Reflexivization in English and Polish: An Arc Pair Grammar Analysis (Linguistische Arbeiten) [Reprint 2010 ed.] 3484301783, 9783484301788

Over the past few decades, the book series Linguistische Arbeiten [Linguistic Studies], comprising over 500 volumes, has

214 6 31MB

English Pages 133 [136] Year 2010

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Chapter 1. The premises
1.1. Introduction
1.2. The framework, its primitives and basic definitions
1.3. PN laws, grammatical rules and tertium comparationis
1.4. The data
1.5. Earlier analyses of the data
Chapter 2. The hypothesis
2.1. Unaccusative hypothesis
2.2. Arc Pair Grammar and the unaccusative hypothesis
2.3a Other PN laws relevant for the analysis of Polish and English reflexives
2.4. Relevant Genetic Pair Networks
2.5. The hypothesis
Chapter 3. Verification
3.1. Total and part reflexives
3.2. Middles or initial unaccusatives
3.3. Antipassives
3.4. Impersonal constructions
3.5. Inversion constructions
3.6. Reciprocals
3.7. On the limitations of the analysis and its possible extentions
3.8. Conclusions
References
Index
Recommend Papers

Reflexivization in English and Polish: An Arc Pair Grammar Analysis (Linguistische Arbeiten) [Reprint 2010 ed.]
 3484301783, 9783484301788

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Linguistische Arbeiten

178

Herausgegeben von Hans Altmann, Herbert E. Brekle, Hans Jürgen Heringer, Christian Rohrer, Heinz Vater und Otmar Werner

Wojciech Kabinski

Reflexivization in English and Polish: An Arc Pair Grammar Analysis

Max Niemeyer Verlag Tübingen 1987

CIP-Kurztitelaufnahme der Deutschen Bibliothek Kubinski, Wojciedi: Reflexivization in English and Polish: An arc pair grammar analysis /Wojciech Kubinski. -Tübingen : Niemeyer, 1987. (Linguistische Arbeiten ; 178) NE:GT ISBN 3-484-30178-3

ISSN 0344-6727

Max Niemeyer Verlag Tübingen 1987 Alle Rechte vorbehalten. Ohne Genehmigung des Verlages ist es nicht gestattet, dieses Buch oder Teile daraus photomechanisch zu vervielfältigen. Printed in Germany. Druck: Weihert-Druck GmbH, Darmstadt.

CONTENTS

Chapter 1.1. 1.2. 1»3· 1.4. 1.5.

1. The premises Introduction The framework, its primitives and "basic definitions PN laws, grammatical rules and tertium comparationis The data Earlier analyses of the data

Chapter 2.1. 2.2. 2.3.

2 . The hypothesis unaecusative hypothesis Arc Pair Grammar and the unaccusative hypothesis Other PN laws relevant for the analysis of Polish and English reflexives 2.4. Relevant Genetic Pair Networks 2.5. The hypothesis

Chapter 3.1. 3.2. 3.3. 3.4. 3.5. 3.6. 3.7.

3. Verification Total and part reflexives Middles or initial unaccusatives Antipassives Impersonal constructions Inversion constructions Reciprocals On the limitations of the analysis and its extentions 3.8. Conclusions

1 2 10 16 21

43 46 50 57 66

68 76 86 92 103 108 possible 113 116

References

121

Index

125

CHAPTER I THE PREMISES

1·1.

Introduction

This work aims to search, for an explanation of a striking discrepancy between English and Polish in the number of reflexive constructions in the two languages. Such explanation would require postulating rules which would account for the reflexive marking of numerous Polish constructions and for the non-reflexive character of their English translation equivalents» The framework within which such rules will be formulated is Arc Pair grammar ( APG), and the particular element of APG· which will be crucial for the solutions proposed here is the unaccusative hypothesis. As a result, this work may be also viewed as an attempt to test APG as a tool for contrastive analysis and to verify the predictions of the Unaccusative Hypothesis in the comparison of two different languages. Since APG is not a generally known framework this work must necessarily include a short presentation of the basic elements of this theory and of those elements which will be crucial for problems discussed below. Another topic which must be considered is the relation of APG to contrastive analysis since various features of APG may make it a theory particularly well tailored to serve as tertium comparationis in contrastive analysis. A subsequent section must be devoted to the data and past attempts to account for these data within several different frameworks. The unaccusative hypothesis along with its theoretical implications deserves a section of its own. Finally, language specific rules should be proposed for both languages, which would account for the data along with universal PN laws. Obviously, if necessary, revisions

2

in the system of PN laws would also have to be considered. And, last but not least, possible interaction of the unaccusative hypothesis with other linguistic phenomena should also be discussed. It is most fitting to thank at this point all those without whose friendly assistance I would have never conceived this work not to mention completing it. Prom the host of persons I am indebted to I should single out Paul Neubauer who first preached the APG gospel in Poland, Roman Kalisz without whose help I would have never embarked on this project, Aleksander Szwedek who supervised my work at its initial stage, Tomasz P. Krzeszowski who helped me conclude it and David Malcolm who aided me in weeding out some of the stylistic errors. I should also thank the British Council whose generosity enabled me to spend an invaluable year in Britain ruminating on syntax and semantics. Finally, I should thank my wife, Ola, for her patience and unfailing sense of humor. As for the inadequacies and flaws inherent in this work, I am entirely to blame for them. 1.2.

The framework, its primitives and basic definitions.

Johnson and Postal (1980) claim that APG is 35% new although it draws a lot from earlier work in Relational Grammar ( RG ) · One might be at variance with the authors as to the exact percentage involved since APG seems to take much more from RG than Johnson and Postal wish to acknowledge. With two fundamental exceptions all APG primitives are borrowed from RG, albeit redefined in a much more formally rigorous fashion. These primitive constructs include such notions dating back to Transformational Grammar (TG) as: nonterminal nodes, logical nodes, phonological nodes, grammatical category nodes. Two other sets of primitives, relational signs and coordinates, APG owes specifically to RG. Finally there are two primitive relations between arcs, called Sponsor and Erase, which are a unique feature of APG. Nonterminal nodes, which in APG are identified with positive integers beginning with 4, since the first three numbers are traditionally reserved for relational signs, represent clauses, phrases, etc. having no distinguishing features beyond being distinct from each other. Logical nodes are those primitive elements which are needed for semantic description. Phonological nodes will obviously be relevant for phonological

description. Grammatical category nodes will include such elements as Clause, Nominal, Masculine etc. Relational signs will include such well-known elements from RG as Subject or 1, Direct Object or 2, Indirect Object or 3, Chomeur (Cho ) and less familiar ones like Linear Precedence (LP), Labels (L), etc. Finally coordinates represent distinct linguistic levels ( APG answer to TG phrase markers) represented by sequences of elements as c^, c«, c,, etc. An Arc in APG terms will be an ordered pair of nodes combined with a member of the set of relational signs and a non null coordinate sequence. ( 1 ) An example of an Arc

55 C C

1 2

44 In ( 1 ) 55 and 44 are the nodes, 55 is the arc's tail and 44 is its head, 1 is a member of the set of relational signs and are a coordinate sequence. Arbitrary sets of so defined arcs will constitute relational graphs (R-graphs) provided they meet a number of conditions. These include such obvious ones as that this set must be non-empty and finite. A less obvious condition ( so called Pointedness Condition) specifies that an arbitrary set of arcs may form an R-graph iff it contains one and only one node which is not governed by any other node and which remotely governs ( R-governs) all other nodes in this set of arcs( an APG analogue of TG root ) . The definition of an R-graph contains also a number of other conditions which must be met by any arbitrary set of arcs if it is to constitute an R-graph. ( 2 ) TG constituent structure tree

/\VP

NP

Noun I Bill

V | sings

( 3 ) Simplified R-graph equivalent of( 2)

100.

-^σι VP

Bill

Sing

Johnson and Postal (1980) frequently simplify their diagrams, often omitting information which is not crucial for the particular problem they happen to "be discussing. Thus in( 3) some of the IP arcs are suppressed, all logical nodes are missing along with their I arcs, relational signs are left unspecified (TO stands only for Immediate Constituent) and all coordinates are left out. Obviously a complete R-graph would necessarily include all these elements. Practically all diagrams provided by Johnson and Postal (1980) are in some way simplified for the sake of clarity. In RCr R-graphs were considered to be the proper representations of sentences. In APG· R-graphs form only subsets of the proper, formal representations of sentences, called Pair Networks ( PN ) . The formal definition of R-graphs and two primitive relations, Sponsor and Erase, are the necessary prerequisites for defining formally the notion of PN. The primitive relations Sponsor and Erase range over particular R-graphs as their domains. Both relations are binary i.e. they hold between two arcs. Sponsor and Erase account in APG for most of those phenomena which in standard TG· were dealt with by means of transformational rules, though this is by no means a clear-cut analogy. Sponsor and Erase are claimed by Johnson and Postal (1980) to constitute the essence of the grammatical structure of a natural language sentence. Each PN will consist essentially of two finite sets of ordered pairs of arcs, one set interpreted as a Sponsor relation and the other as an Erase

5 relation. An arc A sponsors arc 'B if the occurrence of A partially justifies the occurrence of B ( partially since some arcs will have two distinct sponsors) . In some cases A=B and then we can speak of self-sponsorship. In such cases the self-sponsoring arc is a member of the logical representation of a sentence, provided it is not an LP arc. An arc A erases arc B if the occurrence of A justifies the nonoccurrence of B at a phonologically relevant level. In some cases A=B and then we can speak of self-erasure. Sponsor and Erase relations mediate between logical representation of a sentence (L-graph) and a phonologically interpreted representation (S-graph) . To facilitate representation of Sponsor and Erase relation in PN diagrams, 'Johnson and Postal (1980) pro- · pose graphical notation presented in ( 4 ) . ( 4 )a. Sponsor (A,B) b. Erase

(A,B)

=

(T

=

(T

In ( 4 )A and B represent arcs while the wiggly arrow stands for Sponsor and the double arrow for Erase. Utilizing Sponsor and Erase, Johnson and Postal (1980) define the notion of Primitive Pair Network ( PPN ) . PPN is a set of arcs containing a Sponsor relation and an Erase relation. In an R-graph which forms the arc set of a PPN, all arcs which are not erased are defined as Surface Arcs. All surface arcs of an R-graph form the Surface Arc Set of this R-graph. A Surface Arc Set will constitute an S-graph if it satisfies a number of conditions. Analogously to the R-graph an S-graph must be nonempty and pointed. It must also satisfy the condition that each of its nonterminal nodes R-governs a phonological node. There are also other conditions which must be satisfied by S-graphs. All of these conditions are formally stated in the definition of S-graph. In an R-graph which forms the arc set of a PPN, all self-sponsoring arcs (except LP arcs) are defined as Logical Arcs. All logical arcs of an R-graph form the Logical Arc Set of this R-graph. A Logical Arc Set will constitute an L-graph if it meets a number of conditions such as, inter alia, the nonemptiness condition and the connectedness condition.

All of them are formally stated in the definition of L-graph. A Pair Network (PN) , which in APG is a single formal object representing the whole structure of a sentence, is defined as a ΡΡΤΓ which must meet 4 conditions: C l )every arc in PN must be sponsored, (II) the arc set of a PN is an R-graph, Q, ( III) the Surface Arc Set of Q is an S-graph, (IV) the Logical Arc Set of Q is an L-graph. Johnson and Postal (1980) illustrate their definition of PN with a diagram, here numbered as ( 5 ) . Obviously ( 5 )is simplified since it does not include LP arcs, much of the logical representation of lexical items, coordinate sequences and nonterminal nodes in the form of integers. Also some relational signs are left unspecified (as OR ) . ( 5 )a. John eats meat, b.

CI

Noun

eat (s)

"in [(5b)3 , Erase is the empty set. Both the L arcs with logical node heads, H and K, are self-sponsoring. In contrast, only one of the phonological terminations is self-sponsoring, Q j the others, R and S-., are sponsored by distinct L arcs. The significant difference between 0 and both R and S.. is that the head of 0, John, represents a name, whereas the heads of R and S^, meat and eat ( s ) , respectively, do not, A concomitant difference between 0 and both R and S-. is that R and S.,, but not 0, are sponsored by logical terminations ( H and K, respectively ). Thus, 0,H and K will be in the L-graph associated with

L(5b)3, but R and S^ will not "be, Also, the grammatical terminations D,E,J,S. and S5 will not be in the L-graph ofC(5b)3. The L-graph for[I(5b)Jwill consist of A, B,C,G,H,X,0,S ? and S,. The R-graph is, of course, the entire set of arcs shown in ^ u(5b) U, and the S-graph is, in this case, identical to the R-graph. M (Johnson and Postal 1980: 102-103) . A brief look at the APG analyses of several linguistic phenomena which have well established corresponding analyses in tore traditional TG terms might help to illustrate the internal mechanics of the framework devised by Johnson and Postal (1980) . Analyses of extraposition, reflexivization and passivization might well serve as convenient cases in point. In most TG analyses the derivation of ( 6 ) would involve the rule of extraposition which moves the that clause from the subject position to the final position in the matrix clause, ( 6 ) It surprises me that Jack likes bagels. In the APG framework (6) would correspond to a PN representation along the lines of (7) . (7 )

me

that

Jack

bagels

Surprises

likes

In ( 7 ) aros D, E, G, I, J, Κ are self-sponsoring i.e. their presence in the PN representation may be justified on purely semantic grounds. Therefore arcs D, E, G, I, J, K all belong to the L-graph of ( 7 ) and are all assigned the c1 coordinate thereby roughly corresponding to those levels of analysis characteristic of TG deep

8

structure. Arcs H and M embody the APG analogue of the mechanism proposed by Chomsky (1973) to deal with rules like Wh Movement which move elements under Comp nodes. The presence of arc G in the PN representation is justfied by the presence of arc D (hence D sponsors H) while M is sponsored by both D and H. Arc C is cosponsored by D and M. Finally, C erases D i.e. the absence of D in the noninitial strata of ( 7 ) is the consequence of the presence of C. Arc H is labelled with the relational sign F which stands for flagging while the Marq.relational sign of M stands for marquee. Plagging mechanisms devised by Johnson and Postal (1980) are useful not only for dealing with complementizers but also serve to account for such phenomena as prepositional flagging in case of English indirect or oblique objects. Arcs A, B and C illustrate the APG analogue of the rule of extraposition. The presence of the 1 arc A, headed by the element it, is justfied by the presence of C. C and A together justify the presence of arc B. The presence of B motivates the absence of C from the final stratum of (7 ). Thus, using APG jargon, C sponsors A, A and C cosponsor B and B erases C. Consequently, arcs A, B, E, G, H, M, I, J, K belong to the S-graph of (7 ). Obviously ( 7 ) is a simplified representation with LP arcs, all nonterminal nodes, logical nodes and phonological nodes being omitted. Similar simplification will characterize all graphical representations presented in the remainder of this work. Reflexive constructions may be presented in a more transparent way since they do not involve the additional complications connected with flagging mechanisms. Given all the simplifications of the. relevant PN representation, ( 8 ) would correspond to something like (9) . ( 8 ) Jack shaved himself.

Jack

himself

shaved

In (9 ) arcs A, B and D are self-sponsoring. Arc C is cosponsored by A and B· C is responsible for the absence of B in the final stratum o f ( 9 ) i . e . C erases B. Thus the L-graph of ( 9 ) consists of arcs A, B and D while the S-graph is composed of arcs A, C and D. A passive construction like (10) would correspond to a Pair Network representation like (11) . (10 ) Jack was tickled by Mary. (11 ) 1

Jack

led

Obviously ( 11 ) is only a partial representation of (10) since, apart from the usual simplifications, such important elements as the verbal morphology and the flagging of the chömeur are left unattended to. In $1 ) arcs B, C and E are self-sponsoring and constitute its L-graph while arcs A, D and E comprise the S-graph. A and C cosponsor D while B sponsors A. A erases B and D erases C. The examples presented above may create the false impression that most of the generalizations stated within TG can be easily translated into the APG formulae. This impression is false not because such translation is in principle impossible (indeed Johnson and Postal (1980 ) take great pains to prove that APG can do anything TG can, and more ) but because at this stage of my exposition nothing yet seems to warrant that sentences (6) , ( 8) and do) must be assigned respective Pair Network representations ( 7 ) , (9) and (11 ) . Indeed, I have not mentioned yet any principles that specify which types of arcs can sponsor or erase which types of other arcs and in what ways sponsorship or erasure might be properly constrained. Johnson and Postal (1980 )devote most of their monumental work to an elaboration of such principles which jointly constrain the set of well formed PN representations in such a way that, inter alia, PN representations (7 ), (9 )and (11) indeed are

10

the proper parts of respective representations of ( 6 ) , (8) and (10) . The theoretical status of these principles will be briefly discussed in section 1.3· of this chapter. Those individual principles which are particularly relevant for this inquiry will be mentioned and briefly explained in sections 2.2. and 2.3. of this book. . This short section is certainly a very superficial rendering of the APG theory. One of the basic assets of APG is its rigorous formalization. Each notion is defined in terms of primitives or other defined notions. Definitions subsequently serve as basis for constructing proofs of theorems. All definitions and theorems are presented for the sake of clarity in somewhat modified logical notation. While it makes reading tedious and difficult, it certainly sets up APG as a linguistic theory quite superior in achieved precision to any other current framework, perhaps with the notable exception of Montague Grammar« Nothing of this sense of precision and thoroughness is left in the above sketchy notes. This is however hardly surprising when one tries to give a nine page summary of a 100 page technical text. Therefore directing the more inquisitive reader to Johnson and Postal (1980) for a truly comprehensive introduction to the intricacies of APG, I hope that the above notes along with information presented in sections 1.3, 2.2 and 2.3. will suffice to situate the ensuing chapters of this book within an adequate, if incomplete, scheme of reference. 1.3.

PN laws, grammatical rules and tertium comparationis.

Much of the work in Relational Grammar (RG) was devoted to formulating a host of universal laws. Relational Networks ( roughly RG equivalents of APG R-graphs )were required not to contradict these laws regardless of the language whose sentences they represented. APG borrows this basic insight from RG without taking over actual laws. In APG terms a set of arcs constitutes a PN, well formed at the level of universal grammar, if it model theoretically satisfies the formal definition of PN and the set of PN laws. PN laws will thus serve to limit theoretically possible PNs to just those which represent sentences occurring in any of the natural languages of the world. PN laws are interpreted as material

11

implications in the standard logical sense. As such they check for the cooccurrence or noncooccurrence of specified properties of given PNs. APG· grammatical rules do not differ formally from PN "The essential difference between APG grammatical rules and PN laws is one of scope. The former are language- particular and thus not necessarily an aspect of any particular language. Thus PN laws are material implications which determine wellformedness in all languages; grammatical rules are material implications which determine well-formedness only for individual language s.C·.· 3 More precisely, a grammar in APG terms. for some language L is the union of the set of APG PN laws with some finite set of language- particular material implications specific to L . The set formed "by this union is unstructured in the sense of containing no components. Roughly £···] , a given PN is well-formed with respect to I Just in case it model theoretically satisfies the grammar or L , which is the union of the set of PN laws with those material implications particular to L.." (Johnson and Postal 1980:7) . 3. An example of a PN law, one of the 116 formulated "by Johnson and Postal (1980 ) , is given in (12) . (12) PN Law J> (The Unique Eraser Law ) Erase (A,B) * Erase (C,B )—>A«0 (12) imposes on all PN representations the requirement that an arc may be erased by only one arc i.e. only one are may be held responsible for the absence of another arc in the final strata of a PN. (12 )may also exemplify notational simplifications introduced by Johnson and Postal (1980) into formal logical notation. For example a full unsimplified version of (12) would be (13) · (13) (VA) (VB) ( V C ) (( Arc( A)*Arc(B)*Arc(C]LErase(A,B). Erase(C,B))—*(A=C) For the sake of clarity the authors suppress in a given formal statement universal quantifiers with the entire statement, except other universal quantifiers, in their scope· Furthermore,parentheses around antecedent and consequent expressions in logical implications and biconditionals, are omitted. Finally conventionalized

12

notation for variables is employed rendering superfluous expressions which restrict variables to their appropriate domains (e,g.capital letters (except F f P , L ) are reserved to conventionally denote arcs) , PNs which satisfy the definition of PN and the set of all PN laws are called Genetic PNs. Universal grammar in the form of all PN laws would be required to characterize a priori everything which could be a sentence in a human language. The set of PNs we11-formed for a given language L is a subset of the set of Genetic PNs. "The task for individual grammars is only to exclude from particular languages objects which, though sentences in other languages, happen not to be sentences in these." (Johnson and Postal 1980:656 ) . Neither PN laws nor ATG grammatical rules will "sanction" particular constructions in particular languages (as was the case in all TG frameworks ) . Rather they will block constructions unattested in any human language ( level of universal grammar) or constructions unattested in a particular language (level of grammatical rules ) « For example PN laws allow PNs of both active and passive sentences. Therefore a particular language lacking any passive constructions would need a rule blocking passives while a language with optional passiviaation would rely solely on universal PN laws with no language particular rules apart from those dealing with such details as verbal morphology, chomeur flagging etc. A language with "short" passives would require a rule blocking "long" passives. Finally a hypothetical albeit unlikely language with obligatory passivization would need a rule blocking PNs which represent active constructions. Obviously, should it turn out that particular data from any natural language is absolutely incompatible with the present set of APG laws, reformulations of these laws would necessarily follow. Thus all PN laws and theorems are in principle empirically verifiable. Keeping in mind the APG conception of PN laws and grammatical rules it might be worthwile to briefly explore the relationship of APG to contrastive analysis (CA) and the central for CA notion of tertium comparationis ( TG ) · Krzeszowski (1984) draws attention to the fact that in any comparison it is necessary to have some point of reference in terms of which the comparison is made and by

13

virtue of which the compared, items are comparable i.e. it is necessary to employ in any comparison some form of TO· In past contrastive analyses semantic equivalence or formal correspondence were generally assumed to be adequate TCs. Krzeszowski (1984 ) rightly observes that other types of TCs are conceivable. A more developed and comprehensive analysis of the notion of TC may be found in Krzeszowski (1985 ) where a distinction is made between immediately relevant TC (IRTC ) and ultimately relevant TO (URTC ) . In Krzeszowski's (1985 ) terms URTC corresponds to equivalence of semantic representations of compared structures while IRTC refers to identity of syntactic representations of compared structures at some level of their derivation. While Krzeszowski (1985) develops his framework along generative lines (Gontrastive Generative Grammar ) , his concept of TC is relevant for all contrastive analyses, regardless of their theoretical assumptions i.e. also for APG contrastive analysis. In the context of the APG model of universal grammar the notion of TC may be viewed from yet another perspective, APG, based on graph-theoretic considerations, with specified primitives and all other notions defined formally in terms of these primitives, seems to be well tailored for crosslinguistic analysis. Furthermore, APG, capitalizing on RG decade long tradition of cross-linguistic comparisons, bases its purportedly universal generalizations on a cross-linguistic data corpus which is considerably broader than in most TG based contrastive analyses. Besides, since PNs are formal representations of the whole structure of sentences, in terms of universally defined notions and categories, it is possible to easily compare any PNa as entities of the same status and. similar form. The formal definition of PN may thus be viewed as a necessary TC for APG contrastive analysis. All graphic representations which satisfy this definition are eo ipso comparable. If however two graphic representations have only this much in common that they are PNs, they may only serve as part of corpus for quantitative contrastive analysis. Any systematic contrastive analysis would, require a richer common denominator ( i.e. TO ) for the compared sentences. Johnson and Postal (1980) claim that all the relevant syntactic, semantic and phonological properties of a particular sentence must be encoded in the PN representation of

H that sentence. The PN representation is in turn well-formed if it satisfies all AK universal laws and all grammatical rules of the language whose sentence it represents. Thus Krzeszowski's (1985 ) URTC corresponds to APG equivalence of L-graphs i.e. identity of all arcs comprising the L-graphs of the compared sentences. Various forms of fereal equivalence of two sentences would correspond to identity of subsets of arcs comprising the S-graphs of these sentences. Identity of noninitial strata of two PNs could serve as the equivalent of Krzeszowski's (1985 ) IRTC. Partial identity of arcs headed by phonological nodes would furthermore serve to establish equivalence in phonology. Given a complete set of APG PN laws and a complete set of grammatical rules for two investigated languages, any systematic comparison of two sentences from the compared languages would be reduced to a trivial exercise of checking the identity of arcs in the arc sets of appropriate PNs. Complete identity of all arcs within the arc sets of two PNs would yield their total equivalence. Complete lack of identity of all arcs within the arc sets of two PNs would yield their total inequivalence. All comparisons of PNs would fall in between these two highly unlikely and hardly envisionable extremes, ranging from PNs sharing large subsets of their arc sets in a systematic way and hence being paraphrases of each other (usually with identical or similar L-graphs ) to PNs sharing only some arcs in a nonsystematic way. For example, should we decide that the identifying factor of passivization is a 2arc sponsoring a 1arc successor, then all such instances in all natural languages, regardless of other factors such as verbal morphology, presence or absence of chOmeurs etc., would be classified as formally equivalent examples of passivization. Obviously a crucial determining factor here would be the recognition of certain arc configurations as .more salient than others for the identification and characterization of specific language phenomena. Unfortunately, in reality the set of PN laws proposed by Johnson and Postal (1980) is far from being exhaustive and large areas of syntax and semantics yet await descriptioa in APG terms. Not only do hitherto formulated PN laws require empirical verification, but also gaps existing in the set of PN laws and practical nonexistence of fcets of language particular rules permit too much

15 indeterminacy and hence possible arbitrariness in PNs to make such straightforwardly simple comparisons of PNs possible. Instead, APG contrastive analysis must be, and in the foreseeable future will continue to be, a much more complex enterprise involving the postulation of rules for the two compared languages and reexamination of the validity of PN laws on the basis of language particular data. In this enterprise the set of PN laws suggested by Johnson and Postal ( 1980 )must serve, in spite of its obvious inadequacies, as the temporary tertium comparationis. This is so because all PNs are theoretically well formed in reference to this set. The piecemeal character of the APG formal apparatus has as its unfortunate consequence the rather tentative character of all PN representations. On the other hand the set of PN laws provided by Johnson and Postal (1980 )delimits the set of possible PNs narrowly enough to serve as a TC for want of a more adequate one. Thus, in order to examine in APG terms a particular language phenomenon cross-linguistically, i.e. contrastively, it would be necessary first to identify the phenomenon in universal terms i.e. specify the characteristic features of Genetic PNs under scrutiny along with those PN laws and theorems crucial for these Genetic PNs. Secondly, data from the compared languages should be interpreted in terms of these Genetic PNs and grammatical rules for the languages in question should be formulated to filter out those Genetic PNs which do not represent sentences in the compared languages. Thirdly, should it turn out that data from any of the compared languages will inevitably contradict any of the PN laws, the PN law in question must be abandoned since PN laws are claimed to hold true for all languages and hence are empirically verifiable in terms of data from any natural language. The considerations above were decisive for shaping the scope and form of the present inquiry. The ensuing sections must contain a presentation of data along with solutions, proposed to account for the data, present in linguistic literature. A subsequent section must relate APG universal claims connected with the described data including relevant PN laws and theorems. Finally language specific rules must be given and eventual revisions in the set of PN laws should be considered, if such reformulations

16

turn out to be necessary. 1 «4.

The data.

This work is concerned with reflexive constructions in English and Polish. It will, however, become clear from what follows that there are no clear-cut criteria which would tell us what is and what is not a reflexive construction. I shall therefore initially assume that all Polish constructions which include elements like siebie, sobie, s ig and sobq, as well as English constructions with -self words will have to count as reflexive constructions. While this is undoubtedly a crude way of singling out the data, no other Possibility readily offered itself to me since any attempt to initially delimit the data on semantic grounds would leave some crucial items outside the analysis. One need not be a linguist to notice that so defined reflexive constructions are much more numerous in Polish than in English, both in terms of the number of types of constructions marked with reflexive pronouns as well as in terms of their frequency of occurrence. However Niedzielski (1976) did take the pains to actually count the ratio of reflexive verbs to the total number of verbs in several texts from both languages obtaining figures ranging from 15,996 to 20,5# for Polish and from 0,4% to 996 for English c.f. Niedzielski(1976:172) . These figures more than adequately prove the point that there exists a discrepancy between the two languages in the occurrence of reflexive constructions. This relative abundance of reflexive constructions characterizes in general all Slavic and Romance languages. Such widespread tendency to mark many constructions, traditionally regarded as semantically non-reflexive, with reflexive morphemes, cannot be accidental. Linguistic theory should thus specify similarities and differences between various constructions marked with reflexive morphemes furthermore shedding some light on cross-linguistic differences in the range of these phenomena. Polish constructions marked with the reflexive morpheme sig have been grouped by various linguist· into a number of types. These will vary from author to author depending on particular theoretical assumptions, frameworks and employed criteria. Thus a hapless reader will encounter in linguistic literature a confusion

17 of labels such as "true" reflexives, reciprocal constructions, inchoative constructions, middle constructions, psych movement verbs or verba affecta,notional passives, impersonal constructions etc. Saloni ( 1976 ) comes up with the following list of sig constructions: (14) Czyta sie, te, ksi^zke, z przyjemnosci^. "one reads this book with pleasure." (15) Pracuje mu sie, dobrze. "He finds his work enjoyable." ( 1 6 ) Janek umyl sie,. "John washed himself

."

(17) Jan i Piotr bij^ sie.. "John and Peter are fighting

each other

."

(18) zupa gotuje sie,. "The soup is cooking." (19) Janek sie, wrocil. "John returned." (20) Janek skrzywii sie,. "John scowled." (21) Po co sie, z nim zadajesz? "Why do you have any dealings with him?" ( 2 2 ) Pawei boi sie,. "Paul is afraid." Saloni (1976) draws a sharp line between constructions like (U) and ( 1 5 ) on the one hand and ( 16 ) , (17) , (18) , ( 1 9 ) , (20) , ( 2 1 ) , and ( 2 2 ) on the other, claiming that there exist two distinct sie, morphemes: nominative sie, and accusative sie. Nominative sie, in constructions of type (14) would represent an unspecified human subject not unlike man in German, one in English and on in itench. Type (15 ) constructions differ from type (14) since here the logical subject is specified and appears in the

18

dative case as superficial indirect object. All other sie_ constructions, according to Saloni (1976) , contain accusative sie« Thus type (16) is obviously comprised of "true" reflexive, presumably. formally equivalent to English reflexives. Type (17 ) is generally labelled reciprocal. Type (18) constructions are called by Saloni ( 1976 ) passive. Under this heading Saloni (1976 ) includes explicitly inchoatives and implicitly psych movement verbs. In each of these constructions the subject appears to be coreferential with a direct object in a corresponding non-reflexively marked construction as shown in ( 23) * ( 2 4 ) and( 25 ) . Hence the label "passive". (23) a. b.

(24) a. b.

(25 ) a. b.

Szklanka zbila sie.. "The glass broke." Janek zbil 3zklank§. "John broke the glass." Zupa gotuje "The soup is cooking." janek gotuje zupe,. "John is cooking the soup." Janek interesuje sie_ tym. "John is interested in this." To interesuje Janka. "This interests John."

Saloni (1976) claims that the distinction between reflexive and passive si^ constructions is not sharp and some examples may be difficult to classify. Group (19) will include reflexive verbs which seem not to differ semantically from their non-reflexive counterparts e.g. wrocic v.s. wrocid sie. Saloni (1976) also includes in this group examples from child language like (26 ) where the morpheme sie, seems to occupy the place reserved for a direct object not present in this case· ( 2 6 ) On sie. kopie. "He is kicking ( someone ) ."

19

In constructions of type (20) sig, according to Saloni (1976 ), has a similar semantic function to sie, in "true" reflexives. Type (20) differs from "true" reflexives in that the action is reflected not on its instigator or performer but rather on a possession or part of the instigator or performer. Niedzielski (1976 ) labels similar constructions as part reflexives while Wilczewskä (1966) calls them narrowing reflexive verbs. Group (21) is a waste-basket category where Saloni (1976) places reflexive verbs which seem not to be in a regular semantic relation to their non-reflexive counterparts e.g. zadawac v.s. zadawao sie« Finally type (22) will contain reflexiva tantum such as bad sie« dowiedziec sie, przejsc sig etc. A somewhat different list of generally recognized sig constructions might be found in the introductory chapter of Kanski'a (1984) brilliant dissertation ( Kanski 1984:1-2 ) : (27) a. b.

(28) a. b. c.

Janek "John Janek "John

my;je sie,. washes ( himself ) ." ubiera sie,. is dressing ( himself ) . "

Drzwi otworzyly sie,. "The door opened." Szklanka zbiia sig. "The glass broke." Lekcja skonczyla sig. "The class finished."

(29) a. Bawelna dobrze sig pierze. "Cotton washes well," b. Ta ksia,zka dobrze sig czyta. "This book reads well." (30) a. b.

Janek sig bije. "Johnie beats (other children) ." Nie pchaj sig. "Don*t push (me ) ."

20

(31) a. b«

(32) a. b. c.

Tu sie_ mow! po angielsku. "Here one speaks English." W Polsce si§ nie je slimakow. "In Poland one doesn't eat snails." Janek skaleczyl sie_. "John hurt himself /got hurt." Janek zabil si§ ( w wypadku ) . "John killed himself/got killed (in an accident) ." Maria cze.sto fotografuje si§. "Mary often photographs herself." or "Mary often gets photographed."

Constructions of type( 27 ) , which correspond-to Saloni*s (1976) type (16) would be labelled "true" reflexives and generally assumed to be the formal equivalents of the English reflexive constructions. Constructions like ( 2 8 ) are usually known as inchoatives while sentences like ( 29 ) are discussed by Wolczynska-Sudol (1976) under the heading notional passives and by Horn (1977) as middle constructions. Kanski (1984) is in agreement with Napoli (1976) and Saloni ( 1976 ) who calls such constructions passive. Type( 30) according to Kanski (1984) represents only a small, probably non open class of sentences in Polish. "The distinguishing characteristics of this group is the apparently "detransitivizing" function of the pronominal element C···! " ( Kanski 1984:5 ) . The subjects of these sentences are not interpreted as direct objects of any corresponding transitive sentences, unlike the sentences of type ( 28 ) and ( 29 ) · Obviously this group corresponds to some of Saloni's 1976 group ( 14 ) . Usually the pronominal element is understood in such constructions to represent an unspecified human subject. Unlike all the other constructions on Kanski's (1984) list, sentences similar to (31) contain sig regardless of lexical transitivity of the verb. The last class of constructions (32) listed by Kanski (1984) is according to him ambiguous between the reflexive and middle readings. The two lists are partly overlapping and partly complementary.

21

On their basis I shall initially and informally adopt the following labels for various sig constructions: total reflexives( so called in opposition to part reflexives ) corresponding to examples (16) and ( 2 7 ) , part reflexives corresponding to (20) , reciprocal constructions corresponding to 07 ), middles corresponding to (18) , (28) and (29) , impersonal sig constructions corresponding to (14) and (31) , inversion constructions corresponding to (19) and (30) . The rationale for the last two labels will become clear in the ensuing chapters. Examples similar to (32) will be covered by the analysis of either total reflexives or middles, while in the case of groups ( 21 ) and ( 22 ) I can only hope that some of these examples will be analyzable within the other paradigms while the remainder will be characterized in terms of word-formation rule s. 1.5.

Earlier analyses of the data.

Contrastive works on reflexivization in English and Polish are less numerous than one would expect. Three short articles by Niedzielski ( 1 9 7 6 ) , Wolczynska-Sudol (1976) , Kubinski ( 1 9 8 2 ) , a somewhat longer one by Horn (1977) and two doctoral dissertations by Kardela (1979) and Kanski (1984) practically exhaust the field. Thus in order to obtain richer material for theoretical considerations one must follow in the footsteps of Kanski (1984 ) and discuss contrastive works on related phenomena in other languages like Italian and Russian, such as Babby (1975) , Cranmer (1976 ) and Napoli ( 1976 ) . Unfortunately, works of Polish linguists differ quite considerably from this thesis in methodology and research aims and few references will be made to Polish linguistic literature. However one argument between Wilczewska (1966) and Saloni ( 1976 ) must be mentioned. Wilczewska (1966) wants to analyze sie^ as a word-formation morpheme, exempting only two types of sig constructions from such analysis: impersonal s ig as in (15 ) and sie^ in constructions like ( 33) , where sie would be analyzed in syntactic terms.

(33) Jan dal sie. bic. "John allowed himself to be beaten."

22

Saloni (1976) disagress with Wilczewska (1966) , arguing that most of the other sig constructions may be properly described in terms of syntactic rules and not word-formation rules since combinations of verbs with sig exemplify such semanto-syntactic regularities that they may not be discussed under the label of word formation. Thus, according to Saloni (1976) , most sig^ constructions are syntactic constructions that contain an independent lexeme which, not unlike other pronouns, has a fixed meaning not due to the properties of extralinguistic reality but due to the relations between reality and the construction in which the form of the given lexeme occurs, (c.f. Saloni 1976:118) . I shall side in this dispute with Saloni (1976) and assume that all sie^ constructions may be properly described in terms of syntactic regularities with the possible exception of some reflexiva tantum ( 2 2 ) and some verbs of type ( 2 1 ) . Within the field of Polish-English contrastive studies the first to characterize contrastively and attempt to solve problems of reflexivization in these two languages was Niedzielski (1976) . Niedzielski (1976) called the various sig constructions which according to him were semantically nonreflexive, pseudo-reflexives, using the same label for English constructions which though semantically reflexive were not marked as such by reflexive pronouns. On the basis of somewhat loose and informal semantic criteria Niedzielski (1976) discerned altogether seven types of reflexively marked constructions in Polish. These he labelled total reflexives, part reflexives, reciprocals, directed benefactives, observed benefactives, passives and emissives. In order to analyze these groups and account for the differences between Polish and English examples, Niedzielski (1976 )constructs a number of rules within the framework of case grammar as conceived by Fillmore (1968 ). The results are by no means convincing. For all constructions, apart from emissives, Niedzielski( 1976) attempts to find some form of coreferentiality in their deep structures in order to apply reflexivization in Polish. This yields highly counterintuitive structures especially in case of constructions labelled by Niedzielski (1976) passives. Strangely enough under the single label of jassives Niedzielski (1976 ) mentions not only inchoatives and verba affecta but also impersonal constructions

23 very different from them. Clearly, Niedzielski (1976) fails to notice some of the basic problems in accounting for the Polish data, such as the status of impersonal constructions troublesome for most other analyses. J5missives, under which label Niedzielski (1976) includes such verbs as budzic si£ "to wake up", receive a completely different analysis from other reflexive verbs. In other case, in contrast to all other discussed constructions, reflexivization applies after the application of a rule of copying. Hence here reflexivization is not conditioned by deep structure coreference. Niedzielski»s ( 1976) solution to the Polish data problem is thus nonhomogenuous, often counterintuitive and finally not as general as required (e.g. impersonal constructions escape proper characterization in terms of Niedzielski's (1976) rules ). In spite of its shortcomings Niedzielski»s (1976) article includes valuable insights. One should not, however, look for them in the sphere of the actual formalization and proposed rules, but rather in the area of informal observations on the semantics of sie constructions. Niedzielski (1976 ) asks interesting questions, echoed in this thesis, quickly puts forward unsatisfying answers and in process of doing so notices, for example, that there seems to be something inherently passive about some sie constructions. It is certainly to Niedzielski'3 ( 1976 )credit that he undertook pioneering work in attacking a long neglected issue in the sphere of Polish English contrastive studies, although his solution may be judged as somewhat misguided. Wolczynska-Sudol's ( 1977^analysis has much less ambitious aims. Withun her sphere of interests is only a subset of sig constructions, namely the subset including sentences like (34) and (35) , which in spite of their reflexive marking have been analyzed as passive in meaning. Wolczynska-Sudol ( 1977) calls such constructions notional passives contrasting them with what she calls formal passives as in (36) and (37) » where ( 36a ) and (37a) represent notional passives and ( 36b ) - ( 3?b ) formal passives, (34) Mie.so gotuje sie,. "The meat is cooking."

24

(35) Krowy do ja. sie, dobrze. "The cows are milking well." (36) a* b.

(37) a. b.

Brzwi otwariy sie,. "The door opened." Drzwi zostaiy otwarte. "The door was opened." Pacing zatrzymal sie,. "The train stopped." Pociajy zostal zatrzymany. "The train was stopped."

Notional passives resemble formal passives in taking patients as subjects. They differ morphologically from formal passives since English notionally passive verbs formally do not differ from their semantic equivalents in regular transitive, active sentences while Polish constructions usually involve the presence of the morpheme sie.. Neither English nor Polish notional passives allow agents, which also differentiates them from formal passives. Analyzing such examples as ( 34 ) - ( 37 ) Wolczynska-Sudoa: (1977) posits within the framework of case grammar a number of rules for Polish and English in order to capture the similarities and differences between formal passives and notional passives in the two languages» She also tries to relate impersonal sie, constructions to notional passives claiming that the difference between the two lies in subjectivilization not having occurred in the former and applying in the latter. But, as in the case of Niedzielski (1976 ), it is not in the sphere of formalization and rules that Woiczynska-Sudoi's ( 1977) analysis merits greatest attention. Her rules are not meant to provide any explanation for the reflexive marking of the Polish constructions. Relations between various reflexive and pseudoreflexive constructions are in no way clarified and indeed it is not Woiczynska-Sudol's (1977) aim to seek such clarification. She merely limits herself to echoing the observation by Szober (1953) and Gaertner (1933) that reflexive forms are often used in the passive sense in Polish, furthermore quoting Zandvoort (1962) on the existence of similar, though

25

non-reflexive, constructions in English, later discussed by Lakoff (1977) as patient-subject constructions. It is especially this informal observation which constitutes a modest, though not insignificant, contribution to the contrastive analysis of reflexivization in Polish and English. Kubinski (1982) attempts to combine some of the considerations brought up by Niedzielski ( 1976 ) and Wolczynska-Sudol (1977) with the much celebrated discussion of patient-subject constructions by Lakoff (1977) within gestalt theory. Capitalizing on Lakoffs and Wolczynska-Sudol's observations, he tries to adapt their analyses to other sig constructions. Claiming that various sig constructions involve either patients in subject positions ( "true" reflexives, reciprocals, inchoatives ) or lack of agents in subject positions (inchoatives, impersonal constructions) , Kubinski (1982) tries to construct a prototype to which all sig constructions would be instances of partial pattern matching. Within such a prototype sie, is claimed to function as a particle associated with verbs whose subjects have some properties of agent and some of patient. Impersonal constructions would have very "weak" agents, not present in surface structure at all. While it is not at all unrealistic to attempt to construct a chain of reflexively marked constructions in order to establish their analysis in cognitive grammar terms, it should be pointed out that Kubinski's (1982) article certainly falls short of such aim. His rather sketchy and informal analysis does not contribute much to the discussion of reflexivization in English and Polish apart from the belief that all sie_ constructions have some common denominator thus deserving an analysis in uniform terms and that the apparently widespread usage of reflexively marked Polish constructions requires a systematic explanation. All of the above analyses represent cases of "beating" about the "bush" and do not provide satisfactory, overall solutions to the problem of reflexivization in English and Polish. Other, more polished attempts have been made to .solve the problem of pseudoreflexives in other languages. Kanski (1984) discusses solutions proposed by Babby (1975) and Cranmer (1976) on the one hand and Napoli ( 1976 ) on the other. I shall follow Kanski (1984 ) and lavishly quote from his highly critical evaluation of these pro-

26

positions. In his analysis of Russian reflexive constructions Babby( 1975) presented his concept of derived intransitivity (DI) · This notion was used by Cranmer (1976 ) for contrastive analysis of Russian, German and English data. Kanski ( 1984 ) adopts DI to the analysis of Polish sie, verbs and presents the following assumptions underlying DI: ( 38) "a. b. c.

for each sie-verb to which DI applies there must be a corresponding transitive verb without sie.; there must be no change in lexical meaning between the verb with sie, and the corresponding verb without differences in meaning between various sig verbs result from different syntactic processes which accompany the introduction of sie into the phrase marker." (Kanski 1984:8)

Babby (1975) proposes for Russian examples a rule of -sja introduction by means of which -sja 'is obligatorily attached to a transitive verb just in case the direct object accusative NP dominates a dummy i.e. is not filled lexically. Presumably for Polish a corresponding rule of sie, insertion could be devised. The direct object position may be vacated in two cases: either because of an object preposing rule which moves direct objects to subject position whenever this position is lexically not filled e.g. in case of sentences like (28 ) or ( 29 ) ; or because the direct object NP node dominates only a semantic feature [[Animate] which deletes thereby leaving the NP node empty e.g. in case of sentences like (30) . Kanski (1984) points out some undesirable consequences of DI, such as high redundancy stemming from reliance on obligatory transformations predictable from information in the lexical component (c.f. Kanski 1984: 9-11 ) , controversial feature marking of direct objects in type ( 30 ) constructions and the impossibility of relating this feature to the application of sie./-sja insertion rule which in principle is blind to the. way the direct object node becomes empty, as well as the inability to prevent unspecified object deletion from feeding such an obligatory rule of sieV-sja insertion ( c.f. Kanski 1984: 11-15) , and finally the quite arbitrary way in which "true" reflexives relate

27

to the constructions of type (28) , (29) and ( 3 0 ) ( c . f . Kanski 1984: 15-18 ) . Another drawback of DI is that it does not account for all data, leaving impersonal constructions totally unexplained, A somewhat more successful analysis of reflexively marked constructions in Italian may be found in Napoli ( 1976 ) . Napoli's ( 1976 ) point of departure is the very same observation made earlier by proponents of DI, that the subjects of pseudo-reflexive constructions appear as direct objects in corresponding transitive constructions. Napoli (1976) proposes for Italian middles a deep structure in which the subject NP is lexically not filled, just as it is the case with DI deep structure phrase marker. However, besides moving the direct object NP to the subject position, in Napoli's (1976 )analysis a special rule of copying leaves a copy of the moved lexical material under the direct object NP node. The resulting structure "feeds" reflexivization just like constructions with two coreferential, deep structure NPs ("true" reflexives) , (39) - (42) exemplify such derivational history. Wszystkie drzwi otworzyly "All doors opened."

(39) (40 )

NP

X

DET

A

N

otworzyc wszystkie

drzwi

DET

wszystkie

drzwi

otworzyc

wszystkie

drzwi

28

DET

wszystkie

drzwi

otworzyc

sie.

Napoli's (1976 )copy hypothesis seems certainly superior to DI since it offers a non-arbitrary way of relating "true" reflexives to middles, something DI is incapable of doing. Unfortunately, Napoli (1976 )cannot explain impersonal si/sig constructions by means of the same transformational mechanism just as proponents of DI could not explain them. Consequently she assumes that Italian possesses two homophonous si's; one for "true" reflexives, reciprocals and middles and one for impersonal constructions. The latter is called by Napoli (1976 ) indefinite si. This solution, which is not unlike the proposal of Saloni ( 1 9 7 6 )to distinguish accusative and nominative sie^ in Polish, does not, unfortunately, explain why in two quite loosely related, if not practically unrelated, languages like Italian and Polish reflexive pronouns occur in both "true" reflexives and indefinite-subject/ impersonal constructions. Obviously, a systematic explanation of this fact would be superior to one relying solely on mere homaphony. Kanski (1984) is thus fully justified in criticizing this aspect of Napoli's (1976) solution to the problem of pseudo-reflexives. In order to prop up her theory of two different si's in Italian, Napoli (1976) attempts to characterize the semantics of the underlying element Pro for which si purportedly substitutes in impersonal constructions. She comes up with a list of features supposedly characterizing Pro. Kanski ( 1984 )literally obliterates this list using as counterevidence data from Italian and Polish (c.f. Kanski 1984: 28-33) . Kanski ( 1984 )later also argues against the last feature on Napoli's list which he does not dispose of immediately, the "humanness" of Pro. While I agree with Kanski's criticism of the other purported features, I shall postpone discussion of the obligatory human character of the subject NP underlying sie_ in

29

impersonal constructions until the discussion of Kanski (1984) in this chapter. Besides the problem of impersonal sig/si Kanski (1984 )claims to have noticed a number of other drawbacks of Napoli's (1976) analysis. For example Napoli (1976) claims that in English Just as in Italian the rule of ΟΟΡΪ moves the direct object NP into the subject position but, in distinction to Italian, it does not leave a copy. Hence the difference, between English on the one hand and Italian and Polish on the other, exemplified in ( 43 ) and ( 44 ) · ( 43 ) La porta si e aperta. "The door opened." ( 44 ) Drzwi otworzyiy sie,. "The door opened." Kanski ( 1984 )convincingly argues that while not as widespread as in Italian and Polish, reflexive expressions which do not involve coreference neverthless exist in English as shown in ( 45) ( c . f . Kanski 1984: 22-24) ( 45 ) The solution presented itself Napoli ( 1976) has also problems with left and right dislocation which she claims are other instances of the application of COPY rule. Napoli claims that such structures do not »feed* reflexivization because "...a reflexive pronoun is yielded when the two coreferential NPs have the same case." (Napoli 1976:98) . Kanski (1984 )however presents examples like ( 46 ) which serve as Polish counterevidence to this claim. ( 46 ) Ten idiota, widzialem go wczoraj w kinie. "That idiot -NOM, I saw him- AGO in the cinema yesterday." Finally Kanski (1984 )claims that Napoli (1976 ) is bound to run into the same problems, as all other analyses of reflexivization based on coreference, with sentences containing anaphoric pro-

30 nouns controlled "by quantified antecedents such as (41) . ( 4-7 ) Kazdy Polak chwali siebie samego "Every Pole praises himself."

.

yellowing Jackendoff (1972) t Kanski (1984- ) correctly observes that any analysis of reflexivization based on coreference will have problems in distinguishing (47 ) from (48 ) . ( 4 8 ) Kazdy Polak chwali kazdego Polaka. "Every Pole praises every Pole." Horn (1977 )attempts to analyze Italian and Polish data in the context of a somewhat different framework, interpretivist Revised Extended Standard Theory, distinguishing both semantically and syntactically sentences like ( 28 ) from sentences like ( 29 ) . Horn's (1977) main argument for providing such distinction is connected with cooccurrence restrictions. Thus ( 51 ) , which Horn (1977) calls middle, will cooccur with the same adverb as transitive (49) , while inchoative ( 5 0 ) will not. ( 49 ) Luigi ruppe la finestra inavvertitamente. "Luis broke the window inadvertently." ( 50 )*La finestra si ruppe inavvertitamente. "The window broke inadvertently." ( 51 ) Le mele si vendono inavvertitamente. "The apples are sold inadvertently." According to Horn (1977) inchoatives like ( 5 2 ) should have as underlying representation structure like (53 ) while middles like (54 ) should look in their deep structure like (55 ) . ( 52 ) La finestra si ruppe. "The window broke."

31 (53 )

la finestra

rompere A

( 54 ) Le mele si vendono. "The apples are sold." (55 )

Λ

vendere

le mele

Horn ( 1977 )claims that si/sig in nonreflexive uses will occur only with verbs which otherwise can occur in transitive structures i.e. with two lexical NPs, whenever these verbs occur with one lexical NP. If only one lexical WP cooccurs with such a verb ( e . g . (53) and ( 5 5 ) ) , a copying rule fills in the unlexicalized direct object NP with element PRO ultimately realized as si/sie ( 53) , for ( 55) the application of the copying rule being preceded by a rule moving direct object to the subject position. Polish impersonal sie^ constructions like (56 ) are according to Horn (1977) essentially similar to Italian middles ( 54 ) and hence their deep structures look like ( 5 5 ) . ( 56 ) Jablka sprzedaje εις w opakowaniu. "Apples are sold wrapped up." Existence of agreement between the original direct object and verb in Italian examples ( 5 4 ) and lack of such agreement in Polish examples ( 5 6 ) Horn (1977) attributes to the supposedly late application of the copying rule in Polish after the application of case assignment and verb agreement rules. Horn (1977 ) further claims

32 that deep structure like ( 53 ) underlies in Polish not only inchoatives but also examples like ( 57 ) , which according to Horn (1977) are "more or less 'passive* in meaning (i.e. they correspond to English agent less passive ".)( Horn 1977:33 ) (57) a. b. c. d.

Jan urodzii si$ w 1948 roku. "Jan was born in 1948." Szkola utworzyia sie, w Poznaniu. "The school was founded in Poznan." Jan wychowal sie, u dziadka. "Jan was brought up at his grandfather's." Miasto ulokowalo sie. na brzegu rzeki. "The town was located on the bank of the river."

Unfortunately for Horn's (1977) analysis he hopelessly misinterprets both Polish and Italian data. Kanski ( 1984 )mercilessly demonstrates it by presenting Polish impersonal sie_ constructions with clearly intransitive verbs ( 58 ) , constructions which cannot be derived from structures like(55 ) by any stretch of imagination. ( 58 ) a» b.

Tak sie. nie mowi. "One doesn't/shouldn't speak like this." Siedzialo sie, i sie, gadaio. "One was sitting and one was talking."

Kanski (1984) further proves that Italian also possesses genuine impersonal constructions ( 5 9 ) which exactly mirror the properties of polish impersonal constructions and similarly cannot take structure like ( 55) as their underlying representation. ( 59 ) a. b.

Quando si e tristi, si beve. "When one is sad, one drinks." Si va a teatro. "One goes to the theater."

Also the adverb cooccurrence restrictions argument is based on false assumptions. Horn (1977) fails to notice that examples exactly like ( 51) but without the adverb are ambiguous between

33 impersonal and 'middle' reading. Only by virtue of that ambiguity, may (51) contain the same adverb as transitive (49) . Polish also possesses such ambiguous sentences e.g. ( 60) . (60) To okno zamyka sie, latwo. "This window shuts easily." or "One shuts this window easily." In case of (51) the adverb simply disambiguates it, imposing on it an impersonal reading. Impersonal constructions simply share adverb cooccurrence restrictions with transitive sentences unlike inchoatives and middles. Thus Korn»s ( 1977 ) identification of Polish impersonal sie^ nominativeless constructions like (56) with what he calls middles in Italian ( 54) » with surface nominative NPs agreeing with verbs like subjects, is absolutely unmotivated. "There are three immediate consequences of Horn's observational errors: a) there is no motivation for distinguishing between "middle" and "inchoative" structures with si/sie. on the ground of adverb cooccurrencies facts; b J on the grammatical reading of the test sentences, the lexical NP( other than si/sie itself) is accusative, and the sentences are assigned an "impersonal" reading, hence c) these sentences cannot be derived by copy from L here (53)3 or hereC(55)3 ; for a lexical NP ( other than si/sie^ itself) is not obligatory in such structures, which can freely occur with intransitive verbs ( c.f, Chere (58)3 and Chere (59)3 . Thus the whole of Horn's analysis and its motivation collapses." (Kanski 1984: 48-49) . In the light of Kanski's (1984) arguments one must fully agree with his harsh criticism of Horn (1977) . However, it is not beside the point to remark that in spite of the general inadequacy of Horn's ( 1977) analysis there are grains of observational truth present in it, such as the general observation that reflexively marked constructions often involve some form of detranzitivation or that some reflexive constructions have a more or less "passive" character. For the record it also ought to be mentioned that kanski ( 1984 ) with considerable zest characterizes the theoretical muddle Horn ( 1977 )falls into because of his proposals. This however need not unduly concern us, since Horn's ( 1977 )theoretical inconsistencies are meaningful only in relation to the particular lexicalist framework he embraces. Kardela ( 1979 )does not introduce anything new to the discussion

34 of Polish sie_ constructions. As only to "be expected from an adherent of trace theory, Kardela ( 1979 ) introduces both "reflexive" and "non-reflexive" sie, pre-transformationally as base lexical items. However, since he distinguishes two completely unrelated lexical items sie_, one a particle in inchoative, middle and impersonal constructions and the other a reflexive pronoun in reflexive and reciprocal constructions, he must introduce them under different category labels. As Kanski (1984) observes, it is not clear what the label "particle" means. Kanski (1984) also points out other theoretical inadequacies of Kardela's (1979) analysis such as the highly inconsistent use of χ convention, ( c . f . Kanski 1984: 61-65) . I shall therefore further· dismiss Kardela (1979) as a nonuniform and highly uninspiring analysis of Polish sig constructions. The most thorough analysis· of Polish sie^ constructions, to date is certainly offered by Kanski ( 1984) . ^anski's (1984) critical examination of Granmer (1976) , Napoli (1976 ) , Horn (1977) and Kardela (1979) allowed him to state that none of these analyses were capable of uniformly characterizing Polish sie, constructions or analogical constructions in Russian and Italian. lanski (1984) attributes this fact to a number of assumptions shared by all these authors, assumptions which, accordingly to Kanski (1984) , jointly spell out a misconceived strategy of interpreting the data. These assumptions Kanski ( 1984: 68-69 )represents as follows: (61) a. b.

c.

d.

Sie/si/-sja CAN be reflexive pronouns. There are sets of sentences in which sie/si/-sja DO NOT function as reflexive pronouns, e.g. middle, inchoative, impersonal. The relationship between REFLEXIVE and NON-REFLEXIVE (uses of ) sie/si/-sjj, is not only that of phonological identity. If the latter assumption is entertained, i.e. if reflexive and non-reflexive sie,/si/-sja are not mere homophones, then There are syntactic-transformational and/or interpretive-lexical procedures which EXPLAIN the non-accidental phonological identity of the two (or more ) uses of the item.

35

e.

There is a level of analysis where the non-reflexive and reflexive 3ieVsi/-sja are "assimilated" (in the form of identical inputs to transformational and/or interpretive rules) .

Assumptions in (61) , if Jointly entertained, yield a set of contradictions which Kanski( 1984: 69-70 ) represents in the form of ( 62) . ( 6 2 ) a. b.

sie/si/-sja both are and are not reflexive pronouns, sie/si/-sja both are and are not ( pronominal) NPs.

Kanski ( 1984) avoids contradictions (62 ) by abandoning assumption ( 61 a ) . Thereby in his analysis sig is never a product of coreference. In consequence Kanski's( 1984) analysis is closest to Cranmer*s (1976) derived intransitivity analysis in spirit. Needless to say Kanski (1984) avoids the problems connected with DI. ^anski's (-1984) analysis is carried out within the framework of Montague grammar. It is outside the scope of this thesis to get into technicalities of this highly formalized and complicated framework. The remarks which follow will be thus a very simplified, unformal and nontechnical rendering of Kanski's ( 1984) account of sig constructions. In Kanski's (1984) sieVsi/-sja do not have the status of reflexive pronouns but are basic lexical items of the category NP. These NPs are non-referential or in more formal jargon are free individual variables. Occurring as arguments of transitive verbs these NPs satisfy transitive verbs with required syntactic arguments of appropriate category, while in disambiguated language LI , into which all natural language sentences are translated , the same free variables when combined with verbs functioning as nonlogical contants create an effect as if nothing really has been added to the verbs. Thus a two place relation will characterize at a given context only one set of individuals. "The latter requirement is met without contradiction in the following ways:

36

(-63) A« a two-place relation R characterizes at an index two IDENTICAL seta» identified as the projections of R; this result includes reciprocity/symmetry and its narrower case-reflexivity. ΒΦ a two-place relation R characterizes at i either of its two POSSIBLE projections." ( Kanski 1984Γ 3H ) ( 63B) is intended to characterize "both constructions of type (28) and(29), which Kanski (1984) calls Left-middles, and constructions of type (30) called by Kanski ( 1984) Right-middles. In impersonal constructions sie a non-referential argument of a verb will likewise in L yield an interpretation identical to that of a verb without sig, while simultaneously satisfying the verb with the required syntactic argument, this time the subject NP argument. Furthermore since slg_ often is the only syntactic argument of the verb in such constructions and the denotation of an intransitive verb by itself does not enable truth evaluation, otherwise optional complements such as various adverbials are required to be present via Obligatory Complement Constraint. Thus, very roughly, sig is a semantically empty( because of its non-referentiality)item which fills out syntactically either the subject (impersonal sig )or direct object (middle, reciprocal, reflexive sie.) slots, thereby inducing relation reduction. Kanski (1984) undoubtedly puts forwards, good explanation for the reflexive marking of "genuinely" reflexive, reciprocal and middle constructions. However his greatest achievement, as he himself acknowledges, is providing an explanation for impersonal sig constructions within the same analysis. It is precisely at this point that his analysis turns out to be most vulnerable. Kanski (1984) observes that impersonal sie constructions seem to require the presence of otherwise optional complements. To ensure that this actually takes place he formulates Obligatory Complement Constraint (OCC) which formally imposes the presence of such complements in impersonal sie^ constructions as a necessary requirement for their well-formedness. Examples (64) and( 65) may serve as data confirming this solution. (64)*Idzie si?. "One walks."

37 (65) W dol idzie si

( 123) is a representation of an unaccusative construction. Like (121 ) and (122) it involves a 2 arc sponsoring a 1 arc successor. Unlike ( 1 2 1 )and ( 1 2 2 ) it does not involve a Cho arc sponsored by a 1 arc since it has no initial 1 arc to be overrun. Similarly to D in ( 122) B in (124) may be replaced by a copy arc headed by a

61

reflexive pronoun. PN Laws 45 and 4-6 ensure that arcs B in (123) and ( 124) must have 1 arc successors. Thus the isomorphism of reflexive and passive constructions carries over to initially unaccusative constructions.

(125)

C C C

1 2 3

(126)

C C C

1 2 3

(127 )

C C C

1 2 3

62

Diagrams (125) - (130) illustrate PNs with non-initial unaccusative strata. ( 1 2 5 ) and (126 ) exemplify antipassive constructions in which the initial subject is demoted to direct object. This yields unaccusative strata whose properties are similar to those of initially unaccusative strata. Hence the possibility of a reflexive-headed copy arc D replacing B in (126) . It ought to be added that, in principle, antipassivization is also possible with transitive clauses, in which case the initial stratum direct object is overrun and must have a cho arc successor (e.g.(127)). (128 )

C C C

1 2 3

( 129 )

,°IC2C3 ( 130)

C C C

1 2 3

Dummy

Refl

63

Diagrams (128) - (130) are instances of inversion, where initial subjects sponsor indirect object successors. Resulting unaccusative strata will be unstable like initially unaccusative strata. Thus in their final strata the presence of 1 arcs will be required. APG PN Laws allow a number of alternative ways in which that might happen. Either no copy is involved ( 128 ) , or a copy arc replaces the initial direct object arc leading to reflexivization ( 1 2 9 ),or the initial subject arc sponsors a dummy-headed direct object ghost arc which upon being promoted to subjecthood as required by PN Law 84 may also leave a reflexive copy (130) , although as (131) shows this is not necessary.

(131)

Numerous languages will require self-erasure of the dummy-headed arcs thereby making them final arcs but not surface arcs. To further complicate this analysis there are no universal principles which would block chomeurization of potentially overrun initial stratum indirect objects e.g. (132 ) or preclude overrunning" and chomeurization of initial stratum direct objects by dummy ghost arcs in structures corresponding to (128) e.g. in ( 132 ) . ( 132 )

64

(133)

(132)and (133) represent relevant subsets of appropriate PNs which depending on the presence or absence of dummy-headed nominal arc or presence or absence of copy arc headed by reflexive pronoun, will correspond to one of the diagrams (128 )- (130) .

(134)

(135)

Dummy

65 Diagrams (134 ) and ( 135 ) illustrate impersonal constructions. Both (134) and (135 ) involve "0"-headed arcs sponsoring dummy-headed ghost arcs. PN Law 84 requires ghost arcs in both diagrams to have 1 arc successors. Principle (119 )permits the erasure of the overrun "0"-headed arcs. The occurrence of successor-predecessor parallel arcs allows for reflexivization to surface in (134 ) in the form of replacer D. Both (134) and (135) illustrate intransitive impersonal constructions. In transitive ones the initial 2 arc may be either overrun by a dummy ghost arc, in which case either of the diagrams (134 ) - (.135) would additionally contain a c 1 2 arc and its c? cho arc successor, or itself have a 1 arc successor, in which case there is no dummy ghost arc. Genetic PNs (136) and (137) are appropriate partial PN representations of these cases.

(136)

(137)

66

(136) would represent so-called reflexive passives while (137) patient subject constructions. All of the PNs discussed above contain parallel arcs, either self-sponsoring or successor-predecessor pairs. While it is not necessary for successor-predecessor pairs unlike self-sponsoring pairs to involve replacer arcs ( e . g . (121) , (123) , (125) , (128) , (131) , (135) and (137 ) > t is nonetheless possible (e.g. (122) , ( 1 2 4 ) , (126) , ( 1 2 9 ) , ( 1 3 0 ) , (134) and (136)). Given the isomorphism between successor-predecessor parallel arcs and self-sponsoring parallel arcs, it ceases to be strange that one should encounter in numerous languages reflexive passives, that impersonal constructions will often take reflexive or passive marking, that initially unacousative predicates will often be marked with reflexive particles and that inversion and antipassive constructions will likewise often take reflexive marking. Given the APff framework all of the Pair Networks discussed above are in principle possible. It is a matter of language particular rules to eliminate from them those unattested in particular languages. 2.5.

The Hypothesis

The hypothesis to be verified in this work might be summarized as follows. In comparison to English, Polish abounds in reflexively marked constructions. Previous attempts to explain this fact have been unsuccessful. Kanski's( 1984 ) analysis, which comes closest to a comprehensive explanation, is not without its drawbacks. APG analysis of reflexivization involves pairs of parallel area or at least overlapping arcs. These pairs may be self-sponsoring or may consist of successors and their predecessors. APG analysis of semantically nonreflexive reflexively marked constructions is essentially a Copy analysis. Contrary to Kanski's (1984 Assessment, the APG variant of Copy analysis is capable of coping with impersonal reflexives. APG Copy analysis ties in with the unaccusative hypothesis. Some intransitive verbs would be classified as initially unaccu-'.. · sative and will require initial 2 arcs to have 1 arc local succes-; sors. PNs of initially unaccusative clauses, noninitially unaccu--

67

sative clauses (e.g. inversion or antipassive constructions) , impersonal clauses, and passive clauses will all include pairs of predecessor-successor arcs, isomorphic with self-sponsoring parallel arcs of semantically reflexive constructions. The set of PN Laws, theorems, and definitions presented in this chapter constitutes a viable tertium comparationis for analyzing reflexivization in English and Polish. Genetic PNs, presented above, exhaust the set of PNs with parallel 1 arcs and 2 arcs which are either self-sponsoring or constitute successor-predecessor pairs. Any PN representation of any polish or English reflexively marked construction should theoretically contain as its proper subsets one of the arc sets in (120 )- (137)(unless the reflexive pronoun heads a 3 arc, oblique arc or Gho arc replacer, or an immigrant arc is one of the two parallel arcs) . The comparison of Polish and English constructions in terms of these Genetic PNs should yield sets of rules for both languages, which would at least partly account for the discrepancy between English and Polish in the number and frequency of reflexively marked constructions. The last chapter of this book is devoted to the verification of the above hypothesis.

68

CHAPTER

III

VERIFICATION

3*1.

Total and Part Reflexives

According to Johnson and Postal ( 1980 ) , semantically reflexive constructions are best represented by Pair Networks with parallel self-sponsoring arcs. Thus both (138 )and (139) will have the similar representations, ( 140 ) and (141) · (138) Janek my je sie.. "John is washing himself." (139) John scratched himself.

(140)

C..C,

Janek

(H1)

John

scratched

69 In direct object positions Polish will invariably have sie or siebie as reflexive pronouns while English will have myself, yourself, himself, herself, itself, oneself, ourselves, yourselves and themselves. Polish indirect object reflexive pronoun will be always sobie (or substandard se ) while English will employ the same items as in direct object position. It thus follows that in Polish the form of the pronoun which heads the replacer arc depends on the relational signs of the cosponsoring parallel arcs while in English only the properties; of the node which heads the two parallel arcs are relevant for the form of the pronoun which heads the replacer arc. Polish rules therefore need not mention any properties of the node which heads the cosponsoring parallel arcs while English rules need not refer to relational signs of the parallel cosponsoring arcs. A first approximation to the set of rules for reflexivization in English might be formulated as in (142) - (150) while the first partial rendering of corresponding Polish rules is given in (151 ) - (153 ). ( 142) English Rule 1 Head ( a , A ) A H e a d ( a, Β )Λ Parallel ( Α,Β) Λ Not (Erase ( Α , Β ) ) A Node Label ( I , a ) — » ( 3 C ) ( Replace ( σ,Β )Λ Head (b,C) Λ Node Label (myself, b ) ) (143) English Rule 2 Head (a,A)A Head (a,B ) A Parallel ( Α,Β ) A Not (Erase ( Α , Β ) ) Λ Node Label( you,a) Λ Node Label ( singular,a)—»(3 C ) (Replace ( Ο , Β ) Λ H e a d ( b , C ) A Node Label (yourself,b))

( 144) English Rule 3 Head (a, Α) Λ Head( a,B )A Parallel ( Α , Β ) Λ Not (Erase ( Α , Β ) ) Λ Not (Node Label( I , a ) V Node Label (you,a ) )ANode Label ( singular,a) A Node Label ( ftminine,a ) — > ( 3 C ) (Replace (G,B) A Head ( b,C ) Λ Node Label (herself,b)) ( 145) English Rule 4 Head (a,A) A Head (a,B) A Parallel ( Α,Β) A Not (Erase ( Α , Β ) ) Λ Not (Node Label ( I , a ) v .Node Label (you,a)) A Node Label (singular,a) A Node Label ( masculine,a ) ^ ( 3 0) (Replace (c,B) A Head ( b , C ) A Node Label (himself,b))

70

( 146) English Rule 5 Head (a,A) Λ Head (a,B) A Parallel ( Α , Β ) Λ Not (Erase(A,B)) Λ Not (Node Label (I,a) V Node Label (you,a)) A Node Label (singular,a) Λ Node Label ( neuter,a ) *(30) (Replace ( C , B ) A Head (b,C ) Λ Node Label (itself,b)) ( 147) English Rule Head (a,A) Λ A Node Label Λ Node Label

6 Head (a,B ) Λ Parallel ( Α,Β) Λ Not (Erase (Α,Β)) ( 0 , a ) — * ( 3 C ) ( Replace ( C,B) A Head (b,C) (oneself,b ))

(148) English Rule Head ( a , A ) A Λ Node Label A Node Label

7 Head (a,B )A Parallel ( Α,Β } Λ Not (Erase ( Α , Β ) ) ( w e , a ) — » ( 3 0 ) ( Replace ( C,B ) A Head ( b , C ) (ourselves,b ) )

( 149) English Rule 8 Head ( a , A ) A Head (a, B )A Parallel ( Α,Β ) Λ Not (Erase ( Α , Β ) ) Λ Node Label (you,a) Λ Node Label ( plural,a ) —>( 30 ) (Replace ( C , B ) A Head ( b , C ) A Node Label (yourselves,b)) ( 150) English Rule 9 Head (a, A) A Head (a,B) A Parallel ( Α,Β ) A Not (Erase ( Α , Β ) ) A Node Label (plural,a) A Not ( Node Label (we,a ) ν Node Label (you,a ) ) — > ( 3 3 ) ( Replace ( C,B ) A Head ( b , C ) A Node Label (themselves,b)) (151) Polish Rule 1 Parallel ( Α , Β ) Λ 1 arc (A) A 2 arc ( Β ) Λ Not (Erase ( Α , Β ) ) > (3 C ) ( Replace (C,B)AHead (b,c)A( Node Label (sie.,b ) V Node Label (siebie,b )) ( 152) Polish Rule 2 Parallel (Α,Β) Λ 1 arc (A) A 3 arc ( B ) A Not ( Erase ( Α , Β ) ) => ( 3 C ) (Replace (C,B) A Head (b,c) A Node Label ( sobie,b )) (153 ) Polish Rule 3 Parallel (Α,Β) A 1 arc (Α) Λ Inst arc ( B )'A Not (Erase(A,B)) 3 C)(Replace (C,B) A Head ( b , C ) A N o d e Label( sobq.,b) )

71

Strictly speaking, it is redundant to state tiW.ce that node a heads two arcs in English rules 1-9, since by virtue of the fact that A and B are parallel arcs the head of A must also be the head of B. The Polish set of rules is anything but complete. This is due to the fact that reflexivization in polish is sensitive to the relational signs of the cosponsoring pair of parallel arcs. As it is not at all clear what other members the set of the oblique relational signs might include besides Instrumental, Locative, Genetive or Benefactive, it would be premature to formulate any final set of rules for all arcs with oblique relational signs. Thus the rule governing the behavior of replacer arcs with Inst relational sign is only given as an example of how to formulate appropriate rules for other oblique signs and Oho arc sign replacer arcs, once the set of all oblique relational signs is fully specified. The formal differences between the Polish and English rules have some curious implications for the analysis of English emphatic reflexives (154) - (155 ) and their Polish nonreflexive equivalents (156) - (157) . (154) Mr Brown himself came to the party. (155) I'll

do it by myself.

(156) Sam pan Kowalski przyszedi na przyj^cie. (157) Zrobie, to sam. It is not clear what is the relational sign of arcs headed by sam (or sama, samo, sami, same, depending on case, gender and number )in the Polish examples and himself/myself (or any other 'self words )in the English ones. It is, however, quite likely that they are replacer arcs( especially English data seems to indicate that) . Should this be the case, the PN representation of (154) and (156) would look like (158 ) .

72

(158)

Mr Brown Pan Kowalski na Arc B in (158) would either require a relational sign not included in the set of relational signs discussed "by Johnson and Postal (1980) ( e.g. Emph for emphatic ) or would be an immigrant arc i.e. a foreign successor of an arc in another clause (incidentally, Johnson and Postal ( 1980 ) hint that all oblique arcs may ultimately turn out to be immigrants ) . Whichever is the case, English rules 1-9, sensitive only to the properties of the node heading the two parallel arcs A and B, will allow for an appropriate reflexive pronoun to head the replacer arc 0 just as in the case of semantic reflexives. On the other hand, Polish rules, sensitive to relational signs of the parallel arcs, will not handle -emphatic cases since, whatever the relational sign of B in (158) is, it is not 2,3 or any of the hitherto known oblique relational signs. In consequence, the fact that English employs.,reflexive pronouns in emphatic constructions and Polish does not, ties in very neatly with the divergent formulation of rules for reflexivization in the two respective languages. Another problem is connected with constructions called by Niedzielski (1976) "part reflexives". Unlike total reflexives, part reflexives inform about actions which seem to be reflected not on their performers or instigators but on parts or possessions of the performers or instigators. This category will include the following examples:

73 (159) Janek skrzywil sie,. "John scowled." ( 160) Kowalski buduje sie,. "Kowalski is building (his house)." Examples like (159) and(160) may be analyzed as either closer to total reflexives like (140) or as more similar to possessive reflexives as in (161) . ( 161) Kowalski maluje swoj dorn. "Kowalski is painting his house." The first option would have to rely on an analogue of Kalisz*s (198J) analysis of beheading constructions. Kalisz (1983) discusses examples like (162) , which according to Borkin (1971) involve a transformation called "beheading". (162) Chomsky takes up a foot on my bookshelf. Kalisz (1983) argues that (162) does not involve any chopping transformation, akin to Borkin's ( 1971 ) beheading, but rather should be analyzed in terms of properties characterizing the semantic content of the lexical item Chomsky. As properties gain or lose in salience, lexical items may be used more or less prototypically. Sometimes lexical items denoting particular entities will be used in contexts where they actually refer to parts or possessions of these entities. Since one of the properties of the referent of the lexical item Chomsky is writing books, we may safely use this word instead of the more explicit circumlocution books by Chomsky. By virtue of analogical reasoning we may say that the lexical item Kowalski in (160) may stand for whatever Mr Kowalski is constructing for himself, be it a barn, house, or garage, and therefore may function as both subject and direct object in (160) thus heading two parallel selfsponsoring arcs as is customary for total reflexives. In terms of this analysis part reflexives like ( 159 ) and (160 )would have PN representations (163) and (164) .

74 (163>

Jane k

sie. skrzywil

( 164)

Kowalski

sie.

buduje

One of the predictions of the above analysis is that part reflexives would mirror all the syntactic properties of total reflexives. Unfortunately this is not the case since, unlike total reflexives, part reflexives do not allow sieMe to head the replacer arc. (165) Kowalski buduje siebie. "Kowalski is building himself." (166) Kowalski skrzywil siebie. "Kowalski scowled himself." Both ( 165) and (166) look at least strange and under no circumstances would they mean what (159) and ( l 6 C > ) mean. Yet there is

75 no way to preclude polish Rule 1 from applying to (163) and (164). There is also no way to formally exclude siebie from appearing as head' of re placers in structures like (163) and (164 ). Furthermore, Polish part reflexives do not have corresponding reflexive equivalents in English. Yet, given PFs ( 163 ) and (164) , there is no formal way of blocking similar English part reflexives since there is no formal difference between total and part reflexives. Thus theory would clearly contradict data. An alternative analysis of part reflexives, closer in spirit to Borkin' s (1971) transformation! of beheading, would avoid these problems. Under this analysis ( 160 ) will have a PN representarion like (167) .

(167 )

Kowalski buduje

In (167) B is immigrant arc which together with A sponsors replacer G. D, overrun by B, has a Cho arc successor. P is erased by its foreign successor B and G self-erases by virtue of PN Law 80 ( UN Node Headed Self-Erased Law ) . E would selferase due to the erasure of F and G (a PN Law should be formulated to ensure the self-erasure of undermined arcs just like PN Law 81 ensures self-erasure of most overmined arcs ) . PN (167) differs both from PNs with successor-predecessor pairs and self-sponsoring parallel arcs. In spite of the differences it is sufficiently similar to them to contain a reflexive headed replacer arc. If Polish part reflexives were to be analyzed in terms of PNs similar to (167) the Polish system of rules would have to be supplemented by a rule which would disallow siebie

76 as the head of the replacer in (167) · ( 168 ) shows how such rule could be formulated. (168) Polish Rule 4 ParalleK A,B) A 1 arc (Α) Λ 2 arc ( Β ) A Foreign Successor (B,C) Λ Gen arc (C ) — ^ ( 3 D ) ( Replace (D,B) Λ Head (b,D) A Not (Node label( s i e b i e , b ) ) )

English will on the other hand need a rule disallowing part reflexives altogether. Possibly (169) could do the job. (169) English Rule 10 G-en arc (A)—> Not( a B) ( 2 arc ( Β ) A Successor ( B , A ) ) At first glance the second analysis seem to be better tailored to cope with the Polish part reflexives in APG terms. Time and future research will show whether it is the proper analysis. I leave unanswered a number of undoubtedly important problems, such as to what extent the predicate in the main clause will determine the possibility of a Gen arc having a 2 arc successor, what regularities underlie the behavior of Gen arcs in general, how the PN law imposing self-erasure on undermined arcs like Ε in 067 ) is to be formulated, etc. 3.2.

Middles or Initial Unaccusativss

The label'middles» usually refers to reflexively marked inchoative verbs. KaAski (1984) would include under the same label so-called reflexive passives. Furthermore, psych movement verbs discussed by Fisiak, Grzegorek and Zabrocki ( 1978) , may perhaps also belong here. Like passives, middles will take patients as subjects, unlike passives they will not take chomeur agents and will have verbal morphology characteristic for the active voice. In Polish middles will take sig in direct object positions. Characteristically, English translations of Polish middles will never be reflexive. Since the agent in Polish middles is unspecified, and indeed in some cases it is hard to visualise any hypothetical agent,

77 e.g. (170) , Genetic PN (124) seems to best reflect all their properties. A partial representation of ( 170) would look like ( 171) . ( 170 ) Ten mur rozsypie sie_ kiedys. "This wall will crumble one day."

(171 )

Ten mur

rozsypie

kiedys

The English translation of (170) would "be represented by a PN like (172>. ( 172)

This wall

will crumble one day

Middles will thus correspond to initially unaccusative clauses, although not all initially unaccusative clauses will be reflexively marked in Polish and hence not all initial unaccusatives would be traditionally labelled as middles.

78 In English initial unaccusatives will never appear in the guise of reflexive marking. The simplest way to ensure that English initially unaccusative clauses will always be analyzed in terms of PNs like (172) and not (171) would be to introduce into the system of rules for English a rule like (173) · ( 173) English Rule 11 Parallel (Α,Β)Λ Successor ( A,B )-»Not (3 G )(Replace( C , B ) ) English Rule 11 would disallow in English all reflexive inchoatives as well as reflexive passives, reflexive inversion and antipassive constructions, and reflexive impersonal constructions. To my knowledge none of the above listed constructions have been attested in English. ( 173) would thus have the desirable consequence of blocking all of them at a stroke. In Polish the situation is much more complex. The existence of reflexive inchoatives clearly shows that reflexive headed replacers should be allowed to replace predecessors in some successor-predecessor pairs. Equally numerous nonreflexive initial unaccusatives show that in other cases replacer arcs should be blocked. It is not at all clear what principles govern these phenomena. A first step toward unravelling the mysteries of reflexively marked initial unaccusatives was taken by Horn (1977) when he claimed that sie in nonreflexive uses will occur with verbs which otherwise can occur in semantically transitive clauses. Kanski (1984) points out that this is not the case since semantically intransitive verbs freely cooccur with sig in impersonal constructions. But this is where Kanski's (1984) criticism is only partly justified for,if we exclude impersonal constructions and a few isolated cases of initially unaccusative predicates such as rozpadac slg "to crumble" or ρο,-jawiac sie "to appear", Horn's (1977) hypothesis seems to hold true. In particular, I do not know of a single case where a verb may occur in transitive clauses and yet its unaccusative homophone will not cooccur with sig. Horn's (1977 ) observation might be translated into an APG rule as (174 ) .

79

(174) Polish Rule 5 Unaccusative (A,ck (a ")) A Tail ( a>A ) Λ V-stemmed (a,ID )A ( 3Q) ( R-graph( Q ) A Node Extraotable ( d , Q ) A V-stemmed ( d , b ) A Tail (d,E ) Λ -Accusative ( Ε,ο., ( d ) ) ) — * ( 30) (Replace ( C , A ) ) (174) exemplifies an APG equivalent of a transderivational rule as defined by Sadock (1975) . The notation "V-stemmed (a,b) " is used by Postal (1979) to denote the relation between nodes like a and b in (175) ·

(175)

J The reason why arc E in (174) must be a c^ arc is that the precondition for having reflexively marked unaccusatives is the existence of their semantically transitive ( i . e . transitive at c^ stratum) homophones and not noninitial, and thus nonsemantic, transitivity of the type exemplified in impersonal constructions like (134) . Polish Rule 5 ensures that such initially unaccusative predicates as: palic sie., unosic sie., lad sie, sq,czyc sie, husta sig, przewrocic aig, rozwijac si^, toczyc sig, suszyc sig, gotowac SJQ, giq.c sig, rozjasniac sie^, dusio sie,, rumienic sie^, otwierac sie, zamykac sig, iamac sig, roztrzaskac sie, etc. will all be accompanied by sie,. A separate rule must account for the fact that no initially unaccusative predicates may cooccur with siebie instead of sie.« ( 1 7 6 ) will disallow siebie as the head of a replacer arc which replaces a member of a successor-predecessor pair. ( 176 ) Polish Rule 6 Parallel ( Α , Β ) Λ Successor ( A , B ) A ( 30 )( Replace (C,B)A Head ( a , C ) ) — > Not (Node Label ( s i e b i e , a ) )

80

Polish Rule 6 will not only block siebie in initially unaccusative clauses but also in reflexive passives, inversion and antipassive clauses and impersonal constructions. Since siebie never occurs instead of sie, in any of these, a rule like (176 ) is a welcome addition to the grammar of Polish. While (174) correctly accounts for most Polish reflexively marked predicates in initially unaccusative clauses it will not account for all cases. Some predicates will require a rule disallowing the replacer altogether ( e . g . predicates like wrzec, umieradt zamarzac, trzeszczeo, etc· ) while others require a rule making rfehe replacer obligatory, even if the predicates do not have transitive homophones and therefore do not fall under rule 074) (e.g. predicates like pojawiaö sie, rozpadac sie., ulatniac siq, etc.). Unfortunately there do not seem to exist any clear semanto-syntactic principles which would determine why one initially unaccusative predicate, which has no corresponding transitive homophone, should cooccur with sie and another should not. Perhaps to find an answer one should indulge in diachronic studies of the problematic predicates. In synchronic terms the only possibility available to formulate appropriate rules seems to be to list lexical items to which the rules apply in the rules themselves. Incidentally, this is exactly what Postal (1979) does when he formulates rules for French extraposition of indefinite (EXI) construction. While it is obviously preferable to find in languages deep set regularities, sometimes no alternative to mere listing is possible« Linguistics is after all not mathematics« The best approximation to the necessary rules which I can presently envision is embodied in (177 ) and (178 ) . ( 177) Polish Rule 7 Unaccusative (A,c k ( a ) ) pojawiad rozpadao b 6 s ulatniac

Tail ( a,A )AV-stemmed (a,b )A

C )( Replace ( C,A ) )

81

( 178) Polish Rule 8 Unaccusative (A,c k ( a ) ) A T a i l ( a,A )AV-stemmed ( a , b ) A pojawiac rozpadao ulatniac > A Not( gQ ) ( R-graph ( Q Wiode Extractable

( d , Q )AV-stemmed ( d , b ) A T a i l (d, Ι) Λ Accusative (E,c.j ( d ) ) ) - * Not(a G) (Replace ( C , A ) )

Polish, rules 5-8 seem to cover all the initially unaccusative cases. However several marginal problems need mentioning here. Kanski (1984) lumps under the label "middles" examples like (179) - (181) and (182) - (183) unlike Horn ( 1977) who calls the former inchoatives and the latter middles. ( 179) Drzwi otworzyly εις. "The door opened." ( 180) Szklanka zbiia si$. "The glass broke." ( 181) Lekcja skonczyla 3ΐς. "The class finished." (182) Baweina dobrze 3ίς pierze. "Cotton washes well." ( 183) Ta ksia_aka dobrze εις csyta. "This book reads well."

Should Kanski (1984) "be right in making equivalent structures like ( 1 7 9 ) - ( 181) and (182) - (183) , then the latter examples v/ill fit neatly into the paradigm of initial unaccusatives and fall under rules (174) and (176) . However, while Kanski (1984) is justified in criticizing Horn (1977) for equating structures like

82

( 182) - ( 183) with impersonal constructions like ( 184- )- (185) , this fact need not necessarily mean that examples ( 182) - (185) must be structurally identical with examples (179 )- (181) . ( 184 ) Τς baweine. dobrze sie. pierze. "This cotton is good to wash." ( 185 ) Te, ksi^zke. czyta sie. dobrze. "This book is good to read."

If it turns out that. examples like (182) - (183 ) actually involve unspecified human agents (like (184) - (185 )) and should be included under the label "reflexive passives" ( c.f. Woiczynska -Sudoi's (1977) notional passives) then they obviously will not fall under the above analysis of initial unaccusatives» PN representations of examples (182) - (183) would then look like ( 186) and ( 187 ) , and correspond to Genetic PN ( 136 ) . ( 186 )

Ο

bawelna

sie.

dobrze

pierze

( 187 )

Q

ta ksia.zka

sie,

dobrze

czyta

83

Since examples like (182) and (183 ) will always involve the presence of sie, and, given PN representations (186 )and (18? ), would not fall under rule (174) , a rule analogical to (174) should be devised for these cases, perhaps along the lines of ( 188 ). ( 188 )Hypothetical Polish rule for reflexive passives, 2 arc (A)ASelf-Sponsor( Α )Λ Successor ( B,A )Al arc (B ) Λ Overrun (B,C)AHead Label (0,0 )-»( 3D)( Replace ( D , A ) ) Arc A in (188) is self- sponsoring to enable ghost arc analysis of -no, -to impersonal constructions. I will return to this issue in the section ou impersonal constructions in this chapter, I will leave here unresolved the question whether examples like ( 182) and (183) should be analyzed as initially unaccusative clauses or reflexive passives, although should other things be equal, Occam's razor would suggest the first option. A somewhat different problem arises with so called psych movement verbs. Fisiak, Grzegorek and Zabrocki (1978 ) use this label for predicates like the ones in (189 ) - ( 191 ) . ( 189 ) Jane k zachwyca sie. tym krajobrazem. "John is pleased with this scenery." (190 ) Janek dziwi sie, zachowaniu Zbyszka. "John is surprised at Bob's behavior." ( 191 ) Janek przestraszyi sie, zachowania Zbyszka. "John was frighted by Bob's behavior." C189 )- (191 )are contrasted with their apparent paraphrases (192 ) - ( 194 )respectively. (192 ) Ten krajobraz zachwyca Janka. "This scenery pleases John." (193 ) Zachowanie Zbyszka dziwi Janka. "Bob's behavior surprises John."

84

(194 ) Zachowanie Zbyszka przestraszylo Janka. "Bob's behavior frightened John." Examples (189 )- (191) are similar to middles since they employ 3J£ in direct object position and involve direct objects promoted to subjecthood (in this respect sentences (189 ) - ( 191 ) are also similar to passive constructions) . In examples ( 189 )- (191) Janek is the subject and sig functions as direct object just like in middles. In (192) ten krajobraz is the subject, while in (193) and in ( 194) zachowanie Zbyazka. Janek functions in all these examples as direct object. Polish data could well be accomodated in terms of APG promotion cum copy replacer arc analysis if it had not been for the problematic status of the NPs in (189) - (191) which correspond to subjects in (192) - (194 ) . Each of these NPs takes different case marking. Thus tym krajobrazem in ( 189) is in the instrumental case, zachowaniu Zbyszka in (190 ) is dative and zachowania Zbyszka in (191) is genetive. Current APG laws would require initial subjects, like the ones in (192 ) - (194 ) to have chomeur successors once they are overrun by direct objects promoted to subjects as in (189) - (191 ). It is not at all clear, however, why all these realizations of a single category, chomeur, should be assigned different case endings. (190 ) might be alternatively analyzed as an example of inversion, where a 1 arc has a 3 arc successor. This would tie in nicely with the dative case ending of the nominal zachowaniu Zbyszka in (190 ) . Unfortunately, no analogical explanation can be put forward for the other two cases. (189 )is particularly perplexing for if the nominal tym krajobrazem were to head a surface instrumental arc it could not be a successor of an initial 1 arc due to PN Law 49 (The No Oblique Successor Lav/ }. To add to the confusion, there are reasons to believe that examples (192) - (194) are not necessarily paraphrases of (189) - (191) respectively. By analogy to Ho±n (1977) , it is possible to construct an adverb cooccurrencies argument with precisely that conclusion in mind. For example (195 )means something quite different from (196 ) although if (191) is a paraphrase of (194) , (196)should be a paraphrase of ( 195 ).

85

( 195 ) Zbyszek celowo przestraszyi Janka. "Bob frightened John on purpose." ( 1 9 6 ) Jane k celowo przestraszyl sie_ Zbyszka. "John felt frightened because of Bob on purpose." ( 195 ) straightforwardly intimates that a particular individual schemed to frighten another individual. ( 196 ), somewhat, against common sense, implies that a particular individual on purpose grew afraid of another individual. On the other hand a passive sentence, like (197 ), which involves chomeurization of the initial subject and promotion of the initial direct object ( just as (196) supposedly does )has the same meaning as (195) · ( 197 ) Janek celowo sostal przestraszony przez Zbyszka. "John was frightened by Bob on purpose."

This clearly shows that (196) is not a paraphrase of (197 ) contrary to the argumentation that both (196 ) and (197 ) are variants of (195) , (196) a reflexive passive and ( 197 ) passive, and that both (196) and (197) involve 2 to 1 promotion and chomeurization of the original 1 arc. Similarly ( 198 )ia not a paraphrase of (199 ). ( 198 ) Janek ch^tnie zachwyca sie, tym krajobrazem. "John willingly takes pleasure in this scenery." ( 1 9 9 ) Ten krajobraz ch^tnie zachwyca Janka. "This scenery willingly pleases John." If the adverb cooccurrencies argument is valid constructions like ( 189 ) - (191) will not be paraphrases of ( 192 ) - (194) . For example, (189 ) would be closer in structure, if not in meaning, to ( 200 ) than to (192) which is indeed a claim that Saloni (1976 ) makes . ( 189 ) Janek zachwyca sie. tym krajobrazem. "John is pleased with this scenery."

86

( 200 ) Ten kraj sachwyca Janka swym krajobrazera. "This country pleases John with its scenery." ( 192) Ten krajobraz zachwyca Janka. "This scenery pleases John." ( I89)could then have a partial PN representation like ( 201 ) . ( 201)

Janek

εις

tym krajobrazem zachwyca

) t ( 190 ) and ( 191) would thus be initially unaccusative clauses and as such would be susceptible to rules( 174 ) and (1?6 ) . Predicates like ba ( 30 ) (Replace (C,B)AHead ( b , C ) A N o d e Label ( yourself ,b )) English Rule 3 Head (a,A>AHead (a,B)AParallel (A,B)MIot( Erase ( Α , Β ) )Λ Not ( Node Label (I,a)VNode Label( you,a ) )ANode Label ( singular, a )ANode Label (feminine,a ) — X 30 ) ( Replace (0,3) AHead ( b , 0 )ANode Label (herself , - b ) ) English Rule 4 Head ( a , b ) A H e a d (a, B) A Parallel ( A , B ) A Not ( Erase ( Α , Β ) ) Α Ν ο ΐ ( Node Label ( I , a )VNode Label ( y o u , a ) ) A Node Label ( singular,a ) ANode Label ( Masculine,a)—} (3 0)( Replace ( C,B ) Λ Head ( b , 0 ) Λ Node Label ( himself,b)) English Rule 5 Head ( a,A)AHead (a,B) AParallel ( A,B)/vHot ( Erase ( A,B ) )A Not ( Node Label ( I , a ) V Node Label (you,a ) ) Λ Node Label ( singular,a ) A Node Label ( n e u t e r , a ) — ^ ( 3 σ ) (Replace ( C , B ) A Head ( b,C )Λ Mode Label ( i t s e l f , b ) ) English Rule 6 Head ( a,A) A Head ( a , B ) A Parallel (A,3) A Not ( Erase ( A,3 ))A Node Label ( 0,a )-»( 3 0 ) ( R e p l a c e ( G , B ) Λ Head (b,0 ) Λ Node Label ( oneself,b ) ) English Rule 7 Head ( a,A ) A Head (a,F>) A Parallel (A, B) A Hot (Erase ( A , B ))A Node Label ( we,a )-* ( 30 ) (Replace ( C , D ) Λ Head ( b, 0 )A ITode Label ( ourselves,b )) English Rule 8 Head ( a,A ) Λ Head ( a , 3 )AParallel ( Α , τ ? ) Λ Γ ο 1 , (Erase ( A , B ) ) A I-lode Label ( you,a )ANode Label (Plural,a )—> ( 30 ) ( Replace ( C , B ) A r I e a d ( b , C ) A "Jode Label (yourselves,b ) )

118

English Rule 9 Head( a , A ) A Head ( a,B ) Λ Parallel ( Α , Β ) Λ Not (Erase ( Α,Β ))Λ Node Label ( Plural,a )ANot ( N o d e Label (we,a)v Node Label (you,a))->( a C ) ( Replace ( G,B ) Λ Head ( b , C ) A Node Label (themselves,b ) ) English Rule 10 Gen arc (A)—} Not ( 3 B ) ( 2 arc

(B ) A Successor ( B , A ) )

English Rule 11 Parallel( A,B )A Successor ( A , B )—» Not ( 30 ) (Replace ( C , B ) ) Of these rules, English Rule 11 is crucial in explaining the discrepancy between the widespread usage of reflexivization in Polish and the comparative paucity of reflexives in English. The following rules are suggested for Polish: Polish Rule 1 Parallel ( Α , Β ) Λ 1 arc ( A ) Λ 2 arc ( B )ANot (Erase (A,B))-»(3G) (Replace (C,B) A Head ( b , C ) A N o d e Label ( sie^b ) ν Node Label ( siebie,b)) Polish Rule 2 Parallel ( A , B ) A 1 arc ( A) A 3 arc (Β ) Λ Not (Erase ( Α , Β ) ) — ^ ( 3 G) (Replace ( C , B ) A Head ( b,C ) Λ Node Label ( sobie,b)) Polish Rule 3 Parallel ( A , B ) A 1 arc ( A ) A I n s t arc ( B )ANot (Erase ( A , B ) ) — - » ( G) (Replace( C , B ) A H e a d ( b , C ) A Node Label (soba,b ) ) Polish Rule 4 Parallel ( A , B ) A 1 arc (A) Λ 2 arc ( Β )Λ Foreign successor (B,C) Λ Gen arc (C )—* (3 D) (Replace ( D,B ) A Head ( b, L ) A Not (Node Label (siebie,b ))) Polish Rule 5 Unaccusative ( A , c k ( a))A'Tail (a,A )AV-stemmed ( a , b ) A (3Q )(R-graph (Q )AlIode Extractable ( d,Q ) Λ V-stemmed ( d,b)ATail ( d , E ) A Accusative ( E,c 1 ( d )) )-^( 30 )(Replace ( C , A )

119 Polish Rille 6 Parallel ( A , B ) ASuccessor ( Α , Β ) Α . ( 3 θ ) (Replace ( σ, Β) Λ Head ( a , C ) ) -*Not (Node Label (siebie,a)) Polish Rule 7 Unaccusative (A,c k (a )) ATail (a, A ) AV-stemmed ( a,b ) Λ [pojawiaol b £ •crozgla.dao? _ =, (3C) (Replace ( G , A ) ) julatniao I •

Polish Rule 8 Unaccusative ( A, c k ( a) ) ATail (a, A) KV-stemraed ( a,b) Λ ipo j awiac! ID if /rozpadac/* A Not ( 3Q ) (R-graph CQ) ANode Extractable (d,Q") lulatniac «· j • /sV-stemmed ( d , b ) ATail ( d , E ) Λ Accusative (E,c, (d ) } ) -? Not ( 3C ) C Replace ( C, A) ) polish Rule 9 1 arc (A,c..(a)) 1

i(5

AV-steiomed (a,b) ^b6: jpchac

Λ

Not ( 3B) ( 2 arc ( Β , Ο ^ Α ) ) A(3C) (Cho arc successor ( C,B) ) ) ( 3 D ) ( 2 arc ( D ) ASuccessor ( ϋ , Α ) ) Polish Rule 10 r ~\ 2 arc ( A,c..(a)) Λ V-steramed (a,b) Λ b 6 < przygl^dac/ ( S B ) ( 3 a r c ( B , c 2 ( a ) ) ASuccessor ( B , A ) ) ^ | J

--*

Polish Rule 11 1 arc( A,c k (a)) Λ Head Label ( 0 , A ) ASponsor ( Α , Β ) A&host ( Β) Λ. 2 arc( B,c k + 1 (a)) Λ Pare ( 0 , c k ( a ) > ANot (Head Label ( past,c )) ( 3D) ( Replace ( D,B ) ) Polish Rule 12 ck-0verrun ( Α,Β ) A N o t ( Ghost (A) v ( g C) ((Replace ( A , C ) A G r h o s t Λ Not (3D) (c k (D ) Λ Employed (D) ALocal Assassinate ( D , B ) ) ( 3E) CC^o arc ( Ε ) Λ (c ) (E ) Λ Successor ( Ε, Β ) )

120

Polish Rule 13

Ergative( A,Cj_(a ) ) Λ Not (t SB ) ( 3 arc ( ~ B , c k ( a ) ) ) ) AV-stemmed (a,b) Λ

przypomniec I podobac V —^ ( gC )(3 arc ( C , c R + 1 ( a ) ) A Lsnic , J Successor ( G , A ) )

A 3 arc * · ( B , c k + 1 ( a ) ) A Successor (Ώ,Α) —^ ( 3C)(2 arc ( G , c k + 1 ( a ) ) A Shost ( 0 ) A Sponsor ( A , C ) ) Polish Rule 15

Ghost (A,c, (0,E))

1 (a)

)Λ Sponsor ( B,A )A1 arc ( B , c k ( a ) ) A 3 arc Successor

^ ( a D ) ( Replace ( D,A ))

The system of English rules, elaborated here, should be further supplemented by rules dealing with constructions with nonparallel arcs (i.e. all cases where, in TG terms, the controller of the reflexive pronoun is in a different clause or even a totally different sentence ). To the system of Polish rules should be added rules dealing wibh pairs of arcs where one of them is an oblique R-sign arc. Finally, it is necessary to devise rules for· reciprocal constructions in both languages.

121

REFERENCES

Babby,L.H. 1975. A transformational analysis of transitive -sja verbs in Russian. Lingua 35, 3-4.297-332. Boas,P. and E.Deloria, 1939· Dakota grammar Memoirs of the National Academy of Sciences, Vol.23 * Borkin,A. 1971. dereference and beheaded NP's. An unpublished paper presented at the winter meeting of the Linguistic Society of America in St. Luis, Missouri. Chomsky,N. 1973. Conditions on transformations. In A festschrift for Morris Halle, eds. S.R.Anderson and P.Kiparsky. New York, Holt, Rinehart and Winston. Comrie,G. 1977. In defense of spontaneous demotion: the impersonal passive. In Syntax and semantics 8, eds. P.Cole and J.Sadock. New York, Academic Press. Cranmer,D.J. 1976. Derived intransitivity a contrastive analysis of certain reflexive verbs in German, Russian and English. Tttgingen, Max Niemeyer Verlag. Curme,G. 1952. A grammar of the german language. Mew York, Frederick Ungar Publishing Company. Duranti,A. and E.R.Byarushengo. 1977. On the notion "Direct Object". In Haya grammatical structure, eds. E.Byarushengo, A.Duranti and L.Hyman. Dept. of Linguistics, University of Southern California, Los Angeles, California. Dougherty,R. 1974. The Syntax and semantics of each other constructions. Foundations of Language 12. 1-47. Fillmore,C. 1968. The case for case. In Universals in linguistic theory, eds. E.Bach and R.Harms. New York. Fisiak,J.,Lipinska-Grzegorek f M. and Zabrocki,T. 1978. An introductory English-Polish contrastive grammar. Warszawa, PWN.

122

Gaertner,H. 1933. G-ramatyka wspolczesnego je.zyka polskiego. tfarszawa, Ksia_znica - Atlas. Gary,J. and E.L.Keenan. 19.77· On collapsing grammatical relations in universal grammar. In Syntax and semantics 8, eds;' P.Cole and J.Sadock. New York, Academic Press. Green,G.M. 1976. Governed-rule change and universal grammar. In eds. Relational Grammar. Studies in the linguistic sciences volume 6, number 1, Morgan, Greene and Cole, University of Illinois, Urbana. 152-169. Horn,G.M. 1977. An analysis of certain reflexive verbs and its implications for the organization of the lexicon. Studia Anglica Posnaniensia 9. 17-42. Johnson,D.E. and P.M.Postal. 1980. Arc Pair Grammar. Princeton, Princeton University Press. Kali32,R. 1983. On the so called "beheaded noun phrases" in English and Polish. Papers and Studies in Contrastive Linguistics 16. 43-51. Kanski,Z. 1984. Arbitrary reference and reflexivity: a generative study of the Polish pronoun sie, and its English equivalents. University of Silesia dissertation, Kato\/ice. Kardela,H. 1979· Semantic and syntactic properties of reflexive constructions in Polish and English. University of Maria Ourie-Skiodowska dissertation, Lublin. Kimeryi,A. 1976. A relational grammar of Kinyarwanda. Unpublished doctoral dissertation, University of the City of Los Angeles. KrzeszowskijTeP« 1984. Tertium comparationis. In Contrastive linguistics. Prospect and problems, ed. J.Fisiak. Berlin, Hew York, Amsterdam, Mouton Publishers, 301-312. - 1985. Towards a typology of contrastive studies. To appear in Pragmatics and beyond, ed. V/.Oleksy. Kubinski,W. 1981. Impersonal passives: can Polish even up the score with Dutch? To be published in Linguistica et Anglica G-edanensia. - 1982. Polish sie, constructions and their English counterparts. Papers and Studies in Contrastive Linguistics 15. 55-65. Lakoff,G-o 1977. Linguistic Gestalts. Chicago Linguistic Society 13. 236-287.

123

Napoli,D.J. 1976. The two si's of Italian. Reproduced by the University of Indiana Linguistic Club, Bloomington. Neubauer,P. 1979. She Score on "Impersonal Passive": Dutch-1, Polish-0. An unpublished paper presented at the 17 International . Conference on Polish-English Contrastive Linguistics in Boazkowo. Niedzielski,H. 1976. English and Polish pseudo-reflexives. Papers and Studies in Contrastive Linguistics 4. 167-198; Perlmutter,D.M. 1978. Impersonal Passives and the unaccusative hypothesis. Berkeley Linguistic Society 4. 157-189. Perlmutter,D.M. and P.Postal. 1983. Some Proposed Laws of Basic Clause Structure. In Relational Grammar 1, ed. D.M.Perlmutter. Chicago University Press. 81-128. - 1984. The 1 Advancement Exclusiveness Law. In Relational Grammar 2, eds. D.M.Perlmutter and C.A.Rosen. Chicago University Press. 81-125. Postal,?. 1979. Some Arc Pair Grammar descriptions. An unpublished paper. Pullum,G. 1975. Comitatives via conjunct ascensions, and a possible case of initial-term agreement. In Squibs on Relational Grammar Nos. 1-6. Unpublished papers, University College, London. Ruszkiewicz,P. 1984. Aspects of reflexivization in English. University of Silesia dissertation, Katowice. Sadock,J.M. 1975. The soft interpretive underbelly of Generative Semantics. In Syntax and Semantics 3, eds. p.Cole and J.L.Morgan. New York, Academic Press. Saloni,Z. 1976. Cechy skladniowe polskiego czasownika. Wroclaw, Ossolineurn. Sridhar,S.N. 1976. Dative subjects, rule government and Relational Grammar. In Relational Grammar. Studies in the linguistic sciences volume 6 no. 1, eds. Morgan, Greene and Cole. University of Illinois, Urbana. 130-151· Szober,S. 1953. Gramatyka j^zyka polskiego. Warszawa, PWN. Tokarski,J. 1982. Elements of Polish grammar. In Wielki Slownik polsko-angielski, J.Stanislawski. Warszawa, PWN. Wachtel,T. 1979. The demotion analysis of initially unaccusative impersonal passives. Chicago Linguistic Society 15. 321-330.

124

Wilczewska,Ko 1966.. Czasowniki zwrotne we wspolczesnej polszczyznie. Torun, lowarzystwo Naukowe w Toruniu. Woiczyn3ka-Sudol,A. 1977. Notional passive in English and Polish. Papers and Studies in Contrastive Linguistics 5. 153-164. Zandvoort,R.W» 1962.: A handbook of English grammar.· Bristol, Western Publishing Services. Zwicky,A. 1977» On clitics. Indiana University Linguistics Club.

125 INDEX

absolutive,47,48,49 accusative,46,48,49 1 Advancee Preservation Law, 98-99,101,116 1 Advancement Exclusiveness Law,96-99,115 antipassives,49,86-92,110,114 arc, 3 beheading,73-76 Gho Arc Cosponsor Law,54-55 Cho Arc No Local Successor Law,54,56 Cho Arc Second Sponsor Identity Law,54,56 chomeur,9,56 Cltfmeur Law,54,56,60,100-101,116 clitics,87,114 Con Arc Endpoint Label Identity Law,108-109 contrastive analysis,12-13 coordinates,3 coordination,108-109 COPY analysis,27-29 Demotion No Replacer Law,48,49, 60 Dummy Arc Law,54,56 dummy nominals,54-55,63-64,95, 96,98,104-108 Earliest Strata Uniqueness Law, 47,49,100 emphatic reflexives,71-72,113 ergative,46,48 Erase,4-5 evanescent,47,49 extraposition, 7-8 facsimile,55 Fall Through Law,51,53 Final 1 Arc Law,47,49

flagging,8 foreign successor,56 Genetic PNs,12,57-58 Ghost Arc Law,54,55,63,65,94 ghost arcs,54-55,63,100-102,106-108 Ghost Coordinate Law,53-54,54-55 Ghost Organicity Law,54,56 Ghost Status Theorem,54,55 Graft Overlap Law,50,53 grafts,53 grammatical category nodes,3 immigrant,53 Immigrant Local Sponsor Law,50,53 impersonal constructions,17,18,21, 28-29,31-33,36-40,64-65,83, 92-103,110,115 impersonal passives,55,92-103 inchoatives,17,18,20,30-31,76,81 Inexplicit nodes,56-57,89-91 inversion,21,63,84,103-108,110,114, 115 Internal Survivor Law,51,53,58 left middles,36 L-graph,5-7,H Local Successor Distinct R-sign Law,50,52 logical nodes,2 Maximal Two Sponsor Law,50,53 middles,20,21,27,30-31, ^-33,7b-8b,

1 ιυ, π4 Motivated ChSmage Law,101,116

narrowing reflexive verbs,19 Nominal Arc Successor Law,111,112, 116 non-referential NP,35-36 nonterminal nodes,2 No Oblique Successors Law,47-48,49, 84

126

No Oblique Successors Law,47-48, 49,8? notional passives,17,20,23-24,82 No Vacuous Pall-Through Law,50, 53 nuclear term arcs,56 Nuclear Term Arc Stratal Continuity Law,47,49 Obligatory Complement Constraint, 36,37 oblique,47,71,120 organic arc,56 overrun,56 outrank,48 Pair Network,6-7 parallel arcs,57,66-67,1 H part reflexives,21,72-76,110 passives,9,59,60,94,95,96 PN laws,10-11,47-49,50-57 PN Law 1 Re placer Erase Law, 50,51,58 PN Law 2 Successor Erase Law, 50,52,94 PN Law 3 Unique Eraser Law,11, 50,52 PN Law 4 Self-Sponsor Law, 50,52 PN Law 5 Local Successor Distinct R-sign Law,50,52 PN Law 6 Sponsor Independence Law, 50,53 PN Law 7 Maximal Two Sponsor Law, 50,53 PN Law 12 Immigrant Local Sponsor Law,50,53 PN Law 13 Graft Overlap Law,50,53 PN Law 21 No vacuous Pall-Through Law,50,53 PN Law 23 Replacer Coordinate Law, 53,54,58 PN Law 24 Ghost Coordinate Law, 53-54,54-55 PN Law 26 Pall Through Law,51,53 PN Law 34 Con Arc Endpoint Label Identity Law,108-109 PN Law 44 Final 1 Arc law,47,49 PN law 45 Unaccusative Instability Law,47,49,61 PN Law 46 Unaccusative Law,47,49, 61 PN Law 47 Nuclear Term Arc Stratal Continuity Law,47,49 PN Law 48 Earliest Strata Uniqueness Law,47,49,100

PN Law 49 No Oblique Successors Law,47-48,49,84 PN Law 50 Demotion No Replacer Law,48,49,60 PN Law 51 Nominal Arc Successor Law,111,112,116 PN Law 60 Cho Arc Cosponsor Law, 54-55 PN Law 61 Cho Arc Second Sponsor Identity Law,54,56 PN Law 62 ChQmeur Law,54,56,60, 100-101,116 PN Law 63 Cho Arc No Local Successor Law,54,56 PN Law 79 UN Node Headed Arc Limitation Law,98 PN Law 84 Ghost Arc Law,54,55, 63,65,94 PN Law 85 Stable Ghost Arc Sponsor Law,54,55 PN Law 86 Ghost Sponsor Organicity Law,54,56 PN Law 87 Dummy Arc Law,54,56 PN Law 94 Internal Survivor Law, 51,53,58 predecessor,52 Primitive Pair Network,5 pseudo-reflexives,22 psych movement verbs,17,18,83-86 quantification,109-110 reciprocals,17,18,108-113,120 reflexive passives,14,60,65-66, 82-83,96 relational signs,3 replacer,51-52 Replacer Coordinate Law,53,54,58 Replacer Erase Law,50,51,58 R-graph,3-4 right middles,36,88-89 rules,12 Self Sponsor Law,50,52 S-graph,57,H Sponsor,4-5 Sponsor Independence Law,50,53 stable ghost arc,55 Stable Ghost Arc Sponsor Law,54, 55 stratal uniqueness,92,96,99-102, 106-107,115 Stratal Uniqueness Theorem,47, 49,99 successor,52

127

Successor Erase Law,50,52,94 tertium comparationis,12-14,58,67 total reflexives,21,73,110 true reflexives,17,18,19,20 unaccusative,43-44,47,48,49,60,62 77-83,86,96-103 unaccusative hypothesis,43-44 Unaccusative Instability Law,47, 49,61

Unaccuaative Law,47,49,61 unergative,43-44,46,48,97 Unique Eraser Law,11,50,52 UN Node Headed Arc Limitation Law,98 unstable ghost arc,55 word formation,21-22,92,115