3.1.1. The notion of duration and three different forms of duration

I will adopt here a restricted notion of duration, following Móia (2006, pp. 37–42). In this restricted sense, which is in line with Kamp and Reyle’s (1993) definition of “temporal measurement” (of eventualities), duration involves reference to the amount of time a situation lasts (for atelic eventualities), or takes to culminate (for telic eventualities). In Discourse Representation Theory (DRT) – which I will be using here to formally describe the semantic differences under scrutiny –, duration is represented by a one place functor dur, which maps eventualities (ev) – or intervals (t) – on the amounts of time (mt) they last, occurring in conditions like [dur (ev) = mt] (cf. Kamp & Reyle, 1993, p. 648). This definition excludes from the class of duration adjuncts, for instance, phrases with Portuguese desde or até, and their English counterparts, since and until, which some authors group together with the prototypical (atelic) duration adjuncts (i.e., phrases with Portuguese durante, or English for). Phrases with desde/since or até/until will be regarded as just particular forms of temporal location adjuncts – typically associated with durative location, a concept not to be confused with strict duration – and will therefore not be investigated here (though, as we will see, a partial analysis of desde/since-phrases needs to be made at some point, for comparative purposes). Duration – unlike temporal location – is in principle established independently of the position of the described eventualities on the time axis. However, one particular form of duration (that I will term time-anchored duration) additionally involves systematic anchoring to the time axis. The adjuncts signalling this form of duration, thus function, for all intents as purposes, both as duration and as temporal location adjuncts. However, formally they belong to the grammatical category of duration adjuncts, and are also analysable with the DRT’s functor dur. They correspond to Portuguese adjuncts with the form haver x-time (in the variants há x-time, by far the most common, and havia x-time), where x-time stands for a predicate of amounts of time like dois anos e meio (‘two and a half years’) – cf. (29) below –, and their English counterparts for x-time (when they are equivalent to for x-time now/then).

Assuming the restricted notion of duration above (and the grammatical distinctions discussed in Móia, 2006), three different forms of duration need to be considered: telic duration, non-time anchored (or simple) atelic duration, and time-anchored atelic duration. These are crucial to understanding the grammatical facts explored in this paper. These three forms of duration are typically associated with different prepositional connectives in Portuguese, but two of them – anchored and non-anchored atelic duration – share the same prepositional connective (for) in English, and are therefore not so easily and immediately told apart in that language. The following three sentences illustrate the three forms of duration in question, with the relevant prepositional connectives bolded:

Note, marginally, that the duration connectives em and durante have homonyms that express pure location, as in podem fazer-se visitas guiadas {no Verão/durante o Verão} ‘guided tours are available {in the Summer/during the Summer}’. Whenever necessary, I will use subscripts dur and loc – e.g., durante_dur, durante_loc – to distinguish these homonyms.

Let us briefly summarise the main grammatical aspects to have in mind. The duration of telic and atelic eventualities is expressed with adjuncts headed by different connectives both in Portuguese and in English, a well-known and much-discussed fact in the literature. To refer to the amount of time (mt) it takes, or took, for a given telic situation (ev) to be completed, as in (27), Portuguese uses phrases headed by em, and English uses phrases headed by in. The relevant truth condition imposed on sentences with this type of duration can be represented, in the language of DRT, via the (already mentioned) condition [dur (ev) = mt]. To refer to the whole duration of a given atelic situation, not anchored to any specific point of the time axis, as in (28), Portuguese uses durante (or – mainly in Brazilian Portuguese – por), though the preposition can normally be omitted, and English uses for, which can also sometimes, under the right conditions, be omitted. The relevant truth condition for sentences with this form of duration is represented, in the same logic language, also via the condition [dur (ev) = mt]. Finally, to refer to the duration a given atelic eventuality (ev) reaches at a certain point of the time axis (say, an anchor point, A_pt), as in (29), Portuguese uses a partially grammaticalized form of the verb haver (‘there be’), which behaves much like a prepositional connective. Formally, the asserted amount of time (two and a half years, in example (29)) is not necessarily the duration of the whole described eventuality (ev), but just of the part of it (ev′) that extends up to the anchor point, i.e., sentences with this type of duration state [dur (ev′) = mt] (thus allowing the possibility that ev continues after Apt) (cf. Móia, 2011a). To express this form of duration, English resorts to for-phrases as well (though the preposition can sometimes be omitted). This latter form of duration – that I term time-anchored – is also distinguished in the English literature from the previous one – that I term non-time-anchored (or simple). For instance, Declerck (1991, p. 323) uses the term “continuative” (reading) – as opposed to “indefinite” or “existential” –, and Hitzeman (1997, p. 88) uses the term “position-definite” (reading) – as opposed to “non-position-definite” – to refer to it. However, as said, English, unlike Portuguese, does not distinguish these two forms of atelic duration – time-anchored, and non-time-anchored – via the use of different connectives; rather, it relies on other linguistic cues to tell them apart. For instance, in English, adding a deictic or anaphoric expression (for two years now/then), or preposing the adverbial, leaves only the time-anchored reading available; tense differences are also explored in English to tell the two readings apart – continuous forms are associated with time-anchoring (John has been living in Paris for two years), preterite forms are associated with non-time-anchoring (John lived in Paris for two years), and perfect non-continuous forms typically give rise to ambiguity (John has lived in Paris for two years). Indeed, ambiguities between the two readings are a much-discussed topic in the semantic literature – cf., among others, Dowty (1979, p. 343), Richards (1982, p. 96), Heny (1982, p. 143 ff.), Mittwoch (1988, p. 203 ff.), Declerck (1991, p. 322 ff.), Kamp and Reyle (1993, p. 567 ff.), Hitzeman (1997, p. 87 ff.).

Furthermore, with respect to the ways of expressing duration – in any of its three forms –, two grammatical strategies need to be distinguished: duration can be expressed via a temporal adjunct, as in the three examples above (adjunct duration), or it can be expressed via a predicate-argument combination, as in the three examples below (argumental duration). In this second case, the sentence has as its main predicate a durative verb (e.g., take, spend, last, and their Portuguese counterparts); this verb selects as one of its arguments a predicate of amounts of time and as the other argument a sentence or nominal phrase identifying the relevant (telic or atelic) eventuality (whose duration is asserted).

Typically, different verbs are used to express the different forms of argumental duration (e.g., take for telic duration, and spend for atelic duration), but some verbs can be used to express more than one form of duration, as is strikingly the case of Portuguese levar (‘take’), in (30) and (32) – cf. Móia (2015) on this issue. I will revisit this ambivalence below.

3.1.2. Predicates of amounts of time vs. time-denoting expressions containing predicates of amounts of time

I will discuss here the distinction between two closely related grammatical categories: that of pure predicates of amounts of time like two and a half hours, and that of time-denoting expressions (definite NPs) that contain predicates of amounts of time, like the two and a half hours I was waiting for you at the airport, or those two and a half hours. These categories are sometimes hard to tell apart. However, they describe entities with different ontological statuses: amounts of time, which can be conceived of as properties of intervals, and time intervals, or stretches of the time axis, respectively. These entities are formally distinguished in Kamp and Reyle (1993), who mark them with different discourse referents: mt and t, respectively (a notation I will follow henceforth, and often use in subscript, for clarity purposes). The categorial distinction at stake runs parallel to the distinction between genuine duration adjuncts (for/in two and a half hours), and temporal locating adjuncts (for/in the two and a half hours I was waiting for you at the airport) that are comparable in that they allow systematic inferences concerning duration (cf. Móia, 2000, pp. 135–154). As we will see, these categorial distinctions appear to play a crucial role in some of the grammatical regularities and restrictions we will be discussing later in this section (cf. Table 2, at the end of this section).

The antecedents of all the relative clauses I will be analysing here either include or coincide with predicates of amounts of time. As said, predicates of amounts of time are expressions that represent properties of intervals of equal duration, marked by discourse referents of type mt (cf. Kamp & Reyle, 1993, p. 648). They can be quantified phrases without a determiner (abbreviated here as x-time), like duas horas ‘two hours’, duas semanas e meia ‘two and a half weeks’, or quase cinco anos ‘nearly five years’. For the purposes of this paper, I will focus on structures where this type of predicate is modified by a relative clause, i.e., where it integrates definite NPs like (33) (though it can also be modified by PPs, as in as duas horas de conversações ‘the two hours of talks’). As we will see, the whole NP in (33) can itself be a complex predicate of amounts of time (mt), though it can also be a time-denoting expression (t).

Furthermore, predicates of amounts of time can be formed with the hypernymic noun tempo, or with the hypernymic lexeme quantidade de tempo ‘amount of time’, as their nucleus. In this case, they always correspond to a definite NP with a modifier (e.g., a relative clause, in the cases we will be interested in), as (34). Since the hypernymic noun tempo (though not quantidade de tempo ‘amount of time’) is ambivalent, NPs with the superficial form of (34) and the noun tempo can also be time-denoting expressions, rather than predicates of amounts of time, as we will see.

As said, the complex NPs with the structure in (33) are ambivalent, since they can globally identify a time interval (t discourse referent, in DRT) – qualifying as a time-denoting expression –, or an amount of time (mt discourse referent, in DRT) – qualifying globally as a predicate of amounts of time. A phrase like the two hours I spent in the library (like its Portuguese counterpart as duas horas que eu passei na biblioteca) can refer to either type of temporal entity, with the wider context (showing the predicates the NP combines with) normally allowing for an non-ambiguous interpretation – cf., e.g., during [the two hours I spent in the library]_t nobody called me/n[as duas horas que eu passei na biblioteca]_t, ninguém me telefonou (definite NP as a time-denoting expression) vs. [the two hours I spent in the library]_mt were not enough to study the subject/[as duas horas que eu passei na biblioteca]_mt não chegaram para estudar o assunto (definite NP as a predicate of amounts of time).

In the first case, the definite NP refers to the stretch of the time axis that corresponds to the location (or loc, sensu Kamp & Reyle, 1993) of the eventuality described in the relative clause (ev). The DRS-conditions associated with the interpretation of the nominal phrases at stake are those in (35). Note that, in this case, the relativized constituent – if adjunct – does not classify as a duration adjunct, but rather as a temporal location adjunct, though the distinction (to which I will return later) is subtle, and in many cases inconsequential.

In the second case (perhaps less frequent, but not at all rare), the definite NP refers to an amount of time, namely the amount of time associated with the eventuality described in the relative clause (ev). The DRS-conditions associated with the interpretation of the nominal phrases at stake are as follows:

The following two examples from the corpus cetempúblico illustrate each of the values at stake: time intervals in (37), amounts of time in (38).

As for the complex NPs with the structure in (34) – with amount of time, or with time meaning ‘amount of time’ –, they always refer to amounts of time, i.e., they always associate with the conditions in (39), as is clear in the corpora examples (40)–(41):

However, since the noun tempo ‘time’ can also be the hypernym of a time-denoting expression, meaning ‘interval’, or ‘period’, rather than ‘amount of time’, a phrase superficially identical to (34) (with time, though not with amount of time) can have a different interpretation, namely one parallel to (35). Observe, for instance, the following example from the corpus cetempúblico:

When amounts of time are present, in adjuncts or as arguments of durative verbs, genuine duration (formal condition [dur (ev) = mt]) is conveyed – cf., e.g., o cientista passou dois anos e meio a estudar o problema ‘the scientist spent two and a half years studying the problem’. When time-denoting expressions containing predicates of amounts of time are present, in adjuncts or as arguments of durative verbs, formally, temporal location rather than genuine duration is conveyed – cf., e.g., o cientista passou os dois anos e meio que viveu em Stanford a estudar o problema ‘the scientist spent the two and a half years he lived in Stanford studying the problem’; however, in these structures, duration is inferable: the locating condition [loc (ev) = t] (if exact location is expressed), or [t ⊆ ev] (if inexact location is expressed, allows the inference [dur (ev) = dur (t)], or [dur (ev) ≥ dur (t)], and dur (t) is known, since it coincides with the given predicate of amounts of time (i.e., two and a half years in the example above). In other words, the dividing line between the two domains – duration and durative location – is somehow blurred (and somehow conflates), when predicates of amounts of time are involved, as is the case with the structures I will be analysing in this paper.

Let us now consider – in the next three subsections – the expression of the mentioned three forms of duration, and comparable forms of temporal location, both via adjuncts and via argument-predicate combinations, in relativized constituents in Portuguese. I will start with the most complex case – time-anchored (atelic) duration – and then analyse the other two.

3.2.1. Relativization of adjuncts expressing time-anchored atelic duration

Relativization of time-anchored duration adjuncts is by far the most interesting, and complex, situation from a grammatical perspective. At least four constructions (some with subvariants), plus two very closely related structures not directly involving relative clauses (but rather adjuncts with desde ‘since’ or de ‘of’), need to be considered (identified below as A–F). All of them are found, with more or less frequency, in the corpora analysed, but – crucially – not all of them are regarded as standard. Let us go through the menu of possibilities for expressing time-anchored duration in relativized constituents, and comparable constructions.

Consider, as a starting point, the meaning conveyed by the following (non-relative) sentence with a time-anchored duration adjunct há x-time (há dois anos e meio ‘for two and a half years’).

• Relativization with há que (há que_mt and há que_t, in constructions A1 and A2, respectively)

If the predicate of amounts of time in (45) is relativized, since Portuguese, contrary to English, (i) does not have preposition-stranding (hence, the preposition-like connective há cannot stay in situ), and (ii) does not allow omission of the relative pronoun (que), one might expect that a relative clause conveying the relevant part of the meaning of (45) would be the one in the second line of (46), with the relativized duration constituent há que_mt¹¹ (henceforth, referred to as Construction A1). (46) is, however, not standard Portuguese.

In (46), the relative pronoun stands for a genuine predicate of amounts of time (two and a half years), whence the subscript mt in há que_mt. But we need to consider a variant of this construction (say Construction A2), where the relative pronoun stands for a time-denoting expression of the type described in (33) above (i.e., a definite NP like os dois anos e meio que eles viveram juntos, ‘the two and a half years they lived together’), in a context that elicits the relevant time-denoting interpretation, as (47) below. Since this type of time-denoting phrases – in fact, any type of time-denoting phrases – cannot be the complement of há (a connective that only combines with genuine predicates of amounts of time), the resulting construction would be expected to be even more anomalous than the one in (46).

As said, the constructions in (46) and (47) are felt as odd by Portuguese speakers. However, they seem to be emerging (at least in Portugal) as a novelty. In the 190 million-word Portuguese corpus cetempúblico, relative clauses with há que only occur twice: example (48), which illustrates construction A1, and example (49), which illustrates construction A2. These are the only two examples found in that big corpus, and – very curiously – these are the only two examples found in all corpora of the Linguateca website, that totals more than 1.3 billion words (and includes, besides cetempúblico, very large corpora of Brazilian Portuguese, one with nearly 1 billion words). Its rarity is likely correlated with the sense of anomaly it brings about, but I believe that this construction (at least the variant A1) may become more widespread in the future. Indeed, it is very useful in face of the restrictions affecting the other relativization strategies (that we will discuss below – cf. also Table 2 at the end of section 3). Furthermore, this construction seems to represent a step forward in the grammaticalization process affecting há, which – as said before – is turning it into a preposition-like connective (and as a preposition is expected to regularly move together with relative pronouns).

The only standard alternative to construction A1 using a relative clause seems to be the one where a predicate of amounts of time selected by a durative verb, like ter or levar, is relativized, i.e., structures with argumental – rather than adjunct – duration (cf. 3.2.2 below) – cf. dois anos e meio é o tempo que a Ana e o Pedro têm de amizade, dois anos e meio é o tempo que a Ana e o Pedro já levam como amigos. A similar alternative to construction A2 is also possible – cf. {nos dois anos e meio que a Ana e o Pedro têm de amizade/nos dois anos e meio que a Ana e o Pedro já levam como amigos}, muita coisa aconteceu. However, with respect to construction A2, no less than three other alternatives involving relativization (constructions B, C and D below) can be considered.

Furthermore, two constructions without relativization, using adjuncts with the connectives desde (‘since’) or de (‘of’), which can convey similar information via different syntactic means, need also be taken into account (constructions E and F below).

• Relativization with que and em que (constructions B and C, respectively) + tenses expressing overlapping to TPpt

Let us now consider two variants of the construction in (47), which are also non-standard. In one, say Construction B, illustrated in (50), only a relative pronoun representing the time-denoting expression surfaces at the head of the sentence. In another one, say Construction C, illustrated in (51), the locating preposition em (‘in’) is used before the relative pronoun. It must be emphasized that these two constructions (just like construction A2, for that matter) include a tense that expresses overlapping of the described eventualities to the temporal perspective point of the sentence (i.e., typically, present or imperfective past). In all three constructions, time-anchoring (to the perspective point) is guaranteed via the interpretation of that verb tense (bolded in the examples below).

The construction with the simple relativized constituent que and tenses expressing overlapping (B) is probably the one most likely found in informal oral speech (to convey the relevant meaning), and is perhaps the one with better acceptance among speakers. Curiously, this construction is often used in automatic translation platforms – like Google Translate, Systran Translate, or DeepL Translator – to translate the equivalent (standard) English constructions. Indeed, though simple predicates of amounts of time are never used in non-relative clauses of the type of (45) (cf. *a Ana e o Pedro são amigos dois anos e meio), using a relativized simple predicate of amounts of time is a (non-standard) strategy often used to sidestep the limitations we have been discussing. In written formal Portuguese, it’s a rarity though, and in my opinion the structure is highly anomalous. I came across only 1 example in the corpus cetempúblico (in a newspaper, but reproducing oral speech) – given in (52) –, and found another example in a translated text (a published Portuguese translation of a Norwegian thriller, Sorgenfri; it is not stated in the publication if the original version was the source of the Portuguese translation) – given in (53).

The construction with the relativized constituent em que and tenses expressing overlapping (C) is particularly odd, in my opinion. I found no examples in the corpus cetempúblico, but I heard one use in Portuguese National TV (Telejornal), from a Brazilian ambassador:

Furthermore, I came across one comparable example in the translation of the same book referred to above in (53), indicating that the Portuguese translator hesitates between strategies:

This example differs, however, in that strict duration, rather than temporal location, seems to be at stake (i.e., this is apparently a construction akin to type A1, with em que_mt, rather than to type A2). Since em is never used instead of há in atelic duration adjuncts, the use of em que with an imperfective past form vivia, is highly anomalous, and sounds truly ungrammatical.

• Relativization with (em) que + tenses expressing anteriority to TPpt + overlapping to TPpt deduced from pragmatic information (construction D)

In contrast with the previous two non-standard possibilities, a strategy needs to be considered, which is fully standard, and I believe is sometimes used by speakers as an alternative to all the three above (A, B, C): it is similar to constructions B and C, but it uses in the relative clause tenses that express anteriority to TPpt, rather than overlapping (i.e., preterite¹² or pluperfect, rather than present or imperfective past, respectively). In this case, the preposition em is optional, though I’d say it frequently occurs (so, for the sake of simplicity, I will conflate these two possibilities – relativized constituent que, or em que – under the single designation Construction D). It is crucial to note that, in this construction D, the fact that the relevant situation still holds at the perspective point, and has not ended before it (i.e., for the examples under scrutiny, that Ana and Pedro are still friends), is not asserted, but is rather deduced from pragmatic or world knowledge information; therefore, there is room for ambiguity (or vagueness) in the sentences illustrating this construction: between the situation holding at the perspective point (time-anchored duration), or not (simple duration). This ambiguity is, for that matter, parallel to the one that exists in English when the present perfect or the past perfect occur without linguistic cues that coerce a time-anchored interpretation (cf. the classic ambiguous examples Sam has been in Boston for 20 minutes, or Mary had lived in Amsterdam for four years). Formally, this construction is identical to those that will be explored in section 3.3 involving relativized non-anchored duration adjuncts. Not all verbs seem to work well in this construction, though (e.g., conhecer ‘to known’ doesn’t¹³), albeit most do.

Sentence (56) can be uttered either in a situation where Ana and Pedro are no longer friends – expressing simple duration – or, despite the preterite foram ‘were’, in a situation where they are still friends, but somehow only the past stretch of their friendship is relevant¹⁴ – formally, simple duration, but corresponding to a time-anchored duration interpretation, like in constructions A2, B and C. The latter interpretation is the one relevant here. The fact that this construction D is a possibility can be easily seen with the following example, that sounds perfectly natural, even though speakers know that Marcelo Rebelo de Sousa is the current President of Portugal:

As for the use of this construction in the corpus cetempúblico: in many examples with the relevant tenses, contextual information clearly indicates that the described situation no longer holds at the TPpt, i.e., that duration is not time-anchored (and therefore not an example of construction D) – cf. the use of ex (ex-refém ‘ex-hostage’) in the following excerpt:

In many other examples, there is ambiguity, in the sense that they are compatible with anchored and non-anchored duration. The example below, for instance, might be used to refer to a situation where the mentioned former mayor of Lisbon (Abecasis) is still the mayor (i.e., anchored duration). Only a wider analysis of the context of the news would offer clarification:

My suspicion is that in most of these contexts the intended meaning is non-anchored duration, i.e., that the construction D is seldom used to refer to a situation that still holds at the TPpt, though this possibility exists. The following example seems to be one where the intended meaning involves time-anchoring (i.e., where the referred famous broker – Caldeira – is still in prison, though only the stretch preceding the TPpt is highlighted, a fact reinforced by the use of já ‘already’) – cf. the equivalent construction using argumental duration, os meses que já leva de prisão preventiva:

• Use of an adjunct with desde (‘since’) + subordinate clause, describing (respectively in constructions E1 and F1) a telic situation marking the beginning of the location time, and an atelic situation overlapping the end of the location time (i.e., overlapping TPpt)

Other structures deserve consideration here. The use of counterparts of since-adverbials applied within definite NPs containing predicates of amounts of time is particularly interesting, since the resulting NPs can convey the same information as those containing the four constructions analysed before, despite not containing a relative clause. See:

In my opinion none of these two structures is ungrammatical (the value of desde being typical and ordinary here), though the second one is hardly ever used (and therefore may sound slightly unnatural), and the first one is infrequent. These assertions are substantiated by searches in the corpus cetempúblico, where only 3 instances of construction E1 were found (cf. (63) below), and no example of construction F1 seems to occur.

Therefore, though construction E1 is grammatical in Portuguese, it may be the case that it does not sound as natural as its English counterpart (the morphosyntactically parallel construction with since being extremely common – cf. excursus below). A closely-related but arguably more natural-sounding construction involves using the desde-sentence as the complement of the verb passar or decorrer ‘elapse’ inside a relative clause, with que representing an argumental predicate of amounts of time (i.e., formally associated with relativization in the context of argumental duration): nos dois anos que passaram desde que o livro foi editado (‘in the two years that have elapsed since the book was published’). English also accepts this combination with a relative clause, instead of a direct application of the since-phrases. The corpus cetempúblico contains at least 7 examples of these structures. Here is one:

• Use of an adjunct with desde (‘since’) + telic NP marking the beginning of the location time (construction E2), or of an adjunct with de (‘of’) + atelic NP covering the entire location time (construction F2)

For the sake of completeness, two constructions need to be added to the previous two constructions with desde. One (that I will refer to as E2) is very similar to E1, except that it does not involve a clausal complement, but rather a nominal eventuality-denoting one, with telic (or telic-like) properties (e.g., an NP with the nucleus início ‘beginning’, or encontro ‘meeting’ – cf. nos dois anos e meio desde o seu primeiro encontro ‘in the two and a half years since their first meeting’).

A second one (that I will refer to as F2) is similar to F1, but also with a nominal eventuality-denoting complement, rather than a clausal one. However, when the whole relevant atelic situation is described by nominal means (e.g., a amizade entre a Ana e o Pedro ‘Ana and Pedro’s friendship’), the preposition desde is not used. Rather, the preposition de (‘of’) precedes the relevant NP – cf. (66).

Constructions E2 and F2 are perfectly grammatical, and can be used as alternatives to the problematic constructions explored up to now (whenever an adequate equivalent nominal expression is available). cetempúblico contains at least 5 examples of construction E2 (i.e., this construction does not seem to be very frequent either) – cf. example in (67); its counterpart with a relative clause with passar or decorrer (‘elapse’), instead of with direct application of the desde-phrase (comparable to (64) above) is more common (20+ examples in the corpus cetempúblico), and therefore may potentially sound more natural – cf. example in (68). As for construction 6B, it is extremely frequent; I provide just one example, in (69):

• Excursus. The expression of time-anchored atelic duration in relative clausesand comparable since-adjuncts in English (vs. Portuguese)

For the sake of comparison – and with the subsidiary aim of contributing to English-Portuguese translation studies –, let us now briefly consider constructions similar to all those observed so far (in 3.2.1) in English, using the British National Corpus for illustration.

When subordinate clauses within the relevant NPs are involved, English constructions come in essentially two forms, both very common. The choice between them seems to be dictated by the telicity of the described eventualities. When an atelic eventuality (still holding at the TPpt) is represented, a situation comparable to Portuguese constructions A-D above, English resorts to relative clauses, with perfect verb forms (e.g., present perfect, as in (70), or past perfect, as in (71)). Observe the following illustrative examples:

As for these constructions, a few notes are in order: (i) the relative pronoun (that) is often omitted in English, but is sometimes explicit – cf. the bolded pronouns in (70); (ii) present perfect and past perfect are used in sequences with present and past perspective points, respectively – cf. (70) vs. (71); (iii) it is possible to use continuous verb forms, or deictic or anaphoric cues (now, then), which coerce a time-anchored duration reading (i.e., preclude ambiguity by excluding a simple duration reading involving a past, bounded, atelic eventuality) – cf. the bolded suffixes and adverbs in (70) and (71).

When a telic eventuality is represented, a situation comparable to Portuguese construction E1 above, English regularly resorts to subordinate since-clauses (with preterite or pluperfect forms), often directly applied to the predicate of amounts of time – as in (72a–b) –, though it is possible (just like in the Portuguese construction (64)) to use the since-clause inside a relative clause involving a relativized argumental predicate of amounts of time, as in (72c).

The similarity between these two groups of constructions – (i) with relative clauses, and atelic descriptions, as in (70)–(71), and (ii) with since-clauses, and telic descriptions, as in (72) – can be evinced by means of paraphrases: for instance, the seven years I have been a secretary, in (70a), is equivalent to the seven years since I became a secretary, and the six years since he joined the company, in (72a), is equivalent to the six years he has been in the company.

Still with respect to English, constructions similar to (70)–(71), on the one hand, and (72), on the other hand, but using nominal complements, rather than clausal ones, are possible, and also common – cf. (73) and (74) below, respectively. They are very similar – in form and meaning – to the constructions E2 and F2 above, respectively, and can easily be translated using parallel strategies.

As said, the translation from English to Portuguese is particularly challenging for the clausal structures involving atelic descriptions, of the type exemplified in (70)–(71). In fact, their more or less (morphosyntactically) parallel counterparts are either anomalous or non-standard (constructions A-C), or do require – unlike in English – extralinguistic information to secure the relevant (time-anchored) interpretation, with the possibility that an undesired (non-time-anchored) interpretation is made (construction D). More unproblematic translations could involve various other strategies: (i) exploring the possibility of expressing the relevant eventuality with a telic description rather than with an atelic one, thus being able to use construction E1 (cf., e.g., translating in the two and a half years Ana and Pedro have been friends by the Portuguese parallel of in the two and a half years since Ana and Pedro became friends, i.e., nos dois anos e meio desde que a Ana e o Pedro se tornaram amigos); (ii) exploring the possibility of expressing the relevant eventuality via nominal phrases rather than clauses, thus being able to use constructions E2 and F2 (nos dois anos e meio desde o início da sua amizade, nos seus dois anos e meio de amizade – cf. (65)–(66) above); (iii) exploring the possibility of expressing the relevant eventuality as an argument, rather than as an adjunct, thus being able to explore the much less restricted relativization or arguments in combination with durative verbs (a possibility mentioned in (44) above, and to be better explored in 3.2.2 below – cf. nos dois anos e meio que já têm/levam de amizade); (iv) interestingly still, despite involving more significant changes, simply avoiding subordination, and resorting to juxtaposition instead – cf., e.g., translating (70a), I have heard of three deaths during rugby matches in the seven years I have been secretary, but nothing like this, by the Portuguese parallel of I have been a secretary for seven years (now); in these seven years, I have heard of three deaths during rugby matches, but nothing like this, i.e., sou secretário há sete anos; nestes sete anos, ouvi falar de três mortes em jogos de râguebi, mas nunca nada assim.

3.2.2. Relativization of arguments expressing time-anchored atelic duration

Let us now turn to constructions similar to those observed in 3.2.1, but resorting to relativized arguments, rather than to relativized adjuncts (or to comparable strategies with desde and de). These constructions are quite frequent in the corpus cetempúblico, and appear to be the most common way of circumventing the grammatical restrictions described above. In these structures, an argument – typically corresponding to a simple predicate of amounts of time (or a time-denoting phrase containing predicates of amounts of time) – is relativized, and occurs with a durative predicate, like (mainly) levar (‘to take’), or ter (‘to have’), inflected in a tense that expresses overlapping of the described eventuality to the relevant TPpt (typically, present or imperfective past).

The counterpart of construction A1 above involving argumental duration is illustrated in the following cetempúblico example, containing a perfectly natural – and common – construction. I add for comparison (preceded by the symbol #) the equivalent sequence using the novel construction with adjunct há que, similar to that in (48).

The counterpart of the (somewhat anomalous) constructions A2, B and C, involving relativized arguments is particularly interesting, since it offers a simple standard alternative to those constructions. Its use is widespread. Here are three examples from cetempúblico with the present form of levar, which illustrate various categorial possibilities in the complements of the verb (clausal, and nominal). I add the corresponding equivalent form of the verb ter (‘to have’) in brackets, when it is possible to use it (last example). I also add, for comparison, the equivalent sequence using the novel construction with the adjunct há que, similar to that in (49) (preceded also by the symbol #):

Now, it is important to note that, with respect to time-anchored argument duration, two subsets of verbs need to be distinguished (as shown in Móia, 2015). The first subset includes verbs like levar and ter, exemplified above: they select simple predicates of amounts of time as complements, and the relativization of these complements poses no issues whatsoever. A second subset includes verbs like durar (‘last’), prolongar-se (literally ‘prolong’ + an intrinsic unanalysable clitic se), and arrastar-se (literally ‘drag’ + an intrinsic unanalysable clitic se), the latter two roughly equivalent to English go on for. This second subset behaves differently, in that it selects predicates of amounts of times preceded by a connective (namely há, or havia), i.e., it selects as complements homonyms of time-anchored duration adjuncts.

The very common verb durar (‘last’) is interesting, inasmuch as it belongs only in the second group in European Portuguese, but it is more flexible in Brazilian Portuguese: it can either be used like in European Portuguese, with the connective há (as in (79)), or it can take simple predicates of amounts of time (without any connective) as complements – (79) being equivalent (only in BP) to a guerra já dura vinte anos (cf. Móia, 2020, pp. 69–71). As a matter of fact, this latter use – without há – is much more common in Brazilian newspaper registers (the sequence durar há only occurring twice in the Brazilian corpus NILC/São Carlos).

I did not find any corpora examples of relativization of arguments with há que (in combination with these durative verbs), though this construction – illustrated in (80), for the relativization of a genuine predicate of amounts of time, and in (81), for the relativization of a comparable time-denoting expression – seems as much acceptable as their counterparts with adjunct há que ((48) and (49) above, respectively).

I believe the absence of examples like (80) and (81) is more related with the connective há’s general reluctance to occur before relative pronouns (as a full-fledged preposition would) than with the argument (vs. adjunct) character of the relativized constituent. In other words, I believe the restrictions are more of morphosyntactic nature, having to do with the unfinished grammaticalization process of há, than of semantic nature. The fact that durar (‘last’) is apparently compatible with relativized constituents (of the form que) in Brazilian Portuguese aligns with this conjecture.¹⁶ Though I could only find 3 examples of this structure in the 1+ billion-word Corpus Brasileiro – cf. examples (82) (comparable to (80)), and (83) (comparable to (81)) –, which shows that the structure is not frequently used, various Brazilian speakers consulted said they accept them without problems. Recall that, in BP, durar can belong in the same group of levar and ter, taking simple predicates of amounts of time (which unrestrictedly accept relativization – cf. (75)–(78)) as complements when expressing time-anchored duration.

I found a construction of the subtype in (82) in a translated text published in Portugal (which is anomalous in my grammatical judgement as a native speaker of EP), indicating that this is clearly an area of intense language variation (and hesitation) in contemporary Portuguese: