Subjects and Universal Grammar: An Explanatory Theory available in Hardcover
Subjects and Universal Grammar: An Explanatory Theory
- ISBN-10:
- 0521858542
- ISBN-13:
- 9780521858540
- Pub. Date:
- 08/10/2006
- Publisher:
- Cambridge University Press
- ISBN-10:
- 0521858542
- ISBN-13:
- 9780521858540
- Pub. Date:
- 08/10/2006
- Publisher:
- Cambridge University Press
Subjects and Universal Grammar: An Explanatory Theory
Hardcover
Buy New
$130.00Overview
Product Details
| ISBN-13: | 9780521858540 |
|---|---|
| Publisher: | Cambridge University Press |
| Publication date: | 08/10/2006 |
| Series: | Cambridge Studies in Linguistics , #113 |
| Edition description: | First Edition |
| Pages: | 258 |
| Product dimensions: | 5.98(w) x 9.02(h) x 0.75(d) |
About the Author
Yehuda N. Falk is Senior Lecturer in the Department of English, The Hebrew University of Jerusalem.
Read an Excerpt
Cambridge University Press
0521858542 - Subjects And Universal Grammar - An Explanatory Theory - by Yehuda N. Falk
Excerpt
1 On subjects and explanation
1.1 Overview
Explaining subjects and their properties is an important challenge in contemporary linguistics. For formalist approaches to linguistics, the clustering of properties that subjects display necessitates some special representational properties unique to subjects. Without such representational uniqueness, the properties of subjects that set them apart from other elements of the clause are mysterious. However, this only pushes the need for explanation back one level: such special representation itself calls out for explanation. For functionalist approaches, similar issues are raised, as it is not clear what the functional properties of subjects are that set them apart. From a typological perspective, the mystery of subjects is even deeper, as different language types appear to deploy subject properties in different (but systematic) ways. As a result of the discoveries of ergative languages, Philippine-type languages, active languages, and the like, interesting questions have been raised about the properties of subjects, the representation of subjects, and even the cross-linguistic validity of “subject” as an element of linguistic description.
The concept of “subject” is one with a long history in linguistics. As with most othersuch concepts, contemporary linguistics did not invent the subject. Instead, it has taken a traditional concept and attempted to provide it with theoretical content. Problems have arisen because the concept “subject” originates in traditional studies of classical Indo-European languages such as Greek and Latin, languages which are closely related genetically, areally, and typologically. Investing “subject” with theoretical content thus usually depends on either focusing on languages which are typologically similar to classical Indo-European languages or attempting to extend an Indo-European notion to languages which have very different typological properties. As a result, different researchers take varying positions on which languages are examined, and in some languages which element (if any) is to be identified as the subject. Much of the literature on such topics as ergativity and active languages focuses on debates such as these. These issues need to be clarified if a true understanding of subjects and their properties is to be achieved.
All contemporary approaches to linguistics – formalist, functionalist, typological, etc. – appropriately take the goal of linguistics to be the explanation of linguistic phenomena. As such, they depart from merely being satisfied with describing linguistic facts, although proper description is, of course, a prerequisite for explanation. In the realm of subjecthood, this means that simply stipulating the properties of subjects is not sufficient: the properties should follow from a proper characterization of the nature of subjects. Since explanation is possible only in the context of a theory of the linguistic domain in question, the attempts that have been made at explaining subjects have been as varied as schools of linguistics, and have mirrored the drawbacks of the theoretical assumptions made by the researchers. Formal accounts tend to be characterized by a disregard for functional factors and often by inadequate cross-linguistic coverage. Functionalist and typological accounts are typically based on superficial surveys of languages and disregard the nature of the formal devices involved in syntax.
It is the thesis of this study that a truly explanatory theory of subjects has yet to be constructed, and its goal is the proposal of such a theory. A theory of subjects must be formally grounded, functionally aware, and achieve sufficiently broad typological coverage, including all of the types of languages which are potentially problematic. Unlike previous accounts, the theory of subjects to be proposed here meets all of these criteria. Naturally, it draws on insights of earlier approaches, but it synthesizes them in a way which results in true explanation of the properties of subjects as they are revealed in cross-linguistic study.
In this first chapter, we will enumerate the properties generally thought to be subject properties. We will also discuss typological issues related to subjecthood. Finally, we will discuss different types of approaches to subjects.
1.2 Subject properties
1.2.1 First approximation
As mentioned above, subjects display an array of properties which must be accounted for by a theory of subjecthood. Properties of subjects have been enumerated in studies like Keenan (1976) and Andrews (1985). We will review them here briefly, primarily using examples from English. However, before we discuss the properties of subjects, it is necessary to take heed of the following observation by Andrews (1985: 104):
At the outset we must note that there are no properties which in all languages are always exhibited by subjects and only exhibited by them. There may be some properties that are universally restricted to subjects [fn. omitted], but there are certainly none that they always have. Rather, we find properties that are exhibited by subjects in a wide range of languages, and which may be plausibly argued to be restricted to subjects in some of them.
This observation is not surprising – it is in line with the way typological properties typically apply (Comrie 1989). However, it violates the usual formalist preference for absolute universals, and thus is an important caveat for any formally based theory of subjects. In addition, the fact that typological properties typically emerge as tendencies rather than absolutes is itself something that needs to be explained.
The first property is that if a verb has an Agent argument, the Agent is realized as subject.
| (1) |
|
A verb like the putative eat´ in (1c), in which the Patient is realized syntactically as subject and the Agent as object, is disallowed. Of course, while all Agent arguments are subjects, not all subjects are Agents. If the verb does not have an Agent argument, the subject will express some other thematic role. A special case of this is the passive construction, in which the Agent loses its expressed-argument status (Chomsky 1981, Bresnan 2001).
Another property of subjects is that the addressee of an imperative is a subject. This can be seen in each of the following imperatives: the addressee can have a variety of thematic roles, not necessarily Agent, but it must have the syntactic status of subject.
| (2) |
|
Another property which is apparent in the English imperative examples, although more clearly in other languages, is that the subject is more susceptible to being realized as a covert (null or empty) pronoun. It is telling that the empty-pronoun construction (or pro-drop) is often referred to in the theoretical literature as the null-subject construction, a name which is based on this higher susceptibility. We will discuss the facts in more detail in Chapter 2.
A frequently discussed property of subjects is anaphoric prominence. The exact details vary from language to language (as will be discussed in Chapter 2), but one clear consequence which can be seen in all languages with reflexive pronouns1 is that, in a transitive clause in which the subject and object are coreferential, it is the subject which is expressed as a full NP and the object as the reflexive pronoun.
| (3) |
|
In some languages the antecedent of a reflexive must be a subject, while in others (like English) it just has to have higher prominence, but in either case the most prominent element of the clause is the subject.
An anaphoric construction which does not exist in English, but in which the greater prominence of the subject is again apparent, is the switch-reference construction, in which a clause marks the anaphoric relation (coreference or disjoint reference) between its own subject and the subject of a superordinate and/or coordinate clause. This is exemplified in the following Diyari sentences (Austin 1981).
| (4) |
|
In (4a), the clauses have coreferential subjects, so the “same” morpheme is used in the subordinate clause. In (4b), on the other hand, the subjects are disjoint in reference, and the “different” morpheme is used.
Even in a language like English, which has no switch-reference system, subjects have a special status in coordination and subordination. In coordinated clauses, if the subjects of both clauses are identical, the subject can be omitted in the second clause. The object cannot be involved in this kind of relation.
| (5) |
|
More frequently discussed in the theoretical literature is the subordination construction known as control (or equi). In the control construction, the subordinate subject is covert (and modeled as a special null nominal called PRO in the transformational literature) if it is identical to an element of the main clause. While the coreferential main clause element need not be subject, the subordinate clause element must be.
| (6) |
|
A similar construction is raising,2 in which an element which is a thematic (semantic) argument of the subordinate clause is expressed as part of the main clause, in which it is not a thematic argument. The only kind of subordinate clause element which can be raised in this fashion is the subject.
| (7) |
|
Coordination, control, and raising are thus constructions in which the subject has a special status.
Subjecthood interacts in various ways with long-distance (wh) dependencies. One of the best-known cases is the fact, originally observed in Keenan and Comrie’s (1977) classical study of relative clauses, that subjects are more prone to wh-movement cross-linguistically than other elements. In English, paradoxically, subjects appear to be more resistant to wh-movement than other elements of the clause: non-subjects can be extracted from a clause with an overt complementizer while subjects cannot (the that-trace effect).
| (8) |
|
There are other subjecthood-related aspects to long-distance dependency constructions, to be discussed in detail in Chapter 4.
There are other properties that are unique to subjects. For example, many languages require every sentence to have a subject (either overt or covert), a property enshrined in transformational theory’s Extended Projection Principle (and analogous principles in other contemporary theories of syntax). Another property which has been built into the transformational model is that subjects often occupy a special “external” structural position (e.g., outside of VP), a position which provides them with structural prominence relative to other arguments of the verb. Subjects also have semantic and pragmatic prominence. For example, subjects are often definite. They also take wide scope over other elements of the clause.
| (9) |
|
Finally, the subject is usually the discourse topic.
We can summarize these subject properties in the following list.
| (10) | Agent argument in the active voice Most likely covert/empty argument The addressee of an imperative Anaphoric prominence Switch-reference systems Shared argument in coordinated clauses Controlled argument (PRO) Raising Extraction properties Obligatory element “External” structural position Definiteness or wide scope Discourse topic |
This catalog of properties3
represents the reason for the continued interest in the nature of subjects. There is no obvious pretheoretical reason for a single element of the sentence to have all these properties; the fact that one does in many languages calls out for explanation.
1.2.2 Case and subjects
To sharpen the conception of subject properties that we outlined in the previous section, we need to consider the relationship between subjecthood and morphological marking: Case4 and, to a lesser extent, agreement. Subjects in many languages are realized with either no overt Case marking or with the same Case marking that is used with citation forms, two situations we can unify under the heading “unmarked Case.” This unmarked Case, often called nominative, is sometimes taken to be a defining property of subjects in Case-marking languages. However, typological study has shown that this is an overly simplistic view of the situation. We will outline the relevant facts in this section. Similarly, it is often stated that subjects have a special status in terms of agreement. Here again, the facts appear to be more complicated. We will return to the question of Case and agreement in Chapter 3.
Since we need to be able to refer to clausal participants without committing to their status as subjects and objects, we will follow much of the typological literature (see, for example, Comrie 1978, 1989) in using the following terminology:
| (11) | Sole argument of intransitive verb: S Agent-like argument of transitive: A Patient-like argument of transitive: P (sometimes called O) |
The most common (and most familiar) situation is one where S and A have unmarked Case (traditionally called nominative) and P has a marked Case (traditionally called accusative). Such a language is called nominative-accusative. In a nominative-accusative language, the traditional identification of S/A as “subject” and the hypothesis that subjects have the unmarked Case coincide.
However, in some languages, it is the S and P that have unmarked Case. The S/P unmarked Case is usually called absolutive rather than nominative. In this kind of language, the A has a marked Case which is called ergative. For this reason, these languages are usually referred to as ergative languages. As observed by Dixon (1994), ergative languages are found almost everywhere around the globe, including many languages of Australia (Dyirbal, Warlpiri, Diyari, Yidiny, etc.), Eskimo languages (Inuit, Yupik), Basque, Georgian, Avar, Chukchee, Hindi-Urdu, Tongan, Samoan, and many others.5
| (12) | Dyirbal (Dixon 1994)
|
| (13) | Greenlandic Inuit (Marantz 1984)
|
| (14) | Basque (Bittner and Hale 1996b)
|
| (15) | Avar (Mallinson and Blake 1981)
|
The term “ergative language” is also generally used for languages, such as the Mayan languages, in which there is no Case marking, but agreement groups S and P together as opposed to A.
| (16) | Tzotzil (Aissen 1983)
’ On the other hand, some ergative languages, like Warlpiri, have ergative Case marking but nominative–accusative agreement. |
| (17) | Warlpiri(Simpson 1983)
|
As noted above, nominative-accusative languages are plausibly analyzed by calling S/A the subject, and associating subject status with unmarked Case. One way to understand ergative languages would be to hypothesize that the absolutive argument, S/P, is subject instead of S/A. But if P is subject in ergative languages, it should have subject properties. Investigation has shown that things are not that simple. In some ergative languages, such as Basque and Warlpiri, the P argument of a transitive clause has no subject properties. For languages of this kind, called “morphologically ergative,” P is clearly not the subject. (For now, we leave it open whether S/A is the subject in such languages, or whether such languages can be said to have no subject. We will discuss one morphologically ergative language, Warlpiri, in Chapter 6.) It is clear, however, that in morphologically ergative languages unmarked Case cannot be said to be a subject property. For other ergative languages, such as Dyirbal and Inuit, subject properties are split. Some subject properties are properties of (S and) A, just as in English. Other subject properties are S/P (absolutive) properties. So in some sense A and P are both subject-like in these languages, but in different ways. Since the ergative Case marking seems to be in some sense related to syntactic properties, these languages are said to be “syntactically ergative.” We will return to syntactically ergative languages in the next section.
In another class of languages, the Philippine-type languages, any element of the clause can have unmarked Case (usually referred to as nominative); the verb is marked with an affix designating one of its arguments as the nominative. The nominative nominal is referred to by Philippinists as the topic, but as observed by Schachter (1976) it is not a topic in the sense that the term is usually used in linguistics.6
| (18) | Tagalog(Schachter 1987)
|
As with the term “ergative language,” the term “Philippine-type language” can also be used for languages that lack Case marking but have a similar system of marking the verb to designate one of the arguments as having a special status. These other Philippine-type languages are typically Western Austronesian languages not spoken in the Philippines, and with a less extensive marking system than the core case of Tagalog-type languages. One such language is Balinese, where the specially designatedargument is preverbal.
© Cambridge University Press