Portuguese Expressions of Duration and its English Counterparts

This paper focuses on the expression of duration in (European and Brazilian) Portuguese. English is used for comparative purposes, since many of the issues to be discussed have been intently studied for this language. The paper has two main, parallel, goals: (i) a description of the Portuguese subsystem of duration, with an identification of the main structures involved in the expression of this semantic value, and (ii) an attempt to distinguish several conceptually different – though closely related – temporal subdomains within the semantic area under study. This second goal has obviously implications for the grammatical description that go well beyond the analysis of Portuguese. Two distinctions will be explored in some detail: one between strict duration and durative temporal location, which is not assumed by all authors, and another between the duration of telic and atelic eventualities. Other (semantic and/or syntactic) subtypes of strict duration that will be discussed are: argumental vs. adjunct, continuous vs. discontinuous, time-anchored vs. non-time-anchored, and planned vs. non-planned duration.


Different concepts of duration -taxonomic issues
Since the extent of the concept of duration is not unanimous in the literature, the taxonomic issue must be addressed prior to identifying expressions of duration.Consider the following English sentences: (1) a. John was in the office for five minutes.
b. John solved the puzzle in five minutes.
(2) a. John has been in the office for the last two hours.b.John was in the office from 3 to 5.
If we assume a wide concept of duration, all the time adjuncts (italicised) in these sentences are durational, since they convey information about the temporal size of the described eventualities: five minutes, in the first two examples, two hours in the last two.This broad notion of duration phrases is widely assumed (cf.e.g.Huddleston and Pullum 2002: 702ff., just to give an example of a recent English grammar).However, at least two other more restricted views on the category of duration phrases have been adopted in the literature: (i) one that includes the type of time adjuncts in (1a) and ( 2), which combine with atelic eventualities, but excludes the type of time adjuncts in (1b), which combine telic events -cf.e.g.Smith (1991); (ii) one that includes the type of time adjuncts in (1), which directly assert the duration of eventualities, without locating them in the time axis, but excludes the type of time adjuncts in (2), which can at the same type express the duration of an eventuality and position it in the time axis -cf.e.g.Kamp and Reyle (1993) 1 .
The first view is motivated by the observation that substantial linguistic differences (which will be considered in more detail in section 3) set apart the uses of telic and atelic descriptions in the relevant temporal constructions.However, it obscures the fact that important semantic and syntactic properties bring together the time adjuncts in (1a) and (1b), for which reason it will not be adopted here.
As for the second view, it seems linguistically more justifiable.In previous work (cf.Móia 1998Móia , 2000)), I have favoured it, showing that a categorisation that places the time adjuncts in (1) in a different class from those in (2) has greater generalisation power and can better explain several distributional facts about comparable phrases in different languages2 .To be more precise, the categorisation I defend can be summarised as follows: (i) time adjuncts like for five minutes or in five minutes, which directly assert the duration of eventualities, without locating them in the time axis, belong in a class that we can term (true, or strict) duration adverbials; formally, they are set apart by the fact that they contain bare predicates of amounts of time as complements of the temporal preposition; (ii) time adjuncts like for the last two hours or from 3 to 5, which can at the same time express the duration of an eventuality and position it in the time axis, are (members of a subclass of) temporal locating adverbials; formally, they are set apart by the fact that they contain time -denoting expressions -rather than bare predicates of amounts of time -as complements of the heading temporal connective.
As said, this view is in line with e.g.Kamp and Reyle (1993), where "temporal measurement" (of eventualities) -i.e.duration -is taken to involve the determination of the amount of time a situation lasts (or takes to culminate), irrespective of its position on the time axis.In that work, duration is formally represented by a one-place functor dur, which maps eventualities (or intervals, which are not under consideration here) on the amo unts of time they last -cf.Kamp and Reyle (1993: 648).
John has lived in Paris for two years.

Table 1. Duration
As for temporal location, still according to Kamp and Reyle (1993), it involves the association of eventualities (described in matrix structures) with a given interval of the time axis, called "location time".Obviously, the location time may be defined in simpler or more complex ways, depending on the structure of the time adverbial -e.g. by time -denoting phrases that do not include predicates of amounts of time, like 1995, next Summer, the day when the Berlin Wall fell, or the period before the war, by time -denoting phrases that include these predicates in embedded positions, like the last five minutes, or those two hours, or even by situation-denoting nominal or sentential complements, like (during) the Second World War or (while) Mary cooked the dinner.The location relation may take different forms, among which at least the following ones have to be distinguished (cf.Móia 2000):

DRS-conditions associated Examples
[t ⊆ ev] durative location [t = loc (ev)] (if exact location is involved) John played the piano from 3 to 5. John has been playing the piano since 2 o'clock.John has lived in Paris for the last two years.

mere overlapping location
[ev ¡ t] John was ill on Sunday.John was in Paris last weekend.
inclusive location [ev ⊆ t] John got married in 1995.

Table 2. Temporal Location
Now, as we can see, the temporal phrases that are classified here as durative temporal locating adverbials are at the center of the mentioned taxonomic disputes in the literature.Bennett and Partee (1978: 30), for instance, considered phrases like these to be amb ivalent, forming a sort of mixed class of temporal locating and duration adverbials (which they named "frame adverbials" and "durative adverbials", respectively).Smith (1991) or Vlach (1993), on the other hand, consider a broad class of "durative adverbials ", which groups together these ambivalent phrases and those that simply express the duration of atelic situations, without locating them.Kamp and Reyle (1993), as said, separate out these phrases -as temporal locating adverbials associated with the condition [t ⊆ ev] -from those that express strict duration -associated with the condition [dur (ev) = mt] 3 .However, these authors seem to consider phrases like for the last two years -which contain a predicate of amounts of time -as exceptional.In particular, they find it hard to classify them as temporal locating or as duration adverbials, and refer to them as "unresolvably ambivalent" (cf.p. 650).In Móia (1998Móia ( , 2000)), however, I showed (i) that they can be treated as common durative temporal locating adverbials (just like phrases that do not contain predicates of amounts of time, such as from 3 to 5), and, furthermore, (ii) that they exhibit the linguistic properties of locating, not of duration adverbials.In that work, I hypothesise that the two categories of temporal phrases -duration and temporal locatingare essentially told apart by the form of complement of the temporal connectives (bare predicates of amounts of time, and time -denoting phrases, respectively), and that inferentially extracted information about the duration of the located eventuality is what makes some of these locating adverbials look like 3 Some minor differences are being ignored here: e.g.Kamp and Reyle (1993) use the term "temporal measure adverbials" instead of "duration adverbials"; they always introduce an inexact duration condition, [dur (ev) ≤ mt]; and they only consider stative discourse referents (s) in the durative location condition (i.e.[t ⊆ s]), but, as can be easily observed, activities can also occur in these constructions.
ambivalent phrases, whereas, as far as assertion is concerned, they are merely locating expressions.One argument in favour of this view concerns the effect of negation with sentences exhibiting the so-called ambivalent operators, which shows that, given a true negative sentence, the falsity of a corresponding positive necessarily affects the temporal location predication, but not the duration, which can still apply4 .Observe that − if the utterance time is at the beginning of 2006 − the first, but not the second, of the following two sequences is a contradiction: (3) #Ana hasn't lived in Amsterdam for the last two years, but she has lived in Amsterdam from the beginning of 2004 until the end of 2005.
[with utterance time at the beginning of 2006] (4) Ana hasn't lived in Amsterdam for the last two years, but she has lived in Amsterdam for two years.

Asserted vs. inferred duration
If we assume the notion of temporal locating phrase described above, it is clear that the so-called ambivalent adverbials can be analysed as simple locating phrases, the durational information being easily derived from simple (and systematic) inferential patterns.

Temporal location and inferred duration
Observe the following illustrative examples of inferences about duration drawn from (exact) durative locations: (5) John was in the office from 3 to 5. → John was in the office for two hours.
(6) John has been playing the piano for the last two hours.→ John has been playing the piano for two hours.
It must be stressed that inferences about duration may also result from the use of inclusive locating adverbials, although these phrases have not been traditionally associated with duration (or classified as durational), even in texts that take the wide concept of duration alluded to at the beginning of section 1 (e.g.Huddleston and Pullum 2002).Observe the following illustrative examples of inferences about duration drawn from inclusive locations: (7) John solved the puzzle during the (five-minute) break.
→ John solved the puzzle in at most five minutes.
(8) John wrote this novel last August.→ John wrote this novel in at most a month.
Contrary to the examples above, locating phrases that are compatible with both inclusive and durative scenarios (when combined with descriptions of atelic eventualities), do not license any inferences about the duration of those eventualities.Observe that no inference about duration can be drawn from the following sentences that express mere overlapping location: (9) John was ill on Sunday.
In sum, we may say that the issue of asserted vs. inferred duration of eventualities is particularly relevant as regards a clear categorisation of time adjuncts.In general, we may say that inferred duration -as opposed to asserted duration -is pervasive in natural language discourse, since it is regularly associated with temporal locating adjuncts -as expressed in Table 3 -and still, quite often, with discourse structure (in association with the lexical content of propositions) -as in the following example: (11) John was in the office today.He arrived at 5 and left at 7. → John was in the office for two hours.
In this paper, I will essentially deal with strict (i.e.asserted) duration.Within this domain, special attention will be given to the differences between the duration of telic and atelic eventualities.However, before moving to this topic, in section 3, I will briefly consider, in the next section, a major distinction in the domain of duration, which has to do with the grammatical subsystems involved in its representation.

Argumental vs. adjunct duration
In Portuguese, just like in English, the (asserted) duration of eventualities can be expressed with essentially two grammatical means: predicate-argument combinations and time adjuncts.For easier reference, the duration conveyed in these two ways will be referred to as argumental duration and adjunct duration, respectively.
Examples of predicate-argument combinations leading to argumental duration involve predicates like durar and levar, in Portuguese, and last and take, in English.
the meeting general of shareholders lasted three hours 'The general meeting of shareholders lasted three hours.' (13) Os arquitectos levaram três horas a estudar este projecto.the architects took three hours to study this project 'It took the architects three hours to study this project.' The subclass of predicates that occur in these constructions -which can be termed duration predicates -take (at the minimu m) an eventuality-denoting phrase and a predicate of amounts of time as arguments 5 .The predicate itself expresses a duration relation, comparable to the predicate dur of Kamp and Reyle (1993).Adjunct duration is expressed via temporal phrases applied adverbially or adnominally, as in the following two examples, respectively: (14) Os arquitectos analisaram o projecto durante três horas.
the architects analysed the project for three hours 'The architects analysed the project for three hours.' (15) a construção da ponte em apenas um ano the construction of-the bridge in just one year 'the construction of the bridge in just one year' 5 Interestingly, the argument position where predicates of amounts of time occur can often be filled by time-denoting phrases.In these constructions, duration is also obtained by inferential processes: (i) A reunião geral de accionistas durou o Sábado inteiro.the meeting general of shareholders lasted the Saturday whole 'The general meeting of shareholders lasted the whole of Saturday.'(ii) Os arquitectos levaram todo o mês de Março a estudar este projecto.the architects took all the month of March to study this project 'It took the architects the whole month of March to study this project.' As we will see later on, the dividing line between argumental and adjunct duration is somewhat hard to draw in Portuguese, since there are temporal phrases that exhibit adjunct properties but involve a verbal predicate, namely duration haver-phrases (analysed in section 7.3).

Telic vs. atelic duration
As said at the beginning of this text, all the phrases that express the temporal size of an eventuality by directly associating an amount of time to it (via a function like dur) will be grouped together here, in the class of duration adjuncts.Hence, this class includes both the phrases that express the duration of atelic events and those that express the duration of telic events, as in the examples (1a) and (1b), repeated below, respectively: (16) a. John was in the office for five minutes.
b. John solved the puzzle in five minutes.
Although many authors adopt this categorisation (e.g.Kamp andReyle 1993, or Huddleston andPullum 2002), some others place the adverbials in (16a) and (16b) in two independent grammatical categories.For instance, Smith (1991) includes phrases of the kind in five minutes in (16b) in a class she terms "completive adverbials", distinct from that of "durative adverbials", where she includes for five minutes.Her class of completive adjuncts appears to be quite heterogeneous, though.On the one hand, it includes phrases that express the duration of telic events, like (16b), or ( 17) below (which is similar to (16b), but involves a shift from an achievement to an accomplishment by addition of a preparatory phase): (17) John reached the top of the mountain in five minutes.
On the other hand, it also includes homonymous phrases that locate events in the temporal axis, namely in the interval that follows a temporal perspective point by a given amount of time: (18) John will knock on the door in five minutes.
This sentence illustrates what the author calls the "ingressive interpretation": "Adverbials of the interval, completive type are ingressive when they locate instantaneous events.As ingressives, the adverbials indicate an interval at the end of which the event occurs."(p.157).It is clear that sentences like (18) do not express any form of duration, but only positioning on the time axis.Thus in five minutes in (18) can be classified as a simple temporal locating phrase, according to the definitions assumed in this paper, contrary to its homonyms in (16b) and ( 17), where duration, or temporal extent, is involved.As we will see in section 6, the ambiguity between durational and locational in-phrases (cf.Huddleston and Pullum 2002: 710) also exists in Brazilian Portuguese, for the parallel em-phrases, but not in European Portuguese, which systematically resorts to different connectives to express the two readings (em, for duration, and e.g.dentro de or daqui a, for location).Even if one does not follow authors like Smith (1991) in placing the temporal phrases in (16a) and (16b) in completely autonomous categories, it must be recognised that there are striking linguistic differences between time adjuncts that express the duration of atelic -or homogeneous -eventualities (i.e.states and activities) and those that express the duration of telic events (i.e.accomplishments).In fact, even intuitively, two different operations seem to be at stake: if atelic eventualities are involved, to express their duration is to say for how long they lasted; if telic events are involved, to express their duration is to say the amount of time it took for them to be completed, i.e. to reach a culmination (of course, this amount of time coincides with the duration of the preparatory phase of the event, since the culmination is conceived of as punctual).Huddleston and Pullum (2002), for instance, acknowledging the importance of these differences, consider two major subclasses of duration phrases: "bounding" and "non-bounding" (which apply to atelic and telic eventualities, respectively).In this text, I will resort to a hypallage, and use the terms atelic duration and telic duration to refer to this opposition.
Expressions of atelic duration -whether time adjuncts or predicate--argument combinations -are normally different from expressions of telic duration.I will demonstrate this statement with three different groups of expressions, from Portuguese and English, two involving time adjuncts of duration (A, B), and one involving duration predicates (C).To my knowledge, not all the differences described below have been (clearly) stressed in the literature.

A. Time adjuncts headed by different temporal connectives
One the most striking (and long-debated) differences between atelic and telic duration is that they are expressed in many languages -Portuguese and English included -by phrases headed by different temporal connectives.This property is the basis for a classical test to distinguish between atelic and telic eventualities (cf.e.g.Dowty 1979).In (European and Brazilian) Portuguese, atelic duration is expressed by phrases headed by the preposition durante; furthermore, Brazilian Portuguese uses, very frequently, phrases headed by the preposition por as exact synonyms6 .In English, atelic duration is typically expressed by for-phrases.
Xian was capital of-the China for more of thousand years 'Xian was the capital of China for more than one thousand years.'Furthermore, it is possible -both in Portuguese and English -to express atelic duration via bare predicates of amounts of time, i.e. phrases with no overt preposition (though the underlying preposition can normally be spelled out): (20) A Ana tinha dormido (durante) doze horas.the Ana had slept (for) twelve hours 'Ana had slept (for) twelve hours.' As for telic duration, it is expressed by time adjuncts headed by the preposition em in Portuguese, and the preposition in in English.As seen above (cf.( 16b) and ( 17)), the accomplishments whose duration is identified may be either basic or derived by Aktionsart shift.
(21) A Ana escreveu a tese em seis meses.the Ana wrote the thesis in six months 'Ana wrote her thesis in six months.' Neither in Portuguese nor in English can bare predicates of amounts of time express telic duration.Therefore, if a description of an accomplishment is combined with a bare predicate of amounts of time, the only interpretation available is one involving an Aktionsart shift (e.g. from accomplishment to activity by stripping off the culmination).Apparently, this interpretation seems easier to obtain in Portuguese than in English in examples like the following: (22) A Ana leu este livro seis meses (e nunca o acabou).
the Ana read this book six months (and never it finished) ' ?? Ana read this book six months (and never finished it).'

B. Different durational adverbs and adjectives
The duration of eventualities may be expressed by (vague) adverbials phrases, if they are described by sentential means, or by (vague) adjectival phrases, if they are described by nominal phrases: (23) Os operários reconstruíram a ponte muito rapidamente.the workers rebuild the bridge very quickly 'The workers rebuilt the bridge very quickly.' (24) a rápida reconstrução da ponte the quick reconstruction of-the bridge 'the quick reconstruction of the bridge' (25) O rei esteve brevemente exilado num país vizinho.the king was brifly exiled in-a country neighbouring 'The king was briefly exiled in a neighbouring country.' (26) o curto exílio do rei the short exile of-the king 'the short exile of the king' The adverbs and adjectives in these phrases seem to be sensitive to the (a)telicity of the described eventualities.In order to express the duration of telic events, Portuguese resorts typically to adjectives like rápido ('quick') and lento or demorado ('slow'), and to their derived adverbs (rapidamente / lentamente / demoradamente), as exemplified in ( 23) and ( 24); in order to express the duration of atelic situations, Portuguese resorts typically to adjectives like breve or curto ('brief', 'short') and longo or prolongado ('long'), and to some of their derived adverbs (brevemente / longamente / prolongadamente), as illustrated in ( 25) and ( 26) 7 .Note that if the duration phrases in ( 23)-( 24) are substituted by the corresponding ones in ( 25)-( 26), or vice-versa, ungrammaticality arises.These are, therefore, typical examples that illustrate the selectional restrictions at stake (which does not necessarily mean that every (a)telic expression behaves with the same regularity8 ).

C. Different duration predicates
In Portuguese, as well as in English, different aspectual complements are often selected by different duration predicates.If we consider atelic situations expressed by nominal phrases (not by sentences), the typical duration predicate used in Portuguese is durar, a close counterpart of the English verb last9 .
(27) O trabalho de pesquisa sobre esta doença durou um ano. the work of research on this disease lasted one year 'The research work on this disease lasted one year.' Note that, although these verbs do not take directly sentences as arguments, it is marginally possible to combine them with atelic sentences, if they occur embedded, as the complement of n ouns like situação ('situation').The resulting sentences may sound unnatural to some extent, but not utterly ungrammatical: (28) The selectional properties of duration predicates with respect to telic descriptions have been comparatively more emphasised in the literature: English resorts typically to the verb take (cf.Dowty 1979); Portuguese uses at least two verbs with similar restrictions -levar and demorar 11 ; all these verbs can take nominal or sentential eventuality-denoting arguments. 10The Portuguese verb passar often occurs with time-denoting phrases instead of predicates of amounts of time as arguments (whence expressing duration by inference rather than assertion) -cf.fn. 5. 11 Another verb with similar behaviour, but possibly not as common, is gastar.It combines with telic descriptions, and is felt as a bit awkward with atelic descriptions: (i) Gastei duas horas {a / para} resolver o quebra-cabeças.
[I] spent two hours to solve the puzzle (ii) ??
the construction of-the bridge took one year 'The construction of the bridge took one year.' (33) Eles {levaram / demoraram} um ano para construir a ponte.they took one year to build the bridge 'It took them one year to build the bridge.' The preposition that heads the sentential complement of these two Portuguese verbs is typically para, although a can also be used: (34) Eles {levaram / demoraram} um ano a construir a ponte.
In accordance with the distinctions noted above, expressions that are ambiguous between an atelic and a telic reading are unambiguous when combined with the predicates at stake.Observe the following example: (35) A ocupação da ilha pelas tropas {durou / demorou} um ano.
the occupation of-the island by-the troops {lasted / took} one year 'The occupation of the island by the troops {lasted / took } one year.' The NP a ocupação da ilha pelas tropas ('the occupation of the island by the troops') is ambiguous between an atelic reading -the state of the troops being in the island -and a telic reading -the accomplishment of the troops completing the occupation of the island.With durar ('last') only the first interpretation is available; with demorar ('take') -just like with levar -it is the other way round 12 .
12 The combination with duration predicates like durar or demorar may be specially interesting as a possible test for (a)telicity of nominal expressions: (i) O sismo {durou / ?? demorou / * levou} vinte segundos.the earthquake {lasted / took / took} twenty seconds 'The earthquake {lasted / *took} twenty seconds.'In fact, eventuality-denoting nominal phrases are sometimes hard to classify as atelic or telic.This is particularly the case with those that -unlike a presença de tropas na ilha ('the presence of troops in the island') or a construção da ponte ('the construction of the bridge') -do not combine with duration adjuncts headed by durante or em, but rather with duration adjuncts headed by de ('of'): (ii) um sismo de vinte segundos an earthquake of twenty seconds 'a twenty-second earthquake' However, it must be noted that some nounslike e.g.viagem ('trip')seem to accept the combination with all types of verbs without a sharp difference in Up to now we have observed several differences in the ways of expressing telic and atelic duration.To these differences between telic and atelic duration, we can add some more, which have to do with specific properties of homogeneous (i.e.atelic) eventualities, viz.: (i) they are additive, i.e. can be summed up to yield eventualities of the same type (cf.e.g.Bach 1981Bach , 1986) -e.g.two states of Ana being in the office for x and y time can be summed up to yield a state of Ana being in the office that lasts for x+y time; (ii) the same continuous eventuality has different durations at different time intervals -e.g. if the state of Ana being in the office has duration x at the interval t, and is not interrupted, it has duration x+y at the interval that follows t by y time.These properties do not apply to heterogeneous events.This results in linguistic differences such as those highlighted below.

D. Use of imperfective time-anchored duration phrases
As we will see in more detail in section 7.3 below, in order to express the duration that an eventuality has reached at a given (anchor) point in time, Portuguese uses specific duration phrases, namely phrases headed by a form of the verb haver (typically há).The closest English counterparts of these phrases are common duration for-adverbials in combination with an adverb that identifies the anchor point (e.g. now or then).
the museum is closed there-is two months 'The museum has been closed for two months now.' (37) a presença de tropas nesta ilha há dois meses the presence of troops in-this island there-is two months 'the presence of troops in this island for two months now' The described eventualities -which can be expressed either by sentential or by nominal means, as in ( 36) and (37), respectively -are imperfective, in the sense that they are assumed to hold at the anchor point (and, in principle, beyond it).Only descriptions of atelic eventualities occur in these constructions with durational haver 13 , and there seems to be no parallel structure (in Portuguese) to express telic duration.

E. Use of temporal measure quantifiers
Consider the following sentences: meaning.This seems to indicate that the issue of the Aktionsart values of NPs is rather complex and needs further investigation.I will not try to pursue it here. 13Some haver-phrases can combine with telic descriptions, but they are of a different type (non durational), as we will see in section 7.3.
(38) Após três meses de treino diário, o atleta sentia-se em forma.after three months of training daily, the athlete felt in shape 'After three months of daily training, the athlete felt in good shape.' (39) Três meses de disputas legais deixaram-no exausto.three months of disputes legal left him exhausted 'Three months of legal wrangling left him exhausted.' The italicised NPs contain a predicate of amounts of time -three monthswhich can be classified as a temporal measure quantifier since it has a quantifying function similar to expressions like three meters, three litres, or three tons when applied to massive nouns (on this notion of measure quantification, cf.Peres 1992).Though these expressions are not time adjuncts of duration in the same sense that for-phrases are, they exhibit notorious similarities with those adjuncts.Compare, for instance, (38) above with the following equivalent sentence: (40) Após treinar diariamente durante três meses, o atleta sentia-se em forma.after train daily for three months, the athlete felt in shape 'After training daily for three months, the athlete felt in good shape.' In the structures ( 38) and ( 39) above, the temporal quantifier applies to nominal or nominalised expressions -treino diário / daily training, disputas legais / legal wrangling -which represent atelic eventualities 14 .If telic events are involved in these structures, coercion into atelic events seems to occur systematically: (41) Após três meses de construção, a ponte estava quase pronta.after three months of construction, the bridge was almost ready 'After three months of construction, the bridge was almost ready.' There are also structures similar to these, but with quantification over sentential (rather than nominal or nominalised) constituents.In Modern European Portuguese, they involve an infinitive form preceded by the 14 Eventuality-denoting nominal phrases that do not combine with duration adjuncts headed by durante or em, but combine with the verb durar (cf.fn.12), can also occur in these constructions (though Portuguese seems to accept a wider range of phrases than English here): (i) três anos de guerra ('three years of war'), duas horas de incêndio (' ?two hours of fire'), dez anos de uma carreira cheia de sucessos (' ?ten years of a carrier full of successes') preposition a; in Brazilian Portuguese, a verbal -ndo form ("gerúndio"); in English, a verbal -ing form (not preceded by of) 15 : (42) Após três meses a treinar diariamente, o atleta sentia-se em forma.after three months PREP a train daily, the athlete felt in shape 'After three months training daily, the athlete felt in good shape.' The use of telic descriptions in these contexts similarly triggers an aspectual shift; English seems to prefer the use of nominalised of-sequences here: (43) Após três semanas a ler «Guerra e Paz», desisti.
after three weeks PREP a read «War and Peace», [I] gave-up 'After three weeks of reading «War and Peace», I gave up.' This construction with atelic descriptions must not be confused with a superficially similar one where (unshifted) telic descriptions occur.In Portuguese, the construction at stake may differ only in the preposition usedpara instead of a; a close English counterpart involves the preposition to.Compare: As we can see, the duration of atelic situations may be dependent on a time interval in a way that the duration of telic situations is not.In the most natural interpretation of ( 46), with an atelic situation (expressed by Ana worked), the mentioned amount of time -fifty hours -is the duration of the sum of all (possibly discontinuous) subsituations of the type mentioned that happened within the temporal frame (expressed by the second week of May).In other words, there is a maximality requirement, involving everything relevant that happened within the entire time interval.Therefore, if the temporal frame is enlarged, the description in the matrix clause may no longer be quantificationally accurate (since more subsituations of the relevant type may have to be taken into account).Observe the following schematic DRS--representation, where C A ⇒T symbolises the typical shift of atelic into telic associated with the use of duration for-phrases (cf.e.g.Bach 1981, Nerbonne 1983, Moens 1987, Mittwoch 1988, Swart 1998), ev and EV stand for atomic and non-atomic eventualities, respectively, and the subscripts T and A stand for telic and atelic, respectively: n t EV T EV T < n the second week of May (t) Telic situations, which are not homogeneous, have a different behaviour in parallel constructions -cf.(47).In these structures, the mentioned time interval serves merely to locate the described eventuality, which is defined independently of the temporal frame, and taken as atomic (despite the fact that it may involve discontinuity -cf.section 4).Thus, the transitivity of the inclusive location relation guarantees that, even when the frame is enlarged, truth is preserved.
(47′) n t ev T ev T < n the second week of May (t) ev T ⊆ t ev T : Ana read this book dur (ev T ) = mt 50 hours (mt) In sections 6 and 7, I will return to telic and atelic duration.With respect to the former, the main issue will be the distinction between telic duration phrases and several homonymous phrases that belong in different semantic classes.With respect to the latter, the main issue will be the distinction between three subtypes of atelic duration, marked in Portuguese (though not in English) by different temporal phrases.Before proceeding to this analysis, though, two side issues will be addressed briefly in the next two sections: the relationship between duration and (dis)continuity of eventualities, in section 4; the relationship between duration and time -anchoring of eventualities, in section 5.

Duration and (dis)continuity
Telic duration may present different degrees of exactness.Consider the following sentence: (48) A Ana corrigiu os testes em dez horas.
the Ana corrected the tests in ten hours 'Ana graded the tests in ten hours.' 16 When uttering this sentence, the speaker may have in mind two significantly different types of (durational) information: (i) if he focus on the boundaries of the event, ten hours is the time elapsed between the beginning and the end of the accomplishment (all interruptions ignored) -imagine Ana started grading 16 In Móia (2000: 328, fn. 299), I termed these two forms of duration loose and strict duration, respectively.The definitions given there, using the language of DRT were: the tests at 8 a.m and finished at 6 p.m.; (ii) if he focus on the sum of possibly discontinuous subparts of the event, ten hours is the duration of all these subparts (ignoring, for each subpart, pragmatically irrelevant interruptions)imagine that Ana graded the tests in three different days and timed it: 3,5 + 2 + 4,5 hours.I will term these two forms of duration continuous duration and discontinuous duration, again using a hypallage.As we can see, neither in Portuguese nor in English are these differences linguistically marked via different temporal connectives, rather they seem to be contextually determined (and dependent essentially on pragmatic facts).Atelic duration is also compatible with both continuity and discontinuity: (49) A Ana esteve no escritório durante dez horas.the Ana was in the office for ten hours 'Ana was in the office for ten hours.' In the reading that involves continuous duration, a single state (of a given temporal extent) is described.In the reading that involves discontinuous duration, reference is made to a sum of possibly discontinuous subsituations of the relevant type (Ana being in the office, here).The first reading may be coerced by modifying the predicate of amount of times with the adjective seguidas -quinze horas seguidas, 'fifteen hours in a row'.Curiously, no parallel simple form of coercion seems to exist for structures with telic duration, where predicates of amounts of time cannot combine with the adjective seguidas: (50) *A Ana corrigiu os testes em dez horas seguidas.the Ana corrected the tests in ten hours followed '*Ana graded the tests in ten hours in a row.' Note, moreover, the differences between continuous and discontinuous atelic duration that result from adding an adverbial that identifies a time interval: (51) Este fim-de-semana, a Ana esteve no escritório durante dez horas.this weekend, the Ana was in the office for ten hours 'This weekend, Ana was in the office for ten hours.' In the reading that involves continuous duration, and therefore a single state, the adverbial este fim-de-semana ('this weekend') merely provides a frame for inclusive location -a state with the mentioned duration happened anywhere within the relevant weekend (cf.DRS in (47′)).In this case, truth-preservation is guaranteed, even if the temporal frame is widened.In the reading that involves discontinuous duration, and therefore a sum of possibly discontinuous states , the preferred reading is probably the one where the homonymous adverbial este fim-de-semana provides a frame for abstraction over subeventualities -i.e. the sum of all (sub)states of Ana being in the office that happened within the frame has the described duration (cf.DRS in (46′)).As seen above, in this temporally bounded quantification reading, enlargement of the frame does not guarantee truth-preservation.

Duration and time -anchoring
As said before, simple asserted duration can be associated with a binary functor relating eventualities and predicates of amounts of time (dur in Kamp and Reyle 1993).In other words, duration does not involve -just by itselfany form of temporal location.However, it may be the case that the initial or ending point of the eventuality whose duration is specified coincides with a given, or known, time interval (e.g. a temporal perspective point, like the utterance time).In those circumstances, the eventuality is temporally anchored, and -again by hypallage -, w e may speak of time -anchored duration to refer to such situations.Some cases of time -anchored duration have been studied by Hitzeman (1993Hitzeman ( , 1997)).The author distinguishes "p(osition)-definite" and "non--p(osition)-definite" readings of sentences involving duration phrases like for an hour and in an hour.Let us consider each case separately.
(52) Martha will be in her office for an hour.(Hitzeman 1997: 89) According to Hitzeman (1997: 89), in the p-definite reading, for an hour is interpreted as "for the hour immediately following the utterance time"; this is, according to her, the only reading that we obtain in English, if the adverbial is in sentence-initial position.The non-p-definite reading is the one simply involving the duration of some future eventuality of Mary being in her office.These two readings are coincident with what I term here "time -anchored" and "simple" duration, respectively.In Portuguese, a similar situation can occur in sentences with durante-phrases: (53) A Marta vai estar no escritório durante uma hora.
the Marta goes be in-the office for one hour This sentence is ambiguous in a way similar to (52): the time adjunct in sentence-initial position strongly favours the time anchored duration; furthermore, this reading can be made explicit by changing the form of the time adjunct to durante mais uma hora (comparable to English for yet another hour).
Let us now consider the case with telic events.
(54) Smith & Co. will build a bridge in ten weeks.(Hitzeman 1997: 90) When in ten weeks is taken as a duration phrase, this sentence has an ambiguity parallel to the one in ( 52): it can mean that the construction will take ten weeks and will start at some indefinite point in the future -non p--definite reading, or simple duration reading -or that it will take the same amount of time and will start (immediately after) the utterance time.In sum, time adjuncts of duration -both telic and atelic -seem generally compatible with scenarios of temporal anchoring and scenarios of temporal autonomy.The two interpretations seem to be essentially context -dependent, and do not imp ly different time adjuncts (or temporal connectives).This situation is similar to the one observed in section 4, where we noted that the opposition between continuous and discontinuous duration is also essentially dependent on the context and not on the use of different connectives.
There is an interesting fact about Portuguese, though, with respect to time--anchored duration.Contrary to English, Portuguese -just like other Romance languages, for that matter -can use specific markers for time-anchored duration, but only in the case of atelic eventualities viewed both imperfectively and retrospectively.We will return to this issue in section 7.3.

Telic duration phrases and their homonyms
As said, adjuncts of telic duration are typically formed by the temporal prepositions em (Portuguese) and in (English) plus a bare predicate of Clearly, two lexical instances o f the preposition in are at stake herein 1 (durational) in the first complex sentence, and in 2 (locational) in the second one.It seems a mere coincidence that English expresses the two values at stake with the same preposition.Another example of mixing of the two homonyms was given above, concerning the p-and non-p-definite readings of Hitzeman (1997).
Portuguese does not use em-phrases as prospective locating phrases: if the perspective point is the utterance time, it resorts to phrases like daqui a cinco minutos or dentro de cinco minutos, which BP also uses (at more or less free variance with em cinco minutos); these forms correspond roughly to an English structure like five minutes from now (though this form does not seem as common as in five minutes) 18 .
the talk starts {from-here to / inside of} five minutes If the perspective point is not the utterance time, but e.g.some contextually salient time in the past, the phrases may have a slightly different form, or be completely different -e.g.daí a cinco minutos is used instead of daqui a cinco minutos (corresponding roughly to an English structure like five minutes from then 19 ); dentro de is normally not used when the perspective point is not the utterance time; conversely, phrases like passado(s) cinco minutos or cinco minutos depois, which are very common in these contexts, are not used when the perspective point is the utterance time (the most common English counterpart of these phrases is five minutes later).
(61) O orador entrou na sala.A palestra começou {daí a cinco minutos / passado cinco minutos / cinco minutos depois}.the speaker entered in-the room.the talk started {from-there to five minutes / passed five minutes / five minutes after 'The speaker entered the room.The talk started five minutes later.'Ambiguity between the durational and locational em-phrases may arise in Brazilian Portuguese, especially when combined with descriptions of achievements (cf.examples in Móia and Alves 2004).Furthermore, we note that it is not always easy to distinguish the two possible meanings: a simple prospective location of the achievement or the duration of the preparatory phase of that achievement (which has undergone an Aktionsart shift to an accomplishment).I will not explore this issue here.
iii. the quantification-bounding phrase, exemplified in the following sentence: (62) A Ana foi ao Brasil três vezes em vinte dias.the Ana went to-the Brazil three times in twenty days 'Ana went to Brazil three times in twenty days.' 18 Cf.the following examples from the BNC: "Ten minutes from now he would be climbing the stairs in the dark."(B1X 1694); "You get scratching those heads and I'll give you the answer in three minutes from now." (HEW 187). 19"Oh what I'll do when I'm ready I'll give you a bell which means it'll be what ten, fifteen minutes from then to picking you up (...)" (KCY 1441).
In my opinion, the italicised phrase in this sentence does not express duration or location (in the usual sense of these terms, at least).Rather, it provides a temporal frame for circumscribing quantification, much in the same line as temporal adverbials in (46′) above (cf.Móia 2006), with the difference that the frame here is an unspecified interval of the mentioned duration, rather than a definite stretch of the time axis.The meaning of ( 62) can be roughly paraphrased as follows: there is an interval of time [t] of a given extent [mt]twenty days -such that the sum of all events [E] of the described type [ev] -Ana going to Brazil -contained in it has cardinal three.In the language of DRT: (62′) t mt EV 20 days (mt) dur (t) = mt EV = Σev: ev ev ⊆ t ev: Ana go to Brazil |E| = 3 Note that, contrary to most duration phrases, these bounding adverbials readily occur in sentence-initial position: (63) Em vinte dias, a Ana foi ao Brasil três vezes.
in twenty days, the Ana went to-the Brazil three times 'In twenty days, Ana went to Brazil three times.' Note also that they are equivalent to phrases headed by other connectives (which may occur either in sentence-initial or in sentence-final position), viz.no espaço de, in Portuguese, and over 20 , in English, or phrases with indefinite time -denoting phrases like um período de vinte dias / a period of twenty days: Finally, note that the quantification at stake need not be direct quantification over events, expressed by quantifiers like n vezes / n times.It may as well be indirect quantification over events, expressed by e.g.NPs with nominal cardinal quantifiers and distributive readings, or exhaustive enumeration of events, expressed via conjunction (cf.Móia 2000Móia , 2001a for a more complete list of contexts where temporal circumscription of event-quantification emerges):

Subtypes of atelic duration -simple, time-anchored and planned
The three Portuguese sentences below express (atelic) duration.They are roughly parallel to English sentences with the same temporal preposition -for -but involve, in Portuguese, three different temporal connectives: (67) A Ana esteve no escritório durante duas horas.
the Ana was in-the office for two hours 'Ana was at the office for two hours.' (68) A Ana está no escritório há duas horas.the Ana is in-the office there-is two hours 'Ana has been in the office for two hours now.' (69) A Ana saiu por duas horas (mas só regressou ao fim de três).
the Ana left for two hours (but only returned at-the end of three) 'Ana left for two hours (but only returned three hours later).' These sentences illustrate three different subsystems of asserted duration of atelic eventualities via temporal adjuncts, which are typically marked in Portuguese (especially in European Portuguese), contrary to what happens in English, by different temporal connectives: simple duration, time -anchored duration and planned duration.In simple words, they might be distinguished as follows: i. simple duration is independent of the position of the described eventualities in the time axis (i.e.non-time-anchored) and does not have an intensional component; ii.time-anchored duration expresses for how long an eventuality has been going on at a given (anchor) point in time (or, alternatively, for how long it will last from a given anchor point in time onwards); iii.planned duration has an intensional component (as opposed to the simple and the time -anchored duration, which are extensional); it expresses the anticipated duration of the consequent state of a given (telic) eventuality; the actual duration of that state may or may not coincide with it.

Simple (atelic) duration
Simple duration is defined negatively, as non-time -anchored and non--planned (or extensional).As said in section 3, it is expressed by durante--phrases in (European and Brazilian) Portuguese, by synonym por-phrases in Brazilian Portuguese 21 , and by for-phrases in English; furthermore, it is possible -both in Portuguese and English -to express it via bare predicates of amounts of time (with no overt preposition): (70) A princesa dormiu ({durante / por BP }) cem anos.
the princess slept (for) hundred years 'The princess slept (for) one hundred years.' It must be stressed that the Portuguese preposition durante has a locational homonym.This homonym takes time -denoting phrases, rather than bare predicates of amounts of times as complements: durante o ano de 1995 ('during 1995'), durante o período das eleições ('during the election period'), durante os últimos três anos ('for the last three years').In some cases, locational durante-phrases express -exclusively or preferably -a durative reading, allowing inferences about the duration of the located events (according to Table 3, in section 1.2); this is namely the case when they are combined with: (i) universally quantified NPs with eventuality-denoting predicates (e.g.toda a palestra, 'the whole talk'); (ii) time -denoting phrases that contain deictic adjectives combined with predicates of amounts of time (e.g. as últimas três semanas, 'the last three weeks').Curiously, in this type of durative contexts, English often (or preferably) resorts to the preposition for (which I take to be a locational and not a durational connective here) 22 . 21Modern European Portuguese does not normally use por-phrases to express simple atelic duration (unless in some restricted cases, e.g.those expressing very short duration like só a vi por uns segundos ['I only saw her for a second']).However, there are many instances of this use in classical Portuguese writers.For a more thorough analysis of these facts, cf.Móia (2001b). 22English during-phrases are also compatible with durative readings -cf. the following examples from the BNC: "Where were the two girls during the last In these sentences, the described atelic situations are viewed imperfectively, as holding at the utterance time, but in the first case the duration is retrospective -in the sense t hat the situation is said to have reached the mentioned duration at the utterance time -and in the second case it is prospective -in the sense that the situation is said to last for the mentioned amount of time starting from the utterance time.In Portuguese, there is a clear asymmetry between retrospective and prospective (imperfective atelic) duration.The second case, which I will ignore from now on, does not resort to any special form of temporal connectives; rather, it uses the same expressions as simple duration: typically durante-phrases 23  Often, the anchor point is not explicit, and coincides with a temporal perspective point (identified by the tense of the main clause, which coincides with the tense of the verb fazer).(ii) a very similar construction with the verb haver25 , that has the following specificity: the anchor point is never explicit, and necessarily coincides with a temporal perspective point (identified typically by the tense of the main clause), i.e. there are haver-constructions parallel to (79)-( 80) but not to ( 77 Note that, in these sentences, the tense of the verb haver coincides with that of the main clause (just like in the constructions with fazer); however, the formally present form há may occur in any context, even with non-present perspective points: (83) Há dois anos que a Ana estava em Paris.[equivalent to (80)-( 82)] there-is two years that the Ana was in Paris 'Ana had been in Paris for two years then.' This fact seems to indicate that the form haver is losing its verbal character in this construction 27 , the whole phrase it heads possibly being reanalysed as a time adjunct of duration 28 .
It must be noted that the Portuguese sentences with haver, like ( 81)-( 82), are not ambiguous, whereas their English counterparts -as has often been noted -are ambiguous if the anchor-point is not made explicit via adverbs like now or then.Thus, a sentence like Ana has been in Paris for two years may indicate that the situation occurred somewhere in the past (non-anchored duration) or that it reached the mentioned duration at the utterance time occurs in sentence-final position without any complementiser (cf.fn.24): (i) A Ana está em Paris há dois anos.the Ana is in Paris there-is two years 27 Some verbal characteristics still remain, e.g. the possibility of combining haver with adverbs like já ('already') or ainda não ('not yet'). 28If so, a functor slightly different from Kamp and Reyle's dur would have to be introduced in order to represent these time adjuncts of duration -say, a functor A--dur relating an eventuality (ev), an amount of time (mt) and a time-interval that acts as an anchor point (t); the DRS-conditions associated with this form of duration would have to be something like: [A-dur (e, t) = mt].
On the other hand, it is a curious fact that the Portuguese constructions with haver -just like, for that matter, those with fazer -are ambiguous, inasmuch as they may also be used to express temporal location, instead of duration 29 .This occurs notoriously when telic durationless events are involved: (84) O bebé nasceu há dois meses (atrás).
the baby was-born there-is two months (behind) 'The baby was born two months ago.' (85) Faz amanhã dois meses que o bebé nasceu.makes tomorrow two months that the baby was-born 'The baby was born two months ago tomorrow.' Observe that the English counterparts of these sentences involve totally different temporal expressions, namely constructions with ago (instead of constructions with for).In fact, the temporal locating phrases with haver and ago are the retrospective counterparts of the prospective phrases discussed in section 6: phrases with e.g.dentro de or daí/daqui a, in Portuguese, em in Brazilian Portuguese, and in, in English (cf.( 59)-( 60)).
There are many linguistic clues that tell apart the two uses of these temporal expressions with haver and fazer (including Aktionsart and tense differences) 30 , but I will not consider them here for space reasons (cf.Móia 1999, Móia andAlves 2004).I will focus on one difference, though, that involves only haver-phrases.In the locational reading, these phrases are compatible with a postpositional redundant element atrás (literally 'behind') 31 29 They can also behave as temporal quantification adverbs (in the sense of kamp and Reyle 1993) in structures parallel to (81)-( 82), or (i) in fn.26; the relevant structures differ from those presented here in that quantified (discontinuous) intervals or events are used in the complement of haver: a Ana não vai ao ginásio há três fins-de-semana ('Ana hasn't been to the gym for three weekends now'), o Paulo não marca um golo há três jogos ('Paulo hasn't scored a goal for three matches now')cf.Móia (2003). 30In some cases, that I will not consider here, sentence ambiguity may arise (cf.Móia 1999). 31With past and future anchor points, no postpositional element is used in standard Portuguese: (i) O bebé tinha nascido há dois anos (*atrás).the baby had been-born there-is two years behind these variants of the locational constructions with haver, cf.Móia and Alves (2004).

Planned duration (of consequent states)
It has often been noted in the literature about English that for-phrases can refer to the duration of consequent states of telic -typically punctual -events represented in main clauses 32 (cf.e.g.Binnick 1969; McCawley 1971, 1974;  Dowty 1979; Mittwoch 1980; Moens and Steedman 1988; Parsons 1990;  Hitzeman 1993) Moreover, it has been pointed out that, in many cases, these for-phrases do not necessarily indicate the actual duration of the described eventualities; rather, they refer to an intended or planned duration (on the part of some Agent), whence the non-contradictory character of sentences like the following: (93) The sheriff of Nottingham jailed Robin Hood for four years, but he stayed in jail only three days.(cf.Hitzeman 1993: 17) It has also been pointed out that intention does not seem to be a necessary component of English for-constructions that express the duration of consequent states: as Hitzeman (1993) notes, there are felicitous uses of this type of for-phrases which do not involve any intention or purpose whatsoever, like: (94) The hiker lost his way for several hours.(Hitzeman 1993: 15) 32 Not all telic events allow this combination though (cf.Hitzeman 1993, Móia 2000).Hitzeman (1993), inspired by Pustejowsky (1991), proposes to deal with contrasts involving the (un)capacity of for-phrases to combine with different telic eventualities by resorting to a different (basic) Aktionsart classification of the eventuality-descriptions involved: "To explain the ability of the for-phrase to describe the duration of the result state in [the sheriff of Nottingham jailed Robin Hood for four years, and not in *Smith & Co. built a bridge for 10 weeks] (...), I propose that these eventualities are members of different aspectual classes" (p.23); "I propose that eventualities be separated into classes based on the type of interaction they may felicitously have with a for-or until-phrase" (p.108).
Accordingly, sentences like (90) can be taken as ambiguous, allowing both an intentional and a non-intentional interpretation: "Although there is a reading [of the sentence the sheriff of Nottingham jailed Robin Hood for four years] in which the for-phrase describes the intended duration of the result state, it is important to note that there is another reading which does not involve intention" (Hitzeman 1993: 18).It is of interest to note that the difference between actual and planned duration may have a lexical expression in Portuguese (more evidently so in modern European Portuguese than in Brazilian Portuguese, as will be stressed below).In fact, modern EP uses preferably durante-phrases to express actual duration and por-phrases to express planned duration.The use of por-phrases to express actual duration -whether of consequent states or not -is uncommon in modern EP (except for some limited contexts33 ), though it is very common in BP.On the other hand, durante-phrases are seldom used to express planned duration (in both Portuguese varieties), though they might be used without utter ungrammaticality arising: (95) A Ana saiu {por / ?? durante} uma hora mas só voltou ao fim de três.
the Ana left for an hour but only returned at-the end of three 'Ana left for an hour but returned only three hours later.' It is also interesting to note that the type of Aktionsart shift that affects punctual events in these duration constructions (viz.its association with a result state and hence its atelicisation, so to speak), can also affect some temporal locating durative phrases, in particular English phrases with until and Portuguese phrases with até (cf.Mittwoch 1980, Móia 2000).In this case, however, the planned reading seems obligatory, rather than optional (according to e.g.Mittwoch 1980).Consider the following sentence: (96) Ele emprestou-me o livro até segunda-feira.
he lent-me the book until Monday 'He lent me the book until Monday.' (Mittwoch 1980: 220) Até-and until-adverbials define a location time that stretches between the time nailed down by their complement (the mentioned Monday, here) and some contextually determined point in its past.The sentence means that the lending took place at a given point before the mentioned Monday (this point marking the beginning of the location time), and its consequent state − the speaker being in possession of the book − is intended to hold until that Monday.That intention is an essential component of this construction is once more demonstrated by the non-contradictory character of sequences like the following (as stressed by Mittwoch 1980): (97) Ele emprestou-me o livro até segunda-feira, mas eu devolvi-lho no domingo.he lent-me the book until Monday, but I returned-it-to-him on-the Sunday 'He lent me the book until Monday, but I gave it back on Sunday.' Note that, under this analysis, the use of English until, which is normally incompatible with non-durative readings, is not exceptional in (96).

Conclusion
In this paper, a global view on the expression of duration in Portuguese was presented, and a comparison with English was outlined.The first issue to be addressed was the need to distinguish between asserted duration and inferred duration, given that inferences about duration are pervasive in natural language discourse, be it in association with certain temporal locating expressions, or with discourse mechanisms -cf.section 1.2.
In the domain of asserted duration, several distinctions were underlined.Those that are linguistically more prominent in the two languages under consideration are schematised in Table 4 below.One of the most important distinctions is certainly that between the duration of telic and atelic eventualities, which is remarkably similar in Portuguese and English -cf.section 3.As for the different subtypes of atelic duration -simple, time --anchored and planned -it was observed that Portuguese uses a wider variety of linguistic resources (in terms of duration adjuncts or duration predicates) than English.Consequently, some ambiguities that arise in English -like the well-known case of sentences like John has lived in Amsterdam for three years -do not emerge in Portuguese.
It was also observed that homonymy is widespread in the area of durational expressions, which, on the one hand, may give rise to ambiguities, and, on the other hand, may result in disputes about categorisation.In particular, it was noted that the temporal connectives durante, haver and em, in Portuguese, and for and in in English, have locational and durational variants, and, furthermore, that predicates of amounts of times headed by e.g.em or in may serve as temporal bounders for quantification over events, in a construction that is not exactly "locational" or "durational" in the common sense of these words.Also, it was noted that atelic duration has common characteristics with temporal measure quantification over atelic eventualitiescf.topic E, in section 3.In general, this exposition showed that different semantic domains, like temporal location, temporal circumscription of quantification and temporal measure quantification, are intimately intertwined with duration.

( 65 )
Em vinte dias, a Ana fez três viagens ao Brasil. in twenty days, the Ana made three trips to-the Brazil 'In twenty days, Ana made three trips to Brazil.' (66) Em vinte dias, a Ana visitou o Brasil, o México, o Belize e Cuba. in twenty days, the Ana visited the Brazil, the Mexico, the Belize and Cuba 'In twenty days, Ana has visited Brazil, Mexico, Belize and Cuba.'

( 79 )
Faz (agora) dois anos que a Ana está em Paris.makes (now) two years that the Ana is in Paris 'Ana has been in Paris for two years now.' (80) Fazia (então) dois anos que a Ana estava em Paris.makes (then) two years that the Ana is in Paris 'Ana had been in Paris for two years then.' Há dois anos que a Ana está em Paris 26 .[equivalent to (79)] there-is two years that the Ana is in Paris 'Ana has been in Paris for two years now.' (82) Havia dois anos que a Ana estava em Paris.[equivalent to (80)]there-was two years that the Ana was in Paris 'Ana had been in Paris for two years then.' . Observe the following examples: (90) The sheriff of Nottingham jailed Robin Hood for four years.(Binnick 1969) (91) I borrowed the book for a week.(Huddleston and Pullum 2002: 705) (92) I sent him out for half an hour.(ibid.)

Table
Dowty 1979, andHuddleston andPullum 2002) durou um ano. the situation of the Ana study Architecture lasted one year ' ?The situation of Ana studying Architecture lasted one year.'Whensententialatelicargumentsare involved, English uses typically the verb spend (cf.e.g.Dowty 1979, andHuddleston andPullum 2002), and Portuguese the verb passar 10 : (29) A Ana passou um ano a estudar Arquitectura.theAna passed one year PREP a study Architecture 'Ana spent one year studying Architecture.
'Note that the verbs durar and last cannot combine with telic descriptions:(30) *A construção desta ponte pelos Japoneses durou um ano.the construction of this bridge by the Japanese lasted one year '*The construction of this bridge by the Japanese lasted one year.'(31)*A situação de os Japoneses construírem esta ponte durou um ano. the situation of the Japanese build this bridge lasted one year '*The situation of the Japanese building this bridge lasted one year.' Três semanas a ler este livro é muito.three weeks PREP For the sake of simplicity, I will only give examples of European Portuguese below.The Brazilian Portuguese counterparts are obtained by substituting the «gerúndio» (e.g.estudando) for the combination of the preposition a and the infinitive (e.g. a estudar).A Ana trabalhou durante cinquenta horas em Maio. the Ana worked for fifty hours in May 'Ana worked for fifty hours in May.' (47) A Ana leu este livro em cinquenta horas na segunda semana de Maio. the Ana read this book in fifty hours in-the second week of May 'Ana read this book in fifty hours in the second week of May.' → A Ana leu este livro em cinquenta horas em Maio. the Ana read this book in fifty hours in May 'Ana read this book in fifty hours in May.' a read this book is much 'Three weeks (of) reading this book is a long time.' (45) Três semanas para ler este livro é muito.three weeks to read this book is much 'Three weeks to read this book is a long time.' F. Different behaviour in combination with frame adverbials Expressions involving atelic and telic duration are associated with different inferential patterns in constructions with adverbials that identify time intervals (cf.Móia 2000, for more details).Consider the following examples, where fifty is to be interpreted in both examples as an exact cardinal quantifier (i.e. as equivalent to exactly fifty): (46) A Ana trabalhou durante cinquenta horas na segunda semana de Maio. the Ana worked for fifty hours in-the second week of May 'Ana worked for fifty hours in the second week of May.' -/→ 15 Hitzeman (1997)ding is what I term time -anchored reading, but, curiously, it is not whatHitzeman (1997)calls a p-definite reading (in fact, the author does not mention this reading of the sentence at all).What she terms -misleadingly, in my opinion -p-definite reading, is a reading that has nothing to do with duration: the construction of the bridge will take place (or start) ten weeks after the utterance time.As we can see, in this reading, in ten weeks is a temporal locating phrase, that involves reference to the interval that follows the utterance time by the specified amount of time, i.e. it is a different time adjunct that English, by coincidence, expresses with homonyms; European Portuguese, as said already, would use different adverbials: em dez semanas, for duration (time -anchored or not), and dentro de dez semanas or daqui a dez semanas, for location: (55)A Smith & Co. vaiconstruir a ponte em dez semanas.the Smith & Co. goes build the bridge in ten weeks [time -anchored or simple duration] (56) A Smith & Co. vai construir a ponte daqui a dez semanas.the Smith & Co. goes build the bridge from-here to ten weeks [temp oral location] . As for the first case, it is linguistically more interesting in Portuguese, inasmuch as it has specific markers.If one wants to express for how long an atelic situation has been going on at a given point in time, one of two constructions is normally used in Portuguese: (i) a predicate-argument combination with the duration predicate fazer -one argument describing the measured eventuality, a second one the amount of time corresponding to its duration and a third one the temporal anchor, as in the following examples:(77) Faz amanhã dois anos que a Ana está em Paris 24 .makestomorrow two years that the Ana is in Paris 'Tomorrow, Ana will have been in Paris for two years.'(78)Fez ontem dois anos que a Ana está em Paris.made yesterday two years that the Ana is in Paris 'Yesterday, Ana had been in Paris for two years.'