Matthew C. Jones University Writing Program University of Florida Overt Unreliability and the Metarepresentational Frame On the Limits of Henry James’s Unreliable Narration José A. Álvarez-Amorós In this paper, I argue that Henry James’s unreliability in his short fiction shows a recurring peculiarity hitherto undiscussed or, at best, subsumed under the standard approach to this phenomenon. Even when his character-narrators report questionable information at odds with the authorial design as inferred by the reader, they seldom fail to trace such information to their own subjectivities and distinguish it explicitly from authenticated fictional fact. Relying on the metarepresentational capacity of real (and realistic) minds to process information inseparably from its source and aided by key rhetorical notions, I theorize this special kind of unreliability which I call overt, transparent, or selfacknowledged. Then I explore its different manifestations in “The Diary of a Man of Fifty” (1878), “The Aspern Papers” (1888), and “The Way It Came” (1898), three Jamesian tales whose narrators variously manage to keep track of their own minds as the source of their (often unwarranted) representations. On the resulting evidence I conclude that the limits of James’s unreliable narration are narrower and less disruptive than customarily held to be. One third into Henry James’s “The Aspern Papers” (1888), and just as the narrator is gathering data about the arresting oddities of the Bordereau household, he ponders on Miss Tita’s revelations and suspiciously notes AAA - Arbeiten aus Anglistik und Amerikanistik Band 46 (2021) · Heft 2 Gunter Narr Verlag Tübingen DOI 10.2357/ AAA-2021-0009 José A. Álvarez-Amorós 72 that her “story does not hang together” (CS 3: 264) 1 . This paper deals precisely with Jamesian stories that do not seem to hang together, and, more specifically, with the epistemic consequences of deploying narrators who openly admit that they are projecting what could be loosely called their subjectivity onto their reporting, and thereby representing their own mental states (belief, fear, desire, hope, etc.) as if they were the presumed baseline ontology of the fictional storyworld, that is, fictional reality 2 . In what follows, I present a possible approach to the dynamics of unreliability admission based on the metarepresentational ability - a cognitive endowment of the human mind that seems tailor-made for the purpose - and then I suggest how it can be used to probe the limits of James’s unreliable narration in his shorter fiction. 1. Initial assumptions That the names of Wayne C. Booth and Henry James tend to co-occur whenever one revisits the vexed issue of narrative unreliability is, by now, a critical commonplace. Booth identified and first formulated this phenomenon within a rhetorical context, while James employed it in his fiction and came close to theorizing it in rudimentary terms when he famously discussed his reflectors and centres of consciousness as endowed with the right balance of lucidity and bewilderment, of conscience and inconscience. This felicitous combination, he thought, would allow them to preserve the illusion of reality derived from the “general human exposure” (1984: 16) without rendering the story incomprehensible. And yet Jamesian unreliability should not be equated with internal, epistemically limited point of view, even if the line between both is not always sharply defined in the critical literature (e.g. Weinstein 1970: 208) 3 . It is rather a potential 1 References to James’s Complete Stories (1996-1999) will be given in the text as CS indicating volume and page 2 This paper is not on possible-world semantics. However, any reader of Marie-Laure Ryan (1991) and Lubomír Doležel (1998) will recognize approximations to their respective concepts of text actual world (TAW) (Ryan 1991: 24) and factual domain (Doležel 1998: 150) behind my multiple references to baseline fictional ontology, fictional reality, fictional fact(uality), etc. As explained in the text, my view is that the baseline fictional ontology is constructed by the implied author using the whole gamut of narrative resources at his or her disposal (e.g. the narrator’s voice). The information reported by the narrator is processed by readers as true when in agreement with that ontology and as false when opposed to it. 3 Narrative unreliability has been made contingent on a number of factors such as limited point of view, first-person narration, and ethical or psychological flaws in those characters who double as narrators. These kinds of associations have been jocularly called “package deals” by Yacobi (2001). They have empirical, statistical nature at most, she argues, and cannot be taken as universal conditions for unreliability to occur. In fact, she discusses several counter-examples in her paper (2001: 225-28). Here I will stick to Jamesian first-person unreliability to avoid the discussion of yet another On the Limits of Henry James’s Unreliable Narration 73 consequence of the radicalization of internal point of view and the concurrent want of authoritative commentary, plus the necessary development of discrepancies between what narrators actually say and what they honestly think they say. In James’s narrative practice, internal, uncorrected point of view insensibly slides into unreliability when pushed to the limit, thus suggesting that the former is a first step, an indispensable though by no means a sufficient condition for the latter. If discrepancies fail to occur, circumscribed point of view and unreliability remain distinct compositional resources within the narrative text. 1.1. The two faces of narrative unreliability Unreliability comes in two main varieties - rhetorical and cognitive. It first emerged as a rhetorical device from the pages of Booth’s The Rhetoric of Fiction in 1961, and was later revised and considerably enriched by James Phelan on equally rhetorical grounds (e.g. Phelan & Martin 1999, Phelan 2005, Phelan 2007, Phelan 2017). Nowadays, and despite countless tweaks and updates, it is still considered the standard approach. For Booth, an unreliable narrator is one who departs from “the norms of the work (which is to say, the implied author’s norms)” (1961/ 1983: 158), and produces a text that has a surplus of meaning for the reader which the narrator is hardly aware of. In consequence, the three main participants in the narrative communication are not perfectly aligned, the narrator being out of the loop and ironically displaced. What matters in this model is the intervening distance between the implied author and the unwary narrator as perceived by the reader. This formula is utterly reversed in the cognitive approach. For theorists such as Ansgar Nünning (1979, 2008) and Tamar Yacobi (2000, 2005), the implied author is a superfluous entity which Occam’s razor should take care of. In their view, when readers encounter disturbing incongruities between what the text denotes and their own internalized norms, cognitive schemata, and general foreknowledge (for instance, when the act of breaking into an old woman’s rooms to steal a bundle of letters of her property and almost shocking her to death in the process is simply reported as “this last indiscretion” or “these irregularities” [CS 3: 301, 304] by the narrator of “The Aspern Papers”), they tend to naturalize such incongruities by positing an unreliable narrator, “a mere integrative hermeneutic device” (Nünning 2008: 87), on which to hang the blame for the said lack of fit. Since the emphasis falls on the reader and his or her interaction with the text, the implied author becomes a dispensable figure, and the distance which counts now is that between the narrator and the reader. theoretical point. But James’s “The Liar” (1888), for instance, is ample proof that unreliability and third-person narratives are all but compatible. José A. Álvarez-Amorós 74 1.2. Unreliability and the metarepresentational frame Cutting across these two allegedly incompatible models of unreliability (Sternberg & Yacobi 2015: 335), a third one was proposed by Lisa Zunshine in 2006 4 . Unreliability, for her, is a function of the human - and humanlike - capacity to process situated information, that is to say, information closely linked to the circumstances in which it was obtained, whether place, time, or, especially, provenance. Such capacity to store in memory representations of physical or mental states of affairs along with the details of their acquisition is often called metarepresentational ability in the specialized literature on cognitive evolutionary psychology (e.g. Cosmides and Tooby 2000: 59-61, 69-71, 75, 79). Those in possession of it, whether flesh-and-blood individuals or their fictional replicas, enjoy a highly efficient protection mechanism which enables them to monitor the sources of contingent information, assess its truth potential according to the trustworthiness of such sources, and handle it with due precaution so that, if proved false, it will not damage wholesome information mentally stored or encourage unwarranted actions. Confronted with a statement such as “Drastic home lockdown will be enforced as from today”, any individual with normal metarepresentational abilities will most likely process it with an unmistakable source tag, and gauge its truth value differently if it comes from a government agency or from the next-door neighbour. Given the complexity and indirection of the fictional genre, the metarepresentational ability is vital for readers to make sense of narrative texts by tracking at all times who said what, who thought what, and, ultimately, who must answer for the truth value of a piece of information. Consider in this regard two contrasting cases. If emulating Don Quixote himself, and paying no heed to a whole set of explicit textual pointers, a reader of Cervantes’ novel proved unable to trace the existence of the giants to its proper source - the knight’s runaway fancy - he or she would perform a faulty reading of the episode and even risk the comprehension of the entire work. Responsibility for the misreading would lie in this case with the reader’s metarepresentational ability which could have malfunctioned for a variety of permanent or transitory reasons. On the contrary, if a standard metarepresentation such as “I thought she had turned a little pale” (CS 4: 616), to be found in James’s story “The Way It Came” (1898), composed by an unambiguous source tag (“I thought”) and an element of content (“she had turned a little pale”), got speciously reported in the text as “She had turned a little pale” tout court, all sorts of uncanny things would start to happen, as, for instance, that a state of mind - the speaker’s thought - might pass for a fictional fact, thus interfering with the reader’s capacity to ascribe truth values to different kinds of information and reconstruct a 4 See Marsh (2018: especially 1336-38) for a recent application of the metarepresentational frame to the analysis of standard unreliability. On the Limits of Henry James’s Unreliable Narration 75 consistent account of the storyworld. Misreading here would be induced by textual design. Instead of this partially fabricated example, consider now the similar case of Boffer Bings, the much-instanced unreliable narrator of Ambrose Bierce’s “Oil of Dog” (1890). He describes his mother, a murderess of unwanted children, and his father, a manufacturer of dog oil whose raw material are stolen dogs, as “honest parents” (1979: 800), and the nearby river into which he is instructed to throw “the débris of her work” (1979: 800) as “thoughtfully provided [by nature] for the purpose” (1979: 801). The clash of Boffer’s account with what readers infer the implied Bierce would actually endorse is so obvious - this tale being a fairly bold case of unreliability - that they can hardly accept a baseline ontology constructed on such outrageous principles. Rather, they tend to attribute every reported aberration to the problematic subjectivity of a teller who has lost track of himself as the origin of his own representations and has consequently omitted from the narrative text all verbal clues pointing to that origin. (Compare “I was born of honest parents” [1979: 800] with, for example, “I was born of honest parents, or so they seemed to me at that time”.) To all practical effects, Boffer perceives his states of mind as fact and, what is ethically worse, acts on this perception. Authoritative narrators are generally cooperative and, when appropriate, tag information correctly; what we call unreliable narrators, however, make more or less sporadic mistakes in that respect. They often ignore they are reporting their beliefs, fears, or desires in lieu of established facts, and hence they misleadingly drop essential source indicators. This tampers with the phenomenological status of unreliability, since the improper - and deceptive - omission of source tags turns metarepresentations into elliptical metarepresentations (Recanati 2000: 74-75), but not into actual representations. A conception of (un)reliability based on degrees of success or failure of the narrator’s metarepresentational skills forms, to my mind, an ideal theoretical template to discuss the trustworthiness of those narrators, like James’s, who often substitute their mental states for fictional facts, but can still keep track of their responsibility in the substitution by expressing qualms about the provenance and truth value of the information they relay. Carried out either by readers in the process of extracting meaning from a fictional text or by narrators as they come to grips with their material, source tagging is not a binary procedure with static, irreversible results. Information can be stored with varying degrees of precaution, and these can alter as the narrative develops. Take a piece of intelligence such as the existence of the Aspern papers in James’s namesake tale. Most readers and critics will store it in memory with no source tag, that is, as an absolute certainty which can unrestrictedly condition the comprehension of the text. Other readers, however, in view of how peculiarly James introduced and sustained this issue in the tale, will opt for processing it with a weak, hesitant tag pointing to the narrator’s subjectivity; they will still use it as a José A. Álvarez-Amorós 76 basis to infer the meaning of the tale, but will keep it cautiously decoupled from more factual facts such as the existence of the Bordereau women or of Venice itself. Finally, a small number of readers will assimilate it with a screaming tag pointing to the narrator’s fancy and will surround it with an impregnable cordon sanitaire to prevent it from insidiously determining the interpretation of the tale. Jacob Korg, for one, produced in 1962 a reading of “The Aspern Papers” in which the papers themselves only exist in the narrator’s mind, and other critics have often expressed similar doubts (Falconer 1987: 1, Rivkin 1989: 136, Snyder 2004: 135). The intensity of these tags, moreover, is not permanent. It can fluctuate with the informational progression of a narrative and form its metarepresentational dynamics. The default attitude is to accept the narrator’s version and process it without source tags or very weak and general ones at most. But inconsistencies between this version and the norms presumably sponsored by the implied author, or simply the emergence of new, relevant information, will alert our metarepresentational instinct and make us retroactively revise the allocation and intensity of tags and often the global meaning of a narrative, especially when unreliable narrators tend to be introduced as fairly reliable ones whose fallibility only transpires as the story moves forward. And yet this constant revision of the metarepresentational frame of information, integral to any reading of (fictional) narrative, may only achieve partial success. There will often be vital pieces of information recalcitrant to stable, rational tagging, such as the presence of the supernatural in “The Way It Came” or “The Turn of the Screw” (1898), but not, for instance, in “Sir Edmund Orme” (1891). This may result in strong metarepresentational ambiguity, and, contingent on the reader’s disposition, in large, rewarding doses of aporetical pleasure. 2. Jamesian overt unreliability and the dissociation of the teller and the told Framed by the foregoing assumptions, my position in this paper is that James’s narrators mostly tag the information they pass on in correct and cooperative fashion. On the whole, they manage to keep track of themselves as the sources of their representations, and are often aware that what they report may be traced back to their own subjectivities, whether it agrees or not - speaking in rhetorical terms - with what the implied James would say if he had a voice like that of a narrator. This means that his tellers often incur a more attenuated, reader-friendly kind of unreliability, or, in other words, that the gap between what they report and what they think they report is narrower and more easily bridged than in less conscientious narrators. To put the case in these terms, however, has significant consequences as it leads to the dissociation of the teller and the told in the specific area On the Limits of Henry James’s Unreliable Narration 77 of narrative unreliability. This results in a hitherto unexplored duality between reliable/ unreliable reporter and true/ false information whose members do not implicate each other by necessity and can be decoupled and independently discussed 5 . In fact, a simple typology could be built by combining the reporter and information options separately placed on two axes. Reporters diversify according to their trustworthiness, which is, after all, a pragmatic notion, and, for the purposes of this paper, a function of their capacity to sense and verbalize the difference between imaginary states of affairs and the domain of fictional fact by attaching appropriate tags to the former, i.e. by metarepresenting them. In other words, I hold that the tendency to show open concern for the origin of reported content is a basic symptom of reliability. Information, for its part, is more properly depicted as true or false - rather than as reliable or unreliable - which, in a context of rhetorical theory, primarily means aligned or misaligned with the implied author’s project, and so either true or false within the world modelled on that project. In consequence, an unreliable narrator is one who believes to be telling the truth even if he or she is not, and acts in the casual, matter-of-fact way of someone who has no qualms about the content of his or her utterances. A false piece of information, however, is one which conflicts with the implied author’s norms as apprehended by the reader. Boffer Bings communicates as fact his conviction that nature has thoughtfully placed a river close to his mother’s studio to facilitate his task; but readers rightly infer that the proposition “Nature thoughtfully provided a river to facilitate Boffer’s task” is false with reference to the implied Bierce’s ethics and general worldview. Conversely, Jamesian narrators of the kind under discussion here can be reliable and cooperative, can warn readers that they may be voicing mental states rather than fictional facts by metarepresenting them correctly, and supplying textual clues to this effect particularly when the information provided is hardly endorsed by the implied author. Put another way, that a piece of information is appropriately tagged does not mean it is true; it rather means that the narrator is reliable or, at least, that he or she has taken significant steps towards reliability. 3. Information and source tagging - a tentative typology Though its value is purely heuristic and it has no pretensions to any systematic treatment of narrative (un)reliability, the typology suggested 5 Köppe and Kindt provide the only glancing reference to this issue I have been able to find: “What is more, the relationship between the unreliability of an informant and the unreliability of his information is rather complex” (2011: 87). José A. Álvarez-Amorós 78 above may still have some illustrative potential 6 . It is based (a) on two kinds of information, and (b) on the attachment of source tags to each in order to indicate their factuality within the narrator’s mind. Information can be true or false, that is, it can concur or not with the implied author’s design for a story as inferred by the reader; and it can be processed and reported by narrators either as source-free, unconditional fact, or as sourcebound content explicitly contingent on their subjectivities. Ideally, four distinct types result from this intersection, but there are subtle branchings and nuances that call for brief commentary. 3.1. Cases of authorially endorsed information In any narrative, there is a large amount of external, presumably objective information which does not trigger any metarepresentational unrest and is assimilated as established truth in need of no precautionary tag. Such is the case with the existence of Mrs. Prest or the Venetian statue of Bartolommeo Colleone in “The Aspern Papers”, Corvick’s death in “The Figure in the Carpet” (1896), the weekend at Newmarch in The Sacred Fount (1901), and so on. Few readers would wonder at the absence of protective source tags qualifying the certainty of these elements within their respective storyworlds. But in James, as in any other novelist of consciousness, there is also a substantial volume of information about mind which does not challenge the norms of the work and also enjoys factual status. While conversing with Miss Tita, the acquisitive narrator of “The Aspern Papers” describes himself as “not having her rectitude” and as “the reporter of a newspaper who forces his way into a house of mourning” (CS 3: 280). These propositions express self-incriminating content which both the implied James and his readers would subscribe to in view of what happens in the story. We sense the teller is correctly diagnosing his ethical deficit. We read them as wholly factual, and so we do not process them as conveying source-contingent information. Pressing the analogy a little, one could even say that the city of Venice and the narrator’s lack of rectitude have the same ontological status within the fictional universe of “The Aspern Papers”. And yet nothing prevents a narrator from reporting as explicitly sourcebound content a piece of information which, true in his or her mind, is also signalled as true by the implied James. Two instances can illustrate this type. “Then I guessed”, says the narrator, “that nocturnal prowlings were not in the least her [Miss Tita’s] habit” (CS 3: 261). Given her meek character and the overwhelming control Juliana Bordereau exercises over the 6 Evidence of the limitations of this typology is that it employs binary categories when a continuum would reflect the nature of unreliability in more precise terms. For example, appropriate tagging as a criterion for a speaker’s reliability is not a binary phenomenon, but rather admits of degrees and intensities as previously argued. On the Limits of Henry James’s Unreliable Narration 79 whole range of her life, the narrator’s comment feels wholly factual (“Nocturnal prowlings were not in the least her habit” would be a perfectly equivalent alternative in information-processing terms). But he chooses to bind the content to his subjectivity - i.e. to metarepresent it - and this is verbally indicated by the source tag “Then I guessed”. Similarly, after the narrator of “The Figure in the Carpet” has vented his irritation at the pleasure Corvick and Gwendolen find in pursuing Vereker’s secret, he traces back his annoyance to his own mind, admits that his words look “unamiable”, and confesses that “what probably happened was that [he] felt humiliated at seeing other persons derive a daily joy from an experiment that brought [him] only chagrin” (CS 4: 587). Here again the narrator’s mental reconstruction of his own motives seems to agree with the authorial plan for the story. In all these cases, and regardless of the tagging strategy implemented in the text, James’s tellers handle information we deem well aligned with the implied author’s norms and thus endowed with factual status. By a wide consensus, the types of reporting just noted, whether tagged or not, are held to be reliable. 3.2. Cases of authorially disowned information: standard vs. overt unreliability When this default situation is reversed, it yields two further possibilities based on the telling of information which appears to be incompatible with the general drift of the work. Let us consider the following passages from “The Aspern Papers”. They relate to the situation created by Miss Tita’s alleged offer of her hand to the narrator as a condition for him to obtain the coveted documents: My gondolier, behind me, must have seen my ears red as I wondered, sitting there under the fluttering tenda, with my hidden face, noticing nothing as we passed - wondered whether her delusion, her infatuation had been my own reckless work. Did she think I had made love to her, even to get the papers? I had not, I had not; I repeated that over to myself for an hour, for two hours, till I was wearied if not convinced. (CS 3: 315, my emphasis) But I had not given her cause - distinctly I had not. I had said to Mrs. Prest that I would make love to her; but it had been a joke without consequences and I had never said it to Tita Bordereau. (CS 3: 316) Both fragments belong to the same episode and both deal with the narrator’s befuddled reaction to Miss Tita’s offer. But they are not, I think, identical takes on the state of affairs in hand. They report questionable, if not downright false, information to the effect that the narrator had never made love to Miss Tita, a proposition amply refuted by his general behaviour which, by authorial design, accords fairly well with conventional courtship José A. Álvarez-Amorós 80 (e.g. sending her flowers every day, taking her out on his gondola, treating her to ice-creams at Florian’s in Piazza San Marco, dazzling her with his conversation, allusions to Romeo and Juliet, and what not). But while the second fragment presents this unwarranted information as fictional fact, without the least awareness on the teller’s part that he may be reporting a case of wishful thinking, the first one explicitly shows that he has not lost track of himself as the source of his representations and excruciatingly doubts if his alleged restraint counts as true in the baseline ontology of the storyworld or is just a comforting, unrealized desire. These two quotations turn on the narrator’s ability to perceive and express the difference between the world of fictional fact and that of imagination. When such ability fails for a number of reasons, we have standard cases of unreliable telling. Take, for instance, the “scruple[s]” the narrator of “The Aspern Papers” ostensibly attributes to himself four times in the story (CS 3: 251, 266, 290, 305), or his statement that it “would be brutal” to take Frech leave of Miss Tita, as his “idea was still to exclude brutal solutions” (CS 3: 317). In view of his grossly unethical conduct, which includes raiding Juliana’s rooms, both his “scruple[s]” and his professed intention to avoid “brutal solutions” clearly infringe the norms of the work and can hardly exist anywhere but in his mind. Similarly, the narrators of “The Figure in the Carpet” (1897) and “The Way it Came” make trenchant statements that seem out of step with the signals the reader receives from the implied James. One boasts his capacity for fine discernment - as in “waves that promised, I could perfectly judge, to break in the end with the fury of my own highest tides” (CS 4: 608, my emphasis) - which contradicts his customary obtuseness and is reported as deceptively factual; the other makes the unqualified claim that the effective occurrence of the contentious post-mortem interview featured in the tale “was simply a question of evidence” (CS 4: 626), which, of course, it wasn’t. This type of standard unreliability based on instances of covert onelevel mental embedment whereby a state of mind (“[I thought, felt, imagined, etc.] I had scruples about going back”) is reported as fact (“I had scruples about going back” [CS 3: 305]) coexists with more complex cases of covert two-level mental embedment formed by reports of states of mind which are actually reports of states of mind about further states of mind. This usually happens when a narrator forces minds on other characters to fit his or her own aims and preconceptions, and yet makes them pass for authoritative, factual descriptions by removing the primary tag which points to his or her subjectivity. In “A Light Man” (1869), for instance, the narrator says, “Theodore likes him [Mr. Sloane] - or rather wants to like him; but he can’t reconcile it to his self-respect […] to like a fool” (CS 1: 416). That the narrator’s friend Theodore likes Mr. Sloane is perfectly in line with the progression of plot and character in this tale, but the second part of the sentence - as from “or rather wants to like him” - is a groundless statement that betrays the narrator’s frame of mind rather than his friend’s; On the Limits of Henry James’s Unreliable Narration 81 so restoring the missing tag (e.g. “I think”, “in my view”, etc.) seems consistent with the overall thrust of the story. Planting groundless thoughts in other characters’ minds without due narratorial acknowledgement is an occasional source of local pockets of unreliability in Jamesian narrative practice from his earliest experiments to his mature handling of the technique in the late 1880s and 1890s 7 . Standard unreliability, however, gives way to a more transparent variety illustrated by the first passage about Miss Tita’s offer quoted above. The teller’s incapacity to metarepresent his own mental states correctly is replaced here by some form of explicit awareness that what he communicates is an imaginary construction rather than factual information. Flaunting qualms about the origin of a piece of intelligence, and hence about its factual status, does not seem to promote unreliability; instead, it contributes to its deactivation, or, at least, to controlling its disruptive potential. That Holden Caulfield often admits to lying does not make him more unreliable. It rather allows the reader to process his narrative with extra metarepresentational precautions. It would be much more damaging if he lied and had no inkling of it. Likewise, the fact that the narrator of James’s “The Figure in the Carpet” confesses that he “was shut up in [his] obsession forever” (CS 4: 604) does not render him unreliable, as Kellum suggests (1976: 106). At any rate, what could provoke unreliability is his mild mental disorder, not its open recognition, especially when such recognition involves an accurate feat of introspection and closely corresponds with the signals readers get from the implied author. In sum, when the governess of “The Turn of the Screw” (1898) voices her heart-rending self-doubts - “for if [Miles] were innocent what then on earth was I? ” (CS 4: 739) - she is partially moving towards reliability by showing she is not wholly oblivious that vital representations of fictional reality may originate in her mind, at least in one reading of the novella. Overt, self-acknowledged unreliability has, in my view, bonding effects quite similar to those described by Phelan in his 2007 analysis of the ethics of Nabokov’s Lolita (1955), and first suggested by Booth in relation to Tristram Shandy (1759-67) (1961/ 1983: 230-32, 240). In rhetorical terms, standard unreliability results from the growing distance between the narrator and the alignment formed by the implied author and the implied reader. Paradoxically, however, some of the perceived discrepancies that 7 For additional illustration see “The Diary of a Man of Fifty” (1879): “he wondered who I was and what I wanted, and he did me the honour to perceive that, as regards these points, my appearance was reassuring” (CS 2: 457); “The Aspern Papers”: “At this a singular look came into the face of Miss Bordereau’s niece - a kind of confession of helplessness, an appeal to me to deal fairly, generously with her” (CS 3: 279); “The Figure in the Carpet”: “Of course, he knew; otherwise he wouldn’t return my stare so queerly” (CS 4: 601); “The Way It Came”: “He gave a loud sound of derision; but it was not a genuine one” (CS 4: 633) (my emphasis in all cases.) José A. Álvarez-Amorós 82 make narrators unreliable also invite readers to bond with them, generating six subtypes of bonding unreliability identified and discussed by Phelan in his study (2007: 225-32). It seems to me that the cognitive effort of keeping track of oneself as the source of one’s own representations even in distressing circumstances tends to enlist the reader’s sympathetic trust in much the same way. In 1981 Brian McHale asserted that the “history of our poetics of prose is essentially a history of successive differentiations of types of discourse from the undifferentiated ‘block’ of narrative prose” (1981: 185), a dictum that has been approvingly echoed by later narratologists (e.g. Palmer 2004: 7). In this section I have attempted to implement McHale’s insight in the specific area of (un)reliable narrative discourse. By “successive differentiations” based on criteria such as the types of information and the narratorial source-tagging activity, I have arrived at the kind of unreliability that forms the speculative core of this paper, that is, overt unreliability as the blend of questionable information and the narrator’s more or less keen awareness of its counterfactuality. The next step is, of course, to investigate how James deployed this variety of unreliable discourse in his short fiction. 4. Varieties and emphases of overt unreliability in James’s tales It is just obvious that James did not invent unreliable narration. Although we still lack a detailed account of the emergence of this narrative resource in Western literature, it seems to have been deployed, with ups and downs, since modern realistic fiction took centre stage in the late eighteenth century (Zerweck 2001: 159). And yet James’s oeuvre, which is itself a microcosm of the transition between nineteenth-century realism and the experimental writing leading to modernist narrative, is also indicative of the evolution of unreliability at a strategic point in the history of the genre. James’s unreliable narration follows an arguable trajectory from blunt obviousness to subtle complication. His earliest attempts in “A Light Man” or “The Diary of a Man of Fifty” (1879), where the double vision at the root of this technique becomes a conspicuous, so-what cliché, were gradually superseded by the elusive uncertainties of “The Figure in the Carpet”, “The Way It Came”, or “The Turn of the Screw”. This trajectory, however, is not smoothly homogeneous, but rather comprised of two segments which join at about 1888, the year of publication of “The Aspern Papers” and “The Liar”, its heterodiegetic counterpart. Unreliability in the first of these segments is what Booth would call a stable ironic form “in the sense that once a reconstruction of meaning has been made, the reader is not then invited to undermine it with further demolitions and reconstructions” (1975: 6). Despite the progressive refinement that typifies the passage from “The Diary of a Man of Fifty” to “The Aspern Papers”, the ironic treatment of their narrators, once properly detected, supports a single, abiding interpretation. On the Limits of Henry James’s Unreliable Narration 83 In sharp contrast, “The Turn of the Screw” epitomizes the second segment. As is well known, it encodes an ironic view of the governess-narrator in unresolved coexistence with a straight reading of her plight. In metarepresentational terms, those unreliable stories based on stable irony contain a number of propositions denoting misaligned information whose source tags can be rationally, and perhaps even consensually, restored by readers. For this to happen, however, a consistent authorial design must show behind the narrator’s utterances so that elliptical metarepresentations can be identified and processed. When the signals sent by the implied author to his audience do not cohere - e.g. ghosts are both epistemic and ontological - then interpretation - e.g. ghosts are either epistemic or ontological - must inevitably precede any attempt at rationally restoring source tags. But interpretation entails choice, and choice has proved an almost impossible task in some of James’s late nineteenth-century short fictions. Understood as the reporting of a mental state plus some form of explicit awareness that it is a mental state, overt unreliability is not limited to a specific phase of James’s storytelling. Tales from three different decades display phenomena that can be viewed as aspects and modulations of this type of unreliability - as well as so many keys to its nature and functionality. Self-doubt in “The Diary of a Man of Fifty” erupts in a kind of coda that prompts readers to append source tags retroactively to pieces of intelligence such as the alleged factual analogy between the narrator and his young confrère, on the one hand, and between the Countess Salvi and her daughter, on the other. In “The Aspern Papers”, for its part, the narrator parades his own subjectivity as the origin of his representations at least in three ways: first, by standard propositional tagging (e.g. “It seemed to me” [CS 3: 253], “I sat spinning theories” [CS 3: 257]); second, by textual resources that go beyond the sentential boundary; third, and very especially, by a deluge of defactualizers and conjectural pointers - “as if” recurring no fewer than 89 times - which operate as impersonal, condensed metarepresentations referring intelligence to an unspecified source and hence preventing its confusion with hard fictional fact. Finally, “The Way It Came”, a notable instance of James’s conflicted unreliability, merits attention for the frequent intertwining in it of the telling and the experiencing selves, that is, of the narrator and the character as the two existentially linked components of a first-person narrative situation. Considering the visibility of both, a set of important questions beg themselves - who introduces the awareness that a state of mind might be just a state of mind? is it the narrator, the character, both, none? who progresses towards reliability by properly distinguishing fact from fancy? how does the (dis)agreement between these two figures generate the ambivalence of this story? Any attempt to answer these questions, as well as others raised by “The Diary of a Man of Fifty” and “The Aspern Papers”, calls for a closer look at these three stories. José A. Álvarez-Amorós 84 “The Diary of a Man of Fifty” is a tale of warped remembrance. It consists of eighteen diary entries of dissimilar length wherein the diarist-narrator, an aging English soldier, relates his return to Florence twenty-seven years later, revisits his past in the light of his present experience, and strives to prevent young Edmund Stanmer from committing “an act of thumping folly” (CS 2: 454) which he claims to have avoided, at great emotional expense, when he first came to the Tuscan city. On December 12, 1878, James entered in his notebooks a synopsis for this narrative which closely informs the narrator’s version of the events up to its denouement. At this point, synopsis and tale diverge fundamentally. James’s notes conclude with the narrator’s triumph - convinced that Stanmer is going to make a life-long mistake by marrying a countess’s daughter, whose mother he himself loved but renounced in self-protection, he “determines to warn him and [Stanmer] opens his eyes” (James 1987: 9). The tale, however, closes with three entries in which we know of Stanmer’s eventual marriage, of his immense happiness three years later, and, finally, of the narrator’s painful doubts as to whether his actual mistake might not have been to forsake an extraordinary woman for unfounded reasons. As often with James, a set of unconditional, source-free facts in the notebooks become contingent in the tale on a specific subjectivity - in particular that of the first-person narrator - and are thus destabilized and rendered problematic. This is especially so, moreover, when the implied author curtails the narrator’s capacity to discriminate between fictional fact and imaginary construction, and he or she is reduced to reporting the latter as if it were the former creating, as we know, the standard conditions for narrative unreliability. The central piece of information on which “Diary” hinges as an unreliable narrative act is a presumed analogy. This likens it to a much later Jamesian novella, The Sacred Fount, whose basic - and permanent - gap is the truth value of an analogy established by the narrator between what he deems two parallel vampiric processes obtaining between two pairs of characters. In “Diary” the analogy also involves four characters and two temporal planes. Twenty-seven years ago, the narrator had a romance with now deceased Countess Salvi which he allegedly discontinued lest he should be exposed to disastrous consequences given her inconstant, coquettish nature. Now, as he revisits Florence, he meets Stanmer who is in love with Countess Salvi’s daughter. Persuaded that the youth is a replica of his earlier self and that the ethical flaws he attributed to Countess Salvi are also her daughter’s, he schemes hard to draw him away from her. In James’s notebook entry the analogy is presented as fact - he witnesses “a certain situation of his own youth reproduced before his eyes”, “[t]hat episode of his youth comes back to him with peculiar vividness”, “the daughter is a strange, interesting reproduction of the mother”, the “mother had been a dangerous woman […] an unscrupulous charmer”, “her daughter […] strongly resembles her,” “[s]he is a beautiful dangerous coquette”, On the Limits of Henry James’s Unreliable Narration 85 he sees in the young man “the image of his own early innocence”, and so on (James 1987: 8-9). But much as the notebooks support the factual nature of the analogy, the implied James signals against it in the tale. Artistic execution seems to have dissolved its factuality and turned it into a mental state, a mere imaginary construction on the narrator’s part. The issue, of course, is to determine if he is aware of reporting information sourced from his own mind, that is, if his metarepresentational ability is correctly in place or has been disturbed by a sense of loss and regret, possibly compounded by misogyny, jealousy, or, in Bradley’s view (2000: 69-74), by a latent homosexual feeling for Stanmer concealed under an ostensibly heterosexual plot. For most of the tale the narrator remains unconscious that the analogy is a product of his mind. He often wields it as a magic wand to argue his point with Stanmer and Countess Salvi’s daughter, as when the young man protests that she never lied, and the narrator, instead of offering evidence to the contrary, only exclaims, “That’s just what I would have said to any one who should have made the insinuation” (CS 2: 477; see also CS 2: 460, 464, 466). Initially unwary, the reader soon feels drawn into processing the narrator’s countless analogic assertions with growing caution - first, because the analogy obsessively recurs to the point of becoming his only mode of argument; second, because his ascription of mental states to his fictional peers is often conditioned by its terms and speciously directed to strengthen it; third, because it is challenged from within the storyworld by Stanmer himself when he charges the narrator with “‘overdo[ing] the analogy’” (CS 2: 469) and highlights the disparity between his conduct and his words (CS 2: 475), and by Countess Salvi’s daughter when she denies any resemblance between the narrator and his young protégé (“‘And yet you don’t look at all like him! ’” [CS 2: 473]); and fourth, because the narrator himself, as he slowly progresses to his final anagnorisis, admits to shortcomings (e.g. confusion [CS 2: 460] and inconsistency [CS 2: 474]) which, in themselves, do not constitute unreliability, but may stand in a causal relation to it. Of special interest is the ad hoc construction of Stanmer’s mind to satisfy the preconditions of the analogy. “‘You are in love with her, and yet you can’t make her out’”, states the narrator, and promptly adds, “‘that’s just where I was with regard to Madame de Salvi” (CS 2: 468). This can be read as two mutually supporting assertions - “you can’t make her out” and “that’s just where I was” - each based on nothing except on the narrator’s commitment to his fixed idea. The same pattern holds when the narrator, after telling Stanmer that the two women “‘are mother and daughter - they are as like as two of Andrea’s Madonnas”, constructs a mind for the young man ad libitum just to make it match his own, “‘Your state of mind brings back my own so completely … You admire her, you adore her, and yet, secretly, you mistrust her’” (CS 2: 469; my emphasis). The narrator’s early “[h]appiness mitigated by impertinent conjectures” (CS 2: 454) undergoes a sea change in the closing part of “Diary”. Stanmer’s José A. Álvarez-Amorós 86 blissful marriage and the sedate admonition “Depend upon it, you were wrong! … Was it not rather a mistake? ” (CS 2: 484) he addresses to the narrator deactivates the last remnants of the analogy for the reader and undermines the former’s conviction about its factuality. Dithering source tags appended to the analogy and signalling its origin in the narrator’s mind are retroactively validated, and his insistent doubting reveals a new awareness of his epistemic role in the representation of the Salvi affair. But we can still delve somewhat deeper. The analogy is twofold and equates the narrator to Stanmer and the countess to her daughter. The first equation proves to be false - Stanmer is not a reflection of the narrator’s younger self. For one, he follows his instinct in the face of overwhelming pressure and marries his lover, whereas the narrator’s circumspection led him to back out to protect his mental welfare. The second equation, however, remains undecidable, and it is precisely this informational gap which provokes the narrator’s uncertainties and our cognitive reprocessing of the whole tale. If the equation holds, he misrepresented the mother and made the mistake of his life. This is as far as he goes in his brooding. But we can still go further - if it is false, he might have been right or wrong in his past decision, but all his analogic reasoning since he met Stanmer, his rather bullying advice to him, and his mildly offensive attitude to his lover turn out to be baseless and severely misplaced. Either way, however, and at the eleventh hour, the narrator manages to show some evidence that he can keep track of himself as the source of his representations, and the implied James wholly ensures that his true role in this petty Florentine drama is not lost upon the reader. What I have been calling overt, transparent, or self-acknowledged unreliable narration can be observed under a different lens in “The Aspern Papers”, often held as the apex of James’s unreliability as a stable ironic form. The operation of the metarepresentational skill in fictional characters, whether narrators or not, leaves behind verbal traces which I have termed (source) tags following Zunshine’s lead in her book Why We Read Fiction (2006). Consider these four texts taken from “The Aspern Papers”: […] though her [Miss Tita’s] face was in deep shadow […] I thought I saw her smile ingenuously. (CS 3: 297; my emphasis) I was surprised at […] her [Miss Bordereau] having the energy, in her state of health and at her time of life, to wish to sport with me that way simply for her private entertainment - the humor to test me and practice on me. This, at least, was the interpretation that I put upon her production of the portrait […] (CS 3: 288; my emphasis) She became silent, as if she were thinking with a secret sadness of opportunities, forever lost […] (CS 3: 278; my emphasis) I did not know what to do, as I say, but at a venture I made a wild, vague movement in consequence of which I found myself at the door. (CS 3: 315) On the Limits of Henry James’s Unreliable Narration 87 The first three are successful metarepresentational operations effected by the narrative agent in the sense that content in them is explicitly processed as coming from his subjetivity, and should not be confused by readers with the baseline ontology of the storyworld. The last one, by contrast, is an elliptical metarepresentation and makes a fine instance of standard unreliability for the two usual reasons - the information provided is discredited by the authorial design, and yet no tag indicates the narrator’s awareness that what he says may only be true in his mind. This passage occurs just after Miss Tita has offered herself in marriage to the narrator in exchange for the Aspern papers. When he realizes the full meaning of her proposal, he feels a kind of trap has been laid for him and only thinks of escaping. So finding himself “at the door” can hardly be the result of “not know[ing] what to do”, of “a wild, vague movement” made “at a venture”. To readers, this movement seems more deliberate and consistent with his urge to escape than vague or random. It can only be the latter in his mind as he tries to shake off the responsibility for his indelicate treatment of Miss Tita. There is, however, no verbal clue to this effect. (Compare with “but at a venture I made what then it seemed to me a wild, vague movement”.) Each of the three first quotations above illustrates one way in which language can describe metarepresentational operations, and especially how the linkage between content and source is realized. We have, on the one hand, standard propositional subordination whereby a state of affairs (“I saw her smile ingenuously”) is embedded under a metacognitive verb (“I thought”) which conveys the teller’s attitude towards that state of affairs. Equivalent syntactic forms are parenthetical phrases pointing to the source of a piece of information, such as “Miss Tita met them [the narrator’s eyes] quickly and read, I think, what was in them” (CS 3: 294). But the semantics of the metacognitive verb governing the embedded state of affairs is of fundamental import here (Zunshine 2006: 110). Contrast these metarepresentational statements: I guessed that her aunt had instructed her [Miss Tita] to adopt this tone […] (CS 3: 248; my emphasis) I believed for the instant that she [Miss Bordereau] had put it on expressly […] (CS 3: 241; my emphasis) I knew she would come out; she would very soon discover I was there. (CS 3: 308; my emphasis) But poor Miss Tita would have enjoyed one of Florian’s ices, I was sure. (CS 3: 260; my emphasis) […] I was sure she [Miss Tita] was speaking the truth. (CS 3: 279; my emphasis) […] though Miss Tita gave no high-pitched voice to her satisfaction I felt that she surrendered herself. (CS 3: 276; my emphasis) José A. Álvarez-Amorós 88 […] though my friend [Miss Tita] gave no high-pitched voice to her glee I was sure of her full surrender. (James 1908: 76; my emphasis) Although the embedded content is made contingent on the narrator’s subjectivity in all cases, the intrinsic attitudinal meaning of guess, believe, and feel is heavily at variance with that of know and be sure. Given their marked subjective profiles verging on self-confessed uncertainty, guess, believe, and feel make for glaring source tags whose removal would be quite disruptive; on the contrary, know and be sure originate weak, almost transparent source tags that casual reading can more easily pass over as if they introduced factual information in need of no guarded processing by the reader. That these semantic nuances are highly instrumental in the construction of overall meaning is evidenced by James’s own practice. The last two quotations above are respectively taken from the original 1888 publication of “The Aspern Papers” and from its revised version included in volume 12 of the New York Edition of his novels and tales (1908). In line with the general tendency to stress the narrator’s shameful treatment of the Bordereau women, James played his lexical cards to bring Miss Tita’s surrender as close as possible to a fictional certainty without fully erasing its metarepresentational nature. This is attested to by the existence of a third possibility (“[…] though Miss Tita gave no high-pitched voice to her satisfaction, she surrendered herself”) which James chose to discard. Going beyond the bounds of the sentence, we enter the realm of textual interpretation proper. Here source ascription is a distributed phenomenon. It can happen over a whole stretch of discourse as more or less substantial pointers accrue and make the overall effect that a piece of intelligence is contingent on the narrator’s mind rather than absolute fact. At the end of “The Aspern Papers”, there is a passage whose essentials I give below: As soon as I came into the room […] I also saw something which had not been in my forecast. Poor Miss Tita’s sense of her failure had produced an extraordinary alteration in her […] Now I perceived it […] She stood in the middle of the room with a face of mildness bent upon me, and her look of forgiveness, of absolution, made her angelic. It beautified her; she was younger; she was not a ridiculous old woman. This optical trick gave her a sort of phantasmagoric brightness, and while I was still the victim of it I heard a whisper somewhere in the depths of my conscience […] (CS 3: 319; my emphasis) This passage describes Miss Tita’s “transfiguration” into a younger, beautified woman as the narrator’s wishful precondition “to pay the price” (CS 3: 319) of marriage in order to seize the Aspern papers. The information supplied is clearly at odds with fictional reality as endorsed by the implied author and inferred by the reader, for she is instantly “changed back into a plain, dingy, elderly person” (CS 3: 320). Even so, the narrator never On the Limits of Henry James’s Unreliable Narration 89 loses track of himself as the source of his representation of Miss Tita’s metamorphosis; rather, he drops a number of clues - “saw”, “perceived”, “optical trick”, “victim of it” - that collectively ascribe the whole experience to his obsessed frame of mind in a fairly standard case of overt unreliable narration. Though instances of these two tagging methods abound in this tale, they are massively outnumbered by the occurrence of adverbs and adverbial phrases with strong metarepresentational roles, namely, (a) to identify the subjective provenance of a piece of information, and (b) to decouple it from the baseline ontology of the storyworld. Both roles can be performed by adverbials such as “from my point of view” (CS 3: 280) and “in my mind” (CS 3: 302). But what catches the eye is the striking recurrence of tiny metarepresentational systems encapsulated in such terms as apparently, evidently, sort of, perhaps, and, very especially, as if. When associated with the representation of a state of affairs, they tend to fulfil role (b) as stated above, but not role (a), and, for this reason, we might call them defactualizers. The statements “Inga likes Otto”, “[It appears to me that] Inga likes Otto”, “It appears to me that Inga likes Otto”, and “Apparently, Inga likes Otto” respectively denote a representation, a spurious representation, a metarepresentation with an explicit source, and a anomalous metarepresentation whose source is not declared, but whose content could hardly be mistaken for unconditional fact. Most defactualizers are employed in “The Aspern Papers” to relativize mental attributions and show that they come from another mind which remains impersonal even if, with few exceptions, it can be transparently ascribed to the narrative agent. They tend to neutralize the unreliability that may emerge when a primary tag silently vanishes as a signal that the speaker’s metarepresentational ability has been somehow compromised (e.g. in the second Inga-Otto example above). Take, for instance, the following case, “[…] said Miss Tita, as if she had become conscious that her own question might have looked overreaching” (CS 3: 245; my emphasis), which could be unpacked as “It seemed [to the narrator] that Miss Tita had become conscious that somebody might have thought her question overreaching”. This makes a three-level recursive metarepresentation if the adverbial as if amounts, as I suggest, to an impersonal source tag denying factuality to Miss Tita’s mental processes and turning them into (narratorial) inferences. The pattern I have described is general all through “The Aspern Papers”. As if is constantly used to defactualize mental representations of living characters (Miss Tita, Miss Bordereau, Mrs. Prest, the doctor), but also of deceased ones (Jeffrey Aspern himself), and so are the other adverbials listed above, in particular evidently and perhaps (e.g. “as if she were ashamed”, “as if she failed to understand”, “as if she might be thinking”, “with the sense evidently that she had said too much”, “as if she herself appreciated”, “perhaps she found it less genial”, “as if he were amused”, José A. Álvarez-Amorós 90 etc. [CS 3: 230, 249, 262, 268, 270, 273, 312]). It is interesting to note that finite verbal expressions such as appeared and seemed can fill the same role as apparently and seemingly when the dative complement is omitted, which results in a construction with a high rate of occurrence in this tale (e.g. “in no case did she appear to know”, “she appeared to become aware”, “she seemed to wish to notify me”, “He seemed to smile at me”, “Miss Tita appeared to consider”, etc. [CS 3: 248, 248, 300, 312, 314]). The limits of standard unreliability in “The Aspern Papers”, as set by the above remarks on tagging, are narrow indeed. To the more or less expected occurrence of propositional and textual traces of metarepresentational activity, one must add an inordinate number of adverbial defactualizers that decouple information from the baseline ontology of the storyworld and inferentially show the narrator’s awareness of its non-factual nature for the reader’s guidance. As it happens, apparently, evidently, sort of, perhaps, and as if make a 4.04‰ of the total word count in this narrative, whereas they make 2.58‰ in The Sacred Fount, 2.42‰ in “The Way It Came”, 1.88‰ in “The Figure in the Carpet”, and only 1.59‰ in “The Diary of a Man of Fifty” - stories all of them open to be discussed in terms of unreliability. In combination, these three tagging strategies do form in “The Aspern Papers” a fairly obtrusive pattern, a closely-knit metarepresentational tissue attesting that the narrator’s ability to distinguish between fact and fancy is (mostly) in place, and so that he tends to be reliable even when he reports information arguably at odds with the implied James’s design. To close this paper, I will briefly address the duality between the telling and the experiencing selves as it bears on the issue of overt unreliability in “The Way It Came”. It is standard knowledge in narratology that first-person narrative situations entail the existence of two selves who, except in cases of extreme simultaneous narration, are ontologically identical, but functionally distinct. This means that characters become narrators at a given moment and report their experiences from epistemic vantage points which are inaccessible to their former selves. It is frequent, however, that narrators openly refer to the telling NOW in order to confirm, correct, pass judgement on, or otherwise consider their past experiences. Exploring the complexity of the relations between the telling and the experiencing selves clearly exceeds the scope of this paper. But James’s inveterate habit of alternating the reporting of the THEN , the reporting of the NOW , and the reporting of the THEN from the hindsight of the NOW provides yet another angle of approach to overt unreliability. “The Way It Came” turns on an eerie love triangle formed by two women and a man, all unnamed. It is prefaced by a rather skeptical editor’s note which, given the topic and the narrative resources of this story, functionally resembles the introductory section to “The Turn of the Screw”. If we stick to the first-person female narrator’s version, this is a tale of preternatural love “just engulfed in the infinite and still vibrating with human On the Limits of Henry James’s Unreliable Narration 91 emotion” (CS 4: 630). She has two friends, a man and a woman, who suffered alleged visitations from their respective mother and father at the moment of death. They have never met, and, despite many efforts to bring them together, fate has it that they always fail to come face to face. The man becomes the narrator’s fiancé, but he develops a disturbing interest in the woman he has never met, an interest which is requited. Obsessive jealousy erupts in the narrator who now goes to extremes of disloyalty to prevent their meeting, and bitterly regrets having sparked their mutual interest. The woman suddenly dies, but the narrator insists, in the face of the man’s hesitant denial, that both do now what they could not do before - meet and carry on an astral love affair. The narrator is abundantly conscious of her troubled mental condition bred by jealousy and of her shameful manoeuvres to keep her friends asunder, but it never occurs to her at the time the events happened that their growing affection and its supernatural consequences may only be true in her mind. This means that she is unable to keep track of herself as the source of her representations and thus fails to leave behind the usual verbal traces in the form of tags. By accommodating their gaze, however, readers can perceive a different story based on what might be called self-sustaining jealousy. It can be understood as a kind of a priori jealousy, disconnected from evidence. Nothing can dissipate it, not even the rival’s death, for, when this death happens, the subject simply transfers his or her obsession to the supernatural realm. In the case of “The Way It Came”, two factors seem relevant to the narrator’s mental construction - first, that both male and female appear to have medium powers; second, that the narrator herself confesses a taste for supernatural experiences and thinks it “gratifying to be the subject of one of those inexplicable occurrences that are chronicled in thrilling books and disputed about at learned meetings” (CS 4: 630). Since she cannot enjoy this “magnificent distinction” (CS 4: 632), she forces a volitional pattern on fictional reality and shapes it according to her emotional needs. The expected causal sequence is thus altered as in Miss Tita’s transfiguration episode, where the woman adopts in the narrator’s mind the personality and looks that another part of his mind requires to consider marrying her. The problem is, of course, that there is little ground in “The Way It Came” to rationally choose between the narrator’s version as spelt out above and its being an imaginary construction contingent on her subjectivity. The implied James appears to signal alternatively in both directions, and it is quite revealing that this is the only tale whose notebook discussion explicitly features the word “ambiguity” (1987: 153). Rather than a stable ironic form, it displays a shifting pattern which, in metarepresentational terms, means that the same portion of information can be processed either as traceable to the narrator’s - and/ or to the character’s - mind or as a sourcefree, positive fact. And this heavily conditions interpretation and hence any analysis of unreliability. José A. Álvarez-Amorós 92 To say that a first-person narrator reports intelligence consistent (or not) with the implied author’s inferred plans is to say little when there is a gap of variable width between the telling self and the experiencing self, that is, between the narrator proper and the character. Older narrators, for instance, can be at odds with their younger selves, but agree with the implied author’s design, or else both narrator and character can form a monolithic unit jointly at variance with such design. From the vantage point of the telling NOW the narrator can show awareness that the character’s view of factuality was misplaced and put it right by disclosing its subordination to his or her mental processes. But the narrator’s metarepresentational ability can also fail and bring about the spurious treatment of mental states as hard fictional fact. Both situations can alternate, though, and lead to a dual reading of “The Way It Came”. Let us consider four passages where the narrator’s NOW and the character’s THEN emerge and interact: I see now that she gave me no pretext and that I only found one in the way she looked at the fine face in the Bond Street frame. (CS 4: 614; my emphasis) I had thought it best to let her come; singular as this now seems to me I thought it diminished my guilt […] I blush to tell my story - I take it as my penance. (CS 4: 618; my emphasis) He stood there pleading with a candour that now seems to me beautiful for the privilege of having in spite of supreme defeat known the living woman; while I, with a passion I wonder at to-day […] could only reply […] (CS 4: 625; my emphasis) “She was dead! she was dead! ” I asseverated with an energy, a determination that it should be so, which comes back to me now almost as grotesque. (CS 4: 629; my emphasis except in “be”) Though with varying explicitness, all these texts highlight the ironic distance developed between the narrator and the character as the former progresses towards reliability by correctly indicating the role of her younger mind in shaping the reported material, which is, as we know, a standard symptom of overt unreliability. The first passage marks the turning point of this story. From the way the narrator’s friend looks at her fiancé’s photograph, she infers that her unguarded efforts to bring them together have produced unforeseen results. This is the first link in a long chain of suspicion, speculation, vague evidence, and guesswork culminating in the break-up of their engagement on grounds of a “monstrous” secret relation (CS 4: 634). But from the hindsight of the telling NOW the narrator knows better than the character. The word “pretext” is itself a revelation. As employed here, it ascribes the initial impulse for the whole process to her subjectivity, not to a factual state of affairs, and thus contributes to undermining the supernatural reading of the tale from the very outset. The sec- On the Limits of Henry James’s Unreliable Narration 93 ond passage proceeds along the same line. It hints at the narrator’s desperate ruse to wreck a meeting she had herself fixed for her two friends. She lies to the man to prevent him from turning up, but acts differently with the woman and meets her at the appointed time. A mind-easing strategy THEN becomes an eccentric move NOW , a notion which we feel closer, in principle, to the implied James’s narrative plan, and hence consonant with the contrition voiced by the narrator. Though less explicit, the last two passages work similarly. In the third one, the narrator’s corrective perception NOW of the man’s pleading and of her own passion opposes the character’s earlier view of them. The only difference is that readers are not expressly given the terms of contrast as in the two previous examples - e.g. “she gave me no pretext” vs. “I only found one” - even if they are easily inferred. The fourth text poses an enticing question whose answer may bear on the overall reading of “The Way It Came” - i.e. what looks “grotesque” to the narrator NOW ? Grammatically, it is the “determination” of her statement, granted, but it could also be her whole speech act and its content - i.e. that the woman visited the narrator’s fiancé after her death to effect a tryst that chance and deliberate interference had rendered impossible before. One way or another, these passages show a narrator actively engaged in destabilizing her earlier notions by decoupling them from fictional reality and foregrounding their mentalized status. This pattern, however, is occasionally reversed as becomes an ambiguous narrative. Here are two cases where the narrator aligns with the character, bridges the ironic gap between both, and, depending on which reading of the tale we assume, reports information in (dis)agreement with the implied James: I see it all now, I feel it, I live it over. It’s terribly void of joy, it’s full indeed to overflowing of bitterness; and yet I must do myself justice - I couldn’t possibly be other than I was. (CS 4: 631; my emphasis) [In the man’s sudden death] I distinctly read an intention, the mark of his own hidden hand. It was the result of a long necessity, of an unquenchable desire. To say exactly what I mean, it was a response to an irresistible call. (CS 4: 634; my emphasis) The first passage pivots on the striking phrase “I must do myself justice”. Uttered from the narratorial NOW , backed up by a number of present-tense verbal forms of perception and experience, and followed by a trenchant statement of inevitability, it denotes the narrator’s reluctance to correct the character’s habit of losing track of herself as the source of her representations. But the second one is even more categorical in this regard as it obliterates any distance between both figures concerning the supernatural issue. If a few pages earlier the narrator used the loaded term “pretext” to José A. Álvarez-Amorós 94 imply how she felt NOW about her former self’s perceptions, the emphatic “I mean”, strategically placed in the last sentence of the tale, balances prior doubts and misgivings and grants alleged authority to the interpretation of the narrative as a hapless love affair turned ghostly by untimely death. As the implied author’s view of what happens is not univocal, it is hard to infer what baseline ontology sustains the story, and so to decide which epistemic position(s) accord(s) with it. If the authorial plan was to concoct a supernatural tale, the narrator gains reliability as she identifies herself with the character’s version; if it was to tell a story of obsession and selfsustaining jealousy, the narrator grows more reliable as she openly denounces the character’s misperceptions as in the “pretext” case above. Discussing overt, self-acknowledged unreliability in “The Way It Came” is a vexing affair - as it is too, for instance, in “The Figure in the Carpet” - because two variables obtain and generate mutually supporting instabilities. We have, on the one hand, an ambivalent authorial project, and, on the other, the persistent interplay of the narrator’s and the character’s (conflicted) representations of authorial ambivalence, no less! Circularity rules, and it is a fairly arbitrary decision where to break the circle and start attributing truth values to the set of representations that comprise the story. 5. Concluding remarks In this paper, and aided by a blend of metarepresentational and rhetorical theory, I have attempted to make a case for what I call overt, transparent, or self-acknowledged unreliable narration in James’s short fiction as opposed to the more standard, covert, or opaque variety that has been the critical target for most commentators. Overt unreliability results from James’s propensity to deploy first-person narrators who are openly concerned with the existence of interferences between their mental processes and the baseline ontology of the storyworld, and so with the possibility that they should be reporting unwarranted or counterfactual information. That a narrator may be making a number of misplaced representations and at the same time reflecting on their fictional truthfulness does not intensify, in my view, his or her unreliability; it rather contributes to exposing its limits and keeping it under control. My contention here is that the bounds of Jamesian unreliability are rather modest and circumscribed, even in cases of unstable irony. The obstacles posed by such instability lie in opting for one of two plausible interpretations. But once a reading decision has been made, narrators can be argued to punctuate their telling with verbal clues visibly pointing to their subjectivities as the source of their representations. In “The Diary of a Man of Fifty”, a string of such clues culminates in a barrage of narratorial self-doubt that prompts readers to revise how they have assimilated the information previously provided; in “The Aspern Papers”, the sheer abundance of distancing and defactualizing expressions On the Limits of Henry James’s Unreliable Narration 95 underline the narrator’s coyness about claiming as fact what may only exist in his mind; and in “The Way It Came”, a jealousy-consumed woman-narrator has the occasional presence of mind to trace her younger self’s representations to her early subjectivity, even if such tracing fails at other times and promotes the overall undecidability of this tale. This is not all, however. Other first-person Jamesian narratives often branded unreliable - at least in one interpretation - also conform to this pattern. Looking back on the story’s THEN , the anguished narrator of “The Figure in the Carpet” can still tell fact from fancy and acknowledge the role of his foibles in the shaping of his report. “At last they [Corvick and Gwendolen] even bored me,” he says, “and I accounted for my confusion - perversely, I confess - by the idea that Vereker had made a fool of me” (CS 4: 583; my emphasis). Whether Vereker actually deceived the narrator is a moot point in this tale. But it is undeniable that the latter managed to keep track of the source of his representations, correct his earlier views, and even describe the mental attitude that prompted them (see also CS 4: 587). Likewise, the narrator of The Sacred Fount is insistently conscious of his fancy’s role in the construction of “a larger theory (and thereby a larger ‘law’) than facts, as observed, yet warranted”, and admits that this practice is “the common fault of minds for which the vision of life is an obsession” (James 1979: 23). And even tales seldom discussed in terms of unreliability do feature narrators forever preoccupied with their mental states pervading their narratives, as when the young critic-narrator of “The Death of the Lion” (1894) manages to trace his extravagant praise for Neil Paraday’s artistry to the fact that he was “at the end of three days, a very prejudiced critic” (CS 4: 360). As things stand, and despite occasional episodes of bewilderment experienced by James’s “supersubtle fry” (James 1984: 221) of hair-splitting narrators, it seems safe to argue that lucidity won the day. Works cited Bierce, Ambrose (1979). “Oil of Dog”. In: The Collected Writings of Ambrose Bierce. Intr. Clifton Fadiman. Secaucus: Citadel P. 800-803. Booth, Wayne C. (1961/ 1983). The Rhetoric of Fiction. London: Penguin Books. Booth, Wayne C. (1975). A Rhetoric of Irony. Chicago and London: U of Chicago P. Bradley, John R. (2000). Henry James’s Permanent Adolescence. Basingstoke: Palgrave. Cosmides, Leda & John Tooby (2000). “Consider the Source: The Evolution of Adaptations for Decoupling and Metarepresentation”. In: Dan Sperber (Ed.). Metarepresentations: A Multidisciplinary Perspective. Oxford: Oxford UP. 53-107. Doležel, Lubomír (1998). Heterocosmica: Fiction and Possible Worlds. Baltimore/ London: Johns Hopkins UP. Falconer, Graham (1987). “Flaubert, James, and the Problem of Undecidability”. Comparative Literature 39 (1). 1-18. James, Henry (1984). The Art of the Novel. Ed. Richard P. Blackmur. Boston: Northeastern UP. José A. Álvarez-Amorós 96 James, Henry (1987). The Complete Notebooks of Henry James. Leon Edel & Lyall H. Powers (Eds.). Oxford: Oxford UP. James, Henry (1996-1999). Complete Stories. 5 vols. New York: Library of America. James, Henry (1908). “The Aspern Papers”. In: The Novels and Tales of Henry James. New York: Scribner’s. 12. 3-143. James, Henry (1979). The Sacred Fount. Leon Edel (intr.). New York: Grove P. Kellum, Sharon Smart (1976). “The Art of Self-Incrimination: Studies in Unreliable Narration”. Diss. U of California at Berkeley. Köppe, Tilmann & Tom Kindt (2011). “Unreliable Narration with a Narrator and Without”. Journal of Literary Theory 5 (1). 81-94. Korg, Jacob (1962). “What Aspern Papers? A Hypothesis”. College English 23 (5). 378-381. Marsh, Huw (2018). “Narrative Unreliability and Metarepresentation in Ian McEwan’s Atonement; or, Why Robbie Might Be Guilty and Why Nobody Seems to Notice”. Textual Practice 32 (8). 1325-1343. McHale, Brian (1981). “Islands in the Stream of Consciousness: Dorrit Cohn’s Transparent Minds”. Poetics Today 2 (2). 183-191. Nünning, Ansgar (1979). “‘But Why Will You Say that I Am Mad? ’: On the Theory, History, and Signals of Unreliable Narration in British Fiction”. Arbeiten aus Anglistik und Amerikanistik 22. 83-105. Nünning, Ansgar (2008). “Reconceptualizing the Theory, History, and Generic Scope of Unreliable Narration: Towards a Synthesis of Cognitive and Rhetorical Approaches”. In: Elke D’hoker & Gunther Martens (Eds.). Narrative Unreliability in Twentieth-Century First-Person Novel. Berlin/ New York: de Gruyter. 29-76. Phelan, James (2005). Living to Tell about It: A Rhetoric and Ethics of Character Narration. Ithaca: Cornell UP. Phelan, James (2007). “Estranging Unreliability, Bonding Unreliability, and the Ethics of Lolita”. Narrative 15 (2). 222-238. Phelan, James (2017). Somebody Telling Somebody Else: A Rhetorical Poetics of Narrative. Columbus: Ohio UP. Phelan, James & Mary Patricia Martin (1999). “The Lessons of ‘Weymouth’: Homodiegesis, Unreliability, Ethics and The Remains of the Day”. In: David Herman (Ed.). Narratologies: New Perspectives on Narrative Analysis. Columbus: Ohio UP. 88-109. Recanati, François (2000). Oratio Obliqua, Oratio Recta: An Essay on Metarepresentation. Cambridge: MIT Press. Rivkin, Julie (1989). “Speaking with the Dead: Ethics and Representation in ‘The Aspern Papers’”. The Henry James Review 10 (2). 135-141. Ryan, Marie-Laure (1991). Possible Worlds, Artificial Intelligence, and Narrative Theory. Bloomington/ Indianapolis: Indiana UP. Snyder, Katherine V. (2004). Bachelors, Manhood, and the Novel, 1850-1925. Cambridge: Cambridge UP. Sternberg, Meir & Tamar Yacobi (2015). “(Un)Reliability in Narrative Discourse: A Comprehensive Overview”. Poetics Today 36 (4). 327-498. Weinstein, Philip M. (1970). “The Exploitative and Protective Imagination: Unreliable Narration in The Sacred Fount”. In: Morton W. Bloomfield (Ed.). The Interpretation of Narrative: Theory and Practice. Cambridge: Harvard UP. 189-209. Yacobi, Tamar (2000). “Interart Narrative: (Un)Reliability and Ekphrasis”. Poetics Today 21 (4). 711-749. Yacobi, Tamar (2001). “Package Deals in Fictional Narrative: The Case of the Narrator’s (Un-)Reliability”. Narrative 9. 223-229. On the Limits of Henry James’s Unreliable Narration 97 Yacobi, Tamar (2005). “Authorial Rhetoric, Narratorial (Un)Reliability, Divergent Readings: Tolstoy’s Kreutzer Sonata”. In: James Phelan & Peter J. Rabinowitz (Eds.). A Companion to Narrative Theory. Malden/ Oxford: Blackwell. 109-123. Zerweck, Bruno (2001). “Historicizing Unreliable Narration: Unreliability and Cultural Discourse in Narrative Fiction”. Style 35. 151-178. Zunshine, Lisa (2006). Why We Read Fiction: Theory of Mind and the Novel. Columbus: Ohio State UP. José A. Álvarez-Amorós Department of English Studies University of Alicante