2. LITERATURE REVIEW

Benveniste (Reference Benveniste1966) laid the foundations of the linguistic study of subjectivity by saying that the speaker is the centre of the communicative act, and languages are organised around the speaker: the first-person pronoun I, and the speaker’s now and here. With respect to speaker’s now, Benveniste argues that some verbal tenses (the simple past) should be excluded from subjective discourses, where others (the present, the past compound and the future) should not be used in objective discourses. However, other studies have argued that this claim is too strong (e.g. Banfield, Reference Banfield1982; Fleischmann, Reference Fleischman1990; Sthioul, Reference Sthioul, Moeschler, Jayez, Luscher, de Saussure and Sthioul1998, Reference Sthioul2000; Tahara, 2000; Reference Tahara2004; Saussure, Reference de Saussure, Jaszczolt and de Saussure2013; Moeschler, Reference Moeschler and Pavelin Lesic2014; Reference Moeschler2019), and that verbal tenses occurring in the non-prototypical type of discourse provide access to the speaker’s perspective. The question of the link between verbal tenses and speaker’s subjectivity has been widely debated from a variety of approaches. In this work, we focus on those carried out in literary criticism, for which the perspectival usages of verbal tenses are linked to the Free Indirect Discourse (FID),Footnote ⁴ and those within a linguistic and pragmatic framework, for which verbal tenses may have perspectival usages independently of FID.

Studies in literary criticism have drawn on the distinction between an experiencing self and a narrating self, illustrated by the pair of examples in (2) and (3). The difference between the PS in (2) and the IMP in (3) manifests in the objective character of the description of the seeing of the moon (with the PS) provided by a narrating self, compared to the subjective character of the description of the same event as experienced at a certain moment by an experiencing self (with the IMP) (Banfield, Reference Banfield1982; Fleischman, Reference Fleischman1990, Reference Fleischman, Fleischman and Waugh1991).

1. (2) Elle vit la lune.

   ‘She saw the moon.’

2. (3) Elle voyait la lune (maintenant).

   ‘She saw the moon (was seeing the moon now).’

For Banfield (Reference Banfield1982), in FID (containing represented speech and thought), sentences with the IMP are subjective and they integrate the consciousness of an experiencing character within the description of a series of situations. They express, thus, an individual’s evaluations, emotions, judgments, uncertainties, beliefs and attitudes. As Fleischman (Reference Fleischman1990) puts it, the simple past is non-experiential and entirely detached from the speaker. Nevertheless (as we will discuss below), scholars (Reboul, Reference Reboul1992; Vuillaume, Reference Vuillaume1990; Tahara, Reference Tahara2000) pointed out that verbal tenses may have subjective usages in discourse types that are not FID, such as in (4). In the following fragment, the italicized verbs are in the PS and they express the advancement of time seen from Emma’s point of view (she was terrified and exhausted). The last verb, étouffaient, which is in the IMP, completes the description of Emma’s subjective perceptions and feelings.

1. (4) – Monsieur vous attend, madame, la soupe est servie. Et il fallut descendre! Il fallut se mettre à table! Elle essaya de manger. Les morceaux l’étouffaient. (Flaubert, Madame Bovary)

   ‘– Sir (Charles) is waiting for you, madam; the soup is served. And she had to go downstairs! She had to sit to the table! She tried to eat. The bites of food suffocated her.’

In linguistics, to explain the difference between the IMP and the PS, two types of accounts have been put forward: the semantic approach to the contextual values of verbal tenses (such as Desclés, Reference Desclés1990, Reference Desclés2017; Gosselin, Reference Gosselin1996, Reference Gosselin2005) and the perspective-taking approach (point de vue in French) to grammatical aspect (Comrie, Reference Comrie1976; Caenepeel, Reference Caenepeel1989; Fleischman, Reference Fleischman, Fleischman and Waugh1991, Reference Fleischman, Bybee and Fleischman1995; Smith, Reference Smith1991). In the first type of account, Desclés’ (Reference Desclés1990, Reference Desclés2017) semantic model provides fine-grained aspectual and temporal distinctions to describe the semantics of French verbal tenses and their contextual values (“effets de sens”). According to Desclés’ model, the semantics of the PC, PS and IMP needs to be described using the following components: lexical aspect (states, events and processes; cf. also Mourelatos, Reference Mourelatos, Tedeschi and Zaenen1981; Verkuyl, Reference Verkuyl1993, Reference Verkuyl2008), temporal frame of reference (the enunciator’s frame of reference vs. the speaker’s/external frame of reference vs. a narrative frame of reference), intervals and boundaries (open vs. closed situation boundaries) and semantic invariant (the aspectual and temporal semantic content, common to all contextual values that a verbal tense may have). As such, the IMP expresses a temporal differentiation (oriented towards the past) between an unaccomplished situation and the enunciator’s frame of reference while taking either a (stable) state value or an (evolutive) process value. The PS always expresses an event as disconnected from the enunciator’s frame of reference, and it is in contrast to the PC, which expresses either an accomplished event differentiated from the enunciator’s frame of reference or a (resultative) state successive to an accomplished event. In fact, according to Desclés’ model, the whole paradigm of the use of verb tenses in a text depends on the frames of reference. With the PS, for example, there is a change from the enunciator’s frame of reference towards the narrative frame of reference, knowing that the enunciator’s frame is detached from the speaker’s/external frame of reference.

In the second type of account, the distinction between the IMP and the PS comes from the fact that the former is an imperfective verbal tense while the latter is a perfective verbal tense.Footnote ⁵ According to this line of thought, imperfective verbal tenses which take an “internal perspective”, focusing on a sub-interval of an ongoing event, point to the speaker’s experiencing mind, and are therefore associated with subjective interpretations of utterances. Perfective verbal tenses, on the other hand, present the situation from an external perspective by offering a perception of the situation as a whole, and are thus associated with objective interpretations of utterances. According to Fleischman, the alternation PS/IMP has the pragmatic function to signal the change of perspective: “from an external perspective of the narrator to the inner speech and thoughts of the focalizing characters” (Fleischman, Reference Fleischman, Fleischman and Waugh1991: 35). Nevertheless, as Bres (Reference Bres2003) points out, this hypothesis of verbal tenses alternation is too general since in many cases the alternation PS/IMP does not involve a change of perspective, and changes of perspective may arise in the absence of the PS/IMP alternation. In fact, the predicted associations for perfective vs. imperfective verbal tenses and perspective have not found much support in the scant empirical studies to date (Trnavac, Reference Trnavac2006; Grisot, Reference Grisot2017). For Bres (Reference Bres2003), both the notion of grammatical aspect and of perspective have a role to play in how verbal tenses are used in the discourse. Precisely, grammatical aspect takes action at an early stage of the production of an utterance: it determines whether the eventuality is to be viewed as ongoing or as accomplished. The (subjective) perspective, which is optional, is built later and in interaction with the representation given by the verb itself, by grammatical aspect and tense, and by the context among others. As such, the notion of perspective is relevant with respect to the functioning of verbal tenses at the discursive level: it is a meaning effect or a contextual value drawing on grammatical aspect and on the context.

Pragmatic studies argued that most French verbal tenses have perspectival usages,Footnote ⁶ i.e. usages in which a subjective perspective is expressed, namely where there is a change in how they locate eventualities in time. Some of the best-known perspectival usages of French verbal tenses are the perspectival PS (1), futurate PC (5), future Présent (6), historical Présent (7) and narrative IMP (8).

1. (5) Demain j’ai fini mon article.

   ‘Tomorrow I will have finished my paper.’

2. (6) Je pars demain.

   ‘I am leaving tomorrow.’

3. (7) De 1917 à 1921 Jean Galmot vécut ses années les plus ardentes, les plus riches en péripéties, les plus dramatiques aussi. C’est l’ascension. On le laisse faire. Son succès surprend. (Moeschler, Reference Moeschler2019 from Jean Galmot)

   ‘Between 1917 and 1921 Jean Galmot lived his most intense years, the fullest of adventures, and the most dramatic. It is Holy Thursday. They let him do his job. His success was surprising.’ (our translation)

4. (8) Le lendemain, Paul partait. (Saussure, Reference de Saussure, Jaszczolt and de Saussure2013)

   ‘The next day, Paul left.’

In each of these four cases, there is a semantic tension between the temporal adverbial (a deictic expression) preceding the verb and the basic semantics of the verbal tense, a tension which is resolved semantically, textually or pragmatically depending on the scholars. According to scholars adopting the semantic answer, the solution would consist in a type of semantic accommodation of the semantics of verbal tenses in relation to the interval at which the situation is true (e.g. following Dowty’s Reference Dowty1979 proposal for futurate progressives, as in (6)). Scholars adopting a textual approach suggest that in opaque contexts (in the sense of Quine, 1960), in which there is an ambiguity relative to whose point of view is relevant in determining the truth-condition of a proposition (that of the speaker or of another person), temporal deictics signal a change of perspective: from that of the speaker to that of another person (called mediator, using Desclés and Guentchéva’s, Reference Desclés, Guentchéva, Monod-Becquelin and Erikson2000 term; cf. discussion in Apothéloz, Reference Apothéloz, Ferrari, Lala and Stojmenova2015). Finally, for scholars advocating the pragmatic answer, in these perspectival usages, there may be a shift of meaning of tenses under the pressure of linguistic and non-linguistic, contextual, information (as in Saussure, Reference de Saussure, Jaszczolt and de Saussure2013), or the meaning of verbal tenses is given by a core of semantic information (a basic configuration of the Reichenbachian coordinates moment of speech S, reference point R and event moment E; cf. Reichenbach, Reference Reichenbach1947) and a number of pragmatic features (as in Moeschler et al., Reference Moeschler, Grisot and Cartoni2012; Moeschler, Reference Moeschler2019). According to Saussure (Reference de Saussure, Jaszczolt and de Saussure2013), which builds on Sthioul (Reference Sthioul, Moeschler, Jayez, Luscher, de Saussure and Sthioul1998) and Saussure and Sthioul (Reference de Saussure and Sthioul1999; Reference de Saussure, Sthioul, Larivée and Labeau2005), the comprehender (referring to the reader or to the hearer) of the narrative IMP is prompted to change the reference point R (which is not perspectival in the regular usages of the IMP) for a perspectival point (Saussure, Reference de Saussure, Jaszczolt and de Saussure2013: 52).Footnote ⁷ With this perspectival reference point R, the comprehender interprets the eventuality expressed by the narrative IMP as ongoing from the point of view of a third party who perceives that action. For Saussure, this perspectival analysis is also supported by the fact that the narrative IMP cannot be replaced by a simple past when a marker of subjectivity (such as the adverbial déjà expressing the speaker’s surprise) accompanies the verb, as in (10) (as suggested by Sthioul, Reference Sthioul, Moeschler, Jayez, Luscher, de Saussure and Sthioul1998; cf. Saussure, Reference de Saussure, Jaszczolt and de Saussure2013: 52).

1. (9) Le lendemain, Paul partait déjà.

   ‘The next day, Paul has already left.’

2. (10) *Le lendemain, Paul partit déjà.

   ‘The next day, Paul already left.’

Another example of perspectival usage of French verbal tense is the futurate PC, which Saussure’s (Reference de Saussure, Jaszczolt and de Saussure2013) model explains in terms of the instantiation of a perspectival future moment of speech S’ (p. 56). S’ is a future projection of S, from which the eventuality E is represented as past, as illustrated in example (5). So, it cannot be established that a deterministic pattern of the type PS or PC uniquely triggers objective interpretations, nor that IMP uniquely triggers subjective interpretations.

Notwithstanding the technical details which each of these approaches proposes, there is agreement that these non-prototypal usages are perspectival, in the sense that they express the speaker’s perspective (and knowledge) of the eventuality expressed, and that this is key to resolving the tension between the temporal adverbial and verbal tense. In other words, the comprehender is invited to take the speaker’s perspective when reading this type of perspectival usage of verbal tenses in order to interpret them.

3. STUDY 1: ANNOTATION TASK WITH CORPUS DATA

3.3. Results

For each version of the data, we calculated the mean inter-annotator agreement rates by calculating the agreement rates for each pair of annotators (A1-A2, A2-A3 and A1-A3). The mean inter-annotator agreement rate was 68% for the version of data where verbs were inflected and 64% for the version where verbal tenses were concealed. Since these inter-agreement rates are not much higher than chance (0.5 in a binary categorization), we carried out further analyses on the corpus excerpts which had full agreement among the three annotators: 112 for the first version of the data, and 97 for the second.

Figure 1 shows how annotators annotated the corpus excerpts containing an IMP, a PC and a PS as subjective or objective. The left-hand panel provides results for the first version of the data, where annotators had access to the verbal tense, while the right-hand panel shows how annotators annotated the corpus excerpts from the second version of the data, where initial occurrences of PC, PS and IMP were removed.

Figure 1. Results of the annotation task for the first vs. second version of the data.

The first series of results, from the first version of the data, is as follows: corpus excerpts where the IMP occurs were significantly more often annotated as subjective rather than objective; by contrast, those where the PC occurs were significantly more often annotated as objective rather than subjective. Finally, those where the PS occurs were annotated as subjective and objective with equal frequency (X² (8.42), 2, p=0.0148).

The second series of results, from the second version of the data, is as follows: corpus excerpts with a concealed PC were significantly more often annotated as objective rather than subjective; those with a concealed IMP tended more often to be annotated as subjective than objective; finally, those with a concealed PS were annotated as subjective and objective with equal frequency (X² (5.77), 2, p=0.0559).

The third result regards the comparison between the distribution observed in the first and second versions of the data: a chi-square test of independence showed that there were no statistically significant differences between the two groups of data points (X² (0.5), 1, p=0.4795). In other words, corpus excerpts were annotated in the same way, whether or not annotators had access to the verbal tense.

4. STUDY 2: SELF-PACED READING EXPERIMENT WITH THE PS

4.1. Hypotheses and predictions

To investigate whether perspectival usages of verbal tenses lead to subjectivity, we carried out a self-paced reading experiment with a 2x3 within-group design, with two variables: the type of context, with two levels (subjective and objective); and the compatibility with the adverbial status, with three levels (semantically compatible with hier ‘yesterday’, semantically incompatible but pragmatically compatible with aujourd’hui ‘today’, and semantically and pragmatically incompatible with en ce moment ‘in this moment’), as in Table 1.Footnote ¹⁰ We explain the two variables and the predictions we make below.

Table 1. The 2x3 experimental design

The hypothesis is that if perspective usages are subjective, we would expect a facilitation effect regarding the processing of sentences like Aujourd’hui, personne ne lui adressa la parole when they are preceded by a subjective — rather than objective — context. We expect that perspective-taking, as one of the components of subjectivity, will be facilitated in a subjective context. Therefore, we predict longer reading times when the target sentences are read in an objective context than when they are read in a subjective context.

Furthermore, the perspectival usage of the PS is characterised by a semantic incompatibility between a present time adverbial (aujourd’hui ‘today’, which situates an eventuality E as simultaneous to the moment of speech S) and a past time verbal tense (which situates E as preceding and completely Footnote dissociated from SFootnote ¹³ ). According to the pragmatic studies discussed in section 2, this semantic incompatibility is resolved pragmatically by perspective-taking: the comprehender is invited to take the speaker’s perspective on the eventuality expressed by the verb. At the same time, there is no semantic incompatibility between a past time adverbial, such as hier ‘yesterday’, and the PS, where the PS has a regular, referential, usage. We expect an interaction effect between the type of context (subjective vs. objective) and the compatibility status with the adverbial variable, according to which — in objective contexts — sentences with semantic incompatibility between verbal tense and adverbial are harder to process than sentences with no such incompatibility (such as hier, personne ne lui adressa la parole), due to the lack of the facilitating effect of the subjective context. Finally, we established a control condition, in which the PS co-occurs with a different type of present time adverbial (en ce moment ‘in this moment’, which situates E as coinciding with S). This type of semantic clash is not resolvable either semantically nor pragmatically, as is the case with the adverbial aujourd’hui (which is the perspectival usage of the PS). We expect to find significantly longer processing times for en ce moment than for hier and aujourd’hui.

4.4. Analysis

All answers to the comprehension questions were checked. Two participants had a success rate of less than 70% and were removed from the data set. For the remaining 39 participants, the general level of success in the comprehension questions was 86%. Before conducting the analysis, the data were cleaned in two phases. In the first phase, extreme values were removed by deleting observations under 50ms. and over 4000 ms. measured for all the segments starting with the biasing sentence, as these reading times are not considered in the literature to represent normal reading times for short segments like those used in our experiment (cf. Zufferey, Reference Zufferey2014). Then, outliers were removed from the data by removing all observations above and below three standard deviations from the subject’s mean values for each of the three main target regions (adverbial, subject-verb, and complement of the target sentence). This corresponded to a total of 86 observations (8% of the data). The mean reading times (standard errors SE between parentheses) for the two target regions (as illustrated above in examples (17) and (18) in each of the six experimental conditions) are reported in Table 2.

Table 2. Mean reading times and standard errors (ms.) for the target regions

Data were analysed by fitting linear mixed-effects models using the R software (R Development Core Team, Dalgaard, Reference Dalgaard2010, version 3.1.2). This lets both participants and items be included as random factors in all analyses, therefore avoiding the “language-as-fixed-effect-fallacy” by separating F1 and F2 analyses (Brysbaert, Reference Brysbaert2007; Clark, Reference Clark1973). Models were tested using the lmer() function of the lmer4 package of R, and model comparisons were assessed using the anova() function, which calculates the chi-square value of the log-likelihood to evaluate the difference between models, following Baayen’s (Reference Baayen2008) procedure.

Following Field et al. (Reference Field, Miles and Field2014), we built the models by going from the simplest model to the one of interest. In the first analysis, for each of the two target regions, we first tested a model that only encompassed items and participants as random factors (i.e. random intercepts). We then compared this model to one including the compatibility with the adverbial status as a fixed factor, and finally one that incorporated both compatibility status and type of context (and their interaction) as fixed factors.

4.5. Results

In the adverbial region of the target sentence (region 5), adding the compatibility with the adverbial variable to the simplest model (which only included items and participants as random effects) significantly improved the model (Δχ² = 23.41, Δdf = 2, p < .001). As expected, reading times in the control condition with en ce moment (the semantically and pragmatically incompatible condition) were significantly longer (M = 884.58, SE = 19.07) than those recorded with hier (the semantically compatible condition; M = 786.30, SE = 18.11), as well as those recorded in the target condition with aujourd’hui (the semantically incompatible but pragmatically compatible condition; M = 794, SE = 17.95). This result, shown in Figure 2, confirms that our experiment functioned correctly. Furthermore, the processing times in the hier and aujourd’hui conditions were similar (M = 786.30, SE = 18.11 and M = 794, SE = 17.95 respectively), indicating that the semantic incompatibility from the target condition (aujourd’hui) was pragmatically resolved.

Figure 2. Mean reading times for S5 (the adverbial region).

However, adding the type of context did not further improve the model’s fit (Δχ² = 1.58, Δdf = 1, p > .05, ns.), as the processing time of the adverbial region did not vary with respect to the type of context. In section 4.1, we formulated the prediction that a subjective context will facilitate access to the perspective-taking component. As such, we expected longer reading times when the target sentence was read in an objective context than in a subjective context. This prediction is not confirmed by our results, as the reading times measured for the adverbial aujourd’hui in the objective context (M = 804.98, SE = 27.31) were similar to those measured in the subjective context (M = 785.48, SE = 25.40). In addition, we predicted an interaction effect according to which — in objective contexts — sentences presenting the perspectival usage of the PS with aujourd’hui would be harder to process than sentences presenting a regular usage of the PS with hier, because the facilitating effect of the subjective context is lacking. Again, our results did not confirm this prediction, as we found similar reading times for the adverbial region in the objective context when the adverbial was aujourd’hui (M = 804.98, SE = 27.31) as we did when it was hier (M = 787.14, SE = 26.00).

In the subject-verb region of the target sentence (region 6), we found similar results to those found in the previous region. Adding the compatibility with the adverbial variable to the simplest model (which only included items and participants as random effects) significantly improved the model (Δχ² = 20.94, Δdf = 2, p < .001). As expected, reading times in the control condition with en ce moment (the semantically and pragmatically incompatible condition) were significantly longer (M = 1118.68, SE = 30.40) than those recorded with hier (the semantically compatible condition; M = 1014.05, SE = 30.54), as well as those recorded in the target condition with aujourd’hui (the semantically incompatible but pragmatically compatible condition; M = 995.61, SE = 25.08). This result is shown in Figure 3. As with the previous region, the processing times in the hier (M = 1014.05, SE = 30.54) and aujourd’hui conditions are similar (M = 995.61, SE = 25.08), suggesting that the semantic incompatibility from the target condition (aujourd’hui) was pragmatically resolved.

Figure 3. Mean reading times for S6 (the subject-verb-complement region).

However, adding the type of context did not further improve the model’s fit (Δχ² = 0.01, Δdf = 1, p > .05, ns.), as the processing time of the subject-verb-complement region did not vary with respect to the type of context. As with the previous region, our results show that the reading times measured on the subject-verb-complement, when it was preceded by the aujourd’hui adverbial, were similar in the objective context (M = 997.37, SE = 36.09) and the subjective context (M = 993.69, SE = 34.75). As for the interaction effect that we predicted, the results show that — in the objective context — reading times when the subject-verb-complement region was preceded by adverbial aujourd’hui (M = 997.37, SE = 36.09) were similar to those when it was preceded by the adverbial hier (M = 1033.72, SE = 43.13).

5. GENERAL DISCUSSION AND CONCLUSION

The research question we explored in this article is whether French verbal tenses give access to speaker’s subjectivity via perspective-taking. To answer this question, we carried out two studies: an annotation study, and a self-paced reading experiment.

In the annotation study, our aim was to examine predictions made by two groups of studies (the first in linguistic literary criticism and the second from a pragmatic approach) regarding the possibility of a deterministic relation between perfective vs. imperfective verbal tenses and objective vs. subjective interpretations of utterances. For the linguistic literary criticism approach, this relation is deterministic, in the sense that perfective verbal tenses give rise to objective interpretations where imperfective verbal tenses give rise to subjective interpretations. For the pragmatic approach, this relation is not deterministic, since verbal tenses have both prototypical and non-prototypical usages, the latter being the subjective ones. The results of the annotation task do not favour this kind of deterministic relation, as we found that the three verbal tenses tested (PC, PS and IMP) have both subjective and objective usages. At the same time, we did find that sentences in which the IMP occurs are significantly more frequently understood as subjective, whereas sentences in which the PC occurs are significantly more frequently understood as objective. The same study also showed that annotators interpreted the corpus excerpts when they had access to verbal tense (the first group of annotators) in the same way as they did when verbal tenses were concealed (the second group of annotators). This finding is important, as it indicates that the subjective interpretations of the corpus excerpts in which the PC, PS or the IMP occur are not triggered by the verbal tense itself, or at least not only by the verbal tense, but principally by other cues. Building on the results from Grisot (Reference Grisot2019), we suppose that the subjective interpretations most probably drew from linguistic cues such as the affective and evaluative lexicon, and non-canonical syntactic structures (like cleft sentences and other types of fronted phrases) which were found to be the most frequent categories of linguistic cue of the speaker’s subjectivity. For example, the literary corpus excerpt from (20) (from Grisot, Reference Grisot2019: 31) was annotated by three independent annotators as expressing the speaker’s subjectivity. According to these annotators, this interpretation draws on the following linguistic cues: the affective-evaluative lexicon profiteur ‘profiteers’, spéculateurs ‘speculators’, nouveaux riches ‘the nouveau riche’ and des douzaines ‘dozens’, as well as the fronted syntactic structures comme tous les profiteurs [.] ‘like all profiteers’, on en attribuait […] ‘people thought’.

1. (20) Comme tous les profiteurs, les spéculateurs, les nouveaux riches de France achetaient des châteaux, on en attribuait des douzaines à Galmot. (B. Cendrars, Rhum) [BC025]

   ‘Like all profiteers and speculators, the nouveau riche of France bought castles, and people thought Galmot had dozens.’ (our translation)

Another example is the journalistic corpus excerpt from (21) (from Grisot, Reference Grisot2019: 33), also annotated by three annotators as subjective, an interpretation which they attribute to the following linguistic cues: the cleft sentence c’est […] que […] ‘it is […] that […]’, the pronouns leurs ‘their’ and nos ‘our’, the lexicon journal de reference ‘benchmark newspaper’, and the future time verbal tense continuera ‘will continue’.

1. (21) C’est grâce à leurs talents, leur expertise et l’acuité de leur regard que Le Monde continuera à répondre à votre légitime exigence et restera le journal de référence que vous lisez sur tous nos supports, papier et digitaux. (Le Monde, year 2012) [LM031]

   ‘It is thanks to their talents, their expertise and their keen eye that Le Monde will continue to meet your legitimate requirements and will remain the benchmark newspaper that you read on all our media, paper and digital.’ (our translation)

In the self-paced reading experiment, we tested a very specific type of perspectival usage of verbal tenses: the usage where the PS co-occurs with the present time adverbial aujourd’hui, as in aujourd’hui, personne ne lui adressa la parole. We manipulated the context (as objective or as subjective) and the adverbial (hier, aujourd’hui and en ce moment) to test whether the processing of this perspectival usage of the PS is facilitated by subjective contexts built with affective linguistic cues. Our results did not support the thesis that this perspectival usage of the PS is subjective, in the sense of giving access to the speaker’s subjectivity.Footnote ¹⁶ We identify two possible explanations of this result.

The first possible explanation which would allow us still to consider perspective-taking as a component of speaker’s subjectivity is the following. Perspective-taking involves two phases: a first phase, in which the comprehender is invited to take the temporal perspective of the speaker (in example (1), it is Julien’s aujourd’hui); and a second phase, in which the comprehender is able further to infer information about the speaker’s thoughts, emotions and psychological states once he has taken in the speaker’s temporal perspective. This explanation draws on the incremental and expectation-based models of language processing (Altmann, Reference Altmann1988; Crain and Steedman, Reference Crain, Steedman, Dowty, Karttunen and Zwicky1985; Kehler and Rohde, Reference Kehler and Rohde2017), according to which, at each step of processing a discourse, comprehenders anticipate the next part by leaning on linguistic cues. In the case of perspective-taking, when coming across the adverbial aujourd’hui, the comprehender anticipates the possibility of taking the speaker’s perspective: he can infer information about the speaker’s thoughts, emotions and psychological states. This account is compatible with the theories of perspective shift suggested in the case of FID (Banfield, Reference Banfield1982; Schlenker, Reference Schlenker2004; Reboul et al., Reference Reboul, Delfitto and Fiorin2016). One of the most striking particularities of FID is that indexicals do not belong to the same context: tenses and pronouns depend on the context of speech (or the context of utterance, in Schlenker’s Reference Schlenker2004 terms), where other deictics (temporal adverbials and demonstratives) belong to the context of thought, in which the speaker’s or narrator’s thoughts are expressed (Schlenker, Reference Schlenker2004). In FID, there is a frequent shift between the context of utterance, which is the actual context of the story, and the context of thought. In cases like Aujourd’hui, personne ne lui adressa la parole, the adverbial aujourd’hui belongs to the context of thought, whereas the PS belongs to the context of utterance. So, this type of utterance is not contradictory, as the comprehender shifts between the two types of context. It is possible that participants in our experiment performed the first phase of perspective-taking but did not go on to the second phase, where they would have been able to infer the character’s thoughts, emotions and psychological state. In other words, they did not shift between the two contexts. It might be the case that the second phase (i.e. the shift from the context of utterance to the context of thought) is easier and more naturally accepted in the case of longer narrative stories or novels. Nevertheless, our current design does not allow us to distinguish between the two phases, so a further experiment is necessary to determine whether this is the case.

The second possible explanation is that perspective-taking is not a component of subjectivity. Recall how our results indicate that the PS might not be semantically incompatible with the adverbial if it is understood to refer to a generic or atemporal period of time (which encompasses sub-periods of time previous, simultaneous and posterior to S) rather than a restricted period of time which is uniquely simultaneous to S. Another point is linked to the presumed semantic incompatibility between the PS and aujourd’hui, predicted to be resolved pragmatically by perspective-taking. Our results seem to indicate that perspective-taking might not be a component of speaker’s subjectivity, but a different type of phenomenon. This proposal is supported by the previous theoretical distinction made by Kerbrat-Orecchioni (Reference Kerbrat-Orecchioni2004) between deictic subjectivity (expressed through a broad class of deictic elements, such as personal pronouns, demonstratives, verbal tenses, grammatical aspect and temporal adverbs, among others) and affective-evaluative subjectivity (referring to “the individual usages of the common [language] code”, and mainly expressed through the evaluative and affective lexicon) (p. 80). For Kerbrat-Orecchioni (Reference Kerbrat-Orecchioni2004: 165), deictic elements are not subjective, as they may not receive different assessments from one person to the next. By contrast, the use of affective and evaluative words and expressions may be questioned because they depend on the individual judgment made by each person. According to this proposal, perspective-taking (deployed for deictic subjectivity) would therefore not belong to speaker’s subjectivity, which might be restricted to its affective and evaluative components (cf. proposal made in Grisot, Reference Grisot2019, as well as in Blochowiak et al. Reference Blochowiak, Grisot and Degand2020, with respect to French causal connectives). As such, the perspectival usage of verbal tenses — including the PS co-occurring with aujourd’hui investigated here — should not be considered as authentic subjective usages (i.e. giving access to speaker’s subjectivity). In order to test this novel idea, that speaker subjectivity consists of two components (speaker emotions and speaker epistemic evaluation of propositions), we need to conduct further experimental research.