Go to Semiotic Review of Books Home Page
Go to SRB Highlights
Go to SRB Archives

SRB Highlights

This review appeared in Volume 9 (2) of The Semiotic Review of Books.

SRB Insights: Metasemiosis

by Paul J Thibault

1. Classical Antecedents: Plato And Aristotle

The distinction that Plato made between the sensible or perceptible world and the world of ideas is generally taken to be the first western attempt to distinguish between knowledge of the true or authentic reality -- viz. the realm of the idea, or pure form -- and knowledge of its mere reflection, viz. the material world. According to Plato, only the former can correctly represent truth. That is, true scientific thinking can only take place in the world of ideas, where the mind is not required to refer to the sensible world. Plato, in The Republic, and especially in his discussion of the divided line and the cave (Plato 1973 [1955]: 274-86), goes to some lengths to refine this line of thinking. To do so, he postulates a hierarchy of knowledge, as follows: (1) illusion, concerning the apprehension of the mere shadows and images of things; (2) commonsense opinion and belief concerning both the physical things themselves, as well as the practical morality of ordinary people; (3) mathematical reason, based on assumptions and deductions derived, nevertheless, from the images of things; and (4) philosophy, or dialectic, based on universal first principles which are entirely freed from the sensible world. Whereas (1) and (2) refer to the world of perceptible phenomena and physical things, (3) and (4) refer to the intelligible world of the idea, or pure form. There is a hiearchical progression from (1), the lowest form of knowledge, to (4), which is the highest.

It was Plato's student, Aristotle, who both critiqued and extended Plato's dualistic conception of the physical world and the world of forms. Aristotle rejected Plato's view that the world of forms is transcendent with respect to things and images. According to Aristotle, the former is immanent in the latter. This modification of Plato's conception means that scientific inquiry is not confined to just one side of Plato's dualistic conception. Instead, Aristotle was interested in investigating the nature of the relationships which link the two realms. That is, the links between the material world and the world of pure forms. Aristotle agreed with Plato thatphilosophy was the highest form of scientific knowledge. However, Aristotle, in investigating that which connects the two realms, divided philosophy into a number of sub-disciplines, viz. physics, mathematics, and metaphysics. In introducing the notion of a metaphysics, Aristotle did more than simply produce an overarching basis for the organization of knowledge into separate areas of inquiry. This much had already been achieved by Plato. With his metaphysics, Aristotle proposed a common basis both for talking about the foundational principles on which all of the various domains of inquiry are based, and for specifying their respective positions in the overall knowledge hierarchy.

Thus, Aristotle distinguishes physics from metaphysics on the following grounds: physics is based on direct observation and a posteriori demonstration by means of syllogistic reasoning; metaphysics, on the other hand, is a speculative mode of inquiry which goes beyond the direct observation of physical phenomena so as to investigate the essences ofthings. That is, metaphysics is a contemplative, rather than a practical or experimental, form of scientific inquiry. In his metaphysical inquiry into the essences or the ultimate realities of the universe, Aristotle proposed in a series of linguistic treatises -- e.g. Topics, Categories -- that preceded his Metaphysics a subtle account of the relationship of the categories of Greek grammar to this speculative metaphysical enterprise (Tanner 1970: 139). According to the classical scholar, R. G. Tanner, Aristotle, through his linguistic analyses, showed how the patterns of Greek grammar are projected, by analogy, onto the world of substance, thereby constituting its further articulation as a complex system of causes (Tanner 1970: 143).

2. Modern Scientific Discourse and Plato's TranscendentalRealism

In the modern era, practitioners of scientific discourse have, on the whole, turned away from Aristotle's insights to embrace a neo-Platonic conception of scientific theory and practice. Rather than concern itself with the essences ofthings, as originally conceived by Plato, modern science, as conceived by its epigones, such as Galileo, Bacon, and Newton, has sought, on the basis of systematic observation and experiment, to discover the law-like regularities which regulate the behaviour of the natural world. Modern science, so defined, has sought to explain observable phenomena on the basis of abstract mathematical principles which have universal applicability, independently of human perception and experience. This has lead to the widespread belief in the scientific community that there are two distinct ways of knowing, viz. by "direct perception" and by the "application of rational reasoning and higher intellectual functions" (Davies 1993: 152).

In taking up a modern form of Plato's transcendental realism, the western scientific discourse has, until recently, lost sight of Aristotle's subtle insights into the ways in which language is not simply a formal calculus for discovering eternal truths about a superior external reality. Rather,language is immanent in the practices of those -- e.g., scientists, philosophers -- who use the patterns of natural language in their endeavours to discover and understand what the universe is like and how we relate to this. The two questions turn out to be not so separate, or separable, after all. Thus, Aristotle's highly original linguistic analyses of essences may be seen as an attempt to understand how the language of observation and demonstration helps to constitute that which is observed or demonstrated, as well as the observer's or demonstrator's relationship to this. In this sense, Aristotle's system of syllogistic reasoning may be considered the first expert system for manipulating observable reality and our knowledge of it.

Rather than shying away from the big questions as to how, for example, the world we perceive with our senses relates to 'reality', Aristotle's linguistic analyses of the relationship between the categories of Greek grammar and the world of substances placed the normative character of our metalinguistic resources for understanding and acting onthe world at the centre of his conception. Aristotle showed that the agent who uses language in this way is also engaged in an act of self-observation in the very act of using language for the purposes of scientific observation and demonstration. That is, both observer and observed are embedded in a language system, and it is this which gives meaning to the act of observation.

For example, Aristotle shows that more is involved than simple deduction from observation and the linguistic representation or symbolization of that which is observed through the categories of Greek grammar. The utterance is also and at the same time a modal assertion of the truth-value of the proposition (Tanner 1970: 141). Aristotle thus drew attention to something more than the proposition's function of symbolizing an essence in the process of deduction from observation. He was also concerned with the normative value of the proposition as a modalized assertion of truth (Tanner 1970: 141). The notion of normativity impliesthe agent's ability to reflect upon the grounds for his or her making a given knowledge claim, for performing a particular action, and so on. Aristotle's concern with the modal status of the proposition indicates his awareness of the intersubjectively recognizeable criteria of what is normatively upheld to be true, possible, necessary, and so on. Further, Aristotle's concern was not with some real-world which transcends language. Rather, he was concerned with the kinds of knowledge claims that can legitimately be made in the different domains of scientific praxis that he postulated. For example, what are the normative rules and procedures of metaphysical speculation, as distinct from those having to do with the direct observation of the physical world?

In Aristotle's praxis-oriented conception, these are distinct, if related, realms of human social activity. For Aristotle, these two distinct domains of scientific praxis show that modal criteria of 'truth', 'possibility', and 'necessity' are not absolute categories. Rather, different criteria govern the useof such categories according to the particular domain in which they are used. In this sense, Aristotle's metaphysics is not a formal theory or a calculus which strives after universally consistent criteria of truth and rationality. Instead, Aristotle's metaphysics is informed by a praxis which is concerned with the ways in which theory is used in specific domains of human activity.

3. Kant and the Emergence of Scientific Metalanguage in the Modern Era

In a number of important respects, Kant's views concerning the systematization of knowledge represent an important bridge linking Aristotle to twentieth century notions of metalanguage. Kant was probably the first philosopher of the modern era to propose a unified system, or 'architectonic', "of all knowledge arising from pure reason (1970 [1781]: 655). In Kant's view, this architectonic unity is a system which permits "the unity of the manifold modes of knowledge under one idea" (1970 [1781]: 653):

This idea is the concept provided by reason -- of the form of a whole -- in so far as the concept determines a priori not only the scope of its manifold content, but also the positions which the parts occupy relatively to one another.(Kant 1970 [1781]: 653)
Kant used a range of biological metaphors to show that the "diverse modes of knowledge" and their organization into a more systematic unity cannot be talked about on the same level as the various modes themselves. That is, he recognized that the higher-order 'idea' imposes constraints on the lower-order relations of the parts to each other. Such constraints are informational rather than physical. Further, they exhibit properties of self-organization such that the higher-order 'idea' provides the systemic environment in which knowledge, on analogy with biological organisms, individuates along a specific developmental trajectory:

Systems seem to be formed in the manner of lowly organisms, through a generatio aequivoca from the mere confluence of assembled concepts, at first imperfect, andonly gradually attaining to completeness, although they one and all have had their schema, as the original germ, in the sheer self-development of reason. Hence, not only is each system articulated in accordance with an idea, but they are one and all organically united in a system of human knowledge, as members of one whole, and so as admitting of an architectonic of all human knowledge...(Kant 1970 [1781]: 655)

In the more familiar terminology of twentieth century philosophy, linguistics, and semiotics, Kant's 'architectonic' is a second-order metalanguage which takes the various modes of knowledge as its first-order object-language. It is important to emphasise the importance of Kant's use of biological metaphors in this connection. The metalanguage is the source of both integration and further development of the object-language. Systems of knowledge, like biological organisms, individuate under informational constraints. Like biological organisms, a system of knowledge undergoesdevelopmental stages which are only implicit in its early stages. The "sheer self-development of reason" refers to the growth and development of uniquely individual systems of knowledge. As "members of one whole", they exhibit properties of self-determination on the basis of principles of top-down conceptual integration, rather than on the basis of a mechanistic, or bottom-up, assemblage of parts.

Kant showed more clearly than any previous philosopher that scientific knowledge is constrained by higher-order or meta-level informational constraints. In the process of its self-organization relative to some cultural environment, both the unity of scientific knowledge and the importance of inquiring into its principles of organization come to the fore. That is, the moment that scientific knowledge begins to take seriously the question of how to describe and model itself, then the familiar epistemological problems of self-reference and recursivity become central. These questions are discussed in the next section.

4. Language and Metalanguage

Our experience of the phenomenal-material world by means of our various senses and our linguistic and other semiotic construals of this are not the same. Language and other semiotic systems are systems of interpretation. For this reason, they constitute a theory, or a complementary plurality of theories, of the phenomenal-material world that they interpret (Halliday 1988). As Gregory Bateson has observed, "language bears to the objects which it denotes a relationship comparable to that which a map bears to a territory" (1973: 153). Like map and territory, language and the objects it denotes exist at two distinct levels of abstraction, or two orders of logical typing. There is not, then, a simple relation of correspondence between the map -- c.f. language -- and the territory -- c.f. the phenomena of experience -- that the former is used to interpret. Rather, systems of interpretation such as natural language incorporate into their own internal design the very principles of organization whereby the relationship between object andsystem of interpretation is represented. They incorporate usually implicit metalinguistic rules which must be learned at the same time that language itself is learned, and which specify how and when particular words are related to particular objects. This means that the systems of interpretation which we use to model the infinitely richer and more complex phenomena of the world are self-referential to the users of the given system of interpretation. In other words, such metaling-uistic principles provide local criteria of meaning which enable language users to contextualize the relationship between object and system of interpretation.

This may be modelled in a number of possible ways, as follows: (1) object a and language form b are related to each other in the context c; (2) the relation of a to b is of the type c; or (3) the relation of a to b is enacted or constituted by c. This simple formalism tells us that the relation (a/b//c) is a contextualization of the (a/b) relation by the higher-order context c (Lemke 1984: 37). Following the earlier suggestionsof Bateson and their further development by Lemke, the single slash '/' denotes a first-order relationship, and the double-slash '//' a second-order relationship in the overall hierarchy of relationships represented by the formalism. The basic assumption underlying the formalism, irrespective of the specific mechanisms involved, is that the first-order relationship (a/b) is only completed or given a local meaning when it is in turn contextualized by context c. That is, c is a context of a type which is recognizeable in the culture in question. If a is some real world object, then b may be the word used to denote that object. In such a case, c would be the metalinguistic rule which, as Bateson puts it, "governs" the relation between word and object (1973: 153).

A further reading shows more clearly the implications of 'meta-' levels of contextualization. In this reading, we may say that c comments on (a/b) in the sense that it classifies this relation as being of a certain type, or as being formed according to a certain normative (metalinguistic) rule. Sucha rule may tell us how and when a is related to b.

In all of these possible readings, the higher-order, or meta-level, context c, in construing the lower-order (a/b) relation, imposes a form of closure on this. It does so in ways that integrate the (a/b) relation into the given system of interpretation whereby social agents make sense of it.

5. The Stratification of Language and Metalanguage

The basic principle illustrated in the previous section shows how a given set of relations, say, (a/b) is interpreted by integrating it into a higher-order or meta-level context c. It is also the foundation upon which the stratified view of semiosis proposed by Saussure and Hjelmslev in the early decades of this century is built. According to this view, the two planes, or strata, of the sign, viz. signifier and signified (Saussure), or expression and content (Hjelmslev), each contribute a layer of organization to a single overall linguistic form arising from the reciprocally defining relations between the two strata. Linguistic signifiers are phonological or graphological patterns which construe the other level ofthe sign, the signified, and vice versa. However, signifiers do not directly construe the semantic or conceptual meanings of their signifieds. The point is that phonological and graphological signifiers construe the lexico-grammatical forms -- morphemes, words, phrases, etc. -- which in turn construe the semantic or conceptual meanings of these lexicogrammatical categories. The signifier signifies the word -- the lexicogrammatical form -- not the meaning of the word in any direct way (Thibault 1996: 248).

In this view, the semantic or conceptual stratum constitutes a higher-order, or meta-level, context for construing the lower-order relation between a particular phonological or graphological pattern and the lexico-grammatical form this signifies. This may be illustrated by means of a simple example. Take the English plural noun cats. The stratification of this simple sign is presented in Figure 1 below.

According to the logic of the formalism presented above, anumber of possible interpretations of this relationship may be proposed. Here are some of these: (1) phonological pattern a and lexico-grammatical form b are related to each other in the context of the conceptual meaning c; (2) the (a/b) relation is a stable form of the type c; (3) the relation between a and b is formed by the higher-order meaning c.

The higher-order relation c may also be a metalevel comment on the lower-level (a/b) relation. If, for example, I say 'the word cat denotes the members belonging to the class feline', then I am saying that in the English language the association of this particular phonological pattern with a particular lexicogrammatical form has a certain meaning. The same is true if I translate from one language to another. Thus: 'cat means "gatto" in Italian' tells us that the phonological pattern /kaet/ in English may be glossed as having the meaning that the corresponding Italian word has. The English signifier is glossed as having the meaning 'gatto' in Italian.

The stratified nature of the relations discussed above lies atthe very core of a major epistemological problem concerning the recursive or self-reflexive character of the levels of relations involved. The (a/b//c) relation is a recursive or an iterative one. This logic may be further extended: if c is the higher-order context which constrains the lower-order (a/b) relation, then what, in turn, is the still higher-order rule of interpretation or convention which specifies when (a/b//c) may be used? And so on. The postulation of a given level of contextualization always implies the potential existence of a still higher level. Potentially, this could lead to an infinite regress of contexts of contexts of contexts ... , or, in other words, to an endlessly receding hierarchy of levels of relations.

In our example, the (a/b) relation is the focal level. This both constrains still lower-level relations at the same time that it is in turn constrained by the level above it, and so on. If we take level a -- i.e., the phonological patterns of the signifier -- as focal, we can show from a synchronic perspective how eachlevel or stratum in the hierarchy of scalar levels has its own principles of structure and organization. Thus, phonological units and relations are distinct from lexico-grammatical ones. The relation between any two levels is both indirect and non-causal. Phonological structures do not cause lexicogrammatical ones. Instead, the units and relations on any given level mutually constrain those on other levels.

Now, the fact that any given level is always constrained by its relations with the levels above and below it means that no level is autonomous with respect to the others. This fact draws attention to the ways in which the mutual constraints operating across levels provide the basis for a system to become distinct from its environment by virtue of its own internal dynamics. As Bateson pointed out, words are not the same as the things, etc. that they 'stand for'. The things, events, etc. we perceive in the world and our systems of interpretation for understanding these are not the same. They represent two distinct orders of logical typing.

There are always higher systemic levels which regulate therelations between a system and its environment. It is for this reason that a language system attains a high level of self-organization. However, this should not be taken to mean that it is autonomous with respect to its environment. The reason why has to do with the self-referential nature of such systems. The decision as to what counts as the metalevel of interpretation in relation to the lower levels always implies a choice from a constrained system of possible alternatives. These may be formalized as a paradigmatic set of options at any given level in the overall hierarchy. The choice of one alternative rather than some other always implies criteria of agency. That is, the emergence of a recursive hierarchy of relations entails an agent who may intervene in this hierarchy so as to adjust it to suit the agent's own purposes. An agent does not only use a given system of interpretation in this way. He or she always observes him- or herself, to varying degrees of awareness, in the process of doing so. This means that the progressive differentiation of a givensystem of interpretation from the world -- i.e., its functional closure -- simultaneously implies a continual process of self-observation, self-monitoring, and self-awareness on the part of the agents involved. That is, the possibility of higher- or metalevel norms entails possibilities of choice from among alternatives. Consequently, there also exists the possibility of constructing alternative interpretations and hypotheses concerning the world we share with others.

None of this alters the fact that such higher-order norms may, as Stuart Shanker (1987: 640) points out, be performed unreflectingly. An agent may be unconscious of the norm or the agent may perform the action in an entirely mechanical and unreflective fashion. However, what matters is the agent's ability to justify or explain his action by reference to some norm (Shanker 1987: 641). This means that the emergence of a semiotic system of interpretation necessarily entails meta-level perspectives. Further, the fact that these contain self-representations of the agents who use them paves the way to increasingly more highly specifiedcriteria of self-observation and self-awareness.

6. Hjelmslev's Distinction Between Connotative Semiotic and Metasemiotic

Hjelmslev (1969 [1943]: 114-25) makes a distinction between two types of higher-order semiotic, which he designates as connotation and meta-semiotic. Both a connotative semiotic and a meta-semiotic are second-order systems which stand in a specific kind of relationship to their respective first-order systems. Both arise from the stratal nature of language and may be seen as an early attempt to formalize the various types of higher-order contextualization involved in semiosis:

... a connotative semiotic is a semiotic that is not a language, and one whose expression plane is provided by the content plane and the expression plane of a denotative semiotic. Thus it is a semiotic one plane of which (namely the expression plane) is a semiotic.

What may be particularly surprising here is that we have discovered a semiotic whose expression plane is a semiotic. For after the development taken by logistics in the work of the Polish logicians, one is prepared for the existence of a semiotic whose content plane is a semiotic. This is the so-called metalanguage (or, we should say, metasemiotic), by which is meant a semiotic that treats of a semiotic; in our terminology this must mean a semiotic whose content is a semiotic. Such a metasemiotic linguistics itself must be. (Hjelmslev 1969 [1943]: 119-20)

For Hjelmslev, the first-order semiotic -- e.g. natural language -- may be reconstrued by the higher-order metasemiotic as a metaphenomenon in the discourse of the metasemiotic. The first-order, or denotative, semiotic, on the other hand, is concerned with the construal of the phenomena that we perceive, experience or imagine in the world. Typically, metalanguage is said to be 'language about language'. However, this formulation is inadequate for it does not specify the need for still higher orders of contextualization which tell us, for example, how a givendiscourse is meta-, and when, i.e., in which context-types (Lemke 1984: 88).

Consider the following example, taken from a pedagogical grammar of English for foreign learners:

You can say I'm being careful or You're being annoying, but not *I'm being happy or *She's being tired.

We use the progressive tenses of be (I'm being, etc.) to talk about people's activity and behaviour, but not about states of mind and feelings. (Swan 1984: 96)

Following the meta-contextualization formalism discussed in section 4, it is possible to show how the specifically meta-character of the above text can only be resolved if we postulate still higher orders of contextualization. Let us consider the expression: "You can say I'm being careful or You're being annoying, but not *I'm being happy or *She's being tired". Now, E designates the italicized locutions in the object-language and C is WHAT YOU CAN AND CANNOTSAY, according to the meta-discourse. Thus, E is redundant with C. However, this may be so in different ways, or according to different principles. If we take the entire sentence to be E', then it is necessary, as Lemke shows, to postulate a still higher-order context above E' in order to specify how E and C contextualize each other in this particular metadiscourse. In this case, the relevant meta-discourse is clearly indexed in the surrounding co-text, as well as elsewhere in the same book. The metalevel explanation used by the author concerns the kind of English that he assumes to be most generally used by native English speakers -- British or American -- rather than foreign speakers of English (see Swan 1984: ix).

A further level of meta-contextualization concerns the social contexts in which the meta-discourse itself is used. In the present case, this depends on notions of the text-book as authority describing idealized uses of the language, rather than actual instances. There is a strong expectation of standardized models of English, as mediated by theauthority of the teacher. The language is broken up into discrete rules to be mastered, rather than presented as an organized whole (Swan 1984: 126). The implied pedagogical context is one in which the student is expected to master discrete rules, constructed as step-by-step procedures to be followed, though without reference to specific situation-types in the culture of speakers of English.

Consider now a second example of meta-linguistic discourse. This example was spoken by the male presenter of a talk-back radio program to a young woman who called in to given her opinions on the topic under discussion -- the views of the Archbishop of Sydney concerning the place of women in the home and the church.

Female Caller: ... and I believe that if it wasn't for women's lib that I would be treated as a woman more fully that's what I want to be treated as

Male Presenter: but what does being treated more fully as a woman mean?

In this case, the responsibilty for providing a metalinguistic gloss on the locution referred to by the presenter is shifted from speaker to addressee. Again, we can say that if E is the locution being treated more fully as a woman, then C is the meaning or the symbolic value of the locution that the speaker is asking his interlocutor to provide. The lexicogrammatical features that are selected to form this question indicate that the speaker is seeking an interpretation of a linguistic event, rather than a material one. In the present case, the question arises at this particular point in the exchange because of a local mismatch between the interpretative procedures of the two interactants. The questioner seeks some meta-linguistic clarification or explanation of the meaning that his addressee attributes to the locution in question. In so doing, he tries to make explicit the ideology and the values which are implicit in the female caller's discourse. The presenter's question is a metalinguistic strategy for explicitly connecting with more global, or meta-, criteria of contextualization when the local interactional procedures have produced, as here, a local mismatch between the viewpoints of the two speakers.

Metalinguistic discourse is not simply a calculus of axioms which defines all and only all the objects of a given metalanguage (Droste 1983). Typically, metalanguage is seen as a taxonomic hierarchy of terms, either folk or scientific, when what is needed is an account of the ways in which metalinguistic discourses are themselves operative in particular context-types. The above example draws our attention to the need for a more dynamic, praxis-oriented approach, rather than a static, taxonomizing one. That is, we need to investigate the relations among the local inter-actional context, the metasemiotic consciousness of the interactants, their always partial viewpoints, and the ways in which the interaction of all of these perspectives serves to bring into or out of focus particular metalinguistic forces of a given utterance, as construed from some social viewpoint. The above example illustrates that that which is 'meta-'about discourse means that participants can construe higher-order relations across local occasions. In other words, the construal of such relations is a means of reconstructing the global system -- the metasystem -- of possible contextualizations of a given utterance, relative to some social viewpoint. There is no viewpoint that stands outside the relevant system of relations and the notion of 'meta-' should not be taken to refer to a neutral or objective point of view. Does 'being treated like a proper woman' mean, for instance, women should be feminists?, they should have independent careers?, they should stay at home and mind the children?, or they should act as sweet and docile wall flowers?

7. Multimodality and Metalanguage

Increasingly, a growing number of semioticians are paying attention to the fact that human social meaning-making rarely, if ever, deploys the resources of a single semiotic system such as language. Generally speaking, the participants in a given occasion of discourse orchestrate intoa single semiotic action structure the resources of two or more semiotic systems. Linguistics has concentrated on the analysis of language in isolation from its co-deployment and co-evolution with other semiotic modalities. This is reflected in the kinds of scientific metalanguages which have evolved in the western tradition since the time of Plato and Aristotle. Consider the flow of speech sounds. Linguistics has deployed the specialized metalanguages of phonetics and phonology so as to analyse the sounds of speech into distinctive phonetic features, e.g., [+ nasal], and phonological units such as the phoneme and the syllable. The analytical procedures of phonetics and phonology allow the practitioners of these disciplines to determine that any given number of speakers may be uttering a sound belonging to the same general class, or phonological category. However, this essentially categorical mode of analysis puts to one side many other linguistically non-salient features that may be relevant to the contextual-ization of the speech event. Thesemay include the affective states of the speaker, individual differences in voice, and so on.

But what of our metasemiotic resources for talking about language and the other social semiotic resource systems and practices? Is verbal metalanguage alone adequate or even desirable for describing the theoretical abstractions which linguists use to model language? How can a purely verbal metalanguage fully or adequately describe graded or continuous phenomena, varying scalar levels and their interrelations, dynamically emergent phenomena, and complex hierarchical levels of organizatio in both language and other semiotic modalities? Let us take the notion of stratification in linguistic theory as an example. In section 4 above, I introduced this notion in connection with the question of meta-levels of contextualization. For those linguists who view language as a stratified system, the following properties may be identified:

  1. he system consists of an unbounded hierarchy of scalar levels, going in both directions with reference to some focallevel;
  2. the units and structures at any given level in the hierarchy entertain principles of organization that are specific to that level;
  3. there is no direct or causal relation linking the units and structures on any given level and those on some other level; instead, each stratum mutually constrains the other strata, as well as contributing its own specific dynamics to the whole;
  4. the further apart two strata are, the weaker the constraints they impose on each other. Thus, lexicogrammar and semantics are strongly cross-coupled in this sense; semantics and phonology or graphology much more weakly so;
  5. no level is reducible to some more essential reality at the level below it. Many of the characteristics cited in (1) to (5) above involve both topological and multi-dimensional criteria. These maynot be adequately described by the primarily digital and categorical criteria of a purely verbal metalanguage. Now consider the diagram presented in Figure 2 (over). This diagram attempts to communicate all of the features referred to in (1) to (5) above using the combined resources of at least three semiotic systems, viz. the visual grammar of the abstract diagram, the graphological resources of print, and the lexico-grammatical and semantic resources of language In so doing, this diagram is an instance of a specific multimodal genre of text for describing abstract scientific concepts, viz. in this case, the metatheo-retical concept of stratification in language.

    First, I shall consider the use of the circle in this visual metalanguage. Circles paradigmatically contrast with angular forms such as squares and rectangles in the grammar of the visual. The curved form of the circle is associated with organic wholeness and harmony. The centre of the circle constitutes the focal point for the concentric rings that 'emanate' from this. However, in Figure 2 the three innermost circles do not emanate concentrically from a shared centre of focus owing to the asymmetric disposition of the circles in relation to each other. Note, too, how all four circles intersect at a common point on the circumference of the outermost circle. These two features suggest: (1) the hierarchical nature of the relations at each level does not entail that any given level is necessarily more important than the others; (2) the four stratal perspectives on language that are represented in the diagram intersect with each other in the creation of the whole.

    The naturalness and wholism of circular forms also suggest here the unbounded and continuous nature of the stratal relations that are represented. It also plays down conflict. No level is inherently more important that the others. However, a given level may be selected as the reference point for a given theoretical purpose. This is the focal level. In Figure 2, the lexicogrammatical stratum is focal in this sense. The useof a larger font size and bold type, along with the curved shape of the word LEXICOGRAMMAR, are instances of specific visual-graphic conventions which the makers of such diagrams use in order to signal the relative prominence of some features rather than others. At the same time, the curvature of the word LEXICOGRAMMAR reminds us that the lexicogrammatical stratum of language is not a reified order, separate from the whole of the language phenomenon, as presented in this diagram. The focal status of this stratum is also suggested by the vectorial nature of the three arrows that are internal to the diagram. Each of these originates from a given stratum other than the lexico-grammatical and converges to a focal point. It is this focal point which coincides both with the centre of the overall figure and, as it happens, the lexico-grammatical stratum. The convergence of these three vectors on this focal point thus coincides with the selection of the lexico-grammatical stratum as the focal point of the diagram and the linguistic discussion that accompanied it.

    The co-deployment of the resources of a number of different semiotic systems produces a complex meaning which could not have resulted if each semiotic modality remained isolated from the others. Thus, the copatterning of the visual regions enclosed by any given circle with the nominalized labels associated with it -- viz. 'contextual view', '(discourse) semantic view', etc. -- represents some of the theoretical objects in this particular scientific metalanguage. The vectors that converge on a focal point and the fore-grounded graphological status of the word LEXICOGRAMMAR construct a specific viewpoint for the reader of the diagram. That is, the overall system of relations in the diagram is to be seen from the perspective of the focal level -- the lexicogrammatical stratum. Further, the embedding of the four circles, the one within the other, suggests the scalar nature of the units and relations at any given level with respect to those on the other levels. The outwardly expanding effect of the circles additionally suggests theopen-ended and potentially infinite nature of the orders of scalar relations that are presented. The schematic character of the diagram interacts with the absence of specific deixis in the nominalized labels to suggest the nomic status of the theoretical statement that the diagram gives voice to. The diagram, as a metatheoretical statement, claims that the property of stratification generally holds for any instance of natural language constituted as its theoretical object.

    What the diagram does, through the co-deployment of the diverse semiotic resources discussed here, is to reconstrue natural language phenomena as a metaphenomenon in terms of the theoretical concepts of the object language -- e.g., theoretical concepts such as strata, context, discourse semantics, lexicogrammar, and so on -- in order that these and the theoretical relations that hold between them may be subject to metatheoretical reflection. The object-language is not, then, natural language per se (Droste 1983). Rather, it is the theoretical categories which the metasemiotic constitutes as the objects of its own theoretical activities. Thediagram does not directly refer to natural language as such. Its function is to selectively reconstitute the objects of the theory in specifiable relations to each other. The multi-modal nature of this metasemiotic shows how the scalar relations involved in stratification are more flexibly enhanced than would be the case in a monomodal metasemiotic.

    8. What is 'meta-' about metalanguage?

    The 'meta-' character of metalanguage -- i.e., language used to talk about itself -- means that metalanguage is always at least two removes from the phenomena of experience. If language is used to construe the phenomena of experience -- c.f. Bateson's territory -- then metalanguage is used to construe the metaphenomena of language itself -- c.f. Bateson's map. As Halliday (1988: 32) points out, the metaphenomenon itself comprises two levels, or strata, of explanation, one grammatical and the other semantic. This means that metalanguage is required to interpret not one but at least two levels of linguistic reality. One of these is the lexicogrammatical form; the other is the meaning this has.

    This is revealed by the analysis in Figure 1. Take the plural morpheme suffix in the word cats. This is a unit of lexicogrammatical form, a grammatical category, viz. [PLURAL]. In turn, we can gloss this as a semantic category, say, [MORE THAN ONE], in an attempt to widen the interpretative circle. The semantic category so derived it itself an interpretation of the phenomena of experience -- real or imagined -- denoting more than one of a thing (Halliday 1988: 32). Halliday formulates this problem as follows:

    The grammatical category of 'plural' was set up in the first place to account for a morpho-logical phenomenon: suppose this had been in English, then the -s / -z / -iz of cats, dogs and horses. At this point, therefore, we ought to have come round in a circle: -s / -z / -iz means -s / -z / -iz. But instead we have tried to escape from the circle by finding a gloss for -s / -z / -iz -- that is, an exact synonym for it, in natural languagewording; and that is an extremely difficult thing to do. We might try glossing it as more than one, or several, or many; but the trouble is we don't actually say I like more than one cat, or I like many cat -- we say I like cats. The meaning of the -s on cats is impossible to gloss in natural language, except by means of itself. The category is, quite simply, ineffable. (Halliday 1988: 32-3)

    The dual status of the meta-phenomenon -- grammatical and semantic -- is also expressed in many different ways by the internal design of the grammar of natural language. That is, language or, better, its lexicogrammar, construes its own processes of meaning-making (Rumsey 1990; Matthiessen 1991). One example which shows this quite clearly is the distinction the grammar of, say, English makes between the quoting of the 'actual' words that someone is presumed to have said and the reporting of someone's thoughts. Compare the following:

    1. He said, "John is coming"
    2. He believed that John was coming
    In (1) the deictic perspective is that of the grammaticalized Sayer (He), i.e. the one who is construed as having uttered the quoted clause. The focus is on the actual words that the Sayer is presumed to have uttered. In (2), there is no such presumption of an observed speech event. Therefore, there are no actual words spoken by someone and that can be quoted. In this case, the deictic centre is the speaker of the entire expression. The focus is on the meaning the speaker attributes to the He, rather than on anything the He is supposed to have said.

    Both (1) and (2) show just one area in which the grammar interprets or models its own processes of meaning-making. The grammar, in other words, is internally designed to interpret its own processes of meaning-making (Matthiessen 1991). The difference between (1) and (2) -- quote vs. report -- further illustrates the dual status of language as metaphenomenon. That is, (1) focusses on theexact words that someone is presumed to have uttered on some occasion and recodes this as a quoted locution. Consequently, the focus is on the actual wording, or the lexicogrammatical forms, that were used. In (2), on the other hand, the focus is not on the actual words used, but on the semantic meaning the speaker attributes to the He. This shows how (1) and (2) are meta-linguistic in different ways. Both are metalinguistic glosses of linguistic phenomena, the one focussing on the level of lexicogrammatical form, the other on the level of the semantics. The grammar itself tells us that metalanguage can be 'meta-' in different ways.

    Following the meta-contextualization formalism first discussed in section 4, it is possible to show how the solution to the problem posed in the preceding paragraph requires that the analyst postulate still higher orders of contextualization. For example, if He said, "John is coming" is E', then E is the quoted locution and C is the Sayer's linguistic construal of E in the quoting clause. Again, it isnecessary to specify how E and C contextualize each other, and in relation to which metadiscourses. In the present, invented example, the lower levels of the hierarchy are in fact specified by the lexicogrammatical distinction between quoting and reporting. In (1), for instance, the expression E' is, then, a metalinguistic construal of the actual words someone is presumed to have said on some occasion. But note that C is always a re-contextualization of the original speech event by E. C is not identical to the 'original' event. Rather, it is taken as standing in a relationship of formal identity to the explicit verbal component of that event. This raises the question as to the various possible ways in which E' is metalinguistic.

    In what way did the Sayer, according to the speaker (E), utter C? Was the quoted locution said in a manner appropriate to the overall speech event? Was the Sayer 'sincere' or 'truthful'? Which kinesic, somatic, and other nonverbal semiotic modalities accompanied the Sayer's uttering of the quoted locution? To whom was it addressed? What moral orother authority did the Sayer have to say it? In which context-type was it spoken?

    Now, the metalinguistic force of E' in turn depends on further levels of contextualization, according to situation-type. Does E' occur in a journalistic report of political or other events? In the context of a witness giving evidence in court? A conversation between intimates? A gossip session between neighbours? A spy reporting to a superior? A private investigator reporting to a client? A confession to one's priest?

    The analysis presented above reveals the same point made by Halliday in the quotation discussed earlier. There is no exact correspondence between E and C in our example. In using language to gloss language, this does not mean we are locked into a pathologically tight circle of interpretation. This problem would arise only if language simply reflected an independent external reality (see also Halliday 1988: 39). But language does not work like this. Instead, each level in thehierarchy of meta-contextualizing relations recursively re-construes the others. Thus, the semantics of natural language does not refer to a set of natural kinds existing independently of language. For this reason, Halliday points out that the categories internal to natural languages are 'ineffable' -- there is no ultimate, language-independent schema whereby they may be interpreted in a definitive way (1988: 39). Rather, the semantics of natural language is internal to language itself. It acts as the interface whereby language is used to interpret the world whenever the language system and the world are brought into contact in some act of meaning-making. This fact means that both language and its meta-languages enable us to construct new alternative theories of both the first-order phenomena of the world we experience and the metaphenomena -- grammatical and semantic -- which we use to act on and interpret the former. We are no more prisoners of language than we are of our metalanguages. Furthermore, the widening of the metasemiotic circle of interpretation toinclude other semiotic modalities helps to ensure against the pathological narrowing of the metalinguistic circle that Michael Reddy (1979) has identified.

    Reddy showed how the pathological ramifications of what he called the 'conduit' metaphor of language ramify through both our folk-theoretical and our scientific explanations of language form and function. These explanations assume, by default, that the basis on which language form and function are to be explained lies outside language itself in some nonlinguistic realm of the 'thoughts', 'ideas', or 'information' that language form, according to the conduit metaphor, simply packages up and transports from the sender to the receiver of the message. Language, in this view, is assumed to be a neutral pipeline, or 'conduit', for doing so. For our present purposes, it doe not matter whether these nonlinguistic 'thoughts' and so on are taken to exist in the mind of the speaker or in the world 'out there'. A metalanguage which is invaded by the conduit view oflanguage limits itself to this view. It assumes that the explanation of language form and function is a causal-deductive one based on the causal mechanisms that link language to extra-linguistic phenomena. In this way, the non-linguistic 'thoughts' and so on constitute a causal explanation of language form and function without, however, being subjected to metalinguistic critique. They cannot be so critiqued for they lie outside language itself and, hence, outside the domain of metalanguage.

    Now, Gdel established in his meta-mathematics that a formal theory cannot include itself in its own domain without bringing in unwanted inconsistencies (Lemke 1984: 72-3). Nevertheless, metalanguage can serve as a guide to a self-reflexive and critical praxis (Lemke 1984: 72-3; Thibault 1991: chaps. 2-4). The previous discussion of quoting and reporting indicates just one area in the grammar of English where the grammar itself, on the basis of the accumulated linguistic work of its users, provides its own metalinguistic models of meaning-making. Both quotes and reports show that the Eand C levels of analysis do not belong to the same domain. Rather, the C level -- the quoting or reporting frame -- is a self-reflexive meta-level in and through which we may observe and interpret our own and others' linguistic praxis. That is, language provides us, ready-made, the means to be the self-reflexive observers of our own and other's practices. In so doing, it provides its users with the resources for creating a new hybrid context of the two perspectives -- one that can, perhaps help us to transcend the limitations of either, considered separately (Thibault 1991: chaps. 2-4). A further widening of the circle also means, as we have seen, that the praxis of metalinguistic inquiry is free to fashion new meta-semiotic tools out of the very multimodal resources that always characterize any act of meaning-making.

    9. Emergent Properties of Metasemiosis

    If metasemiosis is an emergent property of the lower-order systems it comments on, then it seems reasonable to postulate that (1) it shares significant properties in commonwith these; and (2) it bears traces of its emergence from historically prior modalities of semiosis. With respect to its lower-order systems, a metasemiotic is a much more highly specified system of relations. Evidence for this may be found in the ritualistic behaviour of some animal species. Can these provide clues as to incipient metasemiotic capacities?

    Let us consider Konrad Lorenz's famous discussion of the emergence of ritualized forms of activity in the inciting behaviour of the female a number of European species of duck. Lorenz showed how, in these species, ritualistic forms of behaviour emerge from original, purely fixed behaviour patterns and, in time, take on symbolic functions. Lorenz's point was that the original phylogenetic function became modified as the result of the imposition of further symbolic or ceremonial constructs that are different in kind from the original behaviour pattern that the new, ritual pattern emerged from:

    The original behaviour differs from case to case according to the varying force of each separate, independently variableimpulse; the newly arisen, fixed motor co-ordination represents only one stereotyped average case. This has now become 'schematized' in a manner strongly reminiscent of symbols in human cultural history. (Lorenz 1969: 51)
    As Lorenz pointed out, the "newly arisen fixed motor pattern does not suddenly become preponderant but exists first beside the unritualized model over which it predominates only slightly" (1969: 51). The supervenience of the new ritual function does not replace the old one, but is integrated with it in a new hierarchy of functions. The new ritualized behaviour 'copies' the earlier form and in so doing it operates as a further constraint on it. The emergence of the new, fixed pattern or schema takes on proto-semiotic characteristics.

    These are as follows. First, the ritualized schema is replicable; it has generic -- 'schematized' -- properties such that the same schema is instantiated across many specific performances by different individuals. Secondly, the different components of the overall activity sequence do not randomly follow one another. Instead, they constitute functioning parts of an overall whole to which they belong. The consolidation, both phylogenetically and onto-genetically, of the newly emergent ritual activities leads in turn to the enhanced symbolic capabilities of the species in question. The emergence of the higher-order symbolic or ritual level means that the system as a whole -- viz. the forms of incipient social organization of these species -- becomes more tuned into environmental regularities on account of the new environmental information which is stored in the newly emergent level of symbolic organization.

    Further support for this perspective is suggested by the ways in which human speech emerged from prior gestural forms of communication, without, however, totally transcending these. Instead, the more highly specified system of natural language resulted in a new integration of this with the earlier non-linguistic semiotic modalities. Tanner (1994: 9) suggestshow the newly emergent capacity for vocal speech in hunter-gatherer societies derived its own metalanguage from the observed properties of the prior non-vocal forms of communication that were progressively integrated with language itself. The modifications were two-way, as evidenced by the quite uniform way in which, as Tanner's etymological evidence suggests, the various words signifying the activity of talking in Sanskrit, Greek, Latin, and Hebrew both (1) provide a meta-commentary on the gestures from which the words themselves were derived at the same time that (2) gesture directly modified the terms of speech, and in ways that remain sedimented in the etymology of both classical and modern languages. This seems hardly surprising since the newly emergent system of vocal communication would be less precisely determined and for this reason more inclined to organize itself and, hence, to model itself along the lines provided by the already existing non-vocal system. As the new vocal system developed andbecame more highly specified, so, too, did it become more effective in elaborating its own models of itself as it tuned into more and more newly contingent ecosocial demands on its expanding and evolving resources. There are compelling reasons to believe that the pre-verbal and pre-human bases of metasemiosis provide important clues as to how the interpretative circle connecting language to metalanguage can be widened to take account of the co-evolution and co-deployment of language with other semiotic modalities.


    Aristotle. "On interpretation". Trans. Harold P. Cooke. In Aristotle in Twenty-Three Volumes, I, 111-79. Loeb Classical Library. London: Heinemann, 1983 [1938].

    Bateson, Gregory. "A theory of play and fantasy". In Steps to an Ecology of Mind, 150-66. London and New York: Paladin, 1973.

    Davies, Paul.The Mind of God. Science and the search for ultimate meaning. Harmondsworth, Middlesex: Penguin, 1993.

    Droste, Flip G. "Reflections on metalanguage and object-language". Linguistics 21, 675-99, 1983.

    Halliday, M.A.K. "On the ineffability of grammatical categories". In Ed. James D. Benson, Michael J. Cummings and William S. Greaves, 27-51. Linguistics in a Systemic Perspective. Amsterdam and Philadelphia: John Benjamins, 1988.

    Hjelmslev, Louis. Prolegomena to a Theory of Language. Trans. Francis J. Whitfield. Madison, Wilwaukee and London: University of Wisconsin Press, 1969 [1943].

    Kant, Immanuel. Critique of Pure Reason. Trans. Norman Kemp Smith. London and Basingstoke: Macmillan, 1970 [1781].

    Lemke, Jay L. "Action, context, and meaning". In Semiotics and Education (monograph), 63-93. Victoria University, Toronto: Toronto Semiotic Circle Monographs, Working Papers and Prepublications, No. 2, 1984.

    Lorenz, Konrad. On Aggression. Trans. Marjorie Latzke. London: Methuen, 1969 [1963].

    Matthiessen, Christian. "Language on language -- the grammar of semiosis" Social Semiotics 2, 1, 1991.

    Plato. The Republic. Trans. H. D. P. Lee. Harmondsworth, Middlesex: Penguin, 1973 [1955].

    Reddy, Michael J."The conduit metaphor -- a case of frame conflict in our language about language". In Ed. Metaphor and Thought. Cambridge and Melbourne Andrew Ortony, 284-324. Cambridge University Press, 1979.

    Rumsey, Alan. "Wording, meaning, and linguistic ideology". American Anthropologist Vol. 92, 2, 346-61, 1990.

    Shanker, Stuart. "Wittgenstein versus Turing on the nature of Church's thesis". Notre Dame Journal of Formal Logic 28, 4, 615-649, 1987.

    Swan, Michael. Practical English Usage. Oxford and Melbourne: Oxford University Press, 1984.

    Tanner, R. G.. "Aristotle as a structural linguist". T ransactions of the Philological Society, 99-164, 1969.

    --- "Linguistic evidence for pre-vocal communication".Department of Classics, University of Newcastle, N.S.W.: Mimeo, 1994.

    Thibault, Paul J. Social Semiotics as Praxis. Text, social meaning making, and Nabokov's "Ada". Theory and History of Literature series, Vol. 74. Minneapolis and Oxford: University of Minnesota Press, 1991.

    --- Re-reading Saussure. The dynamics of signs in social life. London and New York: Routledge, 1996.

    Paul J. Thibault is an associate professor in English language and linguistics in the Dipartimento di Letterature e Civiltà Anglo-Germaniche at the University of Venice (Italy). He has taught and researched in linguistics, semiotics, literary theory, educational linguistics, and English language at the Universities of Padova, Verona, Bologna, and Sydney and Murdoch University (Western Australia).

    Go to Semiotic Review of Books Home Page
    Go to SRB Highlights
    Go to SRB Archives