined by the matrix clause. Thus, a possible hypothesis would be that what blocks long-distance whmovement out of indicative clauses is independent tense. However, I demonstrate that despite its appeal, this hypothesis cannot be maintained. Before I introduce an alternative account in terms of subjacency in section 5.4, I examine the connection between tense and wh-movement in more detail in light of Richards (2001). Richards observes that languages that lack SOT also lack overt whmovement out of tensed clauses. I conclude that at least in Russian, the absence of SOT is not (directly) related to the unavailability of long-distance wh-movement.
5.2. INDEPENDENT TENSE DOES NOT BLOCK WH-MOVEMENT In this section, I examine the interaction between tense and wh-movement in more detail. In 5.2.1, I briefly review Richards’ (2001) suggestion. It is important to emphasize that Richards does not develop a complete theory of the interaction between tense and whmovement. Rather, he points out a generalization and suggests a possible direction in which a complete theory may be developed. In 5.2.2, I show how Richards’ proposal could be applied for Russian and why it does not work. In 5.2.3, I discuss some data that
172 complicate the connection between tense, and the SOT phenomena in particular, and long-distance wh-movement.
5.2.1. Wh-Movement and Sequence of Tense: Richards (2001:274–281) The sequence of tense phenomena (SOT) arise when the matrix and the embedded verb are both in past tense. According to Kratzer (1998), when SOT takes place the semantic features of the embedded tense are not interpreted at all. Other studies propose that SOT involves agreement between tense features of the matrix and the embedded clauses, which results in the embedded T being dependent for its interpretation on the matrix T (Comrie 1986; Ogihara 1996; Abusch 1988, 1994, 1997; Stowell 1995, 1996; Kondrashova 1999; Kusumoto 1999). For example, the English sentence below is ambiguous between SOT and non-SOT reading:
(4) Gordon said that Josephine was pregnant. a.
Gordon said: “Josephine was pregnant.” (the earlier-than-matrix reading; no SOT)
b.
Gordon said: “Josephine is pregnant.” (the simultaneous reading; SOT) (Kusumoto 1999:49)
According to the interpretation in (4a), Josephine’s pregnancy necessarily precedes Gordon’s reporting it. However, there is no requirement that Josephine’s pregnancy coincides with the time of Gordon’s report: Josephine may or may not be pregnant at the time of Gordon’s report. In contrast, according to the interpretation in (4b), Josephine
173 must be pregnant at the moment when Gordon is speaking. This last interpretation is said to involve SOT. Richards (2001) observes that “overt wh-movement cannot escape tensed clauses in the absence of Sequence of Tense…” (p. 278).2 He demonstrates that languages that have SOT, such as English, tend to allow overt wh-movement out of embedded tensed clauses, whereas languages that do not have SOT, such as Polish, tend to disallow it:3
(5) a. What do you think [ that Mary bought t ]?
(English)
b. Who did you see [ that John kissed t ]?
(6) *Co
Maria myśli, [ że Janek kupił
what Maria thinks
t ]?
(Polish)
that Janek bought
‘What does Maria think that Janek bought?’
2
The complete set of data dealt with in Richards (2001) includes both covert and overt wh-movement. Richards refers to “movement which occurs in the overt syntax as ‘overt movement’, and to movement which occurs in the covert syntax as ‘covert movement’…. [O]vert movement need not have any effect on the pronunciation of the sentence; … The terms ‘overt’ and ‘covert’ refer only to the timing of the derivation” (p. 3). According to Richards, overt movement and covert movement precede and follow Spellout, respectively (p. 3). He also assumes that “wh-in-situ does undergo a movement operation in the covert syntax…”(p. 7). For my purposes here I am only concerned with overt syntactic movement that precedes Spell-out and as such has an effect on the pronunciation. 3
Note, however, that according to Rudin (1988:454), certain bridge verbs permit single wh-words to be extracted. Examples like the following would seem to be problematic for this proposal: i. Co Janek mówił, [ że studenci czytają t ]? what Janek said that students read ‘What did Janek say that the students read?’ I return to the discussion of Polish in section 5.4.2.
174 Russian does not have SOT in embedded indicative clauses (Kondrashova 1999; Kusumoto 1999). In (7) below, the matrix and the embedded verbs are both in past tense, but the example allows only the non-SOT reading, as in (7a):4
(7) Maša
skazala, [ čto Petya byl bolen ].
Masha said
IND
Petya was sick
‘Masha said that Petya was sick.’
(Kondrashova 1999:183)
a. Masha said: “Petya was sick.”
(no SOT)
b. #Masha said: “Petya is sick.”
(SOT)
Thus, Russian patterns with Polish in that it lacks both the SOT phenomenon and longdistance wh-movement out of tensed clauses. Richards suggests a possible way of analyzing the connection between SOT and wh-movement:
“Suppose we assume a theory of Sequence of Tense like those suggested by Hornstein (1990) and Varlakosta (1994), in which Sequence of Tense indicates an instance of head-movement which makes the T head of the lower clause dependent in some relevant sense on the T head of the higher clause. I will not attempt to identify more closely the heads involved in this movement relation; for now, let us simply refer to the relevant head-movement as ‘α-to-β movement’. Now suppose further that α-to-β movement occurs in every language, and that Sequence of Tense indicates overt α-to-β movement, while the lack of Sequence of Tense indicates that α-to-β movement is covert. ... The new generalization, in 4
To express the equivalent of the SOT reading the present tense in the embedded clause is used.
175 other words, is that wh-movement out of tensed clauses is only possible if whmovement and α-to-β movement occur in the same component: they either both must be overt, or both covert.”
(Richards 2001:278–279)
My understanding of this proposal is that overt wh-movement out of tensed embedded clauses is only possible in the presence of overt α-to-β movement, that is, in the presence of SOT. Although Richards does not develop a complete theory that would capture this generalization, he suggests a possible account in terms of the Principle of Minimal Compliance (Richards 2001:199):
(8)
The Principle of Minimal Compliance If the tree contains a dependency headed by H which obeys constraint C, any syntactic object G which H “immediately c-commands” can be ignored for purposes of determining whether C is obeyed by other dependencies.
“Applying this kind of theory to the interactions between tense and wh-movement discussed here would involve positing a constituent which is an island for whmovement, but not for the movement involved in Sequence of Tense (the movement I have labeled ‘α-to-β movement’ above). α-to-β movement would then… [render] the island transparent for wh-movement just in case whmovement and α-to-β movement are both either overt or covert.” (Richards 2001:279–280)
176 Thus, α-to-β movement may be viewed as a movement operation that licenses longdistance wh-movement out of tensed clauses. Richards (2001:280) leaves the following two questions unanswered: First, what participates in α-to-β movement? Second, which constituent is an island for wh-movement? In the next section, I apply his proposal to Russian, and show that at least in this language, the connection between the absence of SOT and the unavailability of long-distance wh-movement is not so straightforward.
5.2.2. Wh-Movement and Tense in Russian At first glance, Russian data seem to support Richards’ suggestion. As discussed above, Russian lacks both SOT and long-distance wh-movement out of tensed embedded clauses, in accordance with Richards’ generalization. Therefore, it is tempting to develop an account of Russian wh-movement along the lines suggested by Richards. As a starting point, consider the sentences in (9), where wh-movement is possible out of tensed embedded clauses when the complementizer čto is omitted:
(9) a. Anna dumala [ Liza ljubila Feliksa ]. Anna thought
Liza loved Felix
‘Anna thought that Liza loved Felix.’ b. Kogo Anna dumala [ Liza ljubila t ]? who Anna thought
Liza loved
‘Who did Anna think that Liza loved?’
177 As proposed in section 4.4 of chapter 4, sentences such as in (9) contain a TP and not a CP complement (the Minimal Structure Principle; Bošković 1997:25). Recall that Richards proposes to posit a constituent which is an island for wh-movement, but not for the movement involved in Sequence of Tense, but he does not identify this constituent. Let us suppose that in Russian, the relevant island constituent is a CP. If the sentences in (8) do not contain a CP projection, α-to-β movement is not required to make this island transparent for wh-movement. Thus, the grammaticality of (9b) is accounted for. However, under this approach Russian subjunctive complements pose a problem because they are obligatorily CPs. As the ungrammaticality of (10b) shows, the subjunctive complementizer cannot be omitted:
(10) a. Anna xotela [ čtoby Felix poceloval Lizu ]. Anna wanted
SUBJ
Felix kissed
Liza
‘Anna wanted Felix to kiss Liza.’ b. *Anna xotela Anna wanted
[ Felix poceloval Lizu]. Felix kissed
Liza
‘Anna wanted Felix to kiss Liza.’
Moreover, although the temporal interpretation of embedded subjunctives is determined by the matrix verb, this temporal dependency differs from SOT in the following respects. First, as the examples in chapter 4 demonstrate, the matrix verb need not be in past tense. Second, embedded subjunctive clauses need not express an event
178 coinciding with the matrix event. Finally, SOT is only observed in embedded clauses, whereas Russian has matrix subjunctives whose temporal interpretation is determined by the previous discourse or pragmatics (see sections 4.2.1–4.2.2 of chapter 4). Given these differences, I conclude that the temporal dependency between the matrix and the subjunctive clause on the one hand and SOT on the other hand arise via different mechanisms. In particular, the former is not a reflex of the overt α-to-β movement. This, in turn, suggests that if in Russian CP is an island for wh-movement this island may not be made transparent in embedded subjunctives.5 Richards’ proposal can also be reinterpreted as follows. Suppose that what blocks wh-movement is independent tense content and that the relevant island constituent is a TP. Furthermore, overt α-to-β movement renders TP transparent for overt wh-movement by making the embedded T dependent on the matrix T. Under this interpretation, the analysis seems more promising for Russian. First, the temporal contrast between indicative and subjunctive clauses correlates with the contrast with respect to whmovement. Second, wh-movement is allowed not because of overt α-to-β movement per se, but indirectly because overt α-to-β movement eliminates the independent tense content of the embedded T. Then, in SOT languages the content of T is dependent on the matrix T as a result of overt α-to-β movement. In Russian, the subjunctive T does not have independent tense content to begin with, and therefore no movement operation is
5
It is possible to posit some other overt movement operation that renders the subjunctive CP in Russian transparent for wh-movement. However, such an approach is problematic because there is no empirical evidence for any other kind of overt movement that takes place out of subjunctives and not out of indicatives. In section 5.4 I show that the account in terms of subjacency follows naturally from other properties of Russian with no need to stipulate any additional movement operations.
179 required to make it dependent on the matrix T. Thus, Russian is expected to allow whmovement out of subjunctives regardless of whether or not it is a SOT language. However, under this interpretation (8b) is problematic. If the unavailability of long-distance wh-movement is due to T with independent content, there should be no contrast between (8b) and its counterpart with an overt complementizer. In the next section, I demonstrate that even in English, that is, a language that has SOT, the connection between SOT and wh-movement is not straightforward.
5.2.3. Further Problems with Richards’ Generalization 5.2.3.1. SOT is Not a Property of Languages Enҫ (1987) points out that the sequence of tense phenomenon may be observed in stative complement clauses, such as Josephine was pregnant in (4) above. However, the sequence of tense is never observed in eventive (non-stative) complement clauses. For example, the eventive sentences in (11) have only the non-SOT interpretation:
(11) a. Mary found out that John failed the test. b. The gardener said that the roses died. c. Sally thought that John drank the beer.
(Enҫ 1987:634)
In (11a), the event of John failing the test must precede the event of Mary’s finding out about him failing the test; in (11b), the event of the roses dying must precede the gardener’s report; in (11c), the event of John (allegedly) drinking the beer must precede
180 Sally’s thinking. The contrast between the ambiguous sentences in (4) and the nonambiguous ones in (11) suggests that SOT is a property of constructions, rather than languages.6,7 In particular, in English SOT is observed in constructions with stative predicates. Now, recall that according to Richards’ proposal, the SOT interpretation arises when there is overt α-to-β movement. This movement creates a dependency between the embedded tense and the matrix tense, and it also licenses long-distance wh-movement out of tensed embedded clause by rendering the relevant island constituent transparent. In light of Enҫ’s (1987) observation, Richards’ (2001) approach predicts that in English, long-distance wh-movement should only be possible out of embedded clauses with stative predicates, such as in (4): Since only stative clauses permit SOT, only stative clauses should involve overt α-to-β movement which licenses long-distance whmovement. Accordingly, since eventive embedded clauses such as in (11) do not exhibit SOT, they do not involve overt α-to-β movement which licenses long-distance wh-movement. As a result, wh-movement out of such embedded clauses is predicted to be
6
Kondrashova’s (1999) discussion of Russian, Japanese, and English also shows that SOT is a property of constructions and not languages. In particular, she demonstrates that Russian exhibits SOT in relative clauses. If her analysis is correct, Russian may not be claimed to be a non-SOT language. 7
Enҫ (1987) and Zagona (2003) propose different accounts of the temporal dependency in (3), which do not rely on a sequence of tense rule. Here I do not discuss their proposals. The point of the discussion in this section is to demonstrate that regardless of the account of the temporal dependency between the matrix and the embedded clauses, this phenomenon cannot be responsible for the properties of long-distance whmovement in a given language.
181 ungrammatical. However, as the grammaticality of wh-questions in (12) shows, this prediction is not borne out:
(12) a. What did Mary find out that John failed t ? b. What did Sally think that John drank t ?
In (12), the wh-movement occurs out of embedded clauses with eventive verbs. The declarative counterparts of these sentences do not allow the SOT reading, which suggests that overt α-to-β movement does not take place. The fact that the sentences in (12) are grammatical suggests that the SOT may not be the licensing factor for long-distance whmovement.8 More data further complicate the relation between SOT and long-distance whmovement. When the complementizer that is omitted, the SOT reading of wh-questions appears to be much better than the non-SOT:
(13) Who did Gordon think t was pregnant?
8
a. Who did Gordon think (last year) was pregnant (last year)?”
(SOT)
b. #Who did Gordon think (last year) was pregnant (two years ago)?
(No-SOT)
Of course, one again can stipulate some other overt movement to render the island transparent for whmovement. Here I do not pursue this idea.
182 However, when that is present, the wh-question receives the SOT reading only if the embedded clause is in progressive. Thus, the wh-question in (14) can only receive the reading in (14b), the one that involves SOT:
(14) Who did Gordon think that Josephine was dating t. a. #Who did Gordon think (last week) that Josephine was dating (last year)?
(No-SOT)
b. Who did Gordon think (last week) that Josephine was dating (last week)?
(SOT)
To sum up, the discussion in this subsection suggests that SOT is not a property of languages; rather, SOT arises in particular syntactic and semantic environments. If longdistance wh-movement is licensed by the same syntactic operation that results in the SOT reading, it is expected to be restricted to SOT environments. More specifically, longdistance wh-movement should be possible only in those constructions, which allow the SOT reading. I have shown that this is not the case. In the next subsection, I present more data that undermine the connection between SOT and long-distance wh-movement.
5.2.3.2. The Obviation Phenomenon in Russian Subjunctives In this section, I consider the obviation phenomenon in Russian subjunctives – a longdistance binding phenomenon, which is observed in tenseless clauses. It is often assumed that the tense properties of an embedded clause are relevant for both wh-movement and
183 binding.9 Therefore, examining the obviation phenomenon in Russian may shed light on the relation between tense and wh-movement. I show, however, that in Russian, longdistance binding facts do not parallel long-distance wh-movement facts. Thus, the obviation phenomenon does not provide independent evidence that tenseless (subjunctive) clauses in Russian are transparent for syntactic operations such as longdistance wh-movement. The obviation phenomenon is defined in terms of a disjoint reference requirement (Avrutin and Babyonyshev 1997):
(15) Disjoint Reference Requirement The pronominal subject of a subjunctive clause may not be co-indexed with the matrix subject.
In the subjunctive clause in (16a), the bound pronoun reading of the pronominal subject of the embedded subjunctive clause ona ‘she’ is unavailable. If it is co-indexed with the matrix subject Liza the sentence is ungrammatical; only the free pronoun reading is acceptable. In contrast, the indicative clause in (16b) is ambiguous between a free and a bound interpretation of the pronominal subject of the embedded clause:
9
For instance, Richards (2001:274–275) cites Boyd (1992), who suggests a parallel between the fact that in English, wh-island effects diminish considerably when the verb in the wh-island is an infinitive, and the conditions on long-distance anaphor binding in a number of languages, which also seem to suggest an important role for tense. For example, in Icelandic long-distance anaphors may be bound out subjunctive clauses, but not out of indicative clauses.
184 (16) a. Lizai xočet, [ čtoby ona*i/j pojexa-la v gorod]. Liza wants
SUBJ
she
(subjunctive)
go-PAST to town
‘Lizai wants her*i/j to go to town.’ b. Lizai dumajet, [ čto onai/j pojedet v gorod]. Liza thinks
IND
she
(indicative)
will.go to town
‘Lizai thinks that shei/j would go to town.’
The obviation phenomenon is observed in many languages, and has been extensively discussed in the literature. In particular, various studies follow the domain extension approach (see, for example, Rochette 1988 for French; Picallo 1984 for Catalan and Spanish; Johnson 1985 for Icelandic; Progovac 1993 for Serbo-Croatian; Terzi 1992 for Balkan languages). According to this approach, the subjunctive T is anaphoric, with its interpretation determined by the matrix T. Although these analyses do not postulate any kind of movement operation that would result in “dissolving” the boundary between the matrix and the embedded clause, they propose that due to the anaphoric nature of the embedded T the binding domain for the embedded subject pronoun is extended to include the matrix clause. The obviation phenomenon arises because the subject pronoun would violate Condition B if it were bound by the matrix subject, that is, if it were bound within its binding domain. The domain
extension
approach
is consistent
with
Richards’
(2001)
generalization. Moreover, it is highly compatible with his idea about the connection between long-distance wh-movement and SOT. Under his approach, long-distance wh-
185 movement becomes possible when a certain head-movement operation takes place. It may be further proposed that the same movement operation extends the binding domain of the embedded subject to include the matrix clause. In other words, the embedded clause boundary becomes transparent for binding of both, the embedded T and pronominal subjects. As a result, the sentence violates Condition B. However, Avrutin and Babyonyshev (1997) show that at least for Russian, the domain extension approach cannot be maintained. First, they point out that although the pronominal subject of the embedded subjunctive cannot be co-indexed with the matrix subject, the pronominal object of the embedded subjunctive can. In (17), the direct object pronoun ego ‘him’ may have the same reference as the matrix subject Volodja:
(17) Volodjai xočet [ čtoby Nadja pocelova-l-a egoi/j]. Volodja wants
SUBJ
Nadja kiss-PAST him
‘Volodja wants Nadja to kiss him.’
(Avrutin and Babyonyshev 1997:232)
The domain extension approach predicts (17) to be ungrammatical. If the binding domain for the embedded subject pronoun is extended to include the matrix clause, the embedded direct object is also within the binding domain of the matrix subject, and the sentence should violate Condition B, contrary to fact. Another problem identified by Avrutin and Babyonyshev (1997) concerns binding of reflexive possessive pronouns and pronominal possessives. In Russian, reflexive
186 possessive pronouns must be bound within their binding domain, whereas pronominal possessive pronouns must be free:
(18) Lizai ljubit svoegoi/*j / ee*i/j kota. Liza loves self’s
/ her cat
‘Liza loves heri/j cat.’
The domain extension approach predicts that the reflexive possessive subject pronoun of the embedded clause can be bound by the matrix subject, whereas the pronominal possessive subject pronoun cannot. However, as the examples below show, the pattern is reversed. In (18a), the reflexive pronoun cannot be bound by the matrix subject: the sentence is ungrammatical. In (19b), the pronoun subject can be bound by the matrix subject:
(19) a. *Volodjai xočet, [ čtoby svojai/j žena pojexa-la v Evropu]. Volodja wants
SUBJ
self’s wife go- PAST
to Europe
‘Volodjai wants self’si/j wife to go to Europe.’ b. Volodjai xočet, [ čtoby egoi/j žena pojexa-la v Evropu]. Volodja wants
SUBJ
his
wife go-PAST
to Europe
‘Volodjai wants hisi/j wife to go to Europe.’ (Avrutin and Babyonyshev 1997:234–235)
187 To conclude, Avrutin and Babyonyshev (1997) demonstrate that the domain extension approach is unable to account for the obviation phenomenon in Russian subjunctives. Therefore, the obviation phenomenon cannot provide evidence that Russian subjunctives are transparent for long-distance syntactic operations.
5.2.4. Summary The data presented in this section suggest that the connection between SOT and longdistance wh-movement is not so straightforward. There may be an indirect relationship between SOT and the availability of long-distance wh-movement. However, it is clear that what blocks long-distance wh-movement is not the independent tense specification. The relevant data involve wh-movement in English, as well as the availability of longdistance wh-movement out of a tensed clause in the absence of the indicative complementizer, and the obviation phenomenon in subjunctive clauses in Russian. 10
5.3. RUSSIAN IS A TP-ABSORPTION LANGUAGE Although the development of an alternative explanation of the crosslinguistic variation in SOT and wh-movement is beyond the scope of this chapter, I propose an alternative explanation for Russian wh-movement. In section 5.4 below, I will argue that longdistance wh-movement violates subjacency. My proposal relies on two components: the
10
It is possible that α-to-β movement proposed by Richards permits the SOT reading, but does not require it. In other words, in languages that have SOT such as English α-to-β movement takes place always, regardless of whether or not we are dealing with the SOT interpretation. The arguments presented in 5.2.3 would not hold up against this alternative. However, the arguments presented in 5.2.2 would still remain valid.
188 inability of Russian CP to have more than one specifiers (TP-Absorption) and the whstatus of the indicative complementizer. In this section, based on syntactic tests from Richards (2001), I demonstrate that Russian is a TP-absorption language.
5.3.1. Multiple Wh-Movement and TP-Absorption versus CP-Absorption Rudin (1988) argues convincingly that languages with overt wh-movement, in which all wh-phrases move into a preverbal position, fall into two categories. In languages of the first category, all wh-phrases appear in SpecCP (for example, Bulgarian or Romanian). In languages of the second category, only one wh-phrase appears in SpecCP, whereas the rest adjoin to TP (for example, Polish, Serbo-Croatian, or Czech). Rudin assumes that in all languages wh-phrases move to SpecCP, but they differ as to whether this movement takes place in overt syntax or at LF. She assumes also that in order for a wh-phrase to be able to undergo long-distance movement in syntax it has to be able to adjoin to the embedded CP in syntax.11 One of Rudin’s arguments concerns multiple wh-extraction. To illustrate her point, I use examples from Bulgarian and Serbo-Croatian. Rudin shows that in Bulgarian, all wh-phrases in a multiple question must move up to closest interrogative SpecCP. Whphrases may neither remain in situ, as in (20b–c), nor move to the specifier of a noninterrogative CP, as in (20d–e):
11
Rudin (1988) uses the term adjunction to CP/TP; for my purposes here I assume that adjunction to CP/TP and substitution into SpecCP/TP are equivalent, and refer to the wh-movement in question as movement into SpecCP/TP.
189 (20) Bulgarian (Rudin 1988:450) a. Koji kŭdej misliš
[ če
e
otišŭl ti tj ]?
who where think-2ND that has gone ‘Who do you think that went where? b. *Koj misliš
[ če e
otišŭl t kŭde]?
who think-2ND that has gone c. *Kŭde misliš
[če koj e
where otišŭl t ]?
where think-2ND that who has gone d. *Koji misliš
[ kŭdej (če)
e
otišŭl ti tj ]?
who think-2ND where that has gone e. *Kŭdej misliš
[koji (če) e otišŭl ti tj]?
where think-2ND who that has gone
The only grammatical example is (20a), where both wh-phrases undergo movement into the matrix SpecCP. In (20b–c), one wh-phrase remains in situ, whereas in (20d–e) one wh-phrase moves to the embedded SpecCP, and as result, all four sentences are ungrammatical. In contrast, in languages such as Serbo-Croatian, multiple wh-extraction is ungrammatical:
190 (21) Serbo-Croatian (Rudin 1988:454, from Steven Franks) a. Ko želite
[da vam šta
kupi t]?
who want-2ND to you what buy ‘Who do you want to buy you what?’ b. Šta
želite
[da vam ko
kupi t ]?
what want-2ND to you who buy ‘What do you want who to buy you?’ c. *Koi štaj
želite
[da vam kupi ti tj ]?
who what want-2ND to you buy d. *Štaj koi želite
[da vam kupi ti tj ]?
what who want-2ND to you buy e. *Koi želite
[štaj da vam kupi ti tj ]?
who want-2ND what to you buy f. *Štaj želite
[koi da vam kupi ti tj ]?
what want-2nd who to you buy
In grammatical examples (21a–b), only one wh-phrase moves into the matrix SpecCP, whereas the other one remains in situ. In (21c–d), both wh-phrases move into the matrix SpecCP, whereas in (21e–f), one wh-phrase moves into the embedded SpecCP, rendering the sentence ungrammatical. Under the assumption that movement across a filled SpecCP is universally barred, Rudin (1988) proposes that in languages such as Bulgarian all wh-phrases move into the
191 matrix SpecCP because these languages allow multiple CP specifiers.12 Therefore all whphrases can move into the embedded SpecCP in syntax on their way to the matrix SpecCP. In contrast, languages such as Serbo-Croatian allow only one specifier of CP, so that only one wh-phrase is able to move into the embedded SpecCP in syntax, whereas the rest can only do so covertly, or at LF. As a result, in these languages only one whphrase can undergo overt movement into the matrix SpecCP. Russian patterns with Serbo-Croatian with respect to long-distance multiple whmovement in that it does not allow more than one wh-phrases in the matrix SpecCP:13,14
(22) a. *Kto čto
ty xočeš, [ čtoby t kupi-l t]?
who what you want b. *Čto kto ty
SUBJ
buy-PAST
xočeš, [čtoby t kupi-l t]?
what who you want
SUBJ
buy-PAST
Note, however, that Russian differs from Serbo-Croatian in that it does not seem to allow multiple wh-movement to SpecCP even at LF. In (23), although the second whphrase remains in situ, the sentences nevertheless remain problematic:
12
This assumption follows from the recent version of the subjacency condition (Chomsky 2000), adopted in section 5.4. 13
Since Russian disallows long-distance wh-movement out of indicative complements, to illustrate the whisland constraint I use subjunctive complement clauses. An extensive discussion of long-distance whmovement in Russian is provided in Bailyn (1995). 14
It must be noted that judgments with respect to multiple wh-movement out of subjunctive clauses vary considerably. The examples in (22) reflect my own judgments. However, according to Stepanov (1998:460), such sentences are grammatical, which means that Russian, as opposed to Serbo-Croatian does not exhibit superiority effects.
192 (23) a. ??Kto ty
xočeš, [ čtoby t kupi-l
who you want b. ??Čto ty
SUBJ
čto]?
buy-PAST what
xočeš, [ čtoby kto kupi-l
what you want
SUBJ
t ]?
who buy-PAST
Examples such as in (23) are marginally acceptable under the interpretation, which Pesetsky (1987) refers to as “paired reading”. Under this reading, the wh-phrases within the embedded clause are D-linked, and, under Pesetsky’s (1987) analysis, are assigned scope in situ, without movement. Pesetsky (1987:118) observes that the same is true for Polish and suggests a possibility that Polish has no LF wh-movement. A grammatical option in Russian contains a single wh-phrase, as in (24):
(24) a. Kto ty
xočeš, [ čtoby t kupi-l
who you want
SUBJ
knigu]?
buy-PAST book
‘Who do you want to buy a/the book?’ b. Čto ty
xočeš, [ čtoby Felix kupi-l t]?
what you want
SUBJ
Felix buy-PAST
‘What do you want Felix to buy?’
Building on Rudin’s (1988) analysis, Richards (2001) proposes that the contrast between languages such as Bulgarian on the one hand and Serbo-Croatian on the other hand follows if languages are divided into CP-absorption languages and TP-absorption
193 languages. In CP-absorption languages, CP can have multiple specifiers, whereas TP can only have one. TP-absorption languages are the mirror image of CP-absorption languages: In such languages, TP can have multiple specifiers, whereas CP can only have one. Richards (2001:13) proposes the clause structures for the two types of languages shown in (25). In (25a), wh-movement takes place into multiple specifiers of CP, whereas in (25b) wh-movement is into multiple specifiers of TP. The former is A-bar movement, whereas the latter is A-movement:
(25) a. CP-Absorption
b.
CP WH1
TP-Absorption CP
CP
WH2
C` CP
WH3 C
C C`
TP WH1
TP
TP
WH2 T`
T A-bar movement
TP
WH3 vP
T` T
vP
A-movement
Languages discussed by Richards (2001) fall into two distinct categories with respect to these tests. Each category includes genetically unrelated languages: CP-absorption languages include Bulgarian, Chinese, and Tibetan, whereas TP-absorption languages include Serbo-Croatian, Polish, Japanese, and Hungarian.15 For ease of exposition, I will
15
Following Richards (2001), there are four logical possibilities. In addition to the distinction discussed here, TP and CP can either both have only one specifier, or they can both have multiple specifiers. This
194 continue to use Bulgarian and Serbo-Croatian as representatives of CP- and TPabsorption, respectively. The application of these tests to Russian shows that it patterns with Serbo-Croatian and is unambiguously a TP-absorption language.
5.3.2. Diagnostics for TP-Absorption and CP-Absorption To distinguish between CP-absorption and TP-absorption languages, Richards (2001:13– 30) proposes a number of syntactic tests. These tests include wh-island effects, scrambling, weak crossover, and superiority. The diagnostic that involves wh-island effects is closely related to Rudin’s (1988) extraction test and relies on the availability of multiple CP specifiers. The diagnostics that involve scrambling, weak crossover, and superiority are based on the distinction between A- and A-bar movement. Under the assumption that movement into SpecCP is A-bar movement, whereas movement into SpecTP is A-movement, these tests provide evidence for whether a given XP moves into SpecCP or SpecTP. It is important to bear in mind that Rudin’s original proposal concerns languages in which all wh-phrases move overtly. Richards extends her analysis to languages such as English, where only one wh-phrase moves overtly and the rest move covertly, or Chinese, where all wh-phrases move covertly. Thus, under Richards’ analysis, while the occurrence of multiple wh-phrases in SpecCP implies that we are dealing with a CPabsorption language, it is not the case that all CP-absorption languages necessarily have overt wh-movement into SpecCP. In other words, it is still possible to classify a given typology is a case study by itself, and I leave it for future research. What is relevant for my purposes here is that Russian allows multiple TP specifiers, but only one CP specifier.
195 language as being CP-absorption even if it does not have overt wh-movement (for example, Chinese).
5.3.2.1. Wh-Islands As already noted, the possibility of extracting out of an embedded interrogative clause is related to multiple wh-extraction. It concerns the island status of embedded questions and is based on the idea that SpecCP is the obligatory escape hatch for movement out of an embedded CP. Rudin (1988:456–461) observes that languages such as Bulgarian do not obey the wh-island constraint. This is expected if in these languages any number of whphrases can move into the embedded SpecCP, and thus cannot be blocked from moving out of that CP. In contrast, languages such as Serbo-Croatian obey the wh-island constraint. In such languages, only one wh-phrase is allowed to move overtly into SpecCP. As a result, movement of other wh-phrases through the embedded CP is blocked. This is illustrated in (26):
(26) a. [CP whi …. [CP [CP whj…[TP [tj ti ]]]]
b. [CP whi …. [CP whj…[TP [tj ti ]]]]
CP-Absorption: Bulgarian
TP-Absorption: Serbo-Croatian
Following Rudin’s (1988) analysis, Richards (2001) proposes that movement out of an embedded question is allowed in CP-absorption languages, that is, languages,
196 which allow multiple CP specifiers. In contrast, such movement is disallowed in TPabsorption languages, that is, languages, in which CP can only have one specifier. The ungrammaticality contrast in (27) suggests that Bulgarian is a CP-absorption language, whereas Serbo-Croatian is a TP-absorption language:
(27) a. Bulgarian, CP-absorption (Richards 2001:12, from Rudin 1988:457) ?Koja ot tezi knigi se čudiš
[koj t znae
koj t prodava t ]?
which of these books wonder-2ND.SG who knows who
sells
‘Which of these books do you wonder who knows who sells?’ b. Serbo-Croatian, TP-absorption *Šta
si
me pitao [ko
what aux-2ND.SG me asked
t može da uradi t]?
who
can
to do
‘What have you asked me who can do?’ (Richards 2001:12, from Rudin 1988:459) c. ?*Kakvu knjigu Marko i what
Ivan znaju [ kada je Petar pročitao t ]?
book Marko and Ivan know
when is Petar read
‘What book do Marko and Ivan know when Petar read?’
(Bošcović 2004:619)
Although according to Richards’ informants the Bulgarian example in (27a) is not perfect, it is better than the Serbo-Croatian example in (27b). Under his analysis, this is due to the fact that in Bulgarian, after one specifier of the embedded CP is filled by koj ‘who’, the wh-phrase koja ot tezi knigi ‘which of these books’ can move into another
197 specifier of the embedded CP. This option is not available in Serbo-Croatian: šta ‘what’ cannot move into the only specifier of the embedded CP, because it is already occupied by ko ‘who’. This, it has to move through a filled SpecCP, rendering the sentence ungrammatical. Note, however, that the Bulgarian example in (27a), unlike the Serbo-Croatian example in (27b), contains a D-linked wh-phrase. Rudin (1988:457) notes that movement of an interrogative wh-word out of an embedded wh-question is usually felt to be less normal than movement of a relative pronoun (which I discuss immediately below), but is fairly acceptable with heavy wh-phrases, such as in (27a). Rudin (1988) does not use the term D-linked, presumably assuming that this should make no difference. To ensure a fair comparison, in (27c) I give a Serbo-Croatian example from Bošcović (2004), which contains a D-linked wh-phrase and is ungrammatical, which supports Richards’ (2001) argument. The contrast between the two types of languages extends to extraction of relative pronouns out of embedded questions. Under the assumption that relativization involves movement of a relative pronoun into SpecCP this is expected:
(28) a. Bulgarian, CP-absorption (Rudin 1988:457) Vidjah edna kniga, kojta se čudja saw-1ST a
[koj znae
[koj prodava ___]].
book which wonder-1st who knows who sells
‘I saw a book which I wonder who knows who sells (it).’
198 b. Serbo-Croatian, TP-absorption (Rudin 1988:459) *… osoba,
koja sam
ti
rekao [gde (on) živi…]
individual who have-1ST you told
where he
lives
‘… the individual who you asked me where (he) lives.’
As (28a) shows, Bulgarian allows wh-extraction even multiple embedded interrogative clauses. Under Richards’ analysis, the relative pronoun kojta ‘which’ passes through the specifiers of both embedded CPs on its way to the higher clause, and the sentence is grammatical. In contrast, the ungrammaticality of (28b) shows that Serbo-Croatian does not allow extraction of relative pronouns out of questions, just as it does not allow extraction of interrogative wh-phrases. Since the embedded SpecCP is occupied by gde ‘where’ the relative pronoun koja ‘who’ cannot move into it and has to undergo longdistance movement across SpecCP. As a result, the sentence is ruled out. In Russian, an extraction out of interrogative complements is ungrammatical. In (29a), the verb znat’ ‘know’ takes an interrogative complement. In (29a), kto ‘who’ occupies the embedded SpecCP. In (29b), the direct object čto ‘what’ undergoes longdistance wh-movement across the embedded wh-subject and the sentence is ungrammatical:
(29) a. Anna znajet, [ kto t prodajet etot dom ]. Anna knows
who
sells
this house
‘Anna knows who is selling this house.’
199 b. *Čtoj Anna znajet [ ktoi ti prodajet tj ]? what Anna knows who
sells
‘What does Anna know that who sells?’
Russian relative pronouns also provide a testing ground for wh-island effects. Here Russian patterns with Serbo-Croatian in that it does not allow extraction of relative pronouns out of embedded interrogative clauses. The example in (30b), which is parallel to the Bulgarian sentence in (28a), is ungrammatical:
(30) a. Mne interesno, [ kto t prodajet etot dom ]. me interesting
who
sells
this house
‘I wonder, who is selling this house.’ b. *Ja videla dom, kotoryj mne interesno I saw
house which
[ kto prodajet t].
me interesting who sells
‘I saw a house, which I wonder who sells.’
The sentence in (30a) is interpreted as an indirect question, and I assume that kto ‘who’ occupies the embedded SpecCP.16 The ungrammaticality of (30b) is expected if Russian is a TP-absorption language: If SpecCP of the embedded interrogative is filled by kto
16
One consideration that justifies this assumption is that overt complementizer may not be inserted into the structure: i. *Mne interesno, čto kto prodaet etot dom. me interesting that who sells this house
200 ‘who’, the relative pronoun kotoryj ‘which’ cannot move into this position on its way into the matrix clause, which results in illicit movement out of the interrogative CP.
5.3.2.2. Weak Crossover and Scrambling Richards (2001) observes that CP-absorption languages and TP-absorption languages contrast with respect to weak crossover, a phenomenon, which is related to variable binding. For a variable to be bound it has to be c-commanded by its antecedent and all of the traces of this antecedent (Mahajan 1990). The weak crossover effect arises when the variable is found either in a position which is not c-commanded by its antecedent, or in a position between the antecedent and its trace. The first configuration is usually basegenerated, as in (31a), whereas the second configuration typically results from A-bar movement of the antecedent across the variable, as in (31b):
(31) a. *ei … XPi b. *XPi… ei … t
Let us first consider the weak crossover phenomenon in a base-generated structure; weak crossover as a result of movement is discussed in section 5.3.2.4. In this case, a weak crossover effect may be repaired by scrambling, a semantically vacuous movement operation (see, for example, Ross 1967; Saito 1989, 1992; Bailyn 1995, 2001; Mahajan 1990). Richards (2001) observes that TP-absorption languages use local scrambling as a repair strategy for weak crossover violations:
201 (32) Serbo-Croatian, TP-absorption (Richards 2001:15) a. ?? Njegovi susjedi his
ne vjeruju nijednom politicarui.
neighbors not trust
no
politician
‘His neighbors trust no politician.’ b. Nijednom politicarui njegovi susjedi no
politician
his
ne vjeruju t .
neighbors not trust
‘His neighbors trust no politician. (=No politician is trusted by his neighbours)’
In (32a), the subject pronoun njegov susjedi ‘his neighbors’ may not be interpreted as a bound variable because at no point in the derivation is it c-commanded by its antecedent, the object nijednom politicaru ‘no politician’. In (32b), the object undergoes movement into a position above the subject, from which it can bind the pronominal variable in it, and the bound variable reading becomes available. The ability to establish new binding relations, and in particular to repair weak crossover violations is traditionally considered a property of A-movement. Under the standard assumption that SpecTP is an A-position, the grammaticality of (32b) provides evidence that nijednom politicaru ‘no politician’ moves into SpecTP. In contrast, in CP-absorption languages like Bulgarian scrambling is either absent, or has properties of A-bar movement. In particular, it is unable to cure weak crossover violations:
202 (33) Bulgarian, CP-absorption (Richards 2001:15–16) a. *Majka mui običa vseki čoveki mother his love every person ‘Hisi mother loves everyonei.’ b. *Vseki čoveki običa majka mui t every person
love mother his
The Bulgarian sentence in (33a) is ungrammatical for the same reasons as its SerboCroatian counterpart in (32a). However, movement of the quantified direct object vseki čovek ‘every person’ into a position above the subject does not result in a grammatical sentence. Under the assumption that SpecCP is an A-bar position, the ungrammaticality of (33b) follows if in Bulgarian, the direct object cannot move into SpecTP, and instead undergoes A-bar movement to SpecCP. This contrast is illustrated in (34):
(34) a. [TP DO [TP S … [VP … t ]]]
TP-Absorption: Serbo-Croatian
A-movement b. [CP DO [TP S …
[VP … t ]]]
CP-Absorption: Bulgarian
A-bar movement
Richards (2001) points out that intuitively it makes sense that if a language allows multiple TP specifiers it would use this kind of movement both for wh-movement and scrambling. In other words, scrambling in TP-absorption languages should be into
203 SpecTP, an A-position, and have properties of A-movement. In contrast, scrambling in CP-absorption languages should be into SpecCP, an A-bar position, and is expected to have properties of A-bar movement. Richards (2001:20) suggests that in TP-absorption languages, wh-movement and scrambling belong to the same movement type. More specifically, in these languages wh-movement is similar to the type of scrambling found in Japanese or Hindi, which can cure weak crossover violations only if it is local. Russian patterns with Serbo-Croatian in that it uses local scrambling as a repair strategy, which again suggests that it is a TP-absorption language:
(35) a. *Egoi mama ljubit každogo mal’čikai. his mother loves every
boy
‘Hisi mother loves every boyi.’ b. Každogo mal’čikai ljubit t egoi mama. every
boy
loves his mother
‘Every boyi, hisi mother loves himi.’
In (35a), the embedded subject ego mama ‘his mother’ cannot be interpreted as a bound variable because at no point in the derivation is it c-commanded by the quantified direct object každogo mal’čika ‘every boy’. In (35b), the quantified direct object moves into a position within the embedded clause, from which it c-commands the subject, and the sentence becomes grammatical. This suggests that in (35b), the direct object undergoes A-movement into the embedded SpecTP.
204 Note that local scrambling in Russian also repairs weak crossover violations in embedded clauses:
(36) a. Liza dumajet, *[čto egoi mama Liza thinks
ljubit každogo mal’čikai].
that his mother loves every
boy
‘Liza thinks that hisi mother loves every boyi.’ b. Liza dumajet, [čto každogo mal’čikai ljubit t egoi mama]. Liza thinks
that every
boy
loves
his mother
‘Liza thinks that every boyi, hisi mother loves himi.’
The grammaticality of (36b) suggests that matrix and embedded clauses in Russian have the same TP-absorption properties. In other words, TP in embedded clauses can have multiple specifiers, whereas CP can only have one. This observation becomes important in section 5.4, where I propose an account of Russian long-distance wh-movement.
5.3.2.3. Superiority Richards (2001) shows that wh-movement in CP-absorption languages obey superiority, that is, the requirement that the hierarchical order of wh-phrases be preserved after movement (for example, subject > object; *object > subject). In CP-absorption languages, the fronted wh-phrases are subject to this rigid ordering constraint:
205 (37) Bulgarian, CP-absorption (Richards 2001:16, from Rudin 1988:472–473) a. Koj kogo t vižda t ? who whom
sees
‘Who does see whom?’ b. *Kogo koj t vižda t? whom who
sees
In contrast, in TP-absorption languages such as Serbo-Croatian, local whmovement does not obey superiority, as shown in (38). Superiority effects obtain only for long-distance wh-movement, as in (39):
Serbo-Croatian, TP-absorption (Richards 2001:17, from Bošković 1995: 5–6, 8) (38) a. Koj
je kogai tj vidjeo ti
who aux whom
see
‘Who saw whom?’ b. Kogai je koj tj vidjeo ti? whom aux who
see
‘Who saw whom?’
(39) a. Koi
si
kogai
tvrdio
who aux whom claimed
[ da je tj istukao ti ]? that aux
‘Who did you claim beat whom?’
beaten
206 b. *Kogai
si
koi tvrdio [ da je tj istukavo ti ]?
whom aux who claimed that aux
beaten
To account for the possibility of multiple long-distance wh-movement in Serbo-Croatian in (39), Richards (2001:25) suggests that multiple CP specifiers may be allowed in this language just in case none of these specifiers will be interpreted as a scope position of a wh-word; multiple CP specifiers are licensed in Serbo-Croatian as long as they are all used only as intermediate landing sites. This diagnostic for CP- versus TP-absorption is valid under the assumption that superiority constraints A-bar movement, but not A-movement. If all wh-phrases move to SpecCP they all undergo A-bar movement and are thus subject to superiority. In contrast, if all but one wh-phrases move to SpecTP it is possible that they undergo A-movement and therefore are not constrained by superiority. However, superiority effects are expected with long-distance wh-movement: having undergone A-bar movement into SpecCP, the wh-phrase may not A-move into the matrix SpecTP, as this would result in improper movement (that is, A-bar movement into the embedded SpecCP followed by Amovement into the matrix SpecTP). With respect to local wh-movement, Russian patterns with Serbo-Croatian in that it imposes no constraints on the ordering of fronted wh-phrases:
207 (40) a. Kton čtoj
komuk tn podaril tk tj?
who what whom
gave
‘Who gave what to whom?’ b. Čtoj
komuk kton tn podaril tk tj ?
what whom who
gave
‘Who gave what to whom?’
The grammaticality of (40) again suggests that Russian is a TP-absorption language.
5.3.2.4. Weak Crossover and Wh-Movement As mentioned in section 5.3.2.2, weak crossover effects may arise when an operator undergoes A-bar movement across the variable, which it binds. Since movement of whpronouns in interrogative clauses is the prototypical example of wh-movement, it is expected to give rise to weak crossover effects. Richards (2001) points out that this is indeed the case in CP-absorption languages:
(41) Bulgarian, CP-absorption (Richards 2001:19) a. Koji t običa majka sii? who
loves mother his
‘Whoi loves hisi mother?’
208 b. *Kogoi običa mojka sui t who
loves mother his
‘Whoi does hisi mother love?’
In the grammatical (41a), the object is c-commanded both by the subject wh-phrase and its trace. In contrast, in the ungrammatical (41b), the object wh-phrase undergoes A-bar movement into SpecCP and as a result, the subject variable c-commands its trace. In contrast, in TP-absorption languages, local wh-movement does not induce weak crossover effects. This provides further support for Richards’ (2001) idea that in TP-absorption languages, wh-movement and scrambling belong to the same movement type. Weak crossover effects only obtain for long-distance wh-movement:
(42) Serbo-Croatian, TP-absorption (Richards 2001:20) a. Tkoi t voli svojui majku? who
loves his
mother
‘Whoi loves hisi mother?’ b. Kogai voli who
njegovai majka t ?
loves his
mother
‘Whoi does hisi mother love?’ c. *Kogai njegovai majka misli da Marija voli t? who
his
mother thinks that Maria loves
‘Whoi does hisi mother think that Mary loves?’
209 If in TP-absorption languages, wh-movement and scrambling belong to the same movement type in the sense that both have properties of A-movement, a wh-phrase may be allowed to adjoin to TP before moving to SpecCP, which would explain the fact that local wh-movement does not induce weak crossover effects. In contrast, Richards (2001) argues that when long-distance wh-movement takes place the wh-phrase must use the embedded SpecCP as an escape hatch even in TP-absorption languages. It follows then that it may not undergo A-movement into the matrix SpecTP, as this would result in improper movement. This explains why long-distance wh-movement induces weak crossover effects even in TP-absorption languages. With respect to this test, Russian poses a problem for the clear-cut distinction between CP- and TP-absorption: It patterns with CP-absorption languages in that local wh-movement does lead to weak crossover violations (Bailyn 1995; Sekerina 1997:55):
(43) a. Ktoi t ljubit svojui mamu? who
loves self’s mother
‘Whoi loves hisi mother?’ b. *Kogoi ljubit t egoi mama? who
loves
his mother
‘Whoi does hisi mother love?’
The sentence in (43a) is grammatical, as expected: The wh-movement of the subject does not affect the c-command relations between the subject and the pronominal variable. In
210 contrast, in the ungrammatical (43b) the wh-movement of the direct object across the pronominal variable within the subject results in a weak crossover violation.17 A solution to this apparent problem would be to analyze Russian wh-movement as focus movement. Bošcović (2002), following Stepanov (1998), and also the work by Horvath (1986) and É. Kiss (1995), argues that Russian wh-phrases are inherently focused, and that they undergo obligatory focus movement. It is well known that a focalized XP leaves a trace that behaves like a variable, as it is sensitive to weak crossover (Rizzi 1997; Beninca and Polletto 2004), which explains the ungrammaticality of (43b). However, this approach implies that focus in Russian is licensed in a projection located between TP and CP, and that all wh-phrases obligatorily move into this position, rather than to SpecTP.18 At this point, I do not have an elegant solution for this problem. For the purposes of this chapter, I assume that Russian is a TP-absorption language, as is shown by three other tests. I leave this issue for future research.
17
Note that, according to my own judgments, the weak-crossover effect diminishes with D-linked whphrases, and especially, when negation is present, as in (ia). However, the weak-crossover effect still remains when long-distance wh-movement takes place: i a. Kakogo mal’čikai ne ljubit t egoi mama? which boy NEG loves his mother ‘Which boyi doesn’t hisi mother love?’ b. *Kakogo mal’čikai Anna xočet čtoby egoi mama ne kormi-la t? which boy Anna wants SUBJ his mother NEG feed-PAST ‘Which boyi does Anna want hisi mother not to feed himi?’ Here I do not address this contrast between D-linked and non-D-linked wh-phrases and leave this issue for future research. 18
It should be mentioned that according to Stepanov (1998), Russian wh-phrases never move to SpecCP, and the focus position is located below the CP projection. If this is correct it makes wh-movement invalid as a diagnostic for the TP-absorption status of Russian. Note however, that Stepanov’s (1998:457) argument is based on the assumption that superiority effects are a necessary and sufficient condition for true wh-movement, that is, movement of a wh-phrase to SpecCP.
211 5.3.3. Summary To conclude the discussion in this section, the behavior of Russian finite clauses with respect to wh-islands, A-scrambling, and superiority suggests that under Richards’ (2001) classification, Russian is a TP-absorption language, that is, Russian finite clauses all contain a TP projection, which can have multiple specifiers, whereas CP can only have one.19 However, Russian patterns with Bulgarian, a CP-absorption language, in that local wh-movement induces a weak crossover effect. This may be accounted for independently of the distinction between TP- and CP-absorption languages if wh-movement in Russian is treated as focus movement. The summary of the tests is given in the table below:
(44)
TEST
CP-ABSORPTION
TP-ABSORPTION
Bulgarian
Serbo-Croatian
Russian
Wh-islands
No
Yes
Yes
A-Scrambling
No
Yes
Yes
Weak Crossover
Yes
No
Yes
Superiority
Yes
No
No
Having established that Russian CP can only have one specifier, in the next section I propose an account of long-distance wh-movement in Russian in terms of subjacency.
19
Note that the status of Russian as a TP-absorption language is consistent with the claim that Russian has two subject positions: the grammatical subject position and the topic position (Schoorlemmer 1995:35). The same claim was made for Japanese, another TP-absorption language (see, for example, Doron and Heycock 1999 and references there). It would be a separate research project to establish whether this is a general property of TP-absorption languages.
212 5.4. RUSSIAN LONG-DISTANCE WH-MOVEMENT VIOLATES SUBJACENCY 5.4.1. Russian Complementizers: ČTO versus ČTOBY Having established that the Russian CP can only have one specifier, I propose an account of long-distance wh-movement in Russian in terms of subjacency, a condition on movement. Here I adopt the recent implementation of this condition within the phase theory. Chomsky (1999, 2000) proposes a “phase-impenetrability condition”, stated in (45):
(45) Phase-Impenetrability Condition In phase α with head H, the domain of H is not accessible to operations outside α, only H and its edge are accessible to such operations.
(Chomsky 2000:108)
Chomsky (2000) suggests that “the Phase-Impenetrability Condition yields a strong form of Subjacency. … [It] requires that A-bar movement target the edge of every phase, CP and vP.” (p. 108).21 Thus, assuming that wh-movement is an instance of A-bar movement, to be able to move to the matrix SpecCP the wh-phrase must stop at the embedded SpecCP to avoid a subjacency violation. Let us consider the relevant data again. In Russian, long-distance wh-movement is problematic out of indicative clauses, introduced by the indicative complementizer čto, as shown in (46):
21
For the purposes of the discussion in this chapter, I ignore the phase status of vP and focus on CP.
213 (46) a. Anna dumajet, [ čto Liza ljubit Feliksa ]. Anna thinks
IND
Liza loves Felix
‘Anna thinks that Liza loves Felix.’ b. ??(*)Kogo Anna dumajet, [čto Liza ljubit t ]? who
Anna thinks
IND
Liza loves
‘Whom does Anna think that Liza loves?’
Note that subjacency violations often have this flavor, as shown by the English examples:
(47) a. Who do you wonder [ t` that John married t ]? b. ??/*Whoi do you wonder [ whyk John married ti tk ]]?
In contrast, long-distance wh-movement is acceptable out of subjunctive clauses, introduced by the subjunctive complementizer čtoby, as shown in (48):
(48) a. Anna xočet, [čtoby Liza ljubi-la Anna wants
SUBJ
Feliksa].
Liza love-PAST Felix
‘Anna wants Liza to love Felix.’ b. Kogo Anna xočet, [čtoby Liza ljubi-la who Anna wants
SUBJ
Liza love-PAST
‘Who does Anna want Liza to love?’
t ]?
214 I propose that the contrast between (46) and (48) can be accounted for in terms of the difference between the complementizers. The Russian indicative complementizer čto is a wh-word, meaning ‘what’. As demonstrated in (49), it functions as an interrogative pronoun in interrogative contexts. In (49a), it functions as an indicative complementizer, whereas in the interrogative (49b), it functions as a wh-phrase:
(49) a. Feliks dumajet, [ čto Felix thinks
Liza kupi-la
knigu ].
what Liza buy-PAST book
‘Felix thinks that Liza bought a/the book.’ b. Čto
Liza kupi-la t ?
what Liza buy-PAST ‘What did Liza buy?’
In contrast, the subjunctive complementizer čtoby appears in C, as argued in chapter 4. Thus, for the purposes of the present discussion it can be viewed as a genuine complementizer. Following this line of reasoning, I propose that due to its wh-status, the indicative complementizer čto occupies SpecCP.22 In contrast, the subjunctive čtoby occupies C. The structures of the two types of embedded clauses are given in (50): 22
The idea that a wh-phrase, which appears in SpecCP can function as a complementizer is not new. For example, according to Kayne (1991:665–666), who follows Katz and Postal (1964:96) and Larson (1985:238), in English whether is a wh-phrase which occupies SpecCP, whereas if is a complementizer of the category C. Similarly, van Gelderen (2004) provides evidence that English complementizer that has developed out of the demonstrative pronoun þæt (Lockwood 1968; Traugott 1972). Furthermore, van Gelderen (2004:89–92) argues that the demonstrative complementizer first occupied the specifier position,
215 (50) a. Indicative
b. Subjunctive
CP čto
CP C`
C
C` TP…
C čtoby
TP…
In chapter 4, I suggested that Russian has no formal feature that represented indicative mood, and that the indicative complementizer čto is semantically empty. The structure in (50) is consistent with this proposal. It implies that in embedded indicative clauses, C has no content. Under this approach, the problematic status of long-distance wh-movement out of indicative clauses is explained as follows. As a TP-absorption language, Russian cannot have multiple CP specifiers.23 Assuming that wh-movement past a filled SpecCP violates subjacency, a TP-absorption language such as Russian would allow long-distance whmovement only when the single embedded SpecCP is free and may be used as an escape hatch. If in Russian the indicative complementizer čto is a wh-phrase that occupies the embedded SpecCP and not C a wh-phrase cannot move out of an embedded indicative clause because it cannot use the embedded SpecCP as an escape hatch. Such movement would lead to a subjacency violation, as illustrated in (51):
and later became a head, in accordance with the Head Preference Principle: “Be a Head rather than a Phrase” (van Gelderen 2004:18). 23
My analysis implies that unlike Serbo-Coratian, Russian cannot have multiple CP specifiers even if they are not used as landing sites (see section 5.3.2.3).
216 (51) Wh-movement out of embedded indicative clause VP V V
CP čto
C` C
TP wh-phrase
In contrast, wh-movement out of embedded subjunctive clauses is possible because the subjunctive complementizer occupies C and not SpecCP. Thus, the embedded SpecCP is free and can be used as an escape hatch, as shown in (52):
(52) Wh-movement out of embedded subjunctive clause VP V V
CP wh-phrase
C` C čtoby
TP t
This account does not make reference to the tense distinction between indicative and subjunctive clauses. Long-distances wh-movement out of indicatives is problematic
217 not because of independent tense but because of the position of the indicative complementizer.
5.4.2. French and Polish My subjacency treatment of Russian makes the following predictions about other languages. Long-distance wh-movement should be disallowed in languages with the same CP properties as Russian, that is, TP-absorption languages with whcomplementizers. I use the term wh-complementizer to refer to a complementizer, which not only has the morphological makeup of a wh-word, but also functions as a wh-word in interrogative contexts. In these languages, the presence of a wh-complementizer means that there is no SpecCP available as an escape hatch. However, long-distance whmovement should be allowed in CP-absorption languages with wh-complementizer because such languages have multiple CP specifiers.
Long-distance wh-movement
should also be allowed in TP-absorption languages with “regular”, or non-whcomplementizers under the assumption that such complementizers appear in C and not in SpecCP. In this case, SpecCP can be used as an escape hatch. In this section, I use French and Polish to test these last two predictions. Let us first consider French. To the extent that French may be a CP-absorption language, wh-movement facts from French are consistent with the proposal outlined above. French has overt wh-movement in some contexts and covert wh-movement in others. Recall that Richards’ (2001) distinction between CP- and TP-absorption languages starts off with Rudin’s (1988) proposal that in some languages with multiple
218 wh-fronting, all wh-phrases move to SpecCP, while in others, only one wh-phrase appears in SpecCP, whereas the rest adjoin to TP. Rudin (1988) observes that overt whmovement in French behaves like wh-movement in English, whereas covert whmovement behaves like that in Chinese. According to Richards (2001), English and Chinese are both CP-absorption languages. Let us suppose therefore that French is a CP-absorption language as well.24 Furthermore, the indicative complementizer in French is the wh-phrase que ‘what’. Let us suppose that by virtue of being a wh-phrase, the indicative complementizer in French occupies SpecCP. However, French differs from Russian in that it allows CP to have multiple specifiers. The prediction is that long-distance wh-movement out of indicative clauses should be possible in French because the wh-phrase can move to an additional specifier of the embedded CP. This prediction is borne out, as shown in (53):25
24
My analysis predicts that in languages with a wh-complementizer, movement past such a complementizer will pattern with extraction out wh-islands. In other words, if such a language allows extraction out of wh-islands it will also allow movement past wh-complementizer. This is the right prediction for French, which permits extraction out of wh-islands (Sportiche 1981:233ff). 25
French disallows multiple wh-movement. Recall that Richards proposes that Serbo-Croatian (a TPabsorption language) allows overt multiple SpecCPs as long as they are not used as landing sites (section 5.3.2.3). My understanding is that only one SpecCP can be used as a landing site. If we extend this proposal to French, a CP-absorption language, then French permits (i) only one TP specifier, (ii) only one SpecCP to be used as a landing site, and (iii) additional CP specifiers if they are not used as landing sites. This solution makes the right predictions. First, French disallows multiple wh-movement because it implies more than one CP specifiers used as landing sites. Second, French permits long-distance wh-movement of a single wh-phrase despite the fact that the embedded SpecCP is occupied by the wh-complementizer because additional specifiers of the embedded CP are not used as landing sites.
219 (53) a. Qui a-t-il
dit
who has-he said
[ que Marie épouserait t]? what Marie would-marry
‘Who did he say that Marie would marry?’ (Richards 2001:276; from Chang 1997:55) b.
…VP V V
CP Qui
CP que
C` C
TP Marie épouserait t
Let us now consider Polish data which at first glance appear to challenge my analysis. However, on closer inspection there are other complicating factors that must be taken into account. According to Richards’ (2001) classification, Polish, like Russian is a TP-absorption language. Moreover, the indicative complementizer in Polish is not a whphrase. The prediction is that Polish should allow long-distance wh-movement out of indicative clauses, since the embedded SpecCP is free and can be used as an escape hatch. This prediction is partially borne out. According to Rudin (1988), although certain bridge verbs permit single wh-words to be extracted, as shown in (54a), “…any extraction at all from a finite clause [in Polish] is normally ungrammatical” (p. 454). This is shown again in (54b):
220 (54) a. Co
Janek mówił, [ że studenci czytają t ]?
what Janek said
that students read
‘What did Janek say that the students read?’ b. *Co
Maria myśli, [ że Janek kupił
what Maria thinks
(Rudin 1988:454)
t ]?
that Janek bought
‘What does Maria think that Janek bought?
To the extent that sentences such as (45a) occur in Polish they confirm my prediction. The question is why sentences such as (54b) are ungrammatical. Note that Polish and Russian contrast with respect to partial wh-movement. According to Richards (2001:276), Polish allows partial wh-movement. Partial wh-movement moves a whphrase overtly into a non-scope position, and presumably covertly to its scope position. The wh-phrase in the Polish example in (55) appears within the embedded clause. However, it takes scope over the matrix clause: 26,27
(55) a. Maria myśli, [ że co Maria thinks
Janek kupił
t ]?
(Polish)
that what Janek bought
‘What does Maria think that Janek bought?’ (Richards 2001:276, from Lasnik and Saito 1984)
26
Note that the wh-phrase co ‘what’ in (55a) follows the complementizer. This is consistent with the analysis of Polish (and Russian), according to which wh-phrases undergo focus movement into a position below C. One can speculate that partial wh-movement is not available for bridge verbs such as in (54a). At this point, I do not have a solution for this problem. 27
However, Pesetsky (1987) suggests that Polish does not have LF movement (see section 5.3.1).
221 In contrast, Russian does not allow partial wh-movement, although it has local wh-movement. I distinguish between partial and local wh-movement as follows: Partial wh-movement results in a configuration where the wh-phrase receives scope interpretation in a position higher than the one that it occupies overtly. For example, it appears in the embedded SpecCP, but is interpreted as if it occupied the matrix SpecCP, just like in the Polish example (55). Local wh-movement results in a configuration where the wh-phrase is both moved and interpreted within a single clause. In (56a), the wh-phrase undergoes local wh-movement; it cannot have scope over the matrix clause. 28 The evidence comes from the fact that the matrix clause in (56a) cannot have an interrogative reading, comparable to that of (56b). To the extent that (56b) is interpretable, the wh-phrase takes scope over the matrix clause:
(56) a. Anna znajet, [ kogo Feliks poceloval t ]. Anna thinks who
(Russian)
Felix kissed
‘Anna knows who Felix kissed.’ b. ??/*Kogo Anna znajet, [ čto Feliks poceloval t ]? who Anna knows [ what Felix kissed ‘Who does Anna know that Felix kissed?’
Given the contrast between Polish and Russian with respect to partial whmovement the absence of long-distance wh-movement in Polish can be accounted for in 28
See Stepanov (1998, 2000) and van Gelderen (2001) for different views on partial wh-movement in other types of constructions in Russian.
222 terms of economy considerations: if the wh-phrase can be interpreted, or assigned the matrix scope within the embedded clause, at least overt syntactic movement to the matrix SpecCP is ruled out (cf. Chomsky’s 1995 Procrastination Principle). Since partial whmovement is allowed in Polish long-distance wh-movement is ungrammatical.
(57)
Wh-movement out of embedded indicative clauses in Polish …VP V V
CP C` C że
TP co
Covert wh-movement
TP Janek
T T
VP kupił t
Overt wh-movement In contrast, since Russian does not allow partial wh-movement the wh-phrase has to move into the matrix SpecCP.
5.4.3. Summary In this section, I proposed an account of long-distance wh-movement in Russian in terms of subjacency. I proposed that the indicative complementizer has a status of a wh-phrase
223 and as such occupies SpecCP, rather than C. As a TP-absorption language, Russian cannot have multiple CP specifiers. Thus, the only SpecCP in indicative clauses cannot be used as an escape hatch by the long-distance wh-movement. This leads to a subjacency violation, which accounts for the problematic status of wh-movement out of embedded indicatives. In contrast, the subjunctive complementizer is a genuine complementizer, which occupies C. Thus, in embedded subjunctives SpecCP is available as an escape hatch for long-distance movement of a single wh-phrase. This analysis predicts that crosslinguistically in TP- absorption languages with wh-complementizers long-distance wh-movement should be problematic. The data from French seem to be consistent with this proposal. However, more data need to be carefully examined to fully verify this prediction.
5.5. CONCLUSION I argued that the contrast between Russian indicative and subjunctive embedded clauses with respect to long-distance wh-movement cannot be captured in terms of the tense properties of such clauses. I showed that Richards’ generalization cannot be maintained, and that the SOT phenomenon cannot be considered a licensing factor for long-distance wh-movement. I proposed an alternative account of long-distance wh-movement in Russian, which relies on two components: the inability of Russian CPs to have more than one specifier and the wh-status of the indicative complementizer. According to this analysis wh-movement out of indicative clauses violates subjacency. I also examined French and Polish data and showed they appear to be consistent with my analysis.
224 CHAPTER 6 CONCLUSION AND ISSUES FOR FURTHER RESEARCH
6.0. INTRODUCTION TO THE CONCLUSION In this chapter, I conclude the dissertation. In section 6.1, I review the data dealt with in chapters 2–5 and summarize the main points of the analysis. In sections 6.2 and 6.3, I discuss theoretical implications and questions for further research, respectively.
6.1. RESEARCH QUESTIONS AND THE ANALYSIS: SUMMARY In this dissertation, I explored the morphosyntactic expression of the relationship between tense, aspect, and mood, focusing on the Russian tense system. I have examined the following aspects of Russian: (i) the temporal interpretation of Russian finite indicative clauses; (ii) the morphological past tense form in the irrealis mood; (iii) the relationship between the independent tense content and long-distance wh-movement. I have demonstrated that the temporal reference of Russian past and non-past clauses arises via different mechanisms. In the past, verbs are morphologically marked for both aspect and past tense. In the non-past, verbs are morphologically marked only for aspect, and the present/future distinction is supplied by the aspectual morphology. I showed also that the morphological past tense verb form (or the l-participle), in combination with the conditional particle by is used to form the subjunctive mood. I argued that this is the same form that receives different interpretations depending on the morphosyntactic environment. Finally, Russian indicative and subjunctive clauses
225 contrast with respect to long-distance wh-movement. Wh-movement is problematic out of embedded indicative clauses, which have independent temporal reference. In contrast, wh-movement is acceptable out of embedded subjunctive clauses, whose temporal reference is determined by the matrix clause. In chapters 3–5 of this dissertation, the following questions were addressed:
Q1: How can the relationship between tense and aspect in the non-past be represented in the morphosyntax (Chapter 3)? Q2: How can the interpretation of past tense morphology be accounted for in terms of the morphosyntactic environment (Chapter 4)? Q3: Can the independent tense content block long-distance wh-movement (Chapter 5)?
The analysis in chapters 3 and 4 is based on the idea that although tense, aspect, and mood are distinct grammatical categories which express different relations, these relations are similar in some relevant respect. I analyzed tense, aspect, and mood as dyadic predicates, each of which orders, or relates two temporal arguments in syntax. In each case, the relation between the temporal arguments can be characterized as a relation of (non-)coincidence. This relation is represented as the interpretable feature [±COIN] on C, T, and Asp. This feature serves as the basis for a semantic interpretation. In addition, C, T, and Asp each contain a purely formal feature, which licenses the temporal arguments in syntax.
226 Following Pesetsky and Torrego (2004b), I assumed that features can be valued or unvalued, and that all unvalued features must become valued to prevent the derivation from crashing. An unvalued feature can receive its value via the syntactic operation of agreement with a valued instance of the same feature. The account of the temporal interpretation of Russian finite indicative clauses relies on the presence or absence of the agree relation between the instances of the feature [COIN] on T and Asp. In the non-past, there is agreement between T and Asp, and the temporal reference is determined by aspect. I proposed that in this case, the value of the feature [COIN] on Asp is recycled, and the aspectual morphology is interpreted twice, in two syntactic locations within the same structure. In the past and future imperfective there is no agreement between T and Asp and the temporal and aspectual interpretation of such clauses are determined independently of one another. The morphosyntactic expression of the irrealis mood in Russian is also accounted for in terms of the syntactic operation of agreement between two instances of the same feature. I proposed that C and T both contain the interpretable feature [+COIN] which receives its value from the past tense morphology on the verb. In the indicative mood, this feature is interpreted on T as past. In the irrealis mood, the feature [–COIN] is interpreted on C as irrealis. Since Russian does not have verbal morphology to indicate the irrealis mood, the feature [–COIN] on C receives its value via the agreement with [– COIN]
on T. Finally, in chapter 5 I argued that the temporal properties of an embedded clause
and the (un)availability of long-distance wh-movement have nothing to do with each
227 other. I proposed an alternative account of the properties of long-distance wh-movement in Russian. The problematic status of wh-movement out of finite indicative complements is accounted for in terms of subjacency.
6.2. THEORETICAL IMPLICATIONS Chung and Timberlake (1985) define the categories tense, aspect, and mood as follows: “Tense locates an event in time. Aspect characterizes the internal temporal structure of the event. Mood describes the actuality of the event in terms such as possibility, necessity, or desirability.” (p. 202). This definition suggests that tense, aspect, and mood are semantic universals, and that the relation they express are conveyable in any natural language. Past/Non-Past, Perfective/Imperfective, and Realis/Irrealis are often regarded as the core grammatical components of the aspectual/temporal systems (Chung and Timberlake 1985; Bybee, Perkins, and Pagliuca 1994; Guéron and Hoekstra 1995; Fassi Fehri 2004). Representing tense, aspect, and mood as the same feature being interpreted on different syntactic heads provides a new way of analyzing temporal, aspectual, and mood relations, and the cross-linguistic variation in their grammatical expression. One prediction concerns the locality of value recycling. Under the analysis developed in this dissertation, value recycling proceeds locally: to receive its value, an unvalued feature F agrees with the first available valued instance of the same feature. Thus, the prediction is that agreement cannot skip heads. As has been discussed throughout the dissertation, in many languages tense and aspect, or tense and mood may be expressed by the same morphology. For example, the same form can be interpreted as
228 both present and imperfective. Similarly, past tense morphology is often used to convey either the past temporal reference or the irrealis mood, depending on a particular morphosyntactic environment. However, to the best of my knowledge, there are no languages that use aspectual morphology to express mood. Assuming that the structure in which CP dominates TP and TP dominates AspP is universal, this pattern follows from my theory: in order to interpret aspectual morphology as mood, the agreement has to take place between the relevant features on C and Asp, skipping T. Such an agreement operation would violate locality. Another prediction concerns the direction of value recycling, or reinterpretation of a morpheme. Under the analysis developed in this dissertation, value recycling can never proceed downwards. Assuming that a derivation may not “look ahead”, a feature F on a head H can “look” for a valued instance of the same feature within the lower domain, but not within the higher domain, which has not been built yet. Therefore, a morpheme interpreted on a lower head may be reinterpreted on a higher head, but never the other way around. For example, if the structure in which TP dominates AspP is universal the prediction is that there should be no languages in which tense morphology is reinterpreted on Asp and thus receives aspectual interpretation. Fassi Fehri’s (2004) analysis of Arabic temporal/aspectual system seems to challenge this prediction. Fassi Fehri argues that in Arabic a specified T can induce a specific Asp: simple past is interpreted as perfective, actual present is interpreted as imperfective, etc. This means that tense morphology is reinterpreted on Asp; that is, the
229 reinterpretation proceeds downwards. However, upon closer inspection, Fassi Fehri’s view of Arabic does not appear correct. Let us first consider the relevant data. As demonstrated in (1), in neutral (nondependent, non-embedded) contexts the form katab-a ‘he wrote’ expresses “past” and cooccurs with an appropriate deictic adverb. Note that the sentence in (1) obligatorily receives a past tense interpretation. This suggests that Arabic is a tensed language, that is, a finite indicative clause in Arabic must have temporal reference:1
(1) katab-a r-risaalat-a
(ʔamsi,
*ġad-an).
wrote-3 the-letter-ACC yesterday, *tomorrow ‘He wrote the letter (yesterday).’
(Fassi Fehri 2004:237)
In contrast, when the form katab-a is embedded under an auxiliary it can appear in past perfective or future perfective tenses, as shown in (2a) and (2b), respectively:
(2) a. kaan-a (qad)
katab-a r-risaalat-a
lammaa daxal-tu.
was-3 already wrote-3 the-letter-ACC when
entered-1
‘He had (already) written the letter when I entered.’
1
Fassi Fehri (2004:237) claims that the form katab-a can receive a perfect interpretation, as in (i):
i. katab-a r-risaalat-a (l-ʔaan-a, *ġad-an) wrote-3 the-letter-ACC now, *tomorrow ‘He has written the letter (now). It is not clear, however, whether the interpretation is a true equivalent of the English perfect or rather immediate past, which is most naturally translated into English as present perfect.
230 b. sa-yakuunu (qad) FUT-is
katab-a r-risaalat-a
ġad-an.
already wrote-3 the-letter-ACC tomorrow
‘He will have (already) written the letter tomorrow.’
(Fassi Fehri 2004:238)
Thus, the same form katab-a receives different interpretations depending on whether or not the auxiliary is present: in (1) it is interpreted as past, whereas in (2) it is interpreted as perfective. According to Fassi Fehri, Arabic distinguishes between past and non-past verbal forms. Furthermore, he argues that the suffixed tense, that is an inflected form in which person marker is suffixed, as in katab-a, must be conceived as marking past tense:
“In the Semitist literature, ST [suffixed T] has often been designated as Perf, Accompli (a French term), or Prf, and treated like Asp. But given its unique PAST meaning with a deictic past adverb, and its collocation with RT [reference time] adverbs, it is hard to see how ST can be conceived as marking primarily Asp (expressing Pfv or non-durativity), given that the latter is not sensitive to those adverb contrasts (typically Past and Fut). ST is then reasonably construed as T, whether “absolute” as in (2) [(1a)] or as “relative”, as in (3) [1b)].”
(p. 237)
Under Fassi Fehri’s analysis, the aspectual interpretation of finite clauses is secondary to their temporal interpretation. However, (2b) (Fassi Fehri’s 6) is problematic for this claim: if katab-a means “past” it is not expected to co-occur with the future tense auxiliary. Let us suppose then, following the apparently more traditional view, that a
231 suffixed inflected form designates perfective aspect, and not past tense.2 In light of the theory developed in this dissertation, this means that katab-a comes from the lexicon with an uninterpretable valued feature [+PERF]; this feature receives its interpretation via agreement with the unvalued instance of the same feature on Asp, which in turn licenses the semantic feature [–COIN]. Now, the difference in the interpretation of katab-a in (1) and (2) can be captured as follows. In (1), katab-a is the only morphological form. Therefore, it is reinterpreted on T and as a result the perfective aspect gives rise to a past tense interpretation, a phenomenon attested cross-linguistically. In contrast, in (2) the overt auxiliary is present in T. As a result, the form katab-a is only interpreted on Asp. If katab-a is unspecified for tense, it can freely co-occur with either past or future auxiliaries.
6.3. FUTURE RESEARCH One issue not addressed in this dissertation concerns the relationship between tense and structural case, which has been extensively discussed in the literature (Brody 1997; Stowell 1995, 1996; Zagona 1995; 1998; Pesetsky and Torrego 2001, 2004a,b). In particular, Pesetsky and Torrego (2004a,b) argue that structural case is an instance of an uninterpretable tense feature, or u[TENSE], on a DP. For example, the nominative case on the subject DP is an instance of u[TENSE] that has its interpretable counterpart on T. Under the present analysis, however, there is no formal feature [TENSE]. Instead, T contains the feature [PAST], which receives its value from the verbal morphology. If, as 2
Fassi Fehri (2004) cites Cohen (1989) as a recent work that defends the “pure aspect” view of Semitic languages.
232 proposed by Pesetsky and Torrego, the structural case on the subject DP is an unvalued instance of the formal feature on T, the subject DP is expected to carry an unvalued instance of the feature [PAST]. Further, if this feature on the subject DP receives its value via agreement with the valued feature [PAST] on T, the case properties of subjects are expected to reflect the temporal properties of the clause. Just as the negative and the positive value of this feature on T results in the non-past or past tense interpretation, respectively, the negative and the positive value of the same feature on the subject DP should result in distinct case marking. This is obviously not true. Crosslinguistically, the case marking on the subject tend to remain the same across all temporal references, available in a given language. This observation suggests that the relationship between tense and structural case is more complex and requires further investigation. Another issue worth exploring in light of the proposed analysis concerns possible domains of interpretability. It is proposed in this thesis that by virtue of value recycling, Russian aspectual morphology is interpreted twice within the same syntactic structure: on Asp as perfective or imperfective and then on T as present or future. In this particular case, the double interpretation of a morpheme takes place within the functional domain, that is, the domain above VP. The question is, can the same morpheme be interpreted in both the lexical and the functional domains? The case I have in mind concerns the process of prefixation in Russian, addressed briefly in chapter 2. Russian has about twenty verbal prefixes, which have developed out of directional adverbs and prepositions. Most of such prefixes still retain some of their directional meaning. When combined with a morphologically simple imperfective verb,
233 they make it perfective and future. However, in most cases the same prefix also affects the meaning of the verb. This is illustrated below:
(3)
pišet ‘write-3RD.SG(IMPF/PRES)’
a. pod-pišet ‘sign-3RD.SG(PERF/FUT)’ b. za-pišet ‘jot down-3RD.SG(PERF/FUT)’ c. vy-pišet ‘write out-3RD.SG(PERF/FUT)’
The data in (3) suggest that Russian prefixes have both inflectional and derivational properties, which has been a problem for previous analyses of prefixation.3 The analysis developed in this dissertation has consequences for the interpretation of prefixes. If the same prefix can be interpreted on Asp and T, perhaps the same prefix can also be interpreted in the third location: in the VP domain. The idea is that prefixes combine two meaning components: the lexical component and the inflectional component. The prefix combines with the verb in the lexicon, and that’s where its lexical meaning component is interpreted. However, its functional meaning component, that is, aspectual and temporal information, cannot be interpreted unless the prefixed verb becomes a part of the syntactic structure. To establish the validity of this idea, one has to define the domain of interpretability. Can the same morpheme be interpreted in both the lexical and the functional domain? This question is related to the notion of phase: Assuming that VP (or 3
Filip (2000:78) explicitly suggests that it would be undesirable for a morpheme to be classified as both derivational and inflectional. According to Spencer (1991:196), this would be “a contradiction in terms”. However, neither Filip (2000) nor Spencer (1991) gives any explicit arguments for why this would be an undesirable result, apart from the fact that this would contradict the commonly accepted view.
234 vP) is a phase, this proposal implies that the same morpheme can receive its interpretation across the phase boundary. This, in turn, would have consequences for the theory of phases, and in particular, for the phase impenetrability condition (Chomsky 2001). At the moment, I cannot offer well-articulated answers to these questions. This issue will provide an exciting topic for future research.
235 REFERENCES Abush, Dorit. 1997. Sequence of tense and temporal De Re. Linguistics and Philosophy 20.1–50. Abush, Dorit. 1994. Sequence of tense revisited: Two semantic account of tense in intensional contexts. Ellipsis, tense and questions, ed. by Hans Kamp, 87–139. Amsterdam: DYANA deliverable R2.2.B, University of Amsterdam. Abush, Dorit. 1988. Sequence of tense, intensionality, and scope. Proceedings of 7th West Coast Conference on Formal Linguistics, ed. by Hagit Borer, 1–14. Stanford, CA: CSLI Publications. Ackerman, Farrell, and Gert Webelhuth. 1998. A theory of predicates. Stanford, CA: CSLI Publications. Avrutin, Sergey, and Maria Babyonyshev. 1997. Obviation in subjunctive clauses and AGR: Evidence from Russian. Natural Language and Linguistic Theory 15.229–262. Bailyn, John F. 2001. On scrambling: A reply to Bošcović and Takahashi. Linguistic Inquiry 32.635–658. Bailyn, John F. 1995. A configurational approach to Russian ‘‘free’’ word order. Doctoral dissertation, Cornell University. Binnick, Robert. 1991. Time and the verb. New York: Oxford University Press. Beninca, Paola, and Cecilia Poletto. 2004. Topic, focus, and V2. The Structure of CP and IP, ed. by Luigi Rizzi, 52–75. New York: Oxford University Press. Borik, Olga. 2002. Aspect and reference time. Doctoral dissertation, Utrecht University.
236 Borik, Olga. 2001. Aspect in Russian: Different perspectives. A colloquia talk given at the University of Calgary, April 6, 2001. Borik, Olga, Paz González, and Henk Verkuyl. 2003. Comparing tense systems: The primacy of the Pres/Past opposition. Nordlyd 31.13–29 (Proceedings of the 19th Scandinavian Conference of Linguistics, ed. by Anne Dahl, Kristine Bentzen, and Peter Svenonius). Bošković, Željko. 2004. Topicalization, focalization, lexical insertion, and scrambling. Linguistic Inquiry 35.613–638. Bošković, Željko. 2002. On multiple wh-fronting. Linguistic Inquiry 33.351–383. Bošković, Željko. 1997. The syntax of nonfinite complementation: An economy approach. Cambridge, MA: MIT Press. Bošković, Željko. 1995. Superiority effects with multiple wh-fronting in Serbo-Croatian. Ms., University of Connecticut. Boyd, John. 1992. Exceptions to island constraints and syntactic theory. Doctoral dissertation, University of Massachusetts, Amherst. Branigan, Philip. 2004. The syntax of π: A study of subjects and A-bar movement constraints. Ms., Memorial University. Brecht, Richard. 1977. ČTOBY or ČTO and BY. Folia Slavica 1.33–41. Brody, Michael. 1997. Perfect chains. Elements of grammar, ed. by Liliane Haegeman, 139– 167. Dordrecht: Kluwer. Bybee, Joan, Revere Perkins, and William Pagliuca. 1994. The evolution of grammar: Tense, aspect, and modality in the languages of the world. Chicago: The University of Chicago Press.
237 Chomsky, Noam. 2001. Derivation by phase. Ken Hale: A life in language, ed. by Michael Kenstowicz. Cambridge, MA: MIT Press. Chomsky, Noam. 2000. Minimalist Inquiries: The Framework. Step by step: Essays on minimalist syntax in honor of Howard Lasnik, ed. by Roger Martin, David Michaels, and Juan Uriagereka, 89–155. Cambridge, MA: MIT Press. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Chang, Lisa. 1997. Wh-in-situ phenomena in French. MA thesis, University of British Columbia. Chung, Sandra, and Alan Timberlake. 1985. Tense, aspect, and mood. Language typology and syntactic description, vol. 3: Grammarical categories and the lexicon, ed. by Timothy Shopen, 202–258. Cambridge: Cambridge University Press. Cohen, David. (1989). L’aspect verbal. Paris: Presses Universitaires de France. Comrie, Bernard. 1986. Tense in indirect speech. Folia Linguistica 20.265–296. Comrie, Bernard. 1985. Tense. Cambridge: Cambridge University Press. Comrie, Bernard. 1980. Clause structure and movement constraints in Russian. Morphosyntax in Slavic, ed. by Catherine V. Chvany and Richard D. Brecht, 98–113. Columbus, Ohio: Slavic Publishers. Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press. Cowper, Elizabeth. 2005. The geometry of interpretable features: Infl in English and Spanish. Language 81.10–46. Cowper, Elizabeth. 2002. Finiteness. Ms., University of Toronto. Dahl, Östen. 1985. Tense and aspect systems. Oxford: Blackwell.
238 Dahl, Östen. 1997. The relation between past time reference and counterfactuality: A new look. On conditionals again, ed. by Angeliki Athanasiadou and René Dirven, 97–114. Amsterdam: John Benjamins. Demirdache, Hamida, and Myriam Uribe-Etxebarria. 2000. The primitives of temporal relations. Step by step. Essays on minimalist syntax in honor of Howard Lasnik, ed. by Roger Martin, David Michaels, and Juan Uriagereka, 157–186. Cambridge, MA: MIT Press. Doron, Edit, and Caroline Heycock. 1999. Filling and licensing multiple specifiers. Specifiers: Minimalist approaches, ed. by David Adger, Susan Pintzuk, Bernadette Plunkett, and George Tsoulas, 69–89. Oxford: Oxford University Press. Enҫ, Mürvet. 1996. Tense and modality. The handbook of contemporary semantic theory, ed. by Shalom Lapin, 345–358. Oxford: Blackwell. Enҫ, Mürvet. 1987. Anchoring conditions for tense. Linguistic Inquiry 18.633–657. Fassi Fehri, Abdelkader. 2004. Temporal/aspectual interaction and variation across Arabic heights. The syntax of time, ed. by Jacqueline Guéron and Jacqueline LeCarme, 235–258. Cambridge, MA: MIT Press. Filip, Hana. 2000. The quantization puzzle. Events as grammatical objects, ed. by Carol Tenny and James Pustejovsky, 39–96. Stanford, CA: CSLI. Forsyth, James. 1970. A grammar of aspect: Usage and meaning in the Russian verb. Cambridge: Cambridge University Press.
239 Franks, Steven, and Gerald Greenberg. 1994. The functional structure of Slavic clauses. Proceedings of the 1st Annual Meeting on Formal Approaches to Slavic Linguistics (Ann Arbor 1992), ed. by Jindrich Toman, 77–108. Ann Arbor: Michigan Slavic Publications. Franks, Steven, and Gerald Greenberg. 1988. Agreement, tense, and case of subjects. Chicago Linguistic Society 24.71–86. Giorgi, Alessandra, and Fabio Pianesi. 1997. Tense and aspect: From semantics to morphosyntax. New York: Oxford University Press. Givón, T. 1994. Irrealis and the subjunctive. Studies in Language 18.265–337. Guéron, Jacqueline, and Teun Hoekstra. 1995. The temporal interpretation of predication. Small clauses (Syntax and Semantics 28), ed. by Anna Cardinaletti and Maria Teresa Guasti, 77–108. San Diego, CA: Academic Press. Hacking, Jane F. 1998. Coding the hypothetical: A comparative typology of Russian and Macedonian conditionals. Amsterdam/Philadelphia: John Benjamins. Hale, Ken. 1986. Notes on world view and semantic categories: Some Warlpiri examples. Features and projections, ed. by Pieter Muysken and Henk van Riemskijk, 233–254. Dordrecht: Foris. Hornstein, Norbert. 1990. As time goes by. Cambridge, MA: MIT Press. Horvath, Julia. 1986. FOCUS in the theory of grammar and the syntax of Hungarian. Dordrecht: Foris. Jakobson, Roman. 1957/1984. Shifters, verbal categories and the Russian verb. Russian and Slavic grammar: Studies 1931–1981, ed. by Linda R. Waugh and Morris Halle, Berlin: Mouton.
240 Jakobson, Roman. 1932/1971. Zur Struktur des russischen Verbums. Selected Writings, vol. 1. 3–15. James, Deborah. 1982. Past tense and the hypothetical: A cross-linguistic study. Studies in Language 6.375–403. Johnson, Kyle. 1985. Some notes on subjunctive clauses and binding in Icelandic. Papers in Theoretical and Applied Linguistics (MIT Working Papers in Linguistics 6), ed. by Diane Archangeli, Andrew Barss, and Richard Sproat, 102–134 Cambridge, MA: MIT Press. Junghanns, Uwe. 1997. On BYT’ (and BYTI). Formale Slavistik (Leipzig Schriften zur Kultur-, Literatur,- Sprach- und Ubersetzungswissenschaft, 7), ed. by Uwe Junghanns and Gerhard Zybatow, 251–265. Frankfurt am Main: Ververt Verlag. Iatridou, Sabine. 2000. The grammatical ingredients of counterfactuality. Linguistic Inquiry, 31.231–270. Ivanov, V. V. 1964. Istoričeskaja grammatika russkogo jazyka (Historical grammar of Russian). Moscow: Prosveščenie. Kamp, Hans, and Uwe Reyle. 1993. From discourse to logic. Dordrecht: Kluwer. Katz, J. J., and Paul M. Postal. 1964. An integrated theory of linguistic description. Cambridge, MA: MIT Press. Kayne, Richard. 1994. The antisymmetry in syntax. Cambridge, MA: MIT Press. Kayne, Richard. 1991. Romance clitics, verb movement, and PRO. Linguistic Inquiry 22.641–686. Kiss, K. É. 1995. Introduction. Discourse configurational languages, ed. by Katalin É. Kiss. 3–27. New York: Oxford University Press.
241 Klein, Wolfgang. 1994. Time in language. London: Routledge. Klein, Wolfgang. 1995. A time-relational analysis of Russian aspect. Language 71.669–695. Kondrashova, Natalia. 1999. Barenaked tenses and their morphological outfits. Proceedings of the 29th Meeting of the North East Linguistics Society (University of Delaware 1998), ed. by Pius Tamanji, Masako Hirotani, and Nancy Hall, 183–195. Amherst, MA: Graduate Linguistic Student Association, University of Massachusetts. Kratzer, Angelika. 1998. More structural analogies between pronouns and tenses. Proceedings of Semantics and Linguistic Theory VIII. Ithaca, NY: Cornell University, CLC Publications. Kusumoto, Kiyomi. 1999. Tense in embedded contexts. Doctoral dissertation, University of Massachusetts, Amherst. Larson, Richard K. 1985. On the syntax of disjunction scope. Natural Language and Linguistic Theory 3.217–264. Lasnik, Howard, and Mamoru Saito. 1984. On the nature of proper government. Linguistic Inquiry 15.235–189. LeCarme, Jacqueline. 2004. Tense in nominals. The syntax of time, ed. by Jacqueline Guéron and Jacqueline LeCarme, 441–476. Cambridge, MA: MIT Press. LeCarme, Jacqueline. 1999. Nominal tense and tense theory. Empirical issues in formal syntax and semantics 2: Selected papers from the Colloque de Syntaxe et Semantique Paris (CSSP), ed. by F. Corblin, C. Dobrovie-Sorin, and J.-M. Marandin. The Hague: Holland Academic Graphics.
242 Lin, Jo-Wang. 2005. Time in a language without tense: The case of Chinese. Journal of Semantics 23.1–53. Lockwood, William. 1968. Historical German syntax. Oxford: Clarendon. Mahajan, Anoop. 1990. The A/A-bar distinction and movement theory. Doctoral dissertation, MIT. Matthews, W. K. 1960. Russian historical grammar. Oxford: Oxford University Press. Matthewson, Lisa. 2002. On universality and variation in tense systems. A colloquia talk given at the University of Calgary, October 4, 2002. Mezhevich, Ilana. 2006a. The morphological status of Georgian and Russian preverbs: A grammaticalization approach. Paper presented at the 2006 Annual LSA Meeting, Albuquerque, NM, January 5–8, 2006. Mezhevich, Ilana. 2006b. Aspect and complex predicates: Delimiting events in morphology. Paper presented at the 11th Biennial Symposium, Rice Linguistic Society: Intertheoretical Approaches to Complex Verb Constructions, Poster Session: Complex Predicates and Similar Constructions. Rice University, Houston, TX, March 16–18, 2006. Miklosich, Franz. 1883. Vergleichende Grammatik der slavischen Sprachen, vol. 4, Wien. Moore, John, and David M. Perlmutter. 1999. Case, agreement, and temporal particles in Russian infinitival clauses. Journal of Slavic Linguistics 7.219–246. Musan. Renate. 1995. On the temporal interpretation on noun phrases. Doctoral dissertation, MIT. Ogihara, Toshiyuki. 1996. Tense, scope, and attitude ascription. Dordrecht: Kluwer.
243 Paducheva, Elena. 1995. Taxonomic categories and semantics of aspectual opposition. Temporal reference, aspect, and actionality, vol. 1: Semantic and syntactic perspectives, ed. by Pier Marco Bertinetto, Valentina Bianchi, James Higginbotham, and Mario Squartini, 71–89. Torino, Italy: Rosenberg and Sellier. Palmer, Frank R. 1986. Mood and modality. Cambridge: Cambridge University Press. Perlmutter, David. 1971. Deep and surface constraints in syntax. New York: Holt, Rinehart and Winston. Pesetsky, David. 1987. Wh-in-situ: Movement and unselective binding. The representation of (in)definiteness, ed. by Eric Reuland and Alice ter Meulen, 98–129. Cambridge, MA: MIT Press. Pesetsky, David. 1982a. Complementizer-Trace phenomena and the nominative island condition. The Linguistic Review 1.297–343. Pesetsky, David. 1982b. Paths and categories. Doctoral dissertation, MIT. Pesetsky, David, and Esther Torrego. 2004a. Tense, case, and the nature of syntactic categories. The syntax of time, ed. by Jacqueline Gueron and Jacqueline LeCarme, 495– 538. Cambridge, MA: MIT Press. Pesetsky, David, and Esther Torrego. 2004b. The syntax of valuation and the interpretability of features. Ms., MIT, UMass/Boston. Pesetsky, David, and Esther Torrego. 2001. T-to-C movement: Causes and consequences. Ken Hale: A Life in Language, ed. by Michael Kenstowicz, 355–426. Cambridge, MA: MIT Press.
244 Picallo, Carmen. 1984. The Infl node and the null subject parameter. Linguistic Inquiry 15.75–102. Poletto, Cecilia. 1995. Complementizer deletion and verb movement in Italian. University of Venice Working Papers in Linguistics, 5.1–5. Progovac, Ljiljana. 1993. Subjunctive: The (mis)behavior of anaphora and negative polarity. Linguistic Review 10.37–59. Reichenbach, Hans. 1947. Elements of symbolic logic. New York: Macmillan. Richards, Norvin. 2001. Movement in language. Oxford: Oxford University Press. Ritter, Elizabeth, and Martina Wiltschko. 2005. Anchoring events to utterance time without tense. Proceedings of the 24th West Coast Conference on Formal Linguistics, ed. by John Alderete, Chung-hye Han, and Alexei Kochetov, 343–351. Somerville, MA: Cascadilla. Rizzi, Luigi. 2004. On the cartography on syntactic structures. The structure of CP and IP (The cartography of syntactic structures, vol. 2), ed. by Luigi Rizzi, 3–15. New York: Oxford University Press. Rizzi, Luigi. 1997. The fine structure of the left periphery. Elements of grammar: Handbook of generative syntax, ed. by Liliane Haegeman, 281–337. Amsterdam: Kluwer. Rizzi, Luigi. 1982. Issues in Italian syntax. Dordrecht: Foris. Rochette,
Anne.
1988.
Semantic and
syntactic aspects of
Romance sentential
complementation. Doctoral dissertation, MIT. Ross, John Robert. 1967. Constraints on variables in syntax. Doctoral dissertation, MIT. Rudin, Catherine. 1988. On multiple questions and multiple wh-fronting. Natural Language and Linguistic Theory 6.445–501.
245 Saito, Mamoru. 1989. Scrambling as semantically vacuous A-bar movement. Alternative conceptions of phrase structure, ed. Mark Baltin and Anthony Kroch. 182–200. Chicago: University of Chicago Press. Saito, Mamoru. 1992. Long-distance scrambling in Japanese. Journal of East Asian Linguistics 1.69–119. Schoorlemmer, Maaike. 1995. Participial passive and aspect in Russian. Doctoral dissertation, Utrecht University. Sekerina, Irina. 1997. The syntax and processing of split scrambling constructions in Russian. Doctoral dissertation, CUNY. Smith, Carlotta. 1991. The parameter of aspect. Dordrecht: Kluwer. Smith, Carlotta, and Gilbert Rappaport. 1991. The aspectual system of Russian. The parameter of aspect by Carlotta Smith, Chapter 10. Spencer, Andrew. 2001. The paradigm-based model of morphosyntax. Ms., University of Essex. Spencer, Andrew. 1991. Morphological theory. An introduction to word structure in generative grammar. Oxford, MA: Basil Blackwell. Spencer, Andrew and Marina Zaretskaya. 1998. Verb prefixation in Russian as lexical subordination. Linguistics 36.1–39. Sportiche, Dominique. 1981. Bounding nodes in French. The Linguistic Review 1.219–246. Stepanov, Arthur. 2000. Wh-scope marking in Slavic. Studia Linguistica 54.1–40.
246 Stepanov, Arthur. 1988. On wh-fronting in Russian. Proceedings of the 28th North East Linguistic Society, ed. by Pius N. Tamanji and Kiyomi Kusumoto, 453–467. Amherst, MA: University of Massachusetts, GLSA. Stowell, Timothy. 1996. The phrase structure of tense. Phrase structure and the lexicon, ed. by Johan Rooryck and Laurie Zaring, 277–291. Dordrecht: Kluwer. Stowell, Timothy. 1995. What is the meaning of the present and past tenses? Temporal reference, aspect, and actionality, vol, 1: Semantics and syntactic perspectives, ed. by Pier Marco Bertinetto, Valentina Bianchi, James Higginbotham, and Mario Squartini, 381–396. Torino, Italy: Rosenberg and Sellier. Stowell, Timothy. 1981. The origins of phrase structure. Doctoral dissertation, MIT. Švedova, Natalija Ju. 1980. Russkaja grammatika (Russian grammar). Moscow: Nauka. Swift, Mary D. 2004. Time in child Inuktitut. A developmental study of an Eskimo-Aleut language. Berlin: Mouton de Gruyter. Terzi, Arhonto. 1992. PRO in finite clauses: A study of the inflectional heads of the Balkan languages. Doctoral dissertation, CUNY. Timberlake, Alan. 1985. Reichenbach and Russian aspect. The scope of Slavic aspect, ed. by Michael S. Flier and Alan Timberlake, 153–168. Columbus, Ohio: Slavic Publishers. Traugott, Elizabeth Closs. 1972. The history of English syntax: A transformational approach to the history of English sentence structure. New York: Holt, Rinehart and Winston. Tynan, John, and Eva Delgado Lavín. 1997. Mood, tense and the interpretation of conditionals. On conditionals again, ed. by Angeliki Athanasiadou and René Dirven, 115–142. Amsterdam: John Benjamins.
247 van Gelderen, Elly. 2004. Grammaticalization as economy. Amsterdam: John Benjamins. van Gelderen, Véronique. 2001. Partial wh-movement in Russian. Linguistics in the Netherlands 2001.89–100. van Schooneveld, C. H. 1951. The aspect system of the Old Church Slavonic and Old Russian verbum finitum byti. Word 7.96–103. Varlakosta, Spyridoula. 1994. Issues on Modern Greek sentential complementation. Doctoral dissertation, University of Maryland. Vendler, Zeno. 1967. Linguistics and philosophy. Ithaca, NY: Cornell University Press. Verkuyl, Henk. 2005. How (in-)sensitive is tense to aspectual information? Crosslinguistic views on tense, aspect and modality, ed. by Bart Hollebrandse, Angeliek van Hout, and Co Vet, 145–169. Amsterdam: Rodopi. Vinogradov, V. V. 1947/1971. Russkij jazyk (The Russian language). Moscow: Vysšaja Škola. Willis, David. 2000. Verb movement and Slavonic conditionals. Diachronic syntax: Models and mechanisms, ed. by Susan Pintzuk, George Tsoulas, and Anthony Warner, 322–348. Oxford: Oxford University Press. Zagona, Karen. 2003. Tense and anaphora: Is there a tense-specific theory of coreference? Anaphora: A reference guide, ed. by Andrew Barss, 140–171. Oxford: Blackwell. Zagona, Karen. 1998. Structural case and tense construal. Formal perspectives on Romance linguistics, ed. by J.-Marc Authier, Barbara E. Bullock, and Lisa A. Reed, 305–327. Amsterdam: John Benjamins.
248 Zagona, Karen. 1995. Temporal argument structure: Configurational elements of construal. Temporal reference, aspect, and actionality, vol. 1: Semantic and syntactic perspectives, ed. by Pier Marco Bertinetto, Valentina Vianchi, James Higginbotham, and Mario Squartini, 397–410. Torino, Italy: Rosenberg and Sellier. Zagona, Karen. 1990. Times and temporal argument structure. Ms., University of Washington, Seattle. Zalianjak, Anna, and Alexej Shmelev. 1997. Lekcii po russkoj aspektologii (Lectures on Russian aspectology). München: Verlag Otto Sagner.
