toward a unified account of imperatives: elimination …web.clas.ufl.edu/users/potsdam/papers/kondo...

13
590 FNGUSH liNGUISTICS. VOL. 17. NO.2 (2000) Walsh Dicky and Suzanne Urbanczyk, 385-420, GLSA, University of Mas- sachusetts, Amherst. Prince, Alan and Paul Smolensky (1993) "Optimality Theory: Constraint In- teraction in Generative Grammar," ms., Rutgers University and University of Colorado. Pullyblank, Douglas (1997) "Optimality Theory and Features," Optimality Theory: An Overview, ed. by Diana Archangeli and D. Terence Langen- don, 59-101, Blackwell, Malden, MA. Sagey, Elisabeth (1986a) The Representations of Features and Relations in Non- linear Phonology, Doctoral dissertation, MIT. Sagey, Elisabeth (1986b) "On the Representation of Complex Segments and Their Formation in Kinyarwanda," Studies in Compensatory Lengthening, ed. by Leo Wetzels and Engin Sezer, 253-295, Foris, Dordrecht. Smolensky, Paul (1993) "Harmony, Markedness, and Phonological Activity," paper presented at Rutgers Optimality Workshop-1, Rutgers University, October 23, 1993. Suzuki, Keiichiro (1998) A Typological Investigation of Dissimilation, Doctoral dissertation, University of Arizona. Zamma, Hideki (1994a) "Phonological Requirements on Suffixation," Tsukuba English Studies 13, 21-41, University of Tsukuba. Zamma, Hideki (1994b} "Accentuation of -ory, -ive, and -ion," English Lin- guistics 12, 248-271. Department of Anglo-American Studies Kobe City University of Foreign Studies 9-1 Gakuen-higashimachi Nishi-ku, Kobe-shi Hyogo 651-2187 e-mail: [email protected] t 1 1 1 1 i t f 1 J l l I j REVIEW ARTICLE ----·······-· ·-·-- Requestors must comply with \ Copyright law (Title 17 U.S. Code) TOWARD A UNIFIED ACCOUNT OF IMPERATIVES: ELIMINATION OF COMPLEXITY IN GRAMMAR MAKOTO KONDO Shizuoka University* Syntactic Issues in the English Imperative, by Eric Potsdam, Garland Publishing, Inc., New York, 1998, x+428pp. Keywords: imperative, word order, VP ellipsis, subjunctive I. Introduction In the history of generative grammar, English imperative sentences have attracted less attention than declarative and interrogative sen- tences and often received construction-specific accounts for their apparently peculiar behavior that is not observed in finite clauses. The book under review is a close study of English imperative clause struc- ture and the author argues that the English imperatives have largely regular syntactic behavior that falls within a conventional conception of English clause structure. In other words, the main goal of this study is to demonstrate that all the examples in (la-c) have a common clause structure exemplified in (2): ( 1) a. You closed the window. b. Did you close the window? c. Oose the window! ( 2) (cp C [IP DP I VP)] If the imperatives have the same clause structure as the interrogatives and the declaratives, we can eliminate undesirable complication in the • I would like to thank two anonymous EL reviewers for their critical comments and suggestions. All remaining inadequacies are my own. Ellglillt LilrpisJies 17: 2 (2000) 591-613 - 591 · ·· @ 2000 by the English Linguistic Society of Japan

Upload: others

Post on 05-Mar-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: TOWARD A UNIFIED ACCOUNT OF IMPERATIVES: ELIMINATION …web.clas.ufl.edu/users/potsdam/papers/Kondo 2000.pdf · that auxiliaries (i.e. have and be) and modals show parallel behavior

590 FNGUSH liNGUISTICS. VOL. 17. NO.2 (2000)

Walsh Dicky and Suzanne Urbanczyk, 385-420, GLSA, University of Mas­sachusetts, Amherst.

Prince, Alan and Paul Smolensky (1993) "Optimality Theory: Constraint In­teraction in Generative Grammar," ms., Rutgers University and University of Colorado.

Pullyblank, Douglas (1997) "Optimality Theory and Features," Optimality Theory: An Overview, ed. by Diana Archangeli and D. Terence Langen­don, 59-101, Blackwell, Malden, MA.

Sagey, Elisabeth (1986a) The Representations of Features and Relations in Non­linear Phonology, Doctoral dissertation, MIT.

Sagey, Elisabeth ( 1986b) "On the Representation of Complex Segments and Their Formation in Kinyarwanda," Studies in Compensatory Lengthening, ed. by Leo Wetzels and Engin Sezer, 253-295, Foris, Dordrecht.

Smolensky, Paul (1993) "Harmony, Markedness, and Phonological Activity," paper presented at Rutgers Optimality Workshop-1, Rutgers University, October 23, 1993.

Suzuki, Keiichiro (1998) A Typological Investigation of Dissimilation, Doctoral dissertation, University of Arizona.

Zamma, Hideki (1994a) "Phonological Requirements on Suffixation," Tsukuba English Studies 13, 21-41, University of Tsukuba.

Zamma, Hideki (1994b} "Accentuation of -ory, -ive, and -ion," English Lin­guistics 12, 248-271.

Department of Anglo-American Studies Kobe City University of Foreign Studies 9-1 Gakuen-higashimachi Nishi-ku, Kobe-shi Hyogo 651-2187 e-mail: [email protected]

t 1 1 1 1 i

~

t f 1 J

l ~ l I

j

REVIEW ARTICLE ----·······-· -· ·-·--

Requestors must comply with \ Copyright law (Title 17 U.S. Code)

TOWARD A UNIFIED ACCOUNT OF IMPERATIVES: ELIMINATION OF COMPLEXITY IN GRAMMAR

MAKOTO KONDO

Shizuoka University*

Syntactic Issues in the English Imperative, by Eric Potsdam, Garland Publishing, Inc., New York, 1998, x+428pp.

Keywords: imperative, word order, VP ellipsis, subjunctive

I. Introduction

In the history of generative grammar, English imperative sentences have attracted less attention than declarative and interrogative sen­tences and often received construction-specific accounts for their apparently peculiar behavior that is not observed in finite clauses. The book under review is a close study of English imperative clause struc­ture and the author argues that the English imperatives have largely regular syntactic behavior that falls within a conventional conception of English clause structure. In other words, the main goal of this study is to demonstrate that all the examples in (la-c) have a common clause structure exemplified in (2):

( 1) a. You closed the window. b. Did you close the window? c. Oose the window!

( 2) (cp C [IP DP I VP)] If the imperatives have the same clause structure as the interrogatives and the declaratives, we can eliminate undesirable complication in the

• I would like to thank two anonymous EL reviewers for their critical comments and suggestions. All remaining inadequacies are my own.

Ellglillt LilrpisJies 17: 2 (2000) 591-613 - 591 · ·· @ 2000 by the English Linguistic Society of Japan

Page 2: TOWARD A UNIFIED ACCOUNT OF IMPERATIVES: ELIMINATION …web.clas.ufl.edu/users/potsdam/papers/Kondo 2000.pdf · that auxiliaries (i.e. have and be) and modals show parallel behavior

592 ENGLISH LINGUISTICS, VOL. 17, NO.2 (2000)

grammar of English that would otherwise be stipulated by appealing to construction-specific clause structure of the imperatives.

Since the author is claiming that the clause structures are the same in (la-c), he has a burden to prove that distinctive properties of the im­peratives are explained without appealing to structural differences be­tween the imperatives on the one hand and the declaratives and the in­terrogatives on the other. The study centers around the following issues:

( 3 ) a. verb movement in imperatives h. status of subject in imperatives c. status of dol don't/ do not in imperatives

The author's arguments concerning these issues are outlined in the fol­lowing three sections. Discussion on the analysis is provided in section 5. Section 6 is the conclusion. 1

2. Verb Movement

Since the very early stage of generative grammar, it has been noticed that auxiliaries (i.e. have and be) and modals show parallel behavior with respect to do-support and that their behavior makes sharp contrast with that of main verbs:2

( 4 ) a. They will not come. b. *They (do) not will come.

( 5 ) a. They have not left. b. *They (do) not have left.

( 6 ) a. They are not invited. 3

b. *They (do) not be invited.

l A large portion of this review article is devoted to outlining the author's origi­nal arguments because the book under review contains a lot of new empirical findings and the reviewer finds it important to present them as much as possible.

2 In what follows, the reviewer will use the term auxiliary as a cover term exclu­sively referring to progressive be, passive be. copula be, and perfective have for ex­pository convenience; however, the terminology does not mean that the reviewer is making any claim about the semantics of the auxiliaries and the modals .

.l Since progressive be, passive be and copula be behave in exactly the same way with respect to the phenomenon under discussion, the reviewer cites only one type of be. The same holds true for the phenomena to be discussed later.

·l

4 i l

~ i 4 l

1 ~ ., I

.1 ~

i

·1

l

t ...

TOWARD A UNIFIED ACCOUNT OF IMPERATIVES 593

( 7) a. *They smoke not. b. They do not smoke.

The paradigm observed in (5) and (6) is reversed in imperative sen-

tences: ( 8 ) a. *Have not left us a mess to clean up!

b. Do not have left us a mess to clean up! ( 9 ) a. *Be not making noise!

b. Do not be making noise! The most pervasive analysis of the contrast in (4)-(7) attributes it to

the presence or absence of V-to-1 movement across not. Lasnik (1990), Beukema and Coopmans (1989), Pollock (1989) and Zhang (1990) suggest, based on imperative examples with negation like (8) and (9), that the auxiliaries do not undergo V-to-1 movement in the im­peratives. The author agrees with those previous analyses about the unavailability of V-to-1 movement in the imperatives. In chapter 2 of the book under review, the author presents new empirical evidence that supports the analysis. The evidence is obtained through two additional

diagnoses: (10) Adverb Diagnostic for Verb Raising

A verb has undergone Verb Raising if and only if it may appear to the left of an S-adverh.

(11) VP Ellipsis Diagnostic for Verb Raising a. If a verb can be interpreted as part of an elided VP then

it has not undergone Verb Raising in the VPE anteced­

ent. b. A verb can license a null VP within its extended projec-

tion if and only if it has undergone Verb Raising. In addition to the negation data in (8) and (9), the adverb diagnostic

also indicates that V-to-I movement is unavailable in the imperatives: (12) a. ?Have certainly read at least the introduction !4

b. Certainly have read at least the introduction! (13) a. *Be definitely in bed by nine o'clock!

b. Definitely be in bed by nine o'clock! If the auxiliaries were able to undergo V-to-1 movement, (12a) and

4 The contrast in (12) is not as strong as the examples with be.

Page 3: TOWARD A UNIFIED ACCOUNT OF IMPERATIVES: ELIMINATION …web.clas.ufl.edu/users/potsdam/papers/Kondo 2000.pdf · that auxiliaries (i.e. have and be) and modals show parallel behavior

594 ENGLISH LINGUISTICS, VOL. 17, NO.2 (2000)

(l3a) should be as acceptable as (12b) and (13b), respectively. The degraded acceptability of (12a) and (13a), therefore, indicates the inap­plicability of V-to-1 movement to the imperative auxiliaries.

The first half of the VP ellipsis diagnostic also confirms the gener­alization that there is no V-to-I movement in the imperatives:

(14) a. Please have studied the manuscript beforehand if you can 0!

b. *He has studied the manuscript beforehand and you should 0 too.

In ( l4a), the auxiliary can be interpreted as part of the null VP (i.e. 9), which is not the case in (14b). If have moved to I in (14a), the inter­pretation of the null VP should not include the auxiliary, as in (14b). The interpretive difference between (14a) and (14b), therefore, sup­ports the generalization that V-to-I movement is unavailable in the imperatives. The same holds true for be:

(15) a. Be elected if you can 0! b. *John was elected and you can 0 too.

The acceptability of (15a) indicates the absence of V-to-I movement and the unacceptability of ( 15b) shows that was is moved to I.

The second expectation of the VP ellipsis diagnostic is also borne out:

( 16) a. *The Smiths might have reached a decision regarding the matter. Jane, please have 0 too!

b. The Smiths might have reached a decision regarding the matter so we want Jane to have 0 also.

(17) a. *Huey is doing his homework now. When I come home, you be 0 too!

b. Huey is doing his homework now and when I come home, I hope that you are 0.

According to the diagnostic, it should be the case that the imperative auxiliaries remain in situ in (16a) and (17a) since they do not license the null VPs. The acceptability of (16b) and (17b) shows that the aux­iliaries are outside VP in the declaratives since the null VPs are acceptable in these examples.

The author's observation so far indicates that V-to-I movement is not allowed in the imperatives even if nothing seems to block it. The au­thor finds another construction that also precludes V-to-1 movement. The construction is a subjunctive clause. The author assumes, follow­ing Roberts (1985) and Lasnik (1995), that the subjunctive IP has a

~ ·' 1

'i .. ·~

l

l ..

j •

TOWARD A UNIFIED ACCOUNT OF IMPERATIVES 595

zero modal in its head, which prohibits anything else from occupying that position either by movement or by direct insertion. The subjunc­tive clauses show no V-to-I movement according to any of the above diagnoses. Relative word order of not and auxiliaries. for example, in­dicates that V-to-1 movement is prohibited in the subjunctive clauses:

(18) a. *It is advised that everyone be not fooled by his chican­ery.

b. The librarian insisted that the book not be returned in shabby condition.

Other two diagnoses also confirm the claim that V-to-1 movement is not available in the subjunctive clau.'ies. Ba'ied on the general un­availability of V-to-1 movement in those constructions. the author claims that do-support in the imperatives is not a surprising state of affairs and he states (p. 156): " ... whatever the correct analysis of do in finite clauses is, it extends, at no cost, to imperatives because their syn­tactic behavior is fully parallel."

3. Subject

Another salient distinction between the imperatives and the finite clauses is the optionality of subject as in (19):5

(19) a. You take out the trash! b. Take out the trash!

If the clause structure of the imperatives does not differ from that of interrogatives and declaratives, we expect that imperatives like ( 19b) have some empty category in their subject positions. The author dem­onstrates, in chapter 3 of the book under review, that the expectation is borne out. The null imperative subjects control PRO, bind anaphors,

5 Note that the imperatives allow a nonvocative DP to occupy their subject posi­tions:

( i ) Nobody forget his lunch for the picnic tomorrow! The above example has none of the characteristic properties of the vocatives; (a) it is acceptable witbout a pause between nobody and the rest of the sentence, (b) the subject DP does not refer to the addressee of the utterance, which is the only possi­ble referent referred to by the vocatives, and (c) the subject DP binds the third per­son pronoun his. which is an impossible binding relation if the binder is a vocative DP.

Page 4: TOWARD A UNIFIED ACCOUNT OF IMPERATIVES: ELIMINATION …web.clas.ufl.edu/users/potsdam/papers/Kondo 2000.pdf · that auxiliaries (i.e. have and be) and modals show parallel behavior

5% ENGUSH LINGUISTICS, VOL. 17, NO.2 (2000)

and can be associated with such syntactic operations as passive, raising, unaccusative advancement, tough movement and quantifier float. 6 It is, therefore, reasonable to assume the existence of a null subject even if the imperatives do not have an overt subject.

As for the status of the null imperative subjects, the author argues that they are pro rather than PR0. 7 They alternate with a lexical DP without requiring any other overt syntactic consequences as pro does. The alternation of PRO with a lexical DP, on the other hand, requires an overt Case assigner to appear in the structure. The interpretation of the imperative null subjects also favors the analysis under which they arc analyzed as pro. The imperative null subjects always refer to the addressee of the utterance and cannot receive an arbitrary interpreta­tion. PRO, on the contrary, receives an arbitrary interpretation when

1' The distribution of floating quantifiers in the subjectless imperatives is re·

stricted. Consider the following examples: ( i ) Don't both/all go! ( ii ) *Both/ All go!

If an overt head locally c-commanding the quantifier is provided, the acceptability significantly improves:

(iii) • All make your beds before the guests arrive! ( iv) Please have all made your beds before the guest~ arrive!

The phenomenon, however, is not unique to the subjectless imperatives. The float­ing quantifiers associated with PRO show the similar behavior:

( v) *They promised PRO all to leave. (vi) They promised PRO to all leave.

7 The author considers four types of empty categories in accordance with Chom­sky (1982) and argues against the approach which assumes that the imperative null subjects are traces of movement. In particular, he argues against Beukema and Coopmans (1989), who analyze the imperative null subjects as variables bound by a null topic operator topicalized from the subject position. He presents two argu­ments against the movement approach:

( i ) a. The imperative null subjects cannot be interpreted as third person even when the discourse provides a potentially appropriate third per­son.

b. The imperatives allow topicalization whereas multiple topicalization is generally prohibited.

Although the author does not provide any specific arguments against an analysis under which the imperative null subjects are A-traces. the readers will easily see that (ia) can be carried over to arguments against the A-trace approach. Needless to say. assuming an additional A-position above the matrix IP is very difficult and independently unmotivated if not impossible.

·~

l

i

. j :.; •

TOWARD A UNIFIED ACCOUNT OF IMPERATIVES 597

the controller is missing. If the null imperative subjects were PRO. they should receive an arbitrary interpretation, contrary to the fact, since the imperatives with a null subject do not have any overt control­ler of the subject.

If the imperative null subject is best analyzed as pro, an obvious question to ask arises. How is it licensed? English is not a null sub­ject language since it does not allow occurrence of a null subject in finite clauses. Adopting Rizzi's (1986) Licensing and Recovery Condi­tions on pro, the author explores this issue. The conditions consist of two requirements. One is that pro is Case-marked by its licensing head. The other is that pro has the grammatical specification of the features on the licensing head. Putting it differently. the occurrence of pro is licensed by the governing head and the content of pro is re­covered by the features on the licensing head. The author assumes that the licensing head is J.M

In order to spell out how pro is licensed, it is necessary to identify the exact features that are needed to recover pro. The author argues that the relevant feature is person feature, based on the observation by Borer (1983), Kenstowicz (1989) and Farrell ( 1990) among others.9

Farrell (1990) observes that Brazilian Portuguese allows a null object without any agreement on the governing head and that the null object is limited to the third person. He claims that the recovery portion of Rizz's licensing condition should be modified so that pro can be re­covered if it has the grammatical specifications of designated features. According to Farrell, pro in Brazilian Portuguese inherently has the feature for the third person and, consequently, it freely occurs without any grammatical agreement on the licensing head. Extending Farrell's proposal, the author claims that pro is recovered in imperatives because the semantics of the imperatives guarantees that the imperative null

H One may wonder whether the imperative I has the ability to assign Case since it is nonfinite. It has been proposed, however. that at least M>mc nontinite I has capability to assign (or check in the minimalist terminology) Case. Sec Martin (1992, 19%) and Chomsky (1995).

9 Borer (1983) observes that Hebrew allows a null subject only in the case of the first and second person appearing in the past and future tenses. Kenstowicz { 1989) finds that inflection for person is a necessary condition for a null subject in the Bani-Hassan dialect of Arabic .

Page 5: TOWARD A UNIFIED ACCOUNT OF IMPERATIVES: ELIMINATION …web.clas.ufl.edu/users/potsdam/papers/Kondo 2000.pdf · that auxiliaries (i.e. have and be) and modals show parallel behavior

598 ENGLISH LINGUISTIC'S. VOL. 17. NO.2 (2000)

~;ubject is uniquely the second person. Citing Sadock (1974), Leech ( 1983) and others, the author defines the core meaning of the impera­tives as (20):

(20) In an imperative, it is proposed that the addressee bring about an event.

In the canonical case, the addressee will bring about the event by being its agent. Since the typical syntactic realization of an agent role is sub­ject, tht! subject of the imperative includes the addressee. Since a DP is the second person when its referent includes the addressee, the imperative null subject is uniquely identified as the second person. Accordingly, the recovery portion of the licensing condition is satisfied and the occurrence of pro is allowed in the imperatives.

4. Structure of Imperatives

4.1. Do/Don't We have seen, following the author's arguments, that imperative VP

has no significant differences from other VPs with respect to verb movement (section 2) and that there is no essential distinction between imperative subjects and subjects in finite clauses (section 3). The re­maining elements that contribute to the structure of the imperatives are do and don't, which occur in emphatic imperatives and negative im­peratives, respectively. 10 The author discusses the status of these ele­ments in chapter 4 of the book under review.

Do and don't typically occupy the sentence initial position: 11

(21) Do SOMEONE give him the money before he uses that gun!

to The negative imperative of the following form will be discussed later: ( i ) Do not walk on the grass!

II Some researchers claim that the occurrence of subject is restricted in the emphatic and the negative imperatives. Schmerling (19n) and Stockwell et al. (1973), for example, take the position that the emphatic imperatives do not allow subject to appear and Stockwell et al. (1973) implicitly assumes that don't always precedes subject. The author, following the observation made by Davies (1986), takes the position that any of the followings is possible in the imperatives:

( i ) a. Do/Don't subject VP b. subject do/don't VP

' ~~

J

TOWARD A UNIFIED ACCOUNT OF IMPERATIVES 599

(22) Don't anyone open this box! The author argues that they are base-generated in I and then move to the sentence initial position. In null subject imperatives with do/don't, they remain in situ. The empirical basis for this analysis is provided by the similarity between the imperatives and finite clauses:

(23) Those without galoshes do wear their boots instead! (24) Passengers with luggage don't leave their valuables un-

attended! Roughly speaking, (21) and (22) are the imperative counterparts of the corresponding yes-no questions and (23) and (24) arc the imperative counterparts of the corresponding declaratives.

In addition to the apparent similarity between the imperatives and the finite clauses, the author presents another empirical evidence that supports the analysis. The argument is based on VP ellipsis:

(25) a. Don't you be seen with those losers and don't anyone else be 0 either!

b. Someone needs to volunteer to help with the church bazaar. Oh yes, please, do someone If!

If neither do nor don't is present, the imperatives do not allow VP ellipsis:

(26) *Jack started reading the poem, now Jimmy keep If! As discussed in section 2, VP ellipsis requires an overt 1-elcment:

(27) a. John didn't leave but Mary did. b. *John didn't leave but Mary.

The parallel contrast in (25)-(27) supports the claim that do and don't in (25) are first inserted into the head of IP and then moved to their surface positions.

The author further demonstrates that do and don't in the imperatives show semantic parallelism with do/don't in declarative sentences. ln the case of don't, the parallelism is transparent; that is, it serves as a bearer of negation in both clause types.

In the declaratives, do serves to signal contrastive polarity in the sense that it is used to deny or contradict a negative claim which has been stated or implied in the discourse:

(28) Contrary to what you think, we DID enjoy ourselves at the opera.

Do is also used to enforce an assertion; that is, it affirms or emphasize:> the speaker's attitude- his/her enthusiasm. approval or conviction­towards the proposition:

Page 6: TOWARD A UNIFIED ACCOUNT OF IMPERATIVES: ELIMINATION …web.clas.ufl.edu/users/potsdam/papers/Kondo 2000.pdf · that auxiliaries (i.e. have and be) and modals show parallel behavior

600 ENGLISH LINGUISTICS, VOl.. 17, NO.2 (2000)

(29) I DO wish you would listen. Do in the imperatives is also used in similar ways. The use of do in (30) and (31) is comparable with that in (28) and (29), respectively:

(30) I'm doing my best not to make any mistakes. Do make one, so what'! It doesn't matter that much.

(31) DO sit down! Following Bobaljik's (1995) analysis of do-support, the author pro­

poses the following structure (sec also Laka (1990)): (32) IP

~ DP I' ~ I 2:P I~

do +EMPHi

" lj

VP

2:P is the affirmative counterpart of NcgP and it occupies the same structural position as NegP. Semantically, the head EMPH is the locus of the contrastive polarity and the enforcement of affirmation. Phonetically, it is realized as heavy stress. Syntactically it blocks ad­jacency between I and V, resulting in do-support.

4.2. Subject Position I-to-C movement is a prevailing grammatical device in explaining the

different word order patterns between declarative sentences and their interrogative counterparts. Given the VP-Intemal Subject Hypothesis and/or a more articulated structure of IP, 1-to-C movement is not the only possible derivation that explains the word order variation in the imperatives (see Kuroda (1988), Koopman and Sportiche (1991) and Pollock (1989)). The author examines three possible analyses in chap­ter 5 of the book under review.

The first one is to assume that do is inserted into I and that move­ment of subject DP derives the word order variation:

(33) (IP (DP) [r do [vP (DP) V']]] Under the second analysis, the subject does not move and 1-to-C move­ment gives rise to the word order change:

(34) [cp (do) [II' DP (do) VP)] If this analysis is correct, the word order variation in the imperatives is

.. ·;

i

l

TOWARD A UNIFIED ACCOUNT OF IMPERATIVES 601

the exact counterpart of the word order change observed in declaratives and their interrogative counterparts. The third analysis assumes a more articulated structure of IP and posits more than one functional projection between VP and CP:

(35) (cp C [xP (DP) [x- (do) (yp (DP) (y- (do) VP))]}] The author claims that the first analysis makes wrong predictions

about the word order of the elements that appear in the imperatives: (36) a. Do AT LEAST YOU have tried it before you begin to

criticize! b. *Do have at least you tried it before you begin to criti­

cize! (37) a. Don't anyone be fooled by his shoddy argumentation!

b. *Don't be anyone fooled by his shoddy argumentation! According to the first analysis, subjects must stay in situ when do or don't appears in the sentence initial position. The contrast in (36) and (37) clearly shows that the subjects must precede the auxiliaries even if the imperatives contain do/don't. The fact is not compatible with the analysis that docs not assume the movement of do/don't. The struc­ture in (33) wrongly predicts that (36a) and (37a) should be unaccept­able and that (36b) and (37b) should be acceptable.

The author gives two additional empirical arguments against the first analysis. One is that the analysis makes a wrong prediction about the occurrence of floating quantifiers:

(38) a. Don't you be both talking at once! b. Don't you have all made a mess!

If the subjects remained in situ, the examples could not be derived. The other is that the analysis is also denied based on VP ellipsis:

(39) Rick walked out of the lecture, but don't everyone else 0, please!

According to the first analysis, VP ellipsis must delete the subject with­in VP and that is simply not the case .

The second and the third analyses diverge in their prediction with re­spect to topicalization. If topicalization is an operation that adjoins a topic to IP as argued by Baltin (1982) and Lasnik and Saito ( 1992), the former approach derives the following word order but the latter docs not:

(40) (do/don't) topic subject The examples below seem to favor the third analysis since the second analysis does not derive the acceptable (41a) and does derive the un-

Page 7: TOWARD A UNIFIED ACCOUNT OF IMPERATIVES: ELIMINATION …web.clas.ufl.edu/users/potsdam/papers/Kondo 2000.pdf · that auxiliaries (i.e. have and be) and modals show parallel behavior

602 ENGLISH LINGUISTICS, VOL. 17, NO.2 (2000)

acceptable (41b): (41) a. My good wine, don't anybody touch!

b. *Don't, my good wine, anybody touch! In defending the second analysis, the author points out the parallel con­trast observed in interrogatives:

(42) a. The report shouldn't you have read already? b. *Shouldn't, that report, you have read already?

The parallelism between (41) and (42), according to the author, sug­gests general prohibition against cooccurrence of 1-to-C movement and an IP adjoined element. If the argument is on the right track, (41b) is not a problem to the second analysis. As for the derivation of the acceptable example in (41a), the author adopts Rochemont's (1989) analysis of topicalization, which allows adjunction to CP as well as ad­junction to IP.

Another potential problem to the second approach involves what is often called negative inversion:

(43) (*)With no analysis do be satisfied! Notice, however, it is very difficult to tell if (43) is truly derived by negative inversion because topicalization of PP derives (43) from (44):

( 44) Do be satisfied with no analysis! ·n1e reason for the difficulty is clear; topicalization of the adverbial de­rives the same word order as the structure derived by negative inver­sion since do appears in the sentence initial position in (44).

Jn order to avoid the difficulty, the author employs a semantic test. Liberman (1974) observes that (45a) has two readings given in (45b, c):

(45) a. John would be happy with no job. b. There is no job such that John would be happy with it.

('unsatisfiable' meaning) c. John would be happy without a job. ('unemployed'

meaning) Negative inversion and topicalization of PP eliminate the ambiguity in a different way:

(46) a. With no job would John be happy. b. With no job, John would be happy.

The only possible reading of (46a) is (45b) and (46b) has the 'unem­ployed' meaning alone. (44) is associated with the following two meanings:

( 47) a. There is no analysis that you should be satisfied with. b. You should be satisfied with having no analysis.

-'

TOWARD A UNIFIED ACCOUNT OF IMPERATIVES 603

On the other hand, (43) marginally allows the reading in (47b) but does not mean (47a); then, the interpretation of (43) indicates that negative inversion is not compatible with the imperatives. The result seems to support the third analysis again since the second analysis apparently

allows negative inversion in the imperatives. If we examine negative inversion in interrogatives, however, we find

the parallel behavior: (48) With no job would John be happy?

(48) only has the 'unemployed' meaning. Since 1-to-C movement in the interrogatives is well established, (48) indicates that negative inver­sion is not compatible with independent 1-to-C movement. If this is the case, (43) is not a counterexample to the second analysis because the unavailability of negative inversion receives an independent ex­planation; that is, whatever rules out the negative inversion in ( 48) also excludes the negative inversion in (43). The parallel behavior of the imperatives and the interrogatives leads the author to adopt the second analysis though he acknowledges that the third analysis is still compati­

ble with the data given above.

4.3. Do Not Do and not may separately appear in the imperatives:

( 49) Do not expect to get rich immediately! Chapter 6 of the book under review is devoted to the analysis of im­peratives like (49). Given the discussions so far, the null hypothesis about the phrase structure of ( 49) would be as follows:

(50) (Ip pro do [NcgP not VPJ) If the clause structure of the imperatives is exactly the same as that of interrogatives and declaratives, we expect every word order in (51) to

be found in the imperatives: (51) a. You do not like dogs.

b. Do you not like dogs? c. (?)Do not any of you with rural childhoods like dogs?

The author notes that the expectation is not borne out: 12

12 It is important to note that judgments in (52) do not indicate absolute accepta­bility; rather, they show relative acceptability in comparison with other relevant ex-

amples.

Page 8: TOWARD A UNIFIED ACCOUNT OF IMPERATIVES: ELIMINATION …web.clas.ufl.edu/users/potsdam/papers/Kondo 2000.pdf · that auxiliaries (i.e. have and be) and modals show parallel behavior

604 ENGLISH LINGUISTICS. VOL 17, NO.2 (2000)

(52} a. ?SOMEone do not abandon the gate! The fight is not yet lost and we must maintain the security.

b. *Do someone not abandon the gate! c. ?Do not EVERYone stand off by himself in the comer,

come join the party ! (52a) and (52c) are derived in the same way as (51a) and (5lc} are. Specifically, (51a) and (52a) are derived by direct insertion of do into I. (51c) and (52c) arc derived by adjunction of not onto do followed by I­to-C movement. The complete acceptability of (51b) makes sharp contrast with the totally unacceptable (52b). The contrast seems to contradict the author's approach, which tries to assign the imperatives the same clause structure as that of the finite clauses.

The author points out, however, that subjects arc not the only ele­ments that cannot intervene between do and not:

(53) a. *Please do all not talk at once! b. *Do definitely not tell them where you're from!

(53a) and (53b) indicate that neither a floating quantifier nor an S­adverb can intervene between do and not. We can find the same pat­tern even in declaratives:

(54) a. *The twins did both not vote. b. *He does obviously not understand the procedure.

(54) shows that neither the floating quantifier nor the adverb can inter­vene between do and not in the declaratives. The examples improve significantly if do is replaced by a modal:

(55} a. The twins could both not vote. b. He would obviously not understand the procedure.

The parallel between (52b) and (54) and the contrast between (54) and (55) lead the author to propose that the unacceptability of (52b) should not be derived from some unique property of the imperatives; rather, it should be reduced to some peculiar property of do and not. On the descriptive level, the examples in (52)-(54) indicate that nothing inter­venes between do and not if do is in I. Accordingly, the author pro­poses an obligatory PF rule that requires not to attach to do only if do is present in I. If one could successfully argue that do in this type of imperatives always remains in situ, (52b) falls within the generalization that nothing can intervene between do and not if do is in I; according­ly, the proposed PF rule independently accounts for the unacceptability of (52b).

If this argument is on the right track, why is it the case that do in

.. ' .t

4 \

.~:

.j

... ~

TOWARD A UNIFIED ACCOUNT OF IMPERATIVES 605

emphatic imperatives, don't in negative imperatives and do not in (52c) may move to C? One property shared by these three elements and not carried by do in (52b) is that those clements that may move to C have an impact upon the interpretation of a sentence: negation and emphatic affirmation. Do in (52b) is semantically vacuous and has no impact on sentence interpretation. Do in the interrogatives signals a clause type; hence eligible to undergo I-to-C movement.

Based on the consideration above, the author spells out a syntactic analysis that employs Checking Theory (see Chomsky (1995)). He in­troduces an optional feature to be checked in C, which he calls IMP. He further assumes that IMP is checked by the categorial feature of ~. the category that syntactically realizes negation and affirmation. Both IMP and ~ being optional, there arise four cases to consider:

(56) IMP ~ result

a. present present 1-to-C movement of do/don't -·------

b. present absent crash

c. absent present do/don't in situ

d. absent absent imperatives without do/don't If both IMP and ~ are present as in (56a), ~ moves to C in order to check IMP inC; as a result, I-to-C movement takes place, deriving the imperatives with either emphatic do, don't, or do not in C. In the case of (56b), the derivation crashes because unchecked IMP remains. In the case of (56c), either emphatic do or don't is introduced into the phrase structure and they remain in situ because C has no feature to be checked, and this derives the imperatives with subject in the sentence initial positions. (56d) derives the imperatives without do nor don't. ~ being absent, (56d) does not contain anything that induces do­support. (56) covers the all and only grammatical patterns observed so far:

(57) a. (cp do/don't; (lp (subject) [r t; b;P :S VP])]) b. (cp [do not); [IP (subject) [r t; b:P ~ VP]])] c. [IP (subject) [r do/ don't [ ~P ~ VP}J] d. (ip (subject) [r do not b;P :S VP]]) e. [iP (subject) [r I VP])

(57a, b), (57c, d) and (57e) correspond to (56a). (56c) and (56d). re­spectively

Page 9: TOWARD A UNIFIED ACCOUNT OF IMPERATIVES: ELIMINATION …web.clas.ufl.edu/users/potsdam/papers/Kondo 2000.pdf · that auxiliaries (i.e. have and be) and modals show parallel behavior
Page 10: TOWARD A UNIFIED ACCOUNT OF IMPERATIVES: ELIMINATION …web.clas.ufl.edu/users/potsdam/papers/Kondo 2000.pdf · that auxiliaries (i.e. have and be) and modals show parallel behavior

606 ENGLISH LINGUISTICS, VOL. 17, NO.2 (2000)

5. Discussion

Although the author presents a large body of new empirical findings concerning English imperatives, his analysis of the data faces several theoretical difficulties. The following two subsections deals with two of the problems.

5.1. Recovering Pro Subjects One of the most salient properties of English imperatives is that they

allow a null subject which always refers to the addressee. It is neces­sary to account for this property in order for an analysis of the impera­tives to be convincing. The author's analysis, however, faces a tech­nical problem in identifying pro subjects in English imperatives.

According to the author's analysis, the imperative null subject is pro and it is licensed by the imperative I and the content of pro is semanti­cally identified as the second person.

The analysis crucially depends on the following assumption: (58) The imperative I has a specification for either the second or

the third person. If an imperative sentence with a null subject contains I with the second person feature, no problem arises since the referent of the second per­son DP necessarily includes the addressee.

A potential problem arises if a null subject imperative includes I with the third person feature. Note that the semantics of the imperatives does not prohibit the third person subject from appearing in the im­perative sentences:

(59) YOUR soldiers build the bridge, General Lee! How does the third person subject become compatible with the core meaning of the imperatives given in (20), which is repeated here as (60):

(60) In an imperative, it is proposed that the addressee bring about an event.

The author argues that the overt third person subject is allowed on the basis of control relationship between the subject and the addressee. The addressee of (59) is responsible for bringing about the event de­noted by (59) because he or she has control over the subject. Let us ask now what signals the existence of the control relationship in (59). It is the hierarchical relation obtained between the subject and the addressee that provides the basis for the control relationship. How do ~

I I ....

TOWARD A UNIFIED ACCOUNT OF IMPERATIVES 607

we find the hierarchical relation then? It comes from our real world knowledge, which is part of the discourse in the sense that the knowl­edge is shared by the speaker and the addressee. We then come to the conclusion that the discourse may provide the basis for the control relationship between the addressee and the imperative subject. A natural expectation would be that the third person pro should be allowed if an appropriate discourse is provided, which is contrary to the fact. Note that one cannot argue that the imperative I has a feature specification for the second person alone because it would wrongly rule out (59).

The assumption in (58) also has another theoretical problem. The assumption is nothing but a mere stipulation and the author does not provide any principled explanation for why the imperative I does not have a specification for the first person. It is necessary to assume (58) because the second person pronoun is not the only DP that includes the addressee in its referent. As is well known, the first person plural pro­noun, we, may include the addressee in its referent:

(61) Shall we dance together? What this example implies is that pro could not be uniquely recovered by the semantic consideration alone if the imperative I had a feature specification for the first person. The imperative I with the first person feature would allow the first person pro to appear in its subject posi­tion; then the imperative semantics would not uniquely identify the content of pro as the second person since the first person is compatible with the imperative semantics. The imperative semantics requires that the addressee bring about an event and the semantics would be com­patible with the first person subject as far as its referent includes the addressee. Accordingly, the author is forced to assume (58). Need­less to say, (58) is a construction-specific requirement and assuming it is to take a step backward from the original goal of the study.

Another problem is also related to the third person subject in the im­peratives:

(62) Nobody ride off with my bicycle! Since the referent of the subject is an empty set in (62), nobody has a control over the referent. To utter (62) is to require the addressee to bring about an event whose agent is not in the control relation with the addressee. If this is correct, (62) should be as bad as (63):

(63) *Be tall! It is often claimed that the unacceptability of (63) comes from the fact

Page 11: TOWARD A UNIFIED ACCOUNT OF IMPERATIVES: ELIMINATION …web.clas.ufl.edu/users/potsdam/papers/Kondo 2000.pdf · that auxiliaries (i.e. have and be) and modals show parallel behavior

608 ENGLISH LINGUISTICS. VOL. 17, NO.2 (2000)

that nobody can control one's own height. According to the author's analysis, it should be the case that the addressee is not in a control re­lationship with the subject (=the addressee) in the domain of control­ling one's own height. Accordingly, to utter (62) should carry the same effect as uttering (63), contrary to the fact. ,

The intuitive distinction between (62) and (63) is that the addressee in (62) can partly contribute to the realization of the imperative event by not riding off with the speaker's bicycle. In (63), on the contrary, nobody can contribute to the realization of the imperative event. This line of consideration might lead to the claim that the addressee is not necessarily in a control relationship with the subject if he or she can contribute to the realization of the event. This claim, however, makes a wrong prediction again. Recall the discussion concerning (59). The claim amounts to saying that the third person subject is freely allowed in the imperatives if the discourse provides an appropriate circumstance in which the addressee can contribute the realization of the imperative event. Consequently, it wrongly predicts the occurrence of the third person pro in the null subject imperatives.

5.2. Imperative Clause Type In the final chapter of the book under review, the author reaches the

conclusion that logically possible four kinds of interactions between IMP and ~ give rise to three types of well-formed imperatives and one in­appropriate derivation. The analysis crucially depends on the assump­tion that IMP and ~ are optional elements in a sentence. Let us con­sider the optionality of IMP. The optionality of IMP sharply contrasts with the obligatory presence of 0 feature in interrogatives:

(64) a. What did you buy? b. *You bought what?

(64) indicates that the presence of 0 is obligatory in the interrogatives. ( 64b) is ruled out because 0 feature is not checked. If the phrase structure of the interrogatives is the same as that of the imperatives, what is the grammatical clue that marks the imperative clause type? In particular, what distinguishes (65) from (66)?

(65) Don't you give me any lip! (66) Don't you give me any lip?

The difference in their intonation pattern should not be the clue if Chomsky (1995) is right in assuming that there is no PF-LF interaction. The existence of 1-to-C movement of don't is not the clue either since

J 1

:~

i ·~

TOWARD A UNIFIED ACCOUNT OF IMPERATIVES 609

both involve the movement. The best candidate for the clue is 0 fea­ture in the case of the interrogatives since it is an obligatory element that is not present in other clause types. In addition, the status of 0 feature as a trigger for movement also makes it the best probable candidate because the movement is obligatory in the interrogatives.

In the imperatives, ~ cannot be the marker of the clause type be­cause it is the locus of negation/affirmation, either of which is not the unique character of the imperatives. Other syntactic elements in the imperatives are not different from those in the interrogatives with one exception, namely, IMP. IMP therefore should be best analyzed as a marker of the imperative clause type. The author's analysis. however, does not allow IMP to serve as a marker of the imperative clause type since IMP is an optional element in the imperative sentences.

The author, recognizing the dissimilarity between IMP and 0 fea­ture, suggests (p. 392) that " ... interrogative inversion needs to be con­sidered in the larger picture of finite clauses." He points out that the presence of 0 feature is optional if we consider the interrogatives and declaratives altogether. This would mean that the interrogatives and the declaratives constitute a natural class and the imperatives belong to another class. We may ask. what is an empirical basis for that distinc­tion. The notion finite is not the counterpart of the notion imperative. The former is associated with verb form and the latter is related to clause types. The counterpart of the former is the notion nonfinite and the counterparts of the latter are such notions as declarative and inter­rogative. We can distinguish the imperatives from the interrogatives and the declaratives with respect to tensed/nontensed distinction. It may have a welcome consequence since it puts subjunctives into the same category as the imperatives. Recall that the subjunctives be­have in the same manner as the imperatives with respect to V-to-1 movement of auxiliaries. However, it makes identification of a clause type more difficult because more than one clause type is associated with each class. Our current concern is how to distinguish clause types. Although tensed/nontensed distinction may be useful in differentiating clauses that allow verb movement from those that do not. it is of little help, if any. to identify the imperative clause type.

One might wonder whether we could eliminate the optionality of IMP by assuming weak. IMP and strong IMP in the sense of Chomsky (1995) because the obligatory appearance of IMP in the imperatives would make it the best probable candidate for a maker of the impera-

Page 12: TOWARD A UNIFIED ACCOUNT OF IMPERATIVES: ELIMINATION …web.clas.ufl.edu/users/potsdam/papers/Kondo 2000.pdf · that auxiliaries (i.e. have and be) and modals show parallel behavior

610 ENGLISH LINGUISTICS, VOL. 17, NO.2 (2000)

tive clause type. If the author is correct in assuming that IMP is checked by ~. positing weak and strong IMP does not help either. Since IMP must be checked by ::E, even weak IMP requires the .pres­ence of ::S; consequently, it predicts, contrary to the fact, that all the imperatives involve do-support.

Although positing IMP in all the imperatives is conceptually desir­able, it makes a wrong prediction about the occurrence of do-support in the imperatives. The failure is rooted in the assumption that IMP must be checked by ~- Let us consider the possibility that some occurrences of IMP must be checked whereas others can remain un­checked. If we adopt the following stipulations, word order variation in the imperatives can be explained maintaining that IMP appears in all the imperatives:

(67) a. IMP may be inserted into either I or C. b. IMP in I is interpretable while IMP in C is not. c. Uninterpretable features must be checked in overt syn-

tax. (67a) may be justified because nothing a priori requires the matrix C to be projected at least in English. If C is not in the numeration, IP is the matrix 'clause' and IMP in the head of IP determines its 'clause' type. (67b) is not a new idea. Chomsky (1995), for example, assumes that D feature in T is uninterpretable while D feature in D is interpret­able. TI1e interpretability of features is determined by its position. (67c) may be problematic and is not well motivated empirically; how­ever, it is conceptually more preferable to strong/weak distinction since interpretability is a primitive in the minimalist program whereas strong/weak distinction does not have conceptual necessity.

Let us spell out every possibility derived from (67):

(68) IMP ::S result

a. 1 present (subject) do/don't/do not VP --- . ·- ·-- ·---··--- ----------b. I absent (subject) VP

c. c present do/don't/do not (subject) VP

d. C absent crash In (68a. b), IMP does not need to be checked since it is interpretable by virtue of being in I. The derivations corresponding to (68a, b) con­verge because they do not contain any unchecked features that are un­interpretable. In (6&, d), on the other hand, IMP must be checked

j

j

TOWARD A UNII'IED ACCOUNT OF IMPERATIVES 611

by ::E since it appears in C and is uninterpretable. Consequently, movement of ::E to C is obligatory; otherwise, the derivation would crash. In (68c), the derivation converges since ::S moves to C and checks IMP. In (68d), on the contrary, the derivation crashes because the absence of ::E makes it impossible to check IMP. As is evident from (68), we can derive every attested word order without excluding IMP from the imperatives. The most welcome consequence of this alternative is that we can identify the imperative clause type by appeal­ing to its unique feature, namely, IMP.

6. Concluding Remarks

The main goal of the book under review is to demonstrate that the English imperatives do not have any construction-particular clause structure; their clause structure is essentially the same as that of well­studied finite clauses in the sense that verb movement in the impera­tives shows no peculiar construction-specific behavior; the subject of the imperatives is truly the counterpart of the subject in the finite clauses; do, don't and do not in the imperatives behave in the same way as those that appear in the finite clauses. Toward this goal, the author presents a large body of empirical evidence that supports the claim.

The observation covers a wide range of phenomena. In addition to the imperatives, the observation covers declaratives, interrogatives, subjunctives and so on. These constructions are examined with several tests including negation, adverbials, VP ellipsis and floating quantifiers. Through these tests, the author demonstrates the unavailability of verb movement in the imperatives, the syntactic status of the imperative sub­ject as a true subject, the phrase structure of the imperatives and the status of do/don't/do not as counterparts of those found in the finite clauses. Although the reviewer has made some critical comments against the author's analysis, it does not mean that the empirical findings in the book under review are of no interests. Without a doubt, those empirical findings constitute the basis for further research and give rise to several issues to pursue and, needless to say, it is one of the best contributions to the field that one can make.

Page 13: TOWARD A UNIFIED ACCOUNT OF IMPERATIVES: ELIMINATION …web.clas.ufl.edu/users/potsdam/papers/Kondo 2000.pdf · that auxiliaries (i.e. have and be) and modals show parallel behavior

612 ENGLISH LINGUISTICS, VOL. 17, NO.2 (2000)

REFERENCES

Ballin, Mark (1982) "A Landing Site Theory of Movement Rules," Linguistic Inquiry 13, 1-38.

Beukema, Frits and Peter Coopmans (1989) "A Government-Binding Perspec­tive on the Imperative in English,'' Journal of Linguistics 25, 417-436.

Bobaljik, Jonathan David (1995) Morphosyntax: The Syntax of Verbal Inflec­tion, Doctoral dissertation, MIT.

Borer, Hagit (1983) Parametric Syntax: Ca.Ye Studies in Semitic and Romance Languages, Foris, Dordrecht.

Chomsky, Noam (1982) Some Concepts and Consequences of the Theory of Government and Binding, MIT Press, Cambridge, MA.

Chomsky, Noam (1995) The Minimalist Program, MIT Press, Cambridge, MA. Davies, Eirlys E. (1986) The English Imperative, Croom Helm, London. Farrell, Patrick (1990) "Null Objects in Brazilian Portuguese,'' Natural Lan-

guage and Linguistic Theory 8, 325-346. Kenstowicz, Michael (1989) "The Null Subject Parameter in Modern Arabic

Dialects," The Null Subject Parameter, ed. by Osvaldo Jaeggli and Ken Safir, 263-276, Kluwer, Dordrecht.

Koopman, Hilda and Dominique Sportiche (1991) "The Position of Subjects,'' Lingua 85. 211-258.

Kuroda, Shige-Yuki (1988) "Whether We Agree or Not: A Comparative Syn­tax of English and Japanese," Linguisticae lnvestigationes 12, 1-47.

Laka, Itziar (1990) Negation in Syntax: On the Nature of Functional Categories and Projections, Doctoral dissertation, MIT.

Lasnik. Howard (1990) Essays on Restrictiveness and Learnability, Kluwer, Dordrecht.

Lasnik, Howard (1995) "Verbal Morphology: Syntactic Structures Meets the Minimalist Program," Evolution and Revolution in Linguistic Theory: Essays in Jlonor of Carlos Otero, ed. by Paula Kempchinsky and Hector Campos, 251-275. Georgetown University Press, Washington, D.C.

Lasnik, Howard and Mamoru Saito (1992) Move a: Conditions on Its Applica­tion anJ Output, MIT Press, Cambridge, MA.

Leech. Geoffrey N. (1983) Principles of Pragmatics, Longman, London. Liberman, Mark (1974) "On Conditioning the Rule of Subject-Auxiliary Inver­

sion," NELS 5. 77-91. Martin. Roger ( 1992) "On the Distribution and Case Features of PRO," ms.,

University of Connecticut. Martin. Roger (1996) A Minimalist Theory of PRO and Control, Doctoral dis­

sertation, University of Connecticut. Pollock, Jean-Yves (1989) "Verb Movement, Universal Grammar, and the

Structure of IP,'' Linguistic Inquiry 20, 365-424. Ri7.zi. Luigi (1986) ''Null Objects in Italian and the Theory of pro," Linguistic

Inquiry 17, 501-557.

:~ .. ,

j

TOWARD A UNIFIED ACCOUNT OF IMPERATIVES 613

Roberts, Ian G. (1985) "Agreement Parameters and the Development of Eng­lish Modal Auxiliaries,'' Natural Language and Linguistic Theory 3, 21-58.

Rochemont, Michael (1989) "Topic Islands and the Subjacency Parameter," Canadian Journal of Linguistics 34, 145-170.

Sadock, Jerrold (1974) Towards a Linguistic Theory of Speech Acts, Academic Press, New York.

Schmerling, Susan (1977) "The Syntax of English Imperatives,'' ms., University of Texas at Austin.

Stockwell, Robert P., Paul Schacter and Barbara Hall Partee (1973) Major Syn­tactic Structures of English, Holt, Rinehart and Winston. New York.

Zhang, Shi (1990) The Status of Imperatives in Theories of Grammar, Doctoral dissertation, University of Arizona.

Department of Information Arts Faculty of Information Shizuoka University 3-5-1 Johoku, Hamamatsu-shi Shizuoka 432-8011 e-mail: [email protected]