Printer Friendly

Computing perspective: the pluperfect in Dutch *.

Abstract

The proposal by linguists and narratologists that the pluperfect may be used to introduce perspective has not been founded in the (formal) semantics of tense and aspect. Neither has it been related to precise contextual conditions. In order to make a start in overcoming these shortcomings, the perspectivizing function of the pluperfect has been defined in terms of temporal semantic information and contextual properties. On the basis of the theoretical model a method has been developed to compute perspective as introduced by a pluperfect for two different text types: narratives and news reports. The validity of this method has been tested on a corpus of text fragments originating from newspapers and novels.

1. General setting

Common sense tells us that every utterance in every text has to be produced by someone. But narratology shows that not every utterance produced by a narrator has to express truth according to that narrator. Without flawing the maxim of quality (Grice 1975), a narrator may convey information for which he/she should not be held responsible, or, in other words, which should be attributed to another source (Palacas 1993; Caenepeel 1989). The "introduction of a subjective point of view that restricts the validity of the presented information to a particular subject (person) in the discourse" (Sanders and Redeker 1996) is what we mean by "perspective." The narrator may choose from a variety of means to signal his nonresponsibility: he/she may put the relevant text part between quotes, he/she may embed the relevant text part under an explicit reference to the source ("but Chomsky says that ...") or he/she may use less conspicuous ways, like the so called "free indirect mode" or "represented consciousness" (Banfield 1982; Wiebe 1990a, 1990b). The latter manner is not an easily defined -- nor easily recognized -- means of signalling a second source. It expresses in fact a mixture of sources: narrator and character. But there are even more subtle ways of introducing a point of view different from the narrator's. One way is in the form of a referring expression (Ushie 1986; Sanders 1990), another is in the use of tense (Couper-Kuhlen 1987: 21, 22; Caenepeel 1989; Salkie 1989; Wiebe 1990a, 1990b; Sanders 1994; Landeweerd 1998). This faculty of tense as an instrument for the creation of perspective will be the focus of this paper. We will examine the question of when and why a pluperfect in Dutch can introduce perspective. (1)

In both the theory focusing on linguistic properties of narratives and the temporal (discourse) semantics tradition, we see an intricate relation between tense and perspective. Discourse representation theory (DRT) is a theory that combines linguistic insights (semantic and syntactic) on the sentence level with discourse-theoretical insights. It incorporates to a large extent the logico-semantic tradition in which compositionally defined formal representations of language are mapped onto a model of the world. In the representation structures of DRT, called DRSs, tense is defined in terms of relations between the semantic entities point of speech, events, and points of reference -- each reminiscent of Reichenbach's three-point structures for the English tenses (see section 2, Introduction). However, an extension of this basic set was proposed in Kamp and Rohrer (1983) and applied again in Kamp and Reyle (1993), in order to distinguish between the apparently different functions a reference point in a DRS could have. This new type of reference point was called temporal perspective point (TPpt). Although the TPpt functions as a secondary speech moment, Kamp and Reyle do not attribute this secondary speech moment to a speaking or thinking person, nor do they consider other aspects of personal attitude that are usually connected with the notion of perspective. In sum, the notion of temporal perspective that we find in the literature on formal discourse semantics, and most prominently in relation to the pluperfect, is not connected to a notion of narrative perspective. Landeweerd (1998), however, convincingly argues for a connection between temporal perspective and the narrative perspective attributed to some particular individual responsible for the validity of the content of (part of) the discourse.

In linguistic theories focusing on narratives, we find more observations on the special relation between tense and perspective: Caenepeel (1989), Wiebe (1990a), and Sanders (1994) all address the perspectivizing faculty of certain tenses, in particular the pluperfect. In line with Moens and Steedman (1988), Caenepeel argues that the perfect may be taken to be a stative operator that enables the user to focus on the consequential state of an event. Furthermore, Caenepeel ascribes a (potentially) perspectivizing effect to stative sentences in a discourse. The state of affairs described by this type of sentence is argued to be "perceived or contemplated from a particular [personal or subjective] viewpoint" (Caenepeel 1989; see also Caenepeel 1993; Sandstrom 1993; Lascarides and Asher 1992). It follows that both the perfect itself and those interpretations of the pluperfect that are based on it (discussed in section 2.1) have the potential of introducing a perspective other than that of the narrator.

Wiebe explicitly mentions the potentially perspectivizing force of the pluperfect when used as "shifted past" in fiction narratives. This interpretation of the pluperfect differs from the stative one put forward by Caenepeel, thus strengthening the expectation that the pluperfect may indeed cause perspectivization: in both its interpretations in English, the pluperfect is argued to trigger perspectivization. (2)

We also believe that tense, and in particular pluperfect, may be responsible for perspectivization in specific contexts. However, we feel that an instrument is missing to pin down this effect to the actual meaning structure of tense, in the case of pluperfect. Furthermore, it should be possible to identify the contextual properties that activate the perspectivizing property of the pluperfect (and other tenses). In order to accomplish this, we consider it useful to combine insights from the theory of narratives with instruments developed in formal discourse semantics.

In section 2 we start with the latter. First we argue that Dutch needs three different representations for the pluperfect (section 2.1). Then we introduce the framework (section 2.2) stemming from the tradition of temporal discourse semantics. Subsequently we identify the notion that is central to our argument: the free evaluation point (section 2.3). Then we return to the notion of perspective by explaining how it is defined in the literature and how we intend to make it linguistically definable for our purpose (section 2.4). Finally, after discussing the complicating factor "text type" (section 2.5), we present our hypotheses, connecting the formal definition of the pluperfect and some properties of the context to our perspective notion.

2. Theoretical framework and hypotheses

Introduction

Many of the modern accounts of temporal discourse structure are based on Reichenbach's Elements of Symbolic Logic (Reichenbach 1947). In order to represent each of the English tenses in a unique way, Reichenbach introduces tripartite structures. These structures are constituted by the "point of speech" (S), "point of event" (E), and "reference point" (R). These points are connected by a precedence or a simultaneity relation. (3) His model is useful in distinguishing between the present perfect, the simple past, and the pluperfect, forms that all refer to an eventuality in the past -- but with a different focus. The structures he attributes to these tenses are
E < R, S (present perfect)
E, R < S (simple past)
E < R < S (pluperfect)


where "<" represents the precedence relation and the comma represents simultaneity of two points. Reichenbach, in a way, was the pioneer for the so-called anaphoric treatment of tenses. This approach, very common in modern discourse semantics, treats tense as anaphorically referring to an antecedent, usually (but not necessarily) provided by the tense in the preceding clause (Hinrichs 1981, 1986; Partee 1984; Schopf 1984, and many linguists since). It is the reference point as part of the represented tense structure that is supposed to act as antecedent. Reichenbach can be seen as the pioneer of this approach by force of his conviction that the reference point, R, should be kept in a constant position when representing compound sentences. This principle is known as the rule of "permanence of the reference point" (Reichenbach 1947: 293).

In the anaphoric accounts of tensed clauses it appears to be of importance to add the internal structure of the expressed eventualities, in German referred to as Aktionsart, as a relevant factor in the interpretation and representation of temporal discourse structure. A distinction is made between event clauses and state clauses. (4) The discourse dynamics of event clauses and state clauses turn out differently as a consequence of their dissimilar relations to the reference point of the preceding clause in the discourse. Whereas events help the reference time to move on -- thus causing the events to go with the flow of time -- states do not. (5)

In this tradition, the reference point generally serves a certain discourse purpose: it functions as an anaphor to the preceding "current" reference point and helps to determine the temporal location of the eventuality (cf. Hinrichs 1981, 1986; Webber 1988; Schopft 1984). (6) We shall refer to reference points that meet this purpose as temporally motivated reference points. But in some contexts there appears to be no such temporal motivation for a reference point introduced by a pluperfect. (7) We refer to these reference points as free reference points and we expect them to serve another purpose, that is, the introduction of a perspective.

The examples below invite different interpretations of the pluperfect. Assuming permanence of the reference point, in (1) the eventuality in pluperfect is understood to have taken place before the eventuality in simple past in the subclause. (8) In (2) the same relation should hold -- but no precedence relation between the eventualities expressed seems to be intended. In other words, the (structure contributed by the) pluperfect in the main sentence is not entirely temporally motivated. This fact cooccurs with a semantic effect: someone, possibly Boy, appears to have reported the smile. A perspective is introduced.

(1) Toen hun huwelijk begon in te storten had Anita Johnny pas een keer bedrogen.

`When their marriage started to deteriorate, Anita had deceived Johnny only once.'

(2) Toen Boy hem aankeek had Bob teruggelachen.

`When Boy looked at him, Bob had smiled in return.'

In order to put the general expectation to the test, we shall first argue that the Dutch pluperfect may have three different interpretations and should thus be represented in three different -- though related -- ways (section 2.1). Subsequently we shall present a type of representation structure, the so-called two-track structure (henceforth TTS) (section 2.2). Finally, we shall specify our expectations concerning the function of a free reference point in terms of the TTS of the pluperfect and temporal properties of the context of the utterance containing the pluperfect (section 2.3).

2.1. Three interpretations for the Dutch pluperfect

In order to explain the three possible interpretations of the pluperfect in Dutch we will make a comparison with the English use of this tense form, which is treated as an ambiguous category. Since Jespersen (1924), linguists have acknowledged that the English pluperfect can be interpreted either as a perfect-in-the-past or as a past-in-the-past (Salkie 1989; Caenepeel 1989, 1993; Kamp and Sandstrom 1993). In the former interpretation the pluperfect should be analyzed as a combination of tense-aspect information; in the latter interpretation the pluperfect expresses "double tense" information. English and Dutch are alike in this respect. (9)

The difference between Dutch and English use of the pluperfect relates to the aspectual function of the perfect part -- and thus concerns the perfect-in-the-past interpretation. This difference in the interpretation of the aspectual part occurs both in the present perfect and in the pluperfect, as we will demonstrate. (10)

Perfect-in-the-past: the consequential-state reading. It has been argued by several linguists that by using the English perfect, the speaker refers to the state of affairs that follows from the eventuality expressed. This state of affairs is the so-called consequential (or result) state (Moens and Steedman 1988; Caenepeel 1989; Kamp and Reyle 1993). When the auxiliary with which the perfect participle forms the tense combination is marked for the present, the consequential state is said to hold at the moment of speech. When it is in the past tense, the state is said to hold at some moment in the past.

Reichenbach represented the present perfect and the past perfect (pluperfect) E < R, S and E < R < S respectively, where the capital E may represent either an event, or a process or a state. (11) In the consequential-state interpretation, E would stand for the event preceding the consequential state. That is, the consequential state itself is not represented by the traditional Reichenbachian perfect structures. In order to represent this reading, it would be more accurate to adopt an extra notion, denoted by [E.sub.cs]. This [E.sub.cs] would represent the consequential state, starting at E and finally overlapping with the moment in time on which the speaker focuses, the reference point R.

It is worth emphasizing that the consequential-state interpretation of the perfect primarily refers to this consequential state [E.sub.cs] -- and not to the eventuality E preceding the state. However, since the consequential state depends on the occurrence of the preceding eventuality, this eventuality can always be inferred. (12) English and Dutch examples of the present perfect and the pluperfect that can be interpreted in this way are (3)-(6). We will use the abbreviation PP[F.sub.cs] for this PluPerFect-consequential-state interpretation.

(3) Oscar is radeloos: George heeft hem verlaten!

(4) Oscar is desperate: George has left him!

(5) When I came in, Gino had left already.

(6) Toen ik binnenkwam, was Gino al vertrokken.

Perfect-in-the-past: the past-eventuality reading. The consequential-state interpretation is not to be given to all uses of the perfect in Dutch. Quite frequently in Dutch, we find uses of the perfect that seem to refer to the eventuality expressed rather than to its consequential state. This interpretation is supported by the example of sentences containing aspects of meaning that are incompatible with the consequential-state interpretation:

(7) a. Rupert is in 1996 uit Rotterdam vertrokken.

`Rupert has left Rotterdam in 1996.'

b. Rupert heeft (toen) als scheepskok op de "Neeltje Maria" gewerkt.

`Rupert has (then) worked as a cook on the ship "Neeltje Maria."'

c. Rupert is (lang geleden) uit Rotterdam vertrokken maar sinds drie maanden is hij er weer.

`Rupert has left Rotterdam (long ago) but he is back since three months.'

The consequential-state reading of the perfect is not compatible with certain types of adverbials: adverbials that temporally locate the eventuality itself, as in (7a) (Dowty 1979; Oversteegen 1989). This explicit reference to the eventuality, in this case, the event of leaving Rotterdam, does not agree with the claim that the sentence primarily refers to the consequential state of an inferred event (having left Rotterdam).

In (7b), the consequential-state reading is improbable for another reason. Consequential states result from an accomplished eventuality; they imply a change of state. In general, states do not describe or imply a change of state (see for exceptions Sandstrom 1993: 124), and so they do not have a consequential state (Moens and Steedman 1988). (13) The eventuality in (7b) is classified as a state, however, and the example is acceptable in Dutch. This gives us more justification for the rejection of the consequential-state interpretation as the only possibility for Dutch perfect-in-the-past.

In (7c), the second clause explicitly confirms that the first clause does not refer to the consequential state of leaving. The consequential-state interpretation leads to the inference that Rupert is not in Rotterdam at the moment of speech (see Boogaart 1999: 157), and this is obviously not true.

The English translation of the sentence in (7a) is not acceptable, as may be concluded from this statement by Comrie:

In general, the perfect is incompatible with time adverbials that have definite past time reference, i.e. time adverbials that refer to a specific moment or stretch of time located wholly in the past. Thus English excludes sentences like * John has broken his leg yesterday, and this constraint crucially overrides even considerations of present relevance (Comrie 1985: 32).

We interpret the impossibility of the English perfect to combine with an adverbial with definite past time reference as the impossibility of REFERRING to an eventuality that is located wholly anterior to the reference time. (14) This also explains the unacceptability of the translations of (7b) and (7c), and it is almost entirely compatible with the uses of the present perfect discerned in the literature on the English perfect. (15) McCawley (1971, 1981) and Comrie (1976) distinguish four separate types of the present perfect: (1) the universal perfect or perfect of persistent situation; (2) the existential or experiential perfect; (3) the hot-news perfect or perfect of recent past; (4) the stative perfect or perfect of result. The last type is identical to the consequential-state reading that we have already discussed.

The first type should be interpreted as claiming that some state holds at a stretch of time extending from the past into the present. This is not in contradiction with our restriction on the use of the present perfect in English. The second type should be interpreted as claiming that an eventuality occurs at least once in a stretch of time extending from the past into the present. Obviously, this type has a quantificational rather than a referring function. (16) Our example (7b), without the adverb between brackets, can be read as an example of this type, and as such it is acceptable. In fact, it is only the third type that really contradicts our interpretation -- but note that for this type the past is explicitly restricted to the recent past!

Our claim with respect to the Dutch present perfect is that it can be used to refer to an eventuality that is located wholly anterior to the reference time. Let us call this the past-eventuality reading of the present perfect. This comes down to saying that in the Reichenbachian representation of this reading of the present perfect, the E should precede R. So the location of R still distinguishes the present perfect (E < R, S) from the Dutch simple past (E, R < S), just as Reichenbach intended it to. (17,18)

The pluperfect in Dutch has an interpretation parallel to the interpretation of the present perfect just described -- with the obvious difference that the reference point precedes the point of speech instead of coinciding with it. We shall refer to this interpretation as PP[F.sub.e] -- for the PluPerFect-eventuality reading of the pluperfect -- where e may represent any type of eventuality. The eventuality in this interpretation allows for temporal adverbs with definite past-time reference (cf. [8a]). (19) It may also express a state (cf. [8b]). The opposite of the eventuality may be stated at the point of reference, since the eventuality is interpreted as having been completed at that point or, in other words, it is wholly in the past of R (cf. [8c]).

(8) a. Rupert was in 1996 uit Rotterdam vertrokken.

`Rupert had left Rotterdam in 1996.'

b. Rupert had vroeger als scheepskok op de "Neeltje Maria" gewerkt.

`Rupert had worked as a cook on the ship "Neeltje Maria" in earlier days.'

c. Rupert was (lang geleden) uit Rotterdam vertrokken maar inmiddels was hij er weer.

`Rupert had left Rotterdam (long ago) but he was back again now.'

This reading of the pluperfect in Dutch can be represented as E < R < S and it has the following properties:

1. The eventuality is referred to and not merely inferred (as is the case for the PP[F.sub.cs], which refers to the consequential state of the eventuality).

2. The eventuality is "envisaged" from the position in time referred to as R; therefore we may say that the temporal focus is on that moment (as is also the case for PP[F.sub.cs]) and not on the eventuality itself. (20)

3. The eventuality is interpreted as having been completed at a time anterior to this temporal focus. If the eventuality defines a state, this state is not required to hold at the time referred to as R.

These properties provide us with a means of distinguishing the PP[F.sub.e] interpretation of the Dutch pluperfect from other interpretations. The possible addition of a temporal adverbial has already been introduced as an instrument to distinguish the PP[F.sub.e] from the PP[F.sub.cs] interpretation. In our discussion of the past-in-the-past reading of the pluperfect in the following subsection, we will introduce the instruments needed in order to establish the presence (or absence) of the properties 2 and 3. In section 3.2 all instruments used in this paper are described in detail. (21)

Past-in-the-past. The previous paragraph dealt with the aspectual contribution of the perfect, either in present or in past tense. Jespersen (1924) argued in favor of yet another interpretation of the pluperfect, namely in terms of double tense information. This so called past-in-the-past interpretation has been controversial in the literature on tense and aspect. For one thing, it disturbs a compositional view on tense and aspect, and there have also been other objections (see Binnick 1991 for an overview of the discussion on this item). The past-in-the-past interpretation has, however, been adopted by several linguists (Salkie 1989; Caenepeel 1993; Kamp and Reyle 1993; Sandstrom 1993). We shall not defend it for the English pluperfect; we will restrict ourselves to the Dutch case.

Neither pluperfect used in example (9), taken from our corpus, should be analyzed as a perfect-in-the-past, that is, they should not be interpreted as PP[F.sub.cs] nor as PP[F.sub.e]:

(9) Haar vader had smalend gelachen toen ze laatst op een avond die bewering in een avondprogramma hadden gehoord.

`Her father had laughed contemptuously when they had heard that statement recently in an evening program.'

The reason for rejecting both the PP[F.sub.cs] reading and the PP[F.sub.e] reading is that the sentence temporally focuses on the eventualities of laughing and hearing rather than just referring to them (or inferring them), and rather than focusing on the moment of having laughed and having heard. This interpretation of where the temporal focus goes is corroborated by the use of toen `when': the temporal domain created by the subclause for the location of the eventuality in the main sentence is the time of hearing -- not some time R afterward.

For this interpretation of the pluperfect we shall introduce an extra reference point, simultaneous with the eventuality, to represent the temporal focus. The Reichenbachian representation would be E, R' < R < S. We refer to this interpretation by PP (for the Past-in-the-Past reading of the pluperfect).

The property shared by each of the now-presented interpretations of the pluperfect (PP[F.sub.cs], PP[F.sub.e], PP) is that they introduce a reference time in the past. Thus we may say that the unmarked use of the pluperfect calls for a past context In addition, the explanation given of the past-in-the-past interpretation boils down to saying that it can be interpreted as a simple past in a past context. If the tense used in the immediate context were the present tense, then the temporal condition for use of a pluperfect would not be met. All other things being equal, the most logical alternative to a PP would be a simple past. This means that the past-in-the-past interpretation can be put to the test by changing the tense used in the context into the present: the pluperfect of the PP type can better be "reset" to the simple past. The fragment in (10) shows how this works for the pluperfect clauses in (9); the form marked by "[check]" is the one preferred.

(10) Haar vader ?heeft smalend gelachen / [check]lachte smalend toen ze laatst op een avond die bewering in een avondprogramma ?hebben gehoord / [check]hoorden.

`Her father has laughed / laughed contemptuously when they have heard / heard that statement recently in an evening program.'

From the preference for simple past in this modified fragment we conclude that Dutch does indeed allow for the past-in-the-past interpretation.

To summarize, we have described three interpretations of the Dutch pluperfect. We shall now use the descriptions to find criteria for recognizing the correct interpretation in individual occurrences of the pluperfect.

First, our argument for acknowledging a PP reading of the Dutch pluperfect can also be used for distinguishing this reading from the perfect-in-the past readings. If, as a consequence of changing the past tense used in the context to the present tense, the pluperfect is reset to the present perfect rather than to the simple past, it has a perfect-in-the-past interpretation, that is, it is either a PP[F.sub.cs] or a PP[F.sub.e]. This test relates to the second property listed in the previous subsection. An illustration is given in (11), a fragment taken from our corpus. (22)

(11) Ze maakte een sopje en hing haar handen en polsen in het warme water. Niet te geloven dacht ze, niet te geloven. Haar hele leven had ze haar handen beschermd tegen alles. [...] Nu stak ze met genoegen haar armen in zeepsop.

`She added soap and hung her hands and wrists in the warm water. Unbelievable, she thought, unbelievable. Her entire life she had protected her hands against everything. [...] Now she stuck her hands in the soapy water with pleasure.'

(12) Ze maakt een sopje en hangt haar handen en polsen in het warme water. Niet te geloven denkt ze, niet te geloven. Haar hele leven [check]heeft ze haar handen beschermd tegen alles / ?beschermde ze haar handen tegen alles. [...] Nu steekt ze met genoegen haar armen in zeepsop.

`She adds soap and hangs her hands and wrists in the warm water. Unbelievable, she thinks, unbelievable. Her entire life she has protected / protected her hands against everything. [...] Now she sticks her hands in the soapy water with pleasure.'

Another criterion connected to the second property listed in the previous subsection employs the notion of temporal focus. Referents focused upon, nominal or temporal, are never new (Prince 1982; Gundel et al. 1993). According to centering theory (Grosz et al. 1995) the backward-looking center of a sentence (comparable to focus) MUST be introduced in the preceding sentence. Webber (1988) argues that both reference points and eventualities may introduce the focus (or backward-looking center) of the following sentence. This implies that, in order to keep the temporal focus constant (a situation referred to as "continuing" in Grosz et al. 1995) a sentence containing a PP should be preceded by a sentence in pluperfect. The pluperfect warrants an appropriate potential temporal focus for the focused R'. This may be either the eventuality in the preceding pluperfect sentence or its reference point. (23) If the tense of the preceding sentence were anything but a PPF, we would expect a focus shift.

As a consequence, we have yet another means of distinguishing between PP on the one hand and PP[F.sub.cs] or PP[F.sub.e] on the other hand. We call this the temporal-focus test. PP[F.sub.cs] and PP[F.sub.e] most naturally follow a simple past sentence, whereas PP requires a pluperfect left-hand neighbor. Of course, this requirement only holds in the purely temporal interpretation of the pluperfect. When we apply the temporal-focus test to the sentences in the corpus, we neglect the possibility of a free evaluation point and pretend that the pluperfect simply functions in order to express a temporal relation.

Another instrument for distinguishing PP from PP[F.sub.e] is connected to the third property listed in the previous paragraph for PP[F.sub.e]: whereas the eventuality in a PP[F.sub.e] has been completed at the time focused upon (R), in PP this is not the case. The temporal focus for PP is on R' (<R) at which the eventuality is ongoing. This implies that we may add an adverb expressing completion to a PP[F.sub.e] -- but not to a PP. (24)

The examples (13) and (14), a PP and a PP[F.sub.e] reading respectively, demonstrate that an adverb like al `already' discriminates between these two interpretations of the pluperfect.

(13) *Toen ze laatst op een avond die bewering in een avondprogramma hoorden had haar vader al smalend gelachen.

`When they heard that statement recently in an evening program, her father had already laughed contemptuously.'

(14) Toen ze een nieuwe antiseptische handcreme ontdekte had ze haar hele leven haar handen al beschermd tegen alles.

`When she discovered a new antiseptic hand cream, her entire life she had already protected her hands against everything.'

Conclusion. We conclude that there are three possible interpretations of the pluperfect in Dutch (but not necessarily in English). We did not explicitly take a stance in the discussion about the compositional nature of the pluperfect. Nevertheless, the first two interpretations, PP[F.sub.cs] and PP[F.sub.e], were analyzed as being derived from a combination of a past tense and perfect aspect. The last one, PP, was found to lack the perfect aspect. In fact, it was analyzed as a double past and as such it cannot be seen as compositionally derived. This noncompositional nature of the PP may explain the fact that the pluperfect in English does allow for a temporal adverbial specifying the time of event -- whereas the present perfect does not (Comrie 1985: 66). (25)

A summary of the Dutch situation is given below:

[ILLUSTRATION OMITTED]

At this point it should be remarked that the distinction between PP[F.sub.cs] and PP[F.sub.e] will appear to be of no consequence to the results of our investigation. Nevertheless we shall treat them separately, as two different values for the variable "type of pluperfect." We have several reasons for doing this. First, we are convinced that they represent two individual interpretations. But these interpretations do not have fundamentally different semantic representations. Most importantly in this context, they have the same number of reference points. Second, claims can be found in the literature about the nature of the perfect as a stative operator. As a stative operator the perfect would turn each perfective sentence into a perspectivized sentence. Since our PP[F.sub.cs] interpretation of the pluperfect but not our PP[F.sub.e] interpretation would contain such a stative operator, not making the distinction would lead to the false inference that all occurrences of a Dutch perfect-in-the-past could be expected to introduce perspective. Because this expectation would do no right to our sources, we shall isolate the PP[F.sub.e] readings, and we shall be aware that nothing is claimed in the literature about their pespectivizing ability.

In the following section we present a representation format for these readings. The format is meant to make it easier to perform the "computations" required to test our expectations with respect to the perspectiveintroducing faculty of the Dutch pluperfect.

2.2. Two-track representations

In this section we shall introduce a certain style of representation. We will only partly discuss the theory from which they originate, the two-track theory (Oversteegen 1989). The part of the theory, henceforth TTT, required to understand the two-track representations as a tool in attaining our research goal consists of section 2.2.1, a sketch of the general idea behind the two tracks; section 2.2.2, an explanation of the way tense, perfect, and event type are represented in TTT, and section 2.2.3, the introduction and explanation of the three-fold representation of the pluperfect in Dutch.

We are convinced that TTT is a tool in attaining our research goal but not the only tool that would enable us to prove our point. Other ways of building a representation for the temporal meaning of pieces of discourse may be equally suitable provided they obey certain conditions such as, for example (1) allowing for vagueness in the relations between eventualities, and (2) making use of some kind of reference or evaluation points.

2.2.1. The general idea. The motivation for the form of the representations used in TTT (Oversteegen 1989) comes from McTaggart (1927), who argues that time can be thought of as something dynamic, constantly advancing, and it can also be considered from a God-like perspective covering the history of all eventualities. Oversteegen proposes a linear, two-track representation of temporal meaning in which each of these conceptualizations of time can be expressed. For each sentence, in this representation, called TTS for two-track structure, both the dynamic and the static temporal meaning aspects can be reflected. In the former view, an eventuality is seen from a particular point in time and the predicates PAST, PRESENT, and FUTURE are relevant. Some expressions of time have a deictic nature, and the truth of a tensed statement depends on the temporal point of view taken. Natural-language elements expressing this dynamic and deictic temporal meaning are represented on the S-track. Here S stands for "speaker" or "point of speech." The S-track has a point structure and harbors points of speech and points of evaluation. Points of temporal perspective in general are represented on the S-track (McTaggart referred to this type of temporal ordering as the "A-series").

The static conceptualization of time is independent of a point of view. Eventualities are temporally ordered by the precedence relation but not with respect to the point of speech or with deixis (McTaggart referred to this type of temporal ordering as the "B-series"). It is referred to as static because, once an eventuality E1 is ordered before some other eventuality E2, this ordering is permanent and will not change over time. In TTT, natural-language elements establishing this static and nondeictic temporal ordering are represented on the E-track, where E stands for "eventuality."

By means of a TTS contribution, temporal expressions are represented according to both these views, and the representation contributions are allowed to interact. Since tense is a deictic category characterized by the predicates PAST, PRESENT, and FUTURE, it is primarily represented on the S-track. Event type is associated primarily with the E-track because it is independent of temporal location with respect to the point of speech; a process located in the future will be just as much a process when it is present or past. Perfect in itself is not dependent on the point of speech either: it can combine with past, present, or future and it is therefore primarily considered an E-track notion. Temporal adverbials can be deictic (cf. yesterday) and consequently located on the S-track, or can express relative position (cf. an hour earlier) and consequently be located on the E-track. For each clause, the linguistic categories expressing temporal meaning independently contribute to a representation structure called the two-track structure, which consists of an S-track, an E-track, and interaction between points and intervals on these tracks. Sequences of clauses (or larger portions of discourse) can be represented by means of a (complex) TTS.

This principle of independent contribution makes TTT convenient for our purpose. When we have defined the contributions of the different interpretations of the pluperfect to a TTS, the contribution of event type, and the way two TTSs representing consecutive sentences are combined, we shall be in the position to formally rephrase our expectation with respect to the relation between reference points and perspectivization and to put it to the test in a scientifically acceptable way.

2.2.2. Tense, perfect, and event type in TTT. Tense generally involves the S-track because its most important function is to locate the eventuality expressed by the sentence with respect to the point of speech. More relations are possible than just the three relations that a Priorean tense logic can express: E < S, E = S, and E > S (where E stands for eventuality time and S for point of speech). TTT, with its double track, is capable of expressing more subtle differences. (26)

Tense also establishes a relation between an eventuality on the E-track and evaluation points on the S-track; we will refer to this as a "bridge" between the two tracks. Semantic aspect or type of eventuality is represented on the E-track, but, because of the interaction between tense and aspect, the representation of aspect may affect the set of evaluation points of the eventuality on the S-track.

In his discussion of the past perfect tense, Reichenbach (1947) talks about both E and R as if they were times of eventualities in the past. Dowty (1982) uses two indices, i and j, corresponding to Reichenbach's R and S respectively, to solve certain scope problems in traditional tense logic. In the same spirit, Oversteegen (1989) replaces R with S', which we call an imaginary speech point. Both S' and S can function as evaluation points.

Tense establishes a relation between an eventuality on the E-track and an evaluation point on the S-track. This may be the S-point itself but it may also be an S'-point, depending on the tense used, since S' only plays a part in past tenses; the present-relevance quality of present tenses is reflected by the fact that S itself functions as the relevant evaluation point.

Consequently, we get the basic TTSs below for the contribution of present and past tense respectively to the TTS of the entire sentence.

(16) [ILLUSTRATION OMITTED]

In TTT, the perfect is not treated as an autonomous tense but as a temporal component that combines with a basic tense form to result in a "full" tense such as the present perfect. For example, the value of the relevant evaluation point will be S if the perfect is combined with a present tense, and it will be S' if the perfect is combined with a past tense.

When the perfect combines with one of the basic tense forms, its contribution to the final representation is the constraint that the eventuality precedes the E-track element that corresponds to the relevant evaluation point. The evaluation points for the eventuality are provided by the basic tense form. Combining the perfect with the present and the past results in the derived forms shown in (17) below.

(17) [ILLUSTRATION OMITTED]

One thing is missing from the representation structures in (16) and (17): a specification of the relation between E- and S-point. The position of S' to the left of S should be interpreted as precedence -- but the lack of an immediate connection between E- and S-track implies a temporal nonrelation. The reason for this is that the nature of the "bridge" between E- and S-track, contributed by tense, can only be determined in interaction with the event type expressed in the sentence.

In the case of a terminative eventuality, or an event, the entire E-interval is mapped onto one single evaluation point [diamond] S on the S-track. A sentence containing an event in one of the simple tenses expresses the occurrence of the event as whole -- not part of it. In the case of a process or a state, it is not excluded that the eventuality is still ongoing at the evaluation point. The evaluation point need not be connected in any way with the right-hand boundary of the event interval. Therefore, the TTS contribution of such an eventuality is a set of potential evaluation points [diamond] S corresponding to the entire E interval.

(18) [ILLUSTRATION OMITTED]

The symbol [diamond] S stands for potential evaluation point. The tense form in the sentence determines whether the evaluation point remains potential (i.e. in most cases of a perfect) or whether it turns into an evaluation point (simple present or past tenses).

In summary, the components for a TTS are E (the interval representing the eventuality), S (point of speech and evaluation point), S' (imaginary point of speech and evaluation point), and [diamond] S (potential evaluation point, located on the S-track and temporally corresponding [partly or entirely] to the E-interval). The tense of a sentence contributes to its TTS in terms of the order of points on the S-track, and both tense and event type may contribute to specifying which eventualities correspond to which points on the S-track. In order for a sentence to be true, these ordering and correspondence relations must be reflected in the real-world temporal ordering of eventualities and times, and the eventualities must have occurred at the times indicated by the correspondences shown in the TTS, that is, the eventualities must be "evaluated" at these time points. We will now turn to the TTS contribution of the pluperfect.

As was argued in section 2.1, we should specify three types of TTS contributions, reminiscent of the tentative Reichenbachian structures presented, for the pluperfect in Dutch corresponding to the three interpretations we distinguished. The interpretation represented as (17b) above is an adequate formalization of the perfect-in-the-past PP[F.sub.e]. It presents an eventuality as seen from an evaluation point S' in hindsight and is repeated here as (19). This eventuality may be of any type: state, process, or event. It is the eventuality itself, and not its consequential state, that is reported.

The two interpretations that the Dutch pluperfect shares with the English pluperfect, PP[F.sub.cs] and PP, look slightly different. In the state-like interpretation, the relevant E is not the eventuality expressed in the sentence itself. It is the consequential state (cf. Moens and Steedman 1988) following that eventuality. The relevant evaluation point, S', has to correspond temporally to this consequential state. The TTS contribution of the PP[F.sub.cs] is given in (20) below.

The last interpretation discussed is the PP. Whereas the two previous interpretations of the pluperfect can be seen as transpositions of the perfect, the PP is clearly a past transposition of a past tense. Therefore, we have to combine two past tense TTS contributions in order to obtain the correct representation for the PP: (21).

(19) [ILLUSTRATION OMITTED]

(20) [ILLUSTRATION OMITTED]

(21) [ILLUSTRATION OMITTED]

Note that in the case of (21) it is important whether the eventuality expressed is an event, a process, or a state. In the former case the eventuality maps entirely onto the relevant evaluation point. This should be interpreted as indivisibility: the entire event is presented as taking place at S". In the latter cases, the eventuality only partly corresponds to the relevant evaluation point. The eventuality as a whole is not presented as taking place at the moment of evaluation, S".

Having defined the three TTS contributions for the interpretations of the Dutch pluperfect, and having substituted the notion of evaluation point for the notion of reference point, we can proceed to specify our expectations with respect to the perspective-introducing faculty of a free evaluation point (see section 2, Introduction).

2.3. The free evaluation point

A pluperfect structure in itself cannot contain a free evaluation point. An evaluation point can only be considered free in relation to context. Our question therefore is, when does this occur?

As we mentioned earlier, "free" means that an evaluation point is not temporally motivated. It does not mean that there is no motivation whatsoever. Our research goal concerns a specific function of such a point. Consequently, our question may also be formulated as follows: when is an evaluation point NOT temporally motivated?

When an evaluation point is used to construct the TTS in such a way that the eventuality expressed gets its correct temporal relation (usually precedence) with respect to the eventualities in the preceding context, it is obviously temporally motivated. But when it lacks such a function, it is not temporally motivated. Compare examples (22) and (23).

(22) Toen hun huwelijk begon in te storten had Anita Johnny pas een keer bedrogen.

`When their marriage started to deteriorate, Anita had deceived Johnny only once.'

(23) Toen Boy hem aankeek had Bob teruggelachen.

`When Boy looked at him Bob had smiled at him.'

In (22), the imaginary evaluation point S' of the subclause and the imaginary evaluation point S' of the main sentence coincide (this is a consequence of the use of toen -- we shall not discuss the function of temporal connectives here; see instead Oversteegen 1989). The eventuality in the main sentence is not the focus: it is the consequential state that matters. So the TTS for the combined sentences dictates an overlap between this consequential state and the eventuality expressed in the subsentence. To both, the actual eventuality of deceiving is anterior. In order to obtain this intuitively correct representation for (22), the evaluation point of the pluperfect is required.

In (23) this mechanism would give us the wrong result: the time of the smiling would most likely not be anterior to the looking. If we identify the imaginary evaluation point introduced by the subclause and the S' introduced by the main clause, we end up with a false relation between the eventualities expressed. What would be the correct TTS? To start with, (23) is interpreted as a PP. The temporal focus of the main clause is at the time of smiling. Furthermore, the act of looking and the act of smiling should coincide. So S' of the subclause and S" of the main sentence should be identified. This results in a free evaluation point, however, the S' of the pluperfect is not motivated.

The temporal focus of a PP is on the third evaluation point, S", whereas the temporal focus of the perfect-in-the-past (PP[F.sub.cs] and PP[F.sub.e]) is on S'. Note that the notion of a temporal focus is not a decisive factor for considering the evaluation point free. The factors distinguishing between these two interpretations of the pluperfect are very similar to the factors that distinguish between a simple past and a present perfect, and there may be good reason to shift the attention to a certain point in the past, by using simple past (or past-in-the-past), instead of looking back, with the help of a present perfect (or a perfect-in-the-past).

What, then, is the decisive factor in considering the evaluation point free? In order to identify the relevant factors, we can construct TTT configurations that contain free evaluation points and subsequently determine how they can occur in each case. In other words, we determine the relevant factors by induction. First, however, we have to determine how two subsequent sentences are temporally related (see also section 1) and how this is reflected in TTT and other representation formats.

In a coherent text, every sentence is somehow related to the preceding sentence. Traditionally, the temporal relation between two subsequent sentences in a Reichenbachian approach is realized by "anchoring" the reference (or evaluation) point of the second sentence to the reference (or evaluation) point of the first sentence. This idea is also fundamental to the tense-as-an-anaphor approach that has been popular since Hinrichs (1986) and Partee (1984). In TTS terms this would mean identifying the relevant evaluation point of the sentence that was represented last and mapping the evaluation point introduced by the new sentence onto this point. (27) If we apply this procedure to the TTSs representing the sequence in (24), we get the TTS in (24c) below.

(24) Maria kwam thuis. Ze had de kat's morgens al gevoerd.

`Maria came home. She had fed the cat that morning.'

[ILLUSTRATION OMITTED]

In this combination of a simple past tense and a PP[F.sub.e], no evaluation point is free in (24c): the evaluation points and their identification help construct the correct temporal relation between the event intervals: E2 < E1. Let us now consider some configurations in which this is not the case.

(25) [ILLUSTRATION OMITTED]

In this combination of a simple past tense and a PP we have a free evaluation point. It can easily be concluded that the evaluation point S' in (25c) does not serve the purpose of locating the eventuality E1. The evaluation points S' from (25a) and S" from (25b) are mapped onto each other -- the role of S' in establishing the event structure in (25c) is unclear. A sentence sequence that would correspond to (25) would be (26): (28)

(26) Maria keek uit het raam en had toen haar vader voorbij zien lopen.

`Maria looked out of the window and then she had seen her father walk by.'

The eventualities in (26) should be represented as (partly) coinciding. (29) Both the presence of toen `then' and the inference that the looking and the seeing must have been simultaneous lead to the interpretation that E1 and E2 must -- at least -- overlap. In this paper we use the intuitive interpretation of the intended relation of eventualities expressed. We refer to it as "interpreted temporal relation." We will not elaborate on how this type of inference comes about: we selected only the clear-cut cases for our research.

A second example of a configuration containing a free evaluation point is given in (27).

(27) [ILLUSTRATION OMITTED]

In (27), the evaluation point S' is not required for warranting a correct relation between E1 and E2. Only the S-points are mapped onto each other and the S' contributed by the PP[F.sub.e] in (27b) is not temporally motivated. A sentence sequence that would be represented as (27c) is given in (28).

(28) Al Maria's planten zijn dood. Ze hadden aan te lage temperaturen blootgestaan.

`All of Mary's plants are dead. They had been exposed to temperatures too low.'

The eventualities in (28) are interpreted as following each other: the plants are dead as a result of low temperatures. The reasons why the tense in the second sentence has to be represented as a PP[F.sub.e] (27b) and not as a PP[F.sub.cs] or a PP are (1) it will not be a PP[F.sub.cs] because the eventuality expressed is a state itself (and states do not invoke consequential states); and (2) the temporal focus is not on the time of the eventuality but on some (unidentified) later time.

We are now in a better position to determine what ingredients we need to "calculate" whether we will end up with a free evaluation point. The sentence containing the pluperfect will be referred to as the target sentence. This target sentence is itself crucial: we have to know what type of pluperfect it contains. A simple comparison between (24) and (25)/(26) confirms this observation. The former TTS, representing a PP[F.sub.e] in the target sentence, does not contain a free evaluation point. The latter TTS, containing a PP in the target sentences, does.

The left-hand neighbor of the target sentence plays a part. Since the relation between the evaluation points of this sentence and the target sentence is central in building the combined TTS, the evaluation point in the target sentence can only be free relative to a tense in the preceding sentence.

The interpreted relation between the eventualities expressed in the target sentence and its left-hand neighbor is decisive in calling an evaluation point free. If, for example, the interpreted temporal relation between E2 and E1 in (26) were a precedence relation, the S' in (25b) and (25c) would not be free.

In short, the information we need to compute a free evaluation point consists of

-- the type of pluperfect in the target sentence;

-- the tense used in the sentence to the left of the target sentence;

-- the interpreted temporal relation between the eventualities in the target sentence and the sentence to the left of it.

Before formulating a hypothesis, we have to make another distinction: one concerning text types. As mentioned in the Introduction to section 2, fictional narrative texts and nonfictional news reports are not identical with respect to the way they temporally relate individual sentences in a sequence. Narratives are strongly "anaphoric" with respect to temporal text organization: as a rule, the eventualities are connected by coherence and temporal relations. In news reports, the eventualities are preferably linked to a deictic center, provided by the text. This difference has an impact on the way the hypothesis can be tested for each of these text types.

2.4. The impact of text type

Reichenbach (1947) introduced the permanence-of-the-reference-point principle. We may interpret this principle as requiring that two consecutive sentences, by default, search for a shared reference point (or: evaluation point) -- unless they are explicitly modified in different ways. An example of this principle is (29) below. (30)

(29) Maria came in. She had run an errand.

The moment Maria enters (the S' contributed by S1) and the moment she has run an errand (the S' contributed by S2) coincide. This principle needs to be modified in two important ways. Both modifications have to do with text-type properties.

The first modification concerns the identity relation. A well-known property of narratives is that consecutive sentences, for example in simple past, are interpreted as if the evaluation point moves slightly in the direction of the time arrow. So the eventualities expressed are interpreted as happening "immediately after" each other (the term comes from Partee 1984):

(30) Maria came in. She sat down. She took a pen from her purse and asked for a piece of paper.

This property of narratives is not reserved for simple past sentences; consecutive sentences in the pluperfect may be interpreted likewise. (31) And since we are investigating sentences in the pluperfect, this modification is relevant to our research.

The second modification concerns the general applicability of the principle. As Caenepeel (1995) convincingly argues, the "anaphoric" interpretation of tense in texts (cf. Hinrichs 1986) -- where "anaphoric" refers to how the evaluation point looks for an antecedent to adhere to -- is typical for narrative texts. In news reports, for instance, the mechanism is "deictic" rather than anaphoric. The eventualities (and the evaluation points underlying them) are ordered with respect to S rather than with respect to each other. This difference between narrative texts and news reports is an instantiation of a difference of a more general nature.

Narratives build an episodic structure in which a strong coherence is presumed between the eventualities represented. Lascarides and Asher, in many publications, argue that each sentence in a (narrative) discourse must be related to a preceding sentence by means of a coherence relation. The speech moment S is much less important in narratives: Caenepeel and Sandstrom (1992) refer to this phenomenon as the "bracketed speech point." In news reports, however, eventualities are, by default, related to S (Caenepeel 1993, 1995). The coherence relation and temporal relations between eventualities expressed in consecutive sentences are less strong here. This implies that the continuity-of-the-reference-point principle is not valid in that case. And this, in turn, influences the computation of a free evaluation point.

The continuation of the evaluation point as found in narrative texts enables us to define a procedure, a qualitative procedure, as we shall call it, to determine to what extent the evaluation points can be held responsible for defining the interpreted temporal relation between the eventualities. The lack of such a principle in news reports forces us to find another way to determine the temporal relations between eventualities. We shall refer to this as a quantitative procedure. It does not take into account the qualitative (identity) relations between evaluation points, it simply "measures the distance" between an eventuality and the speech point S. The units of measurement are steps: an eventuality in simple past is one step away from S, an eventuality in pluperfect, either two steps (PP or PP[F.sub.e]) or one step (PP[F.sub.cs]). (32) This difference is due to the interposition of S': an event in PP or PP[F.sub.e] is one step from S'; an event in PP[F.sub.cs] is simultaneous with S'. The number of steps is crucial in determining the relative positions of eventualities.

We shall return to this issue in section 3.2. At this stage suffice it to say that for both narrative texts and news reports the notion of free evaluation point can be defined -- although the way it is calculated differs for each.

2.5. Perspective

In section 1, we gave a definition of the notion of perspective in the way we use it here. But the term has been used by scientists in narratology and cognitive science, at least, under interpretations that were not quite identical. In this section we shall further describe our adopted interpretation of the notion, thus identifying our position in the field.

In narratology, the notion relevant to us is focalization (Genette 1980; Bal 1990). Focalization can be used to distinguish the vision presented in a narration from the voice imparting the story to the reader. Both voice and vision may be the narrator's, but they need not be. Rather, the vision imparted need not be the vision of the narrator. It may also be the (reported) vision of a character in the story. Bal (1990) calls this deviation of voice and vision "embedded focalization." Genette (1980) refers to it as "internal focalization." Interpreting the term literally, we may expect to find two important characteristics in (internally) focalized pieces of narration: (1) the truth of the propositions expressed is a "truth in the eyes of the beholder," not necessarily adopted by the narrator; and (2) they may be evaluative, "coloring" the perceived world according to some actor other than the narrator. As Sanders and Redeker (1996) appear to do, we shall mark all expressions establishing or helping to establish one of these characteristics as linguistic signals of perspective.

Sanders and Redeker (1996) propose restricting perspective to embedded or internal focalization. They defend their position by pointing at the so-called "implied author," the ever-present narrator (Booth 1983; Chatman 1978). This narrator is always present in the narration, explicitly or implicitly, by force of his choices (what to say and what to leave unsaid), wording, and (possibly) evaluations. Or, in other words, the narration is always "externally focalized" (Bal 1990). To what extend his/her focalization affects the presentation (or is supposed to do so), the reader does not know.

For our research purpose, the restriction Sanders and Redeker propose (only embedded perspective is called perspective) is tempting. But we do not entirely adopt it for the following reasons. The meaning of perspective in optics would be something like "the way visible objects present themselves from a certain position to the eye." Interpreted in the narrative situation, perspective would be the way objects or events present themselves from a certain position to someone's eye. Could this position have a temporal dimension as well? It seems that perspective may also change for a person over time (cf. Banfield 1982; Fleischman 1990). In cases where the narrator wishes to distance himself from a previous view, he may want to mark a change in perspective (cf. Sanders 1994).

It is a fact that the person of the narrator embodies an always-present filter (cf. Sanders and Redeker 1996), but it is also a fact that the presence of the narrator may or may not be made manifest. The narrator can present the content of the text as subjective -- and thus the validity as restricted -- or he can present the facts as objectively valid. For the latter case we shall reserve the term zero perspective. For the former case we shall use a second notion, mono-perspective.

As we already discussed under the label "embedded focalization," a piece of text can be channelled through only one source (the narrator), but it can also contain elements (interpretations, judgments, evaluations, descriptions) from a source outside the storyteller, and inside the story. We shall say that these text parts have poly-perspective. This label is chosen for its suggestion that a piece of text communicating the attitude of a protagonist is usually a blend of voices, indistinguishably entangled. (33)

The final reason for our threefold distinction within the notion of perspective is simple: we do not want to exclude the narrator entirely from the notion of perspective because we do not want to exclude the possibility that the narrator's subjectivity (or mono-perspective) can be triggered by a free evaluation point. However, our main expectation concerns the relation between a free evaluation point and poly-perspective.

2.6. The hypotheses

If a pluperfect in a sentence in a (narrative or news) text fragment introduces an evaluation point that can be characterized as "free" with the help of information provided by the type of pluperfect used in the target sentence, the tense in the left-hand neighboring sentence, and the interpreted temporal relation between the events of target sentence and left-hand neighboring sentence, we expect this evaluation point to function as a means of introducing perspective.

This hypothesis is text-type independent. Although the value for free evaluation point is obtained differently for both text types, the general hypothesis remains identical. We tend to envisage this generality as an advantage: if the computations of free evaluation points have been executed separately, it seems to us that a relation between free evaluation point and perspectivization irrespective of text type, differences in temporal text properties notwithstanding, only strengthens our hypothesis:

Hypothesis (i)

A free evaluation point introduces perspective.

Note that this first hypothesis is very strict: it does not say "may introduce perspective." Consequently, our prediction is that we shall not find cases in which no perspectivization can be detected, where one of the evaluation points should be characterized as free.

The first hypothesis says that a free evaluation point is sufficient for introducing perspective, but we shall not claim, of course, that it is necessary. In the linguistic literature, many expressions and constructions have been noted to have a perspectivizing function. In order to optimize the chance that the free evaluation point is indeed the factor causing the perspectivization we shall eliminate any other potential source as much as possible. All linguistic expressions and constructions that, to our knowledge, can introduce perspective are analyzed. We call these expressions and constructions perspectivization cues. (34) We distinguish the cases in which perspectivization might be due to perspectivization cues from those in which no such cues occur. Our second hypothesis concerns the joint capacity of free evaluation points and perspectivization cues, thus giving a necessary condition for perspectivization:

Hypothesis (ii)

All cases of perspectivization can be attributed to either a free evaluation point or a perspectivization cue.

3. The research

3.1. The variables

Source. This is the variable used to distinguish the two text types from which the pluperfect may be taken: a narrative (fiction) or a news report.

Free evaluation point. This is a calculated variable. The calculation is based on the values of three variables:

-- type of pluperfect (see section 2.1),

-- tense left (formally determined),

-- interpreted temporal relation (see section 2.3).

In section 3.2 we shall demonstrate how the free evaluation point was derived from these three variables.

Perspective.

1. If neither the narrator nor any character manifests him/herself by means of evaluative statements, expectations, or other attitudinal linguistic signals, the text has neutral or zero perspective;

2. If the narrator manifests him/herself by means of evaluative statements, expectations) or other attitudinal linguistic signals, the text has (at that point) mono-perspective;

3. If the narrator voices evaluative statements, expectations, or other attitudinal linguistic signals of some other character, the text has (at that point) poly-perspective. (35)

Perspective cues. These are the linguistic signals that tell us that the subjective view of either the narrator or some character is presented. In the next section we shall elaborate on the types of perspective cues that we have distinguished. They fall roughly into the following categories: modality of the senses, modality of the mind, typographical signals, or explicit indications of source.

As a second, orthogonally related distinction, we analyzed the position of the cue. Cues may be either in the target sentence or in the (left-hand) context. If the target sentence appeared to be in the scope of a perspective cue elsewhere, it seemed incorrect to label it "no cue." Instead, its type was determined and it was additionally characterized as "outside target sentence."

3.2. Operationalization of the variables

Type of pluperfect. In section 2.1 three interpretations of the Dutch pluperfect were given; these interpretations determine the values of the present variable: PP[F.sub.e], PP[F.sub.cs], and PP. In the same section we introduced tests, based on the literature and adapted to our purpose, to select one of the three possible interpretations for individual pluperfects. The tests are described below. The tests are performed on the target sentences as these occur in the fragments, and they are performed in the order in which they are presented. If, for some occurrence of a pluperfect, the tests taken together did not unequivocally yield a value for type of pluperfect, the occurrence was excluded from the research.

The type of eventuality mentioned in the tests as condition is the aspectual type of the target clause "before" the temporal inflection of the pluperfect makes its contribution to the temporal meaning of the clause. The instantiations of the condition are event, process, or state. These assignments are based on "classical" tests, such as possible type of temporal modification (Vendler 1957; Verkuyl 1972) and (im)possibility of quantification (Dowty 1979; Krifka 1987).

Notice that, as for states, no tests are needed to distinguish the PP[F.sub.e] or the PP interpretation from the PP[F.sub.cs] interpretation: PP[F.sub.cs] is excluded in advance. According to Moens and Steedman (1988), a state does not express any change and hence no consequential state is attached to it (see also section 2.2.3). Consequently, a pluperfect based on a state need only be subjected to the stripping and the confirmation test in order to decide on either a PP[F.sub.e] or a PP interpretation.

The examples given are taken from our corpus. Possibly relevant context is given between square brackets. The words in italics indicate the application of the tests.

i. Temporal location test

Condition: The eventuality of the target sentence is an event process, or state.

Description: Specify the temporal location of the eventuality by means of a temporal adverb with definite past time reference (e.g. gisteren `yesterday', drie jaar geleden `three years ago') to the target sentence.

Result: If the addition is acceptable and does not change the meaning of the target sentence: PP[F.sub.e] or PP. If the addition is unacceptable or changes the meaning of the target sentence: PP[F.sub.cs].

Account: A PP[F.sub.e] directly refers to the eventuality of the target clause; a PP[F.sub.cs] primarily refers to the consequential state of an event. Consequently, in the case of a PP[F.sub.e] or a PP, the eventuality is always in the past of S, and therefore a past location adverb is allowed; in the case of a PP[F.sub.cs], the consequential state and S coincide, hence this state cannot be located in the past of S. See also section 2.1 of this paper.

Examples: a. [De vleugel hing in de lucht en tekende zich als een geblakerde karbonade af tegen de besneeuwde bergtoppen.] Tussen het zwartgelakte hout en de kabels die het instrument omklemden, was ?die ochtend een grijze deken geschoven. [De gele hijskraan torende als een eenarmige stijve reus boven her huis uit en begon langzaam zijn last neer te laten.]

`[The grand piano hung in the sky and it showed like a burned chop against the snowy mountains.] Between the black enamelled wood and the cables that gripped the instrument was put a grey blanket that morning.'

Comment: If acceptable at all, the addition of the adverb die ochtend `that morning' to the target clause changes the reading of this clause. This is evidence for a PP[F.sub.cs] interpretation.

b. Hij wist niet eens zeker wanneer zijn dochter voor het laatst had gegeten. [Vaag herinnerde hij zich een boterham met komkommer.]

`He didn't even know for sure when was the last time his daughter had eaten. [He vaguely remembered a cucumber sandwich.]'

Comment: The use of wanneer voor het laatst `when was the last time' in the target sentence is evidence against a PP[F.sub.cs] reading: the pluperfect is interpreted either as PP[F.sub.e] or as PP.

ii. Al test

Condition: The eventuality of the target sentence is an event.

Description: Put the adverb al `already' in front of the pluperfect.

Result: The addition of the adverb is acceptable and does not change the meaning of the original sentence: PP[F.sub.e] or PP[F.sub.cs].

The addition of the adverb is not acceptable or changes the meaning of the pluperfect considerably: PP.

Account: The al test uses the property of completedness that is characteristic for PP[F.sub.cs] and PP[F.sub.e] -- but not for PP: see section 2.1.

Examples: a. De verhuizers hadden de poten al onder de vleugel geschroefd en vroegen Wanda waar het instrument moest staan.

`The movers had already screwed the legs underneath the grand piano and asked Wanda where the instrument should stand.'

Comment: Since the addition of al `already' does not pose any problem to the interpretation of the target clause, the pluperfect must have either the PP[F.sub.e] or the PP[F.sub.cs] interpretation.

b. [Ze zijn heel streng. 's Avonds fluistert Suze over hoe het is op school. Er was een jongen die zijn pikkie liet zien onder de bank] en iedereen was ?al geschrokken, [maar het was een stompje kaars].

`[They are very strict. At night Suze whispers about how things are at school. There was a boy who showed his willy under the desk] and everyone had ?already had a shock, [but it was a candle stump].'

Comment: In this fragment the addition of al `already' poses a problem to the interpretation of the target clause. From this observation it follows that the pluperfect is to be interpreted as a PP.

iii. Pluperfect-sequence test

Condition: The eventuality of the target sentence is an event.

Description: Invent an additional pluperfect (that expresses an event and, for the sake of coherence, is lexically plausible), and make a conjunction with the target pluperfect. Alternatively, regard a pluperfect that is extant in the immediate context. Determine the temporal relation between the eventualities expressed by the two pluperfects.

Result: The eventualities are interpreted as at least partly following one another: PP.

The eventualities are interpreted as temporally coinciding: PP[F.sub.cs].

No relation between the eventualities can be derived: PP[F.sub.e].

Account: It has been demonstrated by Hinrichs (1986), Kamp and Rohrer (1983), and Partee (1984) that juxtaposed events yield a temporal sequence relation, whereas a state is to be interpreted as temporally overlapping with a neighboring eventuality. This result was obtained for two sentences in the simple past. We expect to find the same pattern for the pluperfect structures PP and PP[F.sub.cs], in which an evaluation point is simultaneous with the eventuality (as is the case in simple past) -- but not for PP[F.sub.e] because in that case there is no immediate relation between the evaluation point and the eventuality.

Examples: `[De interviewer had gevraagd: "Hebt u als auteur grenzen?"] Stront, was door haar hoofd geflitst, [stront vond ze zo vies dat ze er niet over wilde schrijven.]

`[The interviewer had asked: "Do you have limits as a writer?"] Shit had crossed her mind, [shit disgusted her so much that she did not want to write about it.]'

Comment: The eventuality described in the target clause obviously follows the one described in the matrix clause of the preceding sentence: the pluperfect is interpreted as PP.

iv. Stripping test

Condition: The eventuality of the target sentence is a state, event, or process.

Description: To execute the test, strip one "layer of the past" from the contextual temporal discourse structure; simple past will then become simple present. (36) The pluperfect is preferably substituted either by present perfect or by simple past.

Result: If the pluperfect is preferably substituted by present perfect, the interpretation of the pluperfect is PP[F.sub.e] or PP[F.sub.cs]. If the pluperfect is substituted by simple past, the interpretation of the pluperfect is PP.

Account: This test is based on the fact that the pluperfect is to be interpreted either as a perfect-in-the-past or as a past-in-the-past (see sections 2.1 and 2.2.3). In the TTSs (19)-(21), deleting the rightmost S results respectively in structures of the present perfect and the simple past.

Examples: a. De verhuizers hebben de poten onder de vleugel geschroefd/?schroefden de poten onder de vleugel en vragen Wanda waar instrument moet staan. `The movers have screwed / ?screwed the legs underneath the grand piano and ask Wanda where the instrument must stand.'

Comment: As the preference is for substitution with the present perfect instead of the simple past, we conclude that the pluperfect must be understood as a PP[F.sub.e] or a PP[F.sub.cs].

b. [Het vervolg vertelt hij er nooit bij.] De zon ging die avond pas tegen zevenen onder / ?is die avond pas tegen zevenen ondergegaan. Om hem heen klonken de geluiden van een milde voorjaarsavond / ?hebben de geluiden van een milde voorjaarsavond geklonken, [...]

`[He never relates the sequel.] The sun set / ?has set that night not before seven o'clock. Around him rang / ?have rung the sounds of a mild spring night.'

Comment: In both clauses there is a preference for the use of a simple past. This means that the pluperfect is to be interpreted as a PP.

v. Temporal-focus test (confirmation)

Condition: Assumed PP[F.sub.e], PP, or PP[F.sub.cs].

Description: Establish temporal focus on S' by introducing a sentence in the simple past to the left of the target sentence.

Result: PP is confirmed if there is a temporal focus shift from S' to S".

PP[F.sub.cs] is confirmed if there is no temporal focus shift and the focus overlaps with the consequential state of the eventuality expressed in the target sentence.

PP[F.sub.e] is confirmed if there is no temporal focus shift and the focus follows the eventuality expressed in the target sentence.

Account: The test is based on the observed difference in terms of the relation between the temporal focus (S' or S") and the eventuality expressed (see section 2.1).

Examples: a. Confirmation of the interpretation of pluperfect as PP[F.sub.cs]:

De vleugel hing in de lucht. Tussen het zwartgelakte hour en de kabels die het instrument omklemden, was een grijze deken geschoven.

`The grand piano hung in the sky. A grey blanket was put between the black enamelled wood and the cables encircling the instrument.'

Temporal focus remains on S', and it overlaps with the consequential state of the eventuality expressed in the target sentence.

b. Confirmation of the interpretation of pluperfect as PP[F.sub.e]:

Ze stak haar handen in her zeepsop. Haar hele leven had ze haar handen beschermd tegen alles. `She put her hands in the soapy water. Her entire life she had protected her hands against everything.'

Temporal focus remains on S', and it follows the eventuality in the target sentence.

c. Confirmation of the interpretation of pluperfect as PP:

Ze herinnerde het zich nog duidelijk. Haar vader had smalend gelachen toen ze laatst op een avond die bewering in een avondprogramma hadden gehoord. `She remembered it clearly. Her father had laughed contemptuously when they had heard that statement recently in an evening program.' There is a temporal focus shift from S' to S" or, in other words, from the time of remembering to the time of laughing.

Tense left. The tense form of the verb dictates the values for this variable: simple past, simple present, future, pluperfect, present perfect, future perfect, past future and past future perfect.

Interpreted temporal relation (Inttemp). The value of Inttemp provides more detailed information for the exact mapping of the pluperfect structure onto the contextual structure: we infer the temporal relation between the eventuality expressed by the pluperfect (in the case of PP[F.sub.cs], this concerns the consequential state) and the eventuality of the left-hand context. This inference is based on lexical information and real-world knowledge. The relation is possibly already made explicit in the text by a temporal adverb. If no temporal adverb is present, we add one that corresponds to our inference and decide whether it fits well. One of the following relations is attributed to the pluperfect eventuality relative to the left-hand eventuality: completely precedes; partially precedes; coincides; partially follows; or completely follows (Allen 1983).

Free evaluation point. The value of this variable is dependent on source and derived from the values of type of pluperfect, tense left, and interpreted temporal relation in the following way:

I. If source is narrative (qualitative relation between evaluation points in the past):

i. If tense left is a present tense, and the evaluation point(s) of the target sentence is/are required to render the correct interpreted temporal relation between the intervals E1 and E2 (as one of the possibilities), the value for this variable is negative (note that this is never the case when the left-hand sentence is a present perfect). If this is not the case, the value is positive.

ii. If tense left is a past tense and the evaluation points introduced by the left-hand sentence and the target sentence are identified, and if there is an evaluation point S' or S" contributed by the target sentence that is not required to render the correct interpreted temporal relation between the intervals E1 and E2 (as one of the possibilities), the value of the variable is positive. If there is no such evaluation point, the value of the variable is negative.

II. If source is news reports (quantitative relation between evaluation points in the past):

i. If tense left is a present tense, and the evaluation point(s) introduced by the target sentence is/are required to render the correct interpreted temporal relation (possibly indicated by a temporal [adverbial] specification), the value of the variable is negative. If this is not the case, the value of the variable is positive. (37)

ii. If tense left is a past tense, and

a. If the type of pluperfect is PP[F.sub.cs], evaluation point is not free and so the value is negative, independent of the interpreted temporal relation.

b. If the type of pluperfect is PP[F.sub.e] or PP, and both S' and S" are required to render the correct interpreted temporal relation (because it dictates the correct number of positions for E2 to the left of S), the value of the variable is negative. (38,39) If this is not the case, the value is positive.

Perspective. The values of perspective are zero perspective, monoperspective, poly-perspective. They are attributed on the basis of the following tests, based on Palacas (1993).

-- Attribution of responsibility for the information to a person. The person who is responsible for the information presented in the sentence is either explicitly mentioned in the text, see (i), or remains implicit, see (ii).

(i) a. Explicit mention of the author as the responsible source of information: mono-perspective.

b. Explicit mention of some character as the responsible source of information: poly-perspective.

(ii) If no source is explicitly mentioned, the decisive answer is given by using attributive paraphrase techniques:

a. Volgens mij `according to me' + target sentence: interpretation remains the same: mono-perspective; interpretation changes into a restricted view: zero perspective.

b. Volgens personage X `according to character X, + target sentence: poly-perspective.

-- Deniability test. This test makes explicit the restriction of the validity of the presented information.

(i) Target sentence + maar ik geloof het niet / ik weet wel beter `but I don't believe a word of it / but I know better'.

a. The continuation is possible and does not change the meaning of the target sentence: poly-perspective.

b. The continuation is not possible without changing the meaning of the target sentence: zero perspective or mono-perspective.

(ii) Target sentence + maar personage X hoeft dit niet te geloven `but character X may not believe a word of it'.

a. The continuation is possible and does not change the meaning of the target sentence: mono-perspective or zero perspective.

b. The continuation is not possible without changing the meaning of the target sentence: poly-perspective.

Perspectivization cues. The values of this variable register the occurrence of (potential) perspectivization cues other than the pluperfect. We distinguish four categories. The first refers to lexical elements or syntactic constructions expressing modality of the senses, that is, the proposition is presented as being true, but someone is using an evaluating term to express his personal attitude toward the proposition. The second refers to lexical elements or syntactic constructions that express modality of the mind, that is, someone presents his personal attitude as to the (degree of) truth of the proposition. The third category concerns typographical cues such as parentheses, italics, or quotes. The fourth category includes source cues, which are lexical elements that relate the information presented to the source.

Lexical elements or syntactic cues that do not fit into one of the preceding categories but are generally known as perspectivizing elements are collected in the miscellaneous category. (40) For instance, exclamations, interjections, or syntactically incomplete clauses are also considered to have perspectivizing ability but do not form a separate value for this variable (for more cues see Caenepeel 1989; Wiebe 1990a, 1990b).

3.3. Examples

In order to give the reader an impression of the actual analysis of the corpus, we shall present two fully analyzed examples. The first one is given in (33).

(33) Ze zijn heel streng. 's Avonds fluistert Suze over hoe het is op school. Er was een jongen die zijn pikkie liet zien onder de bank en iedereen was geschrokken, maar het was een stompje kaars (A. Enquist, Het Geheim).

`They are very strict. At night Suze whispers about how things are at school. There was a boy who showed his willy under the desk and everyone had had a shock, but it was a candle stump.'

-- Source:

The fragment is taken from a fictional narrative text.

-- Type of pluperfect:

The type of eventuality is an event. It therefore is subject to (i) the temporal location test, (ii) the al test, (iii) the pluperfect-sequence test, (iv) the stripping test, and (v) the paraphrase test.

Test (i) excludes the PP[F.sub.cs] interpretation, because a definite temporal adverb is allowed: Iedereen was die ochtend geschrokken `Everyone had had shock that morning'.

Test (ii) favors the PP-interpretation by its negative outcome: the addition of the adverb al `already' is not appropriate. Test (iii) yields a sequence interpretation, thereby confirming the PP interpretation: Iedereen was geschrokken en alle jongens en meisjes waren weggerend `Everyone had had a shock and all the boys and girls had run away'. Test (iv) again points to PP: in a present context the use of a simple past would be preferable. To conclude, test (v) confirms the PP interpretation. This example was given in the description of the temporal focus test.

-- Tense left:

The tense in the preceding sentence is a simple past.

-- Interpreted temporal relation:

We understand the eventuality of everyone having a shock as having occurred immediately after the boy showed his willy; a causal relation is suggested. Therefore a precedence relation is assumed.

-- Free evaluation point:

The PP interpretation of the pluperfect implies that the temporal focus is on S'; the precedence relation between the left-hand eventuality and the pluperfect eventuality means that S" is next to the relevant evaluation point of the left-hand sentence. This leaves S' of the target sentence temporally unmotivated: S' is a free evaluation point.

-- Perspective:

In applying the attribution test, poly-perspective becomes evident: Volgens hen / haar was iedereen geschrokken `According to what they / she said, everyone had had a shock'. The deniability test gives the same result: Iedereen was geschrokken -- maar in feite schrokken ze helemaal niet (ik weet wel beter) `Everyone had had a shock -- but in fact they were not shocked at all (I know better)'

-- Perspective cues:

If it is not in the immediate left context (but one sentence to the left) there is a perspectivization cue, Suze fluistert `Sue whispers about', indicating the source of information. It is registered as "source cue in context." (41)

The second example is given in (34).

(34) De rechtbank in Amsterdam achtte het merendeel van de beschuldigingen -- overtredingen en misdrijven ten aanzien van de bedrijfsveiligheid en het milieu -- bewezen. Tegen Cindu was een miljoen gulden boete geeist en voorwaardelijke stillegging van het bedrijf. De oorzaak van de ramp was een tikfout op een receptuur voor een harsketel.

`The court of Amsterdam considered the majority of the complaints proven -- trespasses and crimes against company security and the environment. From Cindu a million guilders had been demanded and conditional closing-down of the company. The cause of the disaster was a typo in the proportions for a resin kettle.'

-- Source:

The fragment comes from a news report.

-- Type of pluperfect:

The input type is an event, so this pluperfect is also subject to the tests (i)-(v). The addition of a definite temporal adverb would change the meaning of the target clause: evidence for interpretation of the pluperfect as a PP[F.sub.cs]. It is not a problem to insert the adverb al in front of the pluperfect: PP[F.sub.cs], or PP[F.sub.e]. Applying the sequence test is a matter of adding a pluperfect to the second conjunct. The eventualities expressed in the target sentence, imposing a fine and conditional closing-down, are not interpreted as ordered with respect to each other; they are simply enumerated: PP[F.sub.cs]. Applying the stripping test (iv) the pluperfect is substituted by present perfect: PP[F.sub.cs] or PP[F.sub.e]. The test results so far all share the interpretation of the pluperfect as a PP[F.sub.cs]. Confirmation of this interpretation is given by the temporal focus test: the state of having been demanded coincides with the temporal focus.

-- Tense left:

The tense in the preceding sentence is a simple past.

-- Interpreted temporal relation:

The period of having been demanded and the period of judging are interpreted as overlapping but not coinciding. Obviously, the former will have started before the latter.

-- Free evaluation point:

If the consequential state expressed in the pluperfect and the eventuality in simple past are to hold at S', one step away from S, they will overlap. (The

[presumed] eventuality causing the consequential state expressed in the target sentence must have occurred before the eventuality expressed in the left-hand sentence.) This is the interpreted temporal relation. A PP[F.sub.cs] does not introduce an evaluation point other than S'. Since S' is required to derive the correct interpreted temporal relation, the value for this variable is negative.

-- Perspective:

Zero perspective.

-- Perspective cues:

A perspectivization cue is present in the sentence preceding the target sentences, bewezen achten `consider proven'.

3.4. Description of the corpus

In total, 267 Dutch fragments containing pluperfects were selected: 142 fragments came from several novels and short stories -- representing fictional narrative -- written by Dutch authors (Dorrestein 1997; Enquist 1997; Hemmerechs 1993; Melchior 1993; Reisel 1993; Romijn Meijer 1983; De Winter 1993), and 125 fragments came from a major Dutch newspaper (de Volkskrant). (42) The fragments from the newspaper were considered a news report if they occurred on the front page, domestic page, foreign page, economics page, or (a few) sports page, and they were not editorial in nature.

Restrictions were placed on the position of the target sentence. This sentence had to be surrounded by sufficient context. Most importantly, it could not be a first sentence of a text or even of a paragraph.

Restrictions were placed on the type of target sentence as well: relative clauses containing a pluperfect were not selected because they do not necessarily play a part in the episodic structure of the text. The majority of the pluperfect sentences were either a main clause or a coordinate clause; other categories were complement clauses (5), subject or object clauses (13), or adverbial subclauses (42).

The procedure for analyzing the fragments was as follows: all fragments were analyzed by both authors. The results were compared and if there was no consensus, the fragment was eliminated from the corpus. The results are presented and discussed in section 3.5 following a demonstration of the analysis.

3.5. Results and conclusions

The results are organized as follows. First, we shall present the data immediately relevant to the hypotheses, that is, the tables for free evaluation point in relation to perspective. While interpreting the data we shall demonstrate how they confirm both hypotheses.

Subsequently, we shall follow up a consequence related to the confirmation of the hypotheses. The consequence concerns the distribution of perspectivization cues in narratives and news reports and their effect in general. No hypothesis was formed on this topic, but our results allow us to draw some tentative conclusions. Finally, we shall identify an alternative explanation for the results -- and prove it to be false.

The first results address the first hypothesis:

Hypothesis (i)

A free evaluation point introduces perspective.

Three conclusions can be readily drawn from Table 1:

Conclusion 1. The most striking result appears to be that the number of cases with a free evaluation point but no form of perspectivization was zero, both in narratives and in news reports.

Conclusion 2. In narratives the majority of the cases have perspectivization, even if no evaluation point is free. In most cases this is poly-perspective.

Conclusion 3. In news reports the majority of the cases have no perspectivization if no evaluation point is free. However, 30% of these cases do have perspectivization.

In our investigation, two causes of perspectivization were distinguished: perspectivization cues and free evaluation points. In this table the perspectivization cues have not yet been taken into account. However, we predict that the presence of one of the causes of perspectivization is sufficient. Therefore ALL cases in the rows showing the results for free evaluation point in narrative texts and news texts should display one of the forms of perspectivization. They do, and conclusion 1 therefore supports our hypothesis (i).

According to conclusions 2 and 3, in both text types there are many perspectivized cases in which no free evaluation point is present. This may be due to the presence of a perspectivization cue. In order to be able to comment on the functioning of the free evaluation point as an autonomous cause for perspectivization, we shall have to extract from the table the cases where the second cause, a perspectivization cue, is present.

In Table 2, the data of Table 1 are repeated, but for a restricted set: only for those cases in which no perspectivization cue could be distinguished. The set resulting when all cases having some kind of a perspectivization cue have been abstracted is small. Nevertheless, the results could not be clearer. On the one hand, the cases in which a free evaluation point is present all have perspectivization, both in narratives and in news reports. These results unambiguously confirm hypothesis (i).

On the other hand, NONE of the cases in which there is no free evaluation point has perspectivization, whether in narratives or in news reports. In other words, when perspectivization was found, it could be attributed to either a free evaluation point or a perspectivization cue. So, according to these results, the two causes we distinguished exhaust the range of possible causes and there is no set of perspectivized cases that is still not explained. Hypothesis (ii) is confirmed by these data:

Hypothesis (ii)

All cases of perspectivization can be attributed to either a free evaluation point or a perspectivization cue.

Note that the mono-perspective column in Table 2 is remarkably empty compared to Table 1. (43) This should be interpreted as follows: whereas perspectivization cues may signal mono-perspective, a free evaluation point typically corresponds to poly-perspective. This is as we expected.

Another prediction can be made from conclusions (2) and (3). If hypothesis (ii) is correct, and either a free evaluation point or a perspectivization cue causes perspectivization, we would expect a larger number of perspectivization cues in narratives than in news reports. Table 2 suggests that this is indeed the case and Table 3 confirms this. The table presents the results in absolute numbers since we are interested in the number of cases where perspectivization was caused by a cue in order to compare the scores of the narrative texts with those of the news reports.

The number of perspectivized cases with a perspectivization cue is much greater in narratives than in news reports. The figures in the lower left cells, 9 and 11, can be recognized as the perspectivized cases without cues but with a free evaluation point alone causing perspectivization.

The figures in the upper right cells need explanation. Although we did not present a hypothesis concerning the sufficiency of a perspectivization cue as a cause for perspectivization, the fact that a free evaluation point appeared to be sufficient, together with the fact that a disjunction of both causes appeared to be necessary, may have suggested it to be sufficient. Our results enable us to comment on this issue. It can be concluded from Table 3 that not all cases with a perspectivization cue can be characterized as perspectivized. We have analyzed the specified perspectivization cues in relation to perspective in order to see the differences in perspectivization ability of the cues. In order to be sure that the cue is the only cause for perspectivization, we made a selection of the cases. Only the cases without a free evaluation point were taken into account.

The conclusion is clear: in a vast majority of the cases (95% of 98 cases for narratives and 75% of 44 cases for news reports) the cue cooccurs with perspectivization. Almost all cases of zero perspective combined with a cue can be traced to one type of perspectivization cue: the modality-of-the-senses type. So the cues indicating modality of the senses do not necessarily cause perspectivization. It is not so hard to explain why this should be so: it is often difficult to discriminate between description and subjective predication. For example, "a beautiful girl" can be a girl generally considered beautiful or a girl seen as beautiful through the eyes of the reporter. But for our purpose it is not necessary to explain these "counter-examples" away. On the whole, the conclusion that perspectivization cues autonomously cause perspectivization seems justified. (44)

We introduced the notion of free evaluation point as a notion independent of the type of pluperfect it is associated with: each of the three types of pluperfect had the potential to introduce a free evaluation point. As a final result, we want to present the relations between a free evaluation point and the type of pluperfect, first in a context with perspectivization cues (see Table 4), and second in a context without perspectivization cues (see Table 5). Whereas it is probable that the free evaluation point causes perspectivization in the first case -- Table 2 showed that a free evaluation point always goes together with perspectivization in our corpus -- it is highly probable that it does in the second case.

The fact that almost all cases of free evaluation points causing perspectivization have a PP might suggest that PP was scored simply as an expression of the interpretation of an implicit perspective. The notions "temporal focus" and "free evaluation point" could be two sides of the same coin. If this suggestion is correct, we would expect all PPs to have a free evaluation point and consequently to introduce perspective. Moreover, we would not expect other occurrences of the pluperfect to function as perspectivizers because they would not have a free evaluation point.

The second consequence of the identification of "temporal focus" and "free evaluation point" is falsified by Tables 4 and 5. The first consequence can be proven false as well. The cases with PP (79 in total) have a free evaluation point in 63% of the cases -- and so in 37% no free evaluation point was identified. Moreover, in 13% of the cases the PP did not introduce perspective at all! This implies that the suggestion can be dismissed.

To summarize, the results confirm both the first and the second hypothesis: we did not find any zero perspective cases in which the condition of a free evaluation point was satisfied (hypothesis [i]), nor did we find perspectivized cases in which neither of the conditions, a free evaluation point or a perspectivization cue, was satisfied (hypothesis [ii]).

3.6. Discussion

We started out with the observation that the literature suggests the existence of a tangent plane between tense, that is, the pluperfect, and perspective. In this paper a straightforward strategy for determining the perspectivizing ability of the Dutch pluperfect was developed. The strategy is based on TTT and takes into account both the information given by the immediate context and the restrictions imposed by the text type. Moreover, it makes use of a semantic entity, the free evaluation point, which seems to embody the observed link between tense and perspective. This is a natural option in TTT: the free evaluation point is an imaginary speech point.

The results obtained here support the work of Caenepeel (1989) and Wiebe (1990a) in attributing a perspectivizing ability to the pluperfect. But the conditions allowing for this perspectivizing ability to work are narrowed down as a result of the present research. Not all cases of a PP appeared to be perspectivizing (as might be concluded from Wiebe 1990a), as we concluded in section 3.4. And certainly not all PP[F.sub.cs] occurrences (Caenepeel 1989) proved to be perspectivizing: in 40% of these cases a zero perspective was found. Consequently, identifying an interpretation of the pluperfect as the "cause" of perspectivization is too general. The more refined notion of free evaluation point seems more appropriate.

There are a few striking advantages in attributing the perspectivizing ability of the pluperfect to a special type of evaluation point. For one, generalizing over the results obtained here, we might assume that the method of computing the "motivational status" of evaluation points generated in a temporal structure is applicable to other tense forms as well. For example, the use of a simple past in a present tense context. Since the presence of a free evaluation point is not restricted per se to the pluperfect, we now have an instrument for computing the perspectivizing force of tense in general.

Second, there is no reason to assume that the perspectivizing function of a free evaluation point is a specific characteristic of the Dutch language. It seems that it would be worthwhile to test the hypothesis on other languages as well.

We will be in a position to check the perspectivizing power of tenses in different text types when we have a theory on the temporal discourse mechanisms at work in those text types. In sum, the tiny link between tense and perspective, embodied in an evaluation point, may prove to be applicable to other tenses, contexts, and languages.
Table 1. Free evaluation point in relation to perspective
for two text types (percentages)

 Zero Poly- Mono-
 perspective perspective perspective

Narratives
 free evaluation point 0 93 7
 (n = 30)
 no free evaluation point 17 62 21
 (n = 112)

News reports
 free evaluation point 0 72 28
 (n = 24)
 no free evaluation point 70 9 21
 (n = 101)
Table 2. Free evaluation point in relation to perspective for a
restricted set: no perspectivization cue present (percentages)

 Zero Poly Mono-
 perspective perspective perspective

Narratives
 free evaluation point 0 100 0
 (n = 9)
 no free evaluation point 100 0 0
 (n = 12)

News reports
 free evaluation point 0 91 9
 (n = 11)
 no free evaluation point 100 0 0
 (n = 57)
Table 3. Perspectivization in relation to perspectivization cues
for two text types (numbers)

 No
 perspectivization Perspectivization
 cue cue

Narratives
 no perspectivization 13 6
 perspectivization 9 114

News reports
 no perspectivization 57 10
 perspectivization 11 47
Table 4. Types of pluperfect occurring with free evaluation point
(percentages)

 PP[F.sub.e] PP[S.sub.cs] PP

Narratives (n = 31) 10 0 90
News reports (n = 25) 4 8 88
Table 5. Types of pluperfect occurring with free evaluation point
and no perspectivization cues (percentages)

 PP[F.sub.e] PP[F.sub.cs] PP

Narratives (n = 9) 0 0 100
News reports (n = 11) 0 9 91


Tilburg University
Received 30 May 2000
Revised version received
12 July 2001


Notes

* Correspondence address: Leonoor Oversteegen, Tilburg University, Discourse Studies Group, P.O. Box 90153, 5000 LE Tilburg, The Netherlands. E-mail: E.Oversteegen@Kub.nl. Birgit Bekker, Tilburg University, Discourse Studies Group, P.O. Box 90153, 5000 LE Tilburg, The Netherlands. E-mail: BBekker@Kub.nl.

(1.) The pluperfect may be seen as a combination of a (past) tense and a perfect aspect (see Introduction to section 2). However, we shall refer to it as a (complex) tense. In section 2.1 we discuss the possible interpretations of this complex tense in Dutch.

(2.) At this point it is sufficient to know that the approach presented in this paper is based on the assumption that the pluperfect denotes either a "perfect-in-the-past" or "past-in-the-past," as was proposed by Jespersen (1924). Caenepeel (1989) relates the perspectivizing power of the pluperfect to the former interpretation, whereas Wiebe (1990a) relates it to the latter interpretation. Discussion of this issue is taken up in section 2.1.

(3.) To be precise, the reference point R relates to the speech point S, whereas the event point E relates to R. So E is only indirectly related to S.

(4.) These two types of clause are distinguished by the notion of telicity in Boogaart (1999: 87). As he argues, this notion may well be deployed as the parameter contrasting the class of "events," which have the property [+telic] and thereby are related to the (Vendler 1957) classes of "accomplishments and achievements," with "states," which are [-telic] and relate to Vendlers' "states and activities."

(5.) In narratives there appear to be exceptions to the principle of forward movement of the point of reference, i.e. the reversed-order phenomenon (Caenepeel and Moens 1994; Lascarides and Asher 1991). More importantly, the anaphoricity principle has been relativized with respect to text type in general (Caenepeel 1991, 1993; Oversteegen 1993). We shall return to this issue in section 2.4.

(6.) In the literature we find both events (Hinrichs 1981, 1986) and reference points (R-chaining, as in Couper-Kuhlen 1987: 22) as anaphors. We have to opt for the latter considering the fact that we do not treat E and R as essentially of the same type (as Hinrichs does).

(7.) We do not distinguish kinds of reference points, as Kamp and Reyle (1993) do. In section 2.2 we will even replace the term "reference point" by the more general notion of "evaluation point" (as used in Oversteegen 1989). The introduction of evaluation points by the pluperfect is described in section 2.2.3.

(8.) In this paper we do not make any claims about the English language. The examples are translated into English for the sole purpose of understanding. Only numbered examples in English are meant as examples for the English language.

(9.) In terms of Reichenbach, tense concerns the relation between the reference point(s) and the speech point, and aspect concerns the relation between the eventuality and the reference point.

(10.) We do not pretend to give an exhaustive (or nearly exhaustive) argumentation on Dutch and English readings of the present perfect. This can be found in Boogaart (1999) and Korrel (1991) (among others). The latter ascribes the different interpretations of the Dutch and the English present perfect (or "transpast transcendent," in her terms) to the different choices made for the interpretation of the evaluation point. It is seen either as the endpoint of the already actualized stretch of duration or as the starting point of the yet to be actualized stretch of duration (Korrel 1991: 42).

(11.) In the Introduction to section 2 it was mentioned that Reichenbach considered E to be a point in time. It should be noted that we go with the representation of this notion as a time interval, as was proposed by Bennet and Partee (1978), Dowty (1979), Kamp (1979), and Tichy (1985), among others. See also section 2.2.

(12.) An example of the formal treatment of this interpretation of the perfect can be found in Landeweerd (1998), who defines the construction rules for the temporal discourse representation structures of the French tenses. A more functionally oriented elaboration of the interpretational effects of the consequential state reading of the (plu)perfect is given in Caenepeel (1989, 1993, 1995).

(13.) Consequential states should not be confused with consequences. Of course, other eventualities can be consequentially related to states, for example in "It is cold. I am going to fetch my coat." It would not be correct, however, to interpret the eventuality "fetching one's coat" as the consequential state of the weather being cold (see also Boogaart 1999: 141).

(14.) See also Korrel (1991: section 3.3). In contrast, Binnick (1991) observes for English an ambiguity in "Max has gone." He states, "[...] it can imply either `Max (has gone and) continues to be gone', or `Max (has gone at some time in the past but) has subsequently returned'" (Binnick 1991: 100). This implies that (7c) would be acceptable, if we replace the present tense in the subclause by a present perfect.

(15.) See also Matthews (1987).

(16.) Boogaart claims that the English present perfect may have a reading "in which the situation is assumed to have ended somewhere before S and the state holding at S cannot be identified with any part of the past situation" (Boogaart 1999: 149). This is, in his words, the right-bounded reading of atelic predicates. In our view, the intended reading corresponds to the existential or experiental perfect -- which is not in contradiction with our interpretation of the English present perfect.

(17.) Boogaart (1999) distinguishes three readings for the present perfect -- which in principle can occur in both English and Dutch. He refers to them as (1) resultative reading; (2) right-bounded reading; and (3) continuative reading. The first reading corresponds to the consequential state reading. The second reading is equivalent to our formulation that the eventuality is wholly anterior to the reference point -- and we argued that this is a very special reading in English, which implies quantification over the eventuality (see also note 15). The third reading is only possible in Dutch if the eventuality is bounded to the right by the moment of speech (see Boogaart 1999: 146, 151).

(18.) Korrel (1991) argues that the interpretation of the Dutch transpast transcendent (= present perfect) is equivalent, in our terms, to the consequential state reading of the perfect. But some of the examples she presents are, in fact, good examples of what we call the past-eventuality reading of the perfect. In example (103) (Korrel 1991: 146): "Bij dat meisje aan de overkant hebben ze een naald uit haar voet gesneden, eng he? `They took a needle out of the foot of that girl, that gives you the creeps, doesn't it?'," the perfect does more than merely make an event inferrable. The event is referred to, which is corroborated by the use of eng he? `that gives you the creeps, doesn't it?' And in example (55) (Korrel 1991: 59): "Lieve help! Wat een troep! Gisteren heb ik de boel nog helemaal opgeruimd en gestofzuigd. En moet je nu eens zien! `Good grief! What a mess! I cleaned and vacuumed everything only yesterday! And now look at it'," the consequential state of the event obviously does not hold at the reference point.

(19.) It is worthwhile noting that this is also possible in English (Comrie 1985; Michaelis 1998).

(20.) Poesio and Kameyama (1992) see the Reichenbachian reference points as potential temporal foci or, in their terms, centers. We agree with them: if a sentence focuses on a time coinciding with E, we assume a reference point there (see the following paragraph on the past-in-the-past). As is the case in the distinction between present perfect and simple past, an R marks the temporal location that we focus upon.

(21.) To a great extent, context is decisive in determining the temporal focus of an individual sentence. Therefore, focus can never be used as the sole criterion for selecting a reading of the pluperfect.

(22.) A similar analysis of the ambiguous nature of the pluperfect can be found for English in McCawley (1971) and Michaelis (1998).

(23.) In fact, we very often find the sequence PP[F.sub.e]-PP. This should not be surprising as it is the past variant of the well-known sequence present perfect-simple past. The perfect functions as the introduction of an eventuality and the simple past elaborates on it.

(24.) This means of discriminating between perfect-in-the-past and past-in-the-past can also be found in Vlach (1981) and Caenepeel (1993).

(25.) Another explanation, in terms of pragmatic presuppositions, can be found in Michaelis (1998).

(26.) Motivation for the two-track model -- both in terms of conceptual adequacy and in terms of functionality -- is given in Oversteegen (1989).

(27.) That is, the temporal focus. If there is only S, it is S; if there is both S and S', it is S'; if there is S, S', and S", it is S".

(28.) At this stage we shall not comment on the perspective in the examples. Our prediction, however, would be that a perspective is introduced.

(29.) Another example corresponding to (25) coming from our corpus would be Hij vroeg haar toen waar ze haar inspiratie vandaan haalde en ze was geschrokken van de vraag `Then he asked her where she got her inspiration and she had given a start at the question'.

(30.) We shall use the term narrative to refer to fiction texts that create an imaginary reality. News reports do not belong to the same type: the function of the story-telling instance and its relation to the world (creating or reporting) differs fundamentally for the two types of text.

(31.) See also Dowty (1982), Cooper-Kuhlen (1987), Kamp and Reyle (1993).

(32.) Michaelis (1998, section 4): "the PaP [pluperfect] indicates a greater degree of displacement from speech time, in the direction of the past, than does the preterite [...]." See also Comrie (1985: section 4).

(33.) The notion polyphony, introduced by Ducrot (1984), appears to resemble our notion poly-perspective. We chose not to use polyphony since Ducrot applies it to utterances that would not be poly-perspectival in our view. For example, the canonical example for polyphony is a negated sentence like `This wall is not white'. It would be a refutation of the contrary -- and therefore polyphonous. The second voice, claiming the positive sentence, is not present in the negated sentence; therefore we do not analyze it as poly-perspectival.

(34.) We shall not refer to a tree evaluation point as a perspectivization cue: demonstrating that it may function as such is the purpose of this paper.

(35.) Calling the latter situation a perspective shift would not illustrate the fact that a piece of text communicating the attitude of a protagonist usually represents a blend of voices.

(36.) Present perfect forms, however, will stay present perfect, just like simple present. So, in fact, the test is restricted to a past tense context.

(37.) Note that the value can be negative here even if the tense left is a present perfect. The quantitative determination of the relation between evaluation points causes the E2 introduced by the target sentence to be placed one position to the left of S' on the E-track and two positions left of S. Consequently it precedes E1, which is only one position left of S. Only in the case of a PP[F.sub.cs] is there no precedence relation: the value of the variable is positive in this case.

(38.) This is precedence in the case of a simple past tense left and overlap or immediately after (narration) in the case of a pluperfect tense left. If the interpreted temporal relation is overlap and tense left is simple past the value of the variable is positive.

(39.) One might object that the difference between PP[F.sub.e] and PP is not in the number of positions: both pluperfect forms define the eventuality as being placed two positions away from S. Consequently they would place the eventuality similarly, and so there would always be one free evaluation point with a PP. This is not correct, though. As mentioned earlier, the difference between the two forms is one of temporal focus. PP has the temporal focus at the time of E -- and consequently on S" -- and PP[F.sub.e] has the temporal focus on S'. Evidence that the focus is on S" (e.g. because this point is specified in a certain way) is sufficient for assuming and motivating S". In the case of a PP, often it is not S" but S' that is free. If tense left is simple past and the interpreted temporal relation between the Es is overlap, the S' point contributed by the target sentence is not temporally motivated and, consequently, is free.

(40.) The number of elements or constructions belonging to a specific category was too small to constitute a separate value.

(41.) Note that the pluperfect sentence, having a perspectivized interpretation, lacks perspectivization cues at the local level. The cue we register in the context, in fact, confirms (not enforces) this perspectivized interpretation.

(42.) We did not keep track of the occurrence of pluperfect in relation to the use of other tense forms. But we did, of course, notice that the pluperfect is used abundantly in narratives, whereas authors of news reports use this form more sparingly.

(43.) The only occurrence of a free evaluation point cooccurring with mono-perspective appeared to be a counterfactual sentence.

(44.) This conclusion may seem circular: an expression or construction is labelled as a perspectivization cue because it seems to perspectivize (part of) the text. For this reason we did not "upgrade" this relation to a hypothesis and it is not central to our argument.

References

Allen, James (1983). Maintaining knowledge about temporal intervals. In Communications of the ACM, D. Waltz (ed.), volume 26 (11). New York: Association for Computing Machinery.

Bach, Emmon (1980). Tenses and aspects as functions of verb-phrases. In Time, Tense and Quantifiers, C. Rohrer (ed.), 10-39. Tubingen: Niemeyer.

Bal, Mieke (1990). De Theorie van Verhalen en Vertellen: Inleiding in de Narratologie. Muiderberg: Coutinho.

Banfield, Ann (1982). Unspeakable Sentences: Narration and Representation in the Language of Fiction. Boston: Routledge and Kegan Paul.

Bennet, Michael; and Partee, Barbara (1978). Toward the Logic of Tense and Aspect in English. Bloomington: Indiana University Press.

Binnick, Robert (1991). Time and the Verb: A Guide to Tense and Aspect. Oxford: Oxford University Press.

Boogaart, Ronny (1999). Aspect and temporal ordering: a contrastive analysis of Dutch and English. Unpublished dissertation, Free University of Amsterdam.

Booth, W. (1983). The Rhetoric of Fiction. Chicago: University of Chicago Press.

Caenepeel, Mimo (1989). Aspect, temporal ordering and perspective in narrative fiction. Unpublished dissertation, University of Edinburgh.

--(1991). Event structure versus discourse coherence. In Proceedings of the DANDI Workshop on Discourse Coherence (Centre for Cognitive Science, University of Edinburgh), M. Caenepeel, J. L. Delin, E. Oversteegen, G. Redeker, and J. Sanders (eds.), 1-14. Edinburgh: ESPRIT Project.

--(1993). Aspect and Text Structure. Deliverable R1.3.1, Part I, Esprit Basic Research Project 6665: DANDELION. University of Edinburgh/CLS/CEC.

--(1995). Aspect and text structure. Linguistics 33, 213-254.

--; and Moens, Marc (1994). Temporal structure and discourse structure. In Tense and Aspect in Discourse, C. Vet and C. Vetters (eds.), 5-20. Berlin: De Gruyter.

--; and Sandstrom, Gorel (1992). A discourse-level approach to the past perfect in narrative. In Proceedings of the 4th International Workshop on Semantics of Time, Space and Movement and Spatio-Temporal Reasoning (University of Toulouse), M. Aurnague, A. Borillo, M. Borillo, and M. Bras (eds.), 167-182. Toulouse: Groupe Langue, Raisonnement, Calcule.

Chatman, S. (1978). Story and Discourse. Narrative Structure in Fiction and Film. Ithaca, NY: Cornell University Press.

Comrie, Bernard (1976). Aspect. Cambridge: Cambridge University Press.

--(1985). Tense. Cambridge: Cambridge University Press.

Couper-Kuhlen, Elisabeth (1987). Temporal relations and reference time in narrative discourse. In Essays on Tensing in English, vol. 1, A. Schopf (ed.), 7-25. Tubingen: Niemeyer.

De Winter, Leon (1993). Shabbat. In De Tien Geboden, Bas Heijne (ed.), 151-170. Amsterdam: Prometheus.

Dorrestein, Renate (1997). Want dit is mijn lichaam. Stichting Collectieve Propaganda van bet Nederlandse Boek. Amsterdam: Contact.

Dowty, David (1979). Word Meaning and Montague Grammar. Dordrecht: Reidel.

--(1982). Tenses, time adverbs and compositional semantic theory. Linguistics and Philosophy 5, 23-55.

Ducrot, Oswald (1984). Le Dire et le Dit. Paris: Editions de Minuit.

Enquist, Anna (1997). Het Geheim. Amsterdam and Antwerp: Arbeiderspers.

Fleischman, Susanne (1990). Tense and Narrativity: From Medieval Performance to Modern Fiction. Austin: University of Texas Press.

Genette, G. (1980). Narrative Discourse. Oxford: Blackwell.

Grice, H. P. (1975). Logic and conversation. In Syntax and Semantics 3: Speech Acts, P. Cole and J. L. Morgan (eds.), 41-58. New York: Academic Press.

Grosz, Barbara; Joshi, A. K.; and Weinstein, S. (1995). Centering: a framework for modeling the local coherence of discourse. Computational Linguistics 21, 203-225.

Gundel, Jeanette; Hedberg, Nancy; and Zacharski, R. (1993). Cognitive status and the form of referring expressions in discourse. Language 69, 274-307.

Hemmerechs, Kristien (1993). Ge[??]nterviewd. In De Tien Geboden, Bas Heijne (ed.), 133-150. Amsterdam: Prometheus.

Hinrichs, Erhard (1981). Temporale Anaphora im Englischen. Unpublished Zulassungsarbeit, Universitat Tubingen.

--(1986). Temporal anaphora in discourses of English. Linguistics and Philosophy 9, 63-82.

Jespersen, Otto (1924). The Philosophy of Grammar. London: Allan and Unwin.

Kameyama, Meguni; Passonneau, Rebecca; and Poesio, Massimo (1993). Temporal centering. In Proceedings of the 31st Annual Meeting of the Association of Computational Linguistics, 70-77. Columbus, OH: Association for Computational Linguistics.

Kamp, Hans (1979). Events, instants and temporal reference. In Semantics from Different Points of View, R. Bauerle, U. Egli, and A. Von Stechow (eds.), 376-416. Berlin: Springer.

--; and Reyle, Uwe (1993). From Discourse to Logic. Dordrecht: Kluwer.

--; and Rohrer, Christian (1983). Tense in texts. In Meaning, Use and Interpretation of Language, R. Bauerle, C. Schwarze, and A. Von Stechow (eds.), 250-269. Berlin: De Gruyter.

Korrel, Lia (1991). Duration in English: A Basic Choice, Illustrated in Comparison with Dutch. Berlin: Mouton de Gruyter.

Krifka, Manfred (1987). Nominal reference and temporal constitution: towards a semantics of quantity. In Proceedings of the Sixth Amsterdam Colloquium, 153-173. Amsterdam: Institute for Language, Logic and Information.

Landeweerd, Rita (1998). Discourse semantics of perspective and temporal structure. Unpublished dissertation, University of Groningen.

Lascarides, Alex; and Asher, Nicholas (1991). Discourse relations and defeasible knowledge. In Proceedings of the 29th Annual Meeting of the Association for Computational Linguistics, 55-62. Berkeley, CA: Association for Computational Linguistics.

--; and Asher, Nicholas (1992). The pluperfect in narrative discourse. In Proceedings of the 4th International Workshop on Semantics of Time, Space and Movement and Spatio-Temporal Reasoning (University of Toulouse), M. Aurnague, A. Borillo, M. Borillo, and M. Bras (eds.), 250-259. Toulouse: Groupe Langue, Raisonnement, Calcule.

Matthews, Richard (1987). Present perfect tenses: towards an integrated functional account. In Essays on Tensing in English, vol. 1, A. Schopf (ed.), 111-176. Tubingen: Niemeyer.

McCawley, James (1971). Tense and time reference in English. In Studies in Linguistic Semantics, C. Fillmore and T. Langendoen (eds.), 96-113. New York: Holt Rinehart.

--(1981). Notes on the English present perfect. Australian Journal of Linguistics 1, 81-90.

McTaggart, J. M. E. (1927). The unreality of time. Mind 17, 457-481.

Melchior, Willem (1993). Alleen nog maar stilte. In De Tien Geboden, Bas Heijne (ed.), 107-120. Amsterdam: Prometheus.

Michaelis, L. A. (1998). Aspectual Grammar and Past-time Reference. London: Routledge.

Moens, M.; and Steedman, M. (1988). Temporal ontology and temporal reference. Computational Linguistics 14, 15-28.

Oversteegen, Eleonore (1989). Tracking time: a proposal for the representation of temporal expressions in Dutch. Unpublished dissertation, University of Utrecht.

--(1993). Aspect and Text Structure. Deliverable R1.3.1, Part II, Esprit Basic Research Project 6665: DANDELION. University of Edinburgh/CLS/CEC.

Palacas, Arthur (1993). Attribution semantics: linguistic worlds and point of view. Discourse Processes 16, 239-277.

Partee, Barbara (1984). Nominal and temporal anaphora. Linguistics and Philosophy 7, 243-286.

Poesio, Massimo; and Kameyama, Megumi (1992). Temporal centering. Unpublished manuscript.

Prince, Ellen (1982). Toward a taxonomy of given--new information. In Radical Pragmatics, P. Cole (ed.), 223-255. London: Academic Press.

Reichenbach, Hans (1947). Elements of Symbolic Logic. London: MacMillan.

Reisel, Wanda (1993). Water en wijn. In De Tien Geboden, Bas Heijne (ed.), 151-170. Amsterdam: Prometheus.

Romijn Meijer, Henk (1983). Mijn naam is Garrigue. Amsterdam: Meulenhof.

Salkie, Raphael (1989). Perfect and pluperfect: what is the relationship? Journal of Linguistics 25, 1-34.

Sanders, Jose (1990). Explicit of niet? Referentie-bepalende factoren bij personen in nieuwsberichten. Interdisciplinair Tijdschrift voor Taal- en Tekstwetenschap 9, 159-180.

--(1994). Perspective in narrative discourse. Unpublished dissertation, university of Tilburg.

--; and Redeker, Gisela (1996). Perspective and the representation of speech and thought in narrative discourse. In Spaces, Worlds and Grammar, G. Fauconnier and E. Sweetser (eds.), 290-318. Chicago: University of Chicago Press.

Sandstrom, Gorel (1993). When-clauses and the temporal interpretation of narrative discourse. Unpublished dissertation, University of Umea.

Schopf, Alfred (1984). Das Verzeitungssystem des Englischen und seine textfunktion. Tubingen: Niemeyer.

Tichy, Pawel (1985). Do we need interval semantics? Linguistics and Philosophy 8, 199-220.

Ushie, Y. (1986). "Corepresentation" -- a textual function of the indefinite expression. Text 6, 427-446.

Vendler, Zeno (1957). Verbs and times. Philosophical Review 66, 143-166.

Verkuyl, Henk (1972). On the Compositional Nature of the Aspects. Dordrecht: Reidel.

Vlach, Frank (1981). The semantics of the progressive. In Syntax and Semantics 14: Tense and Aspect, P. J. Tedeschi and A. Zaenen (eds.), 271-292. New York: Academic Press.

Webber, Bonny L. (1988). Tense as a discourse anaphor. Computational Linguistics 14, 61-73.

Wiebe, Jeanette (1990a). Recognizing subjective sentences: a computational investigation of narrative text. Unpublished dissertation, State University of New York.

--(1990b). Identifying subjective characters in narrative. Proceedings Coling 90, 401-406. Morristown, NJ: Association for Computational Linguistics.
COPYRIGHT 2002 Walter de Gruyter GmbH & Co. KG
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2002 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Oversteegen, Leonoor; Bekker, Birgit
Publication:Linguistics: an interdisciplinary journal of the language sciences
Geographic Code:1USA
Date:Jan 1, 2002
Words:20372
Previous Article:Does MANNER count or not? Manner-of-motion verbs revisited *.
Next Article:Self-mockery in Japanese (1).
Topics:


Related Articles
Check-the-box: European options, U.S. and European consequences.
Mobile Insights Announces First Annual Mobile Perspectives '99 Forum on the US Mobile Computing Market to be held for Asian Pacific System...
Computational complexity and the acquisition of functional categories by French-speaking children with SLI (1).

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters