Hokkaido Institute of Pharmacy

1

Transcript of Hokkaido Institute of Pharmacy

Page 1: Hokkaido Institute of Pharmacy

ON EMPHATIC REFLEXIVES

KAORU FUKUDA

Hokkaido Institute of Pharmacy

This paper investigates the distributional property and the emphaticfunction of English reflexive pronouns in the non-reflexive use. It isshown that although reflexives in the emphatic use are lexical anaphors justlike normal reflexives, they are not theta-marked nor assigned Case, that is,they behave as adjuncts syntactically unlike normal reflexives. The distri-bution of emphatic reflexives is generalized by the statement that they m-command and are m-commanded by their antecedents at S-structure. It isargued that the anaphorhood and the adjuncthood of emphatic reflexivesplay a central role in determining their distribution. A percolation opera-tion is proposed to account for a distinction between appositive and floatedemphatic reflexives as to their emphatic function.*

0. INTRODUCTION. Reflexive pronouns in English may be used em-

phatically as well as reflexively.1 Reflexives in the former use are usuallycalled emphatic reflexives (ERs). We will refer to ERs which immediate-ly follow their antecedents as appositive ERs, and those which are notadjacent to their antecedents as floated ERs.2 1 illustrates typical

* This paper is a revised version of the draft I read at the 23rd monthly meeting of

the Sapporo Linguistics Circle on December 19, 1988. I would like to thank KatsuhideSonoda, Kimihiro Ohno, Yoshihiro Yamada, and an anonymous EL reviewer for theirvaluable comments and suggestions. Special thanks go to Joe Luckett for his labori-ous assistance both in linguistics and otherwise. Needless to say, all remaining inade-quacies and errors are my own.1 Ruszkiewicz 1984 notes a third use of reflexives illustrated in (i).

(i) a. As for himself, the president means to stay out of politics this year.b. I told Albert that phisicists like himself were a godsend.

These reflexives can be distinguished from those in the text in that they may occurwith their antecedent missing in the same sentence as in (ii).

(ii) A man like himself can never play the saxophone because he has onlyone lung. (Ruszkiewicz (1984: 122))

2 The terms 'appositive' and 'floated' are used only for expository purposes here. I

do not think that appositive ERs are genuinely appositives nor that floated ERs arefloated by a movement rule. See note 3.

English Linguistics 6 (1989) 36-51(C) 1989 by the English Linguistic Society of Japan

-36-

Page 2: Hokkaido Institute of Pharmacy

ON EMPHATIC REFLEXIVES 37

instances of appositive and floated ERs.

(1) a. He himself did it.b. We sent John himself.c. He did it himself.d. He had never himself had the measles.e. I hope of course that I am myself not mischievous.

Though the anaphoric relation involving reflexives has been discussedextensively as one of the central issues in generative grammar, it seemsthat ERs have been investigated only slightly. The aim of this paper is tooutline the properties characteristic of ERs through a comparison withreflexives in the reflexive use, secondary predicates, and floated quantifi-ers. Section 1 presents a preliminary characterization of ERs in terms ofthe principles of UG. Section 2 generalizes the distribution of ERs andsection 3 discusses why they exhibit such distribution with reference toLF movement of anaphors. Section 4 considers the possibility of ana-lyzing ERs as predicates, and the section also touches on the emphaticfunction of ERs.

1. BASIC CHARACTERIZATION OF ERs1.1. Let us begin with the lexical properties of ERs. We assume that

reflexives in the reflexive use and those in the emphatic use are the samelexical items. Specifically, ERs are base-generated in place as NPs.3,4

3 Base-generation of ERs implies that we do not postulate a movement rule whichshifts the ER to derive (ib) from (ia).

(i) a. John himself is a spy.b. John is a spy himself.

The transformational approach will face a sourceless problem in deriving (iib) from(iia).

(ii) a. *Someone himself is a spy.b. Someone is a spy himself.

4 Based on the morphological identity to normal reflexives, we assume that the lexi-cal category of ERs is an NP. Consequently, we do not analyze an ER as, say, anadverb.

There is empirical evidence against the adverbial analysis. ERs do not pattern withany major class of adverbs classified by Jackendoff 1972 and Roberts 1988. First, ERsmay occur in the sentence-final position, from which merely-class of adverbs are ex-cluded. Second, they differ from sentence-modifying adverbs in that they appear inthe sentence-final position without comma intonation. Third, ERs can co-occur withstate-denoting predicates unlike manner adverbs.

(i) a. John likes beer himself.b. *John likes beer merely.

Page 3: Hokkaido Institute of Pharmacy

38 ENGLISH LINGUISTICS, VOLUME 6 (1989)

They carry a specification [+anaphor, -pronominal] as lexical anaphors,because they lack their own reference and require an antecedent to whichthey are referentially dependent. It is unnecessary, in this discussion, tospecify these lexical properties of ERs separately from reflexives in thenormal use.

1.2. As for syntax, ERs should be distinguished from normal reflex-ives; only the former may be removed from a sentence without affectingthe grammaticality of the sentence.

(2) a. John likes *(himself).b. John likes beer (himself).

The reflexive in 2a is the object of the verb and it bears a theta-role. Bythe Projection Principle the reflexive must be present categorically atevery syntactic representation to project the lexical structure of the verb.Moreover, the reflexive must be present as a recipient of the theta-role tosatisfy the theta-criterion. On the other hand, the ER in 2b is not theta-marked by the verb. Otherwise the object NP would violate the theta-criterion, because the verb takes only one internal argument. The ER isthen an adjunct whose presence is not guaranteed by the Projection Prin-ciple nor by the theta-criterion.

The verb like assigns only one Case, which is assigned to its object NPin 2b. Then the ER is Caseless.5 According to the Case Filter formu-lated in Chomsky (1981: 49), an NP must have Case if it has phoneticcontent. The Case Filter, as it stands, would wrongly exclude 2b as

ungrammatical. Chomsky (1986a: 97) proposes that the Case Filter be

c. John likes beer, probably.d. *John likes beer quickly.

Further, subject-oriented adverbs require agentive subjects but ERs are not subject tothis restriction.

(ii) a. John was deliberately polite/*asleep.b. John was himself polite/asleep.

5 If Case were assigned not to the object but to the ER, then the object NP would be

Caseless in violation of the Case Filter. But in fact, the adjacency condition blocksthe assignment of Case to the ER.

It seems that appositive ERs are also Caseless. It is impossible for the ER in (i) tobe assigned an inherent Case from the head of the antecedent since the head does nottheta-mark the ER. The subject NP John himself is assigned Nominative Case, whichshould be percolated to the head NP John, but not to the ER, for the reasonmentioned above.

(i) John himself is guilty.

Page 4: Hokkaido Institute of Pharmacy

ON EMPHATIC REFLEXIVES 39

superceded by the theta-criterion 3, which incorporates the visibility con-

dition for theta-marking.6

(3) Each argument α appears in a chain containing a unique

visible position P, and each theta-position P is visible in a

chain containing a unique argument α.

The revised theta-criterion requires only argumental NPs to be visible fortheta-marking. Since non-arguments are irrelevant to theta-marking,the ER in 2b does not violate this principle even if it is a Caseless NP.

We have noted two essential properties of ERs. First, ERs are lexicalanaphors and require an antecedent, just like reflexives in the reflexiveuse. Second, ERs behave as adjuncts syntactically, differing fromnormal reflexives. These two characteristics of ERs will be crucial whenwe discuss their distribution.

2. THE DISTRIBUTION OF ERs2.1. The structural notions relevant to our discussion of the distribu-

tion of ERs are the following, which we adopt from Chomsky (1986b: 7-8).

(4) α is dominated by β only if it is dominated by every segment

of β.

(5) α m-commands β iff α does not dominate β and every maxi-

mal projection γ that dominates α dominates β.

Let us determine the constituent structure of the constructions con-taining an ER by means of the following operations.

(6) Pseudo-cleftinga. What John did was finish the job himself.b. What John did himself was finish the job.

(7) VP Frontinga. John wanted to finish the job, and finish the job himself

he did.b.? John wanted to finish the job, and finish the job he did

himself.

(8) Though-Movementa. Like beer himself though John does,...

b.? Like beer though John does himself,...

6 An element is visible only if it is assigned Case. (Chomsky (1986a: 94))

Page 5: Hokkaido Institute of Pharmacy

40 ENGLISH LINGUISTICS, VOLUME 6 (1989)

(9) VP Deletiona. John likes beer himself, and Mary does, too.b.? John likes beer himself, and Mary does herself, too.

(10) Cleftinga. It is John himself that is a spy.b.* It is John that himself is a spy.

(11) NP-Movementa. [Sue herself]i is believed ei to be a spy.b. *[Sue]i is believed ei herself to be a spy.

Pseudo-clefting, VP Fronting, Though-Movement, and VP Deletion serveas VP tests in that they affect only VP constituents. Clefting and NP-Movement serve as NP constituency tests. These tests indicate that sen-tences containing an ER have an adjunction structure.7

(12) a. b.

In 12, NPj is not dominated by VP nor by NPi, so that NPi m-com-

mands NPj, and vice versa, given the definitions of dominance and m-

command above. Now the distribution of ERs may be generalized des-

criptively as follows: an ER has a mutual m-command relation to its

antecedent at S-structure.8

7 Adjunction structures such as 12 can be base-generated by the X-bar schema (i),

which is independently necessary to generate small clauses and modifiers like manneradverbs.

(i) X"→X"*X"X"*

(ii) a. John considers [AP Mary [AP stupid]]b. John [VP [VP speaks Japanese] fluently]

8 As an anonymous EL reviewer has pointed out, it seems that the mutual m-com-

mand relation does not hold in (6-8)a. Though the precise nature of the operations inquestion is unclear to me, it is known that they yield outputs which are equally prob-lematic for other relations. For example, the coreference interpretation between theitalicized NPs in (i) is impossible though it is configurationally expected.

(i) a. *They said Mary was happy that she was going to Paris and happy thatMary is going to Paris she is.

b. *Happy with the man Mary married though she is...(a, b from Gueron (1984: 160))

One possibility of avoiding the disturbing effect of these operations is to assume thatthey are rules of the PF side and do not interact with rules of the LF side such as thoseinterpreting the reference of pronouns and reflexives.

Page 6: Hokkaido Institute of Pharmacy

ON EMPHATIC REFLEXIVES 41

2.2. In addition to appositive and sentence-final positions, ERs mayoccur fairly freely between a modal and aspectual verbs.

(13) a. John must have been being beaten.b. John must himself have been being beaten.c. John must have himself been being beaten.d.? John must have been himself being beaten.e.?? John must have been being himself beaten.

It is generally assumed that aspectual verbs take a VP complement, andwe assume that passive be also takes a VP complement. The structure of13a, on these assumptions, will be represented roughly as in 14.

(14)

The VP*-adjoined positions in 14 provide the loci for the ERs in 13b-e.Note that the ER is not dominated by VP because it is not dominated byevery segment of VP. Then the ER adjoined to any VP* stands in amutual m-command relation to the antecedent, in accordance with ourearlier generalization.9

9 ERs may occur between the main verb be and a predicate nominal, though it

seems that they do not m-command their antecedents.(i) a. John is himself a fool.

b. John must be himself a fool.c. John has been himself a fool.d.? John is being himself a fool.

It may be possible to retain the generalization in the text. The ER in (ia) isadjoined to VP, out of which the main verb be is moved to INFL to become inflected.As for the ERs in (ib-c), I assume following Akmajian, Steel and Wasow 1979 that therule of Be Shift has applied to move the main verb be out of the innermost VP seg-ment, to which the ERs are adjoined.

(ii) a. John bei-I [VP himself [VP ei a fool]]b. John must-I [VP be [VP himself [VP a fool]]]

c. John havei-I [VP ei [VP been [VP himself [VP a fool]]]]

Page 7: Hokkaido Institute of Pharmacy

42 ENGLISH LINGUISTICS, VOLUME 6 (1989)

Our generalization predicts that ungrammatical sentences will resultif there is at least one maximal projection which dominates either anER or its antecedent but which does not dominate the other. In 15 thecircled NPs prevent the potential antecedent from m-commanding theER.

(15) a.* [(NP)[NP John's] mother] was angry at Mary himself.b.* [(NP)[NP John] and Mary] finished the job himself.

c. [(NPi) Friends of [NPj those girls]] are teachers themselvesi, *j.

Failure of an antecedent to m-command an ER gives rise to a subject-object asymmetry. As we saw above, an antecedent in the subject posi-tion m-commands an ER adjoined to VP. But an antecedent in theobject position, which is dominated by VP, does not m-command an ERadjoined to VP. Thus the subject-object asymmetry in 16 is as expectedunder our definitions.

(16) a. The manager [VP [VP hit Mary in the office] himself].b.* Mary [VP [VP hit the manager in the office] himself].

The reverse case, where an ER fails to m-command its antecedent, isillustrated by the following examples from Culicover and Wilkins

(1984: 28).(17) a.* John put the book himself on the table.

b.* Mary persuaded Bill herself to leave.c.* Mary promised Bill herself to leave.

Since the ERs in 17 occur between the complements of the verbs, theycannot occupy the position adjoined to the whole VP. They may insteadoriginate in the position adjoined to either of the complements. Forexample, the structure of 17a may be either 18a or 18b.

(18) a.

Because the main verb be (id) is not restructured due to the presence of the progres-sive be, the sentence seems to be less acceptable than (ib-c), though the judgment isnot so clear.

Page 8: Hokkaido Institute of Pharmacy

ON EMPHATIC REFLEXIVES 43

b.

The ER in either position is dominated by VP. As a result it does notm-command the antecedent in the subject position.

Observe the contrast between 17 and 19.

(19) a. John baked the cake himself for the party.b. John wrote the letter himself to Mary.

(Culicover and Wilkins (1984: 28))The PPs in 19 are not the complements of the verbs. A VP Deletion testshows that they are VP-adjoined elements.

(20) a. Who baked the cake for whom? John did for the chil-dren.

b. Who wrote the letter to whom? John did to Mary.

(Culicover and Wilkins (1984: 27))Now 19a, for example, has the structure illustrated in 21, where the ERoccupies the position adjoined to VP and m-commands the subjectantecedent.

(21)

The definitions of dominance and m-command adopted here enable

us to generalize the distribution of ERs as: an ER m-commands its ante-

cedent, and vice versa.

3. ERS AND BINDING THEORY

3.1. Now we turn to the question of why ERs have the distribution

they do. Principle A of binding theory is considered to determine the

distribution of anaphors. As far as ERs are concerned, it follows from

this principle that an antecedent should m-command an ER to bind the

Page 9: Hokkaido Institute of Pharmacy

44 ENGLISH LINGUISTICS, VOLUME 6 (1989)

latter.10 But the fact that an ER must m-command an antecedent re-mains unexplained.

3.2. Chomsky 1986a proposes that binding theory, at least Principle Aand B, applies after anaphors are adjoined to INFL or VP at LF.11 LFmovement of anaphors has many consequences on binding theory.Relevant to our discussion here are the antecedent-anaphor and the ana-

phor-trace relations after LF movement. As Chomsky (1986a: 175)notes, the antecedent-anaphor relation at LF is a government relation inthe sense that they m-command each other. We assume, followingChomsky, that the relation serves as a condition; that is, at LF a mutualm-command relation must hold between an antecedent and an ana-

phor.12 As for the anaphor-trace relation, the traces of anaphors must beproperly governed to satisfy the Empty Category Principle (ECP).13 Let us assume that ERs are also subject to LF movement. Thus the

ER in 22a is adjoined to VP at LF to yield 22b.

(22) a. John [VP [VP finished the job] himself]b. John [VP himselfi [VP [VP finished the job] ei]]

10 The notions necessary to formulate Principle A are defined in Chomsky 1986a as

follows.

(i) a. α binds β if α c-commands and is coindexed with β.

b. A local domain for an anaphor α is the minimal governing category of

α.

c. A governing category of α is a maximal projection containing α, a sub-

ject, and a lexical governor of α.

d. α governs a maximal projection if α and X" c-command each other.

Note that Chomsky's notion of c-command is equivalent to our notion of m-com-mand. Given these definitions, the VPs in (iia-b) count as the lexical governors ofthe ERs because they m-command each other, and the IPs count as the local domainsof the ERs.

(ii) a. [IP John will [VP himself [VP finish the job]]]b.* John believes that [IP Mary will [VP himself [VP finish the job]]]

11 Following Chomsky 1986b and Pica 1987, we assume that maximal projection

anaphors are adjoined to a maximal projection, obeying the condition on adjunc-tion. Specifically, we assume that the adjunction site for ERs is VP instead of INFLif they are subject to LF movement.

12 Chomsky explains the subject-orientation in long-distance binding such as (i) by

the government condition.

(i) They told us (that) [picture of each other] would be on sale.After LF movement the anaphor is governed by the subject, but not by the object.

13 By ECP, a non -pronominal empty category must be properly governed, that is, it

must be theta-governed or antecedent-governed. Since ERs do not originate in theta-

positions, the traces of ERs must be antecedent-governed to satisfy ECP.

Page 10: Hokkaido Institute of Pharmacy

ON EMPHATIC REFLEXIVES 45

The trace of the ER, ei, is properly governed, and the antecedent andthe moved ER satisfy the mutual m-command condition. But note thatthe mutual m-command relation holds before LF movement of the ER.The fact that LF movement does not affect the antecedent-anaphorrelation in typical cases like 22 seems to cast doubts on the necessity ofthe movement.

Let us examine the motivation of LF movement of anaphors. Pica1987 presents an argument for LF movement of argumental anaphors toadjunct positions. His argument goes as follows: for an NP to denote,the specifier of the NP closes the open position associated with the headN. The specifier of reflexives lacks this ability, like possessive pronounsexpressing inalienable possession. An NP whose head remains open isunsaturated and deficient as an argument. Consequently, when reflex-ives occupy an argument position, they must be adjoined to VP or INFLat LF to suppress the argumenthood. The open position of the head selfis now closed by being bound by an antecedent. Thus the deficient argu-menthood motivates LF movement of argumental anaphors. His argu-ment, however, does not hold of adjunct anaphors. It makes no sense tosuppress the argumenthood of adjunct anaphors.

LF movement of ERs will lead to undesirable results in some cases.We saw in section 2.3. that the ER in 23a may occupy the positionadjoined to either complement. Note that the ER does not m-commandthe antecedent at S-structure.

(23) a.* John put the book himself on the table.b. John [VP himselfi [VP Put [NP [NP the book] ei] on the table]]c. John [VP himselfi [VP put the book [PP ei [PP on the table]]]]

If the ER were adjoined to VP at LF, the antecedent and the ER wouldcome to m-command mutually, meeting the condition on the antece-dent-anaphor relation. Then under LF movement of ERs, 23a would be

grammatical, contrary to the fact.Now we abandon LF movement of ERs and assume alternatively that

ERs are not subject to LF movement due to their adjunct status. But weretain the assumption that the open position of the head of reflexivesmust be closed by being bound by an antecedent at LF under the mutualm-command condition. Now that LF movement is not available, unlikeargumental reflexives, ERs meet the condition at LF only when they havea mutual m-command relation to their antecedents at S-structure. Thusthe two characteristics of ERs, anaphorhood and adjuncthood, combineto determine their unique distribution we observed in the previous section.

Page 11: Hokkaido Institute of Pharmacy

46 ENGLISH LINGUISTICS, VOLUME 6 (1989)

4. PREDICATION AND FOCUS PERCOLATION

4.1. Assuming that ERs do not undergo LF movement, we arguedthat their distribution can be viewed as the reflection of the condition

governing the antecedent-anaphor relation at the level of LF. Anotherpossible way to derive the distribution is to assume that ERs are co-indexed with their antecedents by a predication rule at a relevant level.Rothstein 1985 proposes that every non-theta-marked maximal projec-tion is linked to a subject under the mutual m-command condition at S-structure. Since ERs are NPs not theta-marked by any lexical heads, itseems plausible to suppose that they are predicated of their antecedentsas subject.14

If ERs are to be analyzed as predicates, they qualify as secondary predi-cates, whose defining property is stated by Rothstein (1985: 89) as in 24.

(24) X is a secondary predicate of Y if and only if Y is an NP theta-marked by a lexical head other than X.

It follows from 24 that a secondary predicate cannot be linked to pleo-nastics, which are not theta-marked by anything. Observe the exam-

ples in 25 and 26, where the restriction is found not only between pleo-

14 It seems that ERs require their antecedents to be unique. In this respect they

behave like locally bound reflexives and unlike non-locally bound ones.(i) a. *Johni told Maryj about the picture themselvesi+j.

b. *Johni told Maryj about themselvesi+j.c. Johni told Maryj that there were pictures of themselvesi+j inside.

(b, c from Lebeaux (1985: 346))It is well-known that a predicate must be linked to a unique subject. Lebeaux 1985argues that locally bound reflexives are predicated of their antecedents as subject andhe proposes an analysis which derives the uniqueness of antecedent from the unique-ness of subject. His proposal is too far-reaching to discuss here. Instead of goinginto his analysis in detail, I simply note here that according to his definition of localbinding (ii), ERs are locally bound and exhibit further local binding properties such asnon-free variation with pronouns, c-command requirement at S-structure, and onereading under VP Deletion.

(ii) α is locally bound iff it is bound within the minimal NP or S containing it

and its governor.(iii) a. Johni saw himselfi/*himi.

b. Johni saw the girl himselfi/*himi.

(iv) a. *John's wife respects himself.b. *John's wife likes beer himself.

(v) a. John hit himself, and Bill did, too. (=Bill hit Bill)b. John signed the letter himself, but I didn't. (=I didn't do it myself)

Page 12: Hokkaido Institute of Pharmacy

ON EMPHATIC REFLEXIVES 47

nastics and secondary predicates but also between pleonastics and ERs.

(25) a. *It rained heavy.b. *It snows freezing. (a, b from Rothstein (1985: 92))

(26) a. *It rained itself yesterday.b. *It itself rained yesterday.

We can account for the restriction in a uniform fashion by appealing to24,15 and this makes it more attractive to treat ERs by predication.

It should be noted, however, that the predication analysis raises someempirical difficulties at the same time. First, the fact that an appositiveER and its antecedent form a constituent seems to be inconsistent withthe condition on secondary predicates. Adjunct predicates, as Rothsteinnotes, do not form constituents with their subjects. Then appositiveERs at least pose a problem to the predication analysis.

Floated ERs are difficult to interpret as secondary predicates, thoughthey appear in a position which may be occupied by a secondary predi-cate. Safir (1987: 89-90) observes that the floated ERs in 27 are inter-

preted like manner adverbials but are not interpreted as secondarypredicates without being accompanied by again.

(27) a. Murry ate the cake himself.b.? John returned from the hospital himself* (again).

The fact that ERs exhibit floating phenomena may count as evidenceagainst the predication analysis, in conjunction with the fact that they donot appear in the post-copular position, a typical position for a predi-cate. Iwasawa 1988 observes that those quantifiers which float cannot beused predicatively while those which can be used predicatively do notfloat. 28 are some of his examples.

(28) a. The men have all/both/each/*many picked up a glass.b. His sins were many, and his friends were fewc. *The boys are all/both.

As 29 shows, the correlation between predicatehood and floating holdswith ERs as well.

(29) a. The men have themselves picked up a glass.b. *The boys are themselves.

It is true that there remain a few empirical problems in treating ERs as

15 May (1985: 101) accounts for the ungrammaticality of (i) by assuming that exple-

tives lack agreement features.(i) *It seems to itself that Philby is discreet.

Perhaps 26 may be ungrammatical for the same reason.

Page 13: Hokkaido Institute of Pharmacy

48 ENGLISH LINGUISTICS, VOLUME 6 (1989)

predicates. But it should be stressed that in spite of these difficulties, apredication analysis is worth pursuing for theoretical reasons. A suitabletheory of predication will explain the distribution of ERs, guarantee theanaphoric interpretation by coindexing ERs with their antecedents,16 andallow ERs to be licensed as predicates.

4.2. We have so far characterized an ER as an adjunct anaphor. Athird defining characteristic of an ER is the prominent stress it always car-ries. It is called emphatic for this reason.

It has been observed by several linguists that an appositive ER does not

go with indefinite NPs as an antecedent.(30) a. *With proper tools, one oneself can assemble a bicycle.

b. With proper tools, one can assemble a bicycle oneself.

(a, b from Partee (1979: 46))(31) a. *An engineer himself should know this.

b. An engineer should know this himself.

(32) a. *Who himself signed it?b. Who signed it himself?

(31, 32 from Moravscik (1972: 274))(33) a. *Someone/Each guy himself is a spy.

b. Someone/Each guy is a spy himself.The oddness of (30-33)a is usually attributed to the awkwardness of

giving emphasis to the indefinite NPs. But as (30-33)b show, floatedERs are mysteriously not subject to the restriction. The contrast sug-

gests that an appositive ER should be distinguished from a floated ERwith regard to their emphasizing function.

Let us assume tentatively that the maximal projection containing anER becomes prominent by the upward percolation of the prominence onthe ER.

(34) [+prominent]

[+prominent]

As a result of the percolation operation, an ER emphasizes a constituentto which it is adjoined. An appositive ER emphasizes an immediately

16 An anonymous EL reviewer has pointed out to me this virtue of assuming a pred-

ication analysis of ERs.

Page 14: Hokkaido Institute of Pharmacy

ON EMPHATIC REFLEXIVES 49

preceding NP, an antecedent. On the other hand, a floated ER isadjoined to VP and emphasizes it. The antecedents in the subject posi-tion in 30-33 are emphasized by the appositive ERs but not by the floatedERs. Therefore our analysis accounts for the fact that only appositiveERs are subject to the idefiniteness restriction.

That floated ERs serve to focus the VP to which they are adjoined canbe seen from the following examples.17

(35) a.* Mary [VP has herself bought only one dress].b.* It [VP is the boy that is himself homeless].c.* John will [VP himself buy tomorrow morning the course

textbook on generative grammar].The VPs of 35 contain a focused element as well as a floated ER, resultingin a violation of the prohibition on multiple foci in one constituent. The

prohibition seems to be operative in 36, where the appositive ERs are notcompatible with NPs focused by heavy stress or only.

(36) a. *[JOHN himself] is a spy.b. *[Only John himself] is a spy.

5. CONCLUSION. ERs can be characterized as anaphors lexically andas adjuncts syntactically. They should be distinguished from argumentalreflexives in that they are not theta-marked nor assigned Case. Theadjunct status of ERs has a theoretical implication that the visibility con-dition should be preferred over the Case Filter.

The adjunct anaphor status of ERs seems to determine their distribu-tion. Like argumental reflexives, ERs should satisfy the condition onthe antecedent-anaphor relation at LF. Argumental and non-argumentalreflexives seem to differ in the way they meet the condition. LF move-ment is available for the former, but it is not available for ERs due to theadjuncthood of the latter. It then follows that ERs should have a mutualm-command relation to their antecedents at S-structure to satisfy thecondition.

17 I am indebted to Iwasawa 1988 for this argument. Based on the fact that floated

quantifiers do not co-occur with another means of focusing, he argues that they havethe function of focusing the predicates following them.

(i) a. *The students have all passed only the oral exam.b. *It was the boy students that (have) all passed the exam.c. *The students all passed last semester the oral exam on syntax and

semantics.

(a-c from Iwasawa (1988: 84))

Page 15: Hokkaido Institute of Pharmacy

50 ENGLISH LINGUISTICS, VOLUME 6 (1989)

We have suggested the possibility of analyzing ERs as predicates, there-by deriving their distribution from the condition on predication. Wehave also noted that there are some similarities between ERs and those

quantifiers which float: neither of them is interpreted like secondary pre-dicates, and they both serve to emphasize the VP to which they areadjoined, probably percolating the prominence they carry. The percola-tion operation seems to give a desired distinction between an appositiveER and a floated one.

To capture ERs in the framework of generative grammar, I centered myattention on the most remarkable aspects of ERs, that is, their distribu-tion and emphatic function. The theoretical consequences of the propo-sals and suggestions in this paper should be examined in detail in a wider

perspective, specifically the unavailability of LF movement for adjunctanaphors, the possibility of licensing such maximal projections as predi-cates, and the percolation of prominence to a higher node. Though thediscussion here does not cover all the empirical aspects of ERs and thereremain many theoretical issues concerning binding theory and predica-tion theory, this paper has, I hope, achieved its primary goal of outliningthe salient properties of ERs.

REFERENCES

AKMAJIAN, ADRIAN; SUSAN STEEL; and THOMAS WASOW. 1979. The category AUX

in universal grammar. LI 10.1-64.

CHOMSKY, NOAM. 1981. Lectures on government and binding. Dordrecht: Foris.

-. 1986a. Knowledge of language: Its nature, origin, and use. New York:

Praeger.

-. 1986b. Barriers. Cambridge, MA: MIT Press.

CULICOVER, PETER, and WENDY WILKINS. 1984. Locality in linguistic theory. New

York: Academic Press.

GUERON, JACQUELINE. 1984. Topicalisation structures and constraints on corefer-

ence. Lingua 63.139-74.

IWASAWA, KATSUHIKO. 1988. Yuurisuuryooshi no zyozyutsusei to zyutsugosyooten-

ka-kinoo (The predicative force and predicate-focusing function of floatedquantifiers). Eibungaku Kenkyuu 65.75-87.

JACKENDOFF, RAY. 1972. Semantic interpretation in generative grammar. Cam-

bridge, MA: MIT Press.

LEBEAUX, DAVID. 1985. Locality and anaphoric binding. The Linguistic Review

4.343-63.

MAY, ROBERT. 1985. Logical form: Its structure and derivation. Cambridge, MA:

MIT Press.

Page 16: Hokkaido Institute of Pharmacy

ON EMPHATIC REFLEXIVES 51

MORAVCSIK, EDITH. 1972. Some crosslinguistic generalizations about intensifier

construction. CLS 8.271-7.

PARTEE, BARBARA. 1979. Subject and object in Modern English. New York: Gar-

land.

PICA, PIERRE. 1987. On the nature of reflexivization cycle. NELS 17.483-99.

ROBERTS, IAN. 1988. The representation of implicit and dethematized subjects.

Dordrecht: Foris.

ROTHSTEIN, SUSAN. 1985. The syntactic forms of predication. IULC.

RUSZKIEWICZ, PIOTR. 1984. Topics in the grammar of English reflexives. IULC.

SAFIR, KENNETH. 1987. What explains the definiteness effect? The representation

of (in) definiteness, ed. by Eric Reuland and Alice ter Meulen, 71-97. Cam-bridge, MA: MIT Press.