J.L. Lemke

City University of New York

Brooklyn College School of Education

Brooklyn, New York 11210 USA



NOTE: This paper was written in 1994 but for various reasons was never published. It is a revision of a paper originally written in 1987. For the current 1999 version, I have updated a few references, but the paper is otherwise unchanged.



Specifying Genres


What exactly is it that we want the notion of GENRE to do for us? And how should we specify a genre in such a way as to make it best able to serve these theoretical functions? I want to propose some specific answers to these questions which will force us to extend our notions about genre in what I hope will be useful ways.


Genres are social semiotic formations, that is, they are social constructions, the products of conventional social meaning-making practices that belong to a community's system of intertextuality (Lemke 1983, 1985, 1988a, 1988b, 1988c, 1989, 1990, 1995). We construct genres by construing certain sorts of semantic patternings in what we consider to be distinct texts, and we say that such texts belong to the same genre. Co-generic texts are privileged intertexts for each other's interpretation. When we make meaning with any text, we construe meaning relations that also depend on these other texts. The most privileged intertexts of a given text are not necessarily other texts of the same genre (depending on how we define genre): they may be co-thematic texts (roughly, those which talk about the same topic in the same ways), or co-actional texts (those that belong to the same larger sequence of social activities), or others. Which are most privileged depends on the reading practices, and ultimately the cultural norms, of a community.


It is important to recognize that we do not make meaning with texts apart from a context of culture, and that intertextual meaning-making practices play a large part in mediating between texts as instances and cultural meaning patterns that can be construed in many texts. We can represent what is sometimes called `knowledge' of culture in terms of both specific texts and these patterns, which I will call textual or intertextual formations. When we make meaning with or in a particular text, whether speaking/writing it or interpreting it, our resources for making sense in the community include not only the lexico-grammatical systems of its language, but also both prior deployments of those systems (i.e. other texts) and common cultural patterns of deployment of the systems (i.e. textual semiotic formations, including genres).


We do not make textual meaning without memory. In the production of a particular text, there are not only conditional probabilities for which lexicogrammatical system selections will be made, depending on various features of the context of situation as we are constructing it (e.g. Field, Tenor, and Mode features), there are also successional probabilities for these selections, dependent on what selections were made the last time around (see Lemke 1991). Second sentences, next clauses are more constrained, at least probabilistically (as with all register variation), than first ones. Texts do not simply have semantic features, they have semantic feature topographies, semantic texture, local- and global- scale sequential patterns of shifting semantic features. Texts are not semantically homogeneous. They are potentially semantically heterogeneous on all scales. This phenomenon has been well demonstrated by work on phasal analysis within the framework of systemic grammar (Gregory and Malcolm 1981, Gregory 1985).


Not just texts, but any text-formational semantic patterning we construe within a text that preserves ordering or sequentiality, such as the patterns by which we define genre, will also be semantically heterogeneous, will also have a semantic topography and not just a set of semantic features. I am using the term topography here to remind us that texts and text formation patterns like genres unfold in a space of many semantic dimensions. There are many, many semantic features that characterize any scale in a text or genre pattern (group, clause, clause-complex, rhetorical or dialogical structure, genre element, segment, section, etc.). Some (but, to preserve cohesion, usually not all) of these will change from each unit to the next on each scale (and also in more complex, less segmental ways, cf. Lemke, 1995).


A semiotic theory of meaning tells us that the practices by which we specify any meaningful entity do so in part by specifying its relations to other, different entities on the same order. We do not define genres just by specifying their semantic topographies. Those topographies are patterns we construe in some texts; they foreground or privilege certain combinations of meaning features which both construe similarities between co-generic texts and contrastive differences from other non-cogeneric texts. At the level of the patterns themselves, we do not have simply various genres in isolation, but we necessarily have sets of contrasting genres (cf. Martin 1985, 1991).


Do these sets form paradigmatic systems? or, more specifically, how far can we usefully carry a programme of agnation and cross-classification of genres? One consideration in such a programme is the relation between global semantic features of texts (i.e. features in which the whole genre pattern is homogeneous) and the multi-scale, shifting, semantic topographies of texts. Another way to put this is to ask whether there is a semiotic resource system of Genre in the same sense that there is of lexicogrammar? Or do we use notions of Genre differently, as a different kind of resource for meaning, closer to the way in which we use references and allusions to specific other texts to make meaning (cf. the notion of genre as an `intertextual resource' in Threadgold & Kress 1988, 1989)? Perhaps Genre sits on the border of these two sorts of meaning resources, with features of both.


It is possible, of course, to define genres without systematically relating them to one another (e.g. Hasan 1984, 1989, 1995; Ventola 1987). In this case the elements in the structure of a genre, instead of arising as realizations of system network selections, are each unique to the particular genre under definition. These elements may then have realizations in turn, within some register, by semantic features which in turn have lexicogrammatical realizations. This approach seems to capture something about our genre-defining and genre-using practices which is backgrounded by genre-agnation: there is something unique about genre elements or stages. We do not recognize in general that different genres share major structural-functional units. Such a unit in one genre is not going to turn up in what we would recognize as essentially the same form in some other genre. And there is a very good reason for this: however they may be defined, the major stages or elements in a genre structure derive their functions in part from their semantic relations to the other units at their own rank. They cannot simply by pulled out of the overall genre pattern and used as independent building blocks in some other genre.


There are units, and probably sub-units in some constituency hierarchy, that are unique to a genre. But obviously, in a realization hierarchy, eventually some genre-specific units must be realized by sequences of lexical items and grammatical structures which are not unique to the genre. The highest-ranking non-genre-specific units in terms of which the lowest-ranking genre-specific units are realized form a useful starting-point for considering similarities and differences between genres. For texts and topographic text formations like genre (i.e. those with essential structural-textural properties), similarity and difference need not be limited to paradigmatic, system-like agnation relations: typology can be extended toward a topology of genres as well (cf. Lemke 1987).


A topology, in mathematical terms, is a set of criteria for establishing degrees of nearness or proximity among the members of some category. It turns a `collection' or set of objects into a space defined by the relations of those objects. Objects which are more alike by the criteria are represented in this space as being closer together; those which are less alike are further apart. There can be multiple criteria, which may be more or less independent of one another, so that two texts, for instance, may be closer together in one dimension (say horizontal distance), but further apart in another (vertical distance).


What is essential, obviously, is our choice of the criteria, the parameters, that define similarity and difference on each dimension. These parameters must be chosen so that any two texts for which the criteria are relevant can be represented as more or less alike. The same set of parameters allows us to describe both the similarities and the differences among texts, or genres. Ideal genres can be represented by definite points in such a topological space, marking the centers of clusters of other points representing actual texts. Those clusters (`fuzzy sets' or distributions) may overlap, representing ambiguity (or `multiple inheritance' cf. Stillar 1992) in the genre classification of texts. Texts of different genres may be very much alike by some criteria, even though along other dimensions they are different enough that there is no doubt about assigning them to different ideal genres. It now becomes possible to define possible genres that would be in a definite sense intermediate between actual genres, and to decribe the evolution of genres in terms of changing distributions of actual texts over time.


The semantic topography of a text or genre formation already represents a parametrization in a multi-dimensional semantic space that can form the basis for a topology of text-types or genres. It will perhaps make the nature of the argument simpler and easier to follow if I ignore the fine details of the topography and try to simply look at one sort of gross feature, the succession of rhetorical formations (more generally, rhetorical strategies, to allow for non-segmental realizations of the underlying semantic relations, cf. Lemke 1988b, 1995) through a text.


After a discussion of the topography-topology connection, I will return to some more specific considerations of how we use semantic topography patterns of various sorts to make meaning in ways rather different from those in which we rely on lexicogrammatical systems, even register-specific ones, directly.



Rhetorical Topography and Genre Topology


Let us begin with a collection of individual texts. Each has its own highly complex semantic topography: a structural and textural shifting of many semantic features on many scales within the text. What texts are most alike, according to the system of intertextuality of our community? i.e. between which texts do we construct the greatest number of most significantly privileged similarities? Which texts do we consider to be minor variations on a common theme? Obviously this is a matter of degree, which is what allows us a notion of topology. Two weather reports in the same format for similar days, two slightly differently edited versions of the same speech, two variants of the same folktale, two similar proofs of the same geometry theorem; chapters on Newton's laws from two different introductory physics textbooks, two historical accounts of the same well-documented battle; two weather reports for very different days, two recipes from the same cookbook for slightly different sauces; a story and a paraphrase of the story, a story and a precis of the story, a story and a translation of the story into another language; two nursery tales, two limericks, two sonnets; two accounts of the same event from different observer positions, two accounts from conflicting ideological perspectives.


I will use the term texttype to describe the common semantic pattern we construe in the topographies of very similar texts: in particular, texts that belong to both the same discourse formation and the same text formation. Two texts belong to the same discourse formation when they are similar in three respects:


      (1) they must share the same thematic formations (Lemke 1983a, 1985a);


      (2) they must take the same evaluative stance (value-orientation, cf. Bakhtin 1935; Lemke 1988a, 1989b, 1990b, 1992 ) toward these formations; and


      (3) they must construct the same heteroglossic relations among the formations (Lemke 1988a).


Two texts will be said to belong to the same text formation when they are similar in two respects:


       (1) they have the same Generic Structure Potential (GSP) as defined by Hasan (1984, 1989); and


       (2) they realize the same elements of the GSP using the same rhetorical formations.


This definition already involves us in matters of delicacy. Two texts may share the same GSP, but represent different selections of the options within it. The more similar they are in those selections, the more delicately co-generic they are. The same is true with respect to their selection of particular rhetorical formations.


A rhetorical formation is a multivariate semantic structure whose functional elements and their relations are not specific to the realization of an element of a GSP, but occur in texts of many, very different GSPs. Rhetorical formations are semantic structures of the sort identified by Mann and Thompson, 1983, as `relational propositions'; see other classifications cited therein). They include such structures as: Cause-Consequence, Instance-Generalization, Thesis-Evidence, Thesis-Example, Proposition-Evaluation, Proposition-Contradiction-Alternative, Action-Motivation, Problem-Response, Problem-Solution, Preview-Argument, Argument-Summation, and many others. (Note that rhetorical formations commonly are, but need not be, binary, two-part structures.)


Diana Adam-Smith (1987), building on the work of Eugene Winter and Michael Hoey, notes that a number of identifiable rhetorical formations (most with more than two obligatory elements) can be found in both scientific writing and in other genres as diverse as nursery rhymes, letters to the editor, detective stories, and journalistic reporting. There is a great deal of work to be done identifying rhetorical formations whose semantic differences from genre to genre are only more delicate specializations of common semantic structural relations.


Rhetorical formations constitute an intermediate level of semantic structure in texts between generic structure and lexicogrammatical structure. The semantics of genre elements and their relations are specific to one separately defined genre or another. Lexicogrammatical resources enable us to make semantic distinctions which are potentially relevant in all text-types. The elements and relations of rhetorical formations are neither, though clearly which rhetorical formations are likely to occur as realizations of a particular element in a genre structure may be more predictable than are its grammatical structures.


It is possible that there are rhetorical semantic relations, as proposed by Mann and Thompson, which are universal text resources, part of the `textual' component of the semantic stratum of grammar, potentially at stake in every text. But it may also be that readers' judgments of semantic relations between paratactic clauses, or between pairs of clause-complexes unmarked for semantic relationship, are largely dependent on identification of a rhetorical formation (and perhaps even of a genre). I suspect that the semantic construction of text meaning is not made so much directly in terms of lexicogrammatical selections for each clause, as it is made in a mediated way, with relevant lexicogrammatical selections subserving the functional needs of a rhetorical formation, within a genre element, within a GSP.


It is of course artificial to separate the semantics of rhetorical formations entirely from that of discursive and thematic formations. A thematic formation may well include such relations as Cause-Consequence or Thesis-Evidence as part of the thematics of a particular kind of discourse on a particular topic. Non-segmental patterns of foregrounding and backgrounding of particular thematic formations or discourse `voices' also contribute to the rhetorical strategies of a text and may be genre-specific (cf. Lemke, 1995).


The Topology of Textformations


A textformation is specified first by its GSP, then by its optional selections within that GSP, and finally by the rhetorical formations it employs to realize elements of GSP and their relations. This characterization assumes that definite GSP-genres exist; it reifies them in a way which is convenient for analytical purposes, but presupposes an analysis of the semiotic practices by which we construe a genre; i.e. what we do to assign a text to a genre.


For each genre, there are specific patterns of successional probabilities (i.e. probabilities for each next optional selection given those that preceded). In this sense, dynamic construal of genre is a building up, a continual inference, from the succession of text elements. It is a construal of and from the unfolding semantic topography of the text. Because rhetorical formations also participate in this process, as a distinct level (semantic rank) of their own, we can first construe a formation, and then infer the genre element/s it is realizing, or vice-versa, or do both in parallel.


Two texts of the same genre, especially of the same subgenre (i.e. with the same optional choices within a GSP), will tend to have more patterns of succession of rhetorical formations in common than two texts from different genres. But the representation of a textformation in terms of its sequence of (selections of) rhetorical formations is a multidimensional one. Two texts which belong to different genres may still have many rhetorical strategies in common, even though the same rhetorical formations may be realizing quite different genre elements (e.g. the use of analogical or exemplificational rhetorical strategies in legal briefs and scientific articles). Thus, along some dimensions two texts may be topologically quite `near' to one another, even though they may not be near on all dimensions and may ultimately be assigned to different synoptic genres. Texts of the same genre (subgenre) will be near to one another in rhetorical terms, clustering around some `ideal' abstraction of the normative text of the genre. A textformation is a set of texts sharing the same (sub)genre and the same rhetorical strategies for implementing that (sub)genre. It is a point in the space. Each text is represented in this space by its textformation. Co-generic texts will tend to cluster around one or more ideal points, each representing one of the usual strategies for implementing that genre in the community. 


These standard textformations of the genre define it `from below' and must be dynamically construed in terms of the semantic elements of some GSP. To the extent that the redundancies between rhetorical formation sequences and genres are weak, the standard textformation points of different genres, and certainly their cluster haloes of actual text-points, will interpenetrate and overlap. One could not, and cannot in general, expect to derive genres from an empirical study of distributions of rhetorical formations. We must use the semantic criteria of the GSP to impose another order of patterning on the distributions, to separate otherwise similar structures and to identify otherwise dissimilar ones on functional grounds. The space of textformations is larger (i.e. has many more dimensions) than can be spanned by the parameters of rhetorical topography alone. We project the whole onto the `plane' of rhetorical sequences, creating or exaggerating overlaps between generically distinct textformations. But in doing this we represent part of the dynamic process of semiosis: the semiotic practices by which we not only partly infer genre elements and relations from rhetorical patterns, but also construe intertextual similarities between texts which share rhetorical patterns even though they belong to different genres.


Figure 1 shows a number of textformation clusters representing different usual patterns of rhetorical implementation of the elements of two different GSPs. It is greatly simplified in showing only two dimensions of the rhetorical formation space.



Along each of the dimensions Rh1 and Rh2, we can imagine a number of possible rhetorical strategies capable of implementing the functional need of one or more elements of the GSP of a genre. Genre A shows little variability in its choice of the Rh2 strategy, but three distinct preferences among the possible Rh1 strategies. Genre B is characterized by choosing about the same Rh1 strategy in all cases, but its B3 implementation of its GSP elements has the same sort of rhetorical strategy normally associated with texts of Genre A. When textformation clusters are projected onto the plane of the diagram we see how they overlap in this way. Genre-specific criteria distinguish textformations represented by different point-symbols (here `x' or `o'), but which are united as a genre within the outline of the cluster as a whole. We can still see, however, that texts of different genres may be quite close together on some rhetorical features (i.e. vertically or horizontally near or both).


Topology has implications for typology, and also for the evolution of genres. If genres were to be represented by paradigmatic system networks of semantic selections, realized by insertions, orderings, etc. of genre-specific GSP elements into a genre-syntagm, then if a scale of delicacy could be devised, we would have an implicit typology of genres. We could also construct a principle for at least a one-dimensional topology. Some genres would be more like others in the sense of sharing more network selections at more primary degrees of delicacy (cf. Hasan's 1985, 1986 notions of semantic similarities of lexical items). But suppose that at a given level of delicacy, we had semantic choices leading to different genres or subgenres. How could we answer if asked what sort of genre might fall semantically in between two such choices or their realizations? Paradigmatic choices are ordinarily categorial ones, they are not ordered, nor are degrees of difference recognized among them. The topology of a system network is a global one: the least similar choices in a network are furthest apart. But at a given level of delicacy there is no `local' topology. It is that sort of topology, however that is needed to say what sorts of genres might lie in between others in any typology.


Figure 2 indicates this in the case of a system network. The original choices are B1, B2, or B3. But B1 and B2 are drawn closer together, to show that in the semantic topology they are more alike than either is like B3. We will see, in looking at genre evolution, how this might come about. I have also shown an imaginary B4 option, drawn in a position semantically between B2 and B3, but closer to B3. This shows how a network diagram might reflect a semantic topology.





Topology and the Evolution of Genres


Genres change over historical time. Like species, some may persist over long periods in relatively stable forms and come to co-exist with their own `descendants.' Our first basic assumption will be that all genres existing today are stabilized variations and specializations of previous genres, some surviving, some extinct. This assumption actually applies more broadly to all semiotic formations. We will not be concerned here with the appearance of the most radically different or novel new formations, but with gradual evolution. We will use the topological model of textformations to describe this process in general terms.


A texttype, as we have defined it, is a semiotic formation of great specificity (delicacy). Texttypes show greater diversity and evolve more rapidly than do genres, as human `types' do in relation to a species or subspecies. Genres and rhetorical formations as we have defined them, because they are linked by redundancy relations, will tend to co-evolve. New rhetorical formations will come to be used in old genres, and genres can become differentiated by their tendencies to use different rhetorical strategies. Even without rhetorical formation evolution, genres can borrow rhetorical formations from other genres, and where particular rhetorical patterns become stabilized, a new subtype of the genre may be recognized. Naturally, rhetorical formations will evolve in the context of the genres in which they are used, successor forms diverging because they are evolving in distinct functional contexts.


To make the model more precise, let us consider the general pattern of evolution of a semiotic formation. The formation is initially specified at some time in some community by a set of features by which it is distinguished from other formations of the same kind (one genre from other contemporary genres, say). That would be a purely typological description, inadequate for an evolutionary model. To make it adequate we need terms of description that span its predecessor formations and possible successor formations. This requires the shift from a synchronic, synoptic, typological description to an evolutionary, dynamic, topological one. In biological evolution, this is the shift from taxonomic description to genetic description (either in terms of heritable traits or the underlying gene frequencies). A species is not described as a set of traits (or genes) but as a probability distribution of the relative frequency of various possible values of each trait (phenotypes) or population frequencies of genes. Successor species, like two distinct contemporary species, have distinct distribution patterns, and it is possible to imagine the intermediate stages between them.


Our topological model of textformations provides the same capability. We have a cluster of texts or texttypes, represented by their textformations, distributed in clusters around standard or `ideal' points. In the course of evolution, the practices of use of a formation, leading to actual texts, can, as circumstances change, lead to the production of more texts of one kind than of another, thus shifting the `center' of the cluster. If there are, say, polarizing social pressures, one original formation may become split into two distinct successor formations. In the later absence of this pressure, the two formations may come to coalesce. A single pressure may simply shift the character of the formation steadily in a particular direction. In a topological model one can trace the intermediate stages, or attempt to reconstruct them from the endpoints.


Figure 3 illustrates the topological model of evolution, and Figure 4 shows its possible effect on a system network.



In Figure 3, an initial cluster representing a single genre, G2, has gradually become, over time, two clusters representing distinct genres. On one axis they have changed in the same direction, but along the other, they have diverged. This could represent a situation where changing functional uses of the original genre led, through different rhetorical implementations, ultimately to entirely different genres. Figure 4 shows the possible effect of this evolutionary bifurcation on a system network for the genres.




One could also imagine this mechanism producing a situation such as that in Figure 2, with the similar options B1 and B2 having a common ancestor not shared with B3. For further discussion of these issues, see Lemke (1995).


In less formal terms, new genres arise from old ones by processes of differentiation and specialization, and their reverses (coalescence, neutralization of semantic distinctions). Any particular feature we use to trace the history of a genre will be subject to change. We cannot rely solely on a typological, GSP model because we need a basis for continuity during change. The GSP elements are specific to a genre. If that genre splits into two new genres, with newly defined elements, we have a model which can only show sudden transitions: a segment is either an instance of an element of the old or of the new GSP, we cannot see how change occurs gradually.


It should be recognized that, in general, gradualist vs. saltatory models of change are models with different theoretical emphases. When we are mainly interested in categorial differences, in when a text means one thing vs. another because it is construed as of one genre rather than another, we construct saltatory models: discontinuous change. But to the extent that we are interested in representing formations in terms which allow us to see how they are constructed in terms of elements that are not categorially specific to one formation, but provide a common denominator of comparison between formations, then we can view change as passing through an indefinite number of intermediate stages, a gradualist model. All change is defined against the background of some continuity. To say that a formation changes is to imply that it is still in some sense the same formation, only changed. The notion of continuity underlies that of change; variation only makes sense when there is also invariance. When we foreground difference, we only background residual similarity. A dynamic model must always account for both change and stability in the same theoretical terms (cf. Lemke 1984).



Genre and Ecosocial Semiotics


New meaning potential comes into a social semiotic system as language and other semiotic resources are deployed in new activity types, constructing new contexts, new genres, new situations. The development of new activity types in a community, the historical trajectory of the community's doings, involves the systems of semiotic relations embodied in its cultural formations, for these determine in large part what is considered actual, possible, and desirable. But it depends as well on the distribution of material resources in the ecosystem on which the human social community depends and of which it is an integral part. It depends on how the flows of these resources are coupled to the doings of the human community, not simply by semiotic relations, but by thermodynamic and ecological ones as well. The system of interest, the system in which the trajectories of ecological and cultural change co-develop themselves, is the joint system formed by both kinds of connections, an ecosocial system (Lemke, 1993).


In an ecosocial system the couplings of material processes and those of social semiotic practices are tightly integrated. The dialectic interaction of these two overlapping systems of couplings, which arises jointly from (1) the necessity of every semiotic practice being simultaneously a material process and (2) the role of social semiotic constructions in human ecological behavior, especially materially significant collective action, produces ecosocial change of all types.


We need, then, an ecosocial semiotics, in which all the basic theoretical constructs reflect our understanding that within and between communities there is always diversity of semiotic formations as well as continual processes of change as the ecosocial system develops itself. Constructs like community, dialect and sociolect, semiotic cultural formation and genre, all of which describe systems of practices/processes that directly participate in both semiotic and material types of relations/couplings, must be conceptualized as trajectory-entities, i.e. as being defined in terms of meta-stable, dynamic open systems that are continually in the process of developing themselves, creating conditions that require them to change and re-organize themselves in interaction with their environments.


Ecosocial semiotics also recognizes that theoretical constructs must specify the scale or scales across which they apply, because ecosocial systems are not only dynamic, they are `patchy', mosaic systems-of-systems in which different patches at the same and different scales may have different characteristics, including different ages and histories. Ecosocial systems are fundamentally heterogeneous at all scales, both as the result of the individuations of past developmental processes and as the basis for future plasticity of response to internal and external pressures for further change.


We have already seen how the notion of the semantic topography of a genre tries to (synoptically) incorporate both the trajectory character of genre (its patterns of successional probabilities, so that the meaning potential now is made different by the history of meaning-making up-to-now) and its multi-scale, mosaic character.


Ecosocial semiotics emphasizes the differences between and the essential complementarity of theoretical constructs on the order of system (meaning-potentials, distributional statistical ensembles) and those on the order of text, or semiotic formations. The former are systems of differences, abstracted from texts, that serve as resources for meaning-making. The latter are actually-deployed resources, meaning-making activities themselves as events or processes (with texts and textual formations defined as the products of these processes/practices). The former have no material couplings and constitute single semiotic resource systems; the latter always have material couplings and always necessarily co-deploy resources from multiple semiotic resource systems. The former are relatively slowly changing, and their change is totally dependent on the sets of texts they from which they are abstracted; the latter are inherently dynamic, changing moment to moment, or, more conveniently, they are temporally extended entities (trajectory-entities), defined across multiple scales. The former are few, the latter are legion.


While the system-perspective and the text-perspective on meaning-making are complementary, they do not exhaustively describe the metaredundancy relations that characterize an ecosocial system. System tells what can be done; text tells what has been done; but what tells us what is normally, typically, or usually done in a particular context? System gives us many independent sets of options, which can in principle be combined in many ways. Text tells us how, on each occasion, they were combined. But there is more information in the `context of culture', information about which combinations go together when. This information is represented, not by semiotic resource systems, and not by records of enactments (semiotic productions, texts), but by the cultural semiotic formations of a community.


Among these formations, the most fundamental in ecosocial semiotics are the actional formations, the activity-types of a community. All others can be derived from these (e.g. as products or consequences or participants or contexts constructed in and through them). They form a third order of ecosocial construct, with some features of each of the other two. Formations are types, not tokens (systems are systems of relations among types; texts and enactments are tokens). But they are also specifying (not mere potential, they tell how-it-is-done, not what-can-be-done), deploy multiple semiotics, and are temporally scaled (i.e. they define sequential as well as simultaneous contingencies, they are quasi-dynamic, they unfold semiotic topographies).


Actual meaning-making activities are always already contextualized, always already embedded in other meaning-making activities that have preceded them. They are always elements on some scale in a larger, multi-scale organization of interactions among ecosocial processes. In these contexts, the choices are fewer, and the semiotic formations both indicate the options and how the choices are to be combined.


This view leads to an expanded notion of how we make meaning with intertextual formations like genres.


In the case of the linguistic system and its texts, why is register not enough to do the job of accounting for the specific semantic choices made in texts? Mainly because it is still purely system, purely potential. Registers both say too little and too much. They say too little in that that they allow for many different texts, must allow for many different texts because they are potential. They allow for all the combinations of all their options, but not all those combinations, the metaredundancy description tells us, can be equally likely in a real community. They also say too little because they specify only simultaneous contingencies, not sequential ones; or if interpreted sequentially, they (under-)specify the same contingencies for every phase of the text. Register can give us the grounds for the typology of texts, but not for their typical topographies.


Something else is needed to enable linguistic text analysis to take into account the actual ways in which a particular community deploys its linguistic resources, and the differences between how different communities do so. Something else is need to specify which combinations of experiential, interpersonal, and textual resources will combine to make a text of a recognizable type, and why. Something else is needed to map out the sequential phases or stages of texts on various scales, including the changes in register potential in each phase.


I want to call this something, GENRE. It must be an ecosocial semiotic formation. It must specify the sequential (successional probabilities) as well as the simultaneous contingencies for deploying the resources of multiple semiotic systems (not just language alone) to construct Presentational, Orientational, and Organizational meaning. The fundamental theoretical entity here is still an Action Genre, and linguistic-typographical (or linguistic-kinesic) text genres are to be defined as products of the Actional genre that produces them. These three dimensions of meaning-making (Presentational, Orientational, and Organizational) are the semiotic generalizations of the more familiar Ideational, Interpersonal, and Textual metafunctions in language (see below). GENRE must specify, so far as the purely linguistic component of meaning-making is concerned, which combinations of semantic selections in systems from all three metafunctions are most likely at each step in the multi-scale unfolding of the semantic topography of a text of that genre.


Ecosocial semiotics offers one more reminder about genres as about all semiotic formations: they are not the same everywhere or everywhen. Genres index social subcommunities, they participate in heteroglossia, and this must be reflected in their Orientational specifications. And genres are trajectory-entities, they are constantly (perhaps slowly, perhaps not) changing as texts are actually produced. We need to understand not only what represents a big or small difference between two genres, but what represents a big or small change in one genre. And what meaning-constructing practices of a community construe continuities across changes in genres.



The General Semiotic Metafunctions


To understand better what GENRE has to specify, and so the basic dimensions of the semiotic topography of any meaning-making action (a fortiori, the semantic topography of any meaningful text), we need to consider briefly the semiotic generalizations of the linguistic metafunctions.


Imagine a painter painting a wedding scene from life. The process of painting is a representational semiotic process that is part of the overall ecosocial dynamics that includes all the activity co-present (the wedding, the talk, the music, the hissing of the radiators, the onlookers). By painting-at-the-wedding the painter is co-constructing the social reality of the event, making it what it is and more than what it would have been without his participation. He does so as long as his activity counts culturally as doing-a-painting, and without regard for, for example, the angle from which his canvas appears to depict the couple (a non-criterial feature). His painting-as-product is `addressed' to its subjects, to its prospective owners, to others who may see it. His painting-as-process participates in relations to other co-present processes and their participants, and in being influenced by these relations (distractions, frowns from onlookers, the quality of the room lighting), leaves traces of the context of production in the product, not the least of which are many features of the representation of the couple and the scene. His painting is participating in the construction of many social relations, among individuals, among collective categories.


The painter sees his emerging painting in relation to others he has painted, to other paintings he has seen, and he self-categorizes as he paints, that he is painting in such and such a manner, a style. The painter places engages himself in a material coupling to canvas, and brush, and palette and paints; he constructs a semiotic relation between his perception of the event and his representation from it in the painting. He selects, emphasizes, colors, orients, and does so within the cultural traditions of how paintings like his are categorially seen by viewers. He also paints on commission in an economic relation to his patrons, and with an eye to others who will see the painting and perhaps commission another. Perhaps he imagines how a rival would have painted this scene differently, how someone from another school, another period, another subculture might do it.


The painting is done. It is a physical object and a semiotic cultural product, a visual text. The material resources of a technology have been deployed in ways from which we abstract, many times repeated, the community's semiotics of depiction. It has an origin that includes at least a painter in an ecosocial process of enormous richness and complexity. We, the viewers, are now the primary participants, making sense with this object, categorially interacting with it perceptually. We construe a represented process and its participants, but also relations among them above and beyond those called for by the process-type. We construct a probable point-of-view of the painter toward his subject, and we construct relations between ourselves and our reconstruction of the painter and the painting-process. We construct relations between ourselves and our social categories and those we take to be depicted, and between ourselves and the processes depicted. Perhaps we buy the painting. Perhaps we burn it for firewood. Perhaps it falls from the wall over the bed and kills us as we sleep. The ecosocial dynamics continues to construct interactions between elements, human and nonhuman.


What are some of the kinds of doings that the semiotics of painting-as-process abets? Let us begin from the most culturally canonical and work our way outwards. It abets the pictorial representation of processes, events, participants. There is a wedding: there is the bride, there the groom, there the priest, there the parents, there the ring, there the veil, there the church interior. It abets the construction of relations above and beyond those required by the process-type: the bride's parents smile, the groom's do not; the people on the groom's side of the aisle are well-dressed, on the bride's a little shabby; the bride and her people are fully lit, the groom's people in shadow; the gold of the ring is exactly the same as the gold in the center of the flowers on the altar. It abets the construction of the point of view of painter or viewer toward the represented scene: we look down from above and from the front of the nave, a cherub seems so positioned as to smile directly at the bride. It abets the construction of the placement of this painting by its style in a certain period, a certain school, a certain painter's oeuvre. It abets the construction of social relations in the community of its provenance and in ours, and in the construction of social points of view and values, as well as the categorial systems of action and depiction.


Every event, every action is a nexus in which heterogeneous webs of interactions and relations of ecosocial processes coincide. From it we may follow outwards the threads of its thermodynamic and technological connections, its interpersonal and social connections, its economic and cultural connections, into the past, through the present, to the future. It is all semiotic, all categorizable (but not exhaustively so). What we represent and how we represent it works or can be used to create organizational (i.e. structural, whole-part, etc.) linkages and evaluative orientations. How we organize can affect the definition of what is represented and how it is evaluated. The evaluative orientations we impose can change the what and how that is seen to be organized. All the resources serve all the functions when they are used in conjunction, and they must always all be used. There is no meaningful event, action, or representation for which we do not construct a categorial what, an orientational which, an organizational how. These are the basis of the general semiotic metafunctions.


There is a Presentational metafunction, and all semiotic resources contribute to it. It is the function that determines What-It-Is, What's-Going-On. There is an Orientational metafunction that determines How-It-Looks-to-Whom. And there is an Organizational metafunction that determines How-It-Hangs-Together.


The deeper meaning of the Presentational is that we categorially construct the event-types, process-types, activity-types, participant-types. All the doings and participants that are made and re-made, and even innovated and changed.


The Orientational function of semiosis orients the What in the social space of relations of individuals and collective social categories and types: it tells Whose What it is, from Whose Point of View it is seen, What it does to Whom, and What Who Thinks of It: It is Good? is it True? is Beautiful? is it Likely? is it Usual? is it Desirable? is it Ours? It enables us to construct, deconstruct, and reconstruct all of these.


The Organizational function enables us to create and perceive (to construct) wholes from parts and to distinguish parts within wholes. It produces unity and multiplicity, structure and texture, figure and ground, entity and relation, sequence and chord, constituency, contour, periodicity, chain, warp and woof, nexus, intersection, co-patterning, interruption, boundary, and closure.


Every instance of semiosis performs all three of these metafunctional constructions with all the resources at its command. No function may be performed alone, but only in conjunction with the other two. This is not an empirical fact alone, but the defining characteristic of semiosis: it is a triple construction (or at least a multi-construction, for however we may choose to describe or count its metafunctions, they are plural and interdependent).



Intertextual Meaning-making


The meaning relationships we construe between words, and more generally between `wordings' constitute the familiar meaning-making resources of lexis and grammar. Lexical semantics and more generally lexico-grammatical semantics defines the specific kinds of meaning relations, the cross-classifying similarities and differences, our community construes between the meaning-potentials of words and wordings. But in our community we do not just construe meaning-relations between individual words, isolated or in context, or even between larger lexicogrammatical units like clauses or clause-complexes. We also construe meaning-relations between whole texts , and between parts of a text, and between parts and wholes, both within what we count as the same text and between one text and another.


The intertextual-relation-construing practices of a community are a very significant part of its culture, defining the semiotically constructed `web' that ties together all the texts, all the events which that community recognizes. These practices may also be significant for the texts and events between which NO relations are construed (disjunctions in the system of relations), because, while every culture is a culture only by virtue of its making some relations and not others (selective contextualization, i.e. metaredundancy), which ties are made and not made, between which texts and events, is fundamental.


We construe meaning relations between texts by construing particular kinds of patterns in the semantic topographies of those texts. These patterns (semantic formations or intertextual formations) are theoretical constructs intermediate between system (meaning potential) and text (instantial deployment of that potential). They are like systems OF made  texts rather than systems FOR making texts. But they are not properly Systems in the sense of systemic linguistics, for they are never purely paradigmatic, nor do they fundamentally act to classify texts. There is no agnation of these patterns, no minimal contrast pairs.


One reason why there is no simple agnation between semantic formations is that they are not homogeneous in the features that define them. They are not defined by single features which are consistent throughout, but by complex patternings of features, including sequential as well as simultaneous patternings, in which feature values change or modulate as part of the pattern. That is, they are defined by semantic topographies.


This is because our culture construes relations between texts other than ones based purely on semantic consistency. We can and do construct intertextual relations between texts of different registers (which, as argued above, systematize a consistent meaning potential). The classes of texts which are defined by the metalinguistic text-tie construing practices of a community (its system of intertextuality) cannot in general be entirely accounted for by notions of register for two fundamental reasons:


(1) the semiotic, social, cultural principles of intertextuality are not exclusively based on semantic consistency


(2) they are based, in part, on meaning patterns within texts of a given type (discourse coherence strategies) that cannot be described by register alone because they are defined in terms of features which would require a level of register description so delicate that, at that level, the semantic topography of the text would be constantly changing its micro-register.


We construe relations between stretches of text, within and between `texts', by construing similarities and differences of topographical semantic patterns of a number of types already referred to. These patterns, or intertextual semantic formations, derive necessarily from the application of the three general semiotic metafunctions in the domain of intertextual meaning-making. Thematic formations (Lemke 1983) are construed predominantly on the basis of Presentational meaning. Co-actional intertextual relations (Lemke 1985) foreground Organizational meaning. And heteroglossic intertextual relations (Lemke 1988a) obviously foreground Orientational meaning. When we look in more detail at the Organizational bases of intertextuality, we see that such things as genre structure , rhetorical patterns, and various other text-forming strategies (e.g. Hasan's notion of texture as a complementary Organizational principle to that of structure ), play important roles. We are just beginning to understand the kinds of Orientational patterns that play a role in heteroglossic intertextual relations (e.g. Lemke 1992).


There are intertextual semantic formations of many kinds. They are tools for meaning-making every bit as much as are semiotic resource systems like lexicogrammar. Their metafunctional components reinforce one another just as do those of resources at the order of system . They cannot be reduced to those resources (though they cannot be described without, in part, using those resources to do so). Textual meaning-making cannot be understood, cannot be accounted for, without analysis in terms of formations as well as constructs at the order of system such as register.


Formations are a different order of resource for meaning-making than are the systems of meaning-potential which make them possible, and accordingly we deploy them differently. Recently Threadgold and Kress (1988, 1989), have argued for a generalization of the notion of genre to what they call `intertextual resources'. Not only do they point out, as I have here, that earlier models of genre have tended to emphasize constituency organization over other modes of textual organization, and that we need to pay more attention to the Orientational meaning strategies of genres than we have so far, but they note that such notions as `discourse types' and `cultural narratives', which may not be the same as genres, seem to play similar functions. Martin (1991) has seen this claim as offering a model that is in direct competition with system-based, register-like theories of text analysis, rather than, as I would see it, a necessary complement to them.


Consider the `cultural narrative' construct. The original idea here, a literary one, is that there are in every culture particular stories that have been given central roles in the culture's textual definition of itself. They are part of the community's processes of collective identity construction. They may be myths or legends, they may be idealized historical accounts. The stories of the central events of the life of Christ, or of Buddha, as constructed long after by communities of their followers are cultural narratives in this sense. National stories, like American stories about the Revolution and the making of the Constitution, or the First Thanksgiving story (Pilgrims and Indians), or the Mayflower Compact story, would also qualify. There are also more generic stories:  Indian massacre stories (both kinds), Gunfighter-Sheriff stories, even Rancher-Herder stories. Note that what are meant here are the plots typical of such stories, and not their genre forms as such. There are, finally, `canonical' stories about important kinds of events in people's lives that are often recapitulated and form a sort of ideal: the Falling-in-Love-and-Getting-Married story, the My-Child-Says-Its-First-Word story, etc. And these perhaps shade into the narrative forms in which we describe the consciously known recipes for activity-type performances.


At one end of this continuum are highly specific stories, at the other, genres of stories in which the details change, but most of the plot is predictable. But clearly it is a story, not necessarily a text as such, that matters here. That is, there are many tellings of essentially the same culturally salient stories, many texts that render them in ways that still count culturally as being `the same story'. That story is a specific meaning pattern construed by the community in all those texts; it is an intertextual formation (or, in relation to the reading, or telling, practices that construe it, a product of an actional formation).


Consider the Christian story of The Crucifixion. How easy it is to invoke this story with a word, with a pattern of events, with a reference to historical enmities. How much is missed by readers who may not know it. How much meaning is made intertextually between this story and other stories, between this text and texts of many registers and genres. How important this meaning-making has been to the social semiotic of many communities. The intertextual meaning-making practices that depend on (i.e. build meaning in part by construing) this sort of textual formation are many. To the meaning of how many texts, in how many different ways, would the members of various communities consider the story of The Crucifixion relevant?


Do the set of stories, the set of common plots of stories in a community form a semiotic system as such, any more than do the texts of the community as such? Is it reasonable to look for agnations among stories? Or do we understand their semantic topographies intertextually, rather than systemically, by reference to other stories with similar topographies on various scales?


The notion of cultural narrative as intertextual formation is only one example. I have discussed thematic formations and their roles in intertextual, and intra-textual, meaning-making in many places (see references above). They also cannot be fit into agnation schemes; they are not meaning-potentials, they do not have purely paradigmatic relations to one another, they are more than the sum of the semantic parts they are made from, which, ultimately, do have such relations.


Some notion like `cultural discourse' or `discourse formation' or `heteroglossic discourse voice' also clearly seems to be a candidate intertextual formation type. These are, in the simplest sense, ways of talking about something. They have at least a presentational and orientational consistency from one textual instantiation to another, and they may have many other distinguishing features.


And then there is Genre itself. In its broadest usage, it may encompass most of what I have been calling intertextual formations, but I have pretty well already defined it above in a much narrower sense as a particular sort of intertextual formation: one for which the relevant patterning of semantic choices corresponds to the product of a particular actional formation, and so has presentational, orientational, and organizational specifications. This is one of the narrowest, but probably for that reason, more useful characterizations.


While agnation may not be the right way to approach some, or perhaps all, the kinds of textual formations, that is not to say that the meaning-making practices of communities do not construct larger configurations of intertextual formations. There are indeed formations-of-formations. I have discussed this for the case of heteroglossic formations (e.g. Lemke 1988a, 1990b), and it is likely from the work of Michel Foucault (1972), that such meta-formations are a general cultural phenomenon. (He uses the term `discursive formation' for these larger meta-formations, a source of confusion in many discussions.)


Martin (1991) has indicated that something like agnative cross-classifying relations can reasonably be construed among some genres, at least if these are defined by relatively abstract and general features. Obviously the thesis that genres derive from activity-types, and that these are formations that can be defined in terms of deployment of resources from an actional semiotic resource system, would suggest that at some level genres can be agnated. But there may be a deeper problem here. Can activity-types themselves be agnated, not in small sets, which is surely possible, but in a consistent way that cross-classifies (at various levels of delicacy of course) all activity-types? Probably not. Activity-types, as semiotic formations, are too far from the order of system already, too close to the order of enactments, like texts. What can be rendered in terms of system potential must be some semantic-like elements and relations, the actual features defined in an actional semiotic system, which are the component elements of the actional patterns we call formations, rather than the formations themselves. Formations are not reducible to systems. Both are needed in order to analyze text meaning.


In an ecosocial semiotics, the very processes of using formations lead them to change. This is a large subject I will not develop further here (see Lemke, 1993). But it is important to see that a focus on text rather than on system can go much further even than the formational half-way house I have been building here. It is in the order of text that the processes of change are at work. Insofar as we focus on the unique meanings in texts, rather than on their commonalities with other meanings, we are led away from the reassuring stability of system toward the unpredictable chaos of happenings . But at the same time we are also led away from the rule-governed order of the rulers and the ruled, toward the semogenetic improvisation of the players and their play.






Adams-Smith, Diana. 1987. Source and derived discourse: Popularization in science research papers. Paper presented at the International Congress of Applied Linguistics, Sydney.


Bakhtin, Mikhail Mikhailovich. 1935. Discourse in the novel.  In M. Holquist, Ed. The Dialogic Imagination (1981).  Austin TX: University of Texas Press.


Foucault, M. 1972. The Archeology of Knowledge . New York: Random House.


Gregory, Michael. 1985. Towards Communication Linguistics. In J.D. Benson and W.S. Greaves, Eds.  Systemic Perspectives on Discourse, Volume I . Norwood, NJ: Ablex Publishing.


Gregory, M. and K. Malcolm. (1981) Generic situation and discourse phase. Toronto: Glendon College, Applied Linguistics Papers (mimeo).


Hasan, Ruqaiya. 1984. "The structure of the nursery tale." In  L. Coveri, Ed. Linguistica Testuale . Rome: Bulzoni.


--    1985. "Lending and borrowing: From grammar to lexis" Beitrage zur Phonetik und Linguistik (Special Issue: Essays for Arthur Delbridge, J.E. Clark, Ed.)


--    1986. "The grammarian's dream: Lexis as most delicate grammar"  In  M.A.K. Halliday and Robin Fawcett, Eds., New Developments in Systemic Linguistics . London: Frances Pinter.


--    1989. The structure of a text. In M.A.K. Halliday and Ruqaiya Hasan. Language, Context, and Text . London: Oxford University Press.


--   1995. The conception of context in text. In M. Gregory and P. Fries, Eds., Discourse in Society: Functional Perspectives . Norwood, NJ: Ablex Publishing.


Lemke, J.L. 1983a. Thematic analysis: systems, structures, and strategies. Recherches Semiotiques/Semiotic Inquiry (Toronto) 3 : 159-187.


--  1984. Semiotics and Education . Toronto: Victoria College/Toronto Semiotic Circle Monographs.


--  1985a. Ideology, intertextuality, and the notion of register. In J.D. Benson and W.S. Greaves, Eds.  Systemic Perspectives on Discourse, Volume I . Norwood, NJ: Ablex Publishing.


--  1987. The topology of genre. (Unpublished manuscript)


--  1988a. Discourses in conflict: Heteroglossia and text semantics. In J. D. Benson and W. S. Greaves, Eds. Systemic Functional Approaches to Discourse (pp. 29-50). Norwood, NJ: Ablex Publishing.


--  1988b. Text structure and text semantics. In R. Veltman and E. Steiner, Eds. Pragmatics, Discourse, and Text (pp. 158-170). London: Pinter.


--  1988c. Genres, semantics, and classroom education"  Linguistics and Education 1 : 81-99.


--  1989a. Social semiotics: A new model for literacy education. In David Bloome, Ed. Classrooms and Literacy . Norwood, NJ: Ablex Publishing


--  1989b. Semantics and social values. WORD 40 (1-2): 37-50.


--  1990a. Talking Science: Language, Learning, and Values . Norwood, NJ: Ablex Publishing.


--  1990b. Technical discourse and technocratic ideology. In M. A. K. Halliday, John Gibbons, and Howard Nicholas, Eds.  Learning, Keeping, and Using Language: Selected Papers from the 8th AILA World Congress of Applied Linguistics , Sydney 1987. Vol. II. pp. 435-460. Amsterdam: John Benjamins.


--  1991. Text production and dynamic text semantics. In E. Ventola, Ed. Functional and Systemic Linguistics: Approaches and Uses . [pp. 23-38]. Berlin: Mouton/deGruyter (Trends in Linguistics : Studies and Monographs 55).


--  1992. Interpersonal meaning in discourse: Value orientations. In M. Davies and L. Ravelli, Eds. Advances in Systemic Linguistics: Recent Theory and Practice . [pp.82-104]. London: Pinter.


--  1993. Discourse, dynamics, and social change. Cultural Dynamics 6(1): 243-275. [Special Issue: Language as Cultural Dynamic, M.A.K. Halliday, Ed. Leiden: Brill.]


--  1995. Intertextuality and text semantics. In M. Gregory and P. Fries, Eds. Discourse in Society: Functional Perspectives. [pp.85-114] Norwood, NJ: Ablex Publishing.


Martin, J.R. (1985) Process and text. In J.D. Benson and W.S. Greaves, Eds., Systemic Perspectives on Discourse (pp.248-274). Norwood, NJ: Ablex Publishing.


--    1991. Intrinsic functionality: Implications for contextual theory. Social Semiotics 1 (1): 99-162.


Mann, William and Thompson, Sandra. 1983. Relational Propositions in Discourse. Marina del Rey, CA: Information Sciences Institute.


Threadgold, T. 1989. Talk about genre: Ideologies and incompatible discourses. Cultural Studies 3 (1): 101-127.


Threadgold, T. and G. Kress. 1988. Towards a social theory of genre. Southern Review 21 (3): 215-243.


Stillar, G. 1992. Phasal analysis and multiple inheritance. (Manuscript)


Ventola, E. 1987. The Structure of Social Interaction . London: Pinter