Structures of Tense and Aspect

Elly van Gelderen

Arizona State University

In this paper, I provide an analysis along Minimalist lines (e.g. Chomsky 1995) for parts of the English auxiliary system. I focus on `perfect' have but indicate how progressive and passive can also be incorporated. The English present perfect construction is not perfective in aspect. It expresses past tense with an added present relevance element. I represent the English present perfect by using Reichenbach's (1947) representation of tense as points in time. I argue that the structural representation incudes three positions for tense: the point of Speech occupies C; the point of Reference I and the point of Event occupies ET (a new functional category). There is perfective aspect in English which, based on work by Sanz 1996 and Jelinek 1997, I represent as a Transitive Phrase. Aspect is argued to be structural, e.g. to follow from the presence of a Preposition or a certain kind of object NP (cf. Verkuyl 1993).

In the first section, I provide some background on functional categories and features. In section two, an analysis for the perfect as well as for other tenses is provided by linking certain tenses to certain functional categories. I show that the tense of have is independent from that of its participle in Modern English and indicate this independence structurally. I also show that have is a clitic, but whether it is or not does not change the analysis of its tense. In section three, I claim that there is an additional projection (TrP) responsible for telicity or completion, i.e. for aspect. Aspect is unrelated to have and is represented through an abstract preposition. In section four, I discuss the representation of progressive and passive auxiliaries, again through separate Functional Categories.

Chomsky (1992), based on Pollock (1989) and Chomsky (1989), i.e. `early' Minimalism, argues that all Case is checked (rather than assigned) in a Spec-Head relationship. For this purpose, several functional categories are introduced, such as AGRs and AGRo in (1). NPs move to the Specifier positions and verbs move to the Head positions. Nominative Case is checked against AGRs and objective is against AGRo:

1. AGRsP

Spec AGRs'

AGRs TP

Spec T'

T AGRoP

Spec AGRo'

AGRo VP

Verbal agreement is checked in a Head-Head relationship between V and AGR after the verb incorporates into the AGR Head as in (2). The person and number features of the head are given `content' by the NP (just as the verb gives `content' to the Case: if in AGRs, nominative; if in AGRo accusative):

2. AGRP

Spec AGR'

AGR XP

V AGR .......

The checking of Case and agreement occurs either overtly or covertly, depending on whether the features in the Functional Head are strong or weak. Overt checking of the NP takes place in a Spec-Head relationship as before SPELL-OUT (or at s-structure in earlier frameworks); covert movement will mean that the element must wait till Logical Form (LF) to check its features because this is `cheaper'.

Other functional categories are introduced. Thus, T(ense)P accompanied by V- and N-features is also included. Categories such as ASP(ect)P, VoiceP, Perf(ect)P, Num(ber)P, PersonP and others are also possible (cf. Rivero 1990; Marácz 1991; Mitchell 1993; Ritter 1991) and these will be elaborated in this paper where relevant. I assume a functional category is only activated if an element is selected from the lexicon that projects into that particular functional category. Features are both present in FCs and connected to lexical items. In Chomsky (1995), they are defined as intrinsic or optional and Interpretable or non-Interpretable. The V- and N-features of Chomsky (1992) are reformulated in Chomsky (1995) as categorial features (only D-features are actually discussed), and Case and phi-features are added. All movement at LF (i.e. covert) caused by weak features of a target, is replaced by feature movement. I will not elaborate on each of these changes.

There are many types of features: semantic (e.g. abstract object), phonological (e.g. the sounds), and formal (Chomsky 1995: 230ff; 236; 277ff). The formal ones are relevant to syntax and are divided into intrinsic or optional; and into Interpretable or non-Interpretable. The intrinsic ones are "listed explicitly in the lexical entry or strictly determined by properties so listed" (Chomsky 1995: 231) and include categorial features, the Case assigning features of the verb, and the person and gender features of the noun. Optional features are added arbitrarily and are predictable from UG Principles (e.g. nouns need Case). They include the tense and agreement features of verbs and the number and Case features of NPs.

The "much more important distinction" (p. 277) is that between interpretable and non-Interpretable. The Interpretable ones are relevant for interpretation at LF and include categorial and nominal phi-features. They are not deleted or erased after they are checked because they are relevant to the interpretative component. Non-Interpretable features are deleted and they involve the Case features of NPs and verbs and the phi-features of verbs. There are a number of reasons behind the distinction. Some features (e.g. phi-features of NPs) remain visible after checking and hence cannot be deleted. This is the reason an NP can move cyclically and provide the phi-features along the way (Chomsky 1995: 282). This is not true for the non-Interpretable Case. Once Case has been checked by an NP that same NP cannot move to check Case elsewhere.

The reason for abandoning V- and N-features in favor of categorial, Case and phi-features is that the Extended Projection Principle effects (i.e. that clauses have structural subjects) are accounted for by means of a strong D-feature in I (Chomsky 1995: 232). This is necessary since Minimalist trees do not automatically project a Specifier position and Spec IP must somehow be present. Hence, the assumption that a D-feature exists in I. Wh-movement is triggered by a D-feature in C rather than by a [wh]-feature. The argument that it is D-features that trigger movement comes from expletives. If the expletive there in (3) was present to check the Case features, the Case features of the postverbal four arguments would not be attracted. As a result, the non-Interpretable Case features of the NP would remain unchecked and the sentence would not be well-formed (the term used for well-formedness is convergence):

3. There are four arguments in her head.

If the expletive was present to check the phi-features, the Interpretable plural phi-features of the noun would not be attracted to I(nflection) and again, (3) would not converge. Since (3) is grammatical, there is only inserted to check the categorial features.

Chomsky notes a difference between functional categories such as D(eterminer), I(nflection), or T(ense) and C(omplementizer) on the one hand and AGR(eement with the)s(ubject) and AGR(eement with the)o(bject) on the other hand. AGR lacks phi- and Case-features since V and T provide the latter by moving to AGR. In addition, AGR never has lexical content, unlike D, I and C. The evidence that AGRs and T both exist comes from the analysis for an Icelandic construction. Chomsky argues that for English IP suffices and that the Case to the object can be dealt with through features present in the light verb v in (4) rather than through AGRo. In this paper, I assume the same, namely that AGR is only activated in a language if structural evidence exists for it (i.e. in Icelandic but not in English).

As a summary, an instance is provided of how the features work in a Minimalist system. In (4), an NP (with Interpretable categorial and phi-features, but with non-Interpretable Case) moves to Spec IP overtly to check the strong categorial feature ([D]) in I. The phi-features and Case move along and are checked in due course. For English, assuming the non-listed, categorial V-features are weak, this is the only overt movement necessary. The phi-features of the verb or Auxiliary are attracted to I at LF. The features of the object NP move to the specifier of v and the main verb adjoins to v. The result is that all the non-Interpretable features are checked:

4. IP

. I'

I VP

[D] . .

[Case] v VP

[D] NP V'

Zora V NP

[phi] [phi] [phi]

[Case] [Case] [Case]

reads books

In (4), the Interpretable features are the phi-features of the NPs; the non-Interpretable ones are the features of the targets (C, T and v), the Case features of the NPs and verbs, and the phi-features of the verb.

Minimalist ideas relevant for this paper are (a) movement is triggered by the need to check features, (b) some features are Interpretable (and can check features multiple times), but some are not. There are a number of matters not discussed in Chomsky (1995). Even though categorial features are introduced, it is only the category of the noun, i.e. the D-feature, that is addressed. If categorial features are the only ones that can be strong and trigger overt movement, V-features must be present as well. Otherwise, overt Verb-movement would not take place, as it does in a number of languages. Not much is said about the checking of tense features either and I will argue that tense features of lexical items are interpretable and that is how a tense interpretation is arrived at at LF. The tense features associated with Functional Categories such as I and ET are non-Interpretable. I give instances how both lexical elements are connected with features as well as the FCs.

2 The English Present Perfect

In this section, I discuss the features connected to have and the participle. In 2.1, I quote some literature about the meaning of the English present perfect and provide a structural representation which depends on Reichenbach's (1947) and Hornstein's (1990) insights into the present perfect. In 2.2, I argue that the tense of have is independent from that of the participial form, and show that have is a clitic but this claim is only tangentially related to the analysis of its tense.

Perfect have or -(a)v in English is classified as perfective (Quirk et al 1972: 91). However, this does not mean it achieves a goal or that it can be seen as telic or resultative, as the strangeness of (5) and (6) indicates:

5. I've read.

6. I've gone.

Sentences (5) and (6), if uttered at all, are interpreted as an activity that was done without reaching a particular goal. The same is true for other languages, e.g. in Dutch, the simple past and perfect as in (7) and (8) indicate an unfinished activity:

7. Ik las.

I read, `I was reading'.

8. Ik heb gelezen.

I have read, `I've been reading'.

In what follows, I assume that the English present perfect is indeed past tense, i.e. that the event occured in the past, as the text from the Wasserman cartoon in (9) indicates:

Jespersen (1931: 193) says: "in he has collected much evidence against her nothing is said about the time he collected it, the only thing said being that the act of collecting is finished at the present moment [my emphasis]". Hence, the tense is past but there is present relevance (cf. also Comrie 1976 [1987]: 52ff). The notion of perfectivity or telicity, i.e. reaching a goal, is not connected to have in (5) and (6) but is present in a separate functional category, to be elaborated on in section three.

Thus, the English present perfect by itself indicates a past but with present relevance, as comes out in a Reichenbach (1947: 290ff) schema as in (10), where E is the point of the Event, S the point of Speech and R is the Reference Point. A present perfect would have an Event in the past, but with the Reference at the time of Speech as in (10). In a past perfect, the event precedes the reference point which in turn precedes the time of speech as in (11). In a simple past as in (12), both the event and the reference points are in the past:

10. E___S,R

11. E__R__S

12. E,R__S

Hornstein (1990: 194) puts it as follows: "perfective tenses have three individuated points of temporal interest [S, E, and R]". In my analysis below, I make use of this and argue that the three relevant points in time are structurally present: the time of the Event, of Speech and of the point of Reference. Thus, I link Functional Categories with tenses whereas Hornstein (1990: 111ff.) maps morphemes onto tenses.

As is well-known, the mere use of have or -av does not suffice to express perfect in Modern English: connected to have is the ending on the verb that follows (i.e. the `affix-hop' phenomenon). Below, I show that in older varieties of English (cf. (19) to (25)) these two phenomena are in a way independent of each other, and on occasion only one of the two is expressed. I will now formulate a tree structure for a present perfect compatible with these observations as well as with Reichenbach's (1947) theory outlined above. Reichenbach argues there are three points in time, S, R and E which sometimes coincide. I will argue that the point of Speech is in C (in a main clause always present); the point of Reference in I; and the point of Event in ET (short for `Event Time'):

13. CP

. C'

C IP

[pres] I'

They I PfP

[pres]. Pf'

(ha)ve Pf ETP

t . ET'

ET VP

[past] V

gone t

S R E

In (13), (ha)ve originates in the Perfect Phrase and moves to I to check its present tense features. The ETP has variable tense features depending on the time of the Event, here past. In (13), gone checks its past tense features in ET. Since the time point of I is the same as in C, there is present relevance even though the event happened in the past, indicated by the past participle that moves to ET. The tense of S is situated in C, adapting an idea from Enç (1987: 643) that anchoring is a necessity and that "[i]f Comp does not have a governing category, it is anchored if and only if it denotes the speech time". Hornstein (1990: 14) argues likewise that S "is a deictic element anchored within the discourse situation, often to the moment of Speech". The structure of ET is detailed in (14), where the past features of ET are non-Interpretable but those of gone are Interpretable and survive till LF (this is relevant for the simple past discussed below):

14. ET

gone ET

[past] [past]

Thus, have does not check the past tense features in ET. Have, after checking its present tense features against I (again the features in I are non-Interpretable but those of have are Interpretable), moves to C covertly to check the tense features there. Have is `taken out' of the lexicon as an element with present features that are checked in I; gone has past features and checks those in ET. In cases where have checks other features in addition to present, it is specified in the lexicon differently (e.g. the finite has needs to check person and number). As I will show below, in stages of English such as in (19) and (20), ET occurs without have and the past participle first checks its past tense in ET before moving on to check R in I.

In the case of a simple present, a simple past and a simple future, R plays less of a role. However, in accordance with Reichenbach (1947: 290; 297), I assume (15) is correct for the simple present; (12) above, repeated as (16), for the simple past; and (17) for the simple future. Hornstein (1990: 111) accounts for the relation of R to either E or S by mapping rules, for instance, (15) is arrived at by a morpheme rule "present morpheme: associate S and R"; (16) by "past morpheme: R removed to the left of S"; and (17) by "future morpheme: R removed to the right of S". I will account for the relation of R to the other tenses by arguing that, in simple tenses, the tense in I is the same as that of ET. The verb checking ET has Interpretable tense features and, after checking ET, checks the tense of I. Thus, the structure of a simple past is as in (18):

15. __S,R,E__

16. E,R__S

17. S__R,E

18. CP

. C'

C IP

. I'

[pres]She I ETP

[past]. ET'

ET VP

[past] V

went

S R E

The verb went in (18) moves to ET and from there to I, and the interpretation as in (16) follows. This also works for present and future in that the verb checking ET will be present and future respectively. Since S is present, a present verb such as am moving to I would check the present tense of R and (15) would be the result. In the case of a future event, the future of the ET would be checked by the verb before it moved to I to check future. The result is that both R and E are future, as opposed to S which is present.

Some comments on the actual interpretation of tense at Logical Form. As I mentioned, I assume that the tense features of e.g. have and gone are Interpretable but that those of the Functional Categories are not. As Chomsky (1997) puts it, the latter are not interpreted at either the semantic (i.e. LF) or phonetic interface but they force movement. I assume that the features of C (Speech time) are Interpretable as well and that the tense of have (i.e. of Reference) and that of gone (i.e. of Event) are specified relative to the tense of C. The tense of the simple present, past and future are similarly arrived at.

To conclude 2.1, I argue that the English present perfect construction consists of present tense in I and past tense features in ET as in (13). This explains that present relevance is connected to past tense. This account is also valid for simple tenses as I show, even though that is not the focus of my paper. In the next subsection, I make three points: (a) that have and the past participle have historically been separate, as is suggested by the analysis in (13); (b) that the features are not always separately expressed; and (c) that the auxiliary have is a clitic.

I first examine the form of have and its relation to the past participle. Have is optionally present in certain stages and varieties of English as in (19) and (20). If this construction is a present perfect, it indicates that have is not essential but that the presence of a participle to check tense in I and ET suffices:

19. Chaucer, Prologue 181

We been acorded ...

20. Paston Letters I, 122A (1467)

and for asmuche as we been enfourmed ...

Just as sometimes have is optional, there are stages of English where the past participle ending does not appear as in (21) to (25). This use is said to be the common southern variant (cf. Partridge 1969: 41) in the fourteenth century, for instance, in Chaucer as in (21) and (22), and in 15th century legal Statutes as in (23). In the 15th century Paston Letters, haue be as in (24) occurs 41 times whereas haue been only 6 times. The (perhaps more formal) Chancery Letters from about the same time show understande and be in (25) as well rather than the participial forms. The Northern form is been and it supplants be, especially after Caxton. In Shakespeare, forms of haue are always followed by a past participle fully inflected, even though this is reduced to bin on occasion:

21. Chaucer, Miller's Tale, 3739

What have I do?

If I so ofte myghte have ywedded be.

23. Statutes of the Realm 483

it hath bee dailly used

24. Paston Letters I, #48 (1454)

the sayd ryottys haue be done ...

Thus, have and the past participle are in a way independent of each other. This is represented in (13). I now turn to the status of have.

The auxiliary have, when it only functions as a `perfect', appears as a reduced element in spoken speech as in (26) or as of in informal writing as in (27). When have is a main verb, this reduction is not possible as (28) shows. Rather, (29) is grammatical. In accordance with e.g. Zwicky & Pullum (1983), but see also Kaisse (1983), I will argue that the auxiliary -ve is a clitic and needs to incorporate into another head (should in (26)):

26. I shoulda/shouldav gone.

27. I should of gone.

28. *He shoulda books in his office.

29. He should have books in his office.

Zwicky & Pullum's (1983: 503-4) arguments distinguishing clitics from agreement markers are well-known. They involve criteria such as the selection of a host (clitics attach freely), morphological and semantic ideosyncracies (less common for clitics), and attachment to material already containing clitics (clitics can attach to other clitics). All these point to -av being a clitic. For instance, clitics attach to personal pronouns as in (5) above, full nouns as in (30) below, relative pronouns, complementizers, and auxiliaries as in (26). They also attach to other clitics as in I'll'av eaten by then. Kayne (1997: 43) goes much further and argues non-finite have is a complementizer introducing a participle and "would more accurately be written as of". He thinks there is a difference between the reduction in (27) and that in (30). I will not go that far and will treat both (27) and (30) as having reduced forms of the auxiliary have:

30. The kids've told a lie.

(Kayne 1997: 45)

If the auxiliary is indeed cliticized to should in (31), one would not expect preposings of have without should as in (32) since should've, the non-preposed form, forms one word. This prediction is borne out. However, a preposed main verb have as in (33) is grammatical because should and have do not form one word. In addition, there cannot be a replacement by do so when this splits up the clitic from the element to which it is adjoined, as shown in (34). This ungrammaticality is again expected if have is part of may:

31. They should've washed the dishes.

32. *...and [have washed the dishes] they should t.

33. ...and have headaches they should by now.

When have needs to check other features such as person in addition to tense, these appear. In She's gone, third person present is visible. For all persons and numbers except third singular, the -v suffices; with third person singular, an -s appears.

In forms of English that are less prescriptive, e.g. letters from the 15th century, a(ve) occurs extremely freely, as in (35) to (38). This may be an indication of clitichood, even though one must be careful equating written with spoken text. As in Modern English, the reduction only occurs with perfective have, not with main verbs. For instance, in all the 16 instances of xuld(e) followed by haue in The Paston Letters, the latter is a main verb; in the 10 instances where an auxiliary follows the modal, the form is reduced to a as in (36) and (38). In the late 16th century Shakespeare as in (39) and (40), endings may be reduced as well. Checking the Middle English section of the Helsinki Corpus, I find this reduction in all genres, e.g. sermons as in (41) and drama as in (42). The early 14th century Cursor Mundi has many forms as in (43):

35. Paston Letters I, #399 (1481)

36. Idem, I, #131 (1449)

it xuld a be seyd

`It should have been said'.

37. Idem, I, #176 (1464)

3e wold a be plesyd

`You would have been pleased'.

38. Idem, I, #205 (1469)

there xuld not a be do so mykele

`There should not have been done so much'.

39. Shakespeare, Hamlet

Ophelia: So would I ha done

40. Shakespeare, 2 Henry 4, 720

Iust: I know you ha' practised

41. Capgrave's Sermon 7537

Deuel ha hold him

42. Mankind, 21508

Ha done sere prophete

43. Cursor Mundi 16251

And ha men neuer ben sa bald

Looking at have in earlier stages, its status of reduced form in modern English is not unexpected. Auxiliary have has always been a reduced form, but due to prescriptive pressure, it continues to be written as an independent form.

There is a difference between Modern English and older texts in that there are a number of instances where person and number are not checked, as in (40) to (43). (Perhaps AGRs is optional, since C is the locus of Case and agreement in Old and Middle English, cf. van Gelderen 1993). In (35), the situation is as in Modern English: ave not only checks perfective but also plural and that is the reason ave appears; when only the perfective is checked, as in (36) to (39), the ending is not required.

The fact that perfect have is a clitic explains the data as in (26), (28), and (32) to (34) above but there are two instances where have is not a clitic, namely in fronted constructions such as (44) and in negative ones such as (45). Either one can argue there are two kinds of perfect auxiliary, a clitic and a full form; or one could argue that ha- is an expletive, inserted in cases where there is no appropriate host for the clitic as in (44). I'll opt for the latter, but nothing hinges on this. When the negative n't attaches as in (45), the expletive also appears since the negative affix (see Zwicky & Pullum 1983) cannot attach to a clitic:

44. Has she left.

45. I haven't done that yet.

In the present section, I have provided an analysis of the present perfect in English, as well as given some comments on the forms in which the auxiliary appears. In the next section, I examine how aspect is expressed, i.e. telicity or completed actions.

In this section, I examine how completed action, i.e. perfective aspect, can be expressed through either a locative element (preposition or locational NP) or an NP object. My use of a Tr(ansitive) Phrase is based on work by Sanz (1996) and by Tenny (1987; 1994). I ignore the tense structure for convenience but the instances of past tense as in (46) below would have a tense structure as in (18) above.

It has been known, since e.g. Streitberg (1891), that in languages such as Dutch, slapen `sleep' is imperfective, but inslapen `fall asleep', with a prepositional prefix, is perfective (cf. also Abraham 1996; Brinton 1988; Raith 1951, to name but a few). The same is true in Slavic languages (cf. Comrie 1976 [1987]). In Old English, "[t]he perfective aspect" is "often indicated by means of verbal prefixes" (Mustanoja 1960: 446; see also Quirk & Wrenn 1957: 114ff). For instance, þurh `through', of, and to, as in þurhbrecan `break through', ofsceotan `shoot off', tobrecan `break up', render an imperfective verb perfective by specifying the goal. Even in Modern English, eat is imperfective whereas eat up is not. One could express this relationship by having the prefix occupy a functional head responsible for completed aspect.

There is also a relation in many languages between perfectivity and a particular Case (e.g. Kiparsky 1996 for Finnish; Abraham 1996 for Old High German). This would again be accounted for by a special functional category that, apart from containing aspect, is responsible for Case checking. Just like a tensed verb (in most languages) makes nominative Case checking possible, a perfective verb would enable objective Case checking. Sanz (1996) links perfectivity and transitivity by assuming, based on Tenny (1987), the feature [measure] to occupy a head that is checked by an object NP. Hence, `I ate' is not perfective but `I ate the apple' is. She assumes a Tr(ansitive)P as in e.g. Jelinek (1997), but an AGRo head might contain the same feature. I modify Sanz's tree and assume there is a preposition (that may or not be empty) in the head of Tr that is connected with the perfective features, to account for the perfectivity of eat+NP as well as eat up+NP:

46. IP

. I'

I TrP

[t] . Tr'

Tr VP

P . V'

[goal] V NP

ate the apple

In (46), because a Tr head is present that has perfective features, the verb licenses an object. This is not true when eat lacks an object. Often, speakers will emphasize perfectivity by adding a preposition as in (47) and (48) and I take this as evidence that Tr is present:

47. ASU Librarian: We received it in for you.

48. Do you mind if I open me eyes up slowly.

(Onslow in `Keeping up appearances')

There is also a link, noticed by many, between movement out of the VP and perfectivity. Again, that would be predicted if perfective aspect is represented by a Functional Projection that triggers movement that might be overt in some languages. (cf. de Hoop 1992, Verkuyl 1992). Meinunger (1995: 92ff) lists some German sentences relevant in this respect which I list in Dutch, where the same phenomenon occurs. The perfect participle connected to hebben `have' in Dutch (and German) is ambiguous between perfective and imperfective, for instance, in (49), where the object de bijbel is in the VP, either the bible was read completely or parts of it were. In (50), however, where the object moves out of the VP (considering the adverb to indicate the left-boundary of the VP), the bible has been read completely a number of times. The reason for the difference may be that an NP that moves to a higher position as in (50) moves to one (i.e. Spec TrP) where terminative aspect is checked and must be complete:

49. omdat ik vaak de bijbel gelezen heb

because I often the bible read have

`because I've read the bible often'.

50. omdat ik de bijbel vaak gelezen heb

because I the bible often read have

There is also a association between specificity, movement (cf. Diesing 1992), perfectivity and a special kind of Case (see Abraham 1996; Meinunger 1995; Kiparsky 1996). For instance, (51) sounds odd with the perfectivizing affix up and the unspecific object. Once a TrP has been selected (through up), the NP must be definite and must move out of the VP. The relationship between Case and movement is shown in (49) where the VP-internal NP only gets an (inherent) partitive Case as opposed to the VP-external NP in (50):

51. ?I ate up apples.

In discourse terms, the specific NP moves out of the VP because it is the topic rather than the focus. Following Cinque (1993), who argues that the focus position is the most deeply embedded, the specific NP moves out of the VP because it is the topic.

So far, I have shown that telic aspect can be expressed as in (46), i.e. by means of the simple past, as long as an object is present: I ate the apple. With intransitive (achievement) verbs such as in I went, He died, I wrote, and Elizabeth Taylor arrived, the verb is not properly perfective unless a goal is mentioned explicitly. Thus, one could have been emphasizing the action of the going and then I went is more imperfective than perfective; I went there on the other hand is perfective. (52), for instance, expresses a goal and that is why in one hour can be added to express telicity:

52. I went there in one hour.

The structure I suggest for simple pasts is as in (46) but one where the TrP is triggered by the location/time indication. This is what emphasizes the telicity and renders the construction perfective. Thus, either an object as in (46) or a location as in (52) triggers the presence of perfect features. The structure for (52) would be as in (46) with the locational PP checking the [goal] features in Spec TrP.

Sanz (1996) makes a distinction between Achievement verbs such as die and arrive and Accomplishment verbs such as write a letter and build a house. Both are telic but the former's objects are not measurable, whereas the latter's objects are measurable. She suggests two functional categories, an AktP (from Aktionsart) with [telic] features and a TrP with [measure] features. I argue constructions such as (46) and (52) are really similar.

In section 2, I argue that sentences such as (5) and (6) express past tense and present relevance but not perfective, terminative aspect. However, if one thinks of a location, I have gone, I have lived, I have worked become perfective. This locational sense is what justifies positing an abstract preposition in (46) to express the completion.

As argued in section 2, the perfect in English expresses the past tense but have adds current relevance. The completion of the action is expressed through Tr as in (46). In (46), perfectivity is linked to transitivity and, as mentioned, many people have made this link. Visser (1963: 97ff; 127) argues the ge prefix on the participle in Old English is a transitivizer (but see Lindemann 1970). If correct, this might point to ge being the head of a TrP. In Old English, ge expresses perfectivity without have, e.g. both (53) and (54) are well-formed. The prefix for is also a transitivizer in Old and Early Middle English as in (55). These prefixes would be in Tr and responsible for the relationship between perfective and the presence of objects:

53. Beowulf 1591

Sona þæt gesawon

soon that saw

`As soon as they saw that'.

54. Idem 694

Ac hie hæfdon gefrunen, þæt ...

But they had found-out that ...

55. Idem 804

ac he sigewæpnum forsworen hæfde

But he victorious-weapons cast-a-spell-on had

In this section, I argue that perfective aspect is represented by means of TrP, the head of which contains [goal]-features that are checked by either an NP or PP. The idea of an abstract preposition fits with other processes of grammaticalization that have occurred througout the history of English, e.g. the infinitival tense marker to (van Gelderen 1993); the deontic future marker for; and the progressive marker on (see next section), all originally prepositional. It fits with theories of grammaticalization (Heine et al 1991; Bybee et al. 1994) where spatial meanings gradually are extended to temporal ones.

In keeping with what I claim in section three, I will now argue that the progressive aspect in English is expressed by means of an abstract `preposition' as well. Cross-linguistically, there is evidence for such a preposition. I also indicate how the tense structure follows from assumptions made in section two.

In languages such as Dutch, there is no special progressive as in English: either one uses the simple present or a form of be followed by a preposition and a nominal verb. Thus, (56) is translated either as (57) or as (58):

56. I am writing letters.

57. I schrijf brieven

I write letters

58. I ben brieven aan het schrijven.

I am letters on the write

In Old and Middle English, the same two choices are available and the origin of (56) is often said to be a construction as (59) or (60) (but the debate over the origin is one of the fiercest in the literature on English, see e.g. Brunner 1962, Mossé 1938; some claim Celtic influence). In contemporary dialects, (61) occurs where the Verb moves to the Preposition overtly (to check its progressive features). In standard English, the preposition is further reduced to zero:

59. Brut Caligula 581

heo weren on slæpe

they were on sleep, `they were sleeping'.

60. Idem 3115

þe gode beoð on fihte

who good are in fight

61. I am a-going.

The structure I propose for the progressive (combined with the perfect) is as in (62) (cf. also van Gelderen 1996). Be is selected from the lexicon together with the abstract preposition connected to the Functional Category that carries the progressive features. In some languages, the preposition is not abstract but overt, e.g. Early Middle English on in (59) (and Old Egyptian and German have comparable constructions):

62. CP

. C'

C IP

. I'

She I PfP

[pres]. Pf'

Pf ETP

-as . ET'

ET ProgP

[past]. Prog'

Prog VP

P going

[prgr]

been

S R E

I have added the tense features in (62) as well. In this sentence, the clitic -as moves to I overtly to check present, been checks the past tense features in ET and going checks progressive (either through overt or covert movement). C has present tense and an interpretation as in (13) is the result. To capture Vendler's (1967) insight that only Activity and Accomplishment verbs occur as progressives, the progressive features could be seen as involving process. This would exclude State and Achievement verbs.

Bolinger (1971) argues, in a similar vein, that the progressive is an adverbial nominal, "a prepositional phrase from which the preposition has been deleted" (p. 247). He presents 8 pieces of evidence, one is the use of at to replace progressives as in (63), a second is the use of at in sentences such as (64), and a third (Bolinger's 8th) is that a where-question, normally answered by a locational adverb, can be answered with a progressive as in (65):

63. I am writing a book. What are you at these days.

64. He is working = He is at work.

65. Where's Joe? - He's reading.

(Bolinger's sentences)

The progressive construction in (62) can thus be argued to involve a preposition, even though it is an empty one in Modern English.

The passive is perhaps the most complex since it is the process closest to the verb stem, a process that is sometimes claimed to be a lexical one. Many languages have affixes indicating passivity and as a result, a VoiceP has been suggested (e.g. see Rivero 1990; Jelinek 1997). The VoiceP possibly replaces TrP (or AGRoP). In languages that distinguish in their choice of auxiliaries between lexical passives (e.g. zijn `to be' in Dutch) and transformational passives (worden `become' in Dutch), the ones for the transformational passive as in (66) are the ones that indicate imperfectivity, whereas the lexical passive as in (67) is perfective (sentence (67) is ambiguous between lexical and transformational but that is irrelevant here):

66. De appel wordt gegeten

`The apple is (being) eaten'.

67. De appel is gegeten

the apple is eaten, i.e. `the apple has been eaten'.

In English, the apple was eaten can mean `someone was eating the apple' or `the apple only has a core left'. If perfectivity is related to TrP and if passives lack TrPs, as suggested in (46) above, one expects that the transformational passive cannot be perfective and this is borne out in (66). What this means is that a TrP and a VoiceP are in complementary distribution and that since perfectivity is linked to TrP, perfective aspect cannot be expressed in a passive construction. The structure for a passive is then as in (68):

68. IP

. I'

She I PfP

[pres]. Pf'

may Pf ETP

-av . ET'

ET ProgP

[past]. Prog'

Prog VoiceP

P . Voice'

[prgr] Voice VP

been [pass] warned

being

In (68), the CP is left out for practical reasons, but C has present as well as I whereas ET has past tense making the tense interpretation of (68) the same as that of (13) above.

A few points of interest in (68). (a) The CP, IP, ETP `domain' can be seen as the tense domain which is relatively high up in the structure. (b) From (68), it also becomes clear that lexical items in a particular Functional Category do not check the features of that Functional Category but move and after they adjoin, check their features.

There are several structural arguments to support a tree as in (68) that differentiates between the different auxiliaries. (a) The word order is fixed: modal, perfect, progressive, passive and lexical Verb only occur in one order, as (69) and (70) show. (b) Preposing affects some and not other projections (progressive as in (71) but not perfect and passive as in (72), (73) and (74)). (c) Do so replaces the VP as in (75) but none of the other projections, e.g. perfect in (76):

69. The thief may have been being seen.

70. **The thief may be had being seen.

71. ...and being seen the thief may have been.

72. */?...and have been being seen the thief may.

73. ?...and been being seen the thief may have.

74. *...and seen the thief may have been being.

I will not enter into the `be-shift' debate, the phenomenon that when two forms of be are present, one remains in the VP. This was noticed in early transformational work and indicates a link between the ProgP on the one hand and the VoiceP/VP on the other.

In conclusion, I represent the progressive by means of a functional category with an empty Preposition as its head. The passive is represented as a VoiceP.

Several questions remain, for instance, the reason for the loss of perfective prefixes is unclear as is the relevance of the (short lived) English dative of movement indicating telicity. I briefly discuss these issues here.

The relationship between prepositional prefixes and perfectivity that I mentioned exists in Old English is lost by Middle English (Mustanoja 1960). It is not the introduction of the auxiliary have that causes this loss since have is already used as auxiliary in the earliest works (e.g. (54)). It may, however, be the introduction of the other auxiliaries in addition to have and the possibility of having a total of 4 different auxiliaries in one sentence that makes the checking between the main verb and preposition, i.e. the movement of the Verb to the Preposition, impossible. Old English is a synthetic language and exhibits a minimal use of auxiliaries. As Traugott (1992: 200) puts it: "[n]o two auxiliary verbs may occur in sequence except Pre-modal - Passive (e.g. might be destroyed)". However, it is hard to link the demise of the prefixed preposition and the introduction of multiple auxiliaries because both changes are gradual (e.g. some prefixed prepositions remain present in Early Middle English Brut and there are few multiple auxiliaries). By the time of Chaucer, there are complex auxiliary constructions, e.g. (22) above, and prefixed preposition no longer indicate transitivity/perfectivity. Investigating the precise link between these changes remains for further study.

In Spanish, as argued by Sanz (1996), Bonneau et al (1994) and others, there is a telic clitic se as in (77). In Middle English, there is a period in which third person pronouns indicate telicity as in (78) to (79). It is never a frequent construction (fewer than 10 occurrences in Chaucer) but it could be indicative of an overt telicity in Tr. The same is true in many other languages, e.g. Frajzyngier (1997) mentions (80), from Mina, a Chadic language, even though he argues the (possessive) pronoun is a marker of affectedness of the subject rather than of telicity:

77. Pedro se comió tres manzanas

`Pedro ate three apples' (Sanz's (43))

78. Chaucer, Knight's Tale, I, 2270

And with glad herte he wente hym hoom ful soone

79. Idem, Merchant's Tale IV, 1779

And to his bed he wente hym hastily

80. í-tsù t`tàn á wt' t`tàn

3.P-go 3.P PREP village 3.P

`They went home'.

Other aspectual markings in Middle English use the third person as well, e.g. as in (81), but it is unclear whether that can be accounted for in a similar way:

81. Layamon, Caligula 9939

Forð him gon ride; Arður þe riche

away him began ride; Arthur the great

`Arthur the great began to ride away'.

I have argued that with respect to perfective aspect, one needs to distinguish between the present perfect which is a past tense with an added sense of present relevance and the completive/telic aspect. I represent the former through a tense in C, I and ET and the latter through a TrP with either an object or a location rendering the sentence telic. Thus, tense is located relatively high in the tree structure whereas aspect is closer to the VP, i.e. the argument structure. Progressive aspect in Modern English can best be represented as an auxiliary associated with functional category whose head is a preposition. Passives are represented as VoicePs. I have not attempted to give a diachronic account and have only used older stages and varieties of English to argue points about Modern English.

References

Brinton, L. 1988. The Development of English Aspectual Systems. Cambridge: CUP.

Brunner, K. 1962. Die englische Sprache II. Tübingen: Niemeyer.

Burzio, L. 1981/1986. Italian Syntax. Dordrecht: Kluwer.

Chomsky, N. 1995. The Minimalist Program. Cambridge: MIT Press.

Comrie, B. 1976. Aspect. Cambridge: CUP.

Diesing, M. 1992. Indefinites. Cambridge: MIT Press.

Dikken, M. 1992. Particles. Amsterdam: HIL Publications.

Enç, M. 1987. "Anchoring Conditions for Tense", Linguistic Inquiry 18: 633-57.

Hornstein, N. 1990. As Time Goes By. Cambridge: MIT Press.

Jelinek, E. 1997. "Functional Projections and Clitics", UoA Ms.

Kiparsky, P. 1996. "Partitive Case and Aspect", ms Stanford.

Meinunger, A. 1995. Discourse dependent DP (de)placement. Diss Potsdam.

Mustanoja, T. 1960. A Middle English Syntax. Helsinki.

Partridge, A. 1969. Tudor to Augustan English. London: A. Deutsch.

Quirk, R. & C. Wrenn 1957. An Old English Grammar. London: Methuen.

Raith, J. 1951. Untersuchungen zum englischen Aspekt. Munchen: Max Hueber.

Reichenbach, H. 1947. Elements of Symbolic Logic. New York: MacMillan.

Ritter, E. 1991. "Evidence for Number as a Nominal Head". GLOW talk.

Roberge, P. 1997. "Perfect Auxiliary Reduction in Afrikaans", ICHL talk.

Strang, B. 1970. A History of English. London: Methuen.

Tenny, C. 1987. Grammaticalizing Aspect and Affectedness. MIT PhD.

Vendler, Z. 1967. Linguistics in Philosophy. Ithaca: Cornell University Press.

Visser, Th. 1963. An Historical Syntax of the English Language I. Leiden: Brill

Notes