154 2. Towards a theory of tense and time domain is directly related to t 0 . Together, they form a set whose relations we can represent in a diagram. A temporal domain is typically established by an absolute tense (locating a situation time in a particular time-zone) and is expanded by relative tenses. In Meg will buy a bike that has loads of gears the future tense form will buy establishes a post-present domain. The present tense form has expands that temporal domain because it relates the situation time of the bike having loads of gears to the situation time of Meg’s buying a bike. Specifically, it locates the having simultaneous to the buying. That is, the present tense in this sentence is used as a relative tense Ϫ it does not relate the situation time to t 0 but to another orientation time, in this case the situation time located by the future tense main clause. A temporal domain need not be expanded: it may consist just of a single situation time, as in Meg bought a bike. 2.42.1 A temporal domain is a set of orientation times which are temporally related to each other by means of tenses. Since a domain is always established by a tense form and every tense form refers to a situation time, at least one of the orientation times in the set is a situation time. As we will see, a domain sometimes also contains an orientation time which is not a situation time, such as the unspecified orientation time to which the situation time of had got up is represented as anterior in He had got up early that morning when this sen- tence is used as the first sentence of a novel. An absolute tense form always establishes a domain. This means that a domain can only be expanded by relative tense forms. Thus, in (1) John said he was tired because he had worked hard all day and that he would go to bed early. the reference is to a past domain. This is established by the past tense form said, which is therefore an absolute tense form. The other tense forms are relative tense forms: was [tired] represents its situation time as T-simultaneous with the situation time of said, had worked represents its situation time as T- anterior to the situation time of was [tired] and would go represents its situa- tion time as T-posterior to the situation time of said. See Figure 2.2. 2.42.2 When representing a temporal domain by means of a diagram, we will observe the following conventions. The domain is represented by a Venn-dia- gram because it is a set of orientation times (related to each other by the tense forms). The ‘central orientation time’ (ϭ the situation time establishing the domain Ϫ see 8.15) is the only orientation time which is placed on the time line, since it is the only orientation time that is directly related to t 0 . A vertical V. Temporal domains 155 Figure 2.2. The tense structure of John said that he was tired because he had worked hard all day and that he would go to bed early. line is used to represent the relation of T-simultaneity, whereas a slanting line represents either T-anteriority or T-posteriority. A situation time that is T- anterior to another orientation time is located to the left of the latter; a situa- tion time that is T-posterior to another orientation time is located to the right of the latter. All orientation times and situation times are represented by a cross (or ‘x’), irrespective of whether they are durative or punctual. 2.43 Unexpanded domain A temporal domain is unexpanded if it consists of just one situation time: the set of orientation times forming the domain is a singleton. This is the case in He left at five and I will leave at eight, where both tense forms establish a domain of their own, which is not further expanded Ϫ see Figure 2.3. Figure 2.3. The tense structure of He left at 5 and I will leave at 8. 2.44 Absolute and relative past tenses As is shown in Figure 2.2, English has an absolute past tense, which estab- lishes a past domain, and a relative past tense, which expresses T-simultane- ity in a past domain. The semantics of the absolute past tense is: ‘The situation time is located in the past time-sphere (defined relative to t 0 )’; the semantics of 156 2. Towards a theory of tense and time the relative past tense is: ‘The situation time is represented as T-simultaneous with an orientation time in a past domain’. 20 Arguments for this distinction are adduced in 8.23Ϫ32. 2.45 Past (time-sphere) tenses The past (time-sphere) tenses comprise (a) the absolute past tense (creating a past domain), (b) the relative tenses expressing a single T-relation in a past domain, viz. the past tense (T-simultaneity), the past perfect (T-anteriority), the ‘conditional’ tense (T-posteriority), and (c) the ‘complex relative’ (see 1.18.3) tenses expressing two or more relations in a past domain at once, viz. the ‘conditional perfect’ tense (would have V-en), as well as such (nameless) tenses as are built with had been going to, would be going to, would have been going to, had been going to have V-en, would be going to have V-en and would have been going to have V-en. All these tenses have in common that they show past tense inflectional morphology. 2.46 Present (time-sphere) tenses These comprise the present tense, the present perfect, the future tense and the absolute-relative tenses (see the next section). They all show present tense inflectional morphology. 2.47 Absolute-relative tenses An absolute-relative tense is a tense which both establishes a domain and indi- cates a relation in it. In English, the absolute-relative tenses are the future 20. This is actually a simplification. In section 10.2 we will see that a post-present binding orientation time may be treated as if it were t 0 , i. e. as a ‘pseudo-t 0 ’, and that in that case we can speak of a ‘pseudo-past time-zone’ and ‘pseudo-past subdomains’. A pseudo- past subdomain is established by a ‘pseudo-absolute’ past tense and expanded by the relative tenses typical of (true) past domains: (i) [Even if there are witnesses to the hold-up we are planning] they will no doubt say to the police that they didn’t notice what was going on. In this example, didn’t notice is a pseudo-absolute past tense form establishing a pseudo- past subdomain in the post-present domain established by will say; was going on ex- presses T-simultaneity in that pseudo-past subdomain Ϫ see section 10.2.2. It follows that the correct definition of the semantics of the relative past tense is: ‘The situation time is T-simultaneous with an orientation time in a (pseudo-)past (sub)domain’. V. Temporal domains 157 perfect (will have V-en) Ϫ see Figure 2.4 Ϫ and such (nameless) present time- sphere tenses as are built with has been going to or will be going to (and, at least theoretically, has been going to have V-en and will have been going to have V-en.) Figure 2.4. The tense structure of the future perfect. 2.48 Terminol ogy: relative tenses vs absolute-relati ve tenses The terms ‘relative tense’ and ‘absolute relative tense’ are not used in a uniform manner in the linguistic literature. For example, Comrie (1985) applies the label ‘relative tense forms’ to nonfinite verb forms (which express a single temporal relation with a contextually given orientation time Ϫ see 1.10.4), and refers to the past perfect as an ‘absolute-relative tense’ Ϫ a label which we reserve for the future perfect and for forms built with will be going to (see 2.47). We do not follow this practice. In section 1.10.4 we argued that nonfinite clauses are untensed, i. e. that only finite verb forms are tense forms. 2.49 Temporal subordination or (temporal) binding A relative tense, by definition, requires the presence of some other orientation time in order to locate the time of the situation expressed by the relative-tense clause. (For example, such an orientation time is missing in Meg had bought a bike, rendering the clause difficult to interpret, while in Meg said that she had bought a bike we have a clearly accessible orientation time Ϫ the time of Meg’s saying Ϫ to which the relative tense, the past perfect, can relate its situation.) We say that the situation time located by a relative tense is ‘temporally bound’ by (or ‘subordinated’ to) the orientation time from which the temporal location of the situation time takes its starting point. (In the example above, the situation time of the buying is located by starting at the time of saying and locating the buying anterior to it.) In the same way, we talk of ‘binding orientation times’ and ‘bound situation times’. 158 2. Towards a theory of tense and time When a situation time is T-related to another orientation time in a domain, we have temporal subordination or temporal binding. Thus, in John said he would do it, the situation time of would do is ‘temporally bound by’ (or: ‘temporally subordinated to’) the situation time of said. (We know this because the conditional tense, in its purely temporal use, can only be used as a relative tense representing its situation time as T-posterior to another orientation time in a past domain. In other words, the conditional tense is the grammaticaliza- tion of the expression of T-posteriority in a past domain.) In this case the bound situation time is the time of the predicated situation described by [he] would do [it], whereas the binding orientation time is the time of the predi- cated situation described by [John] said. 2.50 Temporal subdomain The situation time which establishes a temporal domain is the ‘central orientation time’ of the domain. When a temporal domain is expanded, each situation time (if any) that is introduced into the domain is itself the central orientation time of a domain-within- a-domain, which we call a ‘temporal subdomain’, even if that domain is not expanded. Thus in Meg said she bought a bike the situation time of she bought a bike is the central orientation time of an unexpanded domain, whilst in Meg said she bought a bike that didn’t have any brakes, the situation time of bought is the central orientation time of a domain which is expanded by the bound situation time of the relative clause that didn’t have any brakes. In a past zone, tenses function in exactly the same way in temporal subdomains as in the temporal domains which they expand Ϫ i. e. the rules for relating situation times by means of tenses are recursive in a past zone. The same does not necessarily go for the three zones that make up the present time-sphere, but we nevertheless find that specific sets of tenses for expressing the set of temporal rela- tions in a domain are re-used in other zones, so that the total number of tenses required to express temporal relations in all the temporal zones is very small. 2.50.1 A ‘temporal domain’ has been defined as a set of orientation times, containing at least one situation time Ϫ see 2.42. The set may consist of no more than one situation time, in which case the set is a singleton and we speak of an ‘unexpanded’ (see 2.43) domain. A set may also be an expanded domain, in which case it contains several orientation times, each of which is related to another by a tense relation. In the same way as any multiple set contains sub- sets, which may be singletons, an expanded temporal domain contains subdo- mains, which may be unexpanded subdomains. That is, when we expand a temporal domain, each situation time that is introduced into the domain cre- V. Temporal domains 159 ates a temporal subdomain, whose ‘central orientation time’ (i. e. the orienta- tion time from which the first T-relations in the subdomain start) is the newly introduced situation time. 21 It is important to see that the tenses used to ex- press T-relations in a past subdomain are exactly the same as those used to express T-relations in the overall past domain. Consider, for example, the fol- lowing: Bill said something. (Said creates a past domain which is not further expanded.) Bill said that some day I would lose my job. (Said creates a past domain which is further expanded: would lose expresses T-posteriority within the domain; would lose creates a past subdomain which is not further expanded.) Bill said that some day I would lose the job that I had had for over twenty years. (Would lose expresses T-posteriority within the overall past domain established by said and creates a past subdomain; had had expresses T-anteriority in the subdomain created by would lose, thus expanding it; the new subdomain created by had had within the subdomain created by would lose is not further expanded.) Ryan said Sheila had told him that she would resign if she didn’t get her way. (Said creates a past domain. Had told expresses T-anteriority in it and creates a subdo- main. In this subdomain, would resign expresses T-posteriority and creates a new, smaller, subdomain. Within this subdomain didn’t get expresses T-simultaneity and in doing so creates a further subdomain.) Figure 2.5 represents the tense structure of the last example. 2.50.2 There are some interesting theoretical conclusions to be drawn from this. First, in section 2.14 we defined ‘orientation time’ as ‘any time capable of functioning as the origin of a temporal relation expressed by a tense form’ and pointed out that it follows that a situation time (ϭ the time of a predicated situation) is always an orientation time. We can now reformulate this and say that a situation time is always the central orientation time of a domain or subdomain. Secondly, the tenses expanding a subdomain of a past domain are exactly the same as those expanding the overall past domain: both past do- mains and past subdomains use the past perfect for T-anteriority, the condi- tional tense for T-posteriority and the past tense for T-simultaneity. This means that we need only a very limited number of tenses to build quite complex past domains, since the rules for expanding a past domain apply recursively to each of the subdomains. The same thing is not necessarily true of domains in the present time-sphere, but it will be shown in chapter 9 that the expansion of 21. In 2.42.2 we wrote that “The central orientation time is the only orientation time which is placed on the time line, since it is the only orientation time that is directly related to t 0 .” It is clear that this remark only applies to the central orientation time of an overall domain, not to the central orientation time of a subdomain. 160 2. Towards a theory of tense and time Figure 2.5. The tense structure of Ryan said Sheila had told him that she would resign if she didn’t get her way. pre-present and post-present domains is also governed by a limited number of rules. (Present domains cannot be expanded because a tense form relating a situation time to t 0 is by definition an absolute tense form, which therefore establishes a domain of its own Ϫ see 9.14.) . domain. The semantics of the absolute past tense is: The situation time is located in the past time-sphere (defined relative to t 0 )’; the semantics of 156 2. Towards a theory of tense and time the. it relates the situation time of the bike having loads of gears to the situation time of Meg’s buying a bike. Specifically, it locates the having simultaneous to the buying. That is, the present. to) the orientation time from which the temporal location of the situation time takes its starting point. (In the example above, the situation time of the buying is located by starting at the