2.1 Aspect and its relation to the progressive form

Aspect, or viewpoint/outer aspect, provides information about the perspective the speaker takes with respect to an event or a state. In many languages (e.g. the Romance languages), a morphological distinction is made between imperfective aspect and perfective aspect.Footnote ⁵ Perfective aspect is used to indicate that the event is seen as bounded, or as a whole (see e.g. Comrie Reference Comrie1976, Giorgi & Pianesi Reference Giorgi and Pianesi1997), and imperfective is used to indicate that only a sub-part of the event is focused on. The focus in this article is on the imperfective aspect, since this is where we find a difference between Swedish and English, but the perfective aspect will be discussed as well. Even though aspect is not overtly marked in Swedish and English, I will assume that every sentence has an aspect value, either perfective or imperfective (see discussion in the concluding section below).

The imperfective can be split into two sub-categories: progressive and non-progressive, where the non-progressive covers habitual, generic and possibly iterative aspect (see e.g. Filip Reference Filip, Tenny and Pustejovsky2000). In this paper I will use progressive in small capitals for the semantic progressive imperfective aspect, and the term progressive form for the English progressive form, i.e. be V-ing. I will assume that what is common for the sub-types of imperfective aspect is that the ast-t is located inside the Event Time. I will assume that habituals and generics contain derived states (see Parsons Reference Parsons1990), in which ast-t is located in the default case. I will not go into the technical detail of how generic and habitual states are derived here, but only point out that in English, the simple form is crucially used for non-progressive imperfective aspect, as shown in (8).

1. (8)

In the progressive imperfective aspect, ast-t is located inside an ongoing event or a (non-generic, non-habitual) state. As has been suggested earlier, and as will be argued for in great detail below, ast-t in English is punctual, and can thus only be located inside predicates that are true at moments, i.e. states (or derived states) in accordance with the first of Taylor's meaning postulates set out in the introduction above. In Swedish, ast-t cannot be located inside a telic predicate (i.e. an accomplishment or an achievement), since telic predicates will not be true for any intervals inside the event, but only for the whole event, in accordance with the third of Taylor's meaning postulates.

Note that progressive imperfective aspect should not be confused with the English progressive form, which actually can be perfective in certain contexts. The progressive imperfective aspect can be characterized as an event going on, or a state holding, where the endpoint and the starting point is of no relevance. The focus is only on a sub-interval (or a point in time) of a larger event or state. Technically it can be characterized as ast-t located inside ev-t. In English, the progressive form has to be used for dynamic verbs to get this reading, as in (9a), while the simple form is used for stative verbs, as in (9b). In both these examples, no assertion is made about the full event or state described in the main clause, but only about a small portion of the event that is somehow related to the event described in the modifying embedded temporal clause.Footnote ⁶

1. (9)

   1. a. John was writing a letter when I entered the room.

   2. b. When I came back home, my brother owned the company.

In Swedish, the progressive imperfective reading is available for atelic dynamic verbs as well as for states in the simple tenses, as illustrated in (10b) ((10a) is taken from Platzack Reference Platzack1979).

1. (10)

We can tell from the choice of preposition in (10a) that the sentence is atelic (i.e. i instead of pa˚).Footnote ⁷ The main clause in (10b) has a progressive interpretation: we make an assertion about a sub-part of the whole event, and this sub-part is temporally related to the event described in the embedded clause. With a telic predicate, as in (11) (‘fly to X’), the progressive interpretation is not available, making (11b) semantically odd; the only interpretation available is a sequential reading: the bird was shot first and then it flew to my house. That is, we have to see the two events as two completed events.

1. (11)

Again, note that the telicity-restriction in Swedish also follows from Taylor's postulates: an accomplishment like ‘fly to my house’ is only true when the whole event is taken into account, in this case, the stork's flying from some contextually salient location to my house, and can crucially not be true at any sub-intervals of the event.

In English, even though the progressive form has to be used for dynamic verbs to express progressive aspect, it is not the case that the progressive form always gives rise to progressive imperfective aspect. The progressive can also be used in bounded contexts, as in (12) where the adverbials ‘for X hours’ and ‘from A to B’ yield bounded readings, i.e. in the examples in (12), we make an assertion about the whole event. Here, the simple form can be used as well. Lexical states can be bounded as well, as is shown in (12c).

1. (12)

   1. a. Yesterday it rained/was raining for two hours.

   2. b. Yesterday I worked/was working from 9 to 5.

   3. c. He owned the company from 1984 to 1989.

A temporal for-phrase thus modifies atelic predicates, but it provides boundaries for the event. The for-phrase and the verb create a complex predicate with the same temporal properties as a telic verb, i.e. a predicate that denotes an event that is not true for any sub-intervals of the whole event: the event of ‘raining for two hours’ is not true for any sub-intervals of the whole event. Just like a telic event, a verb modified by a temporal for-phrase will get a perfective interpretation unless it is within the scope of a higher stativizing operator, like a habitual or generic operator. Hence, predicates modified by ‘for X hours’ and ‘from A to B’ need to surface in the perfective in the Romance languages in non-generic/non-habitual contexts, as exemplified for Spanish in (13) and (14).Footnote ⁸

1. (13)

2. (14)

As we see, English progressive predicates can be bounded in contrast to Romance imperfectives.Footnote ⁹ In other words, the progressive form in English does not unambiguously signal be unboundedness, or indicate that only a sub-part of the event is in focus. Note also that many of the Romance languages have a periphrastic progressive in addition to the imperfective. The copula in the progressive constructions can be marked either imperfective or perfective (or pretérito, as it is called in Spanish), depending on the boundedness of the event, which shows that the progressive–simple distinction is largely independent of the perfective–imperfective distinction:

1. (15)

This aspect distinction is not expressed in English, which again shows that the progressive is not an imperfective per se. Rather, the progressive, just like states in general, can have a perfective interpretation, or an imperfective interpretation (most commonly a progressive imperfective).Footnote ¹⁰ Stativity is a prerequisite for progressive imperfective aspect in English. The progressive operator stativizes a non-stative predicate, and the derived stative predicate can be assigned a progressive imperfective value. A non-stative predicate could not be assigned a progressive imperfective value, due to the nature of ast-t in English.

One might still want to call the English progressive form a type of ‘aspect’, but it must be remembered that the progressive form does not determine the boundedness, or the viewpoint taken of the event. Rather, it creates an in-progress state (Parsons Reference Parsons1990), i.e. it changes the aktionsart of the predicate, and thus has more in common with, for example, endpoint-inducing verb particles and endpoint-demoting conative constructions than the aspect marking we see in, for example, the Romance languages. The progressive, together with other aktionsart modifiers like, for example, phasal verbs have to be located below the viewpoint aspect projection, in something that we can call the extended verb phrase. The head of the extended verb phrase forms a relation with viewpoint aspect, giving the following functional hierarchy:Footnote ¹¹

1. (16) Proposed functional hierarchy

2. Tense [(Viewpoint) Aspect [Extended VP (Progressive, phasal verbs, etc.) [VP/Lexical Aspect]]]

Dynamic verbs in the simple form can have non-progressive imperfective interpretations, as we have already seen in example (8), but most commonly a simple dynamic verb receives a perfective interpretation. Whereas imperfective aspect is characterized as having ast-t inside ev-t, perfective aspect can be characterized as ast-t being identical to ev-t, as in Demirdache & Uribe-Etxebarria (Reference Demirdache, Uribe-Etxebarria, Martin, Michaels and Uriagereka2000). If this characterization is correct, it simply cannot be true that ast-t is always punctual in English, as the event time in typical perfective predicates, as in (17), is clearly not a point in time:

1. (17)

   1. a. We built the house in five years.

   2. b. He ran for three hours.

If Demirdache & Uribe-Etxebarria (Reference Demirdache, Uribe-Etxebarria, Martin, Michaels and Uriagereka2000) are right, we have to conclude that ast-t in English can be either a point in time or a longer interval. I will assume that ast-t can be an interval in English, but only when it is identical to the event time. The ast-t in English could thus have two possible values: it can either be punctual (default) or be determined by the ev-t.Footnote ¹² This will rule out progressive interpretations of dynamic predicates in English: a progressive interpretation of an dynamic predicate requires a non-punctual ast-t, but ast-t in English can only be non-punctual when it is identical to the ev-t. When ast-t and ev-t are identical, an assertion is made about the whole event and, crucially, not a sub-part of it.

The simplest way of characterizing viewpoint aspect in Swedish and English is to assume that the aspect head can carry either the value identity or within:

• • ast-t is identical to ev-t = Perfective aspect

• • ast-t is within ev-t = Imperfective aspect

The perfective aspect can thus be seen as an anaphoric aspect: ast-t and ev-t are co-referential.Footnote ¹³ The value of ast-t is determined by ev-t. The imperfective aspect can be seen as a pronominal or referential aspect: the category Asp introduces a new temporal entity, disjoint from ev-t. Above I have defined imperfective aspect as ast-t within ev-t, but it would also be possible define imperfective aspect as ‘ast-t not identical to ev-t’. One would then have to give a pragmatic account for the within-relation, for example by stating that it would be uninformative to make an assertion about a time where no event is going on, or no state holds. Note, however, that Reichenbach (Reference Reichenbach1947), Giorgi & Pianesi (Reference Giorgi and Pianesi1997), Demirdache & Uribe-Etxebarria (Reference Demirdache, Uribe-Etxebarria, Martin, Michaels and Uriagereka2000) have claimed that the perfect locates ast-t after ev-t. If this is true, ‘non-identity’ would not be a sufficient definition of imperfective aspect. There are, however, several alternatives to the Reichenbachian analysis of the perfect that are compatible with the ‘non-identity’ analysis, for example, an analysis of the perfect as a derived resultant state, in which the ast-t could be located. For clarity of the argumentation in this paper, I will keep the definition above that imperfective aspect locates ast-t within ev-t. It is, however, possible that the only function of Asp is to introduce a temporal argument (or variable) that is either co-referent with or disjoint from ev-t. The aspectual values ‘identity’ and ‘within’ could follow solely from the value of the index: an anaphoric ast-t would of course give rise to an identity relation, and a referential/pronominal ast-t could plausibly give rise to a within-relation.

If we follow Stowell's idea that ast-t is represented in the syntax as a temporal argument (ZP), we can state the difference between English and Swedish as follows:

1. (18) A non-anaphoric ZP in English can only denote a point in time, while a non-anaphoric ZP in Swedish can denote either a point in time or an interval.

In this account, the difference between the two languages is thus only in the value of a zero pronominal element. This would not be an unexpected locus of variation, as we know that languages may differ when it comes to possible values of pro for example (e.g. Italian vs. Icelandic).Footnote ¹⁴

If, on the other hand, we choose to represent ast-t as an index on Asp, we have to state that the a new index introduced by ast-t in English can only be interpreted as a point in time. I leave it up to the reader to decide which technical solution is the best one. In the rest of the article I will simply talk about progressive imperfective aspect as ‘ast-t within ev-t’, and perfective aspect as ‘ast-t = ev-t’, without reference to the exact syntactic structures.

The definition of the perfective above might need some reformulation to hold cross-linguistically. As discussed in e.g. Ramchand (Reference Ramchand and Svenonius2004) and Romanova (Reference Romanova2006), perfective predicates in Russian (and other Slavic languages) are able to pick out certain sub-parts of the event, for example, a starting point, or possibly an endpoint. In these cases, a temporal relation seems to be established between ast-t and a certain sub-event within the macro-event. We can still view this as an identity relation, only that the identity is not between the time of assertion and the whole event time, but rather a sub-event.Footnote ¹⁵ The choice of perfectivizing prefix will determine which sub-event an assertion is made about. In other words, it seems to be possible for a language to make fine-grained distinctions within the perfective, but even in these cases, we are dealing with an identity relation.Footnote ¹⁶

From the discussion of English above we can conclude that dynamic predicates in the simple form, stative predicates in the simple form and predicates in the progressive form can have either perfective or imperfective interpretations, at least in the Romance sense. There is only one reading missing in English: progressive readings of simple non-stative verbs. If we follow Taylor's meaning postulates as was given in the introduction, the obvious conclusion is that ast-t in English is always punctual unless it is identical to ev-t, and thus cannot meaningfully denote a sub-part of an event that is not true at moments. That is, it should not be possible to talk about a point in time of an event named by a predicate that is only true at sub-intervals larger than a moment, since by definition, the predicate in itself cannot hold at a point in time. It is possible though to derive states from non-states. In English, this can be done with the help of, for example, the progressive, which derives in-progress states from non-states (see Parsons Reference Parsons1990). As will be further argued below, predicates in the progressive form in English, always behave like stative predicates, and not like ‘imperfective’ predicates. As has been mentioned above, and as will be carefully argued for below, both states and activities can get a progressive interpretation in Swedish, while telic predicates cannot. From this we can conclude that ast-t can be longer than a moment in Swedish, even in cases where it is not identical to ev-t, and that there is no phonologically null progressive head in Swedish (that would create states or atelic predicates from telic predicates).

It should be noted that a sentence with an activity verb in the progressive imperfective aspect means something slightly different in Swedish and English. The Swedish sentence in (19) does not really mean the same thing as the English sentence in the translation line:

1. (19)

In the Swedish sentence, ast-t picks out one or several minimal stages of a football-playing event. In English, the progressive has turned the activity predicate into an in-progress state, possibly by erasing boundaries of the minimal stages of the event, and the ast-t picks out one moment of the in-progress state. The Swedish sentence and its English counterpart have the same aspect value, i.e. ast-t is located within ev-t, but they still mean something slightly different. This difference in meaning is parallel to the difference between a perfective activity verb in the simple form and the progressive form in English, as in (20):

1. (20)

   1. a. It was raining for two hours.

   2. b. It rained for two hours.

Both sentences above are perfective, but whereas (20a) is a perfective in-progress state, (20b) is a perfective activity event.

3.1 Past tense and AST-T modifiers

The first context I will look at is past tense sentences containing when-clauses. In English, the interpretation of the temporal relation between the when-clause and the main clause crucially depends on the nature of the events in the two clauses.Footnote ¹⁷ The event denoted in the when-clause is most naturally to be interpreted as taking place within the time of the event/state when the main clause predicate is a lexical state or a derived state (here, the progressive is the derived state that is relevant), while the event in the when-clause is interpreted as preceding the event time of the predicate when the main clause is non-stative. The telicity of the event is of no relevance in English. I will for now label the two different relations as ‘Overlap’ and ‘Sequential’.

1. (21)

In Swedish, both dynamic and stative atelic predicates in the main clause can get an overlap-interpretation, while telic predicates get a sequential reading, as shown in (22):

1. (22)

This section will answer two questions: (i) Why does the state/dynamic distinction give rise to the two interpretations (sequential vs. overlap) in English? and (ii) Why is it the the atelic–telic distinction rather than stative–dynamic distinction relevant in this context in Swedish. First, however, a note about temporal adverbs is in order. I will assume, following Demirdache & Uribe-Etxebarria (Reference Demirdache, Uribe-Etxebarria, Guéron and Lecarme2004), that temporal adverbials like when-clauses do not directly give the ast-t of the main clause, but rather relate the ast-t to the point in time or interval given by the adverbial clause. The relation between the time interval or point given by the adverb is not necessarily one of identity, at least not for a when-clause that triggers a sequential reading. Temporal identity is not imposed by adverbs like last year or yesterday either, as illustrated in the following two examples, one with a simple eventive verb, and one with a progressive verb:

1. (23)

   1. a. I wrote a letter yesterday.

   2. b. Q: How come you didn't see any movies last year?

   3. A: Well, I was writing my dissertation last year.

In (23a), the letter-writing did not necessarily take a whole day. We assume that the assertion time is identical to the event time in (23a), i.e. we make an assertion about an interval that lasts exactly as long as the letter-writing takes. The function of the temporal adverbial is only to locate the ast-t within the time interval denoted by yesterday. Crucially, identity is not required. The example in (23b) is interesting, since the main clause is a (derived) state. Here, a temporal identity between the state and the time span is at least pragmatically plausible, but is not required. The event/state denoted by the main predicate might very well have extended last year (i.e. the writing might have started two years ago, and might have continued until this year). The event/state might have also be shorter than the whole year. The subject might have either finished the dissertation within last year, or just stopped writing it at some point. The ast-t in (23b) is thus not identical to the ev-t, i.e. we are not making an assertion about the full in-progress state of writing a dissertation. It is rather a point in time, located somewhere within the derived state. The ast-t is further located somewhere within time interval denoted by the temporal adverb, i.e. last year. The assertion in (23b) is thus only that there is a point in time located within last year, where the subject was in the in-progress state of writing his dissertation. Our world-knowledge tells us that the duration of the event presumably was fairly long, and since the dissertation-writing is given as the main argument for not seeing any movies last year, we can conclude that the dissertation-writing probably covered most or all of last year. But this is not asserted, only filled in by our world-knowledge and pragmatics. If the simple past tense is used in (23b) (e.g. I wrote my dissertation last year), the whole event must have taken place within last year, since the assertion time now is identical to the event time, and the assertion time/event time is located within last year.

Other temporal modifiers directly specify ev-t, for example the durational adverbials for x time and in x time.Footnote ¹⁸ As has already been mentioned, a for-phrase specifies the time span of an atelic predicate, while an in-phrase specifies the time span of a telic predicate. An atelic predicate with a for-phrase modifier behaves like a telic predicate, which is predicted from Taylor's meaning postulates: there is no sub-interval of the predicate run for five hours for which the predicate run for five hours is true.

The temporal relation between the time denoted by the adverbial and the ast-t (or possibly ev-t) of the main clause can be specified in various ways, for example, before or after. The relation between the time adverbial and the main clause predicate is in these cases quite transparent, and Swedish and English do not show any unpredictable differences here. The semantics of when is less transparent. When relates the ast-t of the embedded predicate to the ast-t of the main clause. In the sequential reading of when, we seem to get a reading that is similar to after, i.e. the event in the main clause takes place after the event in the when-clause takes place, but the two assertion times are much more directly connected when a when-clause is used compared to an after-clause. In (21c) and (21d), both the main predicate and the embedded predicate are perfective, that is, both predicates are non-stative, which forces an identity relation between ast-t and ev-t in each clause. Since we have a sequential reading, rather than an overlapping or simultaneous reading, we must conclude that when established neither an identity relation nor a within-relation. Rather, a temporal while-phrase has to be used to give overlapping or simultaneous reading.Footnote ¹⁹ Let us assume the following semantics for when:

We can assume this definition for the overlapping cases in (21) and (22) as well. When the main predicate is stative, the ast-t is punctual, and it is located somewhere inside ev-t. If the predicate inside the when-clause is eventive, the ast-t is an interval which is temporally identical to the ev-t. When locates the ast-t in its complement just before and abutting the ast-t of the main clause. The ev-t of the main clause presumably spans a much longer interval, starting well before the event in the when-clause. This gives rise to the Overlap interpretation.Footnote ²⁰ The definition of English when can be applied to Swedish när ‘when’ also. The difference we see between activity verbs in English and Swedish is not triggered by differences between English when and Swedish när, but in the nature of the imperfective topic time.

Modification by a when-clause can thus be seen as a test for (im)perfective aspect: an Overlap interpretation is only possible when the main clause predicate is imperfective, i.e. when the ast-t is within the EV-T, and a Sequential interpretation is forced when the main clause predicate is perfective, i.e. when the ast-t is identical to the whole event. In English, only states can be assigned an imperfective value, and therefore an Overlap interpretation is only possible when one of the predicates is stative. In Swedish, both states and activities can be assigned an imperfective value, which makes overlap interpretations possible for both states and activities.

Sequential readings seem to be unavailable for both lexical states and progressives (the sequential reading is unavailable for the progressive in (21b)), as shown for states in (24):

1. (24)

The coerced reading ‘start to love/own’ seems to be unavailable in (24), which is unexpected, given that we have argued that states can be perfective.Footnote ²¹ It should, however, be noted that sequential readings are unavailable for many stative verbs in languages with marked viewpoint aspect as well, as can be seen in the Spanish examples (25a, b). Note, however, that the same Spanish predicate can carry perfective marking in other contexts, as in (26a, b).

1. (25)

2. (26)

Presumably, starting points of states are not salient enough to be picked up by a when-clause. However, a starting point, and an endpoint, can be imposed on states with a delimiting adverbial (such as for X time), yielding a perfective interpretation (with perfective marking as a consequence).Footnote ²²

The perfective and the imperfective aspect can also be teased apart with the adverb still.Footnote ²³ The adverb still adds an implication that the event was going on before ast-t, and that it is still going at ast-t. By adding still, we therefore make sure that ast-t is included in ev-t. Still is thus incompatible with a sequential relation between a when-clause and the main clause, and compatible with an overlap relation, as shown in (27). Throughout the paper, I will use the adverb still as a test for imperfective aspect.

1. (27)

For Swedish, as we have seen, clearly telic predicates get a sequential reading when modified by a punctual when-clause. As expected, the Swedish equivalent of the adverb still is not licensed in this context, as shown in (28):

1. (28)

The overlap reading is available as soon as the predicate is atelic, as in the three contexts below: the conative version of (28) in (29a), a predicate with a non-quantized object in (29b) and a stative predicate in (29c). As shown, the adverb fortfarande ‘still’ is felicitous in these contexts, and there is no sequential reading:

1. (29)

The contrast between (28) and (29) clearly shows that only atelic predicates can get progressive imperfective interpretations. That is, ast-t can only be located inside a predicate that is true at sub-intervals of the event.

One possibly problematic fact for the claimed correlation between telicity and ast-t placement is presented by Tonne (Reference Tonne2007). She notes that both (30a) and (30b) below can receive a progressive reading, i.e. a reading where ast-t is located inside the event time (Norwegian example from Tonne (Reference Tonne2007:195), the corresponding Swedish sentences behave like the Norwegian examples):

1. (30)

The example in (30b) has an indefinite quantized object, and we would expect that the quantized object would make the predicate telic, and therefore non-progressive. However, the verb read is well-known for allowing both telic and atelic interpretations with quantized objects, in both English and the Scandinavian languages, as exemplified here for English:

1. (31)

   1. a. He read a book for an hour/in an hour.

   2. b. I read the newspaper for an hour/in an hour yesterday.

This shows that the correlation between (a)telicity and progressive interpretation still holds. Tonne's example just shows that certain typical accomplishment verbs can behave more like degree achievements in certain context (compare a typical degree achievement, e.g. The hole widened for an hour/in an hour).Footnote ²⁴

Summing up, once carefully investigated, we see that telicity restricts the availability of progressive readings of the simple past in Swedish in the same way as dynamicity restricts the availability of progressive readings in English.

3.2 The universal perfect

In the Reichenbachian system, the perfect realizes a temporal structure where the ast-t/Reference Time is located after the Event Time. However, when it comes to the universal perfect, the ast-t actually needs to be included in the Event Time, as illustrated in (32):

1. (32) I have been sitting here ever since 5 o'clock.

The sitting has to take place at ast-t, which is here identical to Speech Time (since it is a present perfect rather than a past perfect), for (32) to be felicitous. In other words, ast-t can be included in the ev-t in perfects, just as in the present (progressive) and past (progressive) tense.Footnote ²⁵ It should be noted that many languages use the present tense in contexts where the universal perfect has to be used in Swedish and English. This is true both of languages that lack a special present perfect tense form, i.e. Japanese and Hindi (Naoyuki Yamoto and Rajesh Bhatt, p.c.), and of some languages that have a special perfect tense form, like Dutch and Italian, where the perfect tense in general cannot express the universal perfect (Marleen van der Vate and Irene Franco, p.c.).Footnote ²⁶ I will therefore assume that the location of ev-t with respect to ast-t in the perfect tenses is partly underspecified. Thus, the semantic interpretation of the perfect cannot be that of establishing a specific relation between ast-t and ev-t. It is more plausible that the perfect functions to either introduce a so-called perfect time span (see e.g. Mittwoch Reference Mittwoch1988, Iatridou, Anagnastopulou & Izvorski Reference Iatridou, Anagnostopulou, Izvorski and Kenstowicz2001) or create a higher level resultant state or target state (see Parsons Reference Parsons1990, Katz Reference Katz, Alexiadou, Rathert and Von Stechow2003), though the exact characterization of the perfect tense is of no importance for this article. What is important is rather that we see the same difference between English and Swedish in the perfect tense as we see in the past tense.

In English, just like in the present and past tense, non-stative predicates need to carry progressive marking in the universal perfect. It does not make any difference if the predicate is an activity, as in (33a), or an accomplishment, as in (33b). States, as usual, surface in their simple form, as shown in (33c):Footnote ²⁷,Footnote ²⁸

1. (33)

   1. a. He has been running/#run ever since five o'clock this morning.

   2. b. He has been writing/#written the paper ever since five o'clock this morning.

   3. c. He has lived here ever since 1985.

Note that we see the same effect in the past perfect (i.e. pluperfect), which tells us that the effect is not triggered by the nature of Utterance Time/Speech Time, but rather ast-t:

1. (34) When we came to soothe him, he had been crying/#cried ever since his parents left him.

In Swedish, the telicity effect is very strong in the universal perfect, as seen in the contrast between an activity, in (35a), and an accomplishment, in (35b):

1. (35)

States pattern with activity verbs, as expected:

1. (36)

Telic predicates can of course occur in the perfect tense, but only with the interpretation that the whole event has been completed prior to ast-t (as in I have already written the letter). Note that this reading is not available with the temporal modifier ever since (see Mittwoch Reference Mittwoch1988 for discussion).

3.3 Episodic present tense

Let us assume that ast-t and Speech Time are co-temporaneous in the episodic present tense. That is, in the present tense, we make an assertion about an event taking place or a state holding at Speech Time. In the imperfective aspect in English, we would thus have a punctual ast-t that is located within ev-t, and the Speech Time would be identical to ast-t, as schematized in (37):

1. (37) Imperfective present tense

2. sp-t = ast-t, ast-t within ev-t

The progressive form has to be used for all non-stative verbs in order to get an episodic present tense reading, as shown in (38). As we also see, the adverb still is felicitous here, showing that ast-t really is located inside ev-t:

1. (38)

If a non-eventive verb is used in the simple present tense form, a generic or habitual reading is the most salient one. We assume that generic predicates have a phonologically null stativizer. Thus, a generic sentence like the one in (39) has the same tense-aspect structure as the other imperfective present tenses above. The difference is that a different stativizer is used in the generic than in the progressive, and the generic stativizer is not pronounced:

1. (39)

If we assume that Speech Time is a point in time, we can rule out a present perfective: in a perfective predicate ast-t is always identical to the whole event, and the whole event is longer than a point in time (at least it requires a transition from one state to another). An identity relation could not possibly hold between a point in time and a time span that is longer than a point in time. However, there are reasons to suspect that Speech Time can have duration in some cases. If Speech Time has duration, the perfective present tense would have the following structure:

1. (40) Perfective present tense

2. sp-t = ast-t, ast-t = ev-t

A perfective present tense could only pick out an event that takes place as the speaker is speaking. This is pragmatically odd, but this is presumably the right characterization for so-called reportive uses of the present tense, as discussed in Parsons (Reference Parsons1990:Chapters 12 and 13). The reportive present tense is exemplified in (41) (from Parsons Reference Parsons1990:30):

1. (41) And the Maryland delegation goes two to one for the democrats!

   (Uttered by a newscaster)

Another instance where the event takes place in the same time as the utterance is with so-called performative predicates like promise, as discussed by Austin (Reference Austin1962) and exemplified in (42):

1. (42) I hereby pronounce you husband and wife.

We have thus no reasons to rule out a perfective present tense. Rather, we have to state that the present perfective is limited to certain contexts, where the whole event actually takes place at Speech Time.Footnote ²⁹ If we were to assume that the present tense in English established a within-relation between Speech Time and ast-t rather than an identity relation, we would expect fewer restrictions on the perfective present tense, i.e. we would expect to be able to use it to talk about events that have started and will culminate sometime in the future.

Whereas the stative–dynamic distinction is very clearly seen in the English present tense, the telicity effect is much more subtle in Swedish, and superficially it seems that even clearly telic events can surface in the present tense with ongoing readings, as can be seen in (43a), where a present tense telic sentence includes the adverb nu ‘now’. Example (43b) shows that the verb phrase really is telic.

1. (43)

It is tempting to analyse (43a) as an imperfective progressive present tense:

1. (44) Progressive present tense

2. sp-t = ast-t, ast-t within ev-t

If this is the right analysis, we have to give up the claim that assertion time cannot be located inside telic events in Swedish. However, we have established above (Section 3.1) that fortfarande ‘still’ is a test for imperfective aspect and, as shown in (45), fortfarande is not licit in the present tense when the predicate is telic (unless we force a generic/habitual reading):

1. (45)

This shows that ast-t is not located inside ev-t in (43a), and that we need another way to account for the presence of the adverb nu in the example. First, however, note that fortfarande is licit in the present tense contexts with all types of atelic verbs (even in non-generic contexts), that is, present progressive readings are available. This is shown here for the conative version of (43a) in (46a), a simple activity verb in (46b) and a stative verb in (46c):

1. (46)

From this we can conclude that the restriction on having ast-t inside a telic event holds in the present tense as well, i.e. in Swedish we see the same telicity-sensitivity in the present tense as we see in the simple past and the perfect. This is of course expected: we would not expect that tense restricts the way ast-t and ev-t are connected.

We must, however, conclude that the present tense in English and Swedish are different. As we saw above, by assuming that the present tense in English always establishes an identity relation between Speech Time and ast-t, we could account for the available readings of the present tense.

In Swedish, an identity relation will not capture the possible readings of the present tense. The Swedish sentence in (43a) above is not a reportive reading. The most salient reading of (43a) is that the event starts at Speech Time, or even right after Speech Time. It is thus presumably an instance of the future use of the present tense, as illustrated here:

1. (47)

Note that the future use of the simple present tense is not available in English (see e.g. Giorgi & Pianesi Reference Giorgi and Pianesi1997 for discussion). The Swedish present can thus be defined in one of the following ways:

1. (48)

   1. a. Speech Time is not after ast-t (i.e. Speech Time is within, before or identical to ast-t).

   2. b. Speech Time is identical to or before ast-t.

To decide between (48a) and (48b), we have to ask whether a within-relation can hold between Speech Time and ast-t, i.e. whether we can make an assertion about an event that started in the past and will end in the future, when using the simple present tense. The judgements are tricky here, but it seems that time span adverbs cannot pick out an interval that starts before Speech Time and ends after Speech Time, but only an interval that is entirely located after Speech Time. This is illustrated in (49), where the follow-up clause is odd since it implies that the event started before Speech Time:Footnote ³⁰

1. (49)

Note that (49) also has a coerced generic reading, which is available in the English simple present tense as well, though this reading is more felicitous with another type of DP, for example, I write this type of letter in 30 minutes, so I should be done in 15 minutes (since I started 15 minutes ago). Note that (49) is fine without the follow-up clause, in a near-future (or starting now) reading, in a context like ‘I will write this letter in 30 minutes, and then I'll join you guys for a beer in the pub’. Let us then conclude that Speech Time in the Swedish present tense is either identical to ast-t or located before ast-t. In perfective aspect, i.e. contexts where ast-t is identical to ev-t, we are restricted to a reportive, performative or future-shifted interpretation. In the imperfective aspect, i.e. when ast-t is within ev-t, we get an ongoing present tense interpretation, as well as a future interpretation. In the ongoing present tense, we could assume that Speech Time is identical to ast-t, and thus have some duration. The Swedish tense-aspect system thus seems to be similar to the tense-aspect systems in Slavic languages, though aspect is morphologically marked in Slavic, but not Swedish. As is well-known, in the Slavic languages where present tense verb can carry perfective or imperfective marking, perfective forms trigger a future-shifted reading (though performative/reportive uses are also available, as in Swedish), while imperfective verbs are interpreted as progressive (or habitual), as is shown in (50) for Russian (examples from Eugenia Romanova, p.c.):

1. (50)

Summing up, the present tense in Swedish can surface in four different temporal configurations, while in English it can surface only in two. The different configurations are given in the following list:

1. (i) Perfective present tense: sp-t = ast-t, ast-t = ev-t, English and Swedish. This is the structure for the reportive present tense and performative speech acts.

2. (ii) Imperfective present tense: sp-t = ast-t, ast-t within ev-t, English and Swedish. In English, this is the structure for lexical states and derived states (generic and in-progress states) in the present tense. In Swedish, this is the structure for present tense interpretations of all atelic predicates, including states, processes and generics.

3. (iii) Perfective future tense: sp-t before ast-t, ast-t = ev-t, Swedish only. This is the structure for telic predicates in the present tense when they are not used as reportive, performative or generic predicates.

4. (iv) Imperfective future tense: sp-t before ast-t, ast-t within ev-t, Swedish only. Not discussed above, the future interpretation is also available for imperfective predicates. This is predicted, since we do not expect that the tense value should restrict the possible aspect values. An example of an imperfective future tense is given in (51).

1. (51)

It is beyond the scope of this paper to discuss the exact relation between the two different tense values that can be spelled out by the present tense marker in Swedish, i.e. ‘before’ and ‘identity’. The present tense might just be vague, or underspecified; for example, the only information that the present tense marker provides could be that no part of ast-t precedes speech time. But the present tense might also be genuinely ambiguous, for example, it could realize a tense node with either the value ‘before’ or ‘identity’. This option is assumed in the list above. A third option is that present tense always imposes a strict identity relation between speech time and ast-t, even in Swedish, and that the future interpretation has its origin in a covert modal verb (realized by will in English). I will leave this for future investigation.Footnote ³¹

It is of course tempting to draw parallels between the aspect-restrictions and tense-restrictions in English and Swedish. The difference between the Swedish present tense and the English present tense is only seen when there is a non–co-referent relation between ast-t and Speech Time, just like we only see a difference between English and Swedish in the aspectual domain when ast-t is not co-referent with ev-t. However, the difference in tense between English and Swedish does not seem to be located in the nature of Speech Time, but rather in the possible values of tense (e.g. before, after, etc.).Footnote ³²

3.4 Complements of modal verbs

The aktionsart and aspect of a verb embedded under a modal verb influence the interpretation of the modal verb in subtle and intricate ways (see e.g. Eide Reference Eide2005 for a detailed discussion). In this paper I am only interested in one question regarding this interaction between modality and aspect/aktionsart: When is it possible to get a present tense, progressive interpretation of a verb embedded under an epistemic modal?Footnote ³³ That is, when is it possible to get Speech Time and ast-t inside Event Time, in the context of a modal? The structure I am interested in is given in (52):

1. (52) Modal [Speech Time = ast-t, ast-t within ev-t]

The purpose of this section is not to lay out the intricate interactions between modality and tense/aspect but rather to illustrate that the state-sensitivity holds for infinitives in English as well, and further to show that some Swedish modal verbs show state-sensitivity, just like their English counterparts. The second point is important, as it shows that the distinction between states and activities is syntactically relevant in Swedish. This point is also made in Section 3.5 below.

In English, non-stative predicates need to carry progressive marking to get a present tense ongoing reading ((53a) and (54a)). Simple infinitives get a future-shifted reading, usually with deontic force ((53b) and (54b)). The telicity of the main predicate is of no importance (the examples in (53) and (54) are based on examples in Ramchand Reference Ramchand2012).

1. (53)

2. (54)

The sentences in (53a) and (54a) above could both be the answer to the question What do you think John is doing now? while the sentences in (53b) and (54b) could not. Stative verbs can surface in the simple infinitival form:

1. (55) John must own half the company now. (Epistemic, present tense)

The Speech Time and ast-t are presumably identical in the sentences above, and the ast-t is located inside the Event Time, i.e. these are statements about events that hold at Speech Time. The data above shows that infinitives in the complement of modals are sensitive to the stative–dynamic distinction as well, just like finite verbs and perfects.

Turning to Swedish, it is clear that epistemic readings are most easily available for stative verbs and copular constructions, as shown for the modal ma˚ste ‘must’ in (56):Footnote ³⁴

1. (56)

Future oriented deontic readings are also possible in the sentences above, given the right context (e.g. ‘The next type of bread we produce must contain enough fibers to count as a healthy bread’).

For non-stative verbs, deontic, future-shifted interpretations are often preferred, but given the right context, epistemic, Speech Time oriented readings are possible. Assume that a person is standing outside a big building, and he can hear from inside a lot of balls bouncing and people running around. In this context, he may very well utter:

1. (57)

Thus, once again, dynamic atelic predicates pattern with states in Swedish. Note that in English, in this very context, the main predicate has to surface in the progressive form. When it comes to telic predicates, it is harder to get the epistemic reading, even impossible, I would say. The best context I can suggest is the following: Mary sees a slightly blurred picture of a man holding something green in his hand, and holding his hand quite close to his mouth. Someone asks Mary what the man in the picture is doing, and she answers:

1. (58)

Even in this context, the above sentence is marked, compared to the perfectly grammatical (57) (a future or habitual deontic reading of (58) is of course possible). The sentence in (58) improves considerably, in the epistemic use, when the object is introduced by a preposition (i.e. in the conative construction, ‘eat on an apple’), which triggers an atelic reading. When no direct sensory input is present at Speech Time (as the sound in the context of (57) and the picture in the context of (58)), an ongoing present tense interpretation of the event denoted by the complement of the modal seems to be impossible, regardless of the telicity of the predicate. The judgments are very subtle, though I find both (59a) (accomplishment) and (59b) (activity) infelicitous as answer to the questions ‘Why isn't John here today? What do you think he is doing?’:

1. (59)

The sentence in (59a) is slightly more marked than the sentence in (59b), indicating that telicity is a factor here as well. Crucially, neither is as good as the English clauses with the progressive, illustrated in the translation lines.Footnote ³⁵ This again clearly shows that the simple forms in Swedish are not equivalent to the progressive forms in English.

If we probe for an answer containing a stative verb, the epistemic interpretation of must is easily available. The sentence in (60) is a perfectly felicitous answer to the questions ‘Why isn't John here today? Where do you think he is?’:

1. (60)

Other modals, such as kunna ‘can’/‘may’ and borde ‘should’, also show strong preferences for stative complements when they are used as epistemic modals (with Speech Time oriented interpretations). This is shown in (61) for kunna, which has at least two different readings: an ability reading and an epistemic possibility meaning. As an answer to a question like ‘What do you think John is doing?’, only the epistemic interpretation makes any sense. As shown in (61a), an activity verb in the complement of kan (present tense of kunna) is not felicitous in this context (though it is felicitous in the ability reading).

1. (61)

A periphrastic construction, like vara ute å ‘be out and’ works much better (indicating that these periphrastic constructions are stative, just like the English progressive), as shown in sentence (61b), and the simple present tense without any modality is unproblematic, as shown in (61c).

These contrasts, as well as the contrasts seen with måste with stative and dynamic complements, show that the stative–dynamic distinction is relevant in the syntactic/semantic computational system in Swedish. At present, I do not know why modals in their epistemic uses are highly state-sensitive, but it could be argued that the complements of modals are structurally reduced, lacking certain projections related to the temporal interpretation. It could also be argued that epistemic modals in general modify propositions, or facts, and that the temporal properties of propositions are similar to those of states (for example, one could say that there is no need for any special type of assertion time to evaluate states and propositions).

3.5 Complements of punctual ECM/raising verbs

As discussed in Hallman (Reference Hallman2009), the ECM verbs discover and reveal select stative verbs and progressives in English, as shown in (62) (from Hallman Reference Hallman2009):

1. (62)

   1. a. The inspector revealed/discovered Max to be a liar.

   2. b. The inspector revealed/discovered Max to be lying.

   3. c. *The inspector revealed/discovered Max to lie.

The same effect can be seen in the raising predicate turn out, as in (63):

1. (63)

   1. a. I wondered what John was up to, and he turned out to be writing/*write a letter.

   2. b. I thought that John was rich, and correctly, he turned out to own half the company.

It is not obvious what the Reichenbachian structure of these sentences would be. I will for now assume that the predicates discover, reveal and turn out pick out an assertion time (the time of discovery/revelation), and this time is located inside the event time (as denoted by the infinitival verb phrase), i.e. the discovery takes place as the event is going on.Footnote ³⁶

In Swedish, neither upptäcka ‘discover’ nor avslöja ‘reveal’ take ECM complements. However, the Swedish equivalent of ‘turn out’, visa sig, can be used as a raising predicate. Visa sig seems to select for stative predicates, just like English turn out. A search of the Swedish tagged corpus PAROLE (http://spraakbanken.gu.se/parole/) for the string ‘visa/visade/visar/visat sig V’ (‘Show infinitive/present/past/participle) sig Verb’) gives around 700 hits. The most common verb in this context is vara ‘be’, followed by ha ‘have’. Otherwise, stative lexical verbs like inneha˚lla ‘contain’ and betså av ‘consist of’ are also quite common, as is modal kunna ‘can; know’.Footnote ³⁷

If we take the verb lie, as used in (62) above for English ECM constructions, the difference in grammaticality is not very strong, though the copula + NP complement in (64a) is clearly preferred over the verbal complement in (64b). The non-raising version is also perfectly fine with a full verb:

1. (64)

The fairly mild effect seen above might be because of the inference from a habitual, or possibly stative, interpretation of ljuga ‘lie’ in (64b). If, however, an episodic reading is forced, an eventive verb seems to be outright ungrammatical. If we take (65) as a starting point, setting a context that forces an eventive reading, it turns out that a stative/copula verb is fine in the complement (66a), while an eventive verb is not. The eventive verb is, however, fine in the non-raising version (66c):

1. (65)

2. (66)

As pointed out in the discussion of (64), it seems like habitual interpretations sometimes interfere. Still, however, habitual readings are much more marked in the infinitival complement of visa sig than true states, as is shown below. Finite complements are not sensitive to the state–habitual distinction:

1. (67)

Summing up, the punctual raising verb visa sig clearly does not allow episodic interpretations of infinitival activity verbs in its complement, as shown in (66b) above, and even strongly disfavors habitual interpretations, see (67b). However, stative verbs are fine. I do not know why visa sig in Swedish selects stative complements, and for now I will assume that this selection is encoded in the lexical entry of visa sig. It would of course be nice to have a more principled explanation of this selectional restriction, but I will leave this for future research.Footnote ³⁸,Footnote ³⁹

3.6 Summing up and discussion

Above we have looked at five contexts where we see state-sensitivity in English, i.e. contexts where dynamic verbs have to surface in the progressive. Table 2 provides a summary of the English data.

Table 2. State-sensitive contexts in English.

Semantically, it could be argued that ast-t is located inside the Event Time in these five contexts. As far as I am aware, a dynamic verb can never surface in the simple form when ast-t is located inside Event Time. Instead, the progressive form has to be used, and as we have seen above, the progressive behaves like a stative predicate, at least as far as syntax is concerned. In other words, we have every reason to conclude that the English progressive is a derived state, i.e. something that is true at moments rather than minimal stages, according to Taylor's characterization.

The Swedish pattern crucially differs from the English one, as shown in Table 3. Note that there is no column for ‘progressive’ here, since overtly derived in-progress states in Swedish have not been investigated in this paper.

Table 3. State- and telicity-sensitive contexts in Swedish.

The clear telic–atelic split in rows 1–3 in Table 3 shows that there is no covert progressive/stativizer in Swedish, at least not of the English type (which, as we have seen above, applies to both telic and atelic predicates). Rows 4 and 5 in this table show that dynamic atelic predicates do not behave like states across the board. This importantly tells us that there is no special type of progressive operator in Swedish that applies only to dynamic atelic predicates. Rather, atelic predicates are truly ‘dynamic’ in Swedish (as opposed to English progressives). If we accept Taylor's postulates, we must conclude that ast-t can be located inside predicates that are true only at intervals, as well as predicates that are true at moments (states). However, ast-t in Swedish cannot be located inside accomplishments and achievements, i.e. predicates that lack the subinterval properties of activities (and states). As discussed above, this follows from Taylor's meaning postulates, since accomplishments and achievements are not true for any sub-interval of the event: if the ast-t were placed inside the run time of the event, it would have to pick out only a sub-interval of the event.