Beyond Comparison: Picture, Text and Method

W. J. Thomas Mitchell

There is no link that could move from the visible to the statement, or from the statement to the visible. But there is a continual relinking which takes place over the irrational break or crack.

— Gilles Deleuze, Foucault.

Gilles Deleuze has suggested that the antinomy of word and image is something like an historical a priori in Foucault's writing,¹ cropping up like an unruly weed whenever some attempt is made to stabilize and unify the field of representation and discourse under a single master-code (mimesis, semiosis, communication, etc.). One traditional answer to this problem in the (American) academic study of the representational arts has been the comparative method. The tradition of "Sister Arts" criticism, and the pedagogy of "literature and the visual arts," has been the dominant model for the interdisciplinary study of verbal and visual representation. In its ambitious forms "interartistic comparison" has argued for the existence of extended formal analogies across the arts, revealing structural homologies between texts and images united by dominant historical styles such as the Baroque, the Classical, or the Modern.² In its more cautious versions it has been content with tracing the role of specific comparisons between visual and verbal art in poetics and rhetoric, and examining the consequences of these comparisons in literary and artistic practice.³ Although these methods have been mainly associated with the work of literary scholars moonlighting in the visual arts, they have a certain hidden institutional presence in art history as well, where the academic legitimation of the field (at least in the U.S.) has sometimes been grounded in the sense that art history provides a "visual analogue" to the study of literature. Donald Preziosi notes that "one of the earliest formal appearances of art history in the American university system occurred in 1974 when the Harvard Corporation appointed Charles Eliot Norton Lecturer on the History of the Fine Arts as Connected with Literature."⁴ The corporate, departmental structure of universities reinforces the sense that verbal and visual media are to be seen as distinct, separate, and parallel spheres that converge only at some higher level of abstraction (aesthetic philosophy; the Humanities; the Dean's office).

Given this background of tradition and institutional structuring, it's hardly surprising that the comparative method has seemed like the only systematic way to talk about relations of word and image (I lay aside for the moment ad hoc discussions based in historical contingencies such as the friendships of painters and poets). Comparison is the ideal trope for figuring "action at a distance" between different systems. When accompanied by a set of standard and acceptable differentiations between the institutions of "literature" and "visual art" (that is, differentiating figures like the "spatial" and "temporal" arts, the iconic and symbolic, even "the verbal" and "the visual") a conceptual framework for comparative study is established. This framework may then be refined (as Wendy Steiner has shown) by distinguishing three-term ("substantive") from four-term ("relational") analogies. This "structured set of interrelations," according to Steiner

has made the comparison of the two arts seem both a possible and a valuable enterprise. It has allowed contemporary theoreticians to discover what they consider 'real' similarities — real because they correspond to structures in other areas of human experience and because they can be discussed in the privileged terms of science.⁵

The next step is to historicize this framework with a master-narrative that can be seen as common to the internal traditions of the disciplines of literature and art history, and to their bureaucratic subdivisions. The sequence of "periods" from medieval to postmodern (with its inevitable consequences for the definition of academic "positions") is overlaid on the framework of analogy and one has an interdisciplinary "field"— the comparative study of literature and the visual arts. Why is it worth doing this? "The answer," says Wendy Steiner, "is that the interartistic comparison inevitably reveals the aesthetic norms of the period during which the question is asked."⁶ In short, the method will not make any waves: it will simply provide confirmation and elaboration of the dominant historical and conceptual models that already prevail in the discipline, offering the sort of highly general, watered-down historicism that can be extracted to match up visual art and literature.

One might suppose that the scientific and systematic tradition of European semiotics would provide an alternative to the pragmatism of the American "interartistic comparison" model, but in practice the only thing that seems to change is the generality of the claims. A good example is Göran Sonesson's Pictorial Concepts, which is the most ambitious recent attempt to synthesize the semiotic analysis of visual representation and verbal discourse. But Sonesson simply replaces the unifying concept of "art" with more general notions of "picturality," "literarity," and "meaning" to produce a higher level comparative method and a more insistently scientistic rhetoric. The basic task of semiotics, argues Sonneson, is "to render comparable, in their similarities and differences, the results of the different human and social sciences, through the unification of their concepts and methods."⁷

Since the filling in of familiar grids with novel (but unchallenging) detail is what usually passes for "the advancement of learning" in academic research, the practice of interartistic or intersemiotic comparison seems a safe professional option. It is a nice extra in times of budgetary surplus, and it may have a certain survival value in times of retrenchment, when the ability to teach in more than one department could be attractive to a cost-cutting administration. At its best, the comparative method can offer a kind of intellectual housekeeping, sorting out the differences and similarities not only between various kinds of cultural objects, but between the critical languages that are brought to bear on them. It also has the advantage of tradition: there is no question that poetics, rhetoric, semiotics, and aesthetics are riddled with the tropes of interartistic comparison, and that these figures, like any other, deserve analysis.

Any such analysis ought to proceed, however, by acknowledging three basic limitations in the comparative method. The first is the presumption of the unifying, homogeneous concept (the sign, the work of art, semiosis, meaning, representation, etc.) and its associated "science" that makes comparative/differentiating propositions possible, even inevitable. The second is the whole strategy of systematic comparison/contrast that ignores other forms of relationship, eliminating the possibility of metonymic juxtapositions, of incommensurability, and of unmediated or non-negotiable forms of alterity. The third is the ritualistic historicism, which always confirms a dominant sequence of historical periods, a canonical master-narrative leading to the present moment, and which seems incapable of registering alternate histories, counter-memories, or resistant practices. "Interartistic comparison" suffers, in short, from comparison, from artiness, and from its inability to do anything but confirm received versions of cultural history. Inter-semiotic comparison has the same problem, substituting scientism for artiness.⁸

All the same, the impulse to "interartistic comparison" cannot be totally pointless. It must correspond to some sort of authentic critical desire to connect different aspects and dimensions of cultural experience. The challenge is to redescribe the whole image/text problematic that underwrites the comparative method, and to identify critical practices that might facilitate a sense of connectedness while working against the homogenizing, anaesthetic tendencies of comparative strategies and semiotic "science." In Iconology I tried to show why the prevailing tropes of differentiation between verbal and visual representation (time and space, convention and nature, the ear and the eye) do not provide a stable theoretical foundation for regulating comparative studies of words and images. I also tried to suggest the ways in which these tropes function as "ideologemes", relays between semiotic, aesthetic, and formal boundaries, and figures of social difference. In the remainder of this essay I want to map out some practical and methodological implications of these conclusions for the study of words and images. I'm concerned here particularly with questions of pedagogy. How do we teach literature and textual analysis in the era of "the pictorial turn" and the dominance of visuality? How do we teach "art history" when the distinctiveness and identity of "art" is exactly what cannot be taken for granted? What do we want from a course, a curriculum, and a discipline that seeks to connect and cross the shifting boundaries of verbal and visual representation?

My first goal for a course that connects images and texts has always been to confront this topic only where it is necessary and unavoidable. The historicist course in "comparative arts" that compares (say) Cubist painting with the poems of Ezra Pound, or the poems of Donne with the painting of Rembrandt is exactly the sort of thing that seems unnecessary, or whose necessity is dictated mainly by the administrative structure of knowledge. It may enlarge our knowledge of Modernism to study both Pound and the Cubists, but it tends also to reduce that knowledge to a set of abstract propositions about the period aesthetic. The real subject of such a course is not the image/text problem, but Modernism. There is no good reason to stop with the comparison of painting and poetry; one might as well go on to look at music, science, film, philosophy, and historical writing. That, in fact, is what a densely woven interdisciplinary study of Modernism would require, though it would not require a comparative methodology as a starting point, much less a singling out of poetry and painting as the objects for comparison. But if one's subject is not Modernism or the Baroque, but the problem of the image/text — that is, the heterogeneity of representational structures within the field of the visible and readable — then one is led to rather different starting points. One may want to shed the comfort of the historicist security blanket with its preordained rediscoveries and refinements of the period's aesthetic concept. One may be forced to leave the enclave of "the aesthetic" altogether to engage with vernacular forms of representation. One will, above all, be constrained to take on the subject of the image/text, not as a kind of luxury "option" for the amateur, the generalist, or the aesthete, but as a literal, material necessity dictated by the concrete forms of actual representational practices. Perhaps these starting points seem so obvious to me because I began this sort of work in a rather different place, with the composite art of William Blake, a poet-painter whose illuminated books seem absolutely to demand a reader capable of moving between verbal and visual literacy, and whose relation to the prevailing aesthetic concepts of "Romanticism" seems so equivocal. Even with Blake's mixed media, however, I was always struck by the oddness, the arbitrariness of the demand for double literacy. Blake's poetry had been (and still continues to be) taught with only the most cursory attention to his graphic art. I know this for a fact because, despite my commitment to the critical analysis of Blake's composite art, I still find it important occasionally to teach him just as a writer. For certain purposes it might be more important to read Rousseau's Emile next to the Songs of Innocence and Experience than to look at the illustrations. Blake has always served for me then, as a kind of exemplar of both the temptation and arbitrariness of comparative studies of verbal and visual art. Blake's illuminated books invite "comparative" analysis, if ever any works of art did; but even they require it only for the specific purpose of confronting the formal materiality and semiotic particularity of Blake's text. If one wants to see Blake's work for what it is, in short, one cannot avoid the problem of the image-text.⁹

One can and must, however, avoid the trap of comparison. The most important lesson one learns from composite works like Blake's (or from mixed vernacular arts like comic strips, illustrated newspapers, and illuminated manuscripts) is that comparison itself is not a necessary procedure in the study of image-text relations. The necessary subject-matter is, rather, the whole ensemble of between media, and relations can be many other things besides relations similarity, resemblance, and analogy. Difference is just as important as similarity, antagonism as crucial as collaboration, dissonance and division of labor as interesting as harmony and blending of function. Even the concept of "relations" between media must be kept open to question: is radical incommensurability (cp. Foucault on Magritte's Pipe)¹⁰ a relation or a non-relation? Is a radical synthesis or identity of word and image (the utopian calligram) a relation or a non-relation? The key thing, in my view, is not to foreclose the inquiry into the text-image problem with presuppositions that it is one kind of thing, appearing in a certain fixed repertoire of situations, and admitting of uniform descriptions or interpretive protocols.

The best preventive to comparative methods is an insistence on literalness and materiality. That is why, rather than comparing this novel or poem with that painting or statue, I find it more helpful to begin with actual conjunctions of words and images in illustrated texts, or mixed media such as film, television, and theatrical performance. With these media, one encounters a concrete set of empirical givens, an image-text structure responsive to prevailing conventions (or resistance to conventions) governing the relation of visual and verbal experience. Some plays (taking their cue from Aristotle) privilege lexis over opsis, speech over scenery, dialogue over visual spectacle.¹¹ The film medium has passed through a technological revolution involving a shift from a visual to a verbal paradigm in the shift from silent film to the "talkies," and film theory invariably confronts some version of the image/text problem whenever it attempts to specify the nature of "film language".¹² The relative positioning of visual and verbal representation (or of sight and sound, space and time) in these mixed media is, moreover, never simply a formal issue or a question to be settled by "scientific" semiotics. The relative value, location, and the very identity of "the verbal" and "the visual" is exactly what is in question. Ben Jonson denounced the spectacular set designs of Inigo Jones as degradations of the poetic "soul" of the masque. Erwin Panofsky thought the coming of sound corrupted the pure visuality of silent movies.¹³ These are not scientific judgments, but engagements in the theoretical praxis of representation. The image/text relation in film and theatre is not a merely technical question, but a site of conflict, a nexus where political, institutional, and social antagonisms play themselves out in the materiality of representation. Artaud's emphasis on mute spectacle and Brecht's deployment of textual projections are not merely "aesthetic" innovations, but precisely motivated interventions in the semiopolitics of the stage. Even something as mundane and familiar as the relative proportion of image and text on the front page of the daily newspaper is a direct indicator of the social class of its readership. The real question to ask when confronted with these kinds of image/text relations is not "what is the difference (or similarity) between the words and the images?" but "what difference do the differences (and similarities) make?" That is, why does it matter how words and images are juxtaposed, blended, or separated?

The "matter" of the image/text conjunction matters a great deal in the work of Blake, whose illuminated books seem designed to elicit the full range of relations between visual and verbal literacy. In his illuminated books, Blake constructed image/text combinations that range from the absolutely disjunctive ("illustrations" that have no textual reference) to the absolutely synthetic identification of verbal and visual codes (marks that collapse the distinction between writing and drawing).¹⁴ It's not surprising, of course, that Blake's illuminated books, and the whole related genre of the "artist's book" would tend to exhibit flexible, experimental, and "high-tension" relations between words and images. The "normal" relations of image and word (in the illustrated newspaper or even the cartoon page) follow more traditional formulas involving the clear subordination and suturing of one medium to the other, often with a straightforward division of labor.¹⁵ In the typical comic strip, word is to image as speech (or thought) is to action and bodies. Language appears in a speech-balloon emanating from the speaker's mouth, or a thought-cloud emerging from the thinker's head. (In the pre-Cartesian world of the medieval illuminated manuscript, by contrast, speech tends to be represented as a scroll rather than a cloud or bubble, and it emanates from the gesturing hand of the speaker rather than the mouth; language seems to co-exist in the same pictive/scriptive space — hand-writing emanating from hand-gesture — instead of being depicted as a ghostly emanation from an invisible interior).¹⁶ Narrative diegesis (cp. Prince Valiant's "Our Story...") is generally located in the margins of the image, in a position understood to be "outside" the present moment of depicted action, scenes, and bodies.¹⁷

This is not to suggest that "normal" relations of word and image are uninteresting, or that vernacular composite forms like the comic strip or the journalistic photo essay are incapable of experimentation and complex deviations from the norm. Gary Trudeau's anti-cinematic, talky cartoon sequences in Doonesbury defy the normal privileging of the visual image as the place "where the action is" on the cartoon page. Doonesbury is a kind of exercise in visual deprivation, rarely showing bodies in motion, often repeating an identical and empty image (a view of the White House, the back of a television set, a point of light above a Presidential seal) in every frame, displacing all movement onto the bodiless voice indicated by the text. Postmodern cartoon novels like Maus and The Dark Knight employ a wide range of complex and self-reflexive techniques. Maus attenuates visual access to its narrative by thickening its frame story (the dialogue of a holocaust survivor and his son is conspicuously uncinematic in its emphasis on speech) and by veiling the human body at all levels of the visual narrative with the figures of animals (Jews are mice, Germans are cats, Poles are pigs).¹⁸ The Dark Knight, by contrast, is highly cinematic and televisual, employing the full repertoire of motion picture and video rhetoric while continually breaking frames and foregrounding the apparatus of visual representation.

Similar observations might be made on the mixed medium of the photographic essay. The normal structure of this kind of image/text involves the straightforward discursive or narrative suturing of the verbal and visual: texts explain, narrate, describe, label, speak for (or to) the photographs; photographs illustrate, exemplify, clarify, ground, and document the text.¹⁹ Given this conventional division of labor, it's hardly surprising that an aggressively modernist experimental deviation like James Agee and Walker Evans' Let Us Now Praise Famous Men would come into existence. Famous Men is a photo essay whose form resists all suturing of word and image: the photos are physically and symbolically separated from the text; there are also no captions on the photos, and few references to the photos in the text.²⁰ These deviations, moreover, cannot be fully accounted for by notions of "purifying" visual and verbal media. Or, more precisely, the "purist" strategies of Agee and Evans only make sense in relation to their sense of participating in a hopelessly compromised and impure representational practice, one whose political and ethical conventions need to be challenged at every level.

How do we get from the study of media in which visual-verbal relationships are unavoidable — i.e., films, plays, newspapers, cartoon strips, illustrated books — to the traditional subjects of "interartistic comparison," the analogies and differences between poems and paintings, novels and statues. In some ways it should be clear that we ought never to come back to this subject, that it is a non-subject without a real method or object of investigation. But in another sense it should be obvious that the subject of the image/text is just as unavoidable and necessary with these "unmixed" media as it is with mixed, composite forms. Interartistic comparison has always had an intuition of this fact, without really grasping its implications. The fact is that the image/text problem is not just something constructed "between" the arts, the media, or different forms of representation, but an unavoidable issue within the individual arts and media. In short, all arts are "composite" arts (both text and image); all media are mixed media, combining different codes, discursive conventions, channels, sensory and cognitive modes.

This claim may seem counter-intuitive at first glance. Surely, the objection will run, there are purely visual and verbal media, pictures without words and words without pictures. The extension of the concept of the composite image/text to unmixed forms like poetry or painting is surely a kind of figurative excess, extending a model that applies literally to mixed media beyond its proper domain.

There are several answers to this objection. The first focusses on the question of literal and figurative applications of the image/text division. It is certainly true that the division applies literally to mixed media like film, television, and illustrated books. But it is also a fact that "pure" visual representations routinely incorporate textuality in a quite literal way, insofar as writing and other arbitrary marks enter into the field of visual representation. By the same token, "pure" texts incorporate visuality quite literally the moment they are written or printed in visible form. Viewed from either side, from the standpoint of the visual or the verbal, the medium of writing deconstructs the possibility of a pure image or pure text, along with the opposition between the "literal" (letters) and the "figurative" (pictures) on which it depends. Writing, in its physical, graphic form, is an inseparable suturing of the visual and the verbal, the "image/text" incarnate.²¹

But suppose we bracket for the moment the issue of writing. Surely, the objection might continue, there are visual representations in which no writing appears, and verbal discourses (especially oral) which need never be written down. How can we deny the merely figurative status of visuality in an oral discourse, or the merely figurative status of textuality in a painting purely composed of shapes and colors, without legible, arbitrary signs? The answer is that there is no need to deny the figurative status of the image/text, only to dispute the "merely" that is appended to it. To claim that a label only applies metaphorically, notes Nelson Goodman, is not to deny that it has application, only to specify the form of application.²² Figurative labels ("blue" moods and "warm" colors) apply as firmly and consistently as literal ones, and have as much to do with actual experience. The fact that images, pictures, space, and visuality may only be figuratively conjured up in a verbal discourse does not mean that the conjuring fails to occur, or that the reader/listener "sees" nothing. The fact that verbal discourse may only be figuratively or indirectly evoked in a picture does not mean that the evocation is impotent, that the viewer "hears" or "reads" nothing in the image.²³

Perhaps the best answer to the purist who wants images that are only images and texts that are only texts is to turn the tables and examine the rhetoric of purity itself.²⁴ In painting, for instance, the notion of purity is invariably explicated as a purgation of the visual image from contamination by language and cognate or conventionally associated media: words, sounds, time, narrativity, and arbitrary "allegorical" signification are the "linguistic" or "textual" elements that must be repressed or eliminated in order for the pure, silent, illegible visuality of the visual arts to be achieved. This sort of purity, often associated with modernism and abstract painting, is both impossible and utopian, which isn't to dismiss it, but to identify it as an ideology, a complex of desire and fear, power and interest.²⁵ It is also to recognize the project of the "pure image," the unmixed medium, as a radical deviation from a norm which is understood to be impure, mixed, and composite. The purist's objection to the image/text, and to the heterogeneous picture of representation and discourse it suggests, turns out to be a moral imperative, not an empirical description. It's not that the claim that all media are mixed media is empirically wrong, but that these mixtures are bad for us, and must be resisted in the name of higher aesthetic values.

It would be easy to document a similar kind of resistance to visuality in literary discourse, in the name of a similar kind of utopian purification of language, and in response to a similar intuition about the composite, heterogeneous character of normal discourse.²⁶ My aim is not to dismiss the purists, but to redescribe the way their utopian projects invoke the metalanguage of the image/text, understood as a body of figures for the irreducible impurity and heterogeneity of media. Christian Metz demonstrated long ago that cinema cannot be reduced to the models of linguistics, that film is parole but not langue.²⁷ But suppose that language itself were not langue, that its deployment as a medium of expression and discourse inevitably resulted in its contamination by the visible? That is what it means, in my view, to approach language as a medium rather than a system, a heterogeneous field of discursive modes requiring pragmatic, dialectical description rather than a univocally coded scheme open to scientific explanation.

This de-centering of the purist's image of media has a number of practical consequences. It clearly obviates the need for comparison, which thrives on the model of clearly distinct systems linked by structural analogies and substantive differentiations. It also permits a critical openness to the actual workings of representation and discourse, their internal dialectics of form understood as pragmatic strategies within the specific institutional history of a medium. There's no compulsion (though there may be occasions) to compare paintings with texts, even if the text happens to be represented directly (or indirectly) in the painting. The starting point is to see what particular form of textuality is elicited (or repressed) by the painting, and in the name of what values. An obvious entry to the "text in painting" is the question of the title. What sort of title does the painting have, where is it located (inside, outside, or on the frame)? What is its institutional or interpretive relation to the image? Why are so many modern paintings entitled "Untitled"? Why the vigorous, explicit verbal denial of any entitlement of language in painting, the aggressive paradox of the title which denies that it is a title? What is being resisted in the name of labels, legends, and legibility?²⁸

The question of the title is a literalist's entry point to a whole series of questions about the ways that words enter pictures. How, for instance, might we sort out the differences among the following kinds of textuality in painting: a picture that represents (among other objects) a text (like an open book in a Dutch painting); a picture that has words and letters, not represented in, but inscribed on its surface, as in Chinese calligraphic landscape or the large canvases of Anselm Kiefer; a picture in the mode of classical history painting that depicts an episode from a verbal narrative, like a still from a movie or a play; a picture in which the words "speak to" or disrupt the image, occupying an ambiguous location, both in and outside the image (like Dürer's signature monograms inscribed on plaques, or Magritte's inscription in "This Is Not a Pipe"); a picture whose entire composition is designed around a linguistic "character" — a hieroglyphic or ideogram, as in the work of Paul Klee; a picture that eschews all figuration, reference, narrative, or legibility in favor of pure, unreadable visuality. The investigation of these questions doesn't begin with a search for contemporary texts that betray structural analogies in some parallel literary institution or tradition. The starting point is with language's entry into (or exit from) the pictorial field itself, a field understood as a complex medium that is always already mixed and heterogeneous, situated within institutions, histories, and discourses: the image understood, in short, as an image/text. The appropriate texts for "comparison" with the image need not be fetched from afar with historicist or systemic analogies. They are already inside the image, perhaps most deeply when they seem to be most completely absent, invisible and inaudible. With abstract painting, the appropriate texts may well be, not "literature" or "poetry," but criticism, philosophy, metaphysics — ut pictura theoria.²⁹

In a similar way, the visual representations that are appropriate to a discourse need not be imported: they are already immanent in the words, in the fabric of description, narrative "vision," represented objects and places, metaphor, formal arrangements and distinctions of textual functions, even in typography, paper, binding, or (in the case of oral performance) in the physical immediacy of voice and the speaker's body. If it is hard to keep discourse out of painting, it is equally difficult to keep visuality out of literature, though the impulse to do so is adumbrated in the topos of the blind poet, literature's answer to painting as "mute poesy." Not that the situation of literature and visual art as mutual "significant others" is purely symmetrical. It seems easier for painting to re-present and incorporate textuality in a quite literal way than for the reverse to happen. Language becomes "literally" visible in two ways: in the medium of writing, and in the utterances of gesture language, the visible language of the Deaf.³⁰

The most damaging objection to the image/text model for the analysis of either texts or images might be that, like the comparative method, it simply rearranges the deck chairs, and reiterates existing dominant paradigms of analysis in the disciplines of literature and art history. The notion that images may be read as texts is hardly news in art history these days: it is the prevailing wisdom, the latest thing.³¹ On the side of literary study, reading texts for the "imagery" is definitely not the prevailing wisdom: it is as old as the hills.³² It is seen as an outmoded paradigm, a relic of psychologistic approaches to literary experience, and of stultifying routines like motif-hunting, image-counting, and a disportionate attention to figurative and formal analysis, at the expense of real cultural history.

There is some truth to these charges. The concept of the medium (visual or verbal) as a heterogeneous field of representational practices, as an "image/text," is not recommended here for its novelty, but for its persistence as a theoretical tradition, its survival as an abiding feature of poetics, rhetoric, aesthetics, and semiotics. It is this tradition that gave us the models of interartistic comparison, and that opens the possibility of other relations between texts and visual images, and the dedisciplining of the divisions between visual and verbal culture.

It may also seem that, in my zeal to overturn the tedious historicism of the comparative method, I've jettisoned history altogether in favor of a kind of descriptive formalism. This charge is half right. This book is not a history of visual and verbal culture, but a theory. It offers the figure of the image/text as a wedge to pry open the heterogeneity of media, and of specific representations. The aim, however, is not to stop with formal description, but to ask what the function of specific forms of heterogeneity might be. Both the formal and functional questions require historical answers: they are not predetermined by any universal science of signs, and their relation to a historical "period concept" is an open question. There's no doubt that a period concept will probably include some general account of image/text relations. It's a commonplace, for instance, that English neoclassical criticism takes the analogy between poetry and painting very seriously, while Romantic poetic theory often debunks the ut pictura poesis tradition.³³

But this doesn't mean that every image/text will inevitably reveal the aesthetic norms of the period or even be describable in the language of those norms. The first place to look for the appropriate description language for analyzing the formal heterogeneity of a representation is in the representation itself, and in the institutional metalanguage — an immanent vernacular, not a transdisciplinary theory — of the medium to which it belongs. As an example, consider the genre of the "backlot" film, which reflects on the movie industry. The members of this genre are more or less self-conscious about the institutional history of the cinematic medium to which they belong; they carry a kind of institutional memory, a myth of the medium, a picture of the theory of the medium itself. Billy Wilder's Sunset Boulevard, a classic instance of the backlot film, takes a version of the image/text (the division between speech and visual representation) as its explicit theme, embodied in the relationship of a young male writer (Joe Gillis, played by William Holden) and an aging female screen idol (Norma Desmond, played by Gloria Swanson). The film allegorises a number of familiar myths of film history and theory: the two central characters represent, respectively, the New Hollywood of "talking pictures" and the Old Hollywood of silent spectacle; they also incarnate the professional tension between the (invisible) writer and the (visible) star, the split between cinema as a literary and a pictorial institution. The love-hate relationship between Joe and Norma dramatizes the ambivalence of cinema about its own past and parts; Norma's murder of Joe suggests that the dead hand of the past will never allow the institution to escape its grip; Norma's pathetic delusion of a "return" to the movies is no more misguided than Cecil B. De Mille's smug assurance to Norma (conveyed in a scene in which he plays himself on the sound-stage of The Ten Commandments) that "the movies have changed a lot since your day."

At first sight (and hearing), Sunset Boulevard is a film that seems conspicuously experimental and deviant in its formal deployment of voice and visual narration. There is a certain straightforward shock effect in framing the entire story in the voice-over narration of the dead Joe Gillis, whose first visible appearance is as an open-eyed corpse seen from below, floating face-down in Norma Desmond's swimming pool.

Once the bizarre premise of the dead narrator is accepted, however, the film settles into a straightforward and conventional suturing of voice and image. Joe's voice addresses the audience in the most ordinary of narrative contracts: he is sardonic, knowing, and sociable, counting on an audience that shares his experience and values ("it was one of those crazy Hollywood houses"; "let's go back six months to where it all started"; "well, this is where you came in"). He is like the narrator of a nineteenth century novel: "an individual who looks back, all passion spent, the narrator has mastered the world and tells a civilized company of listeners about a series of events which now can be composed and named".³⁴ The visual narrative seems invariably to illustrate Joe's voice in the most straightforward ways: when he describes something, we see it on the screen; when he narrates an action, it is performed for us; when he recalls a memory or a dream, it is projected in full. The histoire is firmly controlled by the recit, the visible by the sayable.

If the thematic image/text in Sunset Boulevard is the unshakable dominance of the visual over the verbal in the film medium, its formal image/text seems to convey precisely the reverse message: voice seems continually to dominate and control the image. The film's formal perfection is precisely not to "mirror" or imitate its theme, unless we take inversion rather than similitude as the chief symptom of mimesis. It is perhaps this perverse form of formal/thematic mirroring that produces the film's overall effect of ideological impasse and paralysis, what one might call the pathological version of the infinite relation between word and image staged by Magritte's Pipe.

The "flickering" of this impasse is most visible and readable at the outer edges of the film, its opening and closing sequences. The opening shot shows the words "Sunset Boulevard" painted on the curb of the actual street in Hollywood, the melodramatic theme music giving way to police sirens and the camera eye lifting from the curb to spot the taillights of the cars rushing to Norma Desmond's mansion where William Holden's unidentified voice refers to himself in the third person as some "poor dope" who has been found dead in a swimming pool. The closing sequence is presumably an hour or so later when Norma Desmond descends her staircase, mistaking the news cameras for signs that her fantasy of a "return" to film stardom has been fulfilled. As she says "I'm ready for my close-up, Mr. De Mille," she advances on the camera until her face fills the screen and dissolves into an illegible blur.

The outer frame of the film, then, stands outside the authority of Joe Gillis' voice: the first word enunciated by the film is an image, the written name of a place, a setting, a street which contains Joe Gillis' voice the way a haunted house contains its spirit; the last words of the film are those of Norma Desmond taking control of her own image, which is an image that defies all optical or textual control. This outer frame, with its subversion of the authority of the voice, is then redoubled by the inner frame with its revelation of the body of the narrator. The image of William Holden's floating corpse seen from the bottom of the swimming pool is without question the most memorable shot in this film. And yet it is precisely the image that has to be forgotten to allow the narrative contract of the film to proceed as if a sociable relation of teller and audience could be assumed. If the audience were to remember this image throughout the film, it would not be able to rely on the authority and security of the spoken narrative, but would continually recall its position at the scene of narration, looking up from the bottom of a pool, as drowned in images as the narrator himself.

Sunset Boulevard's reflections on the film medium may help us to see why, in Deleuze's words, "the most complete examples of the disjunction between seeing and speaking are to be found in the cinema,"³⁵ and how those disjunctions underly (indeed, require) the most seamless forms of image/text suturing. The image/text in Sunset Boulevard is specified not only by a technical account of the relation between its verbal and visual codes, but also by an account of the way it thematizes that relation in the institutional practices of the medium to which it belongs. Joe Gillis' "moral choice" in the film is portrayed as a choice between two women who have two different relations to his screenwriting. Norma Desmond as the fatal female "idol" represents the prostitution of Joe's literary talents, the commodification of his pen (not to mention his penis). Betty Schaefer, the pretty young script-reader from Paramount, is content to remain offscreen and encourage Joe's cultivation of literary authenticity. The impossibility and unreality of this choice is the moral version of the impasse presented by the film's contradictory locations of narrative authority.³⁶

The relation of speech to vision in this movie is thus mapped onto ideologemes like the relation of the sexes and generations (the aging female star versus the young male script-writer), as well as the relation of death and life, vice and virtue. It is also a figure for the past-present dialectic of the film's reflection on its medium as a historically evolving (or devolving) institution. This history would, if pursued far enough, situate the film in relation to post-war anxieties about the decline of Hollywood, and Cold War paranoia about Hollywood's susceptibility to "foreign" influences. (At one moment in the film, when Joe Gillis is seen by a friend riding in Norma Desmond's car, he is asked if he's "working for a foreign power"). In reflecting on its own medium, Sunset Boulevard provides both a description language for and a specific instantiation of the cinematic image/text. It pictures a theory of film, and narrates that theory as an account both of the death of cinema, and of cinema as a kind of love-affair with death.

I hope it's clear that the answers to the questions put by the image/text problem are both formally descriptive and historical. They try to locate in the specific formal divisions of the film text its own sense of institutional history (the cinema), and its representation of the struggles over power and value that give that history a specific form. The point of the image/text is certainly not simply to reinforce the truism that film is a mixed, divided medium, but to specify the particular sensory and semiotic mixtures and divisions that characterize particular films, specific genres and institutional practices, and to account for their function in historical terms. The image/text is neither a method nor a guarantee of historical discovery; it is more like an aperture or cleavage in representation, a place where history might slip through the cracks.

The openness of the image/text problem to historical issues may best be illustrated by looking at a place where a history has itself been constructed with self-conscious attention to the interplay of visual and verbal experience. I'm thinking of Thucydides' Peloponnesian War, which is constructed around the alternation between what Thucydides calls "logoi" and "erga," the representation of set speeches, and the eye-witness description of actions.³⁷ Thucydides insists on the distinctness of these two modes of writing, noting especially the different kind of historical authority they carry. The "precise words" of the speeches, Thucydides admits, are "difficult to remember... so my method has been, while keeping as closely as possible to the general sense of the words that were actually used, to make the speakers say what, in my opinion, was called for by each situation."³⁸ Thucydides' text presents the speeches in a form (direct transcription or quotation) that belies their actual status. We are given the illusion of reading and hearing the actual words of Pericles' funeral oration, or the Athenian ambassadors speech to the Spartans. The "reporting of events," on the other hand, is subjected to more constraints, both in the gathering of information and its presentation. Thucydides has "made it a principle not to write down the first story" that comes his way, "and not even to be guided by [his] own general impressions."(48) He claims to have "checked with as much thoroughness as possible" the various eye-witness reports (including his own), and admits that this skepticism will eliminate "the romantic element" of his narrative, both in the facts it admits, and in its hesitant form of presentation ("my history will seem less easy to read").

The Peloponnesian War, then, is a radically heterogeneous representation of history, alternating between the illusion of oral immediacy and direct transmission of speech, and the presentation of visual experience (eye-witness testimony about events) as highly mediated and unreliable. The "general sense" of words is enough for Thucydides, but his own "general impressions" of events in visual memory are not. The question, of course, is "so what?" Why does Thucydides construct a dialectical representation of history? Where do his assumptions about the relative accuracy of oral and visual memory come from? (It's hard to ignore the fact that they seem precisely counter to modern "rules of evidence" which privilege "eye-witness testimony," and denigrate "hearsay").

A proper answer to these questions would take us well beyond the limits of this essay. It would involve a reconstruction of Thucydides' theory of memory and its relation to the senses of vision and hearing. It would require attention to the links between the art of memory and contemporary Greek rhetorical theory. Above all, it would lead to an inquiry into the ideological and institutional status of "speech" and (visible) "action" in Athenian political culture during the Periclean age.

The verbal/visual division is, as Thucydides himself makes clear, not merely a rhetorical feature of historical writing, but a key to the way history itself is made as a dialectic between "what men did" and "what they said." Thucydides shows us cities being conquered with speeches. He also shows us (in Pericles' funeral oration) a polemic against speech-making and an argument in favor of ceremonial spectacle:

Many of those who have spoken here in the past have praised the institution of this speech at the close of our ceremony. It seemed to them a mark of honour to our soldiers who have fallen in war that a speech should be made over them. I do not agree. These men have shown themselves valiant in action, and it would be enough, I think, for their glories to be proclaimed in action, as you have just seen it done at this funeral organized by the state. (144)

The adjustment of the dialectic between speech and visible action is, in Pericles' view, not just a question about the proprieties of ceremonial occasions, but the key issue in the healthy functioning of Athenian democracy: "we do not think that there is an incompatibility between words and deeds; the worst thing is to rush into action before the consequences have been properly debated" (147). The tragedy of

Athenian democracy might well be described as the breakdown of this logos/erga dialectic in the emergence of demagogues who corrupt the logos, and the greed of the Athenians who crave instant, visible action and unlimited expansion of the empire. Thucydides' text is an attempt to preserve this fragile dialectic as a way of knowing and showing history, and perhaps of making it take a different form. (Pericles' critique of the visible "marks and monuments of our empire," his insistence that "famous men" leave their memorials, "not in any visible form but in people's hearts"[149] is an attempt to resist the craze for spectacular memorials that he himself did so much to foster, and which remain as the chief visible sign of Athenian glory). For Thucydides, the fullness of history doesn't lie either in what men did or what they said, what we could see and describe or what we could hear. It seeps through the cracks between hearing and seeing, speaking and acting.

I've juxtaposed these historically and culturally remote examples (a Hollywood film and a classical history), not in order to "compare" them, but to call attention to the extraordinary generality of the image/text as a figure for the heterogeneity of representation and discourse. I hope it is clear that this sort of generality does not entail any lack of attention to material, formal, and historical specificity. I have not forgotten that The Peloponnesian War is nothing but words, and Sunset Boulevard is nothing but traces on celluloid. The image/text is not a template to reduce these things to the same form, but a lever to pry them open. It might be best described, not as a concept, but as a theoretical figure rather like Derrida's differance, a site of dialectical tension, slippage, and transformation. Sunset Boulevard and The Peloponnesian War employ versions of this figure to reflect on their own heterogeneity, and to connect their formal dialectics to ideological and institutional struggles within their own media (cinema, history), and to the cultural contradictions they mediate. They are, as it were, metapictures of their media.

Why are these metapictures coming into focus for us now? Why does the image/text seem to crop up in contemporary criticism and theory like an historical a priori? One answer would appeal to the historical situation of contemporary culture, the increasing mediatization of reality in postmodernism, and the phenomenon I have

called "the pictorial turn," with its regimes of spectacle and surveillance. Our readings of ancient texts and images cannot help but be inflected by our experiences with television and cinema. The claim that all media are mixed media, all arts composite arts, may actually sound like common sense to a generation raised on MTV. Another answer, however, would stress that the purification of the media in modernist aesthetics, the attempt to grasp the unitary, homogeneous essences of painting, photography, sculpture, poetry, etc., is the real aberration, and that the heterogeneous character of media was well understood in pre-modern cultures. The essentializing of the media was reinforced by the emergence of professional disciplines and the academic administration of knowledge. "Interartistic comparison" was, in its best moments, a way of resisting this compartmentalization; at its worst, it collaborated in the insularity of the disciplines and the amateurishness of interdisciplinary efforts. That is why it is time to get beyond comparison, and immerse ourselves in the dialectics of representation.

Footnotes

¹ Gilles Deleuze, Foucault (Minneapolis: University of Minnesota Press, 1988), 60: "Speaking and seeing, or rather statements and visibilities, are pure Elements, a priori conditions under which all ideas are formulated and behavior displayed, at some moment or other."

² Traditional approaches to the comparative method include Wylie Sypher's, Rococo to Cubism in Art and Literature (New York: Random House, 1960), Mario Praz Mnemosyne: The Parallel between Literature and the Visual Arts (Princeton, NJ: Princeton University Press, 1970); and Chauncey Brewster Tinker, Painter and Poet: Studies in the Literary Relations of English Painting (1938, reprint: Freeport, NY: Books for Libraries Press, 1969). Rene Wellek's skeptical critique of the comparative method in his Theory of Literature (Orlando, FL: Harcourt, Brace Jovanovich, 1964, rev. ed.) remains a sensible antidote, but it does not attempt any analysis of the basic intuitions and motives that drove interartistic comparison in the first place. Wendy Steiner's, The Colors of Rhetoric (Chicago: University of Chicago Press, 1982), is the best example of the application of structuralist and semiotic methods to interartistic comparison in modern literature and visual art. See also Göran Sonneson's Pictorial Concepts (Sweden: University of Lund, 1989), discussed below. For a good survey of recent "interart studies," see the special issue "Art and Literature," edited by Wendy Steiner, of Poetics Today 10:1 (Spring, 1989).

³ Jean Hagstrum, The Sister Arts (Chicago: University of Chicago Press, 1958).

⁴ Donald Preziosi, Rethinking Art History (New Haven: Yale University Press, 1989), 9.

⁵ The Colors of Rhetoric, 19.

⁶ The Colors of Rhetoric, 18.

⁷ See Sonneson, op. cit., 19, 16. The situation with the semiotics of cinema and mass culture in the tradition of Roland Barthes is, for the most part, quite different. There a rich body of connections between semiotic structures and ideology has been developed, especially in feminist and psychoanalytic criticism. It would be impossible to do justice to this work in a single footnote. One of my aims is to open up the traditional problems of interartistic comparison to some of the redescriptions made possible by this more self-critical and nonscientistic tradition of semiotic analysis. As a starting point, one might consult the excellent anthology On Signs, ed. Marshall Blonsky (Baltimore, Md.: Johns Hopkins University Press, 1985).

⁸ For a case in point, see the important article by Mieke Bal and Norman Bryson on "Semiotics and Art History," Art Bulletin 73:2 (June, 1991), 174-208. Bal and Bryson argue that semiotics "challenges the positivist view of knowledge" endemic to art history with an "antirealist" theory of the sign (174). Bal and Bryson seem committed, moreover, to a notion of the semiotic challenge as fundamentally political, allowing questions of gender and power, for instance, to become central to the study of the visual image. Yet they also tend to treat semiotics as a neutral, scientific metalanguage: "since semiotics is fundamentally a transdisciplinary theory, it helps to avoid the bias of privileging language that so often accompanies attempts to make disciplines interact"(175). As should be clear by now, I'm skeptical about the possibility of transdisciplinarity and the avoidance of "bias."

⁹ I will employ the typographic convention of the slash to designate "image/text" as a problematic gap, cleavage, or rupture in representation. The term "imagetext" designates composite, synthetic works (or concepts) that combine image and text. "Image-text" with a hyphen designates relations of the visual and verbal. The necessity of a concept such as the "imagetext" was first made clear to me by Robert Nelson in our team-taught seminar on "Image and Text."

¹⁰ See discussion of Foucault on Magritte in my essay, "Metapictures," chapter two of Picture Theory, (Chicago: University of Chicago Press, 1994).

¹¹ Aristotle's repudiation of opsis is so thorough that he is even willing to jettison performance itself, in favor of a narrative presentation of the action. "The plot should be so constructed that even without seeing the play anyone hearing of the incidents thrills with fear and pity as a result of what occurs... To produce this effect by means of an appeal to the eye is inartistic and needs adventitious aid..." (Poetics, XIV, translated by W. Hamilton Fyfe, 1927; [Cambridge, MA: Harvard University Press, 1973], p. 49).

¹² See Christian Metz, Film Language: A Semiotics of the Cinema tr. Michael Taylor (New York: Oxford University Press, 1974), and footnote 27 below.

¹³ Ben Jonson, "An Expostulation with Inigo Jones," in The Complete Poems, edited by George Parfitt (New Haven, CT: Yale University Press, 1975), 345-47; cp. Erwin Panofsky on space and time in cinema in "Style and Medium in the Motion Pictures Critique 1:3 (Jan.-Feb. 1947), reprinted in Gerald Mast and Marshall Cohen, eds., Film Theory and Criticism* (New York: Oxford University Press, 1978), and my essay, "Visible Language," Picture Theory chapter four.

¹⁴ See Blake's Composite Art (Princeton University Press, 1971) and "Visible Language" in Picture Theory.

¹⁵ I am adapting here the concept of "suture" as developed in psychoanalytic film theory, and am relying principally on Stephen Heath's article, "On Suture," in Questions of Cinema (Bloomington: Indiana University Press, 1981), 76-112. Suture might be described most generally in Lacan's words as the "junction of the imaginary and the symbolic," (quoted in Heath, 86), the process by which the subject (the "I") is constituted both as a division and a unity. "I," as Heath notes, is "the very index of suture." Film theory adapted the notion of suture to describe the construction of the spectator position in cinema (the "I/eye" as it were), and to analyze the specific characteristics of film discourse. Suture might be described as that which "fills in" the gaps between images and shots by constructing a subjective sense of continuity and absent positionality. Shot-reverse-shot, with its interplay of shifting spectatorial positions and self/other identifications, is thus the paradigmatic figure for suture in cinema. My adaptation of this notion to the "imagetext" is very rough and preliminary, but not, I hope completely unfounded. At the root of the idea in both psychoanalysis and film theory is the figure of a heterogeneous field of (self)representation, and the process by which its disjunctions are at once concealed and revealed. The specific form of that heterogeneity (for Lacan, the imaginary and the symbolic, for film theory, the transformation of the sequence of film images into a discourse) is already very close in its formulation to the problem of the imagetext. Film theory's emphasis, not surprisingly, has been on

¹⁶ There's nothing to prevent the re-arrangement of these conventional spaces, of course. Dialogue need not appear in voice balloons, but can simply run along the bottom edge of the frame (the normal convention of the single panel cartoon, and of multiple-panel narrative cartoonists like Jules Pfeiffer). For a very useful insider's account of rhetorical and narrative devices in cartoons, see Will Eisner, Comics & Sequential Art (Tamarac, Florida: Poorhouse Press, 1985).

¹⁷ See Art Spiegelman, Maus I: A Survivor's Tale: My Father Bleeds History (New York: Pantheon, 1973). The effect of Spiegelman's brilliant animal caricatures is, of course, more complex than a simple "veiling" of the human form. The reduction of the Holocaust to a "Tom and Jerry" iconography is at once shocking in its violation of decorum, and absolutely right in its revelation of a fitness and figural realism in the animal imagery. Spiegelman's images insist that the Jews really were "scared mice"; they really were treated as vermin to be hunted down and exterminated; they hid in cellars like and with rats; some of them were traitors and rats, notably when they organized themselves into the collaborationist institution of the "Judenrat" (Jewish Council). The Germans really are heartless predators, but feline and unpredictable, luxuriating in the pleasure of absolute power over abject victims. The Poles really are pigs, fattening themselves on the dispossessed goods of the Jews, sometimes goodhearted, but inevitably gross and unrefined. The hyberbole of the animal imagery enforces a mode of critical realism, while defending the viewer against (or preventing) an unbearable (or voyeuristic) access to the banal human forms of evil and abjection. For more on abjection and visual defense, see my essay, "Narrative, Memory, and Slavery," Picture Theory, chapter six.

¹⁸ See Jefferson Hunter, Image and Word: The Interaction of Twentieth-Century Photographs and Texts (Cambridge: Harvard University Press, 1987), and my essay, "The Ethics of Form in the Photographic Essay," After Image, January, 1989.

¹⁹ See "The Ethics of Form in the Photographic Essay," op. cit., for an extended discussion of Let Us Now Praise Famous Men and its relation to the genre of the photographic essay.

²⁰ I take the equivocal status of writing as image/text to be one of the leading themes of Jacques Derrida's Of Grammatology. See "Visible Language" in Picture Theory for an extended discussion.

²¹ See Languages of Art (Indianapolis: Hackett, 1976), 68-69. See also Derrida's discussion of "literal" and "figurative" senses of writing in Of Grammatology, 15: "the 'literal' meaning of writing as metaphoricity itself."

²² On the psychology of visual response in reading, see Ellen Esrock, The Reader's Eye (Baltimore: Johns Hopkins University Press, 1994). Esrock is the first to connect the "senses" (semantic or cognitive) in which readers might be said to visualize with a thorough survey of the psychological literature on the subject.

²³ The classic argument for the purification of poetry and painting, and the strict segregation of them according to visual/verbal and spatial/temporal categories, is, of course, Lessing's Laocoon (1766). For a critique of Lessing, see chapter four of my Iconology.

²⁴ The most notable Modernist redaction of the "purity" argument is Clement Greenberg's 1940 essay, "Toward a Newer Laocoon," with its self-conscious evocation of Lessing.

²⁵ See my essay, "Tableau and Taboo: The Resistance to Vision in Literary Discourse," CEA Critic 51:1 (Fall, 1988), 4-10.

²⁶ See Metz, Film Language: "When approaching cinema from the linguistic point of view, it is difficult to avoid shuttling back and forth between two positions: the cinema as a language; the cinema as infinitely different from verbal language. Perhaps it is impossible to extricate oneself from this dilemma with impurity" (44). We need not accept Metz's ultimate valorization of narrative as the essence of cinema to see that film can never be explicated fully as a systematic language.

²⁷ On titles, see Gerard Genette, "Structure and Function of the Title in Literature." Critical Inquiry 14:4 (Summer, 1988), 692-720.

²⁸ See my essay, "Ut Pictura Theoria: Abstract Painting and Language", Picture Theory, chapter seven.

²⁹ For more on gesture language, see my essay, "Ekphrasis and the Other," chapter five of Picture Theory.

³⁰ See Mieke Bal and Norman Bryson on "Semiotics and Art History," op. cit., which argues that the key move in semiotic versions of art history is the treatment of the image as a sign or "visual text"(179): "Considering images as signs, semiotics sheds a particular light on them, focusing on the production of meaning in society..."(176). Although Bal and Bryson insist that they are proposing "a semiotic turn for art history" rather than "a linguistic turn," they underestimate, in my view, the extent to which semiotics privileges textual/linguistic descriptive frameworks. Far from avoiding "the bias of privileging language," semiotics continually reinstates that bias.

³¹ See Iconology, chapter one, for a discussion of the attacks on the notion of literary imagery.

³² Roy Park, "Ut Pictura Poesis: The Nineteenth Century Aftermath," Journal of Aesthetics and Art Criticism 28 (Winter, 1969), ISS-69. For a more complex account of this aftermath, see Elizabeth Abel, 'Redefining the Sister Arts: Baudelaire's Response to the Art of Delacroix,' Critical Inquiry 6:3 (Spring, 1980), 363-84.

³³ Jonathan Culler, Structuralist Poetics (Ithaca, N.Y.: Cornell University Press, 1975). 195.

³⁴ Foucault, 64. See also Deleuze's comprehensive discussion of the relation between speech and vision, sound- and image-track in "The Components of the Image," in his Cinema 2: The Time-Image tr. Hugh Tomlinson and Robert Galeta (Minneapolis: University of Minnesota Press, 1989), 261.

³⁵ See my article, "Going Too Far with the Sister Arts," in James Heffernan, ed. Space. Time. Image. Sign: Essays on Literature and the Visual Arts (New York: Peter Lang, 1987), 110. For further discussion of Sunset Boulevard's knowing allusions to the ut pictura poesis tradition, and its connection with allegories of the "Choice of Hercules".

³⁶ For a good discussion of the role of logoi and erga, see Virginia J. Hunter, Thucydides: The Artful Reporter (Toronto: Hakkert, 1973).

³⁷ Thucydides, The Peloponnesian War, trans. by Rex Warner (New York: Penguin Books, 1954), p. 47; further page references cited in the text.