III. Arguments for distinguishing between the absolute and the relative past tense 413 the present Ϫ see the discussion of the ‘Special Present Time-sphere System’ in 3.2Ϫ10), we cannot use a nonprogressive present tense form to locate a single durative dynamic situation at t 0 : I am writing a book. *I write a book. (grammatical only in special uses) This constraint follows naturally from the following principles: (a) t 0 is linguistically conceived of as a point: it has no duration. (b) The present tense represents the situation time as coinciding with t 0 . (c) It follows from (a) and (b) that the situation time of a present tense clause has to be punctual. (d) If the full situation is durative, its situation time can only be conceived of as punctual if the situation is homogeneous, e. g. because it is a state (for example, a habit) or if it is represented as homogeneous by the use of the progressive form. This is not the case in *I write a book, which is bounded and therefore nonhomogeneous Ϫ see 1.45. A nonhomogeneous sentence can only refer to the complete situation (i. e. the situation time is also the time of the full situation). Needless to say, it is pragmatically impossible for such a durative situation to coincide with a point. Note that this explanation is based on temporal as well as aspectual and prag- matic factors. The notions progressive, homogeneous and durative are a ques- tion of (grammatical or ontological) aspect Ϫ see 1.20 and 1.33.1. The tempo- ral factor in this explanation lies in the fact that it defines ‘locating a situation time at t 0 ’ as ‘representing the situation time as T-simultaneous (ϭ coinciding) with t 0 ’. 8.30.2 The restriction on *I write a book also applies to I wrote a book in past represented speech, but it does not render the sentence ungrammatical because there is another interpretation of I wrote a book on which the sentence is acceptable. Compare: John said that he was writing a book. John said that he wrote a book. The first sentence is fine because was writing represents its situation as homo- geneous, which means that the situation time can be a proper subpart of the time of the full situation, so that the situation time can be T-simultaneous (coinciding) with the nearly punctual situation time of said. The sentence John said that he wrote a book does not satisfy these conditions: wrote represents the durative that-clause situation as bounded (ϭ nonhomogeneous), so that wrote cannot be interpreted as expressing T-simultaneity: a durative nonhomo- geneous situation cannot coincide with a nearly punctual time (said). However, 414 8. Temporal domains and relative tenses: theoretical foundations John said that he wrote a book is not unacceptable because it allows another interpretation: wrote can be interpreted as creating a domain of its own, which is pragmatically interpreted as W-anterior to that established by said. (In other words, wrote is interpreted as equivalent to had written.) The relevant part of this argument is that, in the same way as *I write a book cannot locate its situation time at t 0 , he wrote a book cannot locate its situation time at the situation time of said (at least in Standard British English). Since ‘locate the situation time at time X’ in both cases means ‘represent the situation time as T-simultaneous (ϭ coinciding) with time X’, this comes down to saying that the situation time of a durative bounded (ϭ nonhomogeneous) situation (which is a durative situation time, since if the situation is bounded, the situation time coincides with the time of the full situation Ϫ see 8.1.4Ϫ5) cannot be represented as T-simultaneous with an orientation time which is conceived of as punctual. (If one does not want to treat said as referring to a punctual situation, the restriction is that the situation time of a durative bounded situation cannot be represented as T-simultaneous with an orientation time which is conceived of as having a much shorter duration than the T-bound situation time). The reason why John said that he was writing a book is not ungrammatical is that the situation time of was writing can be a punctual portion of the full situation, because the progressive form represents the dura- tive full situation as homogeneous. The above data are perfectly in keeping with our claim that was writing in John said that he was writing a book is a relative past tense form, whose temporal semantics is ‘The situation time is T-simultaneous (ϭ coinciding) with an orientation time in a past domain’. As far as we can see, the data cannot be explained by any analysis which denies the existence of a relative past tense (with this semantics) in English. 8.31 Argument 9: T-simultaneity is a unidirectional relation In a theory adopting the existence of a relative past tense meaning ‘the situation time is T-simultaneous (ϭ strictly coinciding) with a past orientation time’, we can nicely explain why the subclause in a sentence like John knew that I wrote a book cannot be interpreted in terms of T-simultaneity (i. e. as using a relative tense). If wrote were a relative past tense form, the bounded situation of my writing a book would have to derive its temporal specification from the nonbounded situation time of John knowing it, which is pragmatically impossible. III. Arguments for distinguishing between the absolute and the relative past tense 415 The following is another example in which the that-clause cannot be interpre- ted in terms of simultaneity: John knew that I wrote a book. This sentence too can only be interpreted in terms of sequence: my writing of the book was completed before John knew about it. This leads to (at least) the following conclusions. Firstly, wrote cannot be a relative past tense form. If it were, it would express T-simultaneity, so that a W-simultaneity interpretation would be possible (and in fact obligatory). Secondly, there must be a reason why wrote cannot be interpreted as expressing T-simultaneity. This time the explanation cannot be that it is impossible to represent a durative situation time as coinciding with a much shorter orientation time (as in John said that he wrote a book), for the situation referred to by knew is not bounded, which means that it can in principle have any duration and that it is homogeneous. Theoretically, a durative bounded situation can coincide with a durative homo- geneous situation or with a durative subpart of it. So, what is it that rules out the simultaneity interpretation of wrote in John knew that I wrote a book? The explanation we suggest is based on the observation (made in 8.17.2) that T-simultaneity is a unidirectional relation: the bound situation time is represented as coinciding with the binding situation time, not the other way round. This is illustrated by the following examples: At 3.17 a.m. John said that Bill was not in his room. (On the default reading of this sentence, the time of the predicated situation (ϭ the situation time) of the that- clause is taken to coincide with 3.17 p.m. This is because this situation time derives its temporal specification from the T-simultaneity relation (expressed by was) with the situation time of the head clause, which coincides with the punctual Adv-time indicated by at 3.17 a.m. The sentence does not say anything about whether or not Bill was also absent from his room at any other time, though it is pragmatically unlikely that the absence was confined to just a point in time.) John said that Bill was not in his room at 3.17 a.m. (The intended reading is that in which at 3.17 a.m. specifies the time of Bill’s not being in his room rather than the time of John’s reporting this. The precise temporal location of the situation time of the that- clause is indicated by at 3.17 a.m. If T-simultaneity meant that the T-binding situation time derived its temporal specification from the coincidence relation with the T-bound situation time, this sentence should be acceptable on the reading ‘At 3.17 a.m. John said that Bill was not in his room (at that time)’. Actually, if at 3.17 a.m. specifies the time of Bill’s being in his room, the sentence can only mean ‘At some time later than 3.17 a.m. John said that Bill was not in his room at 3.17 a.m.’) The claim that the T-simultaneity relation expressed by a relative tense is unidi- rectional leads to the following conclusions. First, in John knew that I wrote a book, the T-binding situation time (knew) cannot derive its temporal specifi- cation from the T-bound situation time (wrote). Secondly, wrote cannot derive its temporal specification from a T-simultaneity relation with knew, because it 416 8. Temporal domains and relative tenses: theoretical foundations is obvious that a bounded situation (wrote a book) cannot derive its temporal specification from a nonbounded one. For lack of specification of boundaries, the situation of knowing something can in principle go on indefinitely. This renders it impossible to represent a bounded situation time as temporally coin- ciding with it, i. e. to use the relative past tense. 20 In sum, our theory nicely accounts for the fact that the that-clause of John knew that I wrote a book cannot be interpreted in terms of simultaneity. The notion of past relative tense plays a crucial part in this explanation. The fact that I wrote a book is bounded also plays a part, but it is unable to explain the obligatory interpretation of wrote by itself (as might be claimed by support- ers of the idea that the past tense form has a single core meaning, namely ‘anteriority to t 0 ’). The above explanation not only refers to boundedness but also to the idea of temporally representing a situation time as coinciding (ϭ T- simultaneous) with the binding orientation time. The latter element has nothing to do with boundedness but has everything to do with the definition of a relative tense as expressing T-simultaneity. 8.32 Argument 10: the Dutch test English and Dutch are cognate languages. In Dutch, the absolute and the relative past tenses are clearly distinct, since only an absolute preterite form can be replaced by the present perfect. Thus, to use an English gloss, we find Dutch sentences like Yesterday he {admitted / has admitted} that his father {was / *has been} dead. This difference 20. Note that the reverse mechanism is possible: if the binding situation time is bounded and the bound situation time is not, the relation of coincidence imposes temporal bound- aries on the bound situation time. For example: (i) He repeated three times that John was in the kitchen. The head clause is bounded here by the presence of three times. (Specification of the number of times that a situation is repeated induces a bounded-repetitive reading.) The clause John was in the kitchen does not by itself represent its situation as bounded. However, as explained in 8.17, the relation of coincidence expressed by the relative preterite was entails that the situation time of the that-clause is just that part of the full situation (possibly all of it) that is strictly simultaneous with the situation time of the head clause. In other words, (i) makes a statement only about that part of the that- clause situation that coincides with the situation time of the head clause. It does not tell us anything about the length of time actually taken up by the complete situation. This means that the boundedness of the head clause imposes boundaries, not on the time of the full situation referred to in the that-clause but on the latter’s situation time. In other words, by imposing boundaries, the head clause picks out a subinterval from the time of the full situation, and it is this subinterval that is interpreted as being the situation time of the that-clause. III. Arguments for distinguishing between the absolute and the relative past tense 417 in substitutability Ϫ the absolute past tense in the head clause can be replaced by the present perfect, while this substitutability does not exist for the relative past tense in the subclause Ϫ shows that Dutch has two past tenses, an absolute one and a relative one. Given the fact that Dutch and English are cognate languages, this suggests that it is quite possible that in English, too, there are two distinct past tenses. Dutch is a Germanic language which is cognate to English, and whose tense system resembles that of English in many respects. One basic difference, how- ever, is that unlike English, Dutch can use the present perfect to represent a situation as holding at a definite past time: Ik heb Jan gisteren gezien. (‘I have seen John yesterday’) Verleden jaar is het project mislukt doordat Jan niet heeft willen meewerken. (‘Last year the project has failed because John has refused to co-operate’) However, a present perfect can replace a past tense form only when the latter is an absolute tense form 21 . Dutch does not normally allow the use of the present perfect as a relative tense expressing T-simultaneity. Compare: Gisteren gaf hij toe dat zijn vader dood was. (‘Yesterday he admitted that his father was dead’) Gisteren heeft hij toegegeven dat zijn vader dood was. (‘Yesterday he has admitted that his father was dead’) *Gisteren gaf hij toe dat zijn vader dood is geweest. (nonsensical because is geweest does not allow a simultaneity reading) (‘Yesterday he admitted that his father has been dead’) *Gisteren heeft hij toegegeven dat zijn vader dood is geweest. (idem) (‘Yesterday he has admitted that his father has been dead’) It appears that of the English preterites only the absolute one can be rendered by a Dutch present perfect. This means that to refer to the past, Dutch can either use the absolute past tense or (under certain conditions) the present perfect. But to expand a past domain established by either of these tenses, only the past tense can be used. This leads to the conclusion that Dutch has both an absolute past tense and a relative one, only the former of which can al- ternate with the present perfect. This proves the existence of the relative past 21. This is the same situation as in Latin, where both the ‘perfectum’ (e. g. amavi) and the ‘imperfectum’ (e. g. amabam) can be used for absolute past time reference, but only the latter can also function as a relative past tense expressing simultaneity. 418 8. Temporal domains and relative tenses: theoretical foundations tense in Dutch. Though Dutch of course is not English, this at least suggests that English too may make use of two different past tenses. Moreover, this provides us with a possible test to say if a given past tense form in English is an absolute or a relative form: if it can translate as a present perfect in Dutch, it seems safe to say that it is an absolute tense form. Otherwise it is probably a relative past tense form. IV. Theoretical conclusions from these arguments 419 IV. Theoretical conclusions from these arguments The above arguments not only confirm the existence of the relative past tense as a tense in its own right (i. e. semantically different from the absolute past tense) but also lead to some further theoretical conclusions. These will be briefly gone into in the next sections. 8.33 Theoretical conclusion 1 The existence of two distinct meanings (an absolute and a relative temporal meaning) corresponding with a single form (the past tense) is not a case of a single past tense being ambiguous, but of two different tenses having the same formal expression (i. e. being homophonous). We have adduced a number of arguments supporting the claim that the abso- lute and relative readings which can be given to past tense forms are not always pragmatically or contextually induced interpretations, but sometimes corre- spond with distinct semantic (temporal) structures. There are in principle two ways in which this conclusion can be accommodated. One possibility is to assume that English has two past tenses, viz. an absolute and a relative one, which happen to correspond to the same grammatical form (i. e. an absolute past tense form is homophonous with a relative one). (On this assumption, a past tense form can under certain conditions be ambiguous between two mean- ings (absolute vs relative), as in John said that Bill was hungry. This is in keeping with what was argued in Argument 7 in section 8.29.) The other is to assume that there is only one past tense in English, which is ambiguous, i. e. which can express two different temporal meanings. At first sight it might seem that it is not easy to choose between these two possibilities. We are faced with the same kind of problem as when a single phonological form (e. g. gay) corresponds with two distinct meanings: is this a case of two different words being homophonous or an instance of a single word being ambiguous? However, the analysis assuming two homophonous past tenses is preferable to the analysis in terms of a single ambiguous past tense, because the latter runs counter to the principle (defended in 2.7) that each tense has a single invariant semantic (ϭ tense-structural) meaning. Since a tense is an abstract concept, viz. the correlation of a temporal meaning with a grammatical form, it is unsound from a theoretical point of view to entertain the idea of one . that the situation time can be a proper subpart of the time of the full situation, so that the situation time can be T-simultaneous (coinciding) with the nearly punctual situation time of said. The. the situation time of the that-clause is just that part of the full situation (possibly all of it) that is strictly simultaneous with the situation time of the head clause. In other words, (i) makes. a.m. (The intended reading is that in which at 3.17 a.m. specifies the time of Bill’s not being in his room rather than the time of John’s reporting this. The precise temporal location of the situation