Tense and aspect (in)compatibility in Serbian matrix and subordinate clauses

1

Tense and aspect (in)compatibility in Serbian matrix and subordinate clauses*

(Please cite as: Neda Todorović. In Press. Tense and aspect (in)compatibility in Serbian matrix

and subordinate clauses. Lingua (2015), http://dx.doi.org/10.1016/j.lingua.2015.04.007)

Neda Todorović

University of Connecticut

[email protected]

Abstract

This paper proposes a compositional temporal analysis of restrictions on perfective aspect in

morphological present tense contexts in matrix and (certain) embedded clauses in Serbian. The

main idea is that the availability of perfective aspect depends on the duration of the time interval

with respect to which the event is ordered. This time interval is further determined by the

interplay of temporal and aspectual components higher in the structure. Moreover, it is shown

that the observed restrictions on perfective aspect in Serbian are, to a large extent, parallel to

restrictions on non-progressive, episodic, non-stative, non-generic predicates in English.

Following the assumption that these predicates contain a perfective component (Kamp & Reyle

1993, Smith 1991, Cowper 1998, Wurmbrand 2014, inter alia), the proposed analysis allows for

the unification of Aspect in Serbian and the abovementioned predicates in English.

Keywords: Aspect Tense Reference time interval Serbian

1. Introduction

In matrix clauses in Serbian, verbs that occur with morphological present tense and which

denote events that temporally overlap with the Utterance Time (henceforth the UT) are

* For encouragement and extremely helpful discussions, I am grateful to Susi Wurmbrand. I am also grateful to

Dorit Abusch, Jonathan Bobaljik, Željko Bošković, Miloje Despić, John Gajewski, Stefan Kaufmann, Tanja

Milićev, Nataša Milićević, Beata Moskal, as well as the audiences of FASL 21 and Novi Sad LingColl1. I also thank

three anonymous reviewers for very useful and constructive feedback. A shorter version of this paper appeared in

the Proceedings of FASL 21.

2

compatible with imperfective aspect, as illustrated in (1a), but not with perfective aspect, as in

(1b).1,2

(1) a. Jovan prevodi pesmu.

Jovan translate- impfv.3.sg.pres. poem

‘Jovan is translating a poem.’

b. * Jovan prevede pesmu.

Jovan translate- pfv.3.sg.pres. poem

Intended interpretation: ‘Jovan has translated a poem (just now).’

Paunović (2001) briefly discussed the restrictions in (1), arguing that the perfective in Serbian,

in addition to containing an aspectual component, also introduces a temporal notion of

precedence (see e.g., Lin 2006 for similar ideas for Chinese). This notion of precedence forces a

perfective event to precede the reference time interval, i.e. the time interval for which a statement

is restricted (see Reichenbach 1947, inter alia).3 Since the reference time interval is the UT in the

case of (1), given the notion of precedence, the event crucially needs to be completed prior to the

UT.

This account (as is) faces problems with the lack of restriction on the perfective with forms that

receive past and future interpretations, as in (2) - (4). Even if the events denoted by the perfective

can be argued to have been completed prior to some point in the past (2b) (or prior to the UT), or

prior to some point in the future (3b), the account cannot explain the grammaticality of perfective

in (4) – according to one of the salient interpretations, the completion of the event occurs exactly

at the point introduced by the time adverbial at 4:30 p.m. yesterday/tomorrow. Rather than

resorting to an asymmetry between a purely aspectual nature of the imperfective and an

aspectual-temporal nature of the perfective (to comply with Paunović’s notion of precedence), I

will argue that the distribution of the perfective follows straightforwardly from a compositional

temporal analysis of the event location. 4,5

1 This restriction holds for non-stative, non-generic, episodic predicates. The distribution of perfective under other

interpretations is not discussed in the paper. 2 The intended interpretation of perfective in (1b) is the one where the point of finishing the translation overlaps

with the UT. The only scenario under which (1b) would be grammatical would be the one which involves

habitual/generic interpretation. Crucially, such interpretation is felicitous only in the presence of an overt QP, e.g.,

svake nedelje ‘every week’ (or if QP is present in the previous context). This is compatible with the analysis

proposed in the paper outlined in Section 3.

3 Note that I resort to this term as it is a widely-used one. However, assertion time from Klein (1995) and

Demirdache & Uribe-Extebarria (2004 et seq.) would be equally applicable. 4 According to some of speakers, (4a) and (4b) can also receive the inchoative interpretation, i.e., 4:30 p.m. marks

the beginning of the interval within which the event of translating is contained. However, an anonymous reviewer

3

(2) a. Jovan je prevodio pesmu.

Jovan is translated-impfv. poem

‘Jovan was translating a poem.’

b. Jovan je preveo pesmu.

Jovan is translated-pfv. poem

‘Jovan finished the translation of a poem.’

(3) a. Jovan će prevoditi pesmu.

Jovan will translate-impfv.inf. poem

‘Jovan will be translating a poem.’

b. Jovan će prevesti pesmu.

Jovan will translate-pfv.inf. poem

‘Jovan will translate a poem.’

(4) a. Jovan je preveo pesmu juče popodne u 4:30.

Jovan is translated-pfv. poem yesterday afternoon in 4:30

Intended interpretation: ‘Jovan finished the translation of a poem at 4:30 p.m. yesterday.’

b. Jovan će prevesti pesmu sutra popodne u 4:30.

Jovan will translate-pfv.inf. poem tomorrow afternoon in 4:30

Intended interpretation: ‘Jovan will have translated a poem at 4:30 p.m. tomorrow.’

Moreover, the observed restrictions on perfective aspect in Serbian are parallel to the

distribution of non-progressive forms of predicates that receive episodic, non-stative, non-

generic interpretation in English (henceforth eventive predicates, following Pesetsky 1995, Enç

1991, Bošković 1996, 1997, Martin 1996, 2001, Wurmbrand 2014, inter alia). The restrictions

on the distribution of eventive predicates in simple present tense have widely been discussed in

Bennett & Partee (1972), Taylor (1977), Dowty (1979), Enç (1991), Smith (1991), Cowper

(1998), Abusch (2004) and Wurmbrand (2014), inter alia. Non-progressive eventive predicates

are not felicitous in matrix clauses when the event overlaps temporally with the UT, as in (5a);

rather, they can only occur in the progressive form, as in (5b). Conversely, non-progressive

forms are allowed when the predicate receives a past or future interpretation, as in (6) and (7),

respectively. By comparing Serbian (1-3) and English (5-7) matrix clauses, we observe the same

observes that there are speaker-variations with respect to the availability of this interpretation. Moreover, for some

speakers, the availability of the interpretation depends on the type of predicate. For a more detailed overview of the

distribution, see footnote 29. 5 An anonymous reviewer points out that if Paunović’s account is modified such that only the final endpoint of the

perfective is allowed to co-occur with the relevant reference time interval, while the rest of the event occurs prior to

it, then the problematic facts in (4) can be accounted for. While I agree that such modification would make the data

in (4) unproblematic, Paunović’s account would need to be modified even more in order to account for the

inchoative interpretation of (4) accepted by some speakers with certain predicates. The inchoative interpretation

allows for the reference time interval to overlap with the initial point of the event, so Paunović's account would have

to allow both the initial and final boundary of an event to overlap with a relevant reference time interval. Such a

modification of her account would make it considerably different from her original proposal.

4

kind of restrictions with events which temporally overlap with the UT, in Serbian with perfective

aspect, and in English with non-progressive forms.6

(5) a. John translates a poem (*right now).

b. John is translating a poem right now.

(6) John translated a poem yesterday.

(7) John will translate a poem tomorrow.

The same parallel is observed in embedded contexts. In Serbian complement clauses, finite

counterparts of infinitival complements are realized with the complementizer da and

morphological present tense. If the embedded predicate is interpreted as occurring

simultaneously with the matrix predicate, the perfective aspect cannot felicitously occur with

morphological present, as in (8). On the other hand, complements receiving a future-oriented

interpretation, as in (9), allow for either aspectual specification on the embedded verb.7

(8) a. Verujem da Jovan prevodi pesmu.

believe-1.sg.pres that Jovan translates-impfv. poem

Intended interpretation: ‘I believe that John is translating a song (right now).’

b. *Verujem da Jovan prevede pesmu.

believe-1.sg.pres that Jovan translates-pfv. poem

Intended interpretation: ‘I believe that John has translated a poem (just now).’

(9) a. Želim da sutra popodne prevodim pesmu.

want-1.sg.pres that tomorrow afternoon translate- impfv.1.sg.pres. poem

‘I want to be translating a poem tomorrow afternoon.’

b. Želim da sutra popodne prevedem pesmu.

want-1.sg.pres that tomorrow afternoon translate- pfv.1.sg.pres. poem

‘I want to translate the entire poem tomorrow afternoon.’

Correspondingly, in English non-finite complements, non-progressive forms display the

same behavior. Wumbrand (2014) observes that, whereas non-progressive forms are not allowed

with embedded eventive predicates receiving a simultaneous temporal interpretation, as in (10),

no such restrictions arise with future-oriented complements, as in (11).8

(10) *John believed Susan to translate a poem right then.

6 I discuss the English counterparts of (4) in Section 2 and Section 3.2.

7 I return to other possible interpretations of (8b) and (9b) in section 4.

8 Examples in (10) and (11) are a modified version of Wurmbrand’s (2014) examples.

5

(11) John decided to translate a poem (tomorrow).

Given the parallel distribution of perfective eventive predicates in Serbian and non-

progressive eventive predicates in English, the aim of this paper is to provide an empirically

adequate formal analysis of the restrictions on perfective aspect with morphological present tense

forms that temporally overlap with the UT in matrix clauses in Serbian, as well as to extend the

analysis to English non-progressive eventives. Moreover, it will be shown that the analysis can

successfully capture the distribution of perfective and non-progressive predicates in embedded

environments in the two languages. Asumming that restrictions on English non-progressive

eventives stem from restrictions on viewpoint aspect, the observed parallelism between English

and Serbian is intended to provide support for the presence of viewpoint aspect in Serbian. Note

also that I assume that viewpoint aspect is projected on the top of and in addition to situation

aspect in Serbian. However, given that the complexity and breadth of the discussion concerning

situation aspect make it a topic in its own right and, given that it is mostly orthogonal to the

analysis in the paper, the discussion on situation aspect is set aside.9

The paper is organized in the following way: Section 2 briefly sketches the main idea of the

previous approaches to the distribution of non-progressive forms in English, the idea that also

finds its place in the analysis I am proposing for Serbian. Section 3 proposes the mechanism of

Tense and Aspect temporal ordering in Serbian. Assuming a compositional analysis to temporal

location of the event (Klein 1994, von Stechow 1999, von Stechow 2002, Pancheva 2003,

Pancheva and von Stechow 2004), I argue that the restrictions on the perfective arise due to the

conflict between requirements of the perfective aspect and the duration of the time interval with

respect to which Aspect is ordered. This analysis is then further extended to English data. It will

be shown that Serbian and English behave similarly in their distribution with past forms and

when restrictions arise, they are due to the same conflict between perfective aspect and the

reference time interval. English and Serbian, however, differ with respect to some interpretations

of future forms. I argue that this difference is not in terms of structure, but in terms of different

interactions with the time adverbial. In Section 4, the proposed analysis captures the distribution

of aspect in complement clauses. Section 5 concludes the paper.

9 I am grateful to an anonymous reviewer for highlighting the relevance that the observed parallelism between

Serbian and English has for the viewpoint status of the perfective-imperfective distinction in Slavic.

6

2. Non-progressive eventive predicates in English are specified for perfective aspect

Restrictions on non-progressive forms of eventive predicates in English have previously been

explored in Enç (1991), Smith (1991), Kamp and Reyle (1993), Cowper (1996, 1998), inter alia.

All these approaches share (either implicitly or explicitly) the assumption that these forms are

specified for perfective aspect. In addition, in environments where these forms are illicit,

requirements of the perfective conflict with the duration of the reference time interval.

Wurmbrand (2014) takes a similar stand, providing a plethora of contexts in English where these

forms are not felicitous. Assuming the presence of a perfective component with these forms,

Wurmbrand adopts the definition of perfective aspect as in (12a) such that the event time interval

needs to be included within the reference time interval.10,11

Given the definition in (12a), it

follows, that non-progressive forms are not allowed in environments where the reference time

interval is included in the event time interval. For instance, the reference time interval in (5a)

(repeated in (13)) is the UT, and given that the event time interval includes the UT, only an

imperfective interpretation is possible. Assuming that the progressive form is a morphological

reflex of an imperfective aspectual specification in English, only a progressive form is predicted

to be possible in eventive present tense contexts, Wurmbrand argues.

(12) a. Perfective: P.t.w.e [time(e)t & P(w)(e)]

b. Imperfective: λP.λt.w.e [t ⊆time(e) & P(w)(e)]

(13) John eats an apple (*right now). (cf. John is eating an apple right now.)

Wurmbrand’s analysis further accounts for the contrasts between (14) and (15). In (14), the

event time interval is included in the longer reference time interval, yesterday and tomorrow,

respectively, and the non-progressive form is allowed. However, in (15), the non-progressive,

10

In embedded contexts, the time for which the statement is restricted is not necessarily the time interval with

respect to which the event is ordered. Rather, adopting Abusch’s (2004) attitude holder’s contemporary now, the

time which the speaker considers his/her now, I argue that this interval serves as the interval with respect to which

the event is ordered (Wurmbrand 2014 argues similarly for English). I provide the details of the analysis in Section

4.2. 11

The formal definitions of imperfective and perfective adopted here are from Kratzer (1998).

7

perfective form is precluded, since the reference time interval, the time of the mailman’s arrival,

is included in the event time interval, the time of translating.12

As expected, in the ‘mailman’s

arrival’ context, only the progressive is allowed, as in (16).

(14) a. John translated a poem yesterday.

b. John will translate a poem tomorrow.

(15) a. *John translated the poem when the mailman arrived.

b. *John will translate the poem when the mailman arrives.

(16) a. John was translating a poem when the mailman arrived.

b. John will be translating a poem when the mailman arrives.

In the Serbian examples corresponding to (14), the perfective is also allowed (cf. (2b) and

(3b)). Interestingly, unlike in English, the perfective is also allowed in the ‘mailman’s arrival’

examples as in (17) (cf. (4)). This comes as a surprise, given that the reference time interval

seems to be included in the event time interval.

(17) a. Jovan je prevodio /preveo pesmu kada je poštar stigao.

Jovan is translated-impfv./translated-pfv. poem when is mailman arrived

’Jovan was translating a poem when the mailman arrived.’ (imperfective)

Intended interpretation: ’Jovan finished the translation of a poem at the moment at

which the mailman arrived.’ (perfective)

b. Jovan će prevoditi /prevesti pesmu kada poštar stigne.

Jovan will translate- impfv.inf. /translate-pfv.inf. poem when mailman arrives

‘Jovan will be translating a poem when the mailman arrives.’ (imperfective)

Intended interpretation: ‘Jovan will have translated a poem at the moment at which the

mailman arrives.’ (perfective)

Although it initially appears as if the perfective aspect in the two languages is subject to

different requirements, I argue that this is not the case. Instead, by considering a range of

possible interpretations for English and Serbian, it becomes clear that the two languages are not

as different as they seem to be. 13,14

12

The examples in (14)-(16) are based on Wurmbrand’s (2014) examples. Her examples corresponding to (15) are

Leo sang in the shower yesterday and Leo will sing in the shower tomorrow, and she reports that both are fine under

an inchoative interpretation, an interpretation where Leo starts singing after or at the moment of the mailman’s

arrival (2014:fn 20). 13

I follow the abovementioned authors, and assume that non-progressive forms in English are specified for

perfective aspect. This is further supported throughout the paper by the parallel behavior of these forms and

perfective forms in Serbian. 14

I thank the anonymous reviewers for a thorough discussion on the range of possible interpretations.

8

Consider first past interpretations in both languages. Like their English counterparts,

perfective verbs in Serbian are not allowed under the interpretation where Jovan was in the

process of translating a novel at the time of mailman’s arrival, i.e., where he had started the

translation at some time before then and was still not done at the time of mailman’s arrival, as

shown in (17a). Rather, only the imperfective form is felicitous with this interpretation, as in

(18). The restrictions with this interpretation in both English and Serbian follow from the

analysis outlined in section 3.2.

(18) Jovan je prevodio /*preveo pesmu kada je poštar stigao.

Jovan is translated-impfv. /translated-pfv. poem when is mailman arrived

Intended interpretation: ’Jovan was translating a poem when the mailman arrived.’

Secondly, the most natural interpretation for Serbian examples containing perfective verbs in

(17a) is the one where the event of translation was completed at 4:30 p.m.. While this reading

would be best expressed with a Pluperfect form in both English and Serbian, the Pluperfect

appears to be marked (I return to the actual form in Serbian). In order to express this reading,

speakers use the form in (15a) for English and in (17a) for Serbian. Note that although (15a) is

preferred in English by speakers for which the Pluperfect belongs to a higher register and is

rarely used, there are speakers who use the Pluperfect productively. Those speakers reject this

interpretation with the forms in (15a).15

As will be shown in Section 3.2, the compositional

analysis of past forms which postulates an additional Perfect component can account for the

availability of this reading in both languages and also capture the speaker-variations observed in

English.

Finally, the third reading is the one where the translation time was very short and the entire

poem was translated during mailman’s arrival. Under such a scenario, the analysis illustrated in

Section 3.2 in principle also allows this reading. 16,17,18

15

Alternatively, these speakers use the verb ‘finish’ to force this interpretation. 16

The question, however, remains as to how pragmatically plausible is this scenario. 17

Note that, for some speakers, (17a) can also receive the inchoative interpretation, i.e. an event of Jovan translating

the poem is part of the interval which begins with the mailman’s arrival. I believe that this is because of the meaning

of kada ‘when’, which, under this interpretation, essentially means nakon što ‘after’ (but see fn. 29). 18

An anonymous reviewer points out that, for some speakers, not every predicate is equally available in (15a) and

(17a) for both Serbian and English. For instance, some speaker are less ready to accept a predicate that takes a

longer time, e.g. ‘translate a novel’, as available, as opposed to a predicate that would take less time to complete,

e.g. ‘translate a poem’. As it will be clear from section 3.2., and as the reviewer correctly points out, the analysis

predicts both predicates to be equally available under the second reading. However, the fact that speakers have

9

In both English and Serbian, future forms are excluded under the interpretation where Jovan

will be in the process of translating a poem before mailman’s arrival and that the process will

still be ongoing at the point of mailman’s arrival. However, the two languages differ in that

English, with the forms in (15b), does not allow for the interpretation where the event will be

completed at the point when the mailman arrives. This interpretation is, however, available in

Serbian with (17b). Unlike with past forms, I propose that there is no Perfect component with

either English non-progressive future forms in (15b) or with Serbian future perfectives in (17b).

What is then responsible for the discrepancy with respect to the available interpretations with

these forms? I propose that the difference lies in the way these forms interact with the time

adverbial. I show in section 3.2 that, although this form in English is more restricted in terms of

available interpretations, English also makes use of the Future Perfect, which compensates for

the interpretations that are otherwise present in Serbian. Crucially, I argue that both past and

future interpretations can be captured under the step-by-step semantic computation, as shown in

section 3.2. Section 3 provides the compositional temporal analysis of the availability of the

perfective form in Serbian and the non-progressive form in English, capturing their distribution

when the event temporally overlaps with the UT, as well as when it is receives past and future

interpretations. Section 4 extends the analysis to embedded contexts.

3. Temporal location of the event with respect to the reference time interval

This section provides a formal analysis of the incompatibility of present tense and perfective

aspect with eventive predicates in Serbian. It also examines how the distribution of perfective in

Serbian fits with respect to the distribution of non-progressive forms of eventive predicates in

English.

clearer judgments in examples with ‘poem’ may be due to the fact that ‘translating a poem’ provides a context under

which both the second and third meaning are possible. The third meaning is excluded/highly implausible, on the

other hand, when the event typically takes a longer time, as in the case of ‘translating a novel’. Moreover, while the

‘translating a novel’ context is still compatible with the second meaning, this meaning requires a silent Perfect (to

which I return in section 3.2), which could be seen as an uneconomical adjustment and seems not to be available for

all English speakers. Furthermore, the sentence would be odd/impossible under the first meaning, which could

contribute to the hesitation speakers have for such examples. In sum, the first meaning is consistently excluded,

which the analysis predicts. The third meaning is possible, however, only when the event is very short. Situations

that are compatible with both the second and third meaning are less marked, and hence considered more felicitous

than situations that are only felicitous with the second meaning, which might be an explanation for why ‘translating

a poem’ sounds more felicitous than ‘translating a novel’.

10

I am adopting von Stechow’s (1999) proposal that both Tense and Aspect are involved in the

temporal ordering of the event time interval with respect to the reference time interval (see also

Klein 1994, von Stechow 2002, Pancheva 2003, Pancheva and von Stechow 2004). The

reference time interval is a time interval with respect to which the event time interval needs to be

temporally located, and which is the UT, unless contextually specified otherwise. Furthermore, I

suggest that the reference time interval is structurally represented, occupying a position in

SpecTP (see, e.g. Demirdache & Uribe-Extebarria 2004). The hierarchical relations of the

reference time interval, Tense and Aspect are given in (19). Following Demirdache & Uribe-

Extebarria's (2004 et seq.), the event time would be adjoined to VP.

(19) TP ei the reference T’

time interval ei

Tense AspP ei

Aspect vP

I propose a compositional analysis, the computation proceeding in step-by-step fashion. More

specifically, both Tense and Aspect in (19) are ordered with respect to the reference time

interval, i.e. the UT. However, for Aspect and hierarchically lower elements, this is an indirect

relation. All relations are strictly local in the sense that each temporal head establishes a relation

with its sister: Tense in T establishes a direct relation with the reference time interval (the UT),

and then the aspectual projection is ordered with respect to the time interval that T establishes in

its interaction with the reference time interval. In that sense, Aspect is also ordered with respect

to the reference time interval, with the difference being that this ordering is not direct, but rather

mediated via Tense. That Aspect establishes an indirect relation with the reference time interval

will also be also evident in section 3.2 for past and future matrix contexts and in section 4 for

embedded contexts. It will be shown that, in the presence of Perfect and woll in the structure,

respectively, these components also serve as mediators between Aspect and the time interval that

T establishes in its interaction with the reference time interval.

11

The step-by-step compositional nature of the system is similar to Demirdache & Uribe-

Extebarria's (2004 et seq.) approach. Under their model, the relations established between tenses

are mapped in a particular way to the syntax, which I do not necessarily assume. My analysis is

strictly compositional, but I believe that the main contribution of the paper (the incompatibility

of certain elements) can be expressed either via the compositional semantics I followed or it can

also be translated into a Demirdache & Uribe-Extebarria’s system. My choice for the former

approach is not intended to exclude other approaches with a similar compositional flavor.

3.1 The UT interpretations

I refer to present perfective forms which denote events that overlap temporally with the

Utterance Time as in (1) in Serbian as the UT interpretations. Regarding the restrictions observed

with the UT interpretations as in (1), I propose that there is a conflict between lengthy events

and short interval with respect to which the events are temporally ordered. Since this conflict

cannot be resolved, the event time interval cannot be temporally located, and the present

perfective form is expected to be infelicitous. I spell-out the analysis below.

According to the definition in (20) (a somewhat modified version of the definition given in

Pancheva and von Stechow 2004), present tense introduces an interval identical with the

contextually specified reference time interval.19

The reference time interval is the UT (see von

Stechow 1999 for a similar proposal), which I take to be a near-instantaneous time interval (cf.

Giorgi & Pianesi 1997, Cowper 1996, 1998, and contra Enç 1987; see also Ogihara 2007). As for

perfective aspect, I am adopting Kratzer’s (1998) definition in (12a), repeated in (21), according

to which the event time interval must be included within the reference time interval. Importantly,

given the hierarchical relations in (19), the time interval with respect to which Aspect is ordered

is determined by the requirements of the c-commanding Tense. The structural representation for

a present perfective form is given in (22).

19

According to Pancheva and von Stechow (2004), an interval introduced by present should be coextensive with the

UT. This definition would suffice for matrix clauses in Serbian, where the UT indeed serves as the reference time

interval for present tense. However, as illustrated with embedded contexts in Section 4, present tense in Serbian is a

relative tense, and, as such, it does not necessarily refer to the UT. Thus, the definition in (20), (which I also adopt

for the embedded present tense), is more accurate for Serbian since the time interval introduced by present is, rather

than being equal to the UT, equal to whatever serves as the contextually salient specified reference time interval (as

it will become evident in Section 4).

12

(20) ⟦ PRESENT1⟧ = p(it).t(i).t1(i) [t1=t & p(t1)=1]

(21) Perfective: P.t.w.e [time(e)t & P(w)(e)]

(22) CP wo

UT TP wo

⟦PRESENT1⟧ = λp. λt. w. ∃t1[t1 = t & P(t1)] AspP wo

perfective: P.t.w.e [time(e)t & P(w)(e)] w.e. K (w)(e)

⟦CP UT [TP Present [AspP perfective [vP ]]]⟧ = w. ∃t1[t1=tc & ∃e [ time(e) t1 & K (w) (e)]]

Given that Aspect is ordered with respect to the time interval that T establishes in its

interaction with the reference time interval, and given the definitions of present tense in (20) and

perfective aspect in (21), perfective aspect will not be able to satisfy the requirements of

including the event time interval within the short interval established by the interaction of present

tense and the reference time interval. More specifically, present tense introduces a time interval

identical with the contextually specified reference time interval, i.e. the UT in matrix clauses.

Perfective aspect requires the event time interval to be included within the time interval

established by present tense, which, in turn, is determined by the UT. Due to the requirements of

tense, the time interval with respect to which the perfective aspect is ordered remains the same as

the one for present tense, i.e. the UT. Now, if the UT is a near-instantaneous interval, it seems

impossible to satisfy the inclusion requirements of perfective: an event marked by perfective

cannot be included within such a short time interval. Since the conflict cannot be resolved, the

event cannot be temporally located. Consequently, present perfective forms are expected to be

ungrammatical.

Furthermore, the same mechanism accounts for the distribution of English non-progressive

eventive predicates receiving the UT interpretation. Since these forms are specified for present

tense and for perfective aspect, the same conflict between lengthy events and short time intervals

is expected to arise. The prediction is borne out, as illustrated in (5), and repeated in (23). Only

the progressive form is felicitous under the UT interpretation. If the progressive forms are

specified for imperfective aspect, which requires for the event time to include the reference time

13

interval, the duration of the reference time interval should not matter (as long as the event time is

long enough to include the reference time interval).20

(23) a. John translates a poem (*right now).

b. John is translating a poem right now.

3.2 Viewpoint Aspect in past and future interpretations

Unlike the UT interpretations, past and future interpretations allow for the realization of

perfective aspect in Serbian (24), and non-progressive eventive predicates in English (25).

(24) a. Jovan je preveo pesmu.

Jovan is translated-pfv. poem

‘Jovan finished the translation of a poem.’

b. Jovan će prevesti pesmu.

Jovan will translate-pfv.inf. poem

‘Jovan will translate a poem.’

(25) a. John translated a poem yesterday.

b. John will translate a poem tomorrow.

As it was shown in section 3.1, the actual problem for the realization of the perfective in the

forms that receive the UT interpretation was the duration of the time interval with respect to

which Aspect was ordered. Applying the same mechanism of the event time location, I propose

that this issue does not arise in constructions with past and future interpretations because they

both extend the reference time interval, allowing for the requirements of Aspect to be satisfied

and for the form to be realized.

Past forms with the structural configuration as in (26), allow for perfective aspect in (24a),

because there is no conflict between the lengthy events and short time intervals. I take past tense

to introduce a time interval preceding the reference time interval, as in (26). Given that UT is

located in SpecTP, past tense introduces an interval located prior to the UT. Importantly, the

interval that past tense introduces is a potentially indefinitely long time interval. This, I argue,

has a crucial impact on the availability of the perfective with past interpretations. Namely, given

20

In this paper, the emphasis is placed on the forms specified for perfective aspect in both languages, whereas

examples with imperfective aspect are mostly used to illustrate the contrast in the distribution of the two aspectual

forms. Imperfective aspect involves restrictions that do not arise in the examples discussed in here. However, they

are predicted to arise if the event time interval is shorter than the reference time interval.

14

the structure in (26), the time interval introduced by past tense (and located prior to the UT, as

dictated by the reference time interval in the SpecTP) serves as the reference time interval for

Aspect. In this case, unlike with the UT interpretations, perfective aspect requires the inclusion

of the event time interval at the potentially indefinitely long time interval preceding the UT. The

duration of the reference time interval for Aspect with past interpretations does not pose an

obstacle for the requirements of the perfective, simply because the interval is not specified as

short and, as such, it is not incompatible with the requirements of perfective. Thus, the

requirements of Aspect can be satisfied, and the form is correctly predicted to be grammatical.

The computation of the structure is given in (28). The same analysis can account for the

distribution of non-progressive forms in English (25a).

(26) CP wo

UT TP wo

⟦PAST1⟧ AspP wo

perfective w.e. K (w)(e)

(27) ⟦PAST1⟧ = λp. λt. .w. ∃t1[t1 < t & P(t1)] (von Stechow 2009)

(28) CP wo UT TP wo ⟦PAST1 ⟧ = λp. λt. w. ∃t1[t1 < t & P(t1)] AspP wo perfective: P.t.w.e [time(e)t & P(w)(e)] w.e. K (w)(e)

⟦CP UT [TP Past [AspP perfective [vP ]]]⟧ = w. ∃t1[t1<tc & ∃e [ time(e) t1 & K(w) (e)]]

The similar extension of the reference time interval for Aspect occurs with future

interpretations in (24b). Abusch (1985, 1988) argues that forms that have future interpretations in

15

English comprise a tense component [present] and a modal component woll.21

Following

Abusch, I propose that the matrix future contexts in Serbian also contain [present] and woll.

However, I analyze the woll component along the lines of von Stechow (2009), according to

whom the future component is temporal in nature, introducing a time interval that follows a

contextually specified time interval:

(29) ⟦woll1⟧ = λp.λt.w.∃t1 [t1 > t & P(t)] (along the lines of Von Stechow 2009)

Crucially, the time interval introduced by the future component is potentially indefinitely

long. Assuming the same structural configuration for both English and Serbian matrix future

contexts as in (30), present tense will be ordered with respect to the UT, with which it needs to

be identical. The future component introduces a time interval following the UT (as dictated by

the present tense). This potentially indefinitely long time interval now serves as the ordering time

interval for Aspect, allowing for the requirements of the perfective to be met, and correctly

predicting a perfective form under future interpretations to be grammatical in both English and

Serbian.

21

In her analysis of tenses, Abusch (1998) observes that, in examples like This evening I will tell my friend about

everybody I meet today (1998:13), the embedded present tense receives the future-shifted reading. The similar holds

for past embedded under future: At the end of next term, I will give automatic A’s to all students who turned in their

terms papers on time (1998:13), where the embedded past refers to events occurring prior to some point in the

future, rather than prior to the UT. Abusch proposes that that the above interpretations are due to will in the matrix

clause which substitutes for the future-moving interval for the embedded present and past, respectively. More

specifically, will is analyzed as composed of true tense part, present and a temporal substitution operator.

Simplifying it grossly, the following is the structure of matrix clauses (would is decomposed in a similar way, with

the difference being that T is specified for past tense):

(i) ∃e [visit (m,l,e) & e (n, ∞) wo T

0: λP. P(n) λt. ∃t [visit [m,l,e) & e (t, ∞)]

Present Marry woll visit London

In the substitution operator in (i), t is a bound variable that corresponds to the tense argument of will. The top

occurrence of will substitutes t with n located in T, deriving the interval (n, 1). Crucially, this can account for shifted

interpretations illustrated above, because the eventualities corresponding within the main verb complements are

located within this interval (n, 1). More specifically, in the case of embedded present, the event time falls into the

future, rather than the UT. In the case of embedded past, the past is relative to a contextually specified future time

interval, rather than to the UT. Additional motivation for the posteriority component of woll can be found in

Thomason (1970), Condoravdi (2001), Copley (2002), Kaufmann (2005), Wurmbrand (2014), inter alia. I provide a

more detailed analysis in Section 4.3.

16

(30) CP wo

UT TP wo ⟦PRESENT1⟧ = λp.λt.w.∃t1[t1 = t & P(t1)] wollP wo

⟦woll1⟧ = λp.λt.w.∃t1 [t1 > t & P(t)] AspP wo

perfective: P.t.w.e [time(e)t & P(w)(e)]) w.e. K (w)(e)

⟦CP UT [TP Present [wollP woll [AspP Perfective [vP ]]]⟧ =

w. ∃t2[t2=tc & ∃t1 [t1> t2 & ∃e [ time(e) t1 & K(w)(e)]]]

The availability of perfective aspect with future and past interpretations on one hand, and its

unavailability with the UT interpretations on the other, is thus due to the length of the time

interval with respect to which Aspect is ordered a relatively long time interval in the former,

but not in the latter case, correctly predicting the form to be felicitous with past and future

interpretations, but not with the UT interpretations. However, when the reference time interval is

restricted to a relatively short interval, the same restrictions apply as with the UT interpretations,

as it will be clear from section 3.2.1.1 and 3.2.1.2.

At this point, a note is in order regarding the range of interpretation of periphrastic past form

‘be + participle’ in Serbian. As is well-known, periphrastic form can receive past interpretations

as well as perfect interpretations (see e.g. Riđanović 2012, inter alia). In that respect, Serbian

periphrastic form would be a superset of the meanings of English simple past and Present Perfect

form (similar to German, e.g. see Von Stechow 2002), as well as of Pluperfect, the interpretation

discussed in the paper. This is compatible with the current analysis where the difference between

the true past and Pluperfect reading of the periphrastic form stems from the difference in

structural configuration (as discussed in the next section; cf. second reading of (17a)).22,23

22

I would like to thank an anonymous reviewer for useful discussion on the range of interpretations of the

periphrastic form in Serbian. 23

Note that, although Present Perfect is not discussed in this paper, Past Simple and Present Perfect are, from a

temporal perspective, rather similar; Instead, English Present Perfect observes, e.g. current relevance (e.g. Portner,

2003) or life-time effects, which do not necessarily stem from its core temporal meaning. In other words, Present

Perfect is temporally similar to Past Simple modulo quantificational past. In a similar vein, Simple Past

interpretations that I am discussing in Serbian are temporally not necessarily different from Present Perfect.

17

3.2.1 Temporally restricted past and future interpretations

If the availability of Aspect crucially depends on the duration of the time interval with respect to

which it is ordered, then we predict the following: with past and future forms that in some way

restrict the reference time interval for Aspect to a relatively short time interval, the inclusion

requirements of the perfective cannot be met, and the form is predicted to be infelicitous. This is

illustrated for the time adverbial at 4:30 p.m.: in English (31) for past and (32) for future

interpretations only progressive forms are available, and in Serbian, only imperfective forms are

available for the past and future forms, as in (33) and (34), respectively. Crucially, in both cases,

the intended interpretation is the one where the event starts at some point prior to the UT and it is

ongoing at 4:30 p.m..

(31) a. *Jovan translated a poem at 4:30 p.m. yesterday.

b Jovan was translating a poem at 4:30 p.m. yesterday.

(32) a. *Jovan will translate a poem at 4:30 p.m. tomorrow.

b. Jovan will be translating a poem at 4:30 p.m. tomorrow.

(33) a. *Jovan je preveo pesmu juče popodne u 4:30.

Jovan is translated- pfv. poem yesterday afternoon in 4:30

b. Jovan je prevodio pesmu juče popodne u 4:30 .

Jovan is translated-impfv. poem yesterday afternoon in 4:30

Intended interpretation: ‘Jovan was translating the poem yesterday at 4:30 p.m..’

(34) a. *Jovan će prevesti pesmu sutra popodne u 4:30.

Jovan will translate- inf.pfv. poem tomorrow afternoon in 4:30

b. Jovan će prevoditi pesmu sutra popodne u 4:30.

Jovan will translate- inf.impfv. poem tomorrow afternoon in 4:30

Intended interpretation: ‘Jovan will be translating a poem tomorrow at 4:30 p.m..’

Following Ogihara (1995), I assume that „the contribution of the adverb is to restrict the

temporal location of the event in question” (1995:663). In other words, in (31)-(34), the

adverbial restricts the potentially indefinitely long time interval introduced by past tense (and the

woll component) to a rather short interval. Importantly, this newly formed time interval serves as

There is, however, one potential issue with Present Perfect interpretations – periphrastic past forms, although

ambiguous, cannot have the interpretation where the event started at some point in the past and is completed at the

UT, which is what the compositional analysis should in principle allow; if Present Perfect is comprised of present

tense and the Perfect component, present tense would make the UT the evaluation time and the Perfect would extend

this interval backwards; the event whose final stage overlaps with the UT should then be available. However, such

interpretation seem not to be attested . However, see Pancheva and von Stechow 2004 on how this issue can be

overcome by resorting to cross-linguistic variations of the semantics of present.

18

the reference time interval for Aspect, and, given its duration, perfective aspect is predicated to

be ungrammatical. Note that this correctly excludes the non-progressive eventive predicates in

English and perfective aspect in Serbian under the interpretation where John was/will be in the

process of translating the poem at 4:30 p.m.

On the other hand, Serbian and English (subject to speaker-variation) allow for the

interpretation where Jovan had been translating the poem for some time, and he finally finished

the translation at 4:30 p.m. yesterday, as in (35). The availability of this interpretation is

discussed in the next section. There is, however, a difference with the future forms. While

Serbian future forms allow for the translation to be finished at 4:30 p.m. tomorrow, as in (36b),

English forms in (36a) do not. As discussed in section 3.2.1.2., English uses the Future Perfect

form to obtain this interpretation.24

24 As an anonymous review correctly points out, in the examples with perfective verbs as in (i) and (ii) below, no

completion takes place at the time specified by the time adverbial. Moreover, the events denoted by the verbs in (i)

and (ii) are not entailed to be completed at any point, and especially not at the time specified by the time adverbial. I

believe that, in these examples, the prefixes make the crucial contribution, by introducing the initial point of the

event, and making this interpretation the only possible interpretation (in (ii), the ablative meaning indicates the

change of place, which contributes the interpretation of the initial point of traveling). Admittedly, these cases are

problematic for the claims made in this paper, because there is no entailment of the event being completed, or in

more general sense, no entailment of the event being bounded in the way boundedness is currently defined. Even if

these prefixes are operating in the domain of lexical aspect, a current definition of viewpoint aspect in Serbian

would also need to cover this type of boundedness, i.e. where the initial point is bound to a particular reference time,

especially because the inceptive interpretation that prefixes introduce is comparable to inceptive interpretation which

is also observed with (35b), as discussed below. Nevertheless, the current definition of boundedness is retained,

since it is compatible with the examples in the paper. The issue of the contribution of prefixes would need to be left

aside for future research.

(i) Jovan je za-pevao juče popodne u 4:30.

Jovan is Inch-sing.pfv. yesterday afternoon in 4:30

‘Jovan started singing yesterday at 4:30 p.m..’

(ii) Marija je ot-putovala u London juče popodne u 4:30.

Marija is Abl-travelled.pfv in London yesterday afternoon in 4:30

‘Marija left for London yesterday at 4:30 p.m..’

19

(35) a. Jovan translated the poem at 4:30 yesterday.

b. Jovan je preveo pesmu juče popodne u 4:30.

Jovan is translated-pfv. poem yesterday afternoon in 4:30

Intended interpretation: ‘Jovan finished translating the poem yesterday at 4:30 p.m..’

(36) a. *Jovan will translate the poem at 4:30 yesterday.

b. Jovan će prevesti pesmu sutra popodne u 4:30.

Jovan will translate- pfv.inf. poem tomorrow afternoon in 4:30

Intended interpretation: ‘Jovan will finish translating the poem tomorrow

at 4:30 p.m..’ .’

Finally, the role of the duration of the time interval with respect to which the event is ordered

predicts that instantaneous events should be possible to occur with a non-progressive and

perfective form. This prediction is borne out in both languages, as in (37) and (38).25

(37) John flipped a coin at 4:30 yesterday.

(38) Jovan je bacio novčić juče popodne u 4:30.

Jovan is flipped-pfv. coin yesterday afternoon in 4:30

‘John flipped a coin yesterday at 4:30 p.m..’

3.2.1.1 Temporally restricted past interpretations

I argue that, with past interpretations, the availability of interpretations in Serbian and

English in (35) has a structural reflex ‒ in both languages there is a Perfect component present in

the structure. With future interpretations, however, there is no Perfect component with the forms

in (36). Rather, I propose that the difference between the two languages is with respect to the

way these forms interact with the time adverbial. I show below that, although this form in

English is more restricted in terms of available interpretations, unlike Serbian, English, also

makes use of Future Perfect which compensates for the interpretations that are otherwise

available in Serbian.

Regarding the unavailability of non-progressive and perfective forms with past

interpretations in (31) and (33), respectively, I propose the following structure:

25

Note that the issue is somewhat more complex. With modifications such as ‘when the mailman arrived’ as in (15)-

(17), the event of flipping the coin can be interpreted only as occurring after mailman’s arrival and not simultaneous

with it in Serbian. This might be related to the meaning of ‘when’ which in Serbian can also mean nakon što ‘after’.

I leave this issue aside for future research.

20

(39) CP wo

UT TP eo

past AspP wo

at 4:30 p.m. yesterday afternoon AspP wo

perfective vP

[CP UT [TP Past [AspP1 at 4:30 pm yesterday afternoon [AspP Perfective [vP ]]]] =

w.∃t3[t3<tc & t3 at 4:30pm yesterday afternoonc & ∃e [ time(e) t1 & K(w)(e)]]]

In (39), past tense is, as before (cf. (26)), ordered with respect to the UT, introducing a

potentially infinite time interval preceding the UT. What is different, however, is the presence of

a time adverbial which restricts the time interval with respect to which the event time interval

needs to be ordered. Crucially, the viewpoint aspect is ordered directly with respect to the time

given by the time adverbial. Since the time interval introduced by the time adverbial is rather

short, the inclusion requirements of the perfective cannot be satisfied and the forms in (31a) and

(33a) are correctly predicted to be ungrammatical. As expected, as soon as the adverbial

introduces a longer time interval, the forms become felicitous again (cf. (24a) and (25a)).

How do we then obtain interpretations where the event is completed at 4:30 p.m., as in (35)?

Recall that these interpretations are easily available for Serbian perfectives and are also available

for some speakers of English. I suggest that these interpretations require structural modifications.

More specifically, (35) should be decomposed as to include an additional aspectual phrase,

corresponding to the participle in these contexts, and which semantically translates as Perfect. It

is precisely the Perfect component that accommodates for the relevant interpretations of the non-

progressive and perfective forms in (35). (35) thus receives the Pluperfect interpretation (see e.g.

Paunović 2001 for Serbian); although morphologically, these forms are realized as present

perfect forms in Serbian and synthetic past forms in English, their meaning corresponds to the

English example “Jovan had translated the poem at 4:30 p.m. yesterday”. I return to the

relationship between these forms and the actual Pluperfect forms in the two languages at the end

of this section.

21

I am adopting Pancheva’s (2003) definition of Perfect in (40). Perfect introduces a Perfect

Time Span (PTS) (Iatridou et al. 2001), which generalizes over time intervals that extend

backwards from the contextually salient time interval. It is integrated into the structure as in (41)

for past interpretations.

(40) ⟦PERFECT⟧ = P.t.w.∃t' [PTS(t', t) & P(t') (w)] (à la Pancheva 2003)

PTS (t', t) iff t is a final subinterval of t‘

(41) CP wo

UT TP eo

past AspP1 wo

at 4:30 yesterday afternoon AspP1 wo

Perfect AspP2

wo

perfective vP

[CP UT [TP Past [AspP1 at 4:30 pm yesterday afternoon [AspP1 Perfect [AspP2 Perfective [vP ]]]] =

w. ∃t3[t3<tc & t3 at 4:30pm yesterday afternoonc & ∃t1∃t2 [PTS (t1, t2) & t2 at 4:30

yesterday afternoonc & ∃e [ time(e) t1 & K(w)(e)]]]

As observed in the structures in (41), AspP2 specified for viewpoint aspect, i.e. the perfective,

is directly c-commanded by AspP1 headed by Perfect, which yields the time interval with respect

to which the perfective is ordered. Perfect introduces a time interval which, modified by

elements higher in the structure, i.e. the time adverbial, Tense and the reference time interval in

SpecTP, serves as the time interval with respect to which the perfective is ordered. In that

respect, the most important feature of Perfect is its property of generalizing over time intervals,

by the means of the PTS. Perfect extends backwards from the contextually salient time, which in

this case is the relatively short time interval given by the higher temporal restrictor 4:30 p.m.

yesterday. Crucially, now this extended time interval serves as the ordering time interval for the

perfective. Unlike in English (31) and Serbian (33), which do not contain Perfect, the ordering

time interval in (41) is a relatively long interval that accommodates for the requirements of the

perfective to be satisfied. The modified structure correctly predicts the availability of this form in

22

Serbian and English under certain interpretations, even with time adverbials that is a rather short

interval — the crucial difference is that the time adverbial is the ordering time interval for the

Perfect, rather than for the perfective.

There are further predictions given the structure in (41): the event of translating can be

completed either at 4:30 p.m. (since this moment marks the right boundary of the Perfect Time

Span) or it can be completed before that, since the requirements of the perfective are met as long

as the event time interval is included in the extended time interval. The latter interpretation is

indeed attested for in Serbian (with a modification including the adverbial ‘already’):26

(42) Juče popodne u 4:30, Jovan je već preveo pesmu.

yesterday afternoon in 4:30 Jovan is already translated- pfv. poem

Intended interpretation: ‘Jovan had already translated the poem yesterday at 4:30 p.m..’

The availability of the interpretation in (42) provides additional motivation for the presence

of Perfect in the structure. More specifically, this is the same interpretation that is otherwise

obtained by the form that corresponds to the English Pluperfect, as illustrated in (43).

Interestingly, (43) also has the interpretation where the event was completed at 4:30 p.m., which

is a welcome outcome of the proposal outlined in this section. Namely, the Perfect component is

predicted to allow for both anteriority reading and the reading where the interval denoted by the

time adverbial overlaps with the final boundary of the event time.27

(43) Jovan je bio preveo pesmu juče popodne u 4:30.

Jovan is been translated-pfv. poem yesterday afternoon in 4:30

’Jovan had already translated a poem yesterday at 4:30 p.m..’

‘John finished translating a poem yesterday at 4:30 p.m..’

There is a decline in use of the pluperfect form in (43), it becoming more and more archaic in

Serbian. Instead, the interpretations are more commonly obtained with the periphrastic past form.

If past participle of Be in (43) is an overt realization of Perfect and if the interpretations assigned

26 It remains to be established whether this adverbial affects the reference time interval and somehow contributes to

the availability of the perfective and the interpretation in question. 27

I thank an anonymous reviewer for pointing out that the latter interpretation is also attested for. Note, however,

that Pluperfect form is extremely rarely used nowadays, with the periphrastic past form being more common.

Pluperfect is more commonly used for events that occurred in distant past or for expressive anteriority. It is never

the first choice for the events that finished at a particular point (4:30 p.m. in this case).

23

to (43) can also be obtained with the form used in (42) and (33a), this might be taken as an

indication that a Perfect component is indeed present in their structure. 28

Interestingly, with English Past Perfect in (44), the non-progressive form is available.

Moreover, for some speakers, only this form, and not the non-progressive synthetic form, allows

for the interpretation where the event is completed at 4:30 p.m..

(44) John had translated a poem yesterday at 4:30 p.m..

I suggest that, for those speakers, the Perfect component is present only with Pluperfect forms,

extending the ordering time interval for the matrix Aspect, and allowing for the requirements of

perfective aspect to be satisfied. Since the same interpretation can be obtained with Serbian

periphrastic past, I argue that there is a Perfect component with these interpretations in Serbian,

but the component is less and less overtly present. Similarly, for English speakers for whom

Pluperfect is highly marked, and the ‘finishing’ interpretation can be obtained with non-

progressive synthetic past, I assume that there is a Perfect component in the structure (as in (41)),

which accommodates for this interpretation.29

28 One possibility would be that the double participle construction is replaced by the morphologically simpler

present perfect, which, however, retains the past perfect interpretation. 29

According to some speakers, (33a) (repeated below in (i)) can also receive the inceptive interpretation, i.e. 4:30

p.m. marks the beginning of the interval within which the event of eating is contained. However, an anonymous

reviewer observes that this interpretation is not available for all speakers and, in addition, that it is not available for

all types of predicates. More specifically, the reviewer reports that for some speakers this interpretation is

unavailable with the predicate 'translated a novel yesterday afternoon at 4:30', but might otherwise be available

with, e.g. a predicate, such as ‘eat an apple’, while again being restricted with a minimally different predicate ‘eat

10 kg of meat’. Indeed, the duration of the event does seem to affect the availability of this reading. Furthermore, the

reviewer observes that a relatively short event, such as ‘eating an apple’, is more permissive under that interpretation

than a longer event, such as ‘eating 10kg of meat’, which might indicate that the inceptive interpretation is an

illusion. Another reviewer is concerned that, even if this interpretation is available, it might be available due to

pragmatic coercion and not a ‘true’ reading of the sentence. Given the speaker-variations and given that different

predicates might affect its availability, it thus remains to be determined more precisely to which extent this reading

is attested in the language.

One possibility to account for the availability of the reading under the current analysis, would be to assume that

this interpretation, like the Perfect, takes the reference time (4:30 p.m.) as a boundary of the interval within which

the event time is included. In other words, under the inceptive interpretation, the reference time would be the left

boundary of an extended interval which needs to include the event time. However, the inclusion would not be

enough: the onset of the event time needs to overlap with 4:30 p.m.. Although at this point I cannot specify the

details of the temporal analysis, the important point is that u 4:30 requires the overlap with a boundary of the event

time interval.

However, such an analysis would not make different predictions regarding the availability of the reading

depending on different types of predicates. More specifically, if (i) contained ‘novel’ instead of ‘poem’, one could

argue that, for those speakers who do not allow inceptive interpretation, this interpretation is excluded due to the

24

3.2.1.2 Temporally restricted future interpretations

Recall that future forms in English are restricted under the interpretation where John was in the

process of translating the poem at 4:30 p.m., as in (45).

(45) a. *John will translate a poem at 4:30 p.m. tomorrow.

b. John will be translating a poem at 4:30 p.m. tomorrow.

The proposed compositional analysis accounts for the lack of interpretations above in the

following way: present tense introduces an interval corresponding to the UT, and although the

woll component extends that interval (cf. (29)), the presence of an adverbial restricts the time

interval with respect to which the event time interval needs to be ordered, as in (46). Crucially,

the viewpoint aspect is ordered directly with respect to the time given by the time adverbial.

Since the time interval introduced by the time adverbial is rather short, the inclusion

requirements of the perfective cannot be satisfied and the form in (45a) is correctly predicted to

be excluded.

length of the process of translating a novel. In other words, it is difficult to imagine someone starting the translation

at 4:30 yesterday and reaching the end point by the UT. The reviewer observes that, if the short time period for

translation is at stake here, we can easily test the availability of this reading if the translation starts, e.g. a year ago,

because there is enough time to complete the translation, and duration of the process should not be the interfering

factor. The reviewer reports that, for his or her consultants, this reading is still unavailable. This is, however, not

predicted to matter in the current analysis. Given the complexity of the issue, I will not pursue the analysis of this

interpretation any further, confining it for further research which would need to define precisely the level of

acceptability of this reading both in terms of degree of speaker-variations and in terms of the availability of the type

of predicate.

(i) Milan je preveo pesmu juče popodne u 4:30.

Milan is translated- pfv. poem yesterday afternoon in 4:30

Intended interpretation: ‘Milan started translating the poem yesterday at 4:30 p.m..’

25

(46) CP wo

UT TP wo

present wollP wo

woll AspP wo at 4:30 p.m. tomorrow AspP wo

perfective vP

⟦CP UT [TP Present [wollP woll [AspP at 4:30 p.m. tomorrow [AspP Perfective [vP ]]]⟧ =

w. ∃t2[t2=tc & ∃t1 [t1 > t2 & t1 at 4:30 p.m. tomorrowc & ∃e [ time(e) t1 & K(w)(e)]]]

Turning now to Serbian, recall that future perfective forms in Serbian in (34a) can receive the

interpretation where the event is completed at 4:30 p.m., as repeated in (47). This interpretation

is not available with future forms in English in (45). On the other hand, Serbian forms in (47b)

are excluded under the interpretation where the event is in progress at 4:30 p.m., similarly to

their English counterparts.

(47) a. Jovan će prevoditi pesmu u 4:30 sutra popodne.

Jovan will translate- impfv.inf. poem in 4:30 tomorrow afternoon

Intended interpretation: ‘Jovan will be translating a poem tomorrow at 4:30 p.m..’

b. Jovan će prevesti pesmu u 4:30 sutra popodne.

Jovan will translate- pfv.inf. poem in 4:30 tomorrow afternoon

‘Jovan will finish translating a poem tomorrow at 4:30 p.m..’

*‘Jovan will be translating a poem tomorrow at 4:30 p.m..’

Note that, with additional modifications, the example can also receive the anteriority

interpretation:

(48) Sutra u 4:30 Jovan će već prevesti pesmu.

tomorrow in 4:30 Jovan will already translate-pfv.inf. poem

Intended interpretation: ‘At 4:30 p.m. tomorrow, Jovan will have already translated

the poem.’

26

At this point, it seems tempting to posit the Perfect component in the structure of future forms

in Serbian, in order to account for the differences in the interpretation between Serbian and

English. If we take the participle to be the overt realization of Perfect, such an option would not

be entirely uncalled for, since the participle in future interpretations is morphologically visible in

the Kajkavian dialect of Croatian, as in (49a). However, two issues would arise under such an

approach. First, note that, although Croatian and Serbian seem to behave on par with respect to

restrictions on the perfective discussed in the paper, Serbian and major parts of Croatian are part

of the Shtokavian dialect which is different from the Kajkavian dialect in a number of linguistic

properties, and drawing parallels between the two might prove to be problematic. Second, and

this seems to be a more serious problem, under the approach that posits Perfect in the structure,

we would expect participles to also occur with Serbian future forms. However, this is only

possible in temporal and conditional clauses, as illustrated in (49b). In matrix contexts, we only

find periphrastic forms consisting of Auxiliary will and infinitive (cf. (4)), which is is puzzling if

there is a Perfect component in the structure of these forms.30

(49) a. -Budeš došla sutra? -Budem. (Kajkavian/Zagreb-Croatian)

be-2.sg.pres come-past.part. tomorrow? be-1.sg.pres

‘Will you come tomorrow?’ ‘I will.’

b. Kada/ako budem došla, reći ću ti. (Serbian)

when/ if be-1.sg.pres come-past.part. say-inf. will you-cl.acc

‘When/ If I have arrived, I will tell you.’

Thus, rather than positing Perfect in the structure of these forms, I propose that English

future forms in (45) and Serbian future forms in (47) are structural equivalents. However, the

difference stems from the way these forms interact with the adverbial.

30 I thank an anonymous reviewer for a discussion on this issue. The reviewer brings up another problematic side of

such an approach. If Perfect was present in the structure with future forms, we would expect those forms to have

Perfect effects. Namely, the Perfect in Serbian has perfect uses (‘Direktor je trenutno izašao.’ lit. ‘The director has

currently left.’ i.e. ‘The director is currently away.’). This type of interpretation is, however, not licensed by with

Future imperfectives (‘Čitaću knjigu u 4:30.’ – only ’I’ll be reading the book at 4:30.’, not ’I’ll have been reading

the book at 4:30.’), while imperfective verbs in the Perfect do get the Perfect interpretation (A: ’Šta ti je?’ ’What’s

wrong with you?’, B: ’Pio sam.’ ‘I’ve been drinking.’). This effect is, however, present with future perfectives. The

reviewer suggest that, with future perfectives, rather than Perfect being responsible for the Perfect-like effect, it is

more likely that it is the perfective component of the verb’s semantics that triggers that effect. I agree with the

reviewer that the presence of this effect with Future perfectives can actually stem from the properties of perfective

itself, and not from the properties of Perfect. Otherwise, it would be difficult to explain why the effect is absent with

future imperfectives.

27

Recall that, according to the structure proposed in (46), perfective cannot satisfy its inclusion

requirements because the time interval introduced by 4:30 p.m. is too short for the lengthy event

to be included in it. This excludes the interpretation where the event is in progress at 4:30 p.m. in

both English and Serbian. However, suppose that 4:30 p.m. does not necessarily restrict the time

interval introduced by woll, but it can also mark its right boundary. In other words, under such a

scenario, the interval introduced by woll would not be restricted to being short, it would just be

delimited by 4:30 p.m.. This is reminiscent of the structure of past interpretations in Serbian,

where 4:30 p.m. marks the right boundary of the interval introduced by Perfect. If the same holds

with future interpretations, then it should in principle be possible to include the event within this

time interval. I propose that this is exactly the case with future perfectives in Serbian, i.e. 4:30

p.m. marks the right boundary of the time interval introduced by woll and it also overlaps with

the final boundary of the event time interval. This derives the interpretation where the event is

completed at 4:30 p.m.. One further prediction is that, if the event is included in time interval

introduced by woll and delimited by 4:30 p.m., it should also be possible to obtain the anteriority

reading, i.e. the event being completed prior to 4:30 p.m.. However, the adverbial u 4:30 ‘at

4:30’ seems to require the overlap with the event boundary. Unless there is additional

modification (as in (48)), the anteriority reading is not attested, as shown in (50). Instead, the

anteriority reading can be expressed with the time adverbial do 4:30 ‘by 4:30’, as in (51).

(50) Jovan će prevesti pesmu sutra u 4:30.

Jovan will translate- pfv.inf. poem tomorrow in 4:30

*‘By 4:30 p.m. tomorrow, Jovan will have already translated the poem.’

(51) Jovan će prevesti pesmu do 4:30 sutra popodne.

Jovan will translate-pfv.inf. poem by 4:30 tomorrow afternoon

‘By 4:30 p.m. tomorrow, Jovan will have already translated the poem.’

Finally, note that, for some speakers, the example in (50) receives an inchoative

interpretation. The event begins at 4:30, which fits well with the observation that 4:30 requires

the overlap with the event boundary.31

How can this interpretation be derived from the structure?

I propose that, under such an interpretation, 4:30 is adjoined higher in the structure, i.e. to wollP,

as in (52).

31

Again, the same set of restrictions with the inchoative past interpretation might also be operative with future

interpretations (cf. fn. 29).

28

(52) CP wo

UT TP wo

present wollP wo

at 4:30 p.m. tomorrow wollP wo woll AspP wo

perfective vP

In this structure, 4:30 marks the left boundary of the time interval introduced by the woll

component and the perfective event is included within the interval, with the onset of the event

overlapping with 4:30. The idea of different adjunction site of the adverbial being responsible for

different interpretations is not novel and it has been advocated for English in Klein (1992) and

Demirdarche and Uribe-Exterrbaria (2004), inter alia. To illustrate, Demirdache & Uribe-

Extebarria observe for English Past Perfect, as in Maddi had left school at 5 p.m. (2004:157, ex.

(21)) that the example is ambiguous between Maddi’s leaving occurring at 5 p.m. or it occurring

prior to 5 p.m. In Demirdache & Uribe-Extebarria's framework, the ambiguity results from the

adverbial modifying the event time (which derives the reading where the event is completed at

particular reference time) or the assertion time (which derives the reading where the event is

completed prior to particular reference time), and the same would work for Serbian future forms.

In other words, different readings in Serbian would be obtained by different adjunction sites of

the adverbial.

Note, however, that English future forms in (45a) do not cover the same range of

interpretations as their Serbian counterparts. We already noted that, in addition to not being in

progress as 4:30 p.m., they also cannot receive the interpretation of being completed at 4:30

p.m.. The latter reading would in principle be possible if ‘at 4:30 p.m.’ marks the right boundary

of the interval introduced by woll. But the adverbial ‘at 4:30 p.m.’ in English does not interact

with future form in (45b) in such a way that it marks the final boundary of the event. Instead, my

English consultants report that the only interpretation that they can assign to (45b) is the one

where the event begins at 4:30 p.m.. Under such an interpretation, this English example would

29

also have the structure in (52), with 4:30 p.m. marking the left boundary of the interval

introduced by woll within which the event would be included, and whose onset would overlap

with 4:30 p.m., as dictated by the adverbial.

Finally, note that the interpretation where the event is completed at 4:30 p.m. in English is

expressed by Future Perfect and with the adverbial ‘at 4:30 p.m.’, whereas the same form occurs

with the adverbial ‘by 4:30 p.m.’ to derive the anteriority reading. These examples are parallel to

Serbian (50) and (51). Thus, English and Serbian can express the same range of interpretations,

the difference being that Serbian uses one form to express the interpretations where the event is

completed at 4:30p.m., prior to 4:30 p.m., or where the event begins at 4:30 p.m., whereas

English divides these interpretations between Future Perfect and Future Simple.

(53) John will have translated a poem tomorrow at 4:30 p.m..

(54) John will have translated a poem tomorrow by 4:30 p.m..

As an interim summary, I have provided a structural temporal analysis which accounts for

the incompatibility of the UT interpretation and perfective aspect of eventive predicates in matrix

clauses in Serbian, and for the lack of restrictions on the perfective aspect with past and future

interpretations. In order for the event temporal location to be successful, the ordering time

interval for Aspect cannot be too short. In the case of the UT interpretation, the requirements of

Tense can be satisfied, but the time interval with respect to which Aspect is ordered is too short

for the event specified for the perfective to be included within it, and the event cannot be

temporally located.

With past and future interpretations, past tense and the woll component, introduce a

potentially indefinitely long time interval with respect to which Aspect in evaluated; hence, the

perfective form can be realized. The parallel distribution of non-progressive forms of the

eventive predicates in English provides additional support for the claim that these forms are

specified for the perfective aspect. Furthermore, the distribution of perfective aspect in past

contexts with overt time adverbials in English and Serbian indicates the presence of an additional

Perfect component. Perfect extends the ordering time interval for the perfective aspect. The

difference in the distribution of perfective aspect in future contexts with overt time adverbials in

English and Serbian stems from the difference in the way these forms interact with the time

adverbial. Although this form in English is more restricted in terms of available interpretations,

30

English also makes use of Future Perfect which compensates for the interpretations that are

otherwise available with Serbian future perfectives.

4. Embedded clauses

This section extends the analysis of the incompatibility of present tense and perfective aspect

in matrix clauses to the incompatibility in embedded present tense contexts in Serbian. Unlike in

matrix clauses, in embedded contexts, perfective aspect with present tense is not always

precluded: although the perfective is not available in the complements of verbs like believe (as in

(55)), complements of verbs like want allow for perfective aspect (as in (56)).32

(55) a. Verujem da Jovan prevodi pesmu.

believe-1.sg.pres that Jovan translates-impfv. poem

Intended interpretation: ‘I believe that Jovan is translating a song (right now).’

b. *Verujem da Jovan prevede pesmu.

believe-1.sg.pres that Jovan translates-pfv. poem

Intended interpretation: ‘I believe that Jovan has translated a poem (just now).’

(56) a. Želim da sutra popodne prevodim pesmu.

want-1.sg.pres that tomorrow afternoon translate- impfv.1.sg.pres. poem

‘I want to be translating a poem tomorrow afternoon.’

b. Želim da sutra popodne prevedem pesmu.

want-1.sg.pres that tomorrow afternoon translate- pfv.1.sg.pres. poem

‘I want to translate the entire poem tomorrow afternoon.’

I propose that differences in the availability of perfective aspect in embedded present tense

contexts are the result of a difference in the ordering time interval for the embedded Aspect. In

the case of (55), the ordering time interval is relatively short, and the inclusion requirement of

the perfective cannot be satisfied. On the other hand, complements of verbs like want extend the

ordering time interval for the embedded perfective aspect, allowing for its requirements to be

met. This difference, I argue, is the result of structural configurations of two types of

complements.

32

The example (55b) is felicitous under the interpretation where the embedded predicate distributes over a plurality

of situations. Similar to cases in (1b), such interpretation seem to be felicitous only in the presence of an overt QP,

e.g. svaki put ‘every time’ (or if QP is present in the previous context).

31

4.1 Matrix indexicality of embedded present tense in English and the lack thereof in Serbian

In complex clauses in Serbian, the embedded complement is introduced by the complementizer

da and the embedded verb which has present tense morphology.33

Unlike in English, the

embedded present in Serbian is not matrix indexical, i.e. it does not necessarily refer to the UT,

but its interpretation is determined relative to the temporal location of the matrix verb.34

This is

illustrated in the contrast between (57) and (58). If the embedded present tense in English needs

to include the UT, then we correctly predict (57a) to be odd, since it conflicts with the general

knowledge of the duration of a pregnancy: Mary cannot be pregnant at the point when Sally

hears about it and also two years after that. In order to obtain the intended interpretation, the

embedded clause would need to contain the past tense, as in (57b).35

Due to the Sequence of

Tense (henceforth SOT) property of English, i.e. the semantically vacuous interpretation of a

morphologically realized tense, the embedded past allows for the overlapping temporal

interpretation of the embedding and embedded predicate (in addition to the anteriority

interpretation).36

In Serbian, embedded present tense is not matrix indexical and it can be used in

the environment which is otherwise infelicitous in English, as illustrated in (58a). Regarding the

embedded past, like in English (57b), embedded past tense in Serbian (58b) is ambiguous

between the two interpretations: (1) either Jovan found out about Marija’s pregnancy during the

time while she was pregnant or (2) Jovan found out that Marija had been pregnant at some point

prior to that moment. 37,38

33 The present tense form can be replaced with the infinitive, but only in complements of verbs like want and only

when those verbs are restructuring verbs. Propositional complements can also contain past and future forms. I return

to their distribution in section 4.2. 34

For a discussion of English matrix indexicality, see Enç (1987), Abusch (1988 et seq.), Ogihara (1996), Schlenker

(1999). 35

As an anonymous reviewer correctly points out, if the matrix verb is in the Present Perfect or in the future,

embedded present tense can be interpreted either at the UT or at some other reference time:

(i) John will say that he’s tired.

The reviewer is, of course, correct that in present-under-present contexts (recall that finite future is composed of

present tense and the woll component in English), SOT can apply, hence removing/inactivating the embedded

present tense. 36

For a discussion of Sequence of Tense effects in English, see Enç 1987, Dowty 1982, Abusch 1988, 1997,

Ogihara 1996, inter alia. 37

I would like to thank an anonymous reviewer for drawing my attention to the availability of the first interpretation

in Serbian. For my consultants, this interpretation seems not to be the first choice with the embedded past but is

facilitated in the presence of the overt adverbial which forces the simultaneous interpretation, as in (i):

32

(57) a. #Two years ago, Sally found out that Mary is pregnant. (Sharvit 2003a)

b. Two years ago, Sally found out Mary was pregnant.

(58) a. Pre dve godine, Jovan je čuo da je Marija trudna.

before two years Jovan is heard that is Marija pregnant

‘Two years ago, Jovan heard that Marija was pregnant.’

b. Jovan je čuo da je Marija bila trudna.

Jovan is heard that is Marija been pregnant.

‘Jovan heard that Marija had been pregnant.’

’Jovan heard that Marija was pregnant.’

Given the lack of matrix indexicality of the embedded present tense in Serbian, the language

should, in principle, allow for various temporal relation with respect to the matrix predicate. This

is indeed the case: Depending on the lexical properties of the embedding predicate, the

embedded tense is interpreted either as: 1) simultaneous with the matrix event time (I refer to

these complements as propositional complements); or 2) posterior with respect to the matrix

event time (I refer to these complements as future irealis complements). Crucially, in both cases,

the morphological tense is present tense. Regarding the distribution of the perfective with the

embedded present, it is not allowed in propositional complements (as in (55)), while it is allowed

in irrealis complements (as in (56)).

In what follows, I argue that the availability of the perfective follows from the structural

difference between propositional and future irrealis complements. I suggest that for both

complement types, the reference time interval for the embedded Tense is a relatively short time

interval. However, irrealis complements involve a future component, woll, which extends the

time interval for the ordering of the embedded Aspect, allowing for the requirements of the

perfective to be satisfied. Propositional complements, on the other hand, do not introduce such a

(i) Jovan je čuo da je Marija u tom trenu bila trudna.

Jovan is heard that is Marija at that point been pregnant

‘Jovan heard that Marija was pregnant at that point.’

38

Sharvit (2003a) proposes that matrix indexicality and SOT effects are correlated in a language. However, in order

to make the same correlation for both Serbian and English, one would need to consider the effects of aspect. It

should be noted that, although English is typically considered to be an SOT language and Serbian a non-SOT

language, in both English and Serbian, statives pattern with imperfective/progressive eventives in allowing for the

vacuous interpretation of embedded past, whereas perfective/non-progressive eventives allow only for the anteriority

interpretation. Thus, the SOT effects are more complex than they initially might appear to be (the lack of

simultaneous interpretation of perfective eventives is, however, compatible with the analysis proposed in the paper).

Although I will not go into further discussion here, this issue definitely presents a promising area for future research

(see, e.g. Portner 2003; for a recent discussion, see Bjorkman 2015). I to thank the two anonymous reviewers for a

discussion on the effects of aspect in English and Serbian.

33

component, and the ordering time interval for the perfective aspect remains relatively short.

Consequently, the perfective form is attested for in irrealis complements, but not in propositional

complements.

The distribution of perfective aspect in the two type of complements patterns with the

distribution of non-progressive eventive predicates in propositional and irrealis future infinitival

complements in English as observed in Wurmbrand (2014). I briefly sketch her analysis in

section 4.4.

4.2 Propositional complements in Serbian

In propositional complements in English, i.e. clauses embedded under the verbs like believe,

as in (59), the matrix verb does not introduce any presuppositions regarding the time of the

realization of the embedded predicate; it does not impose any requirements on it. According to

the line of research following Stowell (1982) (Chomsky and Lasnik 1993, Bošković 1996, 1997,

Martin 1996, 2001), propositional complements are tenseless, receiving the simultaneous

interpretation with the matrix verb.39,40

In what follows I will assume the presence of a Tense

projection in propositional complements in Serbian.

(59) Tom believes Mary to be finishing her homework.

In propositional complements in Serbian, embedded present tense can either be interpreted

simultaneously with the matrix verb, as in (60a) or it can receive a future-oriented interpretation

with a future adverbial, as in (60b). The lack of a past-oriented interpretation of the embedded

present is illustrated by the impossibility of past adverbials as in (60c). Crucially, all the

embedded verbs are specified for imperfective aspect.

39

However, Wurmbrand (2014) provides a number of arguments for propositional complements containing the zero

tense which is responsible for the simultaneous interpretation with the matrix predicate and is involved in the

distribution of eventive predicates. I return to her analysis of English data in Section 4.4. 40

Note that the interpretation of (59) can be further shifted back by an additional Perfect component, as in (i), which

is compatible with either tensed or tenseless approach:

(i) Tom believes Mary to have finished her homework.

34

(60) a. Jovan veruje da dobro radi domaći.

Jovan believes that well does-impfv. homework .

‘Jovan believes that he is doing his homework correctly.’

b. Jovan veruje da mu sutra dolazi rodbina.

Jovan believes that he-dat.cl. tomorrow arrives-impfv. relatives.

‘Jovan believes that his relatives are arriving tomorrow.’

c. *Jovan veruje da mu juče dolazi rodbina.

Jovan believes that he-dat.cl. yesterday arrives-impfv. relatives.

Jovan believes that his relatives were arriving yesterday.’

Regarding the perfective aspect in embedded present tense contexts, propositional

complements pattern with matrix present tense clauses that receive the UT interpretation ‒ the

perfective is prohibited (cf. (55b)). This is illustrated in (61a) for simultaneous interpretation, in

(61b) for future-oriented interpretation, and in (61c) for past-oriented interpretation.

(61) a. *Verujem da Jovan prevede pesmu.

believe-1.sg.pres that Jovan translates-pfv. poem

Intended interpretation: ‘I believe that John has translated a poem (just now).’

b. *Jovan veruje da mu sutra dođe rodbina.

Jovan believes that he-dat.cl. tomorrow arrives-pfv. relatives.

‘Jovan believes that his relatives will arrive tomorrow.’

c. *Jovan veruje da mu juče dođe rodbina.

Jovan believes that he-dat.cl. yesterday arrives-pfv. relatives.

Jovan believes that his relatives arrived yesterday.’

Focusing first on simultaneous interpretation in (61a), I propose that, given the lack of the

matrix indexicality of the embedded present in Serbian, the reference time interval for

propositional complements is not the UT, but the contemporary now of the attitude holder

(henceforth the contemporary now), in the sense of Abusch (2004) (to be specified immediately).

However, similar to the UT in matrix present tense clauses, the contemporary now is considered

to be a relatively short point in time. Since the contemporary now serves as the reference time

interval for Aspect, it prevents the requirements of perfective aspect from being satisfied.

Abusch’s (2004) contemporary now can be defined as a point in time which the attitude

holder believes to be his/her now. In (62), the contemporary now is compatible with John’s

beliefs at 11 a.m. Thus, the contemporary now is not 10 a.m., the Topic Time (Klein 1994), i.e.

the actual time for which the statement is confined, but rather 11 am, the time which John

believes to be the current time — his now.

35

(62) At 10am yesterday, John believed it to be 11 a.m.

Regarding the temporal interpretation of propositional infinitives in English, Abusch argues

that these complements receive an interpretation simultaneous with the contemporary now. In

(63a), Guido is at Monique’s place at the time compatible with the time of speaker’s beliefs, i.e.

the UT. The examples in (63b) and (63c) illustrate that the embedded infinitive cannot receive an

interpretation inconsistent with the contemporary now.

(63) a. Guido is believed to be at Monique's place (now). (Abusch 2004:29)

b. *Guido is believed to be at Monique's place last night.

c. */ok Guido is believed to be at Monique's place tomorrow night.

Regarding the structural composition of propositional complements in English, Abusch’s

structure for the example in (64), corresponds to (65).

(64) John believes Monique to love Solange.

(65) IP:believes (n,j, λn love (n,m,s)) wo

Pres:n VP: λt believe (t, j, λn love (n,m,s)) wo John V’ wo

believes CP: λn love (n,m,s) wo

λ TP: love (n, m, s) wo

T:n VP: λt love (t, m, s) wo

Monique V wo

love Solange

These complements contain a variable n, a temporal argument in the T node. This variable is

bound by a -operator in C. The -operator and the temporal variable together create a property

of times, a licit argument of intentional verbs like believe. Note that the embedding verb also

36

contains the occurrence of a temporal variable n, which is allowed by the system, since Abusch

assumes that “lexical entries for tenses have free occurrences of n” (2004:51). She also argues

that, given the lexical semantics of attitude predicates, the time argument of the attitude verb

serves as a counterpart of the attitude time. In other words, the time argument of the matrix

predicate is the contemporary now. Importantly, the embedded event can also be interpreted at

the contemporary now: The -operator is under the scope of the matrix verb, and it binds a time

variable n in the embedded TP, accommodating for the embedded event to be interpreted at the

time of believing.41

In the analysis of propositional complements in Serbian, I follow Abusch in positing a -

operator which binds a temporal component n. This variable, however, is in the CP domain of a

clause, rather than in TP; the embedded TP is specified for present tense.42

The -operator, like

in English, contributes the contemporary now interpretation of the complement. All components

put together, the structures containing the illicit perfective as in (66) is represented in (67).

(66) *Marija veruje da Jovan prevede pesmu.

Marija believes that Jovan translates-pfv. poem

Intended interpretation: ‘Marija believes that Jovan has translated a poem (just now).’

41

The example Abusch provides is John believes Monique loves Solange. This, however, does not affect the

analysis since both finite or non-finite complement receive the interpretation simultaneous with the embedding verb. 42

In Abusch’s analysis of John believes Monique loves Solange, the embedded tense is specified for [present].

37

(67) TP wo UT T’ wo Pres VP wo

Marija V’ wo

believes CP wo

λ n TP wo Pres AspP wo

perfective VP wo

Jovan V’ wo

translates poem

⟦TP UT [T’ Pres [VP Mary believe [CP λn [ TP Pres(n) [AspP Perfective [vP Jovan translate

poem]]]⟧=[∃t2[t2=n & believe(n)(m) [λnλw∃t1[t1=n & ∃e[time(e) t1 & translates poem (j)(w)(e)]]]]

Restrictions on perfective aspect are the result of the contemporary now serving as the

ordering time interval for the embedded Aspect. Given the simultaneous interpretation of the

matrix and embedded predicate, the ordering time interval for the embedded Tense is the

contemporary now. The embedded present requires to be identical with the contemporary now;

hence, the contemporary now becomes the ordering time interval for the embedded Aspect. If the

contemporary now is a near-instantaneous interval, then the inclusion requirements of the

perfective cannot be met, and, as a result, the form is ungrammatical.

Regarding the effects of matrix tense on the interpretation of embedded perfective, we

predict that the specification of the matrix Tense should not affect the availability of the

embedded perfective it will always be unavailable. More specifically, if the ordering time

interval for the embedded Tense is the contemporary now, then present tense complements will

always receive the interpretation simultaneous with it, and the perfective will be unavailable,

regardless of the matrix Tense specification. The prediction is borne out, as illustrated in (68b)

and (69b). Like in the cases of (66), the ordering time interval for the embedded Aspect is the

38

contemporary now, and the perfective is correctly predicted to be unavailable. As expected, the

imperfective is available ((68a), (69a)).

(68) a. Verovao sam da Jovan prevodi pesmu.

believed-sg.masc am that Jovan translates-impfv. poem

Intended interpretation: ‘I believed that Jovan was translating a poem right then.’

b. * Verovao sam da Jovan prevede pesmu.

believed-sg.masc am that Jovan translates-pfv. poem

Intended interpretation: ‘I believed that Jovan finished translating a poem (at that

point).’

(69) a. Milan će verovati da Jovan prevodi pesmu.

Milan will believe-inf. that Jovan translates-impfv. poem

Intended interpretation: ‘Milan will believe that Jovan is translating a poem right

then.’

b. *Milan će verovati da Jovan uradi domaći.

Milan will believe-inf. that Jovan does-impfv. homework

Intended interpretation: ‘Milan will believe that Jovan will have translated a poem

(right then).’

On the other hand, the specification of the embedded tense is predicted to affect the

availability of the embedded Aspect. Although the ordering time interval for the embedded

Tense is always the contemporary now, an additional temporal operator in the embedded clause

would extend the ordering time interval for the embedded Aspect, and the perfective should be

available. This is indeed the case in embedded past and future contexts in (70) and (71). The

relevant structures are given in (72) and (73).43,44

(70) Milan veruje da je Jovan prevodio/ preveo pesmu.

Milan believes that is Jovan translated-impfv. / translated-pfv. poem

‘Milan believes that Jovan was translating a poem.’ (imperfective)

‘Milan believes that Jovan translated a poem.’ (perfective)

(71) Milan veruje da će Jovan prevoditi/ prevesti pesmu.

Milan believes that will-3.sg Jovan translate-impfv.inf. / translate-pfv.inf. poem

‘Milan believes that Jovan will be translating a poem.’ (imperfective)

‘Milans believes that Jovan will translate a poem.’ (perfective)\

43

Importantly, only the embedded present yields the interpretation simultaneous to the one of the matrix verb,

whereas embedded past can only be interpreted anterior and future posterior with respect to the contemporary now. 44

Other verbs that pattern with the verb verovati ‘to believe’ are osećati ‘to have a feeling’, znati ‘to be aware’,

aspectual verbs like počinjati ‘to begin’ and prestajati ‘to stop’, etc.

39

(72) TP wo UT T’ wo

Pres VP wo

Milan V’ wo

believe CP wo λ n TP wo

Past AspP1 wo

Perfect AspP2 wo

perfective/imperfective VP

Jovan translate poem

⟦TP UT [T’ Present(n) [VP Milan believe [CP λn [ TP Past(n) [AspP Perfect [AspP Perfective [vP Jovan

translate poem]]]⟧ = [∃t3[t3 =n & believe(n)(m) [λnλw ∃t2[t2 <n & ∃t1 [PTS (t1, t2) &∃e[ time(e)

t1 & translate poem (j)(w)(e)]]]]]]

40

(73) CP wo UT T’ wo

Pres VP wo

Milan V’ wo

believe CP wo λ n TP wo

Present wollP wo

woll AspP wo

perfective/imperfective VP

Jovan translate poem

⟦CP UT [T’ Present(n) [VP Milan believe [CP λn [ TP Pres(n) [wollP woll [AspP Perfective [vP Jovan

translate poem]]]⟧=[∃t2[t2=n& believe(n)(m) [λn ∃t1[t1 >n & ∃e[ time(e) t1 & translate poem

(j)(w)(e)]]]]

Finally, a note is in order regarding the assymetry between the imperfective and the

perfective with respect to the availability of future-oriented interpretation. As indicated by the

contrast between (60b) and (61b), only embedded imperfective verb can receive future-oriented

interpretation with morphological present tense. But this seems not to be predicted by the current

analysis: if there is an extension of the time interval in the embedded clauses that receive future-

oriented interpretation, nothing in principle should prevent perfective aspect from occurring in

these complements, which would incorrectly predict them to be felicitous. However, suppose the

following scenario: there is no extension of the time interval with these complements, i.e. there is

no structural component contributing future orientation in these complements. Rather, future

orientation would only be contributed by the adverbial, which, in the the absence of syntactic

component, cannot extend the time interval within which the perfective would be included.

Regarding the perfective, since there is no extension of the time interval, its requirements would

not be satisfied, preventing the form from occuring in these complements. On the other hand, the

41

duration of the time interval should not affect the availability of the imperfective. As long as the

event marked for imperfective includes the reference time interval, the duration of the latter is

irrelevant.45,46,47

45 Another possibility would be to claim that, more generally, the perfective can receive future interpretation only if

this future interpretation is syntacically anchored in some way. This idea receives further support from the

distribution of aspect with future-oriented interpretation in matrix clauses. In Serbian, matrix clauses receiving

future interpretation are most commonly expressed with a periphrastic form, as in (i) and (ii), in which case, there

are no aspectual restrictions (see also the examples discussed in section 3.2.). In addition, it is also possible to use

morphological present, as in (iii) and (iv). Crucially, morphological present forms are compatible only with

imperfective aspect (iii), while being incompatible with perfective aspect (iv).

(i) Milan će jesti jabuku.

Milan will eat- impfv.inf. apple

‘Milan will be eating an apple.’

(ii) Milan će pojesti jabuku.

Milan will eat- pfv.inf. apple

‘Milan will have eaten an apple (and finish it).’

(iii) *Položim ispit sutra

pass-pfv.1.sg pres. exam tomorrow

’I will pass the exam tomorrow.’

(iv) Polažem ispit sutra.

take- impfv.1.sg.pres. exam tomorrow

’I am taking the exam tomorrow.’

Interestingly, perfective verb can receive future interpretation in (v), which is the equivalent of an antecedent of a

conditional, as indicated with an overt complementizer in (vi). In order to maintain the idea that future

interpretations require syntactic anchoring of some kind, one would need to assume that there is a component in the

structure of conditional clauses that licenses future interpretation (Kaufmann 2005, Romero 2014, inter alia).

(v) Položim ispit sutra i na konju sam!

pass- pfv.1.sg.pres. exam tomorrow and on horse am

’If/when I pass the exam tomorrow, I am good!’

(vi) Ako položim ispit sutra, na konju sam!

if pass- pfv.1.sg.pres. exam tomorrow on horse am

’If/when I pass the exam tomorrow, I am good!’

The idea of a required syntactic anchoring for perfective aspect with future-oriented interpretation receives further

support from Chinese (Lin 2006), Korean (Jungmin Kang, p.c.), Guarani (Tonhauser 2011), Lillooet Salish

(Matthewson 2006), among other languages. I leave further discussion of this interesting topic for future research. 46

Unlike with future interpretations, embedded present tense with imperfective aspect is still unavailable in

propositional complements receiving past-oriented interpretation, as illustrated in (60c). The discussion of such a

contrast is beyond the scope of this paper (but note that, even in English, to obtain a backwards shifted interpretation

in these complements, a Perfect component is required (cf. fn 40)). 47

An anonymous reviewer wonders why the discussion of restrictions does not contain Aorist, an aspectual tense

which refers to punctual and/or completed events, and which is restricted to perfective aspect in Serbian (even when

Aorist is reported to occur with imperfective verbs, these verbs receive perfective interpretation). Aorist forms tend

to restrict embedding, which might be part of the broader restrictions on perfective forms discussed in the paper.

I agree that the behavior of Aorist would be a natural choice to explore given that it is predominantly formed of

perfective verbs in Serbian. However, it is unclear to me at this point why Aorist in general resists embedding in

Serbian, and how these restrictions can be integrated into the proposed analysis. Note, however, that the resistance to

embedding of Aorist seems to be a more general pattern in Serbian. For instance, in complex clauses, Aorist occurs

42

4.3 Irrealis future complements in Serbian

In future irrealis clausal complements, i.e. clauses embedded under verbs like want, the

matrix verb introduces a presupposition regarding the realization of the embedded predicate ‒ the

embedded predicate is unrealized at the time when the desire occurs, typically being temporally

located after the matrix predicate.48

In (74), John’s meeting with his sister is intended to happen

in the future with respect to the time of his desire to meet her. Even in (75a), despite the

additional Perfect component, the interpretation of the complement is one where the time of

in main clauses, e.g. in the consequent of conditional clauses, as in (i) and (ii). It can also occur in the antecedent of

a conditional, as in (ii), in consequential clauses (iii), or in causal clauses as in (iv). Whereas all these examples

illustrate that Aorist is not necessarily restricted to past tense interpretation (see also Arsenijević 2013), none of

them is an example of Aorist in complement clause. So, there might be some independent reason that would exclude

these forms in complement clauses in Serbian. I leave the distribution of Aorist in embedded clauses for further

research. Note, however, that (iv) is especially interesting because it shows that Aorist can be used with the

reference to present moment, but only if it receives habitual interpretation, and not if the moment of completing the

event overlaps with the UT. This is parallel to what we observe with perfective morphological present tense in in

Serbian – the extension of the reference time interval allows for the otherwise unavailable perfective. This is also

evident in matrix clauses which contains Aorist as in (v), where Aorist can only refer to an event that has been

completed prior to the UT and not at the UT. Given that Aorist is predominantly formed of perfective verbs, this

parallel behavior calls for further analysis in the future.

(i) Ako ne budemo odlučni, propadoše nam svi planovi. (Riđanović 2012:317)

if not be decisive fall.through –aor.3.pl we-dat. all plans

‘If we are not decisive, all our plans will fall through.’

(ii) Ako pođoh, nagledah se jada... (Stanojčić and Popović 1992:384)

if go-aor.1.sg. see-aor. se sorrow

‘If I go, I will witness all the suffering...’

(iii) Nema nam spasa, pomrijesmo od gladi! (Riđanović 2012:317)

not.have-3.sg. we- dat. salvation die-1.pl aor. from hunger

‘We can’t be saved – we will starve to death.’

(iv) Ne diraj mi kompjuter – ti pokvari sve što dotakneš. (Riđanović 2012:316)

not touch I-dat. computer you break-aor.2.sg all that touch-2.sg.pres.

‘Don’t touch my computer, you break everything you handle!’

(v) Evo dođe struja! (Riđanović 2012:316)

here come-aor.3.sg electricity

‘Power is on again!’

48

However, irrealis complements do not necessarily receive a future-oriented interpretation. For instance, a verb

such as try introduces an unrealized presupposition, but it does not allow for future interpretation of the

complement. In (i), the felicitous interpretation is one where John’s eating of an apple starts at the point when the

trying occurs, i.e. at the time of the matrix predicate. Crucially, the eating cannot be temporally located entirely after

the time of trying, as shown by the impossibility of adverbials such as tomorrow (see Sharvit 2003b for details on

the semantics of try).

(i) John is trying to eat an apple (#tomorrow).

43

meeting the sister follows the time of the desire.49

In order to derive the interpretation where

John’s meeting his sister is intended to have happened prior to the time of John’s desire, the

embedding verb is typically wish, rather than want, as in (75b).

(74) John wants to meet his sister.

(75) a. John wants to have met his sister.

b. John wishes to have met his sister.

In Serbian future irrealis complements the same situation holds – the embedded predicate is

typically temporally located after the time of the matrix predicate. However, unlike in English,

which allows for an additional Perfect component with these complements, the embedded verb in

Serbian can only be morphologically realized as an infinitive or as present tense, these two forms

being interchangeable, as in (76).50

The embedded past in (77b) and embedded future in (78b)

are illicit.51,52

49

Unlike propositional complements, future irrealis complements are typically temporally following the matrix verb. 50

This holds irrespective of the aspectual specification. Note, however, that want in (76) is a restructuring verb; only

in these environments are the two forms interchangeable. Note also that in the Serbian dialect of Serbo-Croatian,

there is a slight preference for using da + present tense construction, whereas in the Croatian and Bosnian dialects,

speakers prefer using the infinitival form. Nevertheless, both forms are considered grammatical in either dialect. 51

It, however, remains unclear why the embedded verb in the future irrealis complements cannot be realized as a

future form. I tentatively suggest that this might be linked to the structural realization of the complement. Namely,

as shown below, my proposal of the structure of future irrealis complements comprises both present tense in TP and

the future component woll in the wollP, which I argue is responsible for the future orientation of the complement. I

entertain two possibilities for the lack of future morphology on the embedded verb: 1) both present tense and woll

are interpreted at LF, but only the present component is realized at PF; 2) there is no Tense in these complements,

and the woll component is morphologically realized as present. The latter option would need to stipulate that the

realization of woll in the complement clauses is different from the one in matrix clauses (recall that, if there is a woll

component in matrix clauses, it is, together with present tense, realized as the periphrastic future form). Finally, it

would be possible to assume that in future irrealis complements there is neither Tense nor wollP, and that the matrix

verb dictates the future-oriented interpretation of the embedded event. In that case, the present tense on the

complement could be considered a default morphological realization. For present purposes, I follow Wurmbrand

(2014), who, on the basis of English, provide evidence for the syntactic presence of woll in these complements. The

main focus of this paper is the semantic composition of structures with perfective aspect, whereas the discussion of

the interaction between semantics and morphology, specifically the question of whether the arguments for the

syntactic presence of woll can also be replicated in Serbian, needs to be left for further research. 52

Želeti in Serbian can refer to two separate lexical items, i.e. ‘to want’ and ‘to wish’. The latter is counterfactual

and it allows for the periphrastic past complement in Serbian, in addition to the morphological present form in the

complement (cf. (75b) in English)). Irrealis complements of the verb želeti, unlike complements in counterfactual

constructions, cannot occur in a periphrastic past form; they only occur in morphological present form (or as

infinitival complement in control constructions). Note also that complements of counterfactuals and future irrealis

complements are different in terms of their semantic contribution (Iatridou 2000)).

44

(76) a. Želim da prevodim pesmu.

want-1.sg.pres that translate- impfv.1.sg.pres. poem

b. Želim prevoditi pesmu.

want-1.sg.pres translate- impfv.inf. poem

Intended interpretation: ‘I want to be translating a poem.’

(77) a. *Želim da sam prevodio pesmu.

want-1.sg.pres that am translated-impfv. poem

Intended interpretation: ‘I want to have been translating a poem.’

b. *Želim da sam preveo pesmu.

want-1.sg.pres that am translated-pfv. poem

Intended interpretation: ‘I want to have translated a poem.’

(78) a. *Želim da ću prevoditi pesmu.

want-1.sg.pres that will translate- impfv.inf. poem

Intended interpretation: ‘I want to be translating a poem.’

b. *Želim da ću prevesti pesmu.

want-1.sg.pres that will translate- pfv.inf. poem

Intended interpretation: ‘I want to translate a poem.’

Unlike propositional complements, irrealis future complements allow for the perfective

aspect on the embedded present (cf. (56b)):

(79) Želim da sutra popodne prevedem pesmu.

want-1.sg.pres that tomorrow afternoon translate-pfv.1.sg.pres. poem

‘I want to translate the entire poem tomorrow afternoon.’

Similar to past and future interpretations in matrix clauses, the availability of the perfective

with future irrealis complements can be attributed to the duration of the reference time interval

for the embedded Aspect. I suggest that the irrealis future complements, unlike propositional

complements, introduce an additional component (woll) which allows for the extension of the

reference time interval, allowing for the embedded perfective to be realized.

According to Abusch (2004), temporal argument of verbs taking future irrealis complements

is, like in the case of propositional infinitives, the contemporary now of the attitude holder. In the

case of propositional complements, the contemporary now is responsible for the simultaneous

interpretation of the matrix and embedded predicates. However, with future irrealis

complements, Abusch (2004) argues that a future-oriented interpretation comes from a temporal

substitution operator which shifts the time of the (potential) realization of the embedded event

into the future with respect to the matrix time. For instance, as a result of the temporal

substitution operator, the embedded (morphological) present receives a future interpretation in

45

Mary intends to give an automatic A to every student who submits a term paper at least 15 pages

long. Following Abusch, I posit the λ-operator in the embedded C, which binds the temporal n in

the embedded tense. However, I argue that future-orientation of the complement is attributed to a

future woll component in Serbian, which, I suggest, is the part of the embedded clause, rather

than built into the semantics of the matrix verb (along the lines of Wurmbrand’s (2014) analysis

of future irrealis infinitival complements in English). The structure for the example in (80) is the

one in (81).

(80) Jovan želi da Marija prevede pesmu.

Jovan wants that Marija translates-pfv. poem

Intended interpretation: ‘Jovan wants for Marija to have translated the poem.’

(81) TP ei

UT T’ ei

Pres VP ei

John V’ ei

want CP ei

λ n TP ei

Pres wollP ei

woll AspP ei

perfective VP ei

Mary translate poem

⟦TP UT [TP Present(n) [VP John want [CP λn [ TP Pres(n) [wollP woll [AspP Perfective [vP Mary

translate poem]]]⟧ =[∃t3[t3 =n & want(n)(j) [λn λw ∃t2[t2 = n & ∃t1[t1 > t2 &∃e[ time(e) t1 &

translate poem (m)(w)(e)]]]]]]

The Tense and Aspect ordering in the above structure proceeds in the usual way: embedded

present tense introduces a time interval identical with the contextually salient time interval,

which in (80), like in the case of propositional complements, is the contemporary now (facilitated

46

by the presence of the λ-operator). Due to the requirement of the present tense, the reference time

interval for the lower temporal components is the contemporary now, a relatively short time

interval. However, due to the presence of the woll component, which introduces a potentially

indefinitely long time, the interval introduced by the contemporary now is extended.53

Crucially,

this presumably long time interval serves as the reference time interval for the embedded Aspect,

making the inclusion requirements of the perfective aspect possible to satisfy.

Further predictions are that the matrix tense should not affect the distribution of the

embedded perfective; the embedded present tense component will always require to be identical

with the contemporary now, and the future component woll will always extend the time interval

with respect to which the embedded Aspect is ordered, allowing for the requirements of the

perfective to be satisfied. The prediction is borne out, as illustrated in (82a) and (82b), where the

matrix clause receives a past and future interpretation.54,55

53

The present tense, dues to its requirements, makes the contemporary now interval serve as the ordering time

interval for the woll component.

54 Verbs whose complements pattern with the verb želeti ‘to want’ are očekivati ‘to expect’, non-implicatives like

planirati ‘to plan’ and odlučivati ‘to decide’, etc. 55

An anonymous reviewer points out that ‘want’ may also allow a simultaneous interpretation in certain contexts,

like in English, as illustrated in (i) and (ii). Such examples would pose a problem for the analysis advocated in the

paper, because, given the simultaneous interpretation, the duration of the time interval with respect to which the

embedded perfective would need to be included is too short, and the examples would incorrectly be predicted to be

infelicitous. However, I believe that in those cases, ‘want’ is best interpreted as ‘like being in the state of’; it has to

refer to a prolonged state which starts at some point in the past and overlaps with the UT. This is confirmed by the

use of the imperfective and the stative predicate, respectively, in the first part of the sentence in (iii) and (iv), and

also by the impossibility of certain adverbials when perfective is used, as in (v) and (vi) (the examples in (v) and (vi)

are due to the anonymous reviewer). In other words, it seems to be the case that instead of dealing with an

instantaneous interval of desiring something, this ‘want’ introduces a lengthy interval allowing for the inclusion

requirements of the perfective to be satisfied. Note, however, that these examples are marked and require specific

contexts, and I set those interpretations aside and only use the unmarked future interpretation of ‘want’.

47

(82) a. Želeo sam da prevedem pesmu.

wanted- sg. am that translate- pfv.1.sg.pres. poem

Intended interpretation: ‘I wanted to have translated a poem.’

b. Želeću da prevedem pesmu.

want-will.1.sg that translate- pfv.1.sg.pres. poem

Intended interpretation: ‘I will want to have translated a poem.’

4.4 Distribution of non-progressive eventives in propositional complements in English

This section just briefly outlines Wurmbrand’s (2014) analysis of different types of infinitival

complementation in English, illustrating the distribution of non-progressive eventives and

showing that they are parallel to the Serbian data above.

Wurmbrand argues that propositional complements in English are TPs, headed by zero tense,

which contributes to the simultaneous interpretation of the matrix and the embedded eventive

predicate. Zero tense in these complements can be motivated by the temporal interpretations

parallel to the complements of verbs like say. In (83), the SOT effects are argued to arise due to

deletion of past tense in the embedded clause, or rather the presence of the zero/null tense.56

Propositional complements in (84) also derive simultaneous interpretation, which Wurmbrand

(i) Želim da u ovom trenutku pojedem jabuku.

want-1.sg.pres. that in this moment eat- pfv.1.sg.pres. apple

‘I want to eat an apple right now.’

(ii) Želim da u ovom trenutku budem u krevetu.

want-1.sg.pres. that in this moment be- pfv 1.sg.pres. in bed

‘I want to be in bed right now.’

(iii) Jedem jabuku, jer želim

eat- impfv.1.sg.pres. apple because want-1.sg.pres.

da u ovom trenutku pojedem jabuku.

that in this moment eat-pfv.1.sg.pres. apple

‘I am eating an apple because I want to eat (and finish) an apple at this point.’(As an answer to a question ‘Why

are you eating an apple?)

(iv) U krevetu sam, jer želim

in bed am because want-1.sg.pres.

da u ovom trenutku budem u krevetu.

that in this moment be- pfv.1.sg.pres. in bed

‘I am in bed because I want to be in bed at this point.’ (As an answer to a question ‘Why are you in bed?)

(v) Želim da *upravo/ *trenutno pojedem jabuku.

want-1.sg.pres. that currently/momentarily eat- pfv.1.sg.pres. apple

‘I want to be eating an apple at this very moment.’

(vi) Želim da ??upravo/trenutno budem u krevetu.

want-1.sg.pres. that currently/momentarily be- pfv.1.sg.pres. in bed

‘I want to be in bed at this very moment.’

56 In addition, the non-SOT interpretation is also available, in which case the embedded Tense is specified for [past].

48

takes as an indicator of the zero tense in these complements. Crucially, the SOT interpretation of

(84) is impossible if the embedded predicate is non-progressive, as illustrated in (85).

(83) Leo believed Julia to be bringing the toys right then. (Wurmbrand 2014)

(84) John said that Mary was reading Middlemarch. SOT possible (Portner 2003)

(85) John said that Mary read Middlemarch. *SOT (Portner 2003)

Wurmbrand, following Portner (2003), suggests that the problem arises due to the

incompatibility of the zero tense and non-progressive forms. Similarly, if the embedded tense in

propositional complements is also zero, then the impossibility of non-progressive forms is

expected, as in (86):

(86) *Leo believed Julia to bring the toys right then. (Wurmbrand 2014)

Wurmbrand argues that zero tense relates the reference time interval of the embedded clause

with the event time interval of the matrix clause, which in the case of (86) is Leo’s contemporary

now. Since this interval serves as the reference time interval for the embedded tense, and due to

its zero specification, the tense introduces an interval identical to the contemporary now; and the

contemporary now serves as the time interval with respect to which the embedded Aspect is

ordered. Assuming that non-progressive forms are specified for perfective, Aspect cannot satisfy

its requirements, and the form is correctly predicted to be excluded. The same analysis can

explain the lack of the simultaneous interpretation of the embedded and matrix predicate in (85)

with non-progressive forms. In addition, Wurmbrand’s analysis is compatible with the analysis

outlined for the propositional complements in Serbian. As expected, if there were an additional

component between the zero tense and perfective aspect, it should allow for the non-progressive

form to be realized (provided that it extends the time interval with respect to which the

embedded Aspect is ordered). The prediction is borne out, as shown in (87) with the embedded

Perfect.

(87) John believed for Mary to have left.

49

Regarding future ireallis complements, Wurmbrand (2014), following Abusch (1985, 1988),

argues that unlike finite future forms in English which are composed of a present tense

component and the modal woll, non-finite future irrealis complements lack the temporal

component and only contain the modal component woll. She argues that the presence of a tense

component in finite future forms results in their absolute interpretation as in (88), while its

absence in infinitival contexts results in the relative interpretation of future, as in (89) (an

interpretation where the embedded event occurs after the matrix time but potentially before the

UT).57

(88) a. Leo decided a week ago that he will go to the party (*yesterday).

b. According to a report I read last week, it was expected that the bridge will collapse

(*yesterday).

(89) a. Leo decided a week ago to go to the party yesterday.

b. According to a report I read last week, the bridge was expected to collapse yesterday.

She observes that, regardless of the specification of the matrix tense, no restrictions arise

with respect to the distribution of non-progressive eventives in future irrealis complements, as in

(90). This follows from the structural composition of future irrealis infinitives the woll shifts

the reference time interval to an unspecified point in the future with respect to the matrix event

time, but crucially, regardless of the specification of the matrix verb. Once again, these data align

with what we observed in Serbian.

(90) a. John decided to sing in the shower.

b. The printer is expected to work again.

5. Conclusion and future directions

In this paper I argued that restrictions on perfective aspect in morphological present tense

contexts in Serbian in both matrix and embedded clauses arise when Aspect is ordered with

respect to a short interval. Given that perfective aspect requires the inclusion of the event time

interval in the reference time interval, the duration of the time interval makes the inclusion

requirements of the perfective impossible to satisfy. This further predicts that whenever the time

57

The examples in this section are from Wurmbrand (2014).

50

interval with respect to which Aspect is ordered is long enough, the requirement of the perfective

can be satisfied, and the form is predicted to be felicitous. We have seen that this is exactly what

happens in a number of contexts including past and future interpretations, as well as in future

irrealis complements. Once we continued playing with those contexts, and restricting the longer

ordering time interval, the perfective again became infelicitous, which supports both our initial

claim about its requirements, as well as structural approach to the temporal location of the event.

Moreover, the proposal in the paper aligned with the proposals for the distribution of non-

progressive forms of eventive predicates in English, providing further support for the claim that

the forms in English contain a perfective component.

Finally, certain questions remain open for further research. One of the puzzles is related to

Perfect, and, more generally, to the exact mapping between syntax/semantics and morphology.

While the existence of the Perfect with constructions receiving past interpretation is motivated

by the morphology, one question that remains unanswered is why the element corresponding to

semantic Past is realized as morphological present in Serbian. Hopefully, future work will shed

new light on these questions.

References

Abusch, D. 1985. On verbs and time. Doctoral dissertation, University of Massachusetts,

Amherst.

Abusch, D. 1988. Sequence of tense, intensionality, and scope. In Proceedings of the 7th

West

Coast Conference on Formal Linguistics, ed. by Hagit Borer, 1-14. Stanford, Calif: CSLI

Publications.

Abusch, D. 1997. Sequence of tense and temporal de re. Linguistics and Philosophy 20.1:1-50

Abusch, D. 2004. On the temporal compostion of infinitives. In The syntax of time, ed. by

Jacqueline Guéron and Jacqueline Lecarme, 27-53. Cambridge, MA: MIT Press.

Arsenijević, B. 2013. Vremensko i aspektualno značenje aorista (The tense and aspect

components in the semantics of the /Serbian/aorist), Srpski jezik 18, 253–261.

Bennett, M. & B. Partee 1972. Toward the Logic of Tense and Aspect in English. Indiana

University Linguistics Club [Republished in Partee 2004]

Bošković, Ž. 1996. Selection and the categorial status of Infinitival Complements. Natural

Language and Linguistic Theory 14:269-304.

51

Bošković, Ž. 1997. The syntax of nonfinite complementation: An economy approach. MIT

Press.

Bjorkman, B. 2015. Only some “fake pasts are real: contrasting counterfactuals and sequence

of tense. Available at http://ling.auf.net/lingbuzz/002461

Chomsky, N. and H. Lasnik. 1993. Principles and parameters theory. In Syntax: An

International handpoem of contemporary research, ed. by Joachim Jacobs, Arnim von Stechow,

Wolfgang Sternefeld and Theo Vennemann, 506-569. Berlin/New York: Mouton de Gruyter.

Condoravdi, C. 2001. Temporal interpretation of modals: Modals for the present and for the

past. In Stanford papers on semantics, ed. by David Beaver, Stefan Kaufmann, Brady Clark, and

Luis Casillas, 59–88. Stanford, CA: CSLI Publications.

Copley, B. 2002. The semantics of the future. Doctoral dissertation, MIT, Cambridge, MA.

Cowper, E. 1996. The features of tense in English. Toronto Working Papers in Linguistics 14:

19-40.

Cowper, E. 1998. The simple present in English: a unified treatment. Studia Linguistica 52: 1-

18.

Demirdache, H and M. Uribe-Etxebarria. 2004. The Syntax of Time Adverbs.. In J. Guéron and

J. Lecarme (eds.) The Syntax of Time, MIT Press: Cambridge, Massachusetts, 143-179.

Dowty, D. 1979. Word Meaning and Montague Grammar. Dordrecht: Kluwer.

Dowty, David. 1982. Tenses, time adverbs, and compositional semantic theory. Linguistics and

Philosophy 9:405-426.

Enç, M. 1987. Anchoring Conditions for Tense. Linguistic Inquiry 18:633-657.

Enç, M. 1991. On the absence of the present tense morpheme in English. Ms., University of

Wisconsin.

Giorgi, A., Pianesi, F., 1997. Tense and Aspect: From Semantics to Morphosyntax. Oxford

University Press, Oxford.

Iatridou 2000. The Grammatical Ingredients of Counterfactuality. In Linguistic Inquiry 31:

231‒270.

Iatridou, S., E. Anagnostopoulou, and R. Izvorsky. 2001. Observations about the form and

meaning of the perfect. In Ken Hale: A Life in Language, 189-238. Cambridge, MA: MIT Press.

Kamp, H. & U. Reyle 1993. From Discourse to Logic. Dordrecht: Kluwer.

Kaufmann, S. 2005. Conditional truth and future reference. Journal of Semantics 22:231–280.

52

Klein, W. 1992. The Present Perfect Puzzle. Language 68.3: 525-552.

Klein, W. 1994. Time in language. London and New York: Routledge.

Klein, W. 1995. A Time Relational Analysis of Russian Aspect. Language, 71.4: 669-695.

Kratzer, A. 1998. More structural analogies between pronouns and tenses. In Proceedings from

Semantics and Linguistic Theory VIII, ed. by D. Strolovitch and A. Lawson, 92-110. Ithaca,

N.Y.: Cornell University, CLC Publications.Linguistics and Philosophy 26:459–510.

Lin, J-W. 2006. Time in a language without tense: The case of Chinese. Journal of Semantics

23:1–53.

Martin, R. A. 1996. A minimalist theory of PRO and control. Doctoral dissertation, University of

Connecticut, Storrs.

Martin, R. A. 2001. Null case and the distribution of PRO. Linguistic Inquiry 32.1:141-166.

Matthewson, L. 2006. Temporal semantics in a superficially tenseless language. Linguistics and

Philosophy 29: 673–713.

Ogihara, T. 1995. The Semantics of Tense in Embedded Clauses. In Linguistic Inquiry

26: 663-679.

Ogihara, T. 1996. Tense, attitudes, and scope. Dordrecht: Kluwer.

Ogihara, T. 2007. Tense and Aspect in Truth-conditional Semantics. Lingua 117.2: 392-418.

Paunović, Ž. 2001. Aspectual-temporal relations in Serbo-Croatian verbal morphology. Essex

Graduate Student Papers in Language and Linguistics, v.III

Pancheva, R. 2003. The Aspectual Makeup of Perfect Participles and the Interpretations of the

Perfect. In A. Alexiadou, M. Rathert, and A. von Stechow (eds.) Perfect Explorations. Mouton

de Gruyter, 277-306.

Pancheva, R. and Von Stechow, A. 2004. On the Present Perfect Puzzle. In NELS 34, eds. K.

Moulton and M. Wolf.

Pesetsky, D. 1995. Zero Syntax. Cambridge, MA: MIT Press

Portner, P. 2003. The (Temporal) Semantics and (Modal) Pragmatics of the Perfect.

Linguistics and Philosophy 26:459–510.

Romero 2014. Fake Tense in Counterfactuals: A Temporal Remoteness Approach, in The Arts

and Craft of Semantics: A Festschrift for Irene Heim, ed. by L. Crnić and Uli Sauerland, vol. 2,

MITWPL 71: 47-63.

53

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

Riđanović, M. 2012. Bosnian for Foreigners: with comprehensive grammar. Rabic, Sarajevo.

Schlenker, P. 1999. Propositional attitudes and indexicality: A cross-categorial approach.

Doctoral dissertation, MIT, Cambridge, MA.

Sharvit, Y. 2003a. Embedded tense and universal grammar. Linguistic Inquiry 34:669-681.

Sharvit, Y. 2003b. Trying to be Progressive: the Extensionality of Try. Journals of Semantics

20:403-445.

Smith, C. 1991. The parameter of aspect. Dordrecht: Kluwer.

Stanojčić, Ž. and Lj. Popović. 1992. Gramatika sprskog jezika. Zavod za udžbenike i nastavna

sredstva. Beograd.

Stowell, T. 1982. The tense of infinitives. Linguistic Inquiry 13:561-570.

Taylor, B. 1977. Tense and continuity. Linguistics & Philosophy 1: 199-220.

Thomason, R. 1970. Indeterminist time and truth-value gaps. Theoria 36:264–281.

Tonhauser, J. 2011. Temporal reference in Paraguayan Guaraní. Linguistics and Philosophy 34,

257– 303.

Von Stechow, A. 1999. Eine erweiterte Extended Now-Theorie für Perfekt und Futur. Zeitschrift

für Literaturwissenschaft und Linguistik 113:86-118.

Von Stechow, A. 2002. German Seit ‘Since’ and the Ambiguity of the German Perfect. In More

than Words: A Festschrift for Dieter Wunderlich, B. Stiebels and I. Kaufmann (eds.), 393-432.

Berlin: Akademie Verlag.

Von Stechow, Arnim. 2009. Tenses in compositional semantics. In W. Klein & P. Li (eds.), The

expression of time in language, vol. 30, 129–166. Berlin: Mouton de Gruyter.

Wurmbrand, S. 2014. Tense and aspect in English infinitives.Linguistic Inquiry 45.3: 403-447.