Plotting Memory. What Are We Made to Remember When We Read Narrative Texts?

1 Prolegomena to a Theory of ›Emplotted‹ Memory

The activity of remembering in the very process of reading narrative can be regarded from two perspectives: as an empirical psychological problem or as one of implied readership. It goes without saying that a narrative text can give rise to a variety of actual mnestic processes. It can be read with an attitude of distracted forgetfulness or be accompanied by mnemotechnic activities such as notetaking. I will argue, however, that memory is also an aspect of implied readership, i. e. of the conditions of reception provided by the text itself (cf. Iser 1987, 34). This becomes particularly apparent when a narrator explicitly recapitulates prior events in order to make his readers understand what is about to happen, but we will see that there are also more elegant and indirect ways of guiding the recipient’s memory in such a way that the plot can be grasped as a meaningful structure. Implied remembering as a dimension of implied readership means that the text appeals to the memory of its readers, that it employs devices that make them recall what has already been told or make them anticipate that what is being told will have to be recalled later. I will try to show that this kind of ›implied‹ narrative memory can be reconstructed by means of textual analysis.

The earliest and still the most productive comment on this kind of remembering can be found in Aristotle’s Poetics. To be precise, Aristotle writes but one word on the topic and it refers in the first place to the mythos of tragedy, but in the context of the whole treatise it is so rich in implications that it can serve as a starting point for my inquiry.^[2] In the crucial seventh chapter of the Poetics, Aristotle highlights two features of tragic emplotment (systasis): wholeness (holon) and magnitude (megethos). Whereas wholeness is described as a structure consisting of beginning, middle and end, Aristotle uses a comparison in order to determine the accurate magnitude of tragedy. Animals or other objects »with a structure of parts« (Aristotle 1995, 1450b) are beautiful when they are neither too small nor too big and can thus be contemplated as a whole. Having established this analogy taken from the field of visual perception, Aristotle draws the corresponding conclusion for the accurate magnitude of the tragic plot: »So just as with our bodies and with animals beauty requires magnitude, but magnitude that allows coherent perception [eusynopton], likewise plots require length, but length that can be coherently remembered [eumnēmoneuton]« (Aristotle 1995, 1451a).

While the adjective eusynotpon is applied to the structure of events merely by analogy, eumnēmoneuton tells us what this is literally meant to say: The magnitude of the plot should not exceed the limits of human memory. We might infer that the size measured by visual perception refers to space, whereas the magnitudes comprehended by memory refer to time, but Aristotle immediately cautions his readers that the discipline of poetics is not concerned with the empirical duration of the performance. Critics have wondered whether »it is the text or the plot which is in question«, quoting passages that seem to confirm both options (Heath 1989, 44). Given that Chapter 17 of the Poetics considers a similar problem from the perspective of poetic production and asks tragedians to begin by devising an abstract of the mythos that can be contemplated (theōreisthai, Aristotle 1995, 1455b), it seems more likely that the requirement of memorial comprehensibility concerns a skeletal version of the plot that can easily be retained. What is at stake then is not the perception of real time but the comprehension of a structure with a defined telos (cf. Halliwell 1987, 99, and Schmitt in Aristotle 2008, 364).

Chapter 7, however, actually begins with the discussion of another feature of the tragic plot, which is not explicitly related to memory: wholeness. Magnitude concerns the size of the mythos, wholeness defines its inner structure as one of necessity and of probability. When Aristotle, in the second half of the chapter, moves on to discuss magnitude and draws the analogy between the animal body and the tragic plot, the question of inner structure is not completely lost since »beauty consists in magnitude and order [taxei]« (Aristotle 1995, 1450b). Commentators agree that taxis refers back to what has just been discussed, i. e. to the requirement of wholeness (cf. Schmitt in Aristotle 2008, 363; cf. Heath 1989, 41). This implies that the order of the body’s parts corresponds to the concatenation of events ex anankēs and that the holistic structure of beginning, middle and end is another prerequisite of plot-related beauty. The last sentence of the section will draw an even closer tie between magnitude and wholeness: »the size [megethei] which permits a transformation to occur, in a probable or necessary sequence of events, from adversity to prosperity or prosperity to adversity, is a sufficient limit of magnitude« (Aristotle 1995, 1451a). Using wholeness as a standard of magnitude, Aristotle emphasizes once more that the extension of the plot is subject to inherent structural principles and not to the external measure of time.

The question I would now like to raise is one that Aristotle does not explicitly answer, namely how wholeness and ›memorability‹ impact each other. Within the conceptual framework provided by the Poetics, it is possible to argue that the coherent order ex anankēs facilitates the memorial retention of the plot. To take up Aristotle’s biological analogy, an animal of a comprehensible size would still be difficult to contemplate if its inner order (taxis) was excessively complex, for instance if it consisted of a huge number of arbitrarily arranged parts. A labyrinthine plot constructed on this model would be difficult to retain due to its inner structural properties, not due to its extension. According to this line of argument, both the requirements of completeness and of proper magnitude have to be met for a plot to be easily rememberable.

The Poetics enables us to draw yet another conclusion that is not overtly stated. As Stephen Halliwell points out, one can »establish a clear association in [Aristotle’s] theory of poetry between unity and intelligibility« (Halliwell 1987, 100). The structure ex anankēs enables the spectator to understand how the sudden reversal (metabasis) comes about. The cathartic effect of tragedy depends on this process and is best produced »when events occur both contrary to expectation and yet on account of one another« (Aristotle 1995, 1452a). But at the moment of the reversal only the final situation of the tragic concatenation of events can be seen on stage while the series of causes that have led to the catastrophe has to be supplemented by the spectator’s memory. This is, as I think, the major reason why Aristotle wants the tragic plot to be eumnēmoneuton: The rational intelligibility and the emotional effect of tragedy are contingent on the spectator remembering the events that bring about the change.^[3]

One last question remains to be discussed with regard to section 7 of the Poetics: Does the requirement of eumnēmoneuton also apply to epic, i. e. to a narrative genre? After all, both have mythos as their central qualitative element in common. The parallels between epic and tragedy have often puzzled readers of the Poetics since Aristotle states in section 18 that epic has a »multiple plot« (polymython), but then, in section 23, applies the rules of the tragic plot to epic, including the requirements of completeness and proper magnitude as well as the analogy of the animal body. In section 5, we are told that tragedy and epic differ in length (tō mēkei, ibid., 1449b). The tragic plot should not exceed »a single revolution of the sun« (ibid.), whereas epic time is unlimited. And yet, again in section 23, Aristotle praises Homer for not having told the story of the entire Trojan war because such a plot would have been »too large and incoherent [megas kai ouk eusynoptos], or else, if kept within moderate scope, too complex in its variety« (ibid., 1458b).

One can try to reconcile these statements by arguing that epic – like tragedy – centres on a coherent plot, but that – unlike tragedy – it achieves greater magnitude by integrating episodes that entertain a loose connection to the main strand of the plot without being completely independent. This would also imply that epic can preserve its unity even when its length exceeds that of tragedy. But how is this possible? Why is tragedy unable to contain episodes that are hierarchically subordinate to the main plot? Arbogast Schmitt (in Aristotle 2008, 304) has suggested that the reason for this difference lies not so much in the structure of the mythos itself as in the mode of mimesis: Epic involves both narrative (apagellia, Aristotle 1995, 1449b) and character speech, whereas tragedy shows all characters in direct action. Due to the presence of a narrative voice, epic can preserve the comprehensibility of its plot even when the limits of a tragic mythos have been exceeded both in length and complexity. The epic narrator can, for instance, recall what has already been told, summarize a long span of time in a few verses or announce the future relevance of the events that are being told. In other words, what is eumnēmoneuton in epic would not be eumnēmoneuton in tragedy because narrative devices can help to extend the »upper limit imposed by the memory« (Heath 1989, 44). In contrast, the memorability of the tragic plot has to be achieved through its limited scope and its holistic structure alone, i. e. (in Aristotle’s terms) through the ›what‹ of mimesis, not through the ›how‹ (cf. Aristotle 1995, 1447a).

I will now turn to a 20^th-century reader of Aristotle, Peter Brooks, whose concept of narrative plot is similarly predicated on the holistic structure consisting of beginning, middle and end. Brooks also concurs that narrative is a »form of understanding and explanation« (Brooks 1992, 10), but while some passages of the Poetics suggest that Aristotle conceives mythos as a synchronic structure that can be abstracted from its successive unfolding in the play (cf. Aristotle 1995, 1455a–b), Brooks endorses a dynamic vision of plot, arguing that »plotting« is the activity that pulls the reader through the text. The engagement with plot involves a passionate quest for meaning that mimics the desire of the intradiegetic protagonists and whose goal is the end as the moment when the plot’s potential for meaning will be fully realized. Brooks refers to the Aristotelean eumnēmoneuton at the beginning of his study:

Just as in the visual arts a whole must be of a size that can be taken in by the eye, so a plot must be ›of a length to be taken in by the memory‹. This is important, since memory – as much in reading a novel as in seeing a play – is the key faculty in the capacity to perceive relations of beginnings, middles, and ends through time, the shaping power of narrative (Brooks 1992, 11).

Building on Aristotelean teachings, Brooks emphasizes the teleological orientation of narrative by pointing out that the meaning of plot always emerges retrospectively and that both the beginning and the middle of a story have to be reinterpreted once the end has been reached. In other words, the understanding of plot involves a permanent oscillation between anticipation and retrospection. The end is the moment when we will have to look back on previous parts of the plot, but as competent readers we know that this moment will come and thus anticipate the retrospective reassessment of what we read. Brooks calls this activity the »anticipation of retrospection« (ibid., 23), but we could equally call it the anticipation of remembering since the retrospective restructuring of meaning presupposes that the reader be capable to recall what has been told. By consequence, the awareness of having-to-remember accompanies the entire process of reading.

Brooks thus empowers temporality with regard to what is being told and – even more importantly – with regard to the activity of reading itself. Quoting Gérard Genette’s reflections on discourse time, Brooks underlines that narrative »depends on its being gone through in sequence and succession, and that it thus metonymically ›borrows‹ a temporality from the time of its reading« (Brooks 1992, 20; cf. Genette 1972, 77 sq.). Reading unfolds in a linear sequence, but it is also characterized by the movements of remembering and anticipation. The longer the narrative and the more coherent its structure, the longer the temporal intervals that have to be bridged and the more intense the activity of retrospection.

We should not imagine this process as a laborious mnemotechnic exercise, though. Even if narrative plots only grant their meaning to readers who are able to rise to the mnestic level of the text, the process of remembering as conceived by Brooks involves desire and pleasure. What enables the reader to return to previous moments of the plot is the narrative tension that has been induced by the beginning. Drawing on Freud’s Beyond the Pleasure Principle, where the ›plot‹ of life begins when a traumatic shock creates the properties of animate matter, Brooks writes: »For plot starts (or must give the illusion of starting) from that moment at which story, or ›life,‹ is stimulated from quiescence into a state of narratability, into a tension, a kind of irritation, which demands narration.« (Brooks 1992, 103) With a playful move typical to his practice of theoretical bricolage, Brooks combines the coherent structure of the Aristotelean mythos with the psychoanalytical concept of trauma. The plot-related compulsion to repeat is thus endowed with a teleological orientation and serves to absorb the tension of the beginning until the end has been reached.

What matters to the question of memory, though, is that the stimulus of the beginning induces a state of tension within the reader and that this state subsists as long as the end is being deferred. This bio-psychological process serves a mnemonic function in so far as it signals that something past remains to be resolved. On the diegetic level, the tension stimulated by narrative can arise from an unsolved conflict, from a desire that has been denied or from a painful experience of separation. In each case, the readerly experience of unrest points back to something that cannot be forgotten. Narrative tension means that a psychic need drives the reader to continue remembering. It is high time to abandon the idea that reading simply requires a storage of information. In fact, emplotted remembering is entangled with the most basic driving forces of storytelling.

I would like to conclude the theoretical section of this article by briefly glancing at some empirical psychological studies that have examined problems related to emplotted remembering. Since Frederic Bartlett’s pioneering studies in remembering (cf. Bartlett 1954 [1932]), psychologists have been conducting experiments on how stories are comprehended and recalled. The experiments consistently show that some types of stories and certain narrative events are more frequently remembered than others. Arguing that readers encode and retrieve stories using a basic plot schema, studies from the 1970s use a narrative grammar modelled on the one proposed by Vladimir Propp in order to predict which kind of narrative structure will be better recalled (cf. Thorndyke 1977, Mandler/Johnson 1977, Mandler 1978). The more a story conforms to this ideal schema, which consists of simple building blocks such as setting, beginning, reaction, attempt and ending, the more it is likely to be adequately remembered. Studies that followed up on this approach pointed out that causal concatenation and a high density of causal relationships increases the recall of narrative events (cf. Black/Bower 1980, Trabasso/van den Broek 1985).

While the papers mentioned so far examine retrospective recall, studies on inference have probed into the activity of remembering during the process of reading. By means of inference, »information that was not explicitly stated in a text […] becomes part of the readers’ discourse representation« (Gerring/Wenzel 2015, 363). Numerous experiments confirm that inferred connections between information given in different sections of the text are of fundamental importance to comprehension. Memory-based processing theories argue that by means of ›resonance‹, stored information is automatically retrieved from the long-term memory and becomes once again available in the working memory when it is triggered by cues in the text segment that is currently being read (cf. McKoon/Ratcliff 1992, Gerring/McKoon 2001, Gerring/O’Brien 2005 and also Ericsson/Kintsch 1995, 223). While ›resonance‹ takes place without awareness, constructionist theories of inference assume that a number of inferences are made ›on-line‹ (i. e. during the process of reading) because readers engage in a conscious »Search (or effort) after meaning« (Graesser/Singer/Trabasso 1994, 371; italics in the original). The advocates of this view also point out that comprehension is challenged when coherence depends on information that cannot easily be provided:

Access to generic information in LTM [= long-term memory] is very quick and executed in a parallel, rather than a serial, fashion. Access to specific information sources in LTM is accomplished more slowly (in parallel); the utilization of content within a specific information source is comparatively time consuming and, at times in a serial fashion, places more demands on WM [= working memory]. Therefore, it takes a noticeable amount of time to reinstate information that appeared much earlier in the text being read, after the information exits WM. Comprehension processes, including inference generation, slow down to the extent that the demands on WM approach the upper bound of capacity limitations. When demands on the WM exceed the upper bound, there is a catastrophic deterioration of comprehension, and few inferences are constructed (Graesser/Singer/Trabasso 1994, 380).

It should not be forgotten that psychological experiments operate with sample texts that hardly ever surpass a few lines. In long narrative prose, the critical storage limits of working memory are frequently stretched to the limits, and the information to be retrieved from long-term memory is often quite specific. During the case studies, it will be crucial to observe how comprehension is enabled even when inferences are difficult to make or how inferences are facilitated by certain narrative techniques.

Conceiving comprehension as a search, the constructionist view quoted above stresses the finalistic orientation of reading. The teleological impulse of narrative that we have found in Aristotle and Brooks also resonates with other empirical inquiries. In a classical study from the early days of Gestalt psychology, Bluma Zeigarnik argued that test persons are more likely to recall activities when they have been interrupted while performing them. This held true in particular when the activity was interrupted at the moment of maximum affective engagement and when the task itself had a holistic structure that led up to a superordinate goal (»Endhandlungen«). Continuous activities (»fortlaufende Handlungen«) that simply consisted of repetitive steps were less likely to be recalled even when they had been interrupted (cf. Zeigarnik 1927, 54). The so-called Zeigarnik effect has been difficult to reproduce and thus subject to much criticism (cf. MacLeod 2020, 1080). Experiments in the recall of stories, however, suggest at least that causally connected events are more likely to be remembered than purely additive sequences (cf. Mandler/Johnson 1977, 116). Furthermore, the hypothesis that there is a general human need for closure is still on the table (cf. Kruglanski/Webster 1996).

It turns out that some of our basic theoretical assumptions – the importance of remembering for the comprehension of narrative, the correlation between structural coherence and memorability as well as the strife for meaning – are in tune with the present findings of empirical psychology. But this is not the crucial point of this article. As has already been pointed out and as will become even clearer in the course of the case studies, I am focussing on implied readership and not on the average mnestic performance of 20^th- and 21^st-century readers. Furthermore, my model of emplotted memory does not serve as a hypothesis to be confirmed or disproved by the case studies, but as a tool that is necessarily transformed by each reading. After all, it would be an unproductive result – at least from the perspective of literary studies – to discover that a theory is simply able to predict what happens in a given text.

It should thus be clear that the theories discussed so far were helpful to conceptualize some hermeneutic intuitions, but that they also aim to raise new questions. Above all, the suggestions I have gathered aim to open a field for textual analysis. It will be easier now to focus our attention on how the reader’s implied memory is guided by narrative and how it contributes to the construction of plot. The requirement that the plot be eumnēmoneuton neither defines a timeless aesthetic standard nor serves as a prediction of potential empirical results, but fulfils a heuristic function. A closer look into the workings of emplotted remembering has become indispensable in order to push the inquiry forward.

2 Memory Plots: Chrétien, Montalvo, Goethe

The question that most urgently needs to be addressed is probably the one concerning the specifically narrative techniques that help to enhance the memory of the reader. Why are narrative texts able to surpass the memorial horizon of tragedy? I would like to use this question as a starting point in order to embark on a series of three case studies. The texts I have selected all deal with male protagonists faced with mnestic problems that tend to arise in the context of erotic relationships: faithfulness, keeping promises, avoiding distraction, meeting deadlines, etc. Covering a period of more than 600 years and representing key moments in the history of the novel, they provide a useful sample for an exploration of plot-related memory.

3 Conclusion: Towards a Cultural History of Emplotted Memory

The case studies have helped to corroborate some assumptions of the present article. The first conclusion I would like to draw is that narrative remembering can be analysed on a textual level as a dimension of implied readership. Furthermore, narrative has to be understood as a way of ›plotting memory‹, i. e. of creating a temporal sequence of events that turns out to be meaningful only at the condition of being adequately remembered. While narrative plot structures memory as a temporal process, remembering also aims at overcoming time: The Aristotelean notion of eusynopton implies that we simultaneously comprehend things that are told subsequently. Finally – and this is the point where conclusions become much less definite – different narrative structures and techniques embody different kinds of remembering. The mnemotechnic devices we have encountered in Chrétien, Montalvo and Goethe are not only diverse, but also located at different conceptual levels: diverging measures between discourse time and story time, mirroring relationships between the part and the whole, implicit continuations and explicit recapitulations, iconic cross-references. It seems hard, if not impossible, to determine what these narrative and poetic strategies have in common, apart from the obvious fact that they both demand and enhance narrative remembering.

This heterogeneity cannot be accounted for if we squeeze it into some abstract narratological taxonomy or if we reduce it to general psychological laws. Quite on the contrary, the next task for an inquiry into emplotted memory will be to link the analysis of narrative structure and of implied readership to cultural discourses on remembering and to embed it in particular historical contexts. The structures and devices that instantiate different processes of remembering could be interpreted, for instance, in relation to the recurring polemics against distracted reading. To mention but one example, in his Brief über den Roman Friedrich Schlegel criticises the female addressee for her excessive and oblivious consumption of novels and romances: »Your memory scorns this vulgar stuff which has become a necessity through an unfortunate habit of your youth; what has to be acquired so laboriously, is entirely forgotten« (Schlegel 1968, 95). Reading too much and without distinction means that nothing will be retained. Schlegel’s half-ironic critique is directed against a practice that actually existed, namely massive and unsystematic borrowing from lending libraries. As Franco Moretti has pointed out, 18^th-century readers reacted to the exploding production of novels and the increasing number of texts that were available on the literary market by renouncing their claim to narrative coherence (cf. Moretti 2010, 7).^[5] They borrowed the volumes of a novel out of order and did not read the texts in their entirety. A novel such as Wilhelm Meisters Lehrjahre, with its minutely elaborated network of symbolic cross-references and the corresponding importance of implied remembering, clearly discourages such rapid consumption. Similar concerns surface in the 16^th-century debate on the Italian romanzo: Torquato Tasso laments that modern audiences desire an ever-increasing amount of variety (varietà) in narrative and notes that the romances that best satisfy this demand – Boiardo’s Orlando innamorato and Ariosto’s Orlando furioso – violate the Aristotelean prescription that the plot be eumnēmoneuton (cf. Tasso 1977, 26 and 41).

I would hence like to suggest that emplotted remembering is grounded in the structure of the text, but that it also responds to historical practices of reading and to discourses on the appropriate uses of literature. Texts that give massive importance to memory call for empirical readers that are able to retain what is narrated over long periods of time. The forms of subjectivity thus produced have a wider cultural resonance. Why were readers around 1500 requested to recall episodes over hundreds of pages, while the addressee of a Bildungsroman written in the late 18^th century has to retrace connections between shifting positions in an iconic network? Is there a link between the contemporaneous crisis of rhetoric and the shift from narratives based on storage to narratives based on inscription? After all, the memorial metaphor of the thesaurus is inextricably entangled with the rhetorical art of memory that came under attack in modernity (cf. Weinrich 1964, 23). A much larger corpus and a much more sustained inquiry would be indispensable to really ask such questions. And in order to answer them, the approach adopted so far will have to integrate other methods such as discourse analysis, media history and the history of reading. The result could be a history of emplotted memory that accounts both for narrative structure and broader cultural concerns.