3.1The semantic pattern of ‘sollte’ + inf.

As already indicated above, the FoF-reading of sollte + inf. is a ‘special case’ in terms of semantics (Glas 1984Glas, Reinhold 1984 ‘sollen’ im heutigen Deutsch. Bedeutung und Gebrauch in der Schriftsprache. Tübingen: Narr., 99f.; Öhlschläger 1989Öhlschläger, Günther 1989 Zur Syntax und Semantik der Modalverben des Deutschen. Tübingen: Niemeyer., 176) since it denotes a high degree of certainty with respect to a future event realization, which is uncommon both for the deontic and epistemic reading of modal verbs (cf. Example (2) above). In order to describe the grammatical semantics of sollte + inf., it is necessary to look at the semantic components of the form, i.e. (i) the modal verb soll- (‘shall’) and (ii) the suffix -te. Its morphology is not completely transparent since the uses of the form sollte for both the indicative preterite and the subjunctive mood (‘Konjunktiv II’) are homophones since the Middle High German period. In the literature, it is sometimes assumed that the subjunctive form sollte has gained a completely new meaning and split into a lexeme different from the modal verb sollen (see Diewald 1999Diewald, Gabriele 1999 Die Modalverben im Deutschen. Grammatikalisierung und Polyfunktionalität [Reihe Germanistische Linguistik 208]. Tübingen: Niemeyer. , 198 for critical discussion and an argumentation in favor of a compositional account). According to Fritz (1997)Fritz, Gerd 1997 “Historische Semantik der Modalverben. Problemskizze – Exemplarische Analysen – Forschungsübersicht.” In Untersuchungen zur semantischen Entwicklungsgeschichte der Modalverben im Deutschen, ed. by Gerd Fritz, and Thomas Gloning, 1–157. Tübingen: Niemeyer., this split however did not occur before 1600, which means later than the first occurrences of FoF readings. The early examples in Old High German discussed in Section 4 – i.e. examples from a period where indicative (scolta) and subjunctive (scolti) in third person are still morphologically distinguished –, suggest that the form carrying a FoF reading can be compositionally analyzed as an indicative preterite form of sollen, which would also be in line with the cross-linguistic comparison (see Section 3.2).

With respect to the semantic contribution of the modal verb soll-, traditional accounts describe the FoF reading as an isolated relic which goes back to an older periphrastic use to denote future tense reference (see Gloning 2001Gloning, Thomas 2001 “Gebrauchsweisen von Modalverben und Texttraditionen.” In Modalität und Modalverben im Deutschen [Linguistische Berichte: Sonderheft 9], ed. by Reimar Müller, and Marga Reis, 177–200. Hamburg: Buske.; Fritz 1997Fritz, Gerd 1997 “Historische Semantik der Modalverben. Problemskizze – Exemplarische Analysen – Forschungsübersicht.” In Untersuchungen zur semantischen Entwicklungsgeschichte der Modalverben im Deutschen, ed. by Gerd Fritz, and Thomas Gloning, 1–157. Tübingen: Niemeyer., 46). This hypothesis remains, however, insufficient for two reasons. First, Old High German skulan and Middle High German suln do not grammaticalize into a future tense marker but generally retain their modal meanings in all stages of German (see Diewald and Habermann 2005Diewald, Gabriele, and Mechthild Habermann 2005 “Die Entwicklung von werden + Infinitiv als Futurgrammem: ein Beispiel für das Zusammenwirken von Grammatikalisierung, Sprachkontakt und soziokulturellen Faktoren.” In Grammatikalisierung im Deutschen [Linguistik – Impulse & Tendenzen 9], ed. by Torsten Leuschner, Tanja Mortelmans, and Sarah De Groodt, 230–250. Berlin/Boston: de Gruyter. ; Zeman 2013Zeman, Sonja 2013 “Zur Diachronie der Modalverben: sollen zwischen Temporalität, Modalität und Evidentialität.” In Funktionen von Modalität [Linguistik – Impulse & Tendenzen 55], ed. by Werner Abraham, and Elisabeth Leiss, 335–336. Berlin/Boston: de Gruyter. ; Abraham and Nishiwaki 2016Abraham, Werner, and Maiko Nisihiwaki 2016 “Modal Verbs in German and Definiteness Effects in Verbal Complements – Focusing on Modern Standard German sollen and Middle High German suln ‘shall’.” In Definiteness Effects: Bilingual, Typological, and Diachronic Variation, ed. by Susan Fischer, Tanja Kupisch, and Esther Rinke, 244–276. Newcastle Upon Tyne: Cambridge Scholars Publishing.). Second, the future relict hypothesis offers no explanation for the high degree of certainty as the major component of its semantics: Since future events are inherently unknown, reference to the future commonly implies uncertainty on behalf of the speaker (see e.g. Dahl 2006Dahl, Östen 2006 “Future Tense and Future Time Reference.” In Encyclopedia of Language and Linguistics, ed. by Keith Brown, 704–706. Amsterdam: Elsevier. for an overview on the cross-linguistic link between future temporality and epistemic uncertainty). Furthermore, the traditional account only focuses on the future realization of the event while neglecting the complex viewpoint constellation, see (3).

With respect to its temporal value, (3) displays a ‘future-in-the-past’ reading that refers to a future event seen from the perspective of the ‘story-now’, while the verbal event is, at the same time, past as seen from the perspective of the point of speech. This double temporal constellation can be derived from the deictic structure of the modal verb construction and the past marker. First, the projective reading is an inherent semantic feature of the modal verb. As is well known, modal verbs like suln + inf. are biphasic constructions since they integrate two different time intervals: The time interval of the deontic modal (I_MODAL) (i.e. the time for which the obligation is valid), and the time interval covered by the event that should be realized, denoted by the infinitive (I_EVENT) (see Abraham 2008Abraham, Werner 2008 “On the Logic of Generalizations about Cross-Linguistic Aspect-Modality Links.” In Modality-Aspect Interfaces. Implications and Typological Solutions [Typological Studies in Language 79], ed. by Werner Abraham, and Elisabeth Leiss, 3–13. Amsterdam: John Benjamins. ; Maché 2008Maché, Jakob 2008 “The Autopsy of a Modal – Insights from the Historical Development of German.” In Modality-Aspect Interfaces. Implications and Typological Solutions [Typological Studies in Language 79], ed. by Werner Abraham, and Elisabeth Leiss, 385–415. Amsterdam: John Benjamins. , 403f.; Depraetere 2012Depraetere, Ilse 2012 “Time in Sentences with Modal Verbs.” In The Oxford Handbook of Tense and Aspect, ed. by Robert I. Binnick, 989–1019. Oxford: Oxford University Press., 990f.; Zeman 2014Zeman, Sonja 2014 “(C)Overt Modality and its Perspectival Effects on the Textual Surface.” In Modes of Modality. Modality, Typology, and Universal Grammar [Studies in Language Complementary Series 149], ed. by Werner Abraham, and Elisabeth Leiss (eds.), 457–484. Amsterdam: John Benjamins. ). In combination with telic infinitive complements, the deictic structure thus offers the potential for two different temporal reference points. In addition, the past marker denotes a distance between the time interval of the modal and the time of speech, so that in consequence three temporal reference points arise, allowing for three different viewpoints (i.e. the narrator’s, the protagonist at the ‘story-now’, and the protagonist in the future).

These viewpoints are however not equivalent alternatives, but are situated on different levels of the hierarchically ordered discourse structure, as seen in the fact that the narrator’s knowledge includes both the knowledge state on the level of the characters and the knowledge about the further line of discourse. This hierarchy allows for the unexpected combination of a future meaning and a high degree of certainty with respect to the actualization of the future event. Uncertainty is ruled out with the emergence of the narrator’s viewpoint on a higher level.22.A reviewer raised the intriguing question whether FoF readings present an exception for Verstraete’s (2005) cross-linguistic rule of composite mood marking, derived from his investigation on mood-marking in non-Pama-Nyungan languages of northern Australia. Verstraete argues that modal meanings of the past are inherently weaker on the scale of information than their non-modal equivalents and lead to the pragmatic implicature of non-actualization and, based on that, counterfactuality: “Using the epistemically weaker expression of potentiality will trigger the implicature that there was a reason not to use the stronger non-modal expression, i.e., that the event did not take place.” (Verstraete 2005, 236) Taking into account the temporal structure and viewpoint constellation, the FoF reading of sollte is however not an exception to Verstraete’s rule. A past modal per se seems not enough to trigger the implicature, as the examples in Verstraete (2005) suggest, that it is necessary that the time interval of the event is before or simultaneous to speech time (compare ‘You should have shot it (→ but he didn’t)’ and ‘You should be shooting it (→ but you aren’t’)’ vs. ‘You should shoot it’, where no implicature is triggered regarding the actualization of the event). This is also in line with the semantic map for the modal uses of Polish mieć in Holvoet (2012Holvoet, Axel 2012 “Polish mieć and the Semantic Map of Interpretive Deontics.” Zeitschrift für Slawistik 57/2: 129–146. , 144), where ‘unrealised intention’ is supposed to lead to counterfactual uses, but not the ‘destiny type’. Seen within a bigger picture, this calls for further cross-linguistic work in perspectival terms in order to specify in greater detail the grammaticalization pathways.

All in all, the FoF reading of the past form can be analyzed by decomposing it into its core semantic components, which show the result of two different viewpoint splits. First, the biphasicness of the modal verb allows for a horizontal shift on the propositional level with the ‘story-now’ as reference point, leading to a potential difference between two different temporal instances of the protagonist, i.e. the protagonist carrying an obligation at the story-now (expressed in the modal verb), and the protagonist at the time of the future realization of the event to be carried out (as expressed in the infinitive). Furthermore, there is a vertical shift leading to the emergence of a narrator’s level above the proposition. As will be seen in the next section, this grammatical pattern can also be observed for the source constructions of other FoF constructions in different languages.

3.2The semantic pattern in crosslinguistic comparison

Seen from a cross-linguistic point of view, FoF readings are fairly unremarkable insofar as they are also documented in other languages. Next to modal verbs that etymologically trace back to the semantics of obligation like English should, Danish skulle and Swedish skulle (see Zeman 2018bZeman, Sonja 2018b “Zukunft in der Grammatik: Neues vom Schicksalsfutur.” In Die Zukunft der Grammatik – die Grammatik der Zukunft [Studien zur deutschen Grammatik 92], ed. by Elisabeth Leiss, and Sonja Zeman, 449–481. Tübingen: Stauffenburg.), a similar pattern can also be observed for the méllō (‘intend’) + inf. construction in Ancient Greek. Méllō + infinitive is commonly classified as a (semi-)auxiliary that denotes a “present intention of the subject of the realization of a state of affairs in the (near) future” (Wakker 2006Wakker, Gerry C. 2006 “Future Auxiliaries or Not.” In Word Classes and Related Topics in Ancient Greek. Proceedings of the Conference on ‘Greek Syntax and Word Classes’ Held in Madrid on 18–21 June 2003, ed. by Emilio Crespo, Jesús de La Villa, and A. R. Revuelta (eds), 237–255. Louvain-la-Neuve: Peeters.), see (4).

In (4), émelle refers to Menelaus’ intention not to kill Adrestes. Méllein denotes the subject’s mental state of thinking about doing something whereas it is not determined whether the intended action will indeed be realized in the course of the story or not (in fact some lines later, Agamemnon will have changed Menelaus’ mind). (5), in contrast, evokes a FoF reading. The realization of the event in the future is presented as affirmed as certain whereas the intention of the subject is not at issue or even contradicted:

In (5), the construction indicates what will happen in the future course of events against the will of the subject; it is clear from the context (and the common knowledge of the audience) that it is not Dolon’s intention not to return. Méllein + inf. thus invokes the same complex perspectival constellation as sollte + inf. and refers to a time interval that is posterior to a reference point that is already past as seen from the perspective of the narrator. By speaking about the events to come, the construction does not only refer to the propositional but also to the non-propositional level,33.There are different terms used in the literature denoting the difference between the propositional vs. the non-propositional level, such as e.g. the distinction between descriptive and interpretive use in Relevance Theory (Sperber and Wilson 1986Sperber, Dan, and Deirdre Wilson 1986 Relevance: Communication and Cognition. 2nd edition. Oxford: Blackwell., 228), presentational and representational (Kroon 1998Kroon, Caroline 1998 “A Framework for the Description of Latin Discourse Markers.” Journal of Pragmatics 30: 205–223. ) and ideational and interpersonal level (Halliday and Hasan 1976Halliday, Michael A. K., and Ruqaiya Hasan 1976 Cohesion in English. London: Longman.). It will be a task for future research to specify these notions more precisely in relation to the different layers of illocution in the sense of Bergqvist (2017)Bergqvist, Henrik 2017 “The Role of ‘Perspective’ in Epistemic Marking.” Lingua 186/187: 5–20. . i.e. it does not only describe the relations between the states of affairs in the represented world, but also their relation to the representational discourse level.

Sollte + inf. and émellen + inf. thus display a very similar semantic pattern. Both sollen (‘shall’) and méllein (‘intend to’) + inf. are biphasic constructions which involve two different temporal reference points. While the time interval of the modal includes the subject’s present, the realization of the event denoted by the infinitive will take place (or not) after the present reference point. In this respect, méllō just as sollen + inf. “inherently unites, then, two different semantic notions: present intention/arrangement and (relatively) future realization” (Wakker 2006Wakker, Gerry C. 2006 “Future Auxiliaries or Not.” In Word Classes and Related Topics in Ancient Greek. Proceedings of the Conference on ‘Greek Syntax and Word Classes’ Held in Madrid on 18–21 June 2003, ed. by Emilio Crespo, Jesús de La Villa, and A. R. Revuelta (eds), 237–255. Louvain-la-Neuve: Peeters., 245, similarly Wakker 2007Wakker, Gerry C. 2007 “Intentions and Future Realisations in Herodotus.” In The Language of Literature: Linguistic Approaches to Classical Texts, ed. by Rutger Allan, and Michel Buijs, 168–187. Leiden: Brill. , 169). In this respect, also méllein + inf. has been discussed as an auxiliary of the future tense (see Basset 2003 [1966]Bassett, Samuel Eliot 2003 [1966] The Poetry of Homer. Lanham, MD: Lexington Books.; Wakker 2006Wakker, Gerry C. 2006 “Future Auxiliaries or Not.” In Word Classes and Related Topics in Ancient Greek. Proceedings of the Conference on ‘Greek Syntax and Word Classes’ Held in Madrid on 18–21 June 2003, ed. by Emilio Crespo, Jesús de La Villa, and A. R. Revuelta (eds), 237–255. Louvain-la-Neuve: Peeters., 2007Wakker, Gerry C. 2007 “Intentions and Future Realisations in Herodotus.” In The Language of Literature: Linguistic Approaches to Classical Texts, ed. by Rutger Allan, and Michel Buijs, 168–187. Leiden: Brill. ; Allan 2013Allan, Rutger J. 2013 “Exploring Modality’s Semantic Space. Grammaticalization, Subjectification and the Case of ὀφείλω.” Glotta 89: 1–46. , 34 Fn. 44). However, similar to the German form sollen + inf., méllō + inf. seems to retain its modal semantics, see Wakker (2006Wakker, Gerry C. 2006 “Future Auxiliaries or Not.” In Word Classes and Related Topics in Ancient Greek. Proceedings of the Conference on ‘Greek Syntax and Word Classes’ Held in Madrid on 18–21 June 2003, ed. by Emilio Crespo, Jesús de La Villa, and A. R. Revuelta (eds), 237–255. Louvain-la-Neuve: Peeters., 2007Wakker, Gerry C. 2007 “Intentions and Future Realisations in Herodotus.” In The Language of Literature: Linguistic Approaches to Classical Texts, ed. by Rutger Allan, and Michel Buijs, 168–187. Leiden: Brill. ) and Markopoulos (2008)Markopoulos, Theodore 2008 The Future in Greek. From Ancient to Medieval. Oxford: Oxford University Press. who shows that méllō was “not exactly equivalent to the Future Tense, as it implied less commitment on the part of the speaker concerning the realization of the action described, in other words it was more modal than temporal” (Markopoulos 2008Markopoulos, Theodore 2008 The Future in Greek. From Ancient to Medieval. Oxford: Oxford University Press. , 21).

Furthermore, both source constructions display a semantic core of necessity. The semantic source concept of sollen relies in the meaning ‘to owe’ and indicates that the subject is obliged to do something by an external source of obligation. Méllō is oscillating between different modal shades and can correspond to a whole range of modal verbs: ‘a) I think, think about doing, like cogito facere, b) I shall, must do, frequently in Homer: it is destined to do something, c) I can, am able to do, d) I am allowed to do, e) I may’ (Kühner et al. 1870Kühner, Raphael, Friedrich Blass, and Bernhard Gerth 1870 Ausführliche Grammatik der griechischen Sprache. Volume 2/1. Hannover: Hahn., § 387 note 2, 150).44.Original wording: “(a) ich denke, gedenke zu thun, wie cogito facere, (b) ich soll, muss thun, b. Hom. häufig: es ist mir vom Schicksale beschieden Etwas zu tun, (c) ich kann, bin im Stande zu thun, (d) ich darf thun, (e) ich mag wol.” According to Kühner et al. (1870)Kühner, Raphael, Friedrich Blass, and Bernhard Gerth 1870 Ausführliche Grammatik der griechischen Sprache. Volume 2/1. Hannover: Hahn., the original meaning is ‘think about doing something’ whereby the modal force can be the subject’s own will (‘intend to do’) or the will of a third external source (similarly also Markopoulos 2008Markopoulos, Theodore 2008 The Future in Greek. From Ancient to Medieval. Oxford: Oxford University Press. , 21). The latter seems to be the relevant semantic component, as seen in comparison with constructions in French (Schrott 1997Schrott, Angela 1997 “Futur du passé im Französischen der Gegenwart.” Romanistisches Jahrbuch 48/1: 41–68.), Polish (Hansen 2009Hansen, Björn 2009 “The Hypothetical Use of Polish ‘mieć plus Infinitive’ Revisited.” In Von grammatischen Kategorien und sprachlichen Weltbildern – Die Slavia von der Sprachgeschichte bis zur Politsprache (Festschrift für Daniel Weiss zum 60. Geburtstag) [Wiener Slawistischer Almanach, Sonderband 73], ed. by Tilman Berger, Markus Giger, Sibylle Kurt, and Imke Mendoza (eds.), 177–185. München: Kubon & Sagner.; Holvoet 2012Holvoet, Axel 2012 “Polish mieć and the Semantic Map of Interpretive Deontics.” Zeitschrift für Slawistik 57/2: 129–146. ),55.I am grateful to Björn Wiemer for pointing out this ‘destiny type’ and the relevant references to me. English (Sugayama 2006) and Spanish (Butt et al. 2013Butt, John, and Carmen Benjamin 2013 New Reference Grammar of Modern Spanish. London: Routledge. , § 21.4.1) which display FoF readings, see Table 1.

Table 1.Comparison of source constructions for FoF readings (Zeman 2018bZeman, Sonja 2018b “Zukunft in der Grammatik: Neues vom Schicksalsfutur.” In Die Zukunft der Grammatik – die Grammatik der Zukunft [Studien zur deutschen Grammatik 92], ed. by Elisabeth Leiss, and Sonja Zeman, 449–481. Tübingen: Stauffenburg.)

	Ger. sollte	Agr. emellen	Fr. devait	Pol. mieć	Engl. was to	Span. había de
past marker	✓	✓	✓	✓	✓	✓
lexical source	‘owe’ > ‘shall’	‘owe’ > ‘must’	‘intend’/‘is intended’	‘have’	‘be’	‘have’
deontic reading	obligation	necessity	obligation > necessity	necessity	necessity, arrangement	necessity, arrangement
epistemic reading	✓	–	✓	✓	–	–
modal source	3. ps. evidence	under-specified	1. ps. inference	3. ps. evidence	covert modality	covert modality
future tense	–	–	–	–66.Despite the fact that mieć + inf. can refer to the future, it is commonly not seen as a future tense marker, see Łaziński (2001)Łaziński, Marek 2001 “Was für ein Perfekt gibt es im modernen Polnisch? Bemerkungen zum Artikel ‚Gibt es ein Perfekt im modernen Polnisch?‘ von H. Weydt und A. Kazimierczak.” Linguistik online 8/1. URL https://​bop​.unibe​.ch​/linguistik​-online​/article​/view​/979​/1637 (05.042018). for discussion.	–	–
biphasicness	✓	✓	✓	✓	✓	✓

Table 1 reveals the following similarities between the different source constructions:

1. Despite the different grammatical constructions that constitute the base for FoF readings, all of the analysed source constructions denote a semantic component of ‘necessity’/‘obligation’, whereby the deontic source is an external force that lies outside the communicative situation. As seen for the be and have constructions, the necessity reading does not require that the obligation has been expressed.

2. None of the construction can be classified as a ‘pure’ future tense marker. The projective semantics results from an inherent biphasicness of the (semi-)modals which denote a present time interval of the time of obligation for an event that might be realized afterwards. The projective reading is reinforced by telic infinitives and infinitives in combination with a preposition (see Sugayama 2006 with respect to the was to construction in English; Abraham 2012Abraham, Werner 2012 “(Inter)Subjectification or Foreign Consciousness/Other’s Mind Alignment as Synchronic and Diachronic Concepts of Change? Conceptualizations and Data Fidelity.” In Covert patterns of modality, ed. by Werner Abraham, and Elisabeth Leiss, 24–78. Newcastle upon Tyne: Cambridge Scholars Publishing. for the functional correspondence between prepositional and telic infinitives).

3. Morphologically, all constructions exhibit a past marker which seems to be a prerequisite for the dual temporal structure that integrates a prospective view on the future events of the story as seen from the ‘story-now’ and the narrator’s retrospective view.

The semantic prerequisites for FoF-readings seem thus to be (i) projective biphasicness, (ii) an external deontic source, and (iii) past tense. This pattern does not yet explain how the FoF reading is disambiguated on the level of discourse – a question that will be addressed in the next section.

3.3Disambiguating factors

In order to examine the disambiguating factors of FoF readings, we first of all have to acknowledge that these readings are restricted to narrative discourse, while non-narrative contexts elicit deontic (or epistemic) meanings, cf. (6–7).

Narrative and non-narrative contexts differ with respect to their perspectival structure. Whereas in non-narrative discourse the speaking and observing subject are per default the same, narrative discourse is characterized by a structural difference between the narrator and character level, whereby the distinction between narrator and character level can be seen as a projection of the grammatical differentiation between speaker vs. observer, and hence, as the result of a vertical split. This double-layered structure provides a viewpoint potential that can be activated in the dynamics of discourse (see Zeman 2018cZeman, Sonja 2018c What is a Narration – And Why Does It Matter? In Linguistic Foundations of Narration in Spoken and Sign Language [Linguistics today/Linguistik aktuell 247], ed. by Annika Hübl, and Markus Steinbach, 174–206. Amsterdam: John Benjamins. ). The emerging viewpoints are hierarchically ordered, since the narrator’s level includes the character/event level, see Figure 1.

Figure 1.Viewpoint levels of narrative discourse structure (Zeman 2018cZeman, Sonja 2018c What is a Narration – And Why Does It Matter? In Linguistic Foundations of Narration in Spoken and Sign Language [Linguistics today/Linguistik aktuell 247], ed. by Annika Hübl, and Markus Steinbach, 174–206. Amsterdam: John Benjamins. )

Taking into account the hierarchy of perspectival structure of narrative contexts is relevant for an explanation of FoF readings in two respects. First, the hierarchical discourse structure leads to a qualitative difference between the viewpoints as seen in the fact that the knowledge of the narrator structurally includes the protagonist’s knowledge but not vice versa. Second, it allows for a simultaneous activation of viewpoints on more than one hierarchical level, and hence, for the integration of both the narrator’s and the protagonist’s perspective which results in the knowledge asymmetry that is inherent to the FoF reading. The effects of the hierarchical structure also become obvious within the interaction between FoF reading and the perspectival constellation of the context, as seen in the fact that multiperspectival FoF readings are blocked in Free Indirect Discourse (FID), cf. the minimal pairs in (8–9).

If no context is given, both (8) and (9) could be interpreted as FoF readings. Yet, unlike in (8), the context in (9) prevents such an interpretation. In this context of Free Indirect Discourse, the character’s point of view constitutes the reference point for linguistic elements of perspectivization like deliberative questions (aber wo und wann?), evaluative expressions (Wahrhafftig, vergeblich), modal particles (ja) and pauses of reflection, indicated by the dashes in the text (see for the characteristics of Free Indirect Discourse Banfield 1982Banfield, Anne 1982 Unspeakable Sentences: Narration and Representation in the Language of Fiction. Boston: Routledge & Kegan Paul.; Fludernik 1993Fludernik, Monika 1993 The Fictions of Language and the Languages of Fiction: The Linguistic Representation of Speech and Consciousness. London: Routledge.; Eckardt 2014Eckardt, Regine 2014 The Semantics of Free Indirect Discourse. How Texts Allow Us to Mind-read and Eavesdrop. Leiden: Brill. ). In FID, the viewpoint is thus located at the protagonist’s level, while the narrator’s viewpoint is disabled (cf. also Vandelanotte, this issue on the relation of FID to (irregular) perspective structure). In consequence, the perspectival context of FID is incompatible with a FoF reading since the viewpoint positioning does not allow for any activation of its dual structure. In this respect, FoF and FID display two complementary perspectival structures that are reverse to each other.

The examples also show that the high degree of certainty which is a characteristic semantic component of FoF readings is linked to the nature of the hierarchical viewpoint structure, since it emerges as a result from the narrator’s authority and controlling force on the discourse, which rules out any intervention of the character on the story level. This authority is thus not linked to any mental consciousness but emerges structurally from the ‘external’, i.e. higher discourse level.

On the content level, the disambiguation of viewpoints can be supported by indications that the future realization of the event lies outside the control of the focalized character, i.e. the most prominent mental subject in the story that constitutes the reference point for narrative perspectivization. In prototypical examples, contrasts between the character’s intention and the narrator’s knowledge indicate that the event realization lies outside the control of the protagonist. In (5) above, for instance, the events will happen against the will of the protagonists: Dolon intends to come back to the Trojans, but he cannot change his destiny. On the linguistic level, this correlates with a high number of passive and impersonal constructions that accompany FoF readings (see Zeman forthcomingZeman, Sonja Forthcoming. Wer spricht? Disambiguierungsfaktoren bei der Perspektivensetzung im narrativen Diskurs. Linguistische Berichte [Sonderband „Rede- und Gedankenwiedergabe in narrativen Strukturen – Ambiguitäten und Varianz“ ed. by Stefan Engelberg & Irene Rapp ).

With respect to the synchronic disambiguation, it thus becomes obvious that the FoF reading is not triggered by any formal markers but is a result from the interaction between the deictic pattern and the perspectival structure of discourse context.

3.4Interim conclusion: Disambiguating factors and the emergence of viewpoints

The cross-linguistic comparison of FoF readings has thus shown that the emergence of viewpoint complexity relies on the interaction of the following parameters:

1. Viewpoint potential

   FoF readings are the result of the interaction between the deictic structure of the source construction and the perspectival structure of discourse context. The emergence of the narrator’s viewpoint can be seen as the result of the activation of a covert viewpoint potential provided by the narrative context.

2. Hierarchy of viewpoints

   The emergence of the narrator’s viewpoint is the result of a vertical split, which leads to a hierarchy of viewpoints. This is seen in the fact that FoF readings are blocked in contexts where a vertical split between speaker and evaluator or narrator and character is suppressed, as e.g. in Free Indirect Discourse.

3. Contrasting perspectives

   The disambiguation of FoF readings on the discourse level is supported by potential contrasts of truth-incompatible perspectives on the level of propositional content, i.e. for example, the beliefs of two mental subjects as alternative, contrasting viewpoints.

What the examples of FoF readings thus make obvious is that perspectivization as a discourse phenomenon is much more than ‘perspective taking’ in terms of perspective shifts. As has been shown before, the emergence of viewpoints is based on several micromechanisms of perspectivization (Zeman 2017Zeman, Sonja 2017 “Confronting Perspectives. Modeling Perspectival Complexity in Language and Cognition.” Glossa: a Journal of General Linguistics 2/1: 6 [Special issue Perspective-taking, ed. by Stefan Hinterwimmer, and Petra B. Schumacher]: 1–22.) rather than a simple viewpoint shift. This is the case both with respect to the emergence of multiple perspective readings of grammatical constructions as well as to the emergence of multiple viewpoints on the discourse level, as illustrated in Figure 2. While alternative viewpoints (i) are the prerequisite for any act of perspectivization, the case of FoF readings displays a more complex situation by presupposing not only a viewpoint shift on the horizontal axis, i.e. between two viewpoints functioning at the same level with respect to the perspectivized element, (ii) but also a viewpoint hierarchy as a result of a vertical viewpoint shift (iii). This hierarchical viewpoint setting is the prerequisite for the metaperspectival constellation and the integration of divergent viewpoints at the same time.

Figure 2.Emergence of viewpoints in multiple perspective constructions (Zeman 2017Zeman, Sonja 2017 “Confronting Perspectives. Modeling Perspectival Complexity in Language and Cognition.” Glossa: a Journal of General Linguistics 2/1: 6 [Special issue Perspective-taking, ed. by Stefan Hinterwimmer, and Petra B. Schumacher]: 1–22.)

In consequence, it seems inadequate to describe the disambiguation of FoF readings in terms of perspective ‘shifts’ in sense of (ii). Rather, it is the result of the interplay between the perspectival structure of the grammatical source construction and the perspectival structure provided by context. As seen in the next section, this is also supported by a diachronic look on FoF readings.