INTRODUCTION

Scope and Aims of the Work

This study focuses on the transformations of textual strategies and devices in a selection of literary works from Homer to Milton. The works to be looked at are all of “epic proportion,” in the sense that they try to produce some sort of master narrative about serious matters. They are also epics in the sense that they share numerous characteristics and devices traditionally identified with the historical genre of epic. One of these devices, the epic simile, will serve as the main signpost for the investigation of the various works. Unlike other studies of the epic simile, however, the focus here will be on the semiotics of this traditional device, on how texts signify rather than on what they signify. The analysis of the similes will be a sort of symptomology opening out into more general issues of textual production in different historical contexts. The purpose is twofold: to bring to bear on these individual narratives the insights and preoccupations of semiotics; and, at the same time, to make a contribution to semiotics by bringing to bear on it the singularity of these various texts. That is, what follows is not only an application of semiotics to literary works, but also an application of literary works to our current understanding of signifying practices. For this reason, the theoretical framework for the textual analyses will itself be evolved in the course of making those analyses.

I begin by placing the problem of Homer’s similes in the context of reading in general. I then outline a theoretical framework and preliminary terminology for the whole work, including a précis of Umberto Eco’s theory of codes and of Michael Riffaterre’s theory of intertextuality. The work of Eco and Riffaterre will form the basis for discussion of more specific theoretical issues to be taken up as they arise. I then turn to Homer’s Iliad, for us the beginning of the epic tradition in the west. In approaching Homer’s similes, I emphasize their productive character rather than their representational function. The similes of the Iliad, I argue, perform a whole range of roles in propelling the narrative of the poem forward. In fact, they often turn out to be solutions to specific textual complexities and are thus privileged places to see Homer’s poetics at work. As such, Homer’s similes go beyond the ornamental role usually attributed to them and can serve as an index of the textual and thematic problems that the poem tries to overcome.

Subsequent chapters take up the reorientation and redeployment of the simile in the Argonautica of Apollonius of Rhodes, the Aeneid of Vergil, Dante’s Divine Comedy, and Milton’s Paradise Lost, with discussion of other texts wherever that seemed necessary to clarify a context. These texts hardly exhaust the possibilities of discussing the epic simile, but they do make up a series with a rich range of similarities and differences, and they all stand in a tradition of conscious imitation of predecessors which begins with Homer. A close structural and theoretical reading of the similes of these texts will provide a means of explaining how each text relates to its textual tradition and how changing cultural contexts produce different ways of conceiving and constructing meaning. These transformations in the process of meaning formation are traced as the culture in which the epic occurs changes from an integrated warrior culture, to the more heterogeneous and widely diffused culture of the Hellenistic era, to the Roman context, with its universal pretensions, to late medieval Christian culture and finally to a culture which can no longer sustain the epic.

Reading Homer

Since Hellenistic times the epic similes of Homer have been viewed basically as ornamental, for like many aspects of the Iliad and Odyssey, they often seemed to critics of the post-Aristotelian era to add little to the advance of the narrative itself.¹ These critics had the word of Aristotle (AP 1459a) that an epic should present a unified action, with all its parts correlated like a living organism. Indeed, the meandering similes of Homer, often scandalously inapposite to the narrative situation they ostensibly elucidate, have presented Homeric scholars with the same problem raised by the texts as a whole: namely, that they seem to be diffuse, loosely constructed and full of digressions and illogic. Given their presuppositions about what constitutes a proper epic, ancient commentators on Homer divided their activities between, on the one hand, emending, excising and otherwise reconstructing the poems to fit their own expectations for a work of art; and on the other, seeking an underlying unity in the form of allegorical interpretations.² The activity of nineteenth-century critics showed similar concerns: the “analysts” attempted to reconstitute the original text of Homer from the lamentably incoherent one transmitted to us; the “unitarians” to show that the Iliad and Odyssey were indeed unified in their received form if one considered them with sufficient subtlety.³ It is well known how the work of Milman Parry radically changed the terms of this controversy. Parry showed that Homeric diction was a traditional medium and that the inclusion of words and phrases from different dialects and different historical strata was the result of a long process of development ruled by the constraints of “economy of expression.” He concluded that the language of Homer was that of an oral tradition and that the use of noun-epithet combinations exhibited a traditional and unified composition technique.

A great deal of subsequent scholarship has attempted to enlarge on Parry’s findings, assuming in varying degrees that the Iliad and Odyssey cannot be properly understood in terms of the Aristotelian notion of a “well-made story.” Rather, the diffuse and loosely constructed poetic surface of Homer results from the circumstance that oral poetic production is governed by the logic of traditional patterns which underlie that surface.⁴ Numerous studies have tried to identify these underlying patterns in the form of “typical scenes,” usually coupled with the purpose of reconciling narrative inconsistency with unity of authorship by showing the underlying compositional habits of the poet. Thus, Hansen’s work on the “conference sequence” answers the objections of Kirk and Page to certain irrelevancies in the plot of the Odyssey by showing a deeper regularity in the poet’s composition of such scenes.⁵ Fenik challenges not only analyst claims of interpolation, but also the “neo-analytic” position that some motifs of the Iliad must be borrowings from antehomerica, by showing the typicality of most of those motifs throughout the Iliad.⁶ Again, W. C. Scott counters the thesis of Lee and Shipp that many of the similes of the Iliad are late interpolations by showing underlying regularities which characterize their use throughout the poem.⁷

The object of these studies, however, is limited to showing that the poems are the product of a single poet, a single tradition or a single oral poet. Hence the criteria consistently invoked are compositional economy and the autonomy of narrative patterns. What seems to be an inconsistency in the narrative surface is resolved once one has grasped the underlying motivation. The exigencies of oral composition and the pressure exerted by compositional habits combine to account for the inclusion of elements inappropriate to a specific context as well as other types of apparent anomaly. These studies, therefore, tend to treat the problem of Homeric composition negatively: compositional economy and the autonomy of narrative patterns, after all, provide apologies for the poet’s petty failings rather than than explanations of his poetic purpose. “It has gradually become obvious,” writes Nagler, “that even if one could view oral-traditional language as embodying a repertoire of ‘stock formulas’ this would offer no automatic solution to the problem of denotative, to say nothing of poetic meanings in Homeric diction...[and] the problem has been the same on the level of the motif.”⁸

In response to this problem Nagler proposed a generative approach to Homer’s language, replacing the notion of formula with a more open-ended “family” of phrases and words related to each other by poetic signification. Thus, for example, Nagler suggests that the explicit mention of attendants following a character conveyed to the properly cued audience significant information about the status of that individual. But various related accoutrements and activities (veils, ascending and descending, chambers, etc.), as well as certain key syntagms (καὶ ἁμὰ τῷ, ἁμ’ ἕπονται, etc.) can refer to the same nuclear event of attendance. Moreover, the related accoutrements and activities can be realized by a number of “allomorphs” (e.g., “wave” for “veil”), all of which adds up to an extensive family of associated diction referring to the nuclear event of attendance with its poetic signification of conferring τίμη on the principal. The various realizations of this dictional family are generated, according to Nagler, by a “pre-verbal template” or “Gestalt”; that is, each verbal realization is a derivative “not of any other phrase but of some pre-verbal, mental, but quite real entity underlying all such phrases at a more abstract level” (Spontaneity, 12).

For the nature of these pre-verbal templates, Nagler adopts the framework of certain Sanskrit grammarians whose theory of meaning revolved around the concept of sphota (Spontaneity, 13-14):

The term, sphota, is derived from an onomatopoetic root sput “to burst,” its application to language being the intuitive perception of meaning which in our idiom also “bursts” upon the mind in some unknown way either spontaneously or when triggered by a linguistic symbol: word, phrase of sentence. The concept of sphota was defined from two points of view, both of which are useful for my purposes: (a) “that from which the meaning bursts forth,” “the linguistic sign in its aspect of meaning bearer,” and (b) as an entity which itself is (at least partially) manifested by speech. But any one Gestalt or sphota beggars definition, for it is itself undifferentiated with respect to any describable phonological feature. The given word, phrase or sentence is only a kind of hypostasis of this entity--an allomorph, as I have been using the term--as a particular geometric shape is a hypostasis of its Platonic Form.

Between this “timeless, undefinable” sphota and the surface structures of actual utterance lies an intermediate stage consisting of patterns of phonological and syntactic norms imperfectly revealed by individual speech acts. Nagler analyzes Homer’s text by abstracting from a number of examples a pattern of syntactic and semantic archetypes. These ideal configurations make up a sort of grammar by which the underlying sphota is transformed into verbal utterance. For Nagler this process of transformation into poetic utterance is hierarchical: from these pre-verbal templates to textual surface, from an inherent “meaning,” which is ontologically prior to linguistic expression, to individual utterances which manifest, darkly, the presence of meaning.

The crucial issue raised by Nagler’s theory is whether meaning is immanent or produced: whether there is a single meaning in all languages or whether meaning is context dependent. The Indian model adopted by Nagler is a radical example of a “universalist” theory of language, in which every utterance derives its meaning by virtue of being a manifestation of some underlying Gestalt which is ontologically prior to its articulation.⁹ The sphota is “eternal and partless, manifested in time but not affected by time” (Spontaneity, 16); it is located so deep in consciousness “that one cannot properly speak of it as pertaining to, or affected by, a particular tradition” (Spontaneity, 26). Thus, for example, a child learning a language will develop an “intuitive feel” for an underlying Gestalt by hearing a number of its manifestations (Spontaneity, 17). The process of language acquisition is therefore a matter of giving particular form to a predetermined content, Widererzeugung to use Humboldt’s term.¹⁰ The implication of this model is that a text such as the Iliad articulates the same basic meaning articulated by all texts, variations occurring only at a number of more superficial levels of linguistic organization. Thus Nagler feels free to corroborate the identity of a Gestalt from any number of other cultural traditions (Near Eastern texts, Beowulf, etc.) for it is basically the same for all people in all times.

Nagler’s explicitly Platonic theory is a straightforward example of an “essentialist” theory of meaning. Nagler’s theory takes to its logical conclusion the tendency to project from various regularities in a text an ontologically prior generative mechanism, whether one is speaking of formulae, type scenes or mythic paradigms. It is worth giving some scope to his theory because Nagler lays out the assumptions which are usually tacit in essentialist interpretive strategies: namely, the undialectical separation of form and content so that language becomes treated as an instrument which we merely take up and use in order to convey information or feelings which existed prior to the act of communication. Interpretive strategies based on such a notion of language focus on finding meaning “in” the text or “under” the text, and generally end up minimizing the historical determinants of the process of meaning production and conferring some more abstract and ahistorical significance on works of literature. Thus, for example, James Redfield, Cedric Whitman and Nagler all end their studies of the Iliad with a version of Achilles as an “existential” hero who experiences the “true meaning of life.” Such readings, however, are based on an ideology of language which views literature as embodiments of eternal verities still accessible to us today as such.¹¹

To say that a text “has” a meaning is the hallmark of an essentialist theory of language. Meaning, as it will be defined below, must be meaning to someone in some place at some time in some context. This is not to say that essentialist readings such as Nagler’s are useless, but that they are critically incomplete and must be supplemented by an inquiry which takes seriously a fuller range of factors of meaning production. Although clearly the “universal limitation of death”¹² has something to do with the Iliad, to foreground such matters cannot but reduce the poem to a repetition of our own most general anxieties. In the following section, I will outline Umberto Eco’s theory of codes, which deals with the problem of signification within a communicational framework. Having dealt with the notion of the “essence” of meaning, it will be necessary to turn to the problem of the “reader.” Then, in the final section of this introduction, it will be possible to outline an “intertextual” approach to meaning production.

Eco’s Theory of Codes

Semiotics is the unified study of all phenomena of signification and/or communication.¹³ The need for such a discipline became explicit in the last century when the term “language” became extended to include a number of non-verbal communication systems (photography, film, painting, etc.) and to a number of “languages within language” (the language of myth, of dreams, of narrative codes, etc.). The idea that verbal language should serve as the model for the study of all types of signification systems led to a number of false expectations and far-fetched metaphors in these areas of research. It therefore seemed desirable to establish a more broadly based study of signifying systems without assuming that verbal language would be the privileged model. One of the most important and thorough responses to this imperative is Umberto Eco’s A Theory of Semiotics (Bloomington, 1976). In his treatment of semiotic phenomena, Eco centralizes the notion of code. Communication occurs when a system of signification exists, and such a system relies on the functioning of a code for the production of an interpretive response in an addressee. Eco distinguishes four types of code (Eco, 36-37):

(a) “A set of signals ruled by internal combinatory laws.” The signals can have many forms: gestures, graphic symbols, sounds, etc. They can be considered as a “pure combinational structure...an interplay of empty positions.” It is a syntactic system.

(b) “A set of possible communicative contents” comprised of “notions” about something, able to be conveyed by a variety of signals and thus independent of any given system of signals. This is a semantic system.

(c) “A set of possible behavioral responses on the part of the destination.” This is a pragmatic system, and most behaviorists define meaning in terms of such responses.

(d) “A rule coupling some items from the (a) system with some from the (b) or (c) system.”

Only this last is properly a code, and Eco calls the other three system-codes, or s-codes for short. The function of a code, then, is to apportion the elements of a conveying system, an s-code of type (a), to the elements of a conveyed system, an s-code of type (b) or (c). The result of such an apportioning is a sign, or a sign-function as Eco calls it to emphasize its operational status (Eco, 49):

Signs are the provisional result of coding rules which establish transitory correlations of elements, each of these elements being able to enter--under given coded circumstances--into another correlation and thus form a new sign.

Take for instance the expression /plane/: the English language provides many content items for it, i.e. “carpentry tool” or “level” or “aircraft.” In this sense we are faced with three sign-functions: plane=X, plane=Y, plane=K.

Such a conception of codes and their operation emphasizes the conventional nature of signification and communication. Not only are the coupling rules culturally defined, but the s-codes themselves are cultural constructs and eminently open to change. The s-codes are systems or structures

that can also subsist independently of any sort of significant or communicative purpose, and as such may be studied by information theory or by various types of generative grammar. They are made up of finite sets of elements oppositionally structured and governed by combinational rules that can generate both finite and infinite strings or chains of these elements (38).

This structuredness allows s-codes to enter into correlation with one another and is a necessary condition for their semiotic functioning. This is so not only for syntactic systems, like algebraic or phonological systems, but also for semantic systems, which must be thought of as cultural constructs with the same ontological status as any other system-code.

The “content” of an expression must be distinguished from its possible referents (Eco, 58-68). The content of an expression such as the word /dog/ is not this or that particular dog, but a culturally defined class of objects which can be conveniently labeled a “cultural unit.” A cultural unit is anything culturally defined and distinguished as an entity, but it is defined

inasmuch as it is placed in a system of other cultural units which are opposed to it and circumscribe it. A cultural unit “exists” and is recognized insofar as there exists another one which is opposed to it. It is the relationship between the various terms of cultural units which subtracts from each of the terms what is conveyed by the others (Eco, 73).

If then the phenomenal world can be thought of as an undifferentiated continuum of raw data, it is culture which segments that continuum and assigns value to the discrete units by placing them in opposition to each other. The value of the various units can thus be thought of as differential. Taxonomic classification systems are the clearest examples of this process, since they define their objects in terms of specific and generic differences. But the same mechanism operates in any semantic system; any definition according to “essential” qualities is in reality a definition according to differences.

A sign is thus the result of a correlation of elements from two systems of differences. Human language, however, is much too complex to be thought of in terms of a series of binary correlations. Language is best viewed as a number of systems and subsystems, all intricately related by a vast complex of codes. The theory of codes makes it possible to rethink certain notions about language, such as the distinction between connotation and denotation. Although both denotation and connotation rely on codes, a connotative code can be said to rely on another code. Thus, for example, the word /dog/ is correlated by one code to the cultural unit canis familiaris. But that cultural unit, along with its expression unit, is further correlated with the cultural unit “fidelity.” Such a “super-elevation of codes” can be extremely complex and can take place among units of any magnitude. Thus a pastoral setting can be correlated by a connotative code with “setting for love” or “blissful state.” Such connotative codes can be conveniently called “subcodes” or “secondary codes” since they are perceived as being dependent on a prior code. But the distinction between codes and subcodes, denotation and connotation, is not an ontological one, since from a semiotic perspective it is impossible to make an appeal to the referent for such a distinction; that is, canis familiaris is not a more real content than “fidelity.” Out of a complex of interdependent codes, however, a culture will inevitably specify one as more real, “natural” and transparent than others. Denotative codes are established by culture and recognized as such by a semiotics of the code. Any given expression will usually convey a number of intertwined contents and will thus not be a simple message, but a text whose content is a multileveled discourse.

The conception of the semantic plane as a system of differences allows us to define meaning without any reference to “objects” or Platonic entities. And although a system-code can be an object of study as such, it is of interest to semiotics only in terms of its dialectical relationship with another system-code and a set of coding rules which constitutes them into semantic plane and expression plane (Eco, 40-46). Thus, a cultural unit can be thought of as the focal point of a system of signs. Various coding correlations can put a cultural unit into relationship with any number of semantic axes. So, for example, the oppositions dark vs. bright, bad vs. good, stupid vs. smart, are interconnected by such expressions as “bright student,” “dark purposes,” etc. A cultural unit can be analyzed into more elementary components based on the various readings it has in different contexts. These elementary components are called semantic markers, or semic components or semes.¹⁴ Whatever their nomenclature, they must be thought of as differential--that is, they are values issuing from systematic oppositions. A cultural unit is a nodal point arising from a series of criss-crossings of numerous oppositional axes.

A cultural unit conceived of as the inventory of a set of semantic markers is called a “sememe.” The format of a sememe would be, Eco notes, that of an encyclopaedia, specifying all the coding information necessary to establish the correct “reading” in a given circumstance. Since every semantic marker, moreover, is itself a sememe to be explained by its own componential analysis, a complete semantic structure of any but the most simple units would be virtually infinite and hence must remain a “regulative hypothesis” (Eco, 128-29):

There must be a methodological principle of semantic research whereby, in almost all cases, the description of fields and semantic branches can only be achieved when studying the conditions of signification of a given message.... A semiotics of the code can be established--if only partially--when the existence of a message postulates it as an explanatory condition.

This practical limitation will be frustrating to the linguist trying to generalize about semantic facts--particularly if he seeks to identify a limited number of semantic universals, as Chomsky proposed.¹⁵ But for the literary critic, the transitory, unstable nature of the semantic system, the fact that it can be manipulated and changed by new messages which propose unforeseen coding correlations, all this is of utmost interest.

The workings of the process of accommodating new messages Eco explains in part by developing Peirce’s notion of the interpretant. Peirce defined the term as the idea to which a sign gives rise in an addressee.¹⁶ Thus it is some sort of equivalence by which an addressee interprets a sign. An interpretant can be a definition, an emotive response, a synonym or a translation into another semiotic system. Eco defines it simply as “another representation which is referred to the same ‘object.’ In other words, in order to establish what the interpretant of a sign is, it is necessary to name it by means of another interpretant to be named by another sign and so on” (Eco, 68). The notion of the interpretant shows again the inherent circularity of the communicative process, which “by means of continual shiftings which refer a sign back to another sign or string of signs circumscribes cultural units in an asymptotic fashion, without ever allowing one to touch them directly, but making them accessible through other units” (Eco, 71). Given a particular message, an addressee will approach it by proposing various interpretants. It is by this process of circumscribing a new message with a sequence of interpretants that an addressee constitutes a meaning for that message.

The theory of codes outlined here is but a small part of Eco’s semiotic theory. This and other parts of Eco’s work will be expanded later in suitable contexts. For the time it will be sufficient to review the basic points and terminology introduced so far. A code is a set of rules correlating elements of a conveying system with elements of a conveyed system. Considered separately, the conveying system and conveyed system are structures of relationships among a series of elements. These structures Eco calls s-codes, and their elements have value in terms of their opposition to other elements in the same system. A correlation of elements from two s-codes is a sign or sign-function. The elements of a semantic system are called cultural units and, as in any s-code, their value is defined differentially. A cultural unit viewed in terms of its manifold coding correlations is called a sememe. A sememe is thus a cultural unit considered as the locus of a system of signs. A sememe can be analyzed into more elementary components (semes or semantic markers) which are themselves loci of other systems of signs. The interpretant can be defined broadly as that which an addressee brings to a message in order to interpret it. So defined the interpretant may seem to be a rather uninteresting tautology, but it will be more useful in the context of specific messages. Eco’s theory of codes is basically quite simple, yet it is capable of unlimited complication in its own terms whenever a system of signification or a communicative act requires such a complication. The theory of codes lays the groundwork for a theory of meaning production as the production of signs (interpretants); and the second half of Eco’s book examines various modes of producing signs, understanding such modes to encompass both the composition and the interpretation of texts.

Reading and the Reader

Eco’s theory implies that a message does not have an intrinsic meaning, but can give rise to any number of meanings. For this reason, a formal analysis of a text must be supplemented by a consideration of the text in a communicational framework in order to take account of the function of the “addressee.” This notion, now a commonplace in literary criticism, has led to an ever growing literature on “the role of the reader” as the one who activates the play between text and interpretant.¹⁷ Studies with such a focus have as their common denominator the intent of deflecting attention away from the literary work as “object” with intrinsic meaning toward the relationship between text and reader. Studies of the act of reading, however, deserve scrutiny to assure that an essentialism which valorizes the text or author as locus of meaning has not been replaced by one which valorizes the “Reader” as locus of meaning.

Norman Holland, for example, notes that individuals bring certain obsessions to their reading activity.¹⁸ He then examines a number of responses to particular works and shows how each reader recreates the work according to his or her own psychological makeup. The object of literary criticism, however, is not to be found in the psychological properties of particular readers, but in the properties of reading as a cultural activity. To use literary artifacts as a sort of Rorschach test has legitimate implications for the psychology of individuals, but criticism is interested in the public reference of the act of reading, not the individual or pathological.

Stanley Fish is also interested in the response a literary work evokes in its readers.¹⁹ Since meaning for Fish is an “event,” a process involving text and reader, he insists that the proper focus for the critic’s attention is the “developing response of the reader in relation to the words as they succeed one another in time.” Thus, in his essays on seventeenth century texts, Fish poses the question “what does this text do?” and explains the affective consequences of the linear unfolding of the sentences. Unlike Holland, Fish is not interested in the particulars of idiosyncratic readings, but in the response a text would ideally evoke from an ideal reader--a reader capable of responding to a text in the way structured by the author’s conscious or unconscious intent. How do we assess an ideal reader’s response? Fish proposes to “become” this ideal reader, or “informed reader,” as much as possible by finding out everything he can about a work’s context. He explains it in this way (407):

The reader, of whose responses I speak, then, is this informed reader, neither an abstraction, nor an actual living reader, but a hybrid--a real reader (me) who does everything within his power to make himself informed. That is, I can with some justification project my responses into those of “the” reader because they have been modified by the constraints placed on me by the assumptions and operations of the method: (1) the conscious attempt to become the informed reader by making myself the repository of the (potential) responses a given text might call out and (2) the attendant suppressing, insofar as that is possible, of what is personal and 1970ish in my response.

Holland would no doubt object that Fish cannot “suppress what is idiosyncratic and 1970ish” in his response; indeed, it is difficult to imagine how one would go about becoming the seventeenth century (if there is such a thing) and then reporting one’s responses. In fact, Fish’s method is more of a means for explaining an interpretation ex post facto than an approach to texts. Since the “potential responses a given text might call out” are virtually infinite, Fish’s theory is really a “subjective” version of the traditional “objectivity” of the literary critic. It is, in short, a theory of well-informed impressionism; and although the interpretations of outstanding readers will always be valuable heuristically, a theory such as Fish’s has not gotten us far beyond the problem posed by more strictly formalist approaches.²⁰

The problem raised by Fish and Holland is the possible scope of a theory of reading. Out of the totality of any given reader’s response, what part can be subjected to an analysis which would bring out the cultural functioning of literary artifacts? The limiting case for generalizing out of the scope of literary criticism all that is “idiosyncratic” is the notion of a literary competence analogous to linguistic competence. The notion is linked to the idea of an “ideal reader” somewhat more disembodied than Fish’s informed reader. Jonathan Culler, for example, refers to Frye’s call for a “coherent and comprehensive theory of literature, logically and scientifically organized, some of which the student learns as he goes on, but the main principles of which are as yet unknown to us.”²¹ Culler then proposes the linguistic analogy:

It is easy to see why, from this perspective, linguistics offers an attractive methodological analogy: a grammar, as Chomsky says, “can be regarded as a theory of language,” and. the theory of literature of which Frye speaks can be regarded as the “grammar” or literary competence which readers have assimilated but of which they may not be consciously aware. To make the implicit explicit is the task of both linguistics and poetics; and generative grammar has placed renewed emphasis on two fundamental requirements for theories of this kind: that they state their rules as formal operations (since what they are investigating is a kind of intelligence they cannot take for granted intelligence used in applying rules but must make them as explicit as possible) and that they be testable (they must reproduce, as it were, attested facts about semiotic competence).

Culler explicitly connects the notion of literary competence with linguistic competence as developed by transformational grammarians, and this analogy requires some scrutiny.

Chomsky’s opposition of competence and performance, the latter being “almost useless as it stands for linguistic analysis of any but the most superficial kind,”²² the former being the implicit generative rules underlying all human language which are the proper object for a science of linguistics, purifies his linguistic theory of all the messy exigencies of actual language. It is for this reason that George Steiner characterizes transformational grammar as a “meta-mathematical ideal of considerable intellectual elegance, but not a true picture of human language at all.”²³ An analogous opposition of competence and performance with regard to literature seems even more problematic, for there is even less evidence that there are implicit generative rules underlying “human literature” as a whole which could be described as a general literary competence. As it turns out, the only formal reading operation that Culler discusses is “naturalization,” “the intent at totality of the interpretive process” (Structuralist Poetics, 127), a principle, however, not peculiar to reading, but to cognition in general. Again, hypothesizing generative rules has great heuristic value; indeed, such hypothesizing is a necessary preliminary step in any analysis of literary texts. But the objectification of such rules into a “deep structure” of interpretive rules leads to an essentialist poetics like that of Nagler.

A theory of reading which would avoid the Scylla of essentialist notions like literary competence and the Charybdis of impressionism (whether idiosyncratic, well-informed or from “interpretive communities”) is suggested by Eco’s theory of codes. Meaning, as defined by the theory of codes, is differential. The reader does indeed constitute the meaning of a text, but s/he does so in part by bringing together a newly encountered text with other texts or fragments of texts which s/he has read before. In the terms introduced above, these texts act as interpretants. Since all literary texts belong to a tradition in one way or another, a study of the “intertextual” relationships of a work will bring to the fore aspects of its functioning as a cultural artifact. The question will not be how particular readers or ideal readers did or do respond to a text, but how a text, by being put into relationship with other texts, produces meaning in this “intertext.” Generally, texts will privilege in various ways the texts to which they can be most fruitfully related (traditional notions of genre, theme, period, etc., recount these privileged connections) and the epics to be considered in the present study give ample testimony to this fact.

An intertextual approach will not entirely avoid the pitfalls of impressionism, on the one hand, and on the other, the objectification of generative rules. Nevertheless, by focusing on textually established interpretants, it is possible to relegate hypothetical “rules” and individual responses to their proper role as heuristic devices. This delimitation is stated in Eco’s definition of the reader as a “textually established set of felicity conditions to be met in order to have a macro-speech act (such as a text is) fully actualized.”²⁴ Such a notion is intimately bound up with historical specificity, with placing a literary work in an historical context A reading produced by an intertextual focus will not necessarily be one ever ascribed to a work by its historical audience; it is not necessarily one “intended” by its author; such a reading certainly will not be the meaning which is “in” the text. It will rather be a meaning which can be produced by inserting a text into a certain system. Given the level of generality of Eco’s semiotic theory, it will be useful, before going on to the Iliad, to consider the intertextual reading model of Michael Riffaterre, which, although not without problems, does provide a more specific set of terminology and indicates ways in which Eco’s semiotic can be applied to literary works.

Intertextuality

In his studies of intertextual relationships in literary texts, Michael Riffaterre identifies indirection as the fundamental characteristic of poetry; that is, a poem says one thing and means another.²⁵ He therefore distinguishes the meaning of a poem from its significance. The meaning is the “mimetic” content of a text, its attempt to represent some sort of reality. The significance is the content of a text which is articulated indirectly. The act of reading a poem involves, according to Riffaterre, a transformation from meaning to significance, a process carried out by the reader retroactively: one initially reads a poem for its “meaning,” its reference. Such a reading will always be unsatisfactory in a poem, leading the reader to make another attempt to comprehend the poem on a different level of organization. The path which will lead to the significance of the poem is made manifest by some sort of “ungrammatically,” some element or feature which does not seem to fit into the mimetic level of the poem’s structure. The shift from meaning to significance is made possible by the reader’s intervention, but Riffaterre asserts that such a shift is implied by the organization of the poem, its indirection.

A poem is comprised, according to Riffaterre, of a set of lexical transformations of a “semantic given,” which he calls the matrix. The matrix is expressed in the poem by a series of variants determined by a model, which is the first or primary actualization of the matrix in the poem. The organization of a poem at the level of significance is thus a set of lexical sequences circumscribing a semantic nucleus. Now it should be clear from our discussion of cultural units that a semantic element can be displayed in the scope of verbal language only through a series of representations, which we have called the sequence of interpretants. The set of lexical transformations of a semantic given, then, is in Eco’s terms the sequence of verbal interpretants of a cultural unit. Riffaterre calls such a lexical sequence, whether it is part of a poem or not, a hypogram.²⁶ Hypograms are themselves systems of signs related to a “semantic given” (i.e., a content). In semiotic terms, the semantic given is the focal point of a series of coded relations which make up the sememe of a cultural unit. Sememe, sequence of interpretants, lexical transformations of a semantic given, are all terms for the same theoretical concept (its interpretants).

A hypogram can be either “potential, therefore observable in language, or actual, therefore observable in another text” (Semiotics of Poetry, 23). That is, a hypogram outside of poetic discourse is simply the sequence of verbal interpretants of a cultural unit. A hypogram is poeticized, according to Riffaterre, when its semantic nucleus becomes the matrix of a poem. Since the semantic nucleus has no status outside of its hypogrammatic representation, one can say that poetic discourse sets up an equivalence between two sets of verbal sequences (of whatever scope): one is the poem itself, the other is the hypogram as it exists in language or in other texts. It is in this sense that Riffaterre’s approach is intertextual: there is a relationship between a poem and another verbal sequence, which can be summarized as a word, which can be a cliché or a whole text. If a reader does not recognize the “other” verbal sequence around which a text is built, s/he will no doubt constitute some meaning or other in reading the text; but it is basic to Riffaterre’s method that there is one verbal sequence which will give the fullest account of a poem’s production.

An example will perhaps illustrate the matter better. Consider the following poem by Emily Dickinson:²⁷

1. Ended ere it begun--

2. The title was scarcely told

3. When the preface perished from consciousness

4. The story, unrevealed

5. Had it been mine, to print!

6. Had it been yours, to read!

7. That it was not our privilege

8. The interdict of God

There are two hypograms poeticized here: “book” and “lovestory.” The word “book” is not used in the poem but suggested by semic actualization.²⁸ Semantic markers (semes) of the word “book” are present in each line. The most obvious ones are “title,” “preface,” “story,” “to print” and “to read” (lines 2-6), so that the “it” of these lines must be a book. But a book also has a beginning and an end (line 1) and often has an imprimatur, which is its privilege (line 7) to be printed and read. This privilege can be abrogated by an interdict (line 8). The poem then is an articulation of semes of the word “book,” and in each line the seme is negated (scarcely told title, perished preface, unrevealed story, etc.). The second hypogram is “lovestory.” The collocation of “mine,” “yours,” “ours” (lines 5-7) suggests a sort of mutuality between people. The text is formally a lyric poem, a traditional genre for love themes, suggesting that this mutuality is one of love. Specifically, it is a love represented as a lovestory--a story of love which is, moreover, unauthorized, forbidden. The matrix of the poem, its “semantic given,” can be represented by some such statement as “our love is forbidden.” The model of the poem, the determinant of the derivation of the matrix, is something like “the book which is unauthorized.” Forbidden love is presented as an unauthorized book.

This example shows clearly Riffaterre’s distinction between matrix, model and hypogram. It also pinpoints what it is that is “outside” of the text which must be brought to bear on the poem to establish its significance. The two hypograms, book and lovestory, are sign systems which are prior to the poem, or more accurately, their prior existence is implied by the poem. They are coded systems. The fact that the sememe of the word “book” includes such semes as preface, story, title, etc., implies a complex of coded relationships which is already actual or potential in language. Lovestory is also a notion which implies a complex of coded relationships involving lovers, trysts, etc., among which is “love is often thematized in lyric verse.” The handling of these hypograms in the poem (i.e., their negation) allows us to formulate the model and the matrix, which are also conventional.

Riffaterre’s method involves the reader as the participant who must actualize the process of signification. The analysis makes explicit certain operations which are possible in terms of the available codes. It is conceivable that in the future, the eight line stanza will come to be used only for challenges to do battle. In that case, a reader unaware of the historical determinants of Dickinson’s poem might assume that it was the musings of a frustrated general unable to do battle with a hated enemy because of politics. No doubt the poem is capable of eliciting a number of responses from a number of readers. Riffaterre’s method focuses on the cultural codes which can be reconstructed on the basis of texts.

The remarkable thing about Riffaterre’s approach is that poems are mere proliferations of signifiers, periphrases of rather common verbal sequences. Although the process takes place in terms of codes, the poeticizing of verbal sequences--in stark contrast to Nagler’s hierarchical theory--is entirely lateral. A hypogram is not an ideal configuration existing on a higher ontological plane than any of its “realizations”; it is itself a system of signs whose poetic transformation results in the generation of variants of itself. As an extreme example of this, Riffaterre offers the following poem by Athanasius Kircher (Semiotics of Poetry, 20):

Tibi vero gratias agam quo clamore? Amore, more, ore, re.

The matrix here is “thanksgiving,” the model “crying out” (clamore), both standard Christian themes.²⁹ The peculiar thing about this poem is that the various ways of crying out (love, habit, word, deed) are, by a happy coincidence in Latin, all physical (phonological and graphic) as well as grammatical proliferations of the model. Although such a text cannot be taken as typical of poetic discourse, it does show the generative power of the signifier, a power completely ignored in the more abstract approach of Nagler. One need only recall that Nagler uses Latin words (procedo, non sola, etc.) to “avoid confusion of the thing itself with any of its allomorphs” (Spontaneity, 68). For Riffaterre, there are only “allomorphs,” a series of variants derived not from some underlying “thing itself,” but from each other.

The difference between the approaches of Nagler and Riffaterre is a function of their assumptions about the status of meaning and its relationship to culture. For Nagler, to repeat, meaning is inherent and prior to culture. For Riffaterre, meaning is above all a product of cultural dynamics: the process of “defining” in its literal sense of “setting down boundaries,” segmenting the continuum of experience into discrete units which have value by their insertion into a system of oppositions. To name something is to establish that it is significant to someone in some way. To a group of people for whom communication is necessary or desirable, the set of significant units defined by their common needs and interests will become articulated into the semantic plane of a signifying system by means of some set of signals organized into a syntactic system. This is the basic model of a culture. The expanding, changing character of signifying systems is the concomitant of an expanding, changing culture (“concomitant” because the relationship is dialectical).

The workings of this process is the subject of semiotics. Riffaterre’s literary analyses pose the question, what coding correlations are implied as an explanatory condition for this poem? The answer to this question lies in the concrete communicative acts which establish the configuration of the semantic plane and the coding correlations which constitute it into a signifying system. The question Riffaterre asks could be posed for any message, and it is in fact the question which users of a signifying system pose in order to interpret any message. For this reason, Riffaterre’s claim that “indirection” is the proprium of poetry seems rather curious. All messages require an addressee to produce a sequence of interpretants which are variants of the message and circumscribe the content unit of the message. Riffaterre’s insistence on this point is, in fact, indicative of a residual essentialism in his theory. Just as Eco replaces Peirce’s typology of signs with a typology of modes of sign production (Eco, 216-17), we should give up the opposition between poems and non-poems (except as an approximation), and speak of the poetic functioning of a message. Hence Eco notes that a message “assumes a poetic function when it is ambiguous and self-focussing” (262); and we should add that any message can assume such a function given an appropriate context. A complex message will always be produced and interpreted by a variety of modes of sign-production. The “aesthetic” function traditionally associated with poetry and to which Riffaterre gives the name “indirection” will be rethought below in terms of Eco’s notion of ratio difftcilis (Eco, 183-89; 217-60).

The problematic nature of Riffaterre’s definition of indirection is somewhat masked by the scope of his study: all his examples are from modern poems of brief compass with little or no interest at the “mimetic” level. They are thus somewhat prejudicial to the clear-cut distinction between meaning (mimesis) and significance, to the notion of “ungrammaticalities” as the motivating force for the shift from meaning to significance, and to the necessity of a second retroactive reading (which must strike the Homeric scholar as decidedly problematic). These notions reflect Riffaterre’s interest in describing the literary phenomenon as a “dialectic between text and reader” and in formulating rules “governing this dialectic” (Semiotics of Poetry, 1). Thus, it is not surprising that Riffaterre introduces the term literary competence, the reader’s “familiarity with descriptive systems,³⁰ with themes, with his society’s mythologies, and above all with other texts” (Semiotics of Poetry, 5). Riffaterre means by literary competence the encyclopedic knowledge which enables one to produce appropriate interpretants, a much more culture-specific notion than that of Culler. Nevertheless, his use of the term and his tendency to describe how the reader recognizes “ungrammaticalities” and becomes frustrated by a mimetic reading, etc., show that Riffaterre has not entirely left behind certain essentialist assumptions. The Reader, as “he” emerges as a textual function in a communicational model is still a formalist fiction, and reader-response criticism has taught us that real readers always exceed in various ways the position a text provides for them. Readers are “addressees,” but they are also more than that; and messages are not only utilizations of a communicational channel in order to convey information or feelings, but also can be exercises of and submissions to power.³¹

In the semiotic approach to be developed below, both author and audience enter the analysis as textually established entities. That is, they are part of the complex of coded relationships postulated by a message as an explanatory condition for its interpretation. A historical person adopts a certain textual strategy in order to communicate something. At the same time, the adoption of a textual strategy is modulated by consideration of the potential audience (beginning with the choice of a language, genre, style, etc.). The textual strategy of the Iliad and the interpretive apparatus implied as an explanatory condition for it are available to us as examples of various discursive processes,³² and as such are able to be subjected to a semiotic analysis. At the same time, a historical person adopts a certain textual strategy in order to do something; that is, as a kind of social practice. Just as a formalist analysis must give way to a consideration of communicational factors, so too, Riffaterre’s intertextual approach must be supplemented by a more “material” cultural analysis. These are not, however, options from which one can choose beforehand, but different moments on an methodological continuum.³³

It will be useful at this point to draw attention to two complementary forms of investigation: poetics, which studies textual processes, the way texts are built up; and reading, which studies the pragmatic implications of these processes in textual realizations. The emphasis in this study will be on the former, but that will inevitably lead to the latter within the constraints outlined above. The inextricable link between poetics and reading lies in the fact that they are both forms of sign production. One of the unfortunate consequences of Parry’s theory of the formula, as well as the type scene studies based on it, is that they tend to separate these two concerns, resulting in a rather static view of Homer’s poetics.³⁴ Scholars who take exception to this tendency, such as Nagler, have described a more flexible poetics, but one based on a rather static notion of meaning production.³⁵

Riffaterre’s intertextual approach posits that texts are produced from other texts and that meaning is produced not by substance (e.g., sphota) but by difference. What will be useful for us in the method of Riffaterre will be the various types of poetic mechanisms that he discusses. The notion of hypogrammatic derivation, for example, a concept derived from and especially appropriate to short poems, will provide a starting point for the study of Homer’s similes, themselves formally discrete from their narrative context. The continuity of the similes with the rest of the narrative, however, will necessitate an extension of Riffaterre’s terms and the evolution of broader concepts of textual poetics. At the same time, Eco’s semiotic theory will provide a conceptual basis for extending a purely communicational analysis to an analysis of signifying practices as social practices, specifically as examples of ideological production. The investigation of the similes in Homer and his imitators will try to see how they function in the “linguistic economies” which articulate and are articulated by these texts in their various historical contexts.