The Journal of Comparative Germanic Linguistics 1: 4354, 1997. c 1997 Kluwer Academic Publishers. Printed in the Netherlands.

43

The English complementizer of

In this squib, I will try to show that in some varieties of English, the preposition of has a use in which it is strongly similar to the familiar to of innitives: (1) John wants to leave.

Although the to of (1) and (2) differs in important ways from the for of (2), I will use the term complementizer to informally refer to both, as well as to the of in question:1 (2) John wants very much for Bill to leave.

This of appears in some varieties of English in sentences like (3): (3) John should have left.

For my (colloquial) English (but not only mine) this standard orthography is misleading: under normal intonation, it is highly articial for me to pronounce the third word of (3) as a full have, i.e., with an initial /h/ and an unreduced vowel. Rather, the vowel must be reduced, and the /h/ must not be pronounced, as reected in the alternative orthography given in (4): (4) John shouldve left.

This orthography conveys the idea that, along with its reduced pronunciation, ve is a form of the verb have. Although this may be an accurate characterization of some varieties of English, there is evidence that for others what is written as ve would more accurately be written as of:
To was called a complementizer in Rosenbaum (1967, 24ff.) I adopt this terminology because I think that innitival to has more in common with for than is currently thought (and than I thought in Kayne (1981)), but I will not pursue this here. Cf. also Lencho (1992).
1

44 (5)

RICHARD S. KAYNE

John should of left.

First, there are speakers for whom the third word of (3)(5) can clearly be pronounced with no /h/ and with the vowel of the preposition of as it is pronounced, for example, in (6):2 (6) Who are you thinking of?

There is thus some plausibility to the claim I will now make, namely that (5) faithfully reects the syntax, for at least some varieties of English.3 The modal is separated from the past participle by the preposition of, a prepositional complementizer, like the to of (1) and (2). Even more closely, this interpretation of (5) recalls the syntax of ought, which can occur followed by to at the same time that it displays modal behavior:4 (7) Ought John to leave?

Put another way, the of of (5) is to past participles as the to of (7) is to innitives. A second piece of evidence in favor of taking the third word of (5) to be a true of comes from the colloquial English phenomenon whereby the /v/ of of drops, leaving just a reduced vowel. Thus a bunch of grapes can be pronounced without the /v/; I will follow here a common orthographic convention and write this reduced of as an a attached to the preceding word: (8) a buncha grapes

From the point of view of the syntax proper, it might be more straightforward to write a bunch a grapes. On the other hand, this convention is also wellestablished for to in cases like They want to leave/They wanna leave. The relevance of (8) lies in the fact that the third word of (5) allows the same reduction:5
2 As noted by Cheshire, Edwards & Whittle (1993, 66), who stated earlier: Should of appears, then, to be widespread throughout Britain. Cf. also Labovs (1972, 151) black English vernacular example: it wouldnt of been nothing I could do. 3 For relevant discussion, cf. Curme (1931, 403, 473) and Mencken (1937, 4423; 1948, 366). Akmajian, Steele & Wasow (1979, 49) use of in cases like (5) but only to represent the pronunciation schwa+/v/. I use of to indicate a syntactic preposition and do not use it to indicate the schwa+/v/ pronunciation of the (nite) verb have. See the discussion of (13) below. 4 Cf. Pullum & Wilson (1977). 5 That the two reductions are similar was pointed out by Zwicky (1970, 328 n).

THE ENGLISH COMPLEMENTIZER OF

45

(9)

John shoulda left.

Taking the third word of (5) to be the preposition of permits an immediate generalization across (8) and (9). That this generalization is a signicant one receives support from an observation made by Jean-Yves Pollock (p.c.), namely that from the perspective of his British English both (8) and (9) are Americanisms (and not possible for him). Since the of of (5) (and its reduced variant in (9)) is parallel to the to of innitives, we would of course not expect to nd it in nite contexts, i.e., in place of nite have. Although there are varieties of English for which (5) accurately reects the syntax (and almost certainly others for which (3) accurately reects the syntax), no variety of English will have a nite of, just as none have nite to: (10) (11)
 The  The

kids of told a lie. kids to tell a lie.

It follows that in (12) the standard orthography is straightforwardly accurate: (12) a. The kids have told a lie. b. They never have told us the truth.

This nite have can be reduced, dropping the /h/ and reducing the vowel: (13) a. The kidsve told a lie. b. They neverve told us the truth.

But here further reduction seems impossible, or at least very difcult: (14) a. b.
 ?The  They

kidsa told a lie. nevera told us the truth.6

This suggests that although of can reduce to -a, have cannot,7 in which case the contrast between (14) and (9) becomes understandable but only if (9) truly has of. A third piece of evidence in favor of the presence of a true of in (5) and a reduced of in (9) comes from the colloquial construction given in (15):
Note the contrast with They would never a told us the truth, with a a reduction of of. Another case of reduction of of to a not adjacent to the modal itself is Should he a left that soon?; also He shouldnt a left so soon. 7 Reduction of nite have seems easier under (non-initial) inversion: ?Whya they been arrested?, ?Whata you done with them?.
6

46 (15)

RICHARD S. KAYNE

If you hadnta said that, : : :

It seems quite clear that the -a here is not a reduced form of have, given: (16)
 If

you hadnt have said that, : : :

Despite the impossibility of (16), a /v/ can be added back to the -a of (15),8 which should obviously now be taken to be a reduced form of of: (17) If you hadnt of said that, : : :

Summing up so far, there seems good reason to take there to exist a complementizer of introducing past participial phrases in a way strongly similar to the way in which to introduces innitive phrases. In further support of this parallelism, note that both of and to are compatible with VP-Deletion: (18) (19) a. b. a. b. If you hadnt of, you wouldnt of gotten in trouble. If you hadnta, you wouldnta gotten in trouble. If you want to, you can. If you wanna, you can.

Both can be followed by all: (20) (21) If they hadnt of/hadnta all left at the same time, : : : If they want to/wanna all leave at the same time, : : :

Both seem compatible with the following non-standard inversion possibility, discussed by Johnson (1988) for the participial case: (22) (23) %Should of/shoulda the kids left? (my choice of orthography) % Oughtta he leave?

Neither is compatible with emphatic so/too: (24) (25)

a. b.

should of/shoulda SO/TOO left. ought to/oughtta SO/TOO leave.  He claims to SO/TOO be happy.
 He

Although one might think that the ungrammaticality of (24) and (25) is due to the impossibility of so/too occurring within an innitive phrase, I nd the following to have a different status (in non-colloquial English):
8

When the participle begins with a consonant, the reduced form seems preferable here.

THE ENGLISH COMPLEMENTIZER OF

47

(26)

 ?He

claims to have SO told the truth.

Here, there is (marginally), following to, a true non-reduced non-nite have, followed by an emphatic so within an innitive. The examples so far considered for which I have suggested that the past participle is preceded by of or its reduced variant -a have all involved modals. Thinking of (26), one can ask whether of/-a can itself be preceded by to. At least in the case of ought, the answer seems clearly to be positive: (27) He oughtta of/a said so before.

In particular, the possibility of having -a after oughtta (= ought to) indicates the presence of of, as opposed to have.9 With a non-modal in place of ought, the judgment with -a seems less clear: (28) ??He seems to a made a mistake.

It might be that seem+innitive is not sufciently colloquial; alternatively, there might be a problem with having two complementizers in succession (to of ) after a non-modal.10 In my English, as I mentioned earlier, (3), repeated here as (29), is highly articial with a true have: (29) ( ) John should have left.

Somewhat surprisingly, however, its VP-deletion counterpart is natural with a full have:11 (30) John should have.

9 Randolph (1927, 4) states that in Ozark English one can nd could-a-of and might-a-of. Conceivably, these are really could/might-to-of, with to following these modals as it does ought. On the other hand, the Scots example Ah would uh could uh done it given by Miller (1993, 121) may well have two ofs. 10 The Icelandic complementizer adh is compatible with innitival V-raising when embedded under a non-modal control verb, but not when embedded under a modal verb cf. Thr ainsson (1984). See also the discussion of Norwegian below. Ken Sar points out a contrast between I would like to a/uv been hiding behind the curtain when he said that and It was wrong of me to  a/??uv been hiding behind the curtain when he said that; he suggests that the of complementizer may require an irrealis context and notes the relevance of ?He might claim to have SO of told truth, but I doubt him (vs. (26)) and If John hasnt ( of) left, we might win this game. 11 However, I dont think the following is possible (cf. (18a)):  If you hadnt have, : : : , suggesting that in the colloquial English that allows (18), the non-nite auxiliary have might be entirely absent.

48

RICHARD S. KAYNE

Also acceptable to me is the following, with stress on have: (31) I should send it off. No, you should HAVE sent it off.

The generalization seems to be that (in my English) non-nite have is possible as an auxiliary12 only in contexts of the kind in which the auxiliary do is possible. In other words, (29)(31) is paralleled by: (32) (33) (34)
 John

does leave early. John does. John DOES leave early.

Thinking of Chomskys (1995, 140) idea that (32) is excluded because auxiliary do is possible only when necessary, one might relate the possibility of (31) to the impossibility of (35) (similar, perhaps, to (24) and (25)): (35) I should send it off.  No, you should OF sent it off.

Problematic is the fact that (36) seems possible alongside (30):13 (36) He really should of/shoulda.

But if that problem can be solved, the impossibility of (29) in the relevant English might follow from the possibility of (37): (37) John should of/shoulda left.14

To say that non-nite have is possible only when necessary, i.e., when of/-a is not available, would not be to say that it is always possible then. For example, neither is possible with -nt: (38) (39)
 John

a. b.

should ofnt/shouldant left. should havent left.  He claims to havent told any lies.
 John

Here there is of course a contrast with auxiliary do: (40) John didnt leave.

12 Non-auxiliary have does not show the restriction under discussion: John should have more time, John should have his car repaired. Note also that the auxiliary having is possible (although perhaps not in colloquial English): Despite having made a mistake, : : : 13 Cf. Wood (1979, 371, 373). 14 This economy calculation would have a global character to it, however.

THE ENGLISH COMPLEMENTIZER OF

49

On the other hand, (38) patterns as expected with to: (41)

want tont/wannant leave.

A common account of (38), (39), and (41) that distinguishes them from (40) while leaving intact the parallel under discussion between do and the non-nite have would attribute the ungrammaticality of the rst three to a property of -nt itself, namely the requirement that nt be adjoined to by a nite head.15 The of/-a complementizer that I have argued can be found preceding (active) past participles in a way strongly parallel to the more familiar to that occurs with innitives is also reminiscent of the innitival complementizers de and di of French and Italian, respectively.16 Exactly like de and di, of is identical in form to the preposition that occurs in possessive and related constructions in the three languages.17 Given the existence of the complementizers de and di in French and Italian, the existence of a complementizer of in English is not an exotic property of English. What does remain specic to English is the use of an overt complementizer with active participial phrases under modals. Part of that specicity apparently lies in the fact that (some) English is able to embed participial phrases directly (abstracting away from the complementizer itself) under modals, without the intermediary of an auxiliary verb have whereas we normally think of English modals as requiring their complement to be innitival (usually without to, apart from ought), rather than participial. The present proposal claims that that normal way of thinking of English is mistaken (at least for some varieties). At least some English modals can also have their complement be a participial phrase (with the complementizer of ). In this respect English does differ from French and Italian. But it does
15 On niteness, cf. Lapointe (1980, 237), and Zwicky & Pullum (1983, 507). On adjunction to nt, cf. Kayne (1989). This niteness property of -nt may be related to what one nds in other languages, as noted by Payne (1985, 240); cf. Zanuttini (1996). Scots has a bound negation -nae that resembles -nt in being incompatible with contracted auxiliaries (a property that may have a counterpart in Serbian/Croatian, as suggested by Wilder (1996, 174)) and in being excluded from contexts of VP-negation; cf. Brown (1991, 80, 84). (It differs from -nt in not inverting with an auxiliary (Brown 1991, 80).) Brown p. 78 mentions (based on a paper by MacAfee) a variety of Scots in which a modal followed by -nae can occur following used to: He used to widnae let me up the brae (it used to be the case that he wouldnt let me go up the hill). Either -nae in this variety is not subject to the niteness requirement, or the embedded modal here is nite, or used to is adverbial, as suggested by Labov (1972, 56) for useta in BEV (which has, p. 57, : : : to dont throw bottles and rocks). 16 For some relevant discussion of de and di, cf. Kayne (1991, sect. 2.2). The Spec, CP proposal made there needs to be rethought. 17 For relevant discussion, cf. Kayne (1994).

50

RICHARD S. KAYNE

not differ from the Scandinavian languages, which robustly show (certain) modals embedding active participial phrases (with no overt complementizer). An example from Taraldsens (1984) discussion of Norwegian is: (42) Vi skulle gjort det fr. we should done it before

Taraldsen states that relative to his discussion Swedish is similar to Norwegian. Faroese examples are given by Lockwood (1977, 141ff.) (cf. Einarsson (1945, 163ff.) on Icelandic). In Faroese, at least, a modal can even be followed by the participial form of have (Lockwood 1977, 143):18 (43) Eg skuldi havt gjrt : : : I should had done : : :

I take these Scandinavian facts to substantiate the plausibility of English allowing an active participle to be embedded under a modal directly.19 Although I will not pursue the Norwegian-English comparison in detail, it is perhaps noteworthy that the Norwegian counterpart of the questionable (28) above is ungrammatical without (non-nite) have. To the extent that English is signicantly similar to Scandinavian here, we want the analyses to be substantially similar. Let me suggest the following: (44) The Scandinavian construction represented by (42) contains a phonetically unrealized complementizer comparable in other respects to English of.20

Conversely, it might be plausible to conclude, from the fact that many modals in Icelandic take innitival complements preceded by an overt complemen18 This Faroese construction has a parallel in some Irish English. Harris (1993, 159) gives the example: I shoulda haven killed him (= I should have killed him). 19 Taraldsen notes that in Swedish the active past participle can be phonologically distinct from the passive participle. The former is sometimes called the supine in discussions of Scandinavian. I leave open the question of why English does not allow the word-for-word equivalent of (42). Both (42) and Swedish (embedded) sentences with a missing nite have (cf. Andersson & Dahl 1974) bear on Kayne (1993); for an interesting proposal accounting for the Swedish restriction to embedded contexts and for the fact that Norwegian does not allow nite have to be missing, see Taraldsen (1984). 20 Cf. perhaps the fact that at least some Scandinavian has zero instead of of in phrases like the month of May, a pound of corn; see Holmes & Hinchliffe (1994, 451). See also the fact that English differs from Scandinavian in having both of and s in a friend of Johns, as well as differing from at least some in having of+lexical DP in all of my friends; see Strandskogen (1989, 129) and Holmes & Hinchliffe (1994, 452). Similarly, English, but not Scandinavian, has a complementizer for that licenses a lexical subject.

THE ENGLISH COMPLEMENTIZER OF

51

tizer adh,21 that all English (and Icelandic) modals have their innitival complements introduced by a complementizer, which is sometimes phonetically unrealized in Icelandic and usually so (except for ought) in English. If this converse suggestion is correct, then modals in English are similar to the auxiliaries have and be (in English and other languages) in taking complements with more internal structure than was envisaged in the early years of generative syntax.22 My conclusion that (some) English has a complementizer of between modal and past participle might appear to be weakened by the following observation: in some cases English does have a non-nite have, as discussed above. This non-nite have can differ in syntactic behavior from non-nite be. When it does so differ, non-nite have can pattern instead with of. For example, Curme (1977, 473) says : : : we often hear in popular speech : : : if they had have said so : : : in other words, a have (frequently in the contracted form of a or of ) is often inserted : : : . This construction is presumably closely related to (15). The two pronunciations given as frequent by Curme t in straightforwardly, but not the pronunciation with have, which he seems to say is sometimes heard. Similarly, the construction from Johnson (1988) cited in (22) above is given by him with the orthography have, suggesting that for some speakers the auxiliary verb have is possible here: (45) %Shouldnt have Pam remembered her name?

Along with this example, Johnson gives the following, with non-nite be, as impossible: (46)
 Shouldnt

be Pam remembering her name?

Related to this, I think, is the fact that I have the following contrast, with respect to the position of all: (47) They should all speak up. No, they should (all) HAVE (all) spoken up.

Non-nite have is possible for me with a following participial phrase when stressed (cf. (31) above); it can be either preceded or followed by all.
21 Cf. Thr ainsson (1994, 186). Faroese seems to have fewer (cf. Lockwood 1977, 146) but has a notable phenomenon of attracted supines (p. 141) yielding cases of a supine preceded by the innitival marker at; cf. perhaps Anward (1988). 22 Cf. for example Chomsky (1957, 39). The text discussion recalls Ross (1967) and Pullum & Wilson (1977). On have and be, cf. Kayne (1993). How auxiliary do ts in here remains to be determined.

52

RICHARD S. KAYNE

Stressed non-nite be is also possible, but the position of all is not the same:23 (48) They should all speak up. No, they should (all) BE ( all) speaking up.

The fact that all can follow have but not be puts have in a group with of/-a, given (49): (49) They should of/shoulda all spoken up.

Although it would go beyond the bounds of this squib, the solution that comes to mind is the following: have resembles complementizer of more than be does because have, but not be, is a form of complementizer. That is so since have results from the amalgamation of be with an abstract prepositional complementizer.24

References
Akmajian, Adrian, Susan M. Steele & Thomas Wasow: 1979, The Category AUX in Universal Grammar, Linguistic Inquiry 10, 164. Andersson, Anders-B orje & Osten Dahl: 1974, Against the Penthouse Principle, Linguistic Inquiry 5, 451453. Anward, Jan: 1988, Verb-Verb Agreement in Swedish, McGill Working Papers in Linguistics. Special Issue on Comparative Germanic Syntax, 134. Brown, Keith: 1991, Double Modals in Hawick Scots, in P. Trudgill & J.K. Chambers (eds), Dialects of English. Studies in Grammatical Variation, Longman, London, pp. 74103. Cheshire, Jenny, Viv Edwards & Pamela Whittle: 1993, Non-Standard English and Dialect Levelling, in J. Milroy & L. Milroy (eds), Real English. The Grammar of English Dialects in the British Isles, Longman, London, pp. 5396. Chomsky, Noam: 1957, Syntactic Structures, Mouton, The Hague. Chomsky, Noam: 1995, The Minimalist Program, MIT Press, Cambridge, Massachusetts. Curme, George O.: 1977 (1931), A Grammar of the English Language. Volume II: Syntax, Verbatim, Essex, Connecticut. Dikken, Marcel den: 1995, Copulas, unpublished ms., Vrije Universiteit Amsterdam/HIL. Einarsson, Stef an: 1945, Icelandic. Grammar, Texts, Glossary, The Johns Hopkins Press, Baltimore. Harris, John: 1993, The Grammar of Irish English, in J. Milroy & L. Milroy (eds), Real English. The Grammar of English Dialects in the British Isles, Longman, London, pp. 139 186. Judging by Pollock (1989, 382), at least some British English might allow (48). Like (47) vs. (48) for me is (26) vs. the worse  He claims to be SO telling the truth. Cf. also Johnson (1988, 159) on VP-fronting and VP-ellipsis differences between non-nite have and non-nite be. 24 Cf. Kayne (1993) and, with some important differences, den Dikken (1995). This issue might be relevant to Zwarts (to appear) discussion of the fact that Dutch does not allow  : : : hebben moet gelezen (: : : have (inn.) modal read (past participle)).
23

THE ENGLISH COMPLEMENTIZER OF

53

Holmes, Philip & Ian Hinchliffe: 1994, Swedish. A Comprehensive Grammar, Routledge, London. Johnson, Kyle: 1988, Verb Raising and Have , McGill Working Papers in Linguistics. Special Issue on Comparative Germanic Syntax, 156167. Kayne, Richard S.: 1981, On Certain Differences between French and English, Linguistic Inquiry 12, 349371. Kayne, Richard S.: 1989, Notes on English Agreement, CIEFL Bulletin (Hyderabad, India), 4167. Kayne, Richard S.: 1991, Romance Clitics, Verb Movement, and PRO, Linguistic Inquiry 22, 647686. Kayne, Richard S.: 1993, Toward a Modular Theory of Auxiliary Selection, Studia Linguistica 47, 331. Kayne, Richard S.: 1994, The Antisymmetry of Syntax, MIT Press, Cambridge, Massachusetts. Labov, William: 1972, Language in the Inner City: Studies in the Black English Vernacular, University of Pennsylvania Press, Philadelphia. Lapointe, Steven: 1980, A Lexical Analysis of the English Auxiliary Verb System, in T. Hoekstra, H. van der Hulst & M. Moortgat (eds), Lexical Grammar, Foris, Dordrecht, pp. 215254. Lencho, Mark: 1992, Evidence that To is a Complementizer, abstract of paper presented at the University of Troms. Lockwood, W.B.: 1977, An Introduction to Modern Faroese, Froya Sk ulab okagrunnar, T orshavn. Mencken, H.L.: 1937, The American Language. An Inquiry into the Development of English in the United States, fourth edition. Alfred A. Knopf, New York. Mencken, H.L.: 1948, The American Language. An Inquiry into the Development of English in the United States. Supplement II, Alfred A. Knopf, New York. Miller, Jim: 1993, The Grammar of Scottish English, in J. Milroy & L. Milroy (eds), Real English. The Grammar of English Dialects in the British Isles, Longman, London, pp. 99 138. Payne, John R.: 1985, Negation, in T. Shopen (ed.), Language Typology and Syntactic Description. Volume I. Clause Structure, Cambridge University Press, Cambridge, pp. 197 242. Pollock, Jean-Yves: 1989, Verb Movement, Universal Grammar, and the Structure of IP, Linguistic Inquiry 20, 365424. Pullum, Geoffrey K. & Deirdre Wilson: 1977, Autonomous Syntax and the Analysis of Auxiliaries, Language 53, 741788. Randolph, Vance: 1927, The Grammar of the Ozark Dialect, American Speech 3, 111. Rosenbaum, Peter S.: 1967, The Grammar of English Predicate Complement Constructions, MIT Press, Cambridge, Massachusetts. Ross, John Robert: 1967, Auxiliaries as Main Verbs, in W. Todd (ed.), Studies in Philosophical Linguistics. Series 1, Great Expectations Press, Evanston, Illinois. se-Berit & Rolf: 1989, Norwegian. An Essential Grammar, Routledge, Strandskogen, A London. Taraldsen, Knut Tarald: 1984, Some Phrase Structure Dependent Differences between Swedish and Norwegian, Working Papers in Scandinavian Syntax 9, 145. Thr ainsson, H oskuldur: 1984, Different Types of Innitival Complements in Icelandic, in W. de Geest & Y. Putseys (eds), Sentential Complementation, Foris, Dordrecht, pp. 247255. Thr ainsson, H oskuldur: 1994, Icelandic, in E. K onig & J. van der Auwera (eds), The Germanic Languages, Routledge, London, pp. 142189. Wilder, Chris: 1996, English Finite Auxiliaries in Syntax and Phonology, ZAS Papers in Linguistics 6, Zentrum f ur Allgemeine Sprachwissenschaft, Sprachtypologie und Universalienforschung, Berlin, pp. 166191.

54

RICHARD S. KAYNE

Wood, Winifred J.: 1979, Auxiliary Reduction in English: A Unied Account, in P.R. Clyne, W.F. Hanks & C.L. Hofbauer (eds), Papers from the Fifteenth Regional Meeting. Chicago Linguistic Society, University of Chicago, Chicago, Illinois, pp. 366377. Zanuttini, Raffaella: 1996, On the Relevance of Tense for Sentential Negation, in A. Belletti & L. Rizzi (eds), Parameters and Functional Heads. Essays in Comparative Syntax, Oxford University Press, New York, pp. 181207. Zwart, Jan-Wouter: to appear, Verb Clusters in Continental West Germanic Dialects, in J. Black & V. Motapanyane (eds), Microparametric Syntax: Dialect Variation in Syntax, J. Benjamins, Amsterdam. Zwicky, Arnold M.: 1970, Auxiliary Reduction in English, Linguistic Inquiry 1, 323336. Zwicky, Arnold M. & Geoffrey K. Pullum: 1983, Cliticization vs. Inection: English nt, Language 59, 502513.