infectum and perfectum. two faces of tense selection in ... · pdf fileinfectum and perfectum....

Abstract This paper investigates the semantics of tense and aspect in Romance

languages. Its goal is to develop a compositional, model-theoretic semantics for

tense and temporal adverbs which is sensitive to aspectual distinctions. I will

consider durative adverbial distributions and aspectual contrasts across different

morphological tense forms. I will examine tense selection under habitual meanings,

generic meanings and state of result constructions. In order to account for these facts

I will argue that temporal homogeneity plays a fundamental role in tense selection

in Romance languages.

Keywords Tense � Aspect � Homogeneity � Romance languages

1 Introduction

In Romance languages we find a rich system of verbal temporal inflections. On one

hand, verbal inflections in these languages locate the situation described by the

sentence in the past, future or present with respect to the time at which the sentence

is uttered (speech time), on the other hand they give access to the temporal stages

of the situation described by depicting it as completed or as ongoing at a past,

future or present time. For example, consider the following Italian and French

sentences

F. Arosio (&)

Universita degli Studi di Milano–Bicocca, Milan, Italy

e-mail: [email protected]

123

Linguist and Philos (2010) 33:171–214

DOI 10.1007/s10988-010-9078-x

RESEARCH ARTICLE

Infectum and Perfectum. Two faces of tense selectionin Romance languages

Fabrizio Arosio

Published online: 19 January 2011

� Springer Science+Business Media B.V. 2011

Italian French

(1) a. Mario legge b. Mario lit

Mario read-3sgPRESimperfective Mario read-3sgPRESimperfective

Mario is reading Mario is readingTraditional classification: Presente Traditional classification: Present

(2) a. Mario leggeva b. Mario lisait

Mario read-3sgPASTimperfective Mario read-3sgPASTimperfective

Mario was reading Mario was readingTraditional classification: Imperfetto Traditional classification: Imparfait

(3) a. Mario lesse b. Mario lut

Mario read-3sgPASTperfective Mario read-3sgPASTperfective

Mario read Mario readTraditional classification:

Passato Remoto

Traditional classification:

Passe Simple

The different verbal inflections are responsible for a difference in meaning

among (1), (2) and (3). While the contrast in meaning between (1) and (2)

concerns the location of Mario’s reading in the flow of time, which is going on

at speech time according to (1) and before speech time according to (2), the

difference between (2) and (3) concerns a contrast in the way Mario’s reading is

described with respect to a past time since it is depicted as going on at that time

according to (2) and completed according to (3). Romance temporal inflections

have been traditionally classified according to these two semantic criteria, i.e. as

past, present or future, since they locate the described eventuality at a time that

can be before, overlapping or after speech time and as perfective or imperfective

depending on whether the described eventuality is depicted as completed or as

ongoing at that time.

While we find a general consensus in assuming that tense is the grammatical

category that is responsible for locating the eventuality described by the sentence in

the flow of time, the analysis of the aspectual contrasts, which mainly concerns the

opposition between perfective and imperfective temporal meanings, is much more

controversial, especially for Romance languages, where we find a rich inventory of

verbal forms for conveying different aspectual meanings.

According to a fruitful proposal by Klein (1994), aspect concerns a temporal

relation between the time at which the eventuality described by the VP holds and the

time introduced by tense, namely between the Reichenbachian reference time and

event time.

This proposal has been more recently investigated and developed in a formal

framework by Kratzer (1998). According to Kratzer, aspectual distinctions are

conveyed by means of aspectual operators which map properties of events denoted

by the VP into properties of times and that tense provides the time that saturates the

temporal property obtained, as represented in (4) below

172 F. Arosio

123

ð4Þ TP AspectP

Aspect Tense

VP

Kratzer contends that the two major aspect operators responsible for the aspectual

distinctions we have discussed so far are the Imperfective and the Perfectiveoperators, as defined below and that the two most common tenses are the past and

the present tense.1

(5) Perfective ¼:��PFV

�� ¼ kPkt9ev(t ˚ s(ev) & P(ev))

[event time included in reference time]

(6) Imperfective ¼:��IPV

�� ¼ kPkt9ev(s(ev) ˚ t & P(ev))

[reference time included in event time]

where s is a function taking an eventuality and giving its temporal trace.

Kratzer’s suggestion is that verb forms are the spell-out of tense/aspect combina-

tions and that the French Imparfait and Passe Simple are the spell out of the

following tense aspect combinations

PAST

IMPERFECTIVE Imparfait

PERFECTIVE Passe Simple

In her paper Kratzer does not discuss the case of Italian. According to von

Stechow (2002) the Italian opposition between the Imperfetto and Passato

Remoto parallels the French one and we can therefore assume the Imperfetto to

be the spell out of the PAST/IMPERFECTIVE combination and the Passato

Remoto to be the spell out of the PAST/PERFECTIVE combination According to

Kratzer’s proposal, the LF of the Italian/French sentences in (2) we discussed

above will be

(7) (a) Mario leggeva [Imperfetto] ITA

Mario read-3sgPASTimp

(b) Mario lisait [Imparfait] FR

Mario read-3sgPASTimp

Mario was reading

1 I give a slightly modified definition from Kratzer’s.

Infectum and Perfectum 173

123

(c)

IPV λPλt∃e(τ(ev) ⊇ t & P(ev))

TP AspectP

PASTVP

λev(Mario_read (ev))

Sentences in (7) are true if the past time introduced by the tense is included in the

temporal trace of Mario read event.2

In this paper, I will present some facts concerning the distribution of durative

temporal adverbials and tense selection in habitual sentences in French and

Italian and I will argue that it is not enough to analyze the verb forms of these

languages as the spell out of different tense/aspect combinations as proposed by

Kratzer. In particular, I will show that for x time and since x time durative

temporal adverbials have a complementary distribution across different verb

forms in most Romance languages and that this distribution changes when the

sentence is interpreted habitually. I will moreover discuss tense selection in

habitual and generic sentences. I will propose a new analysis of the different

tense forms of Italian and French in order to account for these facts. In par-

ticular, I will propose a new definition of tense according to which it is sensitive

to certain properties of its complement [a similar intuition is found in de Swart

(1998)] and I will argue that the verbal forms are the morphological spell-out of

two different tenses imposing some conditions on the temporal homogeneity of

their complements. In order to implement this idea, I will decompose the tense

projection into a temporal variable, localising the eventuality described in the

flow of time, and a homogeneity condition on the interpretation of the tense

complement. The idea is that the temporal property that is saturated by the time

provided by tense should be licensed by the homogeneity conditions located in

the tense projection. Given these assumptions, the distributive facts and the

semantic ambiguities I will discuss in the next paragraph will follow straight-

forwardly.

2 Italian and French facts

In Italian, durative per- and da-adverbials are found in a complementary distribution

in state sentences.

2 The IPV operator clearly does not give us the complete truth conditions for sentence (2). We know they

should be modalised since the Imperfetto/imparfait tenses give rise to imperfective paradoxes largely

discussed for the English progressive constructions.

174 F. Arosio

123

(8) per-adverbials (9) da-adverbials

(a) ?? E buio per due ore (a) E buio da due ore

Be-3sgPRES dark for 2 hours Be-3sgPRES dark since 2 hours

– It has been dark for two hours

PRESENTE PRESENTE

(b) ?? Era buio per due ore (b) Era buio da due ore

Be-3sgPASTimp dark for 2 hours Be-3sgPASTimp dark since 2 hours

– It had been dark for two hours

IMPERFETTO IMPERFETTO

(c) Fu buio per due ore (c) ?? Fu buio da due ore

Be-PASTperf dark for 2 hours Be-PASTperf dark since 2 hours

It was dark for two hours –

PASSATO REMOTO PASSATO REMOTO

(d) E stato buio per due ore (d) ?? E stato buio da due ore

Be-3sgPRES been dark for 2 hours Be-3sgPRES been dark since 2 hours


PASSATO PROSSIMO PASSATO PROSSIMO

(e) Era stato buio per due ore (e) ?? Era stato buio da due ore

Be 3sgPASTimp been dark for 2 hours Be-PASTimp been dark since 2 hours

It had been dark for two hours –

TRAPASSATO PROSSIMO TRAPASSATO PROSSIMO

As you can see from the sentences in (8), per-adverbials combine with the PassatoRemoto, the Passato Prossimo and the Trapassato Prossimo but not with the Pre-sente or the Imperfetto. On the other hand, da-adverbials combine with the Presenteand the Imperfetto but do not with the Passato Remoto, the Passato Prossimo and

the Trapassato Prossimo as we see in the sentences in (9).3

French patterns with Italian in the distribution of pendant and depuis adverbials,

as you can see below

(10) pendant-adverbials (11) depuis-adverbials

(a) ??La fenetre est sale pendant deux jours (a) La fenetre est sale depuis deux jours

The window be-3sgPRES dirty for 2 days The window be-3sgPRES dirty since 2 days

– The window has been dirty for two days

PRESENT PRESENT

(b) ??La fenetre etait sale pendant deux jours (b) La fenetre etait sale depuis deux jours

The window be-PASTimp dirty for 2 days The window be-PASTimp dirty since 2 days

– The window had been dirty for two days

3 (8) (a) and (b) are fine under a habitual interpretation in an appropriate context. As we will see later,

this is predicted by the proposal I am making.


123

IMPARFAIT IMPARFAIT

(c) La fenetre fut sale pendant deux jours (c) ??La fenetre fut sale depuis deux jours

The window be-3sgPAST perf dirty for

2 days

The window be-3sgPAST perf dirty since

2 days

The window was dirty for two days –

PASSE SIMPLE PASSE SIMPLE

(d) La fenetre a ete sale pendant deux jours (d) ??La fenetre a ete sale depuis deux jours

The window have-3sgPRES been dirty for

2 days

The window have-3sgPRES been dirty

since 2 days


PASSE COMPOSE PASSE COMPOSE

(e) La fenetre avait ete sale pendant deux jours (e) ??La fenetre avait ete sale depuis deux

jours

The window have-3sgPASTimp been dirty

for 2 days

The window have-3sgPASTimp been dirt.

since 2 days

The window had been dirty for two days –

PASSE ANTERIEUR PASSE ANTERIEUR

Exactly like in Italian, pendant-adverbials combine with the Passé Simple, the

Passé Composé and the Passé Antérieur but not with the Présent or the Imparfait, as

you can see from the sentences in (10). On the other hand, depuis-adverbials combine

with the Présent and the Imparfait but not with Passé Simple, the Passé Composé or

the Passé Antérieur, as shown in the sentences in (11). I will call these two classes of

adverbials durative for-adverbials and durative since D-adverbials.4

Another interesting fact to be noticed is that, in French and Italian, ongoing and

habitual meanings are generally conveyed by the use of a morphological imper-

fective tense while perfective readings by the use of a morphological perfective

tense, as shown by the sentences below:

(12) (a) Alle tre Carlo correva nel parco [Imperfetto] ITA

At three Carlo run-3sgPASTimp in-the park

(b) A trois heures Carlo courait dans le parc [Imparfait] FR

At three o’clock Carlo run-3sgPASTimp in the park

(i) Imperfective Reading: Carlo was running in the park at three o’clock(ii) Habitual Reading: Carlo used to run in the park at three o’clock

(13) (a) Alle tre Carlo corre nel parco [Presente] ITA

At three Carlo run-3sgPRES in-the park

(b) A trois heures Carlo court dans le parc [Present] FR

At three o’clock Carlo run-3sgPRES in the park

4 Per/pendant-adverbials correspond to the English durative for-adverbials, while da/depuis-adverbials

do not have a counterpart in English; they rather correspond to the German seit-duration-adverbials

described by Musan (2003).

176 F. Arosio

123

(i) Imperfective Reading: Carlo is running in the park at three o’clock(ii) Habitual Reading: Carlo runs in the park at three o’clock

(14) (a) Ieri Carlo corse nel parco [Passato Remoto] ITA

Yesterday Carlo run-3sgPASTperf in-the park

(b) Hier Carlo courut dans le parc [Pass Simple] FR

Yesterday Carlo run-3sgPASTperf in the park

Perfective Reading: Yesterday Carlo ran in the park

Interestingly, under a habitual interpretation, for-adverbials combine with mor-

phological imperfective tenses, as shown by the sentences below

(15) (a) Il venerdı Carlo correva nel parco per due ore [Imperfetto] ITA

The Friday Carlo run-3sgPASTimp in-the park for two hours

(b) Le vendredi Carlo courait dans le parc pendant deux heures

[Imparfait] FR

The Friday Carlo run-3sgPASTimp in the park for two hours

Habitual Reading: On Fridays Carlo used to run in the park for two hours

(16) (a) Il venerdı Carlo corre nel parco per due ore [Presente] ITA

The Friday Carlo run-3sgPRES in-the park for two hours

(b) Le vendredi Carlo court dans le parc pendant deux heures [Present] FR

The Friday Carlo run-3sgPRES in the park for two hours

Habitual Reading: On Fridays Carlo runs in the park for two hoursAdditionally, when a for-adverbial measures the time span of the habit, habitual

meanings are conveyed by the use of a morphological perfective tense, as shown by

the Italian and French sentences below

(17) (a) Leo ha preso il te‘alle cinque per venti anni [Passato Prossimo] ITA

Leo have-3sgPRES taken the tea at five for twenty years

(b) Leo a pris son the a cinq heures pendant vingt ans[Passe Compose] FR

Leo have-3sgPRES taken his tea at five o’clock for twenty years

Habitual Reading: Leo used to have tea at 5 o’clock for twenty years

(18) (a) Leo prese il te‘alle cinque per venti anni [Passato Remoto] ITA

Leo take-3sgPASTperf the tea at five for twenty years

(b) Leo prit son the a cinq heures pendant vingt ans [Passe Simple] FR

Leo take-3sgPASTperf his tea at five o’clock for twenty years

Habitual Reading Leo used to have tea at 5 o’clock for twenty years


123

When a durative for-adverbial measures the time span at which the habit holds in

the past, a past imperfective tense is bad, as we can see in the following sentences

(19) (a) ?? Leo prendeva il te‘ alle cinque per venti anni [Imperfetto] ITA

Leo take-3sgPASTimp the tea at five for twenty years

(b) ?? Leo prenait son the a cinq heures pendant vingt ans [Imparfait] FR

Leo take-3sgPASTimp his tea at five o’clock for twenty years

–

(20) (a) ?? Leo prende il te‘ alle cinque per venti anni [Presente] ITA

Leo take-3sgPRES the tea at five for twenty years

(b) ?? Leo prend son the a cinq heures pendant vingt ans [Present] FR

Leo take-3sgPRES his tea at five o’clock for twenty years

–

Durative sinceD-adverbials combine with imperfective tenses but not with perfec-

tive ones when measuring the time span at which the habit holds, as will be dis-

cussed in Sect. 6.

2.1 My proposal in short

In order to account for these facts (and some others to be presented in our

discussion) I will argue that:

� sinceD-adverbials combine with temporal predicates to give temporally homo-

geneous predicates;

� for-adverbials combine with temporal predicates to give temporally non-

homogeneous predicates;

� present and past imperfective tenses require their complements to be temporally

homogeneous;

� past perfective tenses require their complements to be temporally non-

homogeneous.

This explains the distribution in sentences (8), (9), (10), (11) a–c. I will moreover

argue that:

� the perfect morphology in (8), (9), (10), (11) d–e is the spell-out of a semantic

tense combining with a temporally non-homogeneous predicate.

This explains the distribution in sentences (8), (9), (10), (11) d–e. Finally, we will

observe that

� habits are temporally homogeneous.

This will explain the fact that habitual readings are conveyed by the use of present

or past imperfective tenses; when they undergo for-adverb modification, habits

become non-homogeneous; for this reason we can find perfective morphology in

habitual sentences.

178 F. Arosio

123

3 Infectum and Perfectum in some Romance languages

In order to develop this analysis, I will propose that the temporal system of Italian

and French has two lexical entries in its inventory of tenses: a tense selecting for

temporally homogeneous predicates and a tense selecting for temporally non-

homogeneous predicates. The insight is that, in these languages, the present and past

imperfective forms are two forms of one and the same tense selecting for temporally

homogeneous predicates, while the past perfective, present perfect and pluperfect

forms are forms of a tense selecting for temporally non-homogeneous predicates.

This idea goes back to an observation of the Latin grammarian M. Terentius Varro

(116 B.C.–27 B.C.), who argued in De Lingua Latina, IX, 48 that we should assume

a basic division of the Latin tense forms into two stems, Infectum and Perfectum.

According to M. Terentius Varro, while the Latin present and past-imperfectiveverbal forms [like lego (I read-PRES) and legebam (I read-PAST.IMP) ] are Infectaforms and are analogous to one another, the Latin past-perfective forms [like legi(I read-PAST.PERF)] are Perfecta forms and are to be opposed to the former [see

Oldsjo (2001) for a further discussion].5 In this paper I want to make explicit thenature of the analogy Terentius Varro is talking about.

I will define the class of temporally homogeneous predicates as the one having the

sub-interval property, i.e. divisible, and the class of temporally non-homogeneous

predicates as the one not having it (Bennett and Partee (1972))6:

Given a temporal predicate P of type hi, ti, denoting a function from times i and

truth values t

(21) P is temporally homogeneous if "t [P(t) fi "t0 [ t0� t fi P(t0)]]

(22) P is temporally non-homogeneous if "t [P(t) fi :"t0 [t0� t fi P(t0)]]where t and t0 denote times.

3.1 Tense architecture

In order to develop a tense architecture appropriate to the proposal I am making,

I postulate that Italian and French tenses are to be decomposed into a homogeneitycondition, which distinguishes the Infectum from the Perfectum, and a temporalvariable, which distinguishes the past from the present and from the future. I will

assume that the homogeneity condition is located in the head of the tense projection

whose specifier position is filled by a time variable as represented by the LFs

sketched below

5 Vide: M. Terentius Varro, De Lingua Latina, LVI–LVIII, Liber IX. Arnim von Stechow pointed out

Terentius Varros’ analogy.6 This definition is slightly different from Bennett and Partee’s (1972):

A verb phrase a is a subinterval verb phrase if a is the main verb of a sentence b such that if

kb]kT,t,i then, for every j ˝ i, kb]kT,t,j, where j and i are intervals.

According to this definition

- live is a subinterval verb phrase

if kCarlo livekT,t,I¼1 then for every j ˝ i, kCarlo live kT,t,j¼1.


123

(23) TEMPORA INFECTA

Imperfetto- Imparfait

Presente - Présent

HOM<it, it>

T'

T

PASTi

TP

VP<i,t>

HOM<it, it>

T'

T

PRESi

TP

VP<i,t>

(24) TEMPORA PERFECTA

Passato Remoto - Passé Simple

Reporters Present?

PASTi

TP

N-HOM<it, it>

VP<i,t>

T'

T

N-HOM<it, it>

T'

T

PRESi

TP

VP<i,t>

According to this classification we find two classes of tenses in (some) Romance

languages: the class of the Temporal Infecta, whose tense head HOM requires the

tense complement to be temporally homogeneous, and the class of the Tempora

Perfecta, whose tense head N-HOM requires the tense complement to be temporally

non-homogeneous. The Perfect constructions (the Italian Passato Prossimo and

Piuccheperfetto and the French Passé Composé and Plus-que-Parfait) do not appear

in charts (23) and (24) despite the fact that they fall into the classification above.

This is because, depending on the aspectual class of the predicate they combine

with, these forms are semantically ambiguous between the spell out of a Tempus

Infectum and the spell out of a Tempus Perfectum as we will see in Sect. 5.

According to this tense architecture, the head of the tense projection is a pred-icate restriction, namely, a partial identity function from predicates of times into

predicates of times, presupposing its complement to be temporally homogeneous or

not. We can therefore define the lexical entries of our tenses by assuming a ‘partial

function’ analysis of presupposition and following the notation proposed by Heim

and Kratzer (1998) as following:

(25) kHOMkg,c ¼ kf ˛ Dhi,ti & "t ˛ Di : f(t) fi "t0 (t0� t fi f(t0) ) . f

180 F. Arosio

123

(26) kN-HOMkg,c ¼ kf ˛ Dhi,ti & "t ˛ Di : f(t) fi : "t0 (t0� t fi f(t0) ) . f

As you can see from the tense architecture given, we are assuming the predicate

restriction associated with the Tempora Infecta is HOM and the one associated with

the Tempora Perfecta is N-HOM. In the case of the Tempora Infecta, the predicate

restriction HOM takes a predicate of times as its argument and it gives it back if the

latter is temporally homogeneous; in the case of the Tempora Perfecta, N-HOM

takes a predicate of times as its argument and it gives it back if it is temporally non-

homogeneous. As we will see later, the temporal predicate restrictions are what

bring the contribution of tense into the aspectual interpretation of a sentence. The

property of times obtained is, in turn, saturated by the denotation of the temporal

pronoun occupying the TP-Spec position.

I will assume Heim’s (1994) straightforward referential analysis of tense for the

temporal pronoun occupying the TP-Spec position, but a quantificational approach

can also be assumed for the proposal I am making. According to a referential

analysis, tenses are referential expressions denoting times and verbs have an extra

argument slot for tenses as represented below

(27) klovek :¼ kykxkt [love(y)(x)(t)]

(28) She loved John

LF: love (John) (shej) (PASTi)

As all other referential expressions, tenses bear an index, and sentence (28) is true if

the individual denoted by shej loves John at the time denoted by PASTi. Clearly the

time denoted by PASTi should be a time before the speech time, according to what

(28) says. An account for this fact has been given by Heim (1994) in her comments

on Abusch’s (1994) theory of tense. Heim observes that, just like a free instance of

shej can only refer to female individuals, a free instance of PASTi must refer to a

time before the time of the utterance. In Heim’s proposal the lexical entries for the

past and the present tenses are therefore formally defined in the following way

(29) kPASTikg,c ¼ g(i) when g(i) < tc undefined otherwise

(30) kPRESikg,c ¼ g(i) when :g(i) < tc undefined otherwise

Definition (29) says that the denotation of a temporal variable PASTi is defined if

the value that the assignment function g assigns to the index i is a time before tc, the

time of the context of utterance; if it is defined, it is equal to the value that the

assignment function g assigns to the index i. Definition (30) says that the denotation

of a temporal variable PRESi is defined if the value that the assignment function gassigns to the index i is a time not before tc; if it is defined, it is equal to the value

that the assignment function g assigns to the index i. The truth conditions for (28)

will therefore be


123

(31) k love(John)(Mary)(PASTi)kg,c ¼ 1 iff John loves Mary at g(i) when

g(i) < tc, undefined otherwise

We can now see how the homogeneity proposal can be developed in this frame-

work. Consider the following Italian and French sentences

(32) (a) Mario amava Eva [Imperfetto] ITA

Mario love-3sgPASTimp Eva

(b) Mario aimait Eva [Imparfait] FR

Mario love-3sgPASTimp Eva

Mario loved Eva

As we did before, let us assume that the Italian predicate amare (to love) and the

French predicate aimer (to love) have an explicit argument slot for times. Given our

temporal architecture, the temporal predicate kt(Mario-love-Eva (t)) is merged with

the tense head and the temporal predicate obtained is in turn filled by the temporal

pronoun PASTi, as shown below

(33) PASTi (HOM( λt(Mario-love-Eva (t)) ))

TP

T'

T

PASTi

HOM<it,it>

VP<i,t>

λt(Mario-love-Eva(t))

Given the definedness conditions introduced by the temporal variable and by the

predicate restriction HOM, (33) is defined if

(34) kPASTi kg,c is defined,

and if

(35) kHOM(kt(Mario-love-Eva(t)))kg,c, is defined.

That is, if g(i) \ tc, and if kt(Mario-love-Eva(t)) is homogeneous, which is the

case, since the temporal predicate kt(Mario-love-Eva(t)) denotes a homogeneous

property of times given that it is derived from the state predicate ‘‘to love’’, which

we know to be divisible. In this case,

(36) kPASTi (HOM(kt(Mario-love-Eva(t))))kg,c ¼ 1 iff the time denoted by

PASTi is a past time at

which Mario loves Eva.

182 F. Arosio

123

3.2 Durative for- and Sinced-Adverbials

I will assume durative for- and sinceD-adverbials to be temporal modifiers, namely

functions from time properties to time properties. Intuitively, while for-adverbials

take a temporal predicate and give it back by saying that it is true of an interval of a

certain length (Dowty 1979), sinceD-adverbials take a temporal predicate and, by

introducing an extended now interval,7 they give back a predicate of times abutting

an interval of a certain length of which the original temporal predicate is true (Musan

2000; Von Stechow 2002). While previous studies mainly concentrated on the input

of these adverbs, i.e. on the particular characteristics of their semantic arguments

(Dowty 1979; Krifka 1986, 1992; Hinrichs 1985), I will focus on their output, namely

on the characteristics of the result of their application. The interesting property of

these adverbs is that while for-adverbials make the temporal predicate they modify

non homogeneous, sinceD-adverbials make them homogeneous.

3.2.1 For-adverbials

That for-adverbials turn all temporal predicates into quantized temporal predicates

(which are, a fortiori, non-homogeneous) is easy to see. I will assume that

for-adverbials introduce universal quantification over times (Dowty 1979; see also

Krifka 1992 and Hinrichs 1985)

(37) For x time :¼ kPkt(dTIME(t) ¼ x & "t0(t0˝ t fi P(t0))where d is function measuring the length of an interval onto a temporal

scale.

The definition in (37) says that a for-adverbial takes a predicate of times P as its

argument and it gives back a temporal predicate denoting a set of times i, which are

x long, and for which every subinterval j of i belongs to the set of times denoted by

the original predicate P. More intuitively, the adverb says that the predicate P is true

of an x-long interval and of every subinterval of this interval. These adverbs

quantize the temporal predicate they modify, namely they give a temporal predicate

which, if true of an interval, is false of every proper subinterval of that interval (see

Krifka 1990, 1992)

(38) P is temporally quantized iff "i[P(i) fi "j [j�i fi :P(j)]]where i and j are intervals.

Notice that being quantized is different from being non-homogeneous given the

different scope of negation in (22) and in (38). In fact, given a temporal predicate P,

Quantized(P) entails NON-HOM(P), but NON-HOM(P) does not entail Quan-

tized(P). Consider now the LF of the time predicate John love Eva for two years, as

represented below

7 Dowty (1979) and McCoard (1978).


123

(39) kt(dYEARS(t) ¼ 2 & John-love-Eva(t) &"t0 (t0 ˝ t fi John-love-Eva(t0)))

As shown by the picture below, if (39) is true of an interval i, it is false of every

proper subinterval j of i trivially because j cannot be 2 years long if it is properly

included in i

John love Eva for two years

i =2-year long

j ≠ 2 years

This shows that for-adverbials quantize the temporal property they modify.

One important observation concerns the status of adverbs like ‘‘per meno di due

anni’’ (for less than two years) since every subinterval of a less than two year long

interval is in turn less than 2 years long and, therefore, these adverbs represent a

potential counterexample to the quantizing nature of durative for-adverbials that we

have argued so far.8 However it is clear that adverbs like ‘‘per meno di due anni’’

(for less than two years) require maximality of the temporal property they modify, if

we consider the Italian contrasts below

(40) (a) Gianni ha amato Eva per meno di due anni . . . ?? e ancora la ama

John has loved Eva for less than two years. . . ?? and still her (he)

loves

Leo has loved Eva for less than two years. . . ?? and he still loves her

(b) Gianni ha amato Eva per due anni . . . e ancora la ama

John has loved Eva for two years. . . and still her (he) loves

Leo has loved Eva for two years. . . and he still loves her

Thus, ‘‘to love Eva for less than two years’’ intuitively means to be in love with Evamaximally for less than two years. The right definition of ‘‘for less than two years’’

is therefore

(41) kPkt[(dYEARS(t) � 2 & P(t) & :9t0(t ˝ t0 & P(t0)) & "t00(t00 ˝ t fi P(t00))]

This temporal modifier quantizes the temporal property it modifies.

3.2.2 SinceD-adverbials

SinceD-adverbials turn all temporal predicates into homogeneous temporal predi-

cates. This is also very easy to see. As proposed by von Stechow (2002), these

adverbials introduce an extended-now interval. I will assume a slightly modified

version of von Stechow’s definition of German durative seit-adverbials, which

works better for Italian and French

8 The objection is from Andrea Bonomi (p.c).

184 F. Arosio

123

(42) since x time :¼ kPkt9I( I abuts t & P(I ¨ t) & dTIME(I) ¼ x)

The definition in (42) says that a sinceD-adverbial applies to a temporal predicate P

and it gives back a temporal predicate denoting a set of time i such that there is an

x-time long interval I abutting i (i is at the right-hand edge of I) and the set union of

I and i belongs to the set of times denoted by the original predicate P. Intuitively, it

gives back a property of times abutting an x-long interval of which the original

predicate is true. Consider now the Italian and French temporal predicates in (44)

obtained by the application of the sinceD-adverbials in (43) to the temporal pred-

icate kt(Mario love Eva(t)) that we know to be homogeneous.

(43) (a) da due mesi (since 2 months) :¼ kPkt9I(I abuts t & P(I ¨ t)

&dMONTH(I) ¼ 2) ITA

(b) depuis deux mois (since 2 months) :¼ kPkt9I(I abuts t & P(I ¨ t)

& dMONTH(I) ¼ 2) FR

(44) kt9I(I abuts t & Mario love Eva (I ¨ t) & dMONTH(I) ¼ 2)

The temporal predicate in (44) denotes a set of times i that are abutted by a 2-month

long interval I for which Mario loves Eva at i plus I, as shown by the picture below

i

Mario love Eva time line

I = 2 month long

The ‘‘homogenizing’’ nature of sinceD-adverbials is easy to understand. Consider

the picture below

i

Mario love Eva time line

jI = 2 month long

If (44) is true of the interval i , it should be true of every subinterval j of i. Since the

duration two months is relative to the time I introduced by existential quantification

and not to i, and I abuts j, I ¨ j will always be included in I ¨ i. Therefore, if Mario

loves Eva at I ¨ I, he loves Eva at I ¨ j, and this shows that the obtained predicate is

homogeneous. One important empirical question concerns whether the measuring

function d should apply to (I ¨ t) and not only to I, namely if the x-long interval

should include the reference time introduced by the tense. If this is the case we have

a potential problem while combining a since-D adverbial with a tempus Infectum,

since in this case the temporal property obtained via since-D modification is no

longer temporally homogeneous. I emphasise that this is a hardly decidable


123

empirical question, the intuition is that the measuring function d should apply to I

and not to (I ¨ t). However, there is an additional different and independent

argument that favors the homogenising nature of since-D adverbials. Consider the

Italian sentences below:

(45) (a) Ogni volta che Maria corse nel parco per un’ora Leo dormiva

Every time that Mary run-3sgPASTperf in-the park for two hours,

Leo sleep-PASTimp

Every time that Mary ran in the park for two hours, Leo was sleeping

(b) ??Ogni volta che Maria corse nel parco per un’ora Leo dormiva da

due ore

Every time that Mary run-3sgPASTperf in-the park for one hours,

Leo sleep-3sgPASTimp since 2 hours

–

(c) Ogni volta che Maria entro Leo dormiva da due ore

Every time that Mary come-PAST.PERF in, Leo sleep-PAST.IMP

since two hours

Every time that Mary came in, Leo had been sleeping for two hours

According to their meanings, the temporal variables introduced by the tenses are

bound by universal quantification in these sentences, their LF being something like:

"i[P(PASTi) fi Q(PASTi)]. The contrast among (45) (a), (b), and (c) shows that

when the reference time introduced by the tense in the embedded clause has some

length (as required by the for-one-hour adverbial modification) the since-D

adverbial in the embedded clause is bad. On the other hand, when the reference time

is a point (as in (c) where the embedded clause describes a punctual event) the

since-D adverbial modification is good. This suggests that a since-D adverbial

requires the reference time that the interval I abuts, to be a point of time. If this is

correct, since-D adverbial modification gives homogeneous properties of times

since the property of a point is vacuously homogeneous, having a point of time no

proper subparts by definition.

3.3 Tense and durative adverbials

While the quantizing character of for-adverbials explains why they do not combine

with a Tempus Infectum but do with a Tempus Perfectum, the homogenizing nature

of sinceD-adverbials explains why they combine with a Tempus Infectum, but not

with a Tempus Perfectum. Consider the following Italian and French sentences

where a for-adverbial combines with the Passato Remoto and with the Passe Simple

but not with the Imperfetto and the Imparfait and the associated LFs.

(46) (a) Mario amo Eva per due anni [Passato Remoto þ for adverbial] ITA

Mario love-3sgPASTperf Eva for two years

(b) Mario aima Eva pendant deux ans [Passe Simple þ for adverbial] FR

Mario love-3sgPASTperf Eva for two years

186 F. Arosio

123

Mario loved Mary for two years

PASTi

TP T'

T |

N-HOM<it,it>

VP<i,t>


VP<i,t>

for two years λPλt(δYEAR(t)= 2 & P(t)& ∀t'(t' ⊆t →P(t'))

Adv<it, it>

|

(47) (a) ?? Mario amava Eva per due anni [Imperfetto þ for adverbial] ITA

Mario love-3sgPASTimp Eva for two years

(b) ?? Mario aimait Eva pendant deux ans [Imparfait þ for adverbial] FR

Mario love-3sgPASTimp Eva for two years

--

PASTi

TP T'

T |

HOM<it,it>

VP<i,t>


VP<i,t>

for two years λPλt(δYEAR(t) = 2 & P(t)& ∀t'(t' ⊆t →P(t'))

Adv<it, it>

|

In both (46) and (47) we start the derivation of the LFs from the state predicate

kt(Mario-love-Eva(t)) and we obtain the non-homogeneous temporal predicate

kt(dYEAR(t) ¼ 2 & Mario-love-Eva(t) & "t¢(t¢ ˝ t fi Mario-love-Eva(t‘)) by

for-modification. In (46) this temporal predicate is merged with the tense head

N-HOM and in turn applied to the temporal variable PASTi. The definedness con-

ditions for the sentence obtained require the denotation of PASTi to be a past time

and kt(dYEAR(t) ¼ 2 & Mario-love-Eva(t) & "t¢(t¢ ˝ t fi Mario-love-Eva(t‘)) to

be non-homogeneous, which is the case. Therefore if PASTi refers to a time before

the time of the context of utterance, the sentence is felicitous when uttered in that

context, and it is true if and only if PASTi denotes a time which is 2 years long and

is characterized by Mario’s loving of Eva. On the contrary, in (47) the presuppo-


123

sition carried by HOM cannot be met, since kt(dYEAR(t) ¼ 2 & Mario-love-Eva(t)

& "t¢(t¢ ˝ t fi Mario-love-Eva(t‘)) is temporally non-homogeneous. A presup-

positional failure in (47) is the cause of its ungrammaticality.

On the other hand, the homogeneous character of a temporal predicate obtained

by sinceD-modification allows it to combine with the Imperfetto in Italian and with

the Imparfait in French for the opposite reason, but it prevents it from combining

with the Italian Passato Remoto and the French Passe Simple. Consider (48), where

the Italian Passato Remoto and the French Passe Simple combine with sinceD-

adverbials.

(48) (a) ?? Mario amo Eva da due anni [Passato RemotoþSinceD adverbial] ITA

Mario love-3sgPASTperf Eva since two years

(b) ?? Mario aima Eva depuis deux ans [Passe SimpleþSinceD adverbial] FR

Mario love-3sgPASTperf Eva since two years

--

PASTi

TP T'

T |

N-HOM<it,it>

VP<i,t>


VP<i,t>

since two years λPλt∃I( t abuts I & P(I∪t) & δYEAR(I)=2)

Adv<it, it>

|

As we can observe from its LF, sentence (48) suffers from a presuppositional failure,

since the non-homogeneity presupposition carried by the N-HOM head cannot be met

by the temporal property kt9I(I abuts t & Mario-love-Eva(I ¨ t) & dYEAR(I) ¼ 2),

which is homogeneous. Clearly, such a temporal property combines with a Tempus

Infectum, since it meets the homogeneity presupposition carried by the HOM head. In

this proposal it is implicitly assumed that adverbials cannot have scope between the

homogeneity restriction and the temporal pronoun, namely that tenses cannot be

decomposed in the logical syntax: k PAST þ HOM kg ¼ kP. HOM(P).kt. t\tc.

P(t) ¼ 1. This rules out the LF: PAST[SINCE x TIME[HOM [VP]]].

However, state predicates combine in Italian with the Passato Remoto and in

French with the Passe Simple without any overt for-adverbial modification, as

shown by the sentences below

(49) (a) Mario amo Eva [Passato Remoto] ITA

Mario love-3sgPASTperf Eva

188 F. Arosio

123

(b) Mario aima Eva [Passe Simple] FR

Mario love-3sgPASTperf Eva

Mario loved Eva

Interestingly, sentence (49) is ambiguous between a perfective and an inchoative

reading. According to its inchoative reading, whose analysis will be given in

Sect. 4, sentence (49) says that Mario fell in love with Eva in the past. According to

its perfective reading, sentence (49) describes Mario’s love for Eva which is entirely

terminated in the past. I assume that this reading is obtained by the application of a

covert operator which maximalizes the time at which the described state holds in

order to meet the non-homogeneity presupposition associated with the Passato

Remoto and the Passe Simple. The maximality operator is defined as follows

(50) MAX :¼ kPkt(P(t) & :9t0(t ˝ t0 & P(t0))

This operator maximalizes, and thus quantizes, a temporally homogeneous state

predicate. As I have said, a quantized predicate is non homogeneous (Krifka 1990,

1992). The application of MAX to the homogeneous predicate ‘‘Mario-love-Eva’’

will give us the following LF for (49)

(51) PASTi(N-HOM(kt(Mario-love-Eva(t) & :9t0(t ˝ t0 & Mario-love-Eva(t0))))

According to (51), the application of the MAX operator correctly gives us that

Mario’s love for Eva is terminated in the past. While the application of the MAX

operator quantizes the temporal predicate it modifies and is responsible for the

perfective interpretation of (49), for adverbials only quantize the temporal predi-

cates they modify without entailing perfectivity. In fact, nothing prevents Mario’ssickness to hold up to the speech time given the LF associated with (49). At first

glance, this would seem to be an unwelcome prediction since in Italian (and French)

we usually understand for-adverbial modified Passato Remoto sentences to describe

events which are no longer going on at the present. However, if we have a brief look

at Italian data available on the net we do find sentences like

(52) La Cupola del Brunelleschi fu per lungo tempo, ed e ancora, il simbolo

visivo dell’intera citta.

[www.operaduomo.firenze.it/storia/ spazio/spaziosacro7.htm]

The Cupola del Brunelleschi be-3sgPASTperf for long time, and it

be-3sgPRES still, the symbol visual of-the whole city.

The Cupola del Brunelleschi was for long, and it still is, the visual symbol ofthe whole city.

which require the for-modified states not to be necessarily terminated before the

speech time. I explain the fact that we understand the eventuality as terminated in

the past as an effect of pragmatic factors associated with for-adverbials. Since for-

adverbials are upward entailing (John loved Mary for three years entails John lovedMary for two years), in order to be maximally informative, the length indicated by

the for-adverbial should be the length of the maximal interval for which the pred-

icate is true (see Krifka 1990, 1992 for further discussion). This explains why we


123

understand that Mario’s love for Eva is terminated in the past while processing the

temporal meaning conveyed by (49).

Until now we have been looking at temporal predicates derived from state

predicates like ‘‘love’’ or ‘‘be sick’’, which we argued are temporally homoge-

neous. But what happens when we have event predicates? Is temporal homogeneity

also relevant for the tense selection in sentences with event predicates?

4 Homogeneity and event predicates

In his 1967 article, Davidson proposes that action sentences like (53) (a) should be

analysed as sentences expressing an existential quantification over events. The idea

is that sentence (53) (a) says that there is a past event of Jones’ buttering the toast.Davidson’s motivations for assuming events in the basic ontology of natural lan-

guage mainly concern an analysis of adverbs which is sensible to logical entail-

ment.9 As Davidson observes, every sentence in (53) is entailed by its predecessors

(53) (a) Jones buttered the toast

(b) Jones buttered the toast with a knife.

(c) Jones buttered the toast with a knife in the bathroom.

Davidson argues that if we assume that the logical form of an action sentence

contains an event variable which stands for the event described by the sentence, the

entailment relations from a. to c. follow straightforwardly from the laws of predicate

logic, since these sentences express relations between individuals and events.

According to Davidson’s analysis, the sentences in (53) will have an LF in which an

event variable fills an extra argument slot of the verb and of the adverbs

(54) (a) 9e[butter(Jones, the toast, e)]

(b) 9e[butter(Jones, the toast, e) & With(a knife, e)]

(c) 9e[butter(Jones, the toast, e) & With(a knife, e) & In(the bathroom, e)]

The entailment relations follow, therefore, from the rule of conjunction elimination.

In his original proposal Davidson, points out that not all verbs have an underlying

Davidsonian argument. While action sentences do have one, fact sentences do not.

According to Davidson, fact sentences are sentences describing states whose LF will

be something like

(55) John loves Mary

LF: love(John, Mary)

In contrast to its original formulation, Davidson’s proposal has been extended to

state verbs as well as event verbs by assuming that the logical form of a state

sentence contains a state variable which stays for the state described by the sen-

tence, as represented below

9 See Thomason and Stalnaker (1973) for an alternative analysis and Parsons (1990) for a discussion

about different approaches.

190 F. Arosio

123

(56) John loves Mary

LF: 9s [love(s)(John)(Mary)]

The development of this approach, which in the literature is called neo-Davidso-nian, also assumes thematic role predicates in the LF of a state sentence (Higgin-

botham 1985; Parson 1990).10 The original Davidsonian proposal has however been

defended and assumed by some other authors (Galton 1984; Lobner 1988; Herweg

1991; Katz 1995; 2000) who believe the stative/non-stative distinction to be based

on the fact that while event verbs have an underlying event argument, state verbs do

not. Katz (2000) has recently given a number of arguments in favour of this thesis.

In general, he claims that on the neo-Davidsonian account we should find some

semantic parallels between state sentences and event sentences given their parallel

logical structure and if these parallels are not found, there is evidence against the

neo-Davidsonian account. Katz shows that these parallels are not found in a number

of important domains such as anaphora, nominalization and perception verbs, but

the most convincing argument concerns adverb modification. As is known, there are

classes of adverbs, such as manner and instrumental adverbials, which appear with

event verbs but not with state verbs

(57) (a) Bill buttered the toast carefully.

(b) ??Bill owned the knife carefully.

There are also classes of adverbs, such as modal adverbials, that appear with both

event verbs and state verbs

(58) (a) Bill probably buttered the toast.

(b) Bill probably owned the knife.

But interestingly, we do not find adverbials which appear with state verbs but not

with event verbs. That is to say, it does not seem to be the case that there is a class of

state adverbials. As Katz observes, on the neo-Davidsonian approach state adverbs

might be expected. These adverbs might denote properties of states, like state verbs

do. On the other hand, according to Katz, if state verbs denote properties of times,

stative adverbs are not expected. This is because, if an adverb appears with a state

verb it should be a temporal modifier or a propositional operator, and therefore it

should also be able to appear with event verbs once the event variable is existen-

tially closed. Therefore, following Herweg and Katz, in this paper I will assume that

while state predicates denote properties of times, event predicates denote properties

of events; I will assume that while a state predicate has an explicit argument for

times, an event predicate has an explicit argument for events, as shown below

10 With the term neo-Davidsonian we usually refer to the semantic implementation of Davidson’s

intuition. In its development, thematic role predicates are also introduced into the logical form of a

sentence. In most of these proposals Davidson’s intuition is extended to the analysis of state sentences as

well. Following Katz (2000) I will refer here to those accounts that assume that all sentences have

underlying Davidsonian arguments, in contrast to Davidson’s original proposal.


123

(59) (a) State Predicate: k to love k :¼ kykxkt(love(y)(x)(t))

(b) Event Predicate: k to sleep k :¼ kxke(sleep(x)(e))

As illustrated, since a state predicate like the one in (59)(a) denotes a property of

times and belongs to the logical type hi, ti, it can be modified by a durative temporal

adverbial or saturated by tense; on the contrary, since an event predicate like the one

in (59)(b) denotes a property of events and belongs to the type hev, ti, it needs to be

type shifted to combine with a temporal adverbial or tense. As I sketched in the first

paragraph, implicit aspectual operators are responsible for type shifting of event

predicates by localizing the event described with respect to a time. Following

Kratzer’s and von Stechow’s definitions of aspect, we assume that the common

Romance aspectual operators are the inclusion operator, and its converse. Let us

discuss them again in order to clearly understand how they are connected with the

temporal homogeneity proposal I am making. The inclusion operator, which is

called the Perfective operator since it is responsible for the perfective reading of an

event sentence, localises the event described within a time: it takes a property of

events and it gives back a property of time including the temporal trace of the event.

Its complement, which is called the Imperfective operator since it is responsible for

the imperfective reading of an event sentence, localises the event described as

surrounding a time: this operator takes a property of events and it gives a property of

times properly included in the temporal trace of the event. The definitions we gave

for the aspectual operators are given again below

(60) (a) Perfective ¼: kPFVk ¼ kPkt9e(t ˚ s(e) & P(e))

(b) Imperfective ¼: kIPVk ¼ kPkt9e(s(e) t & P(e))

where s is a function taking an event and giving its temporal trace.11

According to what (60) (a) and (60) (b) say, a temporal property formed via the IPV

operator will always be homogeneous, as shown by the picture below

||τ(e)|| = i time line

j k

This is because, if j is included in the temporal trace of e, every subinterval k of jwill also be included; therefore we expect a temporal property formed via the IPV

operator to combine with a Tempus Infectum. On the other hand, a temporal

property formed via the PFV operator will always be non-homogeneous, as shown

by the picture below

11 In (60) (b) we do not consider the modal aspect of the Imperfective aspect since it clearly goes beyond

the purpose the paper.

192 F. Arosio

123

||τ(e)|| = i time line

j k

As we can see, if j includes the temporal trace of e, it is not the case that every

subinterval k of j will include the temporal trace of e; therefore we expect a tem-

poral property obtained via the PFV operator to combine with a Tempus Perfectum.

Now, consider the following Italian Passato Remoto and French Passe Simple event

sentences and their associated LF:

(61) (a) Mario mangio una mela [Passato Remoto] ITA

Mario eat-3sgPASTperf an apple

(b) Mario mangea une pomme [Passe Simple] FR

Mario eat-3sgPASTperf an apple

Mario ate an apple

λe(Mario-eat-an-apple(e))

PFV: λPλt∃e(t ⊇ τ(e) & P(e))

TP

T' PASTi

N-HOM<it,it> VP<ev t>

AspP<i, t>

The temporal predicate obtained via the PFV operator meets the non homogeneity

condition associated with the N-HOM head. This explains why event sentences have

perfective interpretations under the Passato Remoto, as shown by the bad contin-

uation of the sentence below

(62) (a) Mario mangio una mela, *e la sta ancora mangiando

[Passato Remoto] ITA

Mario eat-3sgPASTperf an apple, *and it he-is still eating

(b) Mario mangea une pomme, *et il est encore en train de la manger

[Passe Simple] FR

Mario eat-3sgPASTperf an apple, *and he is still in busy of it eating

Mario ate an apple, *and he is still eating it

The Italian Passato Prossimo (perfect) and French Passe Compose (perfect) behave

analogously, as shown by the sentences below and their bad continuation


123

(63) (a) Mario ha mangiato una mela, *e la sta ancora mangiando

[Passato Prossimo] ITA

Mario have-3sgPRES eaten an apple, *and it he-is still eating

(b) Mario a mange une pomme, *et il est encore en train de la manger

[Passe Compose] FR

Mario have-3sgPRES eaten an apple, *and he is still in busy of

it eating

Mario has eaten an apple, *and he is still eating it

The case of the Italian Passato Prossimo (perfect) and Trapassato Prossimo

(pluperfect) will be discussed in Sect. 5.

Going back to state predicates, I mentioned in Sect. 3.3 that they are ambiguous

between a perfective and an inchoative interpretation when they combine with a

past perfective tense in the Romance languages considered here. Consider again the

following Italian and French sentences where the state predicate to love combines

with a past perfective tense:

(64) (a) Mario amo Eva [Passato Remoto] ITA

Mario loved-3sgPASTimp Eva

(b) Mario aima Eva [Passe Simple] FR

Mario loved-3sgPASTimp Eva

(i) Perfective Reading: Mario loved Mary(ii) Inchoative Reading: Mario (suddenly) fell in love with Mary.

These sentences are ambiguous. They can be used to convey that there is some past

terminated Mario’s loving of Eva state or that Mario fell in love with Eva in the

past. In Sect. 3.3 we argued that the perfective reading is obtained via a maxi-

malization of the state predicate in order to meet the non-homogeneity condition

associated with the past perfective tense. In order to account for the inchoative

reading, I propose a similar explanation. In order to meet the non-homogeneity

condition associated with the past perfective tense, I assume the inchoative reading

to be obtained by the application of the covert ‘‘achievementizing’’ operator below:

(65) BECOME ¼: kPke (becomee(P)) - where the event e is a becoming

with result P, where P is a state.12

As shown by the LF of (64) represented in (66) below, the temporal predicate

obtained via the become operator is in turn perfectivized and selected by the past

perfective tense. According to their inchoative reading, the sentences in (64) say

that Mario suddenly fell in love with Mary. According to this reading the appli-

cation of the BECOME operator should denote the property of a punctual

achievement event whose temporal trace is a point of time. This explains why the

IPV operator cannot apply to the result of BECOME; the IPV operator requires the

12 See Dowty (1979) for a further discussion.

194 F. Arosio

123

reference time to be properly included in the temporal trace of the event described

and this is impossible for punctual events like the one described by (64) since their

temporal trace is a point of time, according to their inchoative reading.

ð66Þ

VP<i,t>

λt(Mario-love-Eva(t))PFV<et,it>: λPλt∃e(t⊇τ(e) & P(e))

TP T'<i,t>

PASTi

N-HOM<it,it>

AspP<i,t>

BECOME<i t, ev t> : λPλe (becomee(P))

<ev,t>

(67) PASTi (N-HOM (kt9e(t ˚ s(e) & becomee(Mario-love-Eva))))

Under its inchoative reading, the sentence is true if and only if g(i) is a past time

including the time of a becoming event. Interestingly, the analysis predicts that the

result state can hold up to speech time since the embedded state of result is not

temporally constrained, as shown by the sentence below (Bertinetto 1991)

(68) La sua squadra preferita aveva perso. Gianni ne ebbe un forte mal di

pancia che ancora non gli e passato.

His favourite team had lost. Because of this, Gianni got a belly ache,which is still paining him.

4.1 Event predicates and for-adverbials

It is a known fact that for-adverbials combine with atelic predicates but not with telic

ones. Given Dowty’s definition of for-adverbials, the explanation of this fact runs as

follows in the system I am proposing. Given that for-adverbials are temporal mod-

ifiers and they introduce universal quantification over subintervals, they combine

with state predicates which denote homogeneous properties of times. As we have

seen, the temporal property obtained by the application of a for-adverbial is quan-

tized and it meets the non-homogeneity presupposition associated with a Tempus

Perfectum. Event predicates denote properties of events but they cannot combine

directly with a temporal modifier such as a for-adverbial. They must be type shifted

into a time predicate by the application of the aspectual operators as we have seen in

Sect. 1 before they can combine with tense or with a temporal modifier. According to

the definition given in (60)(a), the PFV operator requires t ˚ s(e), namely, that the

temporal trace of the event is either properly included or equal to the time t.According to these conditions, when the temporal property is obtained via the proper-

inclusion relation, it is non-homogeneous, and it does not meet the universal

quantification associated with for-adverbials. Thus, it does not combine with a for-

adverbial. When the relation associated with PFV is the identity relation, the homo-

geneity character of the obtained temporal property depends on the homogeneity


123

character of the event predicate that the PFV operator modifies; thus, a temporal

property obtained via the identity relation combines with a for-adverbial depending on

the homogeneity character of the modified event property. When the event predicate is

telic, like build a house it is non-divisible. In this case, a subinterval of the temporal

trace of the event for which this predicate is true is not the temporal trace of an event of

the same type. In this case, the temporal property obtained via the PFV operator under

the identity relation will be non homogeneous; therefore, the temporal property ob-

tained via the identity relation from a telic event predicate does not combine with a for-

adverbial. This is because the obtained temporal property does not meet the universal

quantification introduced by this adverbial.13 When the event predicate is atelic like

sleep (and thus divisible down to a certain point), the temporal property obtained via

the PFV operator under the identity relation will be homogeneous (down to a certain

point) therefore, a temporal property obtained via the identity relation from an atelic

event predicate combines with a for-adverbial. This is because it meets the universal

quantification introduced by the adverbial. This explains why activity predicates such

as sleep combine with a for-adverbial but telic predicates such as build a house do not

under PFV. More generally, this explains the distribution of for-adverbials across the

different Vendlerian verb classes under PFV. Let us consider IPV modification now.

According to the definition of IPV given in (60)(b), every event predicate can combine

with a for-adverbial under IPV. This is because IPV requires that t � s(e). According

to this condition, the temporal property obtained via IPV is the property of being a timeproperly included in the temporal trace of an event. This temporal property is

homogeneous and it meets the universal quantification introduced by a for-adverbial.

This shows that every event predicate can in principle combine with a for-adverbial via

IPV under a Tempus Perfectum. This prediction is borne out by the following Italian

facts:

(69) Tullio disegno il suo ritratto per circa dieci minuti; poi dovette smettere

Tullio draw-3sgPASTperf his portrait for circa ten minutes; then he had

to stop.

Tullio had been drawing his portrait for circa ten minutes; then he stopped(Bertinetto 1991)

Here the for-adverbial modifies a subinterval of the temporal trace of the Tullio’sdrawing his portrait event. Tullio draw his portrait is an accomplishment predicate

and the event it denotes in (69) is not completed, as shown by the continuation.

13 Actually, in the case of achievement predicates the explanation runs differently. In fact, the temporal

trace of an achievement predicate, which denotes a punctual event, is the property of a point of time. This

property is vacuously homogeneous and it satisfies the condition associated with the universal quantifi-

cation introduced by for-adverbials. The reason why a temporal property obtained via the identity relation

from an achievement predicate does not combine with a for-adverbial is that the temporal trace of an

achievement event does not have sensible length to be measured by for-adverbials.

196 F. Arosio

123

Notice that in (69) the tense morphology is past perfective though the semantic

aspect occurring in its LF is IPV; to make it clear, see the LF for (69) below:14

(70) [PAST N-HOM for 10 minutes IPV Tullio draw his portrait]

This fact shows that the tense morphology does not depend on the semantic

aspectual operator. As we have seen in (69) and (70), a Tempus Perfectum can

combine with a temporal property obtained via the IPV aspectual operator if this

property is made non-homogeneous by for-adverbial modification. It is the presence

of the adverbial, which quantizes the temporal predicate obtained via IPV that

requires a Tempus Perfectum. One important question concerning these facts is why

sentences like (69) are a little odd without an appropriate context or when uttered

without a pragmatic relevant continuation like the one in (69). This is explained by

pragmatic factors associated with the meaning of for-adverbials. As we have seen

before, since for-adverbials are upward entailing (John ran in the park for fiveminutes entails John ran in the park for four minutes), in order to be maximally

informative, the length indicated by the for-adverbial should be the length of the

maximal interval for which the predicate is true. Under IPV this last condition is not

met. For this reason sentences like (69) are odd without an appropriate context.

However, pragmatic principles can be violated, and this can happen in appropriate

contexts, like the one in (69).

5 The perfect in some Romance languages

The presence of the Perfect changes the distribution of the adverbials under the

same tense morphology (borne by the auxiliary verb) as we have seen in sentences

(8), (9), (10) and (11) d–e repeated below

Italian

(8) per-adverbials (9) da-adverbials

(d) E stato buio per due ore (d) ?? E stato buio da due ore

Be-3sgPRES been dark for 2 hours Be-3sgPRES been dark since 2 hours


PASSATO PROSSIMO PASSATO PROSSIMO

(e) Era stato buio per due ore (e) ?? Era stato buio da due ore

Be 3sgPASTimp been dark for 2 hours Be-PASTimp been dark since 2 hours

It had been dark for two hours –

TRAPASSATO PROSSIMO TRAPASSATO PROSSIMO?

14 IPV is required here by the semantics of the temporal adverbial which introduces universal quanti-

fication over subintervals. The prediction is important here since the analogous English example is

ungrammatical *Tullio drew his portrait for ten minutes; then he had to stop.


123

French

(10) pendant-adverbials (11) depuis-adverbials

(d) La fenetre a ete sale pendant deux jours (d) ??La fenetre a ete sale depuis deux jours

The window have-3sgPRES been dirty for

2 days

The window have-3sgPRES been dirty

since 2 days


PASSE COMPOSE PASSE COMPOSE

(e) La fenetre avait ete sale pendant deux jours (e) ??La fenetre avait ete sale depuis deux

jours

The window have-3sgPASTimp been dirty

for 2 days

The window have-3sgPASTimp been dirt.

since 2 days

The window had been dirty for two days –

PASSE ANTERIEUR PASSE ANTERIEUR

In recent work, von Stechow (2002) argued that Present Perfect morphology in the

Italian (8)(d) and (9)(d) is a variant of the past perfective inflection in Romance

languages and he suggested that the same temporal meaning can be spelled out by

the past perfective or by the Present Perfect in these languages (see also Hornstein

1990). While, on the one hand, this proposal correctly predicts the adverbial dis-

tribution in (d), on the other hand, it does not seem entirely appropriate if one

considers the contrast below

(71) (a) Maria sposera un uomo che ha vissuto a NY. [Passato Prossimo] ITA

Maria marry-3sgFUT a man who have-3sgPRES lived in NY.

(b) Maria epousera un homme qui a vecu a NY. [Passe Compose] FR

Maria marry-3sgFUT a man who have-3sgPRES lived in NY.

Maria will marry a man who has lived in NY.

(72) (a) Maria sposera un uomo che visse a NY. [Passato Remoto] ITA

Maria marry-3sgFUT a man who live-3sg.PASTperf in NY.

(b) Maria epousera un homme qui vecut a NY. [Passe Simple] FR

Maria marry-3sgFUT a man who live-3sg.PASTperf in NY.

Maria will marry a man who lived in NY.

Sentences (71) are temporally ambiguous. They can mean: (i) that Mary will marry

a man in the future with respect to the speech time and that this man lives in NY in

the past with respect to the speech time; or (ii) that Mary will marry a man in the

future with respect to the speech time and that this man lives in NY in the past with

respect to the future marrying event (thus his living in NY may be in the future with

respect to the speech time), as represented by (73) and (74) below

198 F. Arosio

123

ð73Þmarrying living-in-NY speech time

ð74Þmarrying living-in-NY speech time

On the contrary, the sentences in (72) are not temporally ambiguous. According to

what they says, the living-in-NY can only be in the past with respect to the speech

time, i.e., (72) can only have the temporal interpretation represented in (73). If we

consider the Past Perfective and the Present Perfect morphology to be free spell-out

variants of one and the same tense while conveying (73), we have to stipulate that

the Present Perfect is semantically ambiguous in order to account for the meaning

variations of (71). In other words, we have to assume that under the reading (74) the

Present Perfect is the spell-out of something else. One way to cope with this situ-

ation is to assume that the perfect morphology in (71) is the spell-out of a Priorean

(Prior 1957) temporal operator, a special tense head selecting for temporally non-

homogeneous predicates whose specifier position is filled by a past or a present time

variable as represented below:

ð75Þ Romance Present Perfect Romance Pluperfect

TP T'

PRESi T

VP<i,t>

PERFECT<it,it>

λPλt∃t' [t' < t & P(t')]: N-HOM(P)

TP T'

PASTi T

VP<i,t>

PERFECT<it,it>

λPλt∃t' [t' < t & P(t')]: N-HOM(P)

Given that in Sect. 3.3 we assumed that tenses cannot be decomposed in the logical

syntax, we will derive the correct scope order in which temporal adverbials are

always in the scope of the perfect.

According to this proposal, the LF associated with the Italian (8)(d). will be

therefore


123

ð76Þ

for two hours λPλt( t = 2 hours & ∀t'(t' ⊆t →P(t'))

PRESi

TP T'

T |

HOM<it,it>

VP<i,t>

λt(it to be dark(t))

PERFECT

λPλt∃t' [t' < t & P (t')]: P is temporally non-homogeneous

VP<i,t>

Adv<it,it>

|

analogously for the French (10)(d). According to this analysis, the difference

between the Present Perfect and the Pluperfect will depend on whether we find a

past or a present temporal pronoun in the Spec-TP position. The ungrammaticality

of the Italian (d)–(e) sentences in (9) and the French (d)–(e) sentences in (11)

follows from the non-homogeneity presupposition we associated with the Perfect in

(76). Since the Perfect presupposes its complement to be non-homogeneous, it

cannot combine with a temporal predicate obtained via a sinceD-adverbial; this non-

homogeneity presupposition is moreover responsible for the fact that eventualities

described under the Perfect get a perfective interpretation.

This analysis accounts for what has been called the ‘‘inclusive meaning’’

(Bertinetto 1986) of the Italian present perfect illustrated by the sentence below15:

(77) (a) Finora ho abitato a Torino

Until-now have-1sgPRES lived in Torino

(Bertinetto 1986)


(b) Jusqu’a maintenant j’ai habite a Turin

Until now I’have-1sgPRES lived in Torino

(FR version is mine)

[Passe Compose] FR

Until now I have lived in Turin

According to what the sentences in (77) say, there is a past interval, which ends now, at

which the speaker lives in Turin. I will assume this temporal meaning to be obtained as

a result of the interaction of the temporal information conveyed by the use of the Italian

adverbial finora and the French Jusqu’a maintenant (until now) occurring in the scope

of the perfect and of some pragmatic factors concerning state predicates such has tolive. Given the following definition of Romance untilNOW-adverbials

(78) untilNOW (finora) :¼ kPkt( P(t) & t abuts NOW)

the correct LF for (77) is

(79) 9t0 [t0 < PRESi & ( I-live-in-Turin(t0) & t0 abuts NOW)]

Notice that the adverbial defined in (78) makes the temporal predicate it modifies

non homogeneous. This is because, if the temporal property obtained via the

15 Given the analogies between French and Italian, this is true of the French ‘‘inclusive meaning’’ too.

200 F. Arosio

123

untilNOW modification is true of an interval, this interval should abut the speech

time; it is clearly untrue that every sub-interval of this interval abuts the speech

time. The temporal predicate I to live until-now in Turin is therefore non-

homogeneous and meets the non-homogeneity presuppositions associated with the

perfect. Because of its ‘‘non-homogeneizing’’ nature, these adverbs do not combine

with a Tempus Infectum, as shown below

(80) (a) ?? Finora abito a Torino [Presente] ITA

Until-now live-1sgPRES in Torino

(b) ?? Jusqu’a maintenant j’habite a Turin [Present] FR

Until now I’live-1sgPRES in Torino

–

(81) (a) ?? Fino ad allora abitavo a Torino [Imperfetto] ITA

Until to then live-1sgPASTimp in Torino

(b) ?? Jusque la j’habitait a Turin [Imparfait] FR

Until then I’live-1sgPASTimp in Torino

–

As in the case of for-adverbials, these adverbs do not maximalize the temporal

predicate they modify. The embedded temporal predicate, in our case I to live inTurin, is therefore free to hold an interval that is bigger than the one introduced by

the perfect; this means that the embedded temporal predicate can extend up to the

speech time. In fact, (77) can be truly uttered in a situation in which the speaker is

living in Turin at the speech time. The LF (79) is therefore compatible with both a

situation in which the speaker does live in Turin at the speech time and with one in

which it is not true that he lives in Turin at the speech time. This second reading

associated with (77) is more marginal. The explanation for this fact is again a

pragmatic one. To live in Turin is not a situation that you can change with a punctual

event. So, if you have been living in Turin just before now, I will infer that you still

live in Turin right now and that you will still live in Turin just after now. This is

what makes the second reading of (77) pragmatically inaccessible. It is clear that the

including now effect is a pragmatic effect depending on the verb meaning if we

consider the sentences below, uttered in the following situation

SITUATION: I went out for a beer with my friends and I left my children at

home with the baby-sitter. I came back and when I entered the room the

children started crying. The baby-sitter says:

(82) (a) Finora sono stati bravissimi [Passato Prossimo] ITA

Until-now be-3plPRES been very-quiet

(b) Jusqu’a maintenant ils ont ete tres sages [Passe Compose] FR

Until now they have-3plPRES been very quiet

Until-now they have been quietBe quiet is a state that one can change in a snapshot; according to my experience, a

child can suddenly start crying or start making noise. The reading according to


123

which the event described doesn’t hold at the speech time is pragmatically acces-

sible in this case. The sentences mean, in fact, that there is a past interval which

abuts now at which the children were quiet. Notice that the presence of the adverbial

is responsible for the including now interpretation of the sentences in (77). Consider

the sentences below

(83) (a) Ho abitato a Torino [Passato Prossimo] ITA

Have-1sgPRES lived at Torino

(b) J’ai habite a Turin [Passe Compose] FR

I’Have-1sgPRES lived at Torino

I have lived in Turin

As Bertinetto (1986) observes, when the adverb is not there the Italian sentence

means that there is a past time at which the speaker lived in Turin. This is explained

in our system: in order to meet the non-homogeneity presuppositions associated

with the perfect, we maximalize the temporal predicate. If the predicate is max-

imalised it cannot hold for an interval which is bigger than the interval introduced

by the perfect. Since we find a non-homogeneity presupposition associated with the

Italian Passato Remoto and French Passe Simple we expect the temporal property

obtained via the untilNOW-adverbial modification to combine with this tenses as

well. This is borne out by facts:

(84) (a) Giuro, fino ad allora fui una madre esemplare.


I swear, until to then was-1sgPASTperf a mother exemplar.

(b) Je jure, jusque la ce fut une mere exemplaire.

[Passe Simple] FR

I swear, until then I was-1sgPASTperf a mother exemplar.

I swear, until then I had been an ideal mother.

To repeat, in (77) the adverb requires the interval introduced by the perfect to abut

the speech time, and the including now interpretation is dependent on pragmatic

factors concerning the verb meaning. Given that to abut is a limit case of to proceed,

the meaning conveyed by (77) is analysed as a limit case of the one conveyed by (83)

(forced by the presence of the adverbial). This suggests that there is no extended-nowmeaning of the perfect in Italian (and French). Namely, the perfect does not introduce

an interval whose right boundary is the speech time (or a generic reference time).

This seems to be correct if we consider the fact that, in some Romance languages,

this meaning is conveyed by the use of the Tempora Infecta combined with the

sinceD-adverbials and not by the perfect as in English, as shown below

(85) I have lived in Turin for two years

The extended-now reading of (85), i.e. the reading for which there is a two-year-

long interval abutting now at which the speaker lives in Turin, is conveyed by the

202 F. Arosio

123

use of the present tense plus the durative da due anni (since 2 years) adverbial in

Italian and French

(86) (a) Vivo a Torino da due anni [Presente] ITA

Live-1sgPRES in Torino since two years

(b) Je vis a Turin depuis deux ans [Present] FR

I live-1sgPRES in Torino since two years

I have lived in Turin for two years

Going back to a more general discussion, it is however false that durative sinceD-

adverbials never combine with the Perfect. In fact they do when the predicate is a

change of state predicate

(87) (a) Il Parco Disney Studios ha aperto i cancelli da sei ore


The Park Disney Studios have-3sgPRES opened the gates since

six hours

(b) Le Disney parc a ouvert ses portes depuis 6 heures

[Passe Compose] FR

The Park Disney have-3sgPRES opened the gates since six hours

–

The Italian sentence (87)(a) and the French (87)(b) say that the Disney Park opened

its gates, and that the gates have been open for 6 hours (up to now); in (87)(a) and

(87)(b), the adverb modifies the state brought about by the completion of the

opening event and it says that this state holds for a six-hour-long interval abutting

the speech time.

In order to give a correct analysis of sentences like (87)(a) and (87)(b), I will argue

that the perfect morphology is ambiguous between the spell-out of a special Tempus

Perfectum, as defined in (75), and the spell-out of a result state construction, in some

Romance languages. The perfect morphology in (87)(a) and (87)(b) is the spell-out of

the latter; it introduces a state of result brought about by the completion of a telic

event which can be modified by temporal adverbials or saturated by tense.

It is clear that the perfect is ambiguous between these two meanings if we

consider the contrast between the two sentences below:

(88) (a) Alle tre, il Parco Disney Studios ha aperto i cancelli da sei ore


At-the three, the Park Disney Studios have-3sgPRES opened the gates

since six hours

(b) A trois heures le Parc Disney Studios a ouvert ses portes depuis six

heures

At three o’clock, the Park Disney Studios have-3sgPRES opened the

gates since six hours


123

(89) (a) Alle tre, il Parco Disney Studios ha aperto i cancelli per sei ore


gates for six hours [Passato Prossimo] ITA

(a) A trois heures le Parc Disney Studios a ouvert ses portes pendant six

heures


gates for six hours [Passe Compose] FR

At three o’clock, the Park Disney Studios opened the gates for six hours.

In (88) and (89), sinceD- and for-adverbials16 modify the result state brought about

by the completion of the opening event but, interestingly, in (89) the adverb alle tre(at three o’clock) localises the opening event, while in (88) it localises the right

bound of the result state (the being open of the gates). This is a clear indication of

the fact that the perfect conveys two different meanings in the two sentences: in (89)

it localises the at-three-o’clock opening event in the past with respect to speech

time, while in (88) it introduces the state of being open of the gates.

One might object that the contrast we have shown does not prove anything, since

these facts could be explained by assuming that the perfect morphology in (88) and

(89) is the spell-out of a temporal operator introducing the post state of the opening

event, namely the forever holding state of the opening event having culminated,

which Parsons (1990) calls resultant state, and that this post state is what is modified

by the sinceD-adverbial. This explanation cannot be correct if we consider the

unavailability of the continuation of (88) below:

(90) (a) ?? Il Parco Disney Studios ha aperto i cancelli da due ore ma la

polizia li ha chiusi un’ora fa. [Passato Prossimo] ITA

The Park Disney Studios have-3sgPRES opened the gates since

two hours, but the police them

have-3sgPRES closed them one hour ago

(b) ?? Le Parc Disney Studios a ouvert ses portes depuis deux heures mais

la police les a fermees il y a une heure. [Passe Compose] FR

The Park Disney Studios have-3sgPRES opened the gates since two

hours, but the police them

have-3sgPRES closed them one hour ago

–

The sentences in (90) show that the sinceD-adverbial does not modify the post state

of the opening event but rather its result state, since the gates should be open at

speech time according to what the sentences say.

16 See Pinon (1999) and von Stechow (2000) for a discussion about result state modification and for-

adverbials.

204 F. Arosio

123

It has been recently argued that telic predicates like to open the gates, have an

accessible result state that can be modified by a durative temporal adverbial (Kratzer

2000; Stechow 2002). Kratzer (2000) and von Stechow (2002) proposed two

alternative but similar definitions of accomplishment predicates sensitive to result

state modification. While in Kratzer’s analysis transformative predicates denote

relations between events and individual states having a certain property, in von

Stechow’s they denote relations between events and properties of states (in order to

account for the different meanings of adverbs modifying causative verbs (see also

von Stechow 2000). Both Kratzer and von Stechow assume then that special

operators called aktionsart choosers convert this relation into an event predicate or

into a state of result predicate. Given that in my system I have assumed states not to

be primitive entities but state predicates to denote properties of times, if we follow

Kratzer’s proposal we come up with the counterintuitive result that transformative

predicates denote a causative relation between events and times having a certain

property as defined below:

(91) to open :¼ kxkektkt’(CAUSE(e, t) & open(x)(t0)).

The only way of making sense of this in the system I am proposing is to assume

Lewis’s idea that times are event slices of a world; we shall avoid going into such a

discussion, not relevant for our aims. On the other hand, if we follow von Stechow’s

proposal and we reformulate his definition of transformative verbs in temporal

terms, we obtain the following more intuitive definition, which integrates more

naturally into the proposal I am making without touching any ontological question

about the nature of time

(92) to open :¼ kxkekP(CAUSE(e,P) & Become(e,P) & Agent(e,y) &

P ¼ open(x)),

- where P is a state, namely it denotes a property of times,

- and adapting Dowty’s (1979) definition for become:Become(e,P) « [P(RIGHT_BOUND(s(e))) & :P(LEFT_

BOUND(s(e)))].

Reviewing Kratzer’s proposal, von Stechow17 argues that the relation in (92) is

converted into an event predicate or into a result state predicate by the application of

the following ‘‘aktionsart choosers’’,

AKTIONSART CHOOSERS

(93) RESULT :¼ kRkt9e9P(R(e,P) & P(t))

(94) EVENT :¼ kRke9P(R(e,P)) where P is a property of times

According to (94), the eventive aktionsart chooser closes the state variable and gives

us a property of events, which can in turn enter the derivation of the LF of an

17 See von Stechow (2001, pp. 415–420.


123

eventive sentence. On the other hand, according to (93), the RESULT aktionsart

chooser closes the event variable and the state variable and gives us a property of

times which are in a complex relation with the event e. The state of result predicate

obtained by applying (93) to (92) is represented below:

(95) kt9e9P(CAUSE(e,P) & Become(e,P) & Agent(e,The-P-D-S)

& P ¼ open(the gates) & P(t)).

As we can see, the result operator is a stativizer and its application to a transfor-

mative verb gives a homogeneous temporal property that meets the homogeneity

condition associated with the Tempora Infecta. The correct LF for (87)(a) and

(87)(b) will be therefore

ð96Þ

VP

λeλP(CAUSE(e,P) & Become(e,P) & Agent(e,DPS) & P=open(the gates))

da 6 ore since 6 hours

λPλt∃I( t abuts I & P(I∪t) & δHRS(I)=6)

TP T'

T

PRESi

HOM

VP

λRλt∃e∃P(R(e,P) & P(t))

VP <et,it<i,t>>

Adv|

Result|

namely,

(97) kt9I 9e9P(CAUSE(e,P) & Become(e,P) & Agent(e,DPS) & P ¼ open(the

gates) & I abuts t & P(I ¨ t) & dHRS(I) ¼ 6).

According to what (97) says the 6-hour long interval introduced by the sinceD-

adverbial is understood as left abutting the temporal trace of the opening event and

right abutting the speech time. This does not follow from the application of the

sinceD-adverbial. I argue that this left abutting effect is due to pragmatic factors.

Since sinceD-adverbial sentences are upward entailing (for instance the Italian

Mario è malato da due ore (Mario is sick since two hours ‘‘Mario has been sick fortwo hours’’) entails Mario è malato da un’ora (Mario is sick since one hour ‘‘Mariohas been sick for one hour’’)), in order to be maximally informative, the length

indicated by the sinceD-adverbial should be the length of the maximal interval for

which we have evidence that the predicate is true. This explains the left abutting

effect of sinceD-adverbial modification of result state predicates. Under such an

analysis the Romance Perfect morphology will therefore be ambiguous being the

spell-out of a special tense selecting for temporally homogeneous predicates and the

spell-out of the result state construction as I argued in this section.

206 F. Arosio

123

6 Tense selection and habitual meanings

It is important here to account for some facts presented in Sect. 1 concerning

habitual meanings and tense selection in the system I am proposing.18 Let me

summarize them. In Romance languages, eventive-ongoing and habitual interpre-

tations are generally conveyed by the use of the same tense morphology, as shown

by the Italian and French sentences below

(98) (a) Gianni lavora in giardino [Presente] ITA

Gianni work-3sgPRES in garden

(b) Gianni travaille dans le jardin [Present] FR

Gianni work-3sgPRES in the garden

Ongoing reading: John is working in the gardenHabitual reading: John works in the garden

(99) (a) Gianni lavorava in giardino [Imperfetto] ITA

Gianni work-3sgPASTimp in garden

(b) Gianni travaillait dans le jardin [Imparfait] FR

Gianni work-3sgPASTimp in the garden

Ongoing reading: John was working in the gardenHabitual reading: John used to work in the garden

As we observed in Sect. 1, while a sentence with past imperfective verb morphology is

ambiguous between an ongoing and a habitual interpretation, a sentence with past

perfective verb morphology is generally not, as shown by the sentences below (again)

(100) (a) Gianni lavoro in giardino [Passato Remoto] ITA

Gianni work-3sgPASTperf in garden

(b) Gianni travailla dans le jardin [Passe Simple] FR

Gianni work-3sgPASTperf in the garden

John worked in the garden

According to Bonomi (1997)19 different aspectual verb forms are associated with

different structures of quantification in Italian and these are responsible for the

aspectual contrasts among (98), (99) and (100). According to Bonomi’s proposal,

imperfective morphology is the spell-out of an imperfective operator introducing

universal quantification over times which is responsible for both a progressive and a

habitual reading of an event sentence, and perfective morphology is the spell-out of

a perfective operator introducing existential quantification over times which is

18 The habitual interpretation is often seen as a subtype of the imperfective (Comrie 1976; Bertinetto

1986).19 See also Bonomi (1995) and Bonomi’s (1998). ‘‘Semantical remarks on the progressive readings of

the imperfective’’.


123

responsible for the perfective reading of an event sentence as formulated in the

following principle.

(101) IPF Unifying Principle (Bonomi 1997):

The progressive reading and the habitual reading of the imperfective

originate from the same logical form, based on universal quantification

over eventualities.

According to this proposal, the Italian ‘‘Imperfetto’’ morphology is the mark of

the presence of an IPF operator which introduces universal quantification over and

is responsible for both the progressive and the habitual readings (analogously for the

French Imparfait). However, habitual readings are always conveyed by morpho-

logical perfective sentences when a durative for adverbial measures the time span of

the habit in Italian and French; consider the facts we presented at the beginning of

this section represented below again

(102) (a) Leo prendeva il te‘ alle cinque [Imperfetto] ITA

Leo take-3sgPASTimp the tea at five

(b) Leo prenait son the a cinq heures [Imparfait] FR

Leo take-3sgPASTimp his tea at five o’clock

Leo used to drink tea at five

(103) (a) ?? Leo prendeva il te‘ alle cinque per venti anni

[Imperfetto] ITA

Leo take-3sgPASTimp the tea at five for twenty years

(b) ?? Leo prenait son the a cinq heures pendant vingt ans

[Imparfait] FR

Leo take-3sgPASTimp his tea at five o’clock for twenty years

(104) (a) Leo ha preso il te‘ alle cinque per venti anni


Leo have-3sgPRES taken the tea at five for twenty years

(b) Leo a pris son the a cinq heures pendant vingt ans

[Passe Compose] FR

Leo have-3sgPRES taken his tea at five o’clock for twenty years

Leo used to have tea at five for twenty years

(105) (a) Leo prese il te‘ alle cinque per venti anni


Leo take-3sgPASTperf the tea at five for twenty years

(b) Leo prit son the a cinq heures pendant vingt ans

[Passe Simple] FR

Leo take-3sgPASTperf his tea at five o’clock for twenty years

Leo used to have tea at five for twenty years

208 F. Arosio

123

As we can see from these contrasts, when a durative adverbial measures the time

span of the habit, habitual interpretations are conveyed by the use of the Passato

Remoto (Past Perfective) or the Passato Prossimo (Present Perfect) and the

Imperfetto is bad. The same holds for French, as observed above. These important

facts are have always been disregarded in the literature about habituality and aspect

in Italian since in the imperfect, morphology has always been treated as the spell-out

of a semantic operator responsible for the habitual reading. (see for example

Bertinetto 1986 and Lenci and Bertinetto 2000). The data in (102)–(105) show that

this view is not empirically correct.

It is however not true that for-adverbials never combine with the Tempora Infecta

(present and past imperfective). In fact they combine when the sentence gets a

habitual interpretation, as repeated below

(106) (a) Il venerdı Carlo correva nel parco per due ore

[Imperfetto þ for adverbial] ITA



[Imparfait þ for adverbial] FR


On Fridays Carlo used to run in the park for two hours

(107) (a) Il venerdı Carlo corre nel parco per due ore

[Presente þ for adverbial] ITA


(b) Le vendredi Carlo court dans le parc pendant deux heures

[Present þ for adverbial] FR


On Fridays Carlo runs in the park for two hours

These facts follow straightforwardly from an integration of Scheiner’s (2002)

convincing analysis of habitual sentences in the system I am proposing. According

to Scheiner, who develops an intuition found in von Stechow and Paslawska (2000),

habitual meanings are obtained by means of a covert extensional habitual operator

which stativizes its complement.20 Her definition of the habitual operator is the

following

(108) HAB :¼ kQkPkI9J[I ˝ J & (Q(P))(J)],

where I and J are intervals and Q a quantifier.

20 It is easy to see that this operator should be an extensional operator, contrary to what Lenci and

Bertinetto claim. For (106) to be true there should be a habit of Carlo’s which is based on actual Carlo’srunning in the park on Friday past events. If Carlo had never run in the park on Friday in the actual world,

(106) would be false. This distinguishes habitual sentences from generic or dispositional ones. Consider,

for instance, the dispositional sentence ‘‘John sells used cars’’. This sentence can be true even if John

never sold a used car in his life.


123

According to (108), HAB denotes a relation between a set of quantifiers, a set of

temporal properties and a set of times such that these times are included in the

interval including the many-quantifier defined times of which the temporal

property is true. An example of a quantifier found in Scheiner’s (2002) is given

below

(109) 9-many :¼ kPkt[ |t0 : t0 ˝ t & P(t0)| > C] where C is a context

dependent number of t0 for which P(t0).

Let us consider (106) as repeated below and see how Scheiner’s proposal works.

(110) (a) Il venerdı Carlo correva nel parco per due ore

[Imperfetto þ for adverbial] ITA



[Imparfait þ for adverbial] FR


On Fridays Carlo used to run in the park for two hours

The correct LF for (110), according to the temporal architecture I am assuming will

be

ð111Þ

VP

λe(Carlo-run-in-the-park(e))

λQλPλI∃J[I⊆J & (Q(P))(J)]

TP T'

T |

HOM

PASTi

λPλt∃e(τ(e)⊆t & P(e))

AspP

HAB|

PFV|

AdvAdv

on Friday λPλt∃I(P(t) & t⊆ I & I = Friday)

Q |

λPλt[ |t' : t' ⊆ t & P(t') | > C]

for two hours λPλt(δHOUR(t) = 2 & ∀t’(t’⊆t →P(t‘))

(112) 9t[PASTi ˝ t & [ |t0 : t0 ˝ t & 9I(9e(dHOUR(t0) ¼ 2 & "t00 (t00 ˝ t0 fit00 ˝ s(e) & Carlo-run in-the-park (e))& t0 ˝ I & I ¼ Friday) | > C]].

210 F. Arosio

123

(112) says that a contextually given past time is included in a time at which Carlo

has the habit of running for two hours on Friday. The LF for (107) will be anal-

ogous with the difference that the temporal pronoun in SpecTP is PRESi.

As you can see from (108), and (112), the habitual operator gives a homogeneous

temporal property. That this is intuitively correct is quite clear: if it is true that last year I

had the habit of running in the park on Fridays, it is true I had this habit in every month

of last year, in every week of last year, and so on. This explains why habitual readings

are generally conveyed by the use of a Tempus Infectum which requires its comple-

ment to be homogeneous. Thus, this also explains why for-adverbials combine with the

Tempora Infecta under a habitual reading: the habitual operator homogenizes the non

homogeneous temporal property obtained by for-adverbial modification.

As we saw in (104) and (105), when a durative adverbial measures the time span

at which the habit holds in the past, habitual interpretations in Italian are conveyed

by the use of the Passato Remoto (Past Perfective) or the Passato Prossimo (Present

Perfect) and the Imperfetto is bad. In the case of French they are conveyed by the

use of the Passe Simple (Past Perfective) or the Passe Compose (Present Perfect)

and the Imparfait is bad. These facts follow straightforwardly from our analysis.

According to what these sentences say, the adverbial does not measure the time of

each Leo’s having tea at five event, this is pragmatically ruled out by the fact that

tea drinking events usually do not take 20 years. It rather measures the whole

interval at which Leo had the habit of having tea at five. In order to obtain this

reading, the durative adverbial has scope over the habitual operator. In this case, the

temporal property obtained, i.e. the property of being a twenty-year long interval at

which Leo has the habit of having tea at five, is quantized and therefore it cannot

combine with a Tempus Infectum, but it does with a Tempus Perfectum.

Clearly, if the interval at which the habit holds is modified by sinceD-adverbials

the temporal predicate obtained combines with a Tempus Infectum but not a

Tempus Perfectum as shown by the sentences below

(113) (a) Leo prendeva il te‘alle cinque da venti anni

[Imperfetto þ sinceD adverbial] ITA

Leo take-3sgPASTimp the tea at five since twenty years

(b) Leo prenait son the a cinq heures depuis vingt ans

[Imparfait þ sinceD adverbial] FR

Leo take-3sgPASTimp his tea at five o’clock since twenty years

Leo had been having tea at five for twenty years

(114) (a) ?? Leo prese il te‘alle cinque da venti anni

[Passato Remoto þ sinceD adverbial] ITA

Leo take-3sgPASTperf the tea at five since twenty years

(b) ?? Leo prit son the a cinq heures depuis vingt ans

[Passe Simple þ sinceD adverbial] FR

Leo take-3sgPASTperf his tea at five o’clock since twenty years

–


123

Analogously to the Italian Passato Remoto and the French Passe Simple, the Italian

Passato Prossimo and the French Passe Compose do not combine with sinceDadverbials under habitual readings for the same reason.

To sum up, it is wrong to assume that Italian Imperfetto and French Imparfait are

the spell out of a habitual operator. They are on the contrary the spell out of a tense

selecting for a homogeneous complement. The property of being a time at which a

habit holds is homogeneous, if not modified by a quantizing adverbial; therefore it

combines with a Tempus Infectum but not with a Perfectum. When the habit is

modified by a quantizing adverbial such as for twenty years, the pattern of

acceptability changes.

These observations can be extended to generic sentences. In the literature about

Italian (see for instance Lenci and Bertinetto 2000) it is claimed that generic

meanings are always conveyed by the use of the Presente, or the Imperfetto. Once

again, this is not correct if we consider sentences like

(115) (a) Il Neanderthale e stato vegetariano per 30.000 anni. Poi e diventato

onnivoro

The Neanderthaler be-3sgPRES been vegetarian for 30.000 years. Then

he became omnivore. [Passato Prossimo þ for adverbial] ITA

(b) Neanderthal a ete vegetarien pendant 30000 ans., puis il est devenu

omnivore

The Neanderthaler have-3sgPRES been vegetarian for 30.000 years. Then

he became omnivore. [Passe Compose þ for adverbial] FR

The Neanderthaler was a vegetarian for 30,000 years. Then he became anomnivore.

The reading here is clearly a generic one but the tense morphology is perfective.

According to the proposal I am making, the tense morphology in (115) is dependent

on the occurrence of the for-adverbial which quantizes the generic temporal

proposition to be true of a 30,000 year long interval. More generally, tense selection

in generic sentences patterns tense selection in habitual sentences and the durative

adverbial distribution we discussed above.

7 Conclusions

In this paper I have shown that temporal homogeneity plays a fundamental role in the

selection of tense in Romance Languages. We discussed the case of French and

Italian. I proposed a tense decomposition in which tense is sensitive to the temporal

homogeneity of its complement. This accounts for the fact that temporal adverbials,

which homogenize the temporal property they modify, combine with morphological

imperfective tenses, which I called Tempora Infecta. Temporal adverbials, which

quantize the temporal property they modify, combine with the morphological per-

fective tenses, which I called Tempora Perfecta. We have seen that this decompo-

sition explains the fact that perfective readings of event predicates are usually

212 F. Arosio

123

conveyed by the use of a Tempus Perfectum, while imperfective readings, by the use

of a Tempus Infectum. This happens usually, as I said, but not always. In fact, when a

temporal adverbial intervenes above the aspectual projection, the pattern of com-

bination changes. This follows from the fact that, in my proposal, tense morphology

is not the spell-out of the combination of a tense and an aspectual operator but rather

the spell-out of semantic tenses which have some influence in the aspectual inter-

pretation of a sentence. Moreover, we have seen that habitual interpretations are

conveyed by the use of a Tempus Infectum or of a Tempus Perfectum depending on

whether the habit is or is not quantized by a durative temporal adverbial. I have

argued that this also accounts for tense selection in generic sentences. I have pro-

posed an analysis of the Perfect in Romance languages which assumes that perfect

constructions are ambiguous between a past/Priorean reading and a result state

depending on the class of the predicate occurring in the VP. We have seen how

temporal homogeneity plays an important role in Perfect constructions.

References

Abusch, D. (1994). Sequence of tense revisited: Two semantic accounts of tense in intensional contexts.

In H. Kamp (Ed.), Ellipsis, tense and questions. Stuttgart, Dyana-2 Esprit Basic Research Project

6852. Amsterdam: University of Amsterdam.Abusch, D. (1997). Sequence of tense and temporal de re. Linguistics and Philosophy, 20(1), 1–50.Bennett, M., & Partee, B. (1972). Toward the logic of tense and aspect in English. Santa Monica, CA:

System Development Corporation.Bertinetto, P. M. (1986). Tempo, aspetto e azione nel verbo italiano: Il sistema dell’indicativo. Firenze:

Presso l’Accademia della Crusca.Bertinetto, P. M. (1991). Il verbo. In L. Renzi & G. Salvi (Eds.), Grande grammatica italiana di

consultazione. (Vol. II, pp. 13–162). Bologna: Il mulino.Bertinetto, P. M. (2001). On a frequent misunderstanding in the temporal-aspectual domain: the «per-

fective-telic» confusion. In C. Cecchetto, T. Guasti, & G. Chierchia (Eds.), Semantic interfaces.

(pp. 177–210). Stanford, CSLI.Bonomi, A. (1995). Aspect and quantification. In P. M. Bertinetto, V. Bianchi, J. Higginbotham, &

M. Squartini (Eds.), Temporal reference, aspect, and actionality. (Vol. I, pp. 309–324). Torino,

Rosenberg and Sellier.Bonomi, A. (1997). Aspect, quantification and when-clauses in Italian. Linguistics and Philosophy, 20,

469–541.Bonomi, A. (1998). Semantical remarks on the progressive readings of the imperfective. Paper presented

at the conference On Syntax and semantics of tense and mood selection, Bergamo, 2–4 July.Comrie, B. (1976). Aspect: An introduction to the study of verbal aspect and related problems.

Cambridge: Cambridge University Press.Davidson, D. (1967). The logical form of action sentences. In N. Rescher, (Ed.), The logic of Decision

and action. (pp. 81–95). Pittsburgh: University of Pittsburgh Press. (Reprinted by D. Davidson, Ed.,

(1990, Essays on actions and events, pp. 105–122). Oxford: Clarendon Press.de Swart, H. (1998). Aspect, shift and coercion. Natural Language and Linguistic Theory, 16, 347–385.Dowty, D. (1979). Word meaning and montague grammar. Dordrecht: D. Reidel.Galton, A. (1984). The logic of aspect: An axiomatic approach. Oxford: Clarendon Press.Heim, I. (1994). Comments on Abusch’s theory of tense. Ms, MIT: Cambridge, MA.Heim, I., & Kratzer, A. (1998). Semantics in generative grammar. Oxford: Blackwell.Herweg, M. (1991). Perfective and imperfective aspect and the theory of events and states. Linguistics,

29, 969–1010.Higginbotham, J. (1985). On semantics. Linguistic Inquiry, 16, 547–593.Hinrichs, E. (1985). A compositional semantics for Aktionsarten and NP reference in English. PhD

dissertation, Ohio State University.Hornstein, N. (1990). As time goes by: Tense and universal grammar. Cambridge, MA: MIT Press.


123

Katz, G. (1995). Stativity, genericity, and temporal reference. PhD Dissertation, University of Rochester.Katz, G. (2000). Anti neo-Davidsonianism: against a Davidsonian semantics for state sentences. In

C. Tenny & J. Pustejovsky, (Eds.), Events as grammatical objects. Stanford, CA: CSLI Publications.Klein, W. (1994). Time in language. London and New York: Routledge.Kratzer, A. (1998). More structural analogies between pronouns and tenses. In D. Strolovitch &

A. Lawson (Eds.), SALT 8, Cambridge, MA, Ithaca, CLC-Publications.Kratzer, A. (2000). Building statives. Berkeley: University of Massachusetts at Amherst, Berkeley Lin-

guistic Society.Krifka, M. (1986). Nominalreferenz und Zeitkonstitution. Zur Semantik von Massentermen, Indivi-

dualtermen, Aspektklassen. PhD dissertation, The University of Munich.Krifka, M. (1990). Four thousand ships passed through the lock: Object-induced measure functions on

events. Linguistics and Philosophy, 13, 487–520.Krifka, M. (1992). Thematic relations as links between nominal reference and temporal constitution. In

I. Sag & A. Szabolcsi, (Eds.), Lexical matters (pp. 29–53). Chicago: CSLI Publications, Chicago

University Press.Lenci, A., & Bertinetto, P. M. (2000). Aspects, adverbs and events: Habituality vs. perfectivity. In

A. F. P. Varzi, (Ed.), Facts and events. Oxford: Oxford University Press.Lobner, S. (1988). Ansätze zu einer integralen semantischen Theorie von Tempus, Aspekt und Aktion-

sarten. In V. Ehrich & H. Vater (Eds.), Temporalsemantik. Beiträge zur Linguistik der Zeitreferenz.

(pp. 163-191). Tubingen: Niemeyer.McCoard, R. W. (1978). The English perfect: Tense choice and pragmatic inferences. Amsterdam: North-

Holland.Musan, R. (2000). The semantics of perfect constructions and temporal adverbials in German. Berlin:

Humboldt Universitat.Musan, R. (2003). Seit-adverbials and perfect constructions. In A. Alexiadou, M. Rathert & A. von

Stechow (Eds.), Perfect Explorations. (pp. 253–276). Berlin: Mouton de Gruyter.Oldsjo, F. (2001). Tense and aspect in Caesar’s narrative. Uppsala: Uppsala Universitet.Parsons, T. (1990). Events in the semantics of English. A study in subatomic semantics. Cambridge, MA:

MIT Press.Pinon, C. (1999). Durative adverbials for result states. In 18th West coast conference on Formal lin-

guistics. Cascadilla Press, Somerville.Prior, A. (1957). Time and modality. Oxford: Claredon Press.Scheiner, M. (2002). Temporale Verankerung habitualler Sätze. MA Thesis, Universitat Wien. Wien.Thomason, R. H., & Stalnaker, R. (1973). A semantic theory of adverbs. Linguistic Inquiry, 4, 195–220.von Stechow, A. (2000). How are results represented and modified? Remarks on Jager & Blutner’s anti-

decomposition. In C. Fabricius-Hansen, E. Lang, & C. Maienborn (Eds.), Adverbs: Proceedings ofthe Oslo conference on Adverbs, 26–28. October 1999.

von Stechow, A. (2002). German seit ‘‘since’’ and the ambiguity of the German Perfect’’. In K. Ingrid &

S. Barbara, (Eds.), More than words: A Festschrift for Dieter Wunderlich (vol. 53, pp. 393–432).

Berlin: Akademie Verlag.von Stechow, A., & Paslawska, A. (2000). Deriving perfect readings in Russian/Ukainian. Ms, Thessa-

loniki.

214 F. Arosio

123

infectum and perfectum. two faces of tense selection in ... · pdf fileinfectum and perfectum....

Documents