2.1A three-dimensional conceptual model

There are vast traditions of research into how interlocutors mutually negotiate their perspectives on an object of conceptualisation (see e.g. Langacker 1990Langacker, R. W. 1990 Concept, Image, Symbol: The Cognitive Basis of Grammar. Berlin: Mouton de Gruyter. ; Sweetser 1990Sweetser, E. 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge UP. ; Verhagen 2005Verhagen, A. 2005 Constructions of Intersubjectivity. Discourse, Syntax, and Cognition. Oxford: Oxford University Press.; Du Bois 2007Du Bois, John W. 2007 “The Stance Triangle.” In Stancetaking in Discourse: Subjectivity, Evaluation, Interaction, ed. by R. Englebretson, 139–182. Amsterdam: Benjamins. ) as well as into how the perspectives of third parties can be represented in language (see e.g. Fludernik 1993Fludernik, M. 1993 The Fiction of Language and the Languages of Fiction. London: Routledge.; Vandelanotte 2009Vandelanotte, Lieven 2009 Speech and Thought Representation in English: A Cognitive- Functional Approach. (Topics in English Linguistics 65). Berlin/New York: Mouton de Gruyter. ; Dancygier 2012Dancygier, B. 2012 The Language of Stories. A Cognitive Approach. New York: Cambridge UP.; Dancygier et al. 2016Dancygier, B., W. Lu, and A. Verhagen 2016 Viewpoint and the Fabric of Meaning. Form and Use of Viewpoint Tools across Languages and Modalities. Cognitive Linguistics Research [CLR] 55. Berlin/Boston: De Gruyter Mouton. ; cf. also Sandler and Pascual this issue). In Van Duijn (2016Duijn, M. J. van 2016 The Lazy Mindreader. A Humanities Perspective on Mindreading and Multiple-Order Intentionality. PhD Thesis, Leiden University., Chapter 5) and Van Duijn & Verhagen (in press)Duijn, M. J. van and A. Verhagen In press. “Beyond Triadic Communication: a Three-dimensional Conceptual Space for Modeling Intersubjectivity”. Pragmatics & Cognition. we present a conceptual model that seeks to integrate these traditions. Its purpose is not primarily to provide a way of drawing detailed schematic representations of single utterances and interaction events, but to make a structural point about how language is used to mutually coordinate the perspectives of a signaller, addressee, and third-party subjects. Nonetheless, graphic representations of the model can be used to depict how individual viewpoint configurations are organised, and we will do so below wherever this helps to illustrate our argument.

Human linguistic communication is prototypically characterised as ‘triadic’ as it is usually concerned with referents in the shared world of the communicators, unlike (most) animal communication, which is ‘dyadic’, only involving a relation between sender and receiver (Tomasello 2008Tomasello, M. 2008 Origins of Human Communication. Boston: MIT Press. ). Thus, linguistic communication is characterised as a configuration with two dimensions (cf. Figure 1): one of the communicative relation between a Speaker/Signaller and an Addressee (x), and another of the relation of joint attention by these subjects of conceptualisation to some object of conceptualisation (y).

Figure 1.Triadic communication, 2-dimensional. Signaller (S) and addressee (A) both assume a set of shared beliefs (the overlapping part of which is the common ground; cf. Clark 1996Clark, H. H. 1996 Using language. Cambridge: Cambridge University Press. ; Verhagen 2015Verhagen, A. 2015 ‘Grammar and Cooperative Communication’. In: Handbook of Cognitive Linguistics, (Handbooks of Linguistics and Communication Science 39), ed. by E. Dąbrowska and D. Divjak, 232–251. Berlin & Boston: De Gruyter. ) and subsequently negotiate how the common ground should be updated with respect to the object. The signals typically reflect both aspects of and operations on the relationship between S and A (the (x)-axis) and on the relationship between the common ground and the object (the (y)-axis)

Various traditions of research into (inter)subjectivity (Langacker 1987Langacker, R. W. 1987 Foundations of Cognitive Grammar: Volume I: Theoretical Prerequisites. Stanford: Stanford University Press.; Verhagen 2005Verhagen, A. 2005 Constructions of Intersubjectivity. Discourse, Syntax, and Cognition. Oxford: Oxford University Press.; Du Bois 2007Du Bois, John W. 2007 “The Stance Triangle.” In Stancetaking in Discourse: Subjectivity, Evaluation, Interaction, ed. by R. Englebretson, 139–182. Amsterdam: Benjamins. ; see Verhagen 2005Verhagen, A. 2005 Constructions of Intersubjectivity. Discourse, Syntax, and Cognition. Oxford: Oxford University Press., Chapter 1, for a discussion and references) have shown how several elementary semantic and grammatical problems (pertaining to negation, complementation, and others) can be explained in an approach that places their function squarely at the intersubjective axis (x), rather than on the language-to-world axis (y).

As human beings are highly social organisms, much of human communication is about people and their relations to each other and to the communicators (e.g. Dunbar 2008Dunbar, R. I. M. 2008 “Mind the Gap or Why Human Aren’t Just Great Apes.” Proceedings of the British Academy 154: 403–23.). Such communication thus not only involves events and their causes and consequences, but regularly also other human beings and their perspectives (emotions, desires, knowledge, etc.) as well as their communicative interactions. Van Duijn (2016Duijn, M. J. van 2016 The Lazy Mindreader. A Humanities Perspective on Mindreading and Multiple-Order Intentionality. PhD Thesis, Leiden University., Chapter 4–5) argues that these cannot be straightforwardly integrated into the structure of Figure 1: linguistic expressions that typically are about people’s desires, knowledge, etc. (constructions of the type “A thinks/knows/says/claims/… that p”) primarily do not constitute objects of conceptualisation in the discourse, but function as kinds of operators.11.The traditional equation of clausal complements with nominal complements, analysing both as direct objects of matrix predicates, has in modern research been argued to be mistaken (for an overview of arguments, see Verhagen (2005Verhagen, A. 2005 Constructions of Intersubjectivity. Discourse, Syntax, and Cognition. Oxford: Oxford University Press., 84–91)). Verhagen (2005Verhagen, A. 2005 Constructions of Intersubjectivity. Discourse, Syntax, and Cognition. Oxford: Oxford University Press., Chapter 3) presents empirical and theoretical considerations supporting the conclusion that the default function of matrix clauses of complement constructions is that of an operator on the propositional content of the complement clause. For further critical discussion of the issue, cf. Boye & Harder (2007)Boye, K. and P. Harder 2007 “Complement-Taking Predicates: Usage and Linguistic Structure.” Studies in Language 31: 569–606. , Newmeyer (2010)Newmeyer, F. J. 2010 “What Conversational English Tells Us About the Nature of Grammar: A Critique of Thompson’s Analysis of Object Complements.” In Language Usage and Language Structure, ed. by K. Boye and E. Engberg-Pedersen, 3–43. Berlin/New York: De Gruyter Mouton., and Verhagen (2010)Verhagen, A. 2010 “Usage, Structure, Scientific Explanation, and the Role of Abstraction, by Linguists and by Language Users.” In Language Usage and Language Structure, ed. by K. Boye and E. Engberg-Pedersen, 45–72. Berlin/New York: De Gruyter Mouton.. Therefore we argue that they must be seen as representing a separate dimension of ‘other minds’, with the same cognitive and communicative capabilities as S and A, but not participating in the communicative situation. Consider Figure 2:

The front plane is identical to the 2D-model of Figure 1, representing the basic structure of linguistic communicative events. What is added is a third dimension (z) of the Signaller and Addressee coordinating their joint interest in an object of conceptualisation with other parties.

The 3D-model of Figure 2 straightforwardly represents the communicative situation in dialogues as we encounter them on a daily basis in our social lives. Two interlocutors have a joint interest in clarifying some issue, and they do so by mentally coordinating with represented third parties who are considering this very issue. It is the character of ‘mentally coordinating’ that makes the (z)-axis different from that of jointly attending to an object (the (y)-axis). Figures 3a and 3b depict examples of two such cases; one in which only one third party is represented and one in which two third parties are being represented as themselves engaged in communicative interaction.

The examples in Figure 3a and 3b both involve explicit marking of the third-party perspectives (“he thinks…”; “John assured Mary…”), but this is not necessary for the (z)-axis to be active. In, for instance, “Supposedly, it will be sunny tomorrow” there is no third party represented ‘on stage’, yet, arguably, a source (or sources) with a non-specified identity is (are) conjured up in the interaction between S and A using the adverb “supposedly”. Within the model, “supposedly” would be described as an operator on both the (x)- and (z)-axes: it lowers the epistemic commitment of S to the statement that it will be sunny tomorrow by evoking in A the inference that this is what non-specified others think or say (cf. Section 4.1 below and Van Duijn 2016Duijn, M. J. van 2016 The Lazy Mindreader. A Humanities Perspective on Mindreading and Multiple-Order Intentionality. PhD Thesis, Leiden University., Chapter 4–5).

Clearly, not every instance of communicative interaction involves the representation of third-party perspectives. A sentence such as “It will be sunny tomorrow”, uttered in an ordinary interaction event, can be discussed exhaustively in terms of the two dimensions represented by (x) and (y) in the model. This is different, however, as soon as the narrative mode is used, given that narratorship by default implies distancing of the speaker from the ‘voice’ that delivers the contents of the story (for a detailed and convincing discussion of this insight and its history in narrative theory, see Dancygier 2012Dancygier, B. 2012 The Language of Stories. A Cognitive Approach. New York: Cambridge UP., Chapter 2 and 5, and Zeman 2016Zeman, S. 2016 “Perspectivization as a Link Between Narrative Micro- and Macro-Structure.” In Perspectives on Narrativity and Narrative Perspectivization ed. by S. Zeman, and N. Igl, 17–43. Amsterdam: John Benjamins. ). In any narrative, the narrator can appear ‘on stage’ and incarnate as a character of sorts (cf. Humbert Humbert in Lolita, as discussed below), or remain implicit and rather take the form of an abstract instance to which choices of what is recounted, and how, must be attributed. In both cases, representations of the perspectives of other subjects in the narrative mode (i.e. characters) necessarily need to pass the ‘filter’ of the narrator’s overall vantage point. As indicated in Figure 4, we then have a duplication of the second plane of Figure 2, but this does not add to the dimensionality of the configuration: the quality of the (z)-axis remains the same.

In this way, conceptual embedding of viewpoints can be represented along the (z)-axis of the model. Everything that is said or thought by the characters in Figure 4 needs to pass ‘through’ the viewpoint of the narrator before entering the communicative ground of the interaction between S and A. More elaborate examples of viewpoint embedding and their representation in the model follow in Section 3.

2.2Viewpoint embedding: Communicative and cognitive aspects

Quite generally, the topics of viewpoint and language intersect in a variety of disciplines, each with a different focus and set of terms to refer to concepts that are partly parallel and partly distinct. In philosophy of mind and the cognitive sciences it is customary to use language merely as a meta-analytic tool to describe states of mind and their relationships at the cognitive level of agents.22.An example would be the embedded proposition from Zunshine (2006)Zunshine, L. 2006 Why We Read Fiction. Theory of Mind and the Novel. Columbus: The Ohio State University Press. cited below. For more examples in this tradition see Dennett (1987)Dennett, D. C. 1987 The Intentional Stance. Cambridge MA: MIT Press., Dunbar (2005, 2008Dunbar, R. I. M. 2008 “Mind the Gap or Why Human Aren’t Just Great Apes.” Proceedings of the British Academy 154: 403–23.), and Zunshine (2016)Zunshine, L. 2016 “The Commotion of Souls.” Substance 140 (45): 2: 118–142. . For discussions see Van Duijn (2016Duijn, M. J. van 2016 The Lazy Mindreader. A Humanities Perspective on Mindreading and Multiple-Order Intentionality. PhD Thesis, Leiden University., Chapter 1) and Apperly (2011)Apperly, I. 2011 Mindreaders. The Cognitive Basis for Theory of Mind. New York: Psychology Press.. These agents are not necessarily humans and the interaction events that are studied are not necessarily communicative, let alone verbal. Traditions in linguistics and narratology, on the other hand, are primarily concerned with how viewpoints are represented in various forms of communicative interaction, verbally and non-verbally, and with how communicative acts affect the viewpoints of discourse participants and vice versa (e.g. Leech and Short 2007 [1981]Leech, G., and M. Short 2007 [1981] Style in Fiction. A Linguistic Introduction to English Fictional Prose. Second Edition. Harlow: Pearson Education Limited.; Toolan 1998Toolan, M. 1998 Language in Literature: An Introduction to Stylistics. London: Arnold.; Vandelanotte 2009Vandelanotte, Lieven 2009 Speech and Thought Representation in English: A Cognitive- Functional Approach. (Topics in English Linguistics 65). Berlin/New York: Mouton de Gruyter. ; Dancygier and Sweetser 2012Dancygier, B. and E. Sweetser 2012 Viewpoint in Language. A Multimodal Perspective. Cambridge: Cambridge UP. ; Dancygier et al. 2016Dancygier, B., W. Lu, and A. Verhagen 2016 Viewpoint and the Fabric of Meaning. Form and Use of Viewpoint Tools across Languages and Modalities. Cognitive Linguistics Research [CLR] 55. Berlin/Boston: De Gruyter Mouton. ). Our approach in this paper builds on concepts from both the traditions in the cognitive sciences and the linguistic/narratological traditions. Note that we use the terms ‘viewpoint’, ‘perspective’, and ‘mindstate’ interchangeably when referring to a discourse participant’s thoughts and perceptions. For a discussion of usages of these terms across various traditions, and differences between these usages, see Van Duijn (2016Duijn, M. J. van 2016 The Lazy Mindreader. A Humanities Perspective on Mindreading and Multiple-Order Intentionality. PhD Thesis, Leiden University., Chapter 1) and Vandelanotte (this issue).

In work by a wide array of cognitive scientists, psychologists, philosophers, and linguists it is suggested that humans must be capable of dealing with multiple perspectives at the same time. This capability has almost without exception been conceptualised as a matter of recursive viewpoint embedding, and it has been argued to be fundamental to cooperation, linguistic communication, understanding literary fiction, and a range of other activities that are part of living in a socio-cultural environment.33.See e.g. Grice (1957)Grice, H. P. 1957 “Meaning.” The Philosophical Review 66 (3): 377–388. , Bennett (1976)Bennett, J. 1976 Linguistic Behaviour. Cambridge University Press., Dennett (1987)Dennett, D. C. 1987 The Intentional Stance. Cambridge MA: MIT Press., and Corballis (2011)Corballis, M. C. 2011 The Recursive Mind. The Origins of Human Language, Thought, and Civilization. Princeton: Princeton UP. in the philosophical tradition; Sperber (1994)Sperber, D. 1994 “Understanding Verbal Understanding.” In: What is Intelligence?, ed. by J. Khalfa, 179–198. Cambridge: Cambridge UP. and Scott-Phillips (2015)Scott-Phillips, T. C. 2015 Speaking Our Minds. Why Human Communication is Different and How Language Evolved to Make it Special. New York: Palgrave Macmillan. in linguistics; and Shultz & Dunbar (2007)Shultz, S. and R. I. M. Dunbar 2007 “The Evolution of the Social Brain: Anthropoid Primates Contrast with Other Vertebrates”. Proceedings of the Royal Society, London 274B: 2429–2436. , Yoshida et al. (2010)Yoshida, W., B. Seymour, K. J. Friston, and R. J. Dolan 2010 “Neural Mechanisms of Belief Inference During Cooperative Games.” Journal of Neuroscience 30: (32): 10744–51. , and Launay et al. (2015)Launay, J., E. Pearce, R. Wlodarski, M. J. van Duijn, J. Carney, and R. I. M. Dunbar 2015 “Higher-Order Mentalising and Executive Functioning.” Psychology and Individual Differences 86: 6–14. in the cognitive sciences, all putting emphasis on the importance of recursive embedding of viewpoints. Across different traditions, propositions with embedded clauses representing perspective layers have been used as meta-analytic tools to represent multi-viewpoint situations and cognitive tasks that readers have to perform in order to deal with them. For example, cognitive literary scholar Lisa Zunshine has argued that for understanding and appreciating a work such as Virginia Woolf’s Mrs Dalloway (1925Woolf, V. 1925 Mrs Dalloway. London: Hogarth Press.), readers need to complete strands of reasoning of the following form:

Woolf intends us to recognize […] that Richard is aware that Hugh wants Lady Bruton and Richard to think that because the makers of the pen believe that it will never wear out, the editor of the Times will respect and publish the ideas recorded by this pen.(Zunshine 2006Zunshine, L. 2006 Why We Read Fiction. Theory of Mind and the Novel. Columbus: The Ohio State University Press., 33, italics in original)

Elsewhere (see Van Duijn et al. 2015Duijn, M. J. van, I. Sluiter, and A. Verhagen 2015 “When Narrative Takes Over: The Representation of Embedded Mindstates in Shakespeare’s Othello’. Language and Literature 24: (2): 148–166. ; Van Duijn 2016aDuijn, M. J. van 2016a “Van binnenuit bekeken. Gedachtenlezen en ingebedde perspectieven in Mrs Dalloway en De maagd Marino .” Tijdschrift voor Nederlandse Taal- en Letterkunde 132: 383–400.) we have argued that such embedded propositions are problematic, even if they are ‘only’ used as a tool to represent viewpoint complexity at a conceptual level.44.In the cited passage Zunshine clearly does not claim that the text contains such recursively embedded sentences. However, what she does claim is that the different viewpoint phenomena present in the novel’s text amount to a viewpoint arrangement that can be justly represented using a proposition of this form. It is this second claim that we argue against: recursive embedding is certainly an important characteristic when studying viewpoints in language, but its role should not be overstated and other types of relations should not be reduced to it. Besides the fact that they make the task a reader has to perform look unnecessarily complex and opaque, they easily mis-/underrepresent all kinds of nuances and details apparent in the actual text or situation. In Section 3 below we will get back to this point.

In our daily social lives, by definition, others’ viewpoints cannot be accessed directly. They can be appraised through an inferential process based on behavioural cues (including linguistic utterances) and immediate circumstances. This latter process is often referred to as ‘mindreading’ (or ‘using one’s Theory of Mind’; Premack and Woodruff 1978Premack, D. G., and G. Woodruff 1978 “Does the Chimpanzee Have a Theory of Mind?” Behavioral and Brain Sciences 1: 515–526. ; Apperly 2011Apperly, I. 2011 Mindreaders. The Cognitive Basis for Theory of Mind. New York: Psychology Press.). Various scholars have pointed out that the process by which language users form an understanding of the viewpoints and mindstates of people or characters referred to in discourse (i.e. those represented along the (z)-axis in the 3D-model) can be seen as a special case of mindreading (e.g. Palmer 2004Palmer, A. 2004 Fictional Minds. Lincoln: University of Nebraska Press.; Verhagen 2005Verhagen, A. 2005 Constructions of Intersubjectivity. Discourse, Syntax, and Cognition. Oxford: Oxford University Press., Chapter 1; Zunshine 2006Zunshine, L. 2006 Why We Read Fiction. Theory of Mind and the Novel. Columbus: The Ohio State University Press.; Budelmann and Easterling 2010Budelmann, F. and P. Easterling 2010 “Reading Minds in Greek Tragedy.” Greece and Rome 57 (2): 289–303. ; Cefalu 2013Cefalu, P. 2013 “The Burdens of Mind Reading in Shakespeare’s Othello: A Cognitive and Psychoanalytic Approach to Iago’s Theory of Mind.” Shakespeare Quarterly 64 (3): 265–294. ). Thus, in the physical presence of an interlocutor (say, when speaking with John), one has direct access to verbal and non-verbal cues that may guide an inferential process regarding John’s mindstates – the classic case of mindreading as discussed in a psychology textbook. However, when John is not physically present but referred to in a conversation we have with Pete (or in a story narrated by Pete), a similar inferential process, drawing on the same capacity for mindreading, can be triggered through cues in Pete’s speech that enable us to construe aspects of John’s inner life.

In summary, and applied to our object of study in this paper: building up an understanding of a thoughtscape underlying a discourse situation or narrative can be seen as a form of mindreading in which linguistic cues guide the inferential process. Across multiple traditions, viewpoint is often construed as something that is with one party at a time only, and that can ‘shift’ back and forth between different subjects. Furthermore, there has been a tendency to focus on the recursive embedding of viewpoints, often to the exclusion of other relation types. In what follows we will compare these aspects (single viewpoint at a time, shifting, focus on embedded layers) with observations from an analysis of the perspective structure in Nabokov’s novel Lolita and various examples from more ‘daily-life’ discourse situations in Dutch and Russian. We have chosen these examples because we consider them to be illustrative of the argument we develop in this paper. Our approach here should thus mainly be seen as exploratory, and would ideally be followed up with the analysis of more systematically collected cases in the future.

3.1The thoughtscape

Vladimir Nabokov’s classic novel Lolita (1997 [1959]) presents the story of a middle-aged protagonist, literary scholar by profession, who falls in love with his landlady’s underaged daughter named Dolores Haze; Lolita for him. The protagonist, who calls himself Humbert Humbert, marries mother Haze in an attempt to stay as close to Lolita as possible. When mother Haze finds out about Humbert’s fantasies and true intentions she runs out in panic and is lethally hit by a car. Making use of the fact that people regard him as Lolita’s stepfather, Humbert takes Lolita on a road trip through various American states. The two develop a twisted relationship of love, sex, and mutual exploitation, which ends in Humbert killing another man, called Quilty, after Lolita ran off with him. The text presented in the novel is almost entirely constructed as Humbert’s post-factum writings while he is in prison, intended to be disclosed only after his death.

The way in which readers get access to the events and characters (including their perspectives) that constitute the plot is notorious among scholars as well as readers. Are we completely at the mercy of what the imprisoned Humbert tells us? Or are there other voices as well? How, and to what extent, do we as readers actually get to know the title character, Dolores Haze, alias Lolita? The analysis of the novel’s thoughtscape presented below, the network of interrelated viewpoints underlying the plot, can help answer such questions. However, more important in the context of this paper is that various examples from the novel illustrate how conventional viewpoint phenomena can work unpredictably and irregularly when combined to express a complex thoughtscape. It is important to note that the discussion that follows by no means strives to be a complete analysis of viewpoint and narration in Lolita, let alone a just treatment of the ethical implications and problematic reception of the novel. For more in-depth treatments we refer to Zeman (2016)Zeman, S. 2016 “Perspectivization as a Link Between Narrative Micro- and Macro-Structure.” In Perspectives on Narrativity and Narrative Perspectivization ed. by S. Zeman, and N. Igl, 17–43. Amsterdam: John Benjamins. , Phelan (2007)Phelan, J. 2007 “Estranging Unreliability, Bonding Unreliability, and the Ethics of “Lolita”.” Narrative 15 (2): 222–238. , Zunshine (2006)Zunshine, L. 2006 Why We Read Fiction. Theory of Mind and the Novel. Columbus: The Ohio State University Press., and Boyd (1991)Boyd, B. 1991 Vladimir Nabokov: The American Years. Princeton, NJ: Princeton UP. .

The book opens with a “foreword” signed by a certain “John Ray, Jr., Ph.D.”, who claims to be appointed by Clarence Clark, the late Humbert’s lawyer, to edit Humbert’s writings and prepare them for print. John Ray writes that he presents Humbert’s memoirs mostly intact, “save for the correction of obvious solecisms and careful suppression of a few tenacious details that despite H. H.’s own efforts still subsisted in his text as signposts and tomb-stones (indicative of places or persons that taste would conceal and compassion spare)” (Nabokov 1997 [1959]Nabokov, V. 1997 [1959] Lolita. London: Penguin., 5). In other words, throughout the entire novel the reader can never be sure whether any passage is solely Humbert’s work, or also subject to one of Ray’s suppressive interventions. At the same time, Ray also refers to newspapers that report on the ‘real’ events, which means that his editing work may also have the effect of a ‘reality check’ – would he have published Humbert’s memoirs if they differed too much from the news facts?

The perspective structure within Humbert’s memoirs adds more complexity. Humbert acts as a first-person narrator whose ‘here-and-now’ is in a US prison in the period between his arrest in September 1952 and his death on November 16th, 1952, which was “a few days before his trial was scheduled to start” according to Ray (5). At various occasions, he addresses his writings to his lawyer Clarence, for example with the following remark between brackets: “I notice the slip of my pen in the preceding paragraph, but please do not correct it, Clarence” (32). At other points he addresses the jury of his court case: “Ladies and gentlemen of the jury!” (109). Yet at the same time the imprisoned Humbert writes for an audience. This not only becomes clear from his will (relayed by John Ray) that his memoirs should be published after his death, but also because he regularly seeks to address his readers. Sometimes he does so in the third person, for example: “I want my learned readers to participate in the scene I am about to replay; I want them to examine its every detail” (56). At other points addressing occurs more directly, through remarks such as “if you can still stand my style (I am writing under observation)” (10). In these and many more examples we seem to see an erudite detainee who is writing his memoirs in front of a heterogeneous audience, while dramatising his awareness of this audience.

Some memories are relayed directly from the perspective of Humbert-the-detainee, such as the tale of his divorce from Valeria, from which the following fragment is taken:

I cannot say he[r new lover, a White-Russian ex-colonel] behaved insolently or anything like that; on the contrary, he displayed, as a small sideshow in the theatricals I had been inveigled in, a discrete old-world civility, punctuating his movements with all sorts of mispronounced apologies (j’ai demandé pardonne – excuse me – est-ce que j’ai puis – may I – and so forth), and turning away tactfully when Valechka took down with a flourish her pink panties from the clothesline above the tub(29)

“Valechka” is what the ex-colonel calls Valeria – by using it here, Humbert’s account invites us to consider the viewpoint of the ex-colonel. At the same time, by using this form instead of her real name, Humbert gives expression to his disdain for the ex-colonel with his “old-world civility”. The perspective from which the scene of Valeria/Valechka taking down her panties is construed can thus be analysed as a mixed viewpoint with elements of the ex-colonel’s perspective, Humbert’s perspective as a participant in and witness of this scene at the time, Humbert-the-detainee’s perspective remembering and narrating it, and, potentially, John Ray Jr., who may have edited Humbert’s account.55.For an account of mixed perspectives in terms of conceptual blending see Dancygier (2012Dancygier, B. 2012 The Language of Stories. A Cognitive Approach. New York: Cambridge UP., Chapter 4).

In other parts Humbert reconstructs fragments from a 1947 diary which, he says, “was destroyed five years ago […] what we examine now (by courtesy of a photographic memory) is but its brief materialization” (40). In the day-by-day notes presented to the reader it remains largely unclear how much was really recorded by Humbert back then and how much was distorted or even invented by Humbert-the-detainee – let alone the question of whether Ray did intervene here. In addition, there seems to be a Humbert writing in his diary at night and one living through the recorded events at day. Which one of the two is responsible for the exegesis of mindstates and intentions behind Lolita’s, her mother’s, and others’ behaviour? See, for example, the following passage in which Humbert, who is then still Lolita’s mother’s lodger, is in his room, writing, when Lolita comes in and leans over his shoulder:

I felt the heat of her limbs throughout her tomboy clothes. All at once I knew I could kiss her throat or the wick of her mouth with perfect impunity. I knew she would let me do so, and even close her eyes as Hollywood teaches. A double vanilla with hot fudge – hardly more unusual than that.(48)

In some sense, the final sentence of this fragment entails a full perspective shift to Lolita: it is from her perspective that we should understand the imagined kiss to be like a vanilla ice cream. However, clearly, the way in which this view reaches the reader is quite a deal-changer here: Humbert Humbert, via his memory of his diary containing his reading of Lolita’s behaviour, relaying the information that she didn’t mind kissing him at that moment… Or even would have liked kissing him? Or should we perhaps understand the opposite here, would have felt to be forced into an inappropriate kiss? We know that it must be Humbert’s remembered reading of Lolita, because this is marked two times by “I knew” followed by a modal complement clause (“could…”, “would…”).

The evaluation of which of these readings holds best (eventually taking place on the reader’s part) clearly depends on the assessment of the total viewpoint structure at any point in the text. An analysis of this structure in terms of shifts between single viewpoints, whether or not conceptualised in terms of a series of recursively embedded layers, is largely insufficient. Rather, the assessment of the object of conceptualisation constantly seems to involve the ability to assess and compare a network of simultaneously relevant perspectives that are mutually related and mixed in all sorts of ways. The two ways of mapping the structure as embedded layers versus our alternative of the ‘thoughtscape’ are depicted in Figure 6a and 6b below.

3.2Conventional patterns, unpredictable viewpoint structure

Dancygier (esp. 2012Dancygier, B. 2012 The Language of Stories. A Cognitive Approach. New York: Cambridge UP.) has demonstrated for a great deal of literary texts how micro-level viewpoint phenomena (mostly lexical items and grammatical patterns capable of indicating perspective) add up and ‘blend’ (Turner & Fauconnier 1995Turner, M., and G. Fauconnier 1995 “Conceptual Integration and Formal Expression.” Metaphor and Symbolic Activity 10 (3): 183–203. ) as a story progresses, thereby forming a coherent viewpoint structure at the macro-level. Her work reflects and partly encompasses the long tradition in narratology and literary stylistics of studying viewpoint/perspective. A special asset of this tradition is having an analytical vocabulary for mixed viewpoints. The most well-known mixed-viewpoint ‘construction’ from the field of narratology is probably free indirect discourse (FID; Banfield 1978Banfield, A. 1978 “Where Epistemology, Style, and Grammer Meet Literary History: The Development of Represented Speech and Thought.” New Literary History 9 (3): 415–454. ; Fludernik 1993Fludernik, M. 1993 The Fiction of Language and the Languages of Fiction. London: Routledge.; Mey 1999Mey, J. 1999 When Voices Clash. A Study in Literary Pragmatics. Berlin/New York: Mouton De Gruyter. ; Bal 2009Bal, M. 2009 Narratology. Introduction to the Theory of Narrative. Toronto: Toronto UP.), in which multiple voices or strains of thought are interwoven in such a way that it is not always clear which part should be ascribed to whom. As Zeman (2016Zeman, S. 2016 “Perspectivization as a Link Between Narrative Micro- and Macro-Structure.” In Perspectives on Narrativity and Narrative Perspectivization ed. by S. Zeman, and N. Igl, 17–43. Amsterdam: John Benjamins. , 26–32) points out, rather than being just local phenomena at various points in a text, instances of FID are “surface effects” through which an underlying structure becomes visible: the possibility of its existence teaches us something about the structure of a text as a whole, and about the narrative mode more widely (see also Vandelanotte this issue).

In this light, consider a third and final example, including what seems to be a representation of Lolita’s voice using direct discourse:

I had thought that months, perhaps years would elapse before I dared to reveal myself to Dolores Haze; but by 6 she was awake, and by 6.15 we were technically lovers. I am going to tell you something very strange: it was she who seduced me. […]

‘You mean,’ she persisted now kneeling above me, ‘you never did it when you were a kid?’

‘Never,’ I said quite truthfully.

‘Okay,’ said Lolita, ‘here is where we start.’(117)

It is particularly salient that the subject matter of the represented dialogue is relevant to Humbert’s court case: was it ultimately he who forced his stepdaughter into their vexed relationship, or did she take some initiative as well? The same holds as in the case of the “double vanilla with hot fudge” quote: the represented conversation is relayed to the reader under the highly suspicious circumstances of Humbert-the-detainee telling from his memory what an earlier version of himself (‘Humbert-the-character’) experienced. However, what this passage makes clear is that it is insufficient at any point to analyse only ‘local’ patterns of viewpoint construction: in isolation, direct discourse suggests a viewpoint shift in its fullest form, whereas in the current context it is highly questionable to what degree such a shift is realised (cf. also Vandelanotte’s narrative examples in this issue). It can even be argued that there is no perspective shift at all, in which case this is an example of perspective persistence in terms of this issue’s Introduction. This suspense between, on the one hand, the conventional meanings of each of the viewpoint phenomena when considered locally and, on the other hand, the unpredictable total viewpoint structure they add up to is ubiquitous in Nabokov’s Lolita – and, indeed, famously so. We expect that similar cases can be found in many more narrative texts, as well as in more daily forms of language usage. In the next section we discuss two specific cases, but suggest that the line of argumentation has wider applicability.

4.1Citation and inquit constructions in Dutch

The first case in point concerns two different ordering patterns of a reporting clause (of the type President Obama stated) and a reported clause with main clause syntax (President Obama said: “France is our oldest ally”, “France is our oldest ally”, President Obama said). For this study, we looked at Dutch (English has the same, or at least very similar, patterns). In Dutch, the difference between these two is not only the ordering of the two clauses; when the reporting clause is postposed, it always has the verb preceding the subject (so-called ‘inversion’; cf. Schelfhout 2000Schelfhout, C. 2000 “Corpus-Based Analysis of Parenthetical Reporting Clauses.” In Computational Linguistics in the Netherlands 1998; Selected Papers from the Ninth CLIN Meeting, ed. by F. I. Van Eynde, I. Schuurman, and N. Schelkens, 147–59. Amsterdam: Rodopi.; de Vries 2006Vries, M. de 2006 “Reported Direct Speech in Dutch”. Linguistics in the Netherlands 23: 212–223. ): the order Subject-Verb (President Obama zei, as in (3)) is disallowed in patterns like (4) in Dutch, for all verbs and subjects.

In narratology, cases such as these are considered two variants of direct discourse, both equally distinct from indirect discourse as in President Obama zei dat Frankrijk onze/Amerika’s oudste bondgenoot was (“President Obama said that France was our/America’s oldest ally” – with subordinate clause syntax of the reported clause, and the appropriate deictic shifts, partly depending on the author’s ‘virtual’ position inside or outside America). However, although the order of (4) never exhibits subordinate syntax in the (initial) reported clause, it may and does occur with various deictic shifts, as in (4′), (4″) and (4‴) (in several respects resembling so-called free indirect discourse, though not completely identical to it; cf. also Vandelanotte this issue):

A corpus study in Dutch periodicals (SONAR) of the verbs used in the reporting clause in each pattern reveals that, besides the general verbs of communication and cognition that occur in both (and that have a high token frequency), there are large sets of verbs (with low token frequency, but contributing to the type frequency of the verbal slot) that occur in only one of the two (Verhagen, 2019Verhagen, A. 2019 “Grammaticale stilistiek en stilistische grammatica – Varianten van redeweergave in het Nederlands.” Nederlandse Taalkunde/Dutch Linguistics 24.). For example, verbs that indicate an interpretation of the manner of speaking or some accompanying behaviour (not, or not necessarily, intended), such as grinniken (chuckle), verzuchten (sigh), jubelen (cheer), kreunen (moan), scanderen (chant), schimpen (scoff), verdedigen (defend), mopperen (grumble), toesnauwen (snarl), glimlachen (smile), glunderen (beam), occur only in postposed reporting clauses (type (4)). The same is true for a range of verbs indicating an attitude of the reported speaker as assessed by the present speaker/narrator such as filosoferen (ponder), mijmeren (muse), poneren (postulate). It is telling that precisely two impersonal expressions −heet(te) het (“it is/was reported”) and klonk het (“it sounded”)− are limited to postposed reported clauses and occur there relatively frequently: as impersonals, these do not even allow a full shift to the perspective of any specific referent responsible for the speech act being represented and thus clearly make the narrator’s perspective persist.

Thus, both the conditions on the reported clause and those on the predicate of the reporting clause differ between the two patterns exemplified in (3) and (4), with the properties of these two ‘slots’ in the latter indicating a higher degree of involvement of the speaker/narrator: the reported clause allowing certain features of free indirect discourse, the predicates of the reporting predicates often indicating the present speaker’s assessment. On this basis, we conclude that the two patterns actually represent not different variants of the same type of represented discourse, but rather two different (conventional) grammatical constructions, each with its own function. The construction with a preposed reporting clause as in (3) is a ‘full blown’ shifting construction, relegating all responsibility to the represented character, i.e. with no (or minimal) marking of the way the represented discourse impinges on the assessment of the situation in the communicative ground; we propose to characterise it as the ‘Citation Construction’. The other construction, with a postposed reporting clause as in (4), we propose to label ‘Inquit Construction’. It has the function of providing a characterisation, by (whoever is) the present narrator, of the piece of discourse that has just been (re)presented, linking it to the ground, thus making it compatible (unlike the Citation Construction) with the presence of one or more indicators of narrator perspective in this discourse (cf. Vandelanotte this issue, for some other, similar phenomena). At the same time, the Inquit construction does not evoke as high a degree of responsibility as a complementation construction, i.e. what is traditionally called “indirect discourse” (President Obama said that France was America’s oldest ally): in the Inquit construction, the syntax of the reported clause always indicates a speech act performed by the character (see Verhagen 2019Verhagen, A. 2019 “Grammaticale stilistiek en stilistische grammatica – Varianten van redeweergave in het Nederlands.” Nederlandse Taalkunde/Dutch Linguistics 24. for further discussion). Thus, we have a whole range of possible distributions of responsibility between narrator and characters that cannot be captured in a theory that only relates different perspectives in terms of embedding and shifting from one layer to the other.

Our three-dimensional model, on the other hand, can easily accommodate these observations. Consider Figure 2 again, repeated here for convenience, which gives the most elementary form involving Signallers, Addressees, and Others concerned with construing some object of conceptualisation:

The Dutch Citation construction completely shifts the viewpoint from the front plane of S and A along the (z)-axis to some Other party, so that the connection between that party and the object of conceptualisation is put ‘on stage’ and made highly prominent. With a complementation construction (‘indirect discourse’), on the other hand, the Other party’s viewpoint is also activated, but S explicitly maintains control over much of the way the object is presented to A (the (x)-axis): the role of the (y)-axis is never reduced to zero, as it is necessarily S who is describing a character’s utterance (or thought). In an instance of the Inquit construction, finally, the relative prominence of the (y)-axis in the front plane and the connection between Other party and object of conceptualisation is very much dependent on other viewpoint-indicating elements, as we have seen. But the latter connection is never reduced to zero, as it is the performance of a character’s speech act (or ‘thought act’) that is represented in the text (while it may easily be subsequently qualified in one way or another). Thus, the specifics of the grammatical and lexical Dutch ‘toolkit’ for viewpoint marking can all be naturally accommodated in this model, including the relatively high degree of flexibility in certain cases, such as the Inquit construction, in allowing a shifting of perspective from one plane to another or not.

4.2The Russian apprehensive construction

Our second grammatical demonstration involves a ‘classic’ problem in Russian syntax (which has analogues in several other languages (Horn 2010Horn, L. R. 2010 “Multiple Negation in English and Other Languages.” In The Expression of Negation, ed. by L. R. Horn, 111–148. Berlin/New York: De Gruyter Mouton. ), e.g. French – but we are only concerned with Russian here). The relevant phenomenon is illustrated by the contrast between (5) and (6):

The remarkable point is that in Russian the reported clause in (5) contains a marker of negation (ne), i.e. when this clause is introduced by kak by, by being a marker of subjunctive mood (in the Russian grammatical tradition), while it is absent in (6), with a future marker (“I am afraid that he will fall ill”). The ‘problem’ then is that the negation seems ‘superfluous’ or even ‘illogical’ in (5), as what is feared is, of course, his falling ill, not its contrary. It has been noted in the past that the use of such a subjunctive negative clause is also limited to a specific semantic set of predicates (predicates of fear and apprehension), and this is confirmed in corpus research (cf. Baydina 2017Baydina, E. 2017 “The Russian Apprehensive Construction: Syntactic Status Reassessed, Negation Vindicated”. MA Thesis Leiden University. [https://​openaccess​.leidenuniv​.nl​/handle​/1887​/46284], 42–46).66.The only verb that does not include apprehension in its lexical meaning and that occurs with kak by-‘complements’ in the Russian National Corpus with some frequency is dumat (‘think’) (51 of 421 cases, i.e. 12%). The full expressions involved generally have a clear apprehension or a wish-of-negation (‘think that something should not happen’) reading (see Baydina 2017Baydina, E. 2017 “The Russian Apprehensive Construction: Syntactic Status Reassessed, Negation Vindicated”. MA Thesis Leiden University. [https://​openaccess​.leidenuniv​.nl​/handle​/1887​/46284] for details and further discussion). It is also known that kak by clauses can be used independently (cf. Nilsson 2012Nilsson, N. Z. 2012 “Peculiarities of Expressing the Apprehensive in Russian.” The Russian Verb. Oslo Studies in Language 4 (1): 53–70.); (7) basically expresses the same message as (5):

A comparison between (5) and (7) may suggest that we are dealing with a case of so-called insubordination in (7), defined as “the conventionalized main clause use of what, on prima facie grounds, appear to be formally subordinate clauses” by Evans (2007Evans, N. 2007 “Insubordination and Its Uses.” In: Finiteness: Theoretical and Empirical Foundations, ed. by I. Nikolaeva, 366–431. Oxford: Oxford UP., 367). In this way, the single clause expression (as in (7)) is seen as a special kind of use of an apparent subclause (as in (5)). But theoretically, it is just as suitable to reverse the relationship, i.e. to analyse the complex expression (5) as involving a special kind of use of the single clause expression (7). We will now argue that this analysis is actually to be preferred: (5) instantiates a conventionalized combination of an independent negative clause type with a predicate of fear, which may be rendered in (not so idiomatic or conventional) English as I am afraid (of something) – May he not fall ill!, and the relationship between these components does not involve embedding.

A first indication is the fact that the independent use is by far the most frequent (actually accounting for more than 50% of the instances of the pattern in the Russian National Corpus; Baydina 2017Baydina, E. 2017 “The Russian Apprehensive Construction: Syntactic Status Reassessed, Negation Vindicated”. MA Thesis Leiden University. [https://​openaccess​.leidenuniv​.nl​/handle​/1887​/46284]). A kak by construction may even be the main clause in a conditional construction, as in (8), from the RNC.

In this kind of environment, the kak by construction does not look like a subordinate clause being used independently, not even on ‘prima facie’ grounds. Second, while the use as a kind of complement of fear and apprehension predicates is the next most frequent type of use (almost 40% of the corpus instances), there is also a group of cases (almost 10%) in which a kak by-construction should rather be classified as a kind of adjunct; cf. (9), also from the RNC:

Here the fear-predicate does have a complement, viz. “to move”, so that the kak by-construction cannot be assumed to fill a complement slot (hence the translation as an adverbial adjunct, headed by lest).77.The English, somewhat obsolescent, conjunction lest is another instance of a grammaticalized form marking a proposition as an object of fear/apprehension, in this case with the negation semantically incorporated (the conjunction may be glossed as “so that not”). Cf. Lichtenberk (1995)Lichtenberk, František 1995 “Apprehensional Epistemics”. In Modality in Grammar and Discourse, ed. by J. Bybee, and S. Fleischman, 293–327. Amsterdam/Philadelphia: John Benjamins Publishing Company. for discussion and comparison.

On the basis of the distributional data and the apparent autonomy of the kak by-construction, Baydina (2017)Baydina, E. 2017 “The Russian Apprehensive Construction: Syntactic Status Reassessed, Negation Vindicated”. MA Thesis Leiden University. [https://​openaccess​.leidenuniv​.nl​/handle​/1887​/46284] concludes that the underlying problem preventing a proper understanding of expressions like (5) is the implicit assumption that the relation between the two clauses is one of embedding. Applying a construction grammar approach, prefigured by Jespersen (1917)Jespersen, O. 1917 Negation in English and Other Languages. Copenhagen: Host., she proposes that (5) instantiates an autonomous construction combining two clauses more in a paratactic than in a hypotactic way (recall the English paraphrase I am afraid (of something) – May he not fall ill! used above).

In this analysis, the ‘free’ kak by construction we see in (7) expresses a wish for the contrary (motivating the negation) of something undesirable, thereby inviting the inference (ultimately based on a pragmatic principle like relevance) that there is reason to fear that it might actually become reality. It is this inference that is put ‘on stage’ in another conventional construction of Russian, viz. the one we see in (5). The first component of this construction profiles a concept of apprehension, the second consists of the kak by construction, contributing its meaning to the entire complex construction; neither component is subordinated to the other.

Notice that the ‘negative wish’ indicated in the second clause is attributed to the subject of the first clause when this is a third person, as in Example (10):

In other words, the perspective of whoever is the one fearing something that is put ‘on stage’ in the first clause persists in the second clause expressing the situation being feared. This observation supports the hypothesis that the combination of two clauses instantiates a constructional pattern with the kak by-construction as its second component.

Once the underlying assumption that the relation between the expression of apprehension and the kak by clause is one of embedding is dropped, the ‘illogical’ status of the negation disappears. But more importantly in the present context, the analysis implies that the grammar of Russian, similar to the grammar of Dutch, provides tools for a range of connections between different perspectives on the same object of conceptualisation, the variety of which goes beyond what can be captured in terms of embedding. The specifics of the tools in the different languages differ: unlike Russian, Dutch does not have some standard way of putting a feeling of apprehension and a wish to avoid something undesirable side-by-side. But what is common is that each language provides its users with tools that clearly require a kind of flexibility that is hard, if not impossible, to provide when embedding is considered the crucial and stereotypical relation between viewpoints. However, such flexibility is naturally allowed for in our thoughtscape model.