What happens when I produce a sign or a string of signs? First of all I must accomplish a task purely in terms of physical stress, for I have to ‘uitter’. Utterances are usually considered as emissions of sounds, but one may enlarge this notion and consider as ‘utterances’ any production of signals. Thus I utter when I draw an image, when I make a purposeful gesture or when I produce an object that, besides its technical function, aims to communicate something.
In all cases this act of uttering presupposes labor. First of all the labor of producing the signal; then the labor of choosing, among the set of signals that I have at my disposal, those that must be articulated in order to compose an expression, as well as the labor of isolating an expression-unit in order to compose an expression-string, a message, a text. Fluency or difficulty in speaking, insofar as it depends on a more or less perfect knowledge of linguistic codes, must be examined by semiotics, although I do not propose to go into the matter here. Rossi-Landi (1968) has dealt with this aspect of performance.
Suppose now that, instead of uttering words, I draw an image corresponding to an object, as when I draw a dog in order to advise people to ‘beware’ of the dog in my garden. This kind of sign-vehicle production seems to be rather different from choosing the word /dog/. It implies extra work. Moreover, it might be pointed out that, in order to say /dog/, I had only to choose among a repertoire of established types, and to produce a single occurrence of that type, while in order to draw the image of a dog I have to invent a new type. Thus there are different sorts of signs, some of them entailing a more laborious mode of production than others.
Finally, when I ‘utter’ words or images (or whatever else), I have to labor in order to articulate them in ‘acceptable’ strings of sign-functions; thus I have to labor on their semantic acceptability and understandability. In the same way, when receiving a sentence, even though I do not have to labor in order to produce the sign-vehicles, I do have to labor in order to interpret them. Obviously I can send my messages in order to mention things and states of the world, in order to assert something about the organization of a given code, in order to question or to command. Either to send or to receive these messages (or texts) requires that the sender should foresee, and the addressee isolate, a complex network of presuppositions and of possible inferential consequences. In exchanging messages and texts, judgments and mentions, people contribute to the changing of codes. This social labor can be either openly or surreptitiously performed; thus a theory of code-changing must take into account the public reformulation of sign-functions and the surreptitious code-switching performed by various rhetorical and ideological discourses.
Many of these activities are already studied by existing disciplines; others will have to constitute the object of a new general semiotics. But even those already studied by pre-or extra-semiotic disciplines will then have to be included as branches of a general semiotics, even if it proves convenient to preserve their present affiliation for the time being.
Whereas a theory of codes was concerned both with the structure of sign-function and with the general possibility of coding and decoding, a theory of sign production will thus be concerned with all the problems outlined in Table 31. This table concerns the kind of labor required in order to produce and interpret signs, messages or texts (physical and psychological effort in manipulating signals, in considering, or disregarding, the existing codes; time needed, degree of social acceptance or refusal, energy expended in comparing signs to actual events; pressure exerted by the sender on the addressee, and so on).
The interconnecting arrows linking the various kinds of labor try to correct the oversimplification due to the bi-dimensional format of the diagram; each kind of labor interacts with many others and the process of sign production – in its relationships with the life of codes – represents the result of a network of interacting forces. On the right side are listed the various different approaches that may be applied to the different areas of study, and are in fact actually adopted irrespective of the general semiotic framework that the table proposes. The existence of such a diversity of approaches should not be regarded as a methodological limitation for semiotics; it must simply be listed among the so-called ‘political’ boundaries mentioned in 0.4.
Let us now examine the items of Table 31 one by one.
(i) There is a labor performed on the expression continuum in order to physically produce signals. These signals may be produced as mere physical entities without semiotic function; but as soon as they are produced – or selected among pre-existing entities – as the expression-plane of a sign-function, their mode of production directly concerns semiotics. They may be either already segmented discrete units or material clusters somewhat correlated to a content. In both cases their production presupposes different modes of labor or different techniques of production. These modes of production will be listed in Table 39.
(ii) There is a labor performed in order to articulate expression-units (either already established by an expression system or proposed as the somewhat segmented functives of a new coding correlation). This kind of labor concerns the choice and the disposition of sign-vehicles. There can be expression articulation during the act of constituting (or making) an innovatory code; during a discourse in which the senders try to observe all the laws of the existing codes; within a text where the sender invents new expression units, therefore enriching and changing the system (for example when Laforgue invents the word ‘violupté’ or Joyce ‘meandertale’; see for instance Eco, 1971). Obviously the modification on the expression-plane must be correlated with a modification on the content-plane, otherwise it becomes mere grammatical nonsense; therefore the labor of system observing, system making and system changing on the expression-plane must be considered in relation to the corresponding labor on the content-plane, through the mediation of a labor on the correlation of functives (item iii).
(iii) There is labor performed in order to correlate for the first time a set of functives with another one, and thus making a code; an example of such code making is given by the operation constituting the Watergate Model in chapter 1.
Table 31 Labor presupposed in the process of sign production
(iv) There is labor performed when both the sender and the addressee emit or interpret messages observing the rules of a given code, as in the case of ‘common’ semiotic acts such as the expression /the train from London will arrive at 6.00 P.M./. This kind of disambiguation of expressions was dealt with in chapter 2.
(v) There is a labor performed in order to change the codes shared by a given society. It is a complex process which involves both semiotic and factual judgments (see 3.2.) and other forms of textual manipulation; in this sense it directly involves the aesthetic manipulation of codes (see 3.7.).
(vi) There is labor performed by many rhetorical discourses, above all the so called ‘ideological’ ones (see 3.9.) in which the entire semantic field is approached in apparent ignorance of the fact that its system of semantic interconnections is more vast and more contradictory than would appear to be the case. In order to avoid openly acknowledging the contradictory nature of the Global Semantic System (see 2.13.) ideological discourse must switch from one code to another without making the process evident. Code switching is also performed in aesthetic texts, not as a surreptitious device but as a manifest procedure, in order to produce planned ambiguities and multi-levelled interpretations (see 3.7.1.).
(vii) There is labor performed in order to interpret a text by means of a complex inferential process. This process is mainly based on abductions and produces forms of overcoding (on the basis of a first level of pre-established rules new rules are proposed which articulate more macroscopic portions of the text) and of undercoding (in the absence of reliable pre-established rules, certain macroscopic portions of the text are assumed to be the only pertinent units even though the more basic combinational rules and their corresponding units remain unknown). To this important aspect of text interpretation the whole of section 14 of chapter 2 has been devoted.
(viii) There is labor performed by both the sender and the addressee to articulate and to interpret sentences whose content must be correctly established and detected. Section 2 of the present chapter will deal with these propositions (such as meta-semiotic, eternal and standing propositions) commonly called ‘statements’, while section 3 will deal with the index-sensitive sentences used in mentioning or referring, and will therefore also deal with many problems regarding item (ix).
Semiotic judgments predicate of a given semiotic item what is already attributed to it by a code (see 3.2.). They can assume three forms: (a) meta-semiotic propositions presupposing a ‘performative’ format («I state that from now on the word ‘ship’ can also be applied to vehicles for space-travel»); (b) eternal propositions of the type «bachelors are males»; (c) index-sensitive propositions coupling certain objects, taken as representative of a bunch of properties, to certain words («this object is a pencil»); this last kind of semiotic judgment, insofar as it is pronounced about actual objects, is also called a ‘mentioning or referring act’ and can be studied under the same profile as index-sensitive factual propositions (see 3.3.1.).
Factual judgments predicate of a given semiotic item what was not attributed to it by the code. This judgment can be of two types: (a) index-sensitive propositions that attribute to a token occurrence of a semiotic type a factual property that, by definition, does not pertain to other ‘tokens’ of the same ‘type’ (“this pencil is black”); this kind of judgment, otherwise called ‘occasion proposition’ (see 2.5.3.), does not modify the semantic representation of a given semiotic item and in this sense it could be left aside by a semiotic inquiry, being better dealt with by a theory of the extensional verification of correspondences between propositions and states of the world; but all the same it has some semiotic purport insofar as, in order to predicate a semantic item as the property of an object, one will need a survey and a definition of this object’s properties and such an operation has a semiotic aspect (see 3.3.3. – 3.3.6.); (b) Standing non index-sensitive propositions like “The moon has been walked on by human beings”: as will be seen in 3.2.2, this kind of judgment when pronounced for the first time is a factual index-sensitive proposition (something is predicated of a given semantic item that no code attributed to it, and it is asserted for the first time by means of an indexical device of the type “in this moment” or “from now on”); but when these judgments are accepted by a society as true, then they assume a meta-semiotic function and gradually become semiotic judgments.
(ix) There is labor performed in order to check whether or not an expression refers to the actual properties of the things one is speaking of. This labor is strictly linked to the one performed in order to grasp the content of the index-sensitive semiotic and factual sentences, or mentions. To this problem section 3 is devoted.
(x) There is labor performed in order to interpret expressions on the basis of certain coded or uncoded circumstances. This labor of inference is linked both to the inferential labor required in order to understand something (thereby becoming the proper concern of a theory of perception and intelligence) and to the inferential labor performed within the text (see vii) which must be considered as an aspect of the labor of over-and undercoding (see 2.14.).
(xi) There is labor that the sender performs in order to focus the attention of the addressee on his attitudes and intentions, and in order to elicit behavioral responses in other people. This kind of labor (which will be considered in many of the following sections) was usually studied by the so-called theory of speech acts. Provided that in the present perspective the notion of ‘speech act’ is taken as concerning not merely verbal acts but every kind of expression (images, gestures, objects), it may be noted that among these various communicational acts figure not only the so-called locutionary ones, which may correspond to semiotic and factual judgments, but also all those types of expression that do not express any assertion but on the contrary perform an action or ask, command, establish a contact, arouse emotions and so on (illocutory and perlocutory acts).(1)
The present chapter 3 will not deal with all problems concerning a theory of sign-production; it will only deal with such specific problems as require direct, immediate and exclusive attention from a semiotic point of view. Let me stress the order of priorities which governs the organization of the eight following sections.
The labor performed in shaping the expression-continuum in order to produce the concrete occurrence of a given sign brings into immediate evidence the fact that there are different kinds of signs. If a general theory of codes, providing the notion of sign-function along with the notion of segmentation of both the expression and the content levels, seemed to offer a unified definition for every kind of sign, the concrete labor of producing these signs obliges one to recognize that there are different modes of production and that these modes of production are linked to a triple process: (i) the process of shaping the expression-continuum; (ii) the process of correlating that shaped continuum with its possible content; (iii) the process of connecting these signs to factual events, things or states of the world. These processes are strictly intertwined; once the problem of shaping the continuum is posed, that of its relationship with the content and the world arises. But at the same time one realizes that what are commonly called types of sign are not the clear-cut product of one of these operations, but rather the result of several of them, interconnected in various ways.
One also realizes that there are some signs that seem better adapted to the expression of abstract correlations (like symbols) and others that would appear to be more useful in direct reference to states of the world, icons or indices, which are more immediately involved in the direct mentioning of actual objects. In order to understand these points it would seem more profitable to tackle the problem of the various kinds of judgments pronounced about the world or codes and acts of mentioning things straight away. So instead of following the theoretical order outlined in Table 31 I will follow a sort of phenomenological order: in pronouncing judgments and performing mentions one discovers how one is using both verbal devices and other sorts of signs, such as for example, a pointing finger or an object taken as an example; at this point one should be able to single out both their differences and their similarities, and to realize that these differences do not characterize the various kinds of signs in themselves but rather discriminate between modes of sign production, every so-called sign being the result of many such operations.
Thus a typology of signs will give way to a typology of modes of sign production, thereby showing, once again, that the notion of ‘sign’ is a fiction of everyday language whose place should be taken by that of sign-function.
To communicate means to concern oneself with extra-semiotic circumstances. The fact that they can frequently be translated into semiotic terms does not eliminate their continuous presence in the background of any phenomenon involving sign production. In other words, signification is confronted with (and communication takes place within) the framework of the global network of material, economic, biological and physical conditions then prevalent. The fact that semiosis lives as a fact in a world of facts limits the absolute purity of the universe of codes. Semiosis takes place among events, and many events happen that no code could have anticipated. The semiotic creativity allowed by codes thus demands that these new events be named and described. The structure of codes can sometimes be upset by an innovatory statement concerning events which do not fit in with the organization of the content. What happens when messages state something concerning an as yet unorganized and non-segmented content? Does the new set of cultural units thus introduced into the social competence modify the pre-established semantic field? And how? This point prompts a return to an old philosophical distinction, widely discussed in logic and linguistic analysis, between analytic and synthetic judgments.
Considered from the point of view of a referential semantics this distinction is open to the broadest criticism. One might well wonder (cf. White, 1950) why such a statement as «all men are rational animals» is considered by traditional philosophers to be an analytic judgment and «all men are bipeds» a synthetic one. In effect, if one predicates the ‘objectivity’ of certain properties, the reason for the distinction between these two types of judgments is not evident. But Cassirer has already given an answer to this problem in Das Erkenntnisproblem in der Philosophie und Wissenschaft der neueren Zeit, II, 8, II. The analytic judgment is the one in which the predicate is contained implicitly in the concept of the subject, and the synthetic judgment is that in which the predicate is added to the subject as an entirely new attribute, due to a synthesis obtained from the data of experience.
Why then, according to Kant, is «all bodies are extensive» analytic and «all bodies are heavy» synthetic? Simply because Kant referred to the ‘patrimony of thought’ which he presumed to be known to his contemporaries. It is worth noting that «body» for him was not a referent but above all a cultural unit. And from the time of Descartes up to Newton and the encyclopedists, «extension» was attributed to this cultural unit as an essential quality which was a part of its definition, whereas «weight» was considered an accessory and contingent quality which did not therefore enter into the same definition. Judgments are either analytic or synthetic according to the existing codes and not according to the presumed natural properties of the objects. Kant explicitly states in the first Kritik that “the activity of our reason consists largely . . . in the analysis of ideas which we already have with regard to objects”. Since, however, the opposition ‘analytic vs. synthetic’ co-involves too many philosophical problems, let us develop the above suggestion within a more specific semiotic context, in this way proposing a more suitable opposition.
Let us call semiotic a judgment which predicates of a given content (one or more cultural units) the semantic markers already attributed to it by a previous code; let us call factual a judgment which predicates of a given content certain semantic markers that have never been attributed to it by a previous code. Therefore /every unmarried man is a bachelor/ is a semiotic judgment solely because there exists a conventional code which refers to a compositional tree which possesses among its markers «never married». Instead /Louis is a bachelor/ is undoubtedly a factual judgment. On May 5, 1821, /Napoleon died on Saint Helena/ constituted a factual judgment. But from that moment on, the same statement has constituted a semiotic judgment because the code has fixed in the compositional tree of /Napoleon/ the definitional connotation «died on Saint Helena». On the other hand /Napoleon, after the battle of Marengo, drank a cup of coffee/ is a factual statement that can hardly be transformed into a semiotic one. Thus White (1950), criticizing the analytic-synthetic distinction, rightly affirms that a judgment is analytic on the basis of a convention and that, when the convention changes, the judgments which were once analytic can become synthetic, and vice versa. But what he intended as a limitation of the logical distinction between analytic and synthetic is instead the condition for the validity of the semiotic distinction between semiotic and factual judgments.
I shall briefly consider a particular example of these judgments, that is, semiotic (or meta-semiotic) and factual statements, granted that these are not to be confused with index-sensitive judgments or mentions (see 2.5.3., where non-statements are called ‘occasion propositions’; mentions will be examined in 3.3.). It should be recalled that:
a) /This is a one dollar bill/ is not a statement: it is a mention (see 3.3).
b) /One dollar is worth 625 lire/ was a semiotic statement in 1971, thereby expressing a coded signifying relationship.
c) /One dollar is worth 580 lire/ was an astonishing factual statement emitted in a given day during 1972.
d) /One dollar is worth 580 lire/ became a semiotic statement of type (b) during 1972.
e) In order to make the factual statement (c) become the new semiotic statement (d) it was necessary that (c) should take the form of a meta-semiotic statement, presupposing or explicitly stating a performative formula such as: /The President of the United States (or the Bank of Italy, or the European Common Market) establishes that, from today on, everybody must accept the financial convention that one dollar is worth 580 lire/. The fact that since 1972 such a meta-semiotic statement has changed so many times only confirms yet again that many codes are very weak and transient, thus lasting l’espace d’un matin, like the rose. But a rose is no less a rose – witness Gertrude Stein – because it is so short-lived; in the same way a code is a code (is a code is a code) provided that a meta-semiotic statement has conventionally established a certain equivalence and a society has accepted it, and remains so until the arrival of another code-changing meta-semiotic adjustment.
Finally, the example of the dollar is particularly apposite, because the financial market represents a perfect case of coupling between units from different content systems, each unit being semantically defined by the opposition it entertains with every other unit. Therefore factual statements sometimes upset and restructure the codes(2).
Even though to designate these operations of content-articulation I have employed terms borrowed from logic (which is mainly concerned with verbal expressions), all these types of propositions also concern non-verbal expressions. The Encyclopaedia Britannica is a text which sets out a lot of meta-semiotic and semiotic statements not only because it records many verbal definitions of various semantic units but also because it uses drawings and photographs in order to analyze the components of the same semantic units (for example visually describing the parts of human body or the elementary components of a four-stroke engine). The New York Times sets out a lot of factual statements not only by means of words but also of photographs or diagrams.
The visual demonstration of the theorem of Pythagoras is a semiotic statement. A road signal announcing a dangerous crossing is at the same time a factual statement and a mention. Other road signals commanding one to «stop» or «beware!» or forbidding right of way are communicational acts that are listed under item (xi). The drawing of a horse with the caption /horse/ represents an index-sensitive semiotic judgment; the portrait of the last winner of the Nobel Prize with the caption /this man has won the Nobel Prize/ constitutes an index-sensitive factual judgment. The Neapolitan gesture meaning «I am hungry» is an index-sensitive factual judgment. And so on.
This dialectic between codes and messages, whereby the codes control the emission of messages, but new messages can restructure the codes, constitutes the basis for a discussion on the creativity of language and on its double aspect of ‘rule-governed creativity’ and ‘rule-changing creativity’.
Factual statements, as usually performed, are an example of creativity permitted by the rules of the code. One can verbally define a new physical particle using and combining pre-established elements of the expression-form in order to introduce something new in the content-form; one can technically define a new chemical compound using and combining pre-existing content-units in a new way, in order to fill up an empty space within a pre-established system of possible semantic oppositions; one can thus alter the structure of both the expression and the content-system following their dynamic possibilities, their combinational capacities – as if the whole code by its very nature demanded continual re-establishment in a superior state, like a game of chess, where the moving of pieces is balanced out by a systematic unit on a higher level. Thus the possibility of meta-semiotic statements which alter the compositional spectrum of a lexeme and reorganize the readings of the sememe is also based on the pre-established elements and combinational possibilities of the code(3).
Signs are used in order to name objects and to describe states of the world, to point toward actual things, to assert that there is something and that this something is so and so. Signs are so frequently used for this end that many philosophers have maintained that a sign is only a sign when it is used in order to name things. Therefore these philosophers have tried to demonstrate that a notion of meaning as separated from the ‘real’ and verifiable ‘denotatum’ of the sign, that is, the object or the state of the world to which the sign refers, is devoid of any real purport. Thus, even when they accept a distinction between meaning and referent (or denotatum) and do not equate the former with the latter, their interest is exclusively directed toward the correspondence between sign and denotatum; the meaning being taken into account only insofar as it can be made to correspond to the denotatum in specular fashion.
The theory of codes outlined in chapter 2 not only tried to restore the meaning’s autonomous status, but even deprived the term /denotation/ of any extensional or referential relevance. The foregoing section, even though it has considered factual statements, has not linked these judgments to the facts about which they are stated. What is characteristic of a factual judgment of the kind examined in the above section is that although it seems to concern facts it can also be used in order to assert non-existent factual states, and therefore to lie.
If I assert that ‘the man who invented eye-glasses was not Brother Alessandro della Spina but his cell-mate’, I do not challenge an established semiotic statement, for the inventor of eye-glasses is a decidedly imprecise historical entity and the encyclopedias are rather vague and cautious on this subject, but I do make a factual statement, or a ‘standing proposition’. It would be very difficult to check whether my judgment is true or false, and some documentation would clearly be needed; but all the same what I have produced is a factual statement (whether true or false) insofar as it does not assert something definitely recorded by a cultural code. Thus factual judgments of this type are not necessarily verified by an actual state of the world or a present entity. In this sense it is possible to assume that they have a meaning irrespective of their verification, and yet once their meaning is understood they demand verification.
Let us now consider another type of factual judgment, the index-sensitive one, i.e. the act of mentioning something actually present, as in /this pencil is blue/ or /this is a pencil/. As was suggested in 3.1., there is a difference between the two examples, and the second can be registered as index-sensitive but semiotic. Nevertheless both seem to be acts of mentioning (or of referring to) something. It may be assumed that in this case their meaning depends directly on the actual thing they refer to, but such an assumption would challenge the independence of meaning from referent maintained in chapter 2.5.
Strawson (1950) says that “mentioning or referring is not something an expression does; it is something that someone can use an expression to do”. From this point of view ‘meaning’ is the function of a sentence or expression; mentioning and referring, and truth and falsity, are functions of the use of the sentence or expression. “To give the meaning of an expression . . . is to give general directions for its use to refer to or mention particular objects and persons; to give the meaning of a sentence is to give general directions for its use in making true or false assertions”.(4) Let us try to translate Strawson’s suggestions into the terms of a theory of codes. To give general directions for the use of an expression means that the semantic analysis of a given sememe establishes a list of semantic properties that should correspond to the supposedly extra-semiotic properties of an object. If this sounds somewhat Byzantine, one could reformulate it as follows: to give general directions for the use of an expression in referring means to establish to which actual experiences certain names, descriptions or sentences can be applied. Clearly this second definition, despite its correspondence to our normal way of speaking, says very little. Moreover, one has to face the question: how does one establish the rules of such an application?
So one must return to the first formulation of the problem. But at this point a new problem arises: how does one establish a correspondence between the semantic properties of a sememe (which clearly is a matter for semiotics) and the supposedly non-semantic properties of a thing? Can the mentioned thing assume the status of a semiotically graspable entity? For, either semiotics cannot define the act of mentioning or in the act of mentioning the thing mentioned should be viewed in some way as a semiotically graspable entity. So we must re-examine the whole process of mentioning.
The act of referring places a sentence (or the corresponding proposition) in contact with an actual circumstance by means of an indexical device. We shall call these indexical devices pointers. A pointing finger, a directional glance, a linguistic shifter like /this/ are all pointers. They are apparently characterized by the fact that they have as their meaning the object to which they are physically connected. I have shown in 2.11.5. that this is not true. Any pointer has first of all a content, a marker of «proximity» or «closeness» independently of the actual closeness of an object. But for the sake of the present analysis let us retain the common notion of pointer as something pointing toward something else.
Suppose now that I point my forefinger toward a cat, saying: /This is a cat/. Everybody would agree on the fact that the proposition «The object I have indicated by the pointer is a cat» is true (or that the proposition «The perceptum at which I pointed at moment x was a cat» is true; to put the matter simply, everyone would agree that what I had called a cat was a cat). In order that the above propositions be true I must be able to translate them as follows: “The perceptum connected with my forefinger at moment x represents the token occurrence of a perceptual type so conceptually defined that the properties possessed by the perceptual model systematically correspond to the semantic properties of the sememe «cat», and both sets of properties are usually represented by the same sign-vehicles”.
At this point the referent-cat is no longer a mere physical object. It has already been transformed into a semiotic entity. But this methodological transformation introduces the problem of the semiotical definition of the percepta (see 3.3.4.). If the sentence was a semiotic act and the cat an empirical perceptum it would be very difficult to say what the expression /is/ was. It would not be a sign, since /this is/ is the connecting device joining a complex sign (the sentence) to an actual perceptum. It would not be a pointer, inasmuch as the pointer points toward the perceptum to be connected with the sign, while /is/ seems to actually perform the connection itself. The only solution seems to be: /this is a cat/ means «the semantic properties commonly correlated by the linguistic code to the lexeme /cat/ coincide with the semantic properties that a zoological code correlates to that perceptum taken as an expressive device». In other terms: both the word /cat/ and that token perceptum //cat// culturally stand for the same sememe. This solution undoubtedly looks rather Byzantine – but only if one is accustomed to think that a ‘true’ perception represents an adaequatio rei et intellectus or is a simplex apprehensio mirroring the thing, as the Schoolmen maintained. But let us simply suppose that the expression /this is a cat/ is uttered in the presence of an iconic representation of a cat. All the above reasoning immediately becomes highly acceptable; we have a sign-vehicle (a) which is a linguistic expression to which a given content corresponds; and we have a sign-vehicle (b) which is an iconic expression to which a given content also corresponds. In this case we are comparing two sets of semantic properties and /is/ can be read as /satisfactorily coincides/ (that is: the elements of the content plane of a code coincide with the element of the content plane of another code; it is a simple process of transliteration)(5). Why does the mentioning act in the presence of a real cat seem so different to us? Clearly because we do not dare to regard perception as the result of a preceding semiotic act, as had been suggested by Locke, Peirce and many other philosophers.
There is a brief passage from Peirce (5.480) which suggests a whole new way of understanding real objects. Confronted with experience, he says, we try to elaborate ideas in order to know it. “These ideas are the first logical interpretants of the phenomena that suggest them, and which, as suggesting them, are signs, of which they are the . . . interpretants”. This passage brings us back to the vast problem of perception as interpretation of sensory disconnected data which are organized through a complex transactional process by a cognitive hypothesis based on previous experiences (cf. Piaget, 1961). Suppose I am crossing a dark street and glimpse an imprecise shape on the sidewalk. Until I recognize it, I will wonder “what is it?” But this “what is it?” may be (and indeed sometimes is) translated as “what does it mean?” When my attention is better adjusted, and the sensory data have been better evaluated, I finally recognize that it is a cat. I recognize it because I have already seen other cats. Thus I apply to an imprecise field of sensory stimuli the cultural unit «cat». I can even translate the experience into a verbal interpretant (/I saw a cat/). Thus the field of stimuli appears to me as the sign-vehicle of a possible meaning which I already possessed before the perceptual event.
Goodenough (1957) observed that: “a house is an icon of the cultural form or complex combination of forms of which it is a material expression. A tree, in addition to being a natural object of interest to a botanist, is an icon signifying a cultural form, the very same form which we also signify by the word tree. Every object, event or act has stimulus value for the members of a society only insofar as it is an iconic sign signifying some corresponding form in their culture . . . . ” Clearly from an anthropological point of view this position is close to what was said in the Introduction and to what will be said in 3.6.3. on the way in which every object may potentially become a sign within the environment of a given culture; and clearly the theory developed here finds many points of contact with the ideas suggested by Peirce.
As Peirce writes: “Now the representative function of a sign lies neither in its material quality nor in its pure demonstrative application; because it is something which the sign is, not in itself or in a real relation to its object; but which it is to a thought, while both of the characters just defined belong to the sign independently of its addressing to any thought. And yet if I take all the things which have certain qualities and physically connect them with another series of things, each to each, they become fit to be signs. If they are not regarded as such they are not actually signs, but they are so in the same sense, for example, in which an unseen flower can be said to be red, this being also a term relative to a mental affection” (5.287).
In order to assert that objects (insofar as they are perceived) can also be approached as signs, one must also assert that even the concepts of the objects (as the result or as the determining schema of every perception) must be considered in a semiotic way. Which leads to the straightforward assertion that even ideas are signs. This is exactly the philosophico-semiotical position of Peirce: “whenever we think, we have present to the consciousness some feeling, image, conception, or other representation, which serves as a sign” (5.283). But thinking, too, is to connect signs together: “each former thought suggests something to the thought which follows it, i.e., is the sign of something to this latter” (5.284).
Peirce is in fact following a very ancient philosophical tradition. Ockham (in I Sent., 2,8; Ordinatio, 2,8; Summa totius logicae, 1, 1) insists on the fact that if the linguistic sign points back to a concept (which is its content), alternatively the concept is a sort of sign-vehicle able to express (as its content) singular things. The same solution can be found in Hobbes (Leviathan, i, 4), not to speak of Locke’s Essay concerning Human Understanding: here Locke explicitly asserts the identity between logic and semiotics (IV,20) and the semiosic nature of ideas. These ideas are not (as the Schoolmen believed) a mirroring image of the thing; they too are the result of an abstractive process (in which – let it be noted – only some pertinent elements have been retained) which gives us not the individual essence of the named things but their nominal essence. This nominal essence is in itself a digest, a summary, a elaboration of the signified thing. The procedure leading from a bunch of experiences to a name is the same as that which leads from the experience of things to that sign of things, the idea. Ideas are already a semiotic product.
Obviously in Locke’s system the notion of idea is still linked to a mentalistic point of view; but it is sufficient to replace the term ‘idea’ (as something which takes place in the mind) by ‘cultural unit’ (as something which can be tested through other interpretants in a given cultural context) and Locke’s position reveals itself as very fruitful for semiotic purposes. Berkeley too (Treatise, Intr., 12) speaks of an idea as general when it represents or stands for all particular ideas of the same sort.
Obviously this interesting chapter of a future history of semiotics deserves a more careful elaboration. But it was in any case important to undertake this first tentative exploration in order to find some historical roots for the approach here proposed. It will help one to understand why throughout the entire history of philosophy the notion of linguistic meaning has been associated with that of perceptual meaning, by means of an identical term (or of a pair of homonymous terms.)
According to Husserl (Logische Untersuchungen, I, IV, VI) the dynamic act of knowing implies an operation of “filling up” which is simply an attribution of sense to the object of perception. He says that to name an object as /red/ and to recognize it as red are the same process, or at least that the manifestation of the name and the intuition of the named are not clearly distinguishable. It would be worth ascertaining to what extent the idea of ‘meaning’ found in the phenomenology of perception agrees with the semiotic notion of a cultural unit. A rereading in this light of Husserl’s discussions might induce us to state that semiotic meaning is simply the socialized codification of a perceptual experience which the phenomeno-logical epoché should restore to us in its original form. And the significance of daily perception (before the epoché intervenes to refresh it) is simply the attribution of a cultural unit to the field of perceptual stimuli as has been said above. Phenomenology undertakes to rebuild from the beginning the conditions necessary for the formation of cultural units which semiotics instead accepts as data because communication functions on the basis of them. The phenomenological epoché would therefore refer perception back to a stage where referents are no longer confronted as explicit messages but as extremely ambiguous texts akin to aesthetic ones.
This is not the place to study this problem in greater depth. Suffice it to say that we have indicated another of semiotics’ limits, and that it would be worth while to continue research on this in relation to the genesis of perceptual signification.
Let us now return to our example of the expression /this is a cat/. One is now ready to accept the idea that an act of mentioning or referring is made possible by a very complicated previous semiosic process which has already constituted the perceived object as a semiotic entity: (i) I recognize the cat as a cat, that is, I apply a cultural schema (or idea, or concept) to it; (ii) I understand the token cat as the sign-vehicle of the type cat (the corresponding cultural unit) concerning myself only with its semantic properties and excluding individualizing physical properties which are not pertinent (clearly the same happens, with some other mediating processes, when I say /this cat is big, black and white/); (iii) among the semantic properties of the cultural unit «cat» I select the ones which are in accordance with the semantic properties expressed by the verbal sign-vehicle.
I thus compare two semiotic objects, that is, the content of a linguistic expression with the content of a perceptual act. At this point I accept the equation posited by the copula /is/. Inasmuch as the equation represents a sort of metalinguistic act, associating a linguistic expression with the living ‘expression’ of a cultural construct (and thereby trying to establish an equivalence between sign-vehicles coming from different codes), it can either be accepted or refused – insofar as it does or does not satisfy the semantic rules imposing as predicates of a given item certain other items, able to amalgamate together through some common semantic properties. So the copula /is/ is a metalinguistic sign meaning «possesses some of the semantic properties of».(6) In some circumstances the metalanguage might not be a verbal one: as when /is/ is replaced by a pointing finger meaning both «this» and «is».
The above discussion (from 3.3.3. to 3.3.5.) has made clear the status of semiotic index-sensitive judgments. But perhaps the nature of factual index-sensitive judgments, such as /this cat is one-eyed/ remains more obscure. In this case I assign to the token of the type-cat a property which is not recognized by the code, so that we would seem to be back with the problem of the relation between on the one hand two semiotic constructs (the sememe «cat» and the conceptual type «cat» and on the other a mere perceptum. Except that the property of being one-eyed is not a mere perceptum, but rather a sort of ‘wandering’ semantic property, coming from some organized subsystem, which is recognized as such and attributed to this one cat, viewed as the occurrence of a more general model. The single occurrence of a type can have more characterizing properties than its model (thus the occurrence of a word shows a lot of free variants), it cannot however have properties which are incompatible with its type.
To predicate new properties of an object is not so different from producing phrases which are semantically acceptable. I can accept well-formed phrases like /the pencil is green/ or /the man sings/ and I must usually refuse phrases like /the pencil sings/ or /the man is green/; it is simply a matter of semantic amalgamation.
Therefore I can accept factual judgments like /this pencil is blue/, for pencils are usually either black or colored, /this pencil is long/ because pencils are physical objects possessing dimensional properties, and /this man sings/ because men can emit sounds. All of them are acceptable factual index-sensitive judgments. On the contrary /this pencil is two miles long/, /this pencil is vibrating at the speed of 2,000 w.p.s./ or /this man is internally moved by a four-stroke engine/ are abnormal factual judgments for they nourish an inner semantic incompatibility. Thus if I said /this cat is four feet long/ there would be two possibilities: either I see that the cat is not actually that long, and in this sense I am simply associating unappropriate words with the living expression of a semantic property that I can conceptually detect and that I could verbally express in another way; or I am really ‘telling the truth’. But if I have told the truth, I am obliged to ask myself: do four-foot-long cats really exist? All my knowledge about cats tells me that they do not usually share such a property, i.e. that the conceptual construct «cat» (corresponding to the sememe «cat» does not possess such a property. Therefore I must assume that what I have seen is perhaps not a cat but a panther. Suppose that I now check and I discover that it has all the properties of a cat and none of the properties of a panther, but that all the same it really is four feet long; then my perception, once it is conceptualized, does not coincide with the conceptual construct that made it possible. I must therefore reformulate the conceptual construct (and therefore the corresponding sememe); it is possible that a mutation may have changed the size of some cats. So I must emit a factual statement (/some cats are four feet long/) after which, by means of a meta-semiotic judgment, I can change the code.
The case of the four-foot-long cat is one of an actually perceived subject of which a puzzling property must be predicated. There are cases of predication in which the property does not create problems, but the subject does. Such is the case of the famous sentence /the present king of France is bald/. To engage in the Olympic Games that this sentence has provoked in contemporary semantics may help to solve the final problem about mentioning.
Everyone is agreed that the sentence in question, if uttered in the present century, is rather puzzling. It may also be suggested that the sentence is meaningless since ‘definite descriptions’ have a meaning only when there is a single object for which they stand. We have already provided the answer to such an assumption, and a theory of codes demonstrates that a description like /the king of France/ is fully endowed with meaning. It is not necessary to assume that a description like /the king of France/ must be verified by a presupposition, thus asking for an existential verification. This theory holds good when attributing a truth value to a proposition; so that if the description /the husband of Jeanne d’Arc/ does not have a ‘referential index’ a statement like /the husband of Jeanne d’Arc came from Brittany/ arouses a lot of interesting questions in terms of extensional semantics. But /the king of France/ stands for a cultural unit, not a person; not only does it share with /the husband of Jeanne d’Arc/ the quality of meaning something, but also it can correspond or not correspond to somebody who actually existed and who, in a possible world, could continue to exist.
So suppose then that someone states that /the king of France is wise/ as Strawson suggests; the expression is endowed with meaning, so the problem is to know under what circumstances it is uttered; if it is used in order to mention Louis XIV, it can be said to be acceptable, but if used in order to mention Louis XV some might judge it rather over-evaluative.
Suppose that I now say /this is the king of France/ pointing with my forefinger toward the President of the French Republic. This is the same as saying /this is a cat/ while indicating a dog. There is a semantic incompatibility between the properties of the sememe and the properties of the cultural unit represented by the indicated person, taken as an occurrence of a conceptual construct.
Suppose that I now say /this man is bald/ when referring to a long haired pop-singer; this represents a typical case of misuse of language. One need only translate the expression as /this man is a bald man/ for it to become clear that I am attributing certain semantic properties to a percept that cannot be taken as an occurrence of a more general model for bald men.
Suppose that I now say /the king of France is bald/; in itself the expression is meaningful and may become true when I use it in order to mention Charles the Bald, who was elected emperor in 875 A.D. If I use the sentence in order to mention Louis XIV the sentence is false. However, both mentions presuppose an indexical device; if I utter them I must in some way indicate which king I am referring to. The same happens when I say /the present king of France is bald/. The word /present/ is in fact a pointer, and as a pointer is a shifter (see 2.11.5.).
What does /the present king of France is bald/ mean? It has the following deep semantic structure: «there is a king of France. The king of France is bald». But /there/ is an ambiguous device : it has the sense of /there/ in /there are many books in the world/ and that of /there/ in /they are there/.. The first /there/ has an imprecise adverbial function, the second has the meaning «in this precise place» and an almost substantive function as in /he is in there/.
One should thus say: /there is there a king of France/, which would mean: «in the precise historical moment (or in the precise spatial environment) within which the sender of the message is speaking». And this is exactly the meaning of /present/, whose compositional tree could be represented as in Table 32 (according to 2.11.5.)
where the absence of an index suggests an imprecise and multidirectional closeness. In terms of meaning the addressee receives the imperative content «point your attention toward the immediate temporal context». In terms of mentioning the addressee does not discover in such a temporal context a possible perceptum that could correspond to a conceptual type having the properties assigned by a code to a «king of France». Therefore the communication ‘miscarries’; this proposition is neither true nor false, but simply inapplicable to any circumstance, and therefore misused. It is the same when I say /this is the king of France and he is bald/ while pointing my finger toward nothing.
Thus /the present king of France is bald/ is a meaningful sentence that, when considered as a mention, is an example of a misuse of sign production. Whereas /the king of France is bald/ is a meaningful sentence that, when used for imprecise mentions (for instance when uttered without specifying or presupposing any uncoded contextual selection) is simply useless. The proof is that, when hearing it, people will ask: “which one?”, thus demanding an indexical circumstantial marker(7).
Even though a definition of sign-function for every type of signs has been given in 2.1. and the process of sign production has also been examined from the point of view of many non-verbal signs, it would nevertheless be somewhat reckless to maintain that there is no difference between various types of signs. It is indeed possible to express a given content both by the expression /the sun rises/ and by another visual expression composed of a horizontal line, a semicircle and a series of diagonal lines radiating from the imaginary center of the semicircle. But it would seem more difficult to assert that /the sun also rises/ by means of the same visual device and it would be quite impossible to assert that /Walter Scott is the author of Waverley/ by visual means. It is possible to assert both verbally and kinesically that I am hungry (at least in Italian!) but it is impossible to assert by means of kinesic devices that «The Kritik der reinen Vernunft proves that the category of causality is an a priori form while space and time are pure intuitions» (even if Harpo Marx got remarkably near it). The problem could be solved by saying that every theory of signification and communication has only one primary object, i.e. verbal language, all other languages being imperfect approximations to its capacities and therefore constituting peripheral and impure instances of semiotic devices.
Thus verbal language could be defined as the primary modelling system, the others being only “secondary”, derivative (and partial) translations of some of its devices (Lotman, 1967). Or it could be defined as the primary way in which man specularly translates his thoughts, speaking and thinking being a privileged area of a semiotic enquiry, so that linguistics is not only the most important branch of semiotics but the model for every semiotic activity; semiotics as a whole thus becomes no more than a derivation from linguistics (Barthes, 1964).
Another assertion, metaphysically more moderate, but possessing the same practical import, might consist in maintaining that only verbal language has the property of satisfying the requirement of ‘effability’. Thus not only every human experience but also every content expressed by means of other semiotic devices can be translated into the terms of verbal language, while the contrary is not true. The effability power of verbal language is undoubtedly due to its great articulatory and combinational flexibility, which is obtained by putting together highly standardized discrete units, easily learned and susceptible to a reasonable range of non-pertinent variations.
An objection to this approach might run as follows: it is true that every content expressed by a verbal unit can be translated into another verbal unit; it is true that the greater part of the content expressed by non-verbal units can also be translated into verbal units; but it is likewise true that there are many contents expressed by complex non-verbal units which cannot be translated into one or more verbal units (other than by means of a very weak approximation). Wittgenstein underwent this dramatic revelation (as the Acta Philosophorum relate) when during a train journey, Professor Sraffa asked him what the ‘meaning’ of a certain Neapolitan gesture was.
Garroni (1973) suggests that there is a set of contents conveyed by the set of linguistic devices L and a set of contents that are usually conveyed by the set of non-linguistic devices NL; both sets contribute to a subset of contents which are translatable from L into NL or vice versa, but such an intersection leaves aside a vast portion of ‘unspeakable’ but not ‘unexpressible’ contents.
There are many proofs to support this theory. The power of verbal language is demonstrated by the fact that Proust successfully created the impression of rendering through words almost the entire range of perceptions, feelings and values embodied in an impressionist-like painting; but it is no chance that he was obliged to analyze an imaginary painting (by Elstir) for even a summary survey of a real painting could have suggested the existence of portions of content that the linguistic description did not cover. On the other hand it is quite clear that no painting (even if organized as some sort of supremely skillful comic strip with thousands and thousands of frames) could get across all that is conveyed by the Recherche.(8)
Whether there are NL semiotic systems; whether what they convey might be or ought to be called ‘content’ in the sense used up to now; whether as a result semantic markers and their interpretants have to be not only verbal devices but also organized and structured perceptions, habits, behaviors and so on; all this constitutes one of the most fascinating empirical boundaries of the present state of the semiotic art, and demands a great deal of further research.
In order to pursue this research it is absolutely necessary to demonstrate that (i) there exist different kinds of signs or of modes of sign production; (ii) many of these signs have both an inner structure and a relation to their content which is not the same as that of verbal-signs; (iii)a theory of sign production must and can define all of these different kinds of signs by having recourse to the same categorial apparatus.
Such is the aim of the following sections. I shall not attempt an exhaustive coverage of the entire field, but will instead try to define different types of signs, to analyze their constitutive differences and to insert them within the framework of the theory of sign-functions and codes. The conclusion to be drawn from this exploration will be that without doubt verbal language is the most powerful semiotic device that man has invented; but that nevertheless other devices exist, covering portions of a general semantic space that verbal language does not. So that even though this latter is the more powerful, it does not totally satisfy the effability requirement; in order to be so powerful it must often be helped along by other semiotic systems which add to its power. One can hardly conceive of a world in which certain beings communicate without verbal language, restricting themselves to gestures, objects, unshaped sounds, tunes, or tap dancing; but it is equally hard to conceive of a world in which certain beings only utter words; when considering (in 3.3.) the labor of mentioning states of the world, i.e. of referring signs to things (in which words are so intertwined with gestural pointers and objects taken as ostensive signs), one quickly realizes that in a world ruled only by words it would be impossible to mention things. In this sense a broader semiotic inquiry into various equally legitimate types of signs could also help a theory of reference, which has so frequently been supposed to deal with verbal language, as the privileged vehicle for thought alone.
Many different classifications of the various types of signs have been put forward during the development of the philosophy of language, linguistics, speculative grammar, semiotics, etc. All of these classifications served the purposes for which they were established. I shall limit myself here to a brief outline of those that are most relevant to the purpose of the present discussion. First of all, signs may be distinguished according to their channel, or expression-continuum. This classification (Sebeok, 1972) is useful for distinguishing many zoosemiotic devices and examines the human production of signs according to the different techniques of communication involved (Table 34).
This distinction does not seem particularly useful for our present discussion, since it would seem pretty vague to place both Beethoven’s Ninth Symphony and Dante’s Divina Commedia among the acoustically channelled signs, and both a road signal and Manet’s Le déjeuner sur l’herbe among the optical signs reflected by daylight.
Nevertheless it does permit one to isolate a set of important semiotic problems. There is a way in which Beethoven’s music and Dante’s poetry may be considered under the same heading. Both musical notes and words may be defined by means of sound parameters; the difference between a C emitted by a trumpet and a D emitted by a fiddle is detectable through reference to the parameters of pitch and timbre; the same happens for the difference between a velar voiced stop consonant (such as [g]) and a labial nasal one (such as [n1): in both cases, the decisive parameter is timbre. When distinguishing an interrogative utterance from an affirmative one, the essential parameters are pitch, dynamics and rhythm, as happens when distinguishing two different melodies.
On the other hand, in order to distinguish two road signals or two Manet paintings one resorts to both space and color parameters. In the first case the pertinent elements are the normal spatial dimensions, with features such as ‘up/down’, ‘right/left’, larger/smaller’ and so on; in the second one, pertinent elements are wave-lengths, frequencies or, more commonly speaking, cues. The fact that a road signal is enormously simpler than a Manet painting does not matter. Thus in detecting two tactile signals, one has recourse to certain thermal and pressure gradients, while in detecting the difference between two signs channelled through a solid matter, like two gestural signals, one relies on positional or kinesic parameters such as the direction of the gesture, its stress, and so on.
One of the most disturbing features of a lot of semiotic studies past and present has been the interpretation of various signs on the basis of the linguistic model, and thus the attempt to apply to them something metaphorically similar to the sound parameters or the model of double articulation, etc. As a matter of fact we know very little about other parameters, for instance those that govern the distinction between the olfactory signs, which are based on chemical features. Semiotics has a long way to go if it is to clarify all these problems; but even if one cannot map them entirely, one must nevertheless trace their outlines. For instance, the notion of binarism has become an embarrassing dogma only because the only binary model available was the phonological (and therefore the phonetic) one. Thus the notion of binarism has been associated with that of absolute discretedness, since in phonology the binary choice was applied to discrete entities. Both notions were associated with that of structural arrangement, so that it looked to be impossible to speak of a structural arrangement for phenomena that appeared to be continuous rather than discrete.
But ‘structure’ does not only mean opposition between the poles of a two-tuple of discrete elements. It also means opposition between an n-tuple of gradated entitites, resulting from the conventional subdivision of a given continuum, as happens with the color system. A consonant is either voiced or not, but a shade of red is not opposed to its absence; instead it is inserted within a gradated array of pertinent units cut from a wave-length continuum. Frequency phenomena do not allow the same kind of pertinentization as do timbre phenomena. As has been already said à propos of a theory of codes (see 2.8.3.), there is structure when a continuum is gradated into pertinent units and the array of these units has precise boundaries.
Moreover, nearly all the non-verbal signs usually rely on more than one parameter; a pointing finger has to be described by means of three-dimensional spatial parameters, vectorial or directional elements, and so on. So an attempt to establish a complete set of semiotic parameters will involve the entire physical conditioning of human actions, inasmuch as they are conditioned by the structure of the human body inserted within its natural and artificial environment. As will be shown in 3.6., it is only by recognizing such a range of parameters that it is possible to speak of many visual phenomena as coded signs; otherwise semiotics would be obliged to distinguish between signs which are signs (because their parameters correspond to those of verbal signs, or can be metaphorically viewed as analogous to them) and signs which are not signs at all. Which may sound paradoxical, even though it is upon such a paradox that many distinguished semiotic theories have been established.
Signs are also distinguished according to whether they originate from a sender or a natural source. Insofar as there exist a lot of signs that are without a human sender, occurring as natural events but being interpreted as semiotic devices, this classification, also summarized by Sebeok (1972), can be useful for the analysis of communicational processes (Table 35).
Signs are also distinguished according to their semiotic specificity. Some signs are objects explicitly produced in order to signify, others are objects produced in order to perform a given function. These latter can only be assumed as signs in one of two ways; either they are chosen as representatives of a class of objects (see 3.6.3.) or they are recognized as forms that elicit or permit a given function precisely because their shape suggests (and therefore ‘means’, ‘signifies’) that possible function. In this second case they are used as functional objects only when, and only because, they are decoded as signs.
There is a difference (as regards sign-specificity) between the injunction /sit down!/ and the physical form of a chair which permits and induces certain functions (among others, that of sitting down); but it is equally clear that they can be viewed under the same semiotic profile.
At this point it would seem advisable to examine what is, perhaps, the most popular of Peirce’s trichotomy, that by which signs are classified as symbols (arbitrarily linked with their object), icons (similar to their object) and indices (physically connected with their object).
This distinction is so widely accepted that we have used it in the preceding pages in order to indicate certain processes, so that they could be immediately, if vaguely, grasped by everyone. It is nevertheless the basic assumption of the following pages that notions such as Icon’ or ‘index’ are all-purpose, practical devices just as are the notions of ‘sign’ or ‘thing’. They can undoubtedly be used for normal purposes, but no satisfactory definition can be found for them in the present context.
The reason is simple: such a trichotomy postulates the presence of the referent as a discriminant parameter, a situation which is not permitted by the theory of codes proposed in this book. The trichotomy could obviously be used in order to discriminate between different kinds of mentions (as indeed it was within that context), but it becomes more disturbing in a classification of modes of sign production which tries to focus exclusively upon the shaping of the signal (i.e. the expression continuum) and the correlation of that signal (as expression) with a content. Thus the following pages represent a critique of the naive notion of index and icon, and the new classification proposed in 3.6. aims to supersede these categories.
A final distinction concerns the replicability of sign vehicles. The same word can be replicated an infinite number of times, but each replica is without economic value, whereas a coin, even though a replica, has a material value of its own. Paper money has a minimal material value but receives a sort of legal value by a convention, so that it cannot be indefinitely replicated; moreover, the process of replication is so technically difficult that it requires special techniques (the reasons of that difficulty are similar to those which seemingly forbid the reproduction of Michelangelo’s Pietà; oddly enough this, too, has received a sort of conventional and ‘legal’ investiture whereby a replica, no matter how perfect, is refused as a fake). Finally a Raphael painting is commonly considered beyond replication except in cases of exceptional skillfulness – and even in these cases the replica is considered imprecise and unable to deceive a well-trained eye.(9) Thus it seems that there are three kinds of relationship between the concrete occurrence of an expression and its model: (a) signs whose tokens can be indefinitely reproduced according to their type; (b) signs whose tokens, even though produced according to a type, possess a certain quality of material uniqueness; (c) signs whose token is their type, or signs in which type and token are identical.
This distinction can be easily reduced to that proposed by Peirce’s trichotomous distinction between legisign, sinsign, and qualisign (2.243.ff.): signs of type (a) are pure sin signs; signs of type (b) are sinsigns which also are qualisigns; signs of type (c) are sinsigns which also are legisigns.
If these distinctions are considered from the point of view of the commercial value of the replica, then they are more a matter for Treasury Departments, income tax inspectors, art dealers and organized crime than for a theory of sign-functions (in which an object’s only recognized value is its quality as a functive). From the semiotic point of view, the fact that a hundred dollar bill is counterfeit does not matter: every object looking like a hundred dollar bill will stand for the equivalent amount of gold to its addressees: the fact that the bill is a fake merely means that this is a case of lying.
A perfect replica of Michelangelo’s Pietà which rendered each nuance of the material texture of the original with great fidelity would also possess its semiotic properties. Therefore the value accorded to the ‘authenticity’ of the original statue has more relevance for a theory of commodities, and when given undue importance on the aesthetic level it is a matter for social scientists or critics of social aberrations. The lust for authenticity is the ideological product of the art market’s hidden persuaders; when the replica of a sculpture is absolutely perfect, to privilege the original is like giving more importance to the first numbered copy of a poem than to a normal pocket edition. But when one considers the same problem from the point of view of sign production, other factors have to be considered. Differing modes of production of the expression, along with the necessary type/token-ratio, determine a fundamental difference in the physical nature of various types of signs.
At this point we must make a clear distinction between absolute duplicative replicas which produce a double, and partial replicas, which will simply be called replicas.
I mean by an absolutely duplicative replica a token which possesses all the properties of another token. Given a wooden cube of a given size, matter, color, weight, surface structure and so on, if I produce another cube possessing all the same properties (that is, if I shape the same continuum according to the same form) I have produced not a sign of the first cube, but simply another cube, which may at most represent the first inasmuch as every object may stand for the class of which it is a member, thus being chosen as an example (see 3.6.3.).
Obviously, as Maltese (1970:115) suggests, an absolute replica is a rather Utopian notion, for it is difficult to reconstruct all the properties of a given object right down to its most microscopic characteristics; but there is a threshold fixed by common sense which recognizes that, when a maximum number of parametric features have been preserved, a replica will be accepted as another exemplar of the same class of objects and not as an image or representation of it. Two Fiat 124 cars of the same color are not each other’s icon but two doubles.
In order to obtain a double it is obviously necessary to reproduce – to a given extent – all the properties of the model-object, maintaining their original order and interrelationships. But in order to do so it is necessary to know the rule which governed the production of the model-object. To duplicate is not to represent, to imitate (in the sense of making an image of), to suggest the same appearance; it is a matter of equal production conditions and procedures.
Suppose one has to duplicate an object devoid of any mechanical function, such as a wooden cube: one has to know (a) the modalities of production (or of identification) of the material continuum, (b) the modalities of its formation, i.e. the rules governing the relationships between its geometrical properties. Suppose now that one has to duplicate a functional object, such as a knife. One must also know its functional properties. A knife is the double of another knife if, ceteris paribus, it has the edge sharpened to the same degree. This being so, even if there were some microscopis difference in the surface texture of the handle, which could not be detected by sight, touch or a sensitive weighing machine, everybody would say that the second knife was the double of the first.
If the object is a very complex one, the principle of duplication does not change; what changes is the number of rules and the technical difficulties involved, as would be the case when trying to make the double of a Chevrolet, clearly no matter for the ‘do it yourself’ enthusiast.
An object as functionally and mechanically complex as the human body is not duplicable precisely because we are ignorant of many of its mechanical and functional rules, and first and foremost those required in order to produce living matter. Any duplication which does not follow all the rules of production and which therefore produces only a given percent of the mechanical and the functional properties of the model-object is not a double, but at best a partial replica (see 3.4.8.).
In this sense an uttered word is not the absolute duplicative replica of another word of the same lexicographic type, but rather, as we shall later see, a partial replica. If, however, I print the same word twice (for example: /dog/ . . . /dog/) I can say that one is the double of the other (microscopic differences in inking or in the pressure of the type on the paper being more a matter for metaphysical doubts about the notion of identity or equality).
According to this notion of double, it is commonly supposed that a painting is not truly duplicable. This is not completely true for, under given technical conditions, and using the same materials, one could theoretically establish a perfect double of the Mona Lisa by means of electronic scanners and of highly refined plotters. However, the perfection of such a double is determined by a perfect knowledge of even the microscopic texture of the artifact, which is usually unattainable.
Since we have defined as duplicable an object whose productive rules one knows, a painting will not usually qualify as such. What will qualify are such craft products as are traditionally duplicated without appreciable differences, so that nobody will be tempted to consider the duplicate as an iconic reproduction of the original; the duplicate is as much original as is its model. The same happens in civilizations where the representative rules are strictly standardized, so that an Egyptian painter might quite possibly have been able to duplicate a mural painting.
If Raphael’s painting seems beyond duplication, this is because he invented his rules as he painted, proposing new and imprecise sign-functions and thereby performing an act of code-making (see 3.6.7.). The difficulty in isolating productive rules is due to the fact that, while in verbal language there are recognizable and discrete signal-units, so that even a complex text may be duplicated by means of them, in a painting the signal looks ‘continuous’ or ‘dense’, without distinguishable units. Goodman (1968) remarks that the difference between representative and conventional signs resides in this opposition (dense vs. articulate) and it is to this difference that the difficulty in duplicating paintings is due. As we shall see later (3.5.) this opposition is not sufficient to distinguish the so-called “iconic” or “representative” signs, but it may be retained for the moment.
A painting does in fact possess qualisign elements; the texture of the continuum from which it is made counts for a great deal, so that a dense signal is not reducible to a distinction between pertinent recognizable elements and irrelevant variations; even minimal material variations count. It is this quality which makes a painting into an aesthetic text, as will be better explained in 3.6.7. This is undoubtedly one of the reasons why the duplication of a painting is well-nigh impossible and why its rules of production are hardly detectable.
But the other depends on the particular type/token-ratio realized by a painting. In order to make this point clear, we must now consider the case of partial replicas.
In replicas the type is different from the token. The type only dictates those essential properties that its occurrences must display in order to be judged a good replica, irrespective of any other characteristic that they may possess. Thus tokens of the same type can possess individual characteristics, provided that they respect the pertinent ones fixed by the type. It is this kind of type/token-ratio, for example, that rules the production of phonemes, words, ready-made expressions, etc. Phonology establishes certain phonetic properties that a token phoneme must have in order to be recognized as such; everything else is a matter of free variation. Regional or idiosyncratic differences in pronunciation do not matter, provided they do not affect the recognizability of the pertinent properties.
The type/token-ratio obeys different parameters and rules of fidelity according to different sign systems. Maltese (1970) lists ten kinds of ratio, from the absolute duplicate (which, given six properties of the visual-tactile experience of a given object, reproduces all of them) down to the reproduction of a unique property, as happens in a symbolic and schematic representation on a plane surface. This list coincides in some respects with the various ‘scales of iconicity’ (such as that proposed by Moles) and problems connected with these scales will be considered in 3.6.7.; but at present we are concerned with the three first degrees of Maltese’s scale: between the first (6/6), second (5/6) and third (4/6) one could easily classify the various kinds of type/token-ratio operating within the relationship between an expression and its type. For instance a road signal commanding “stop” is a 6/6 reproduction of its type; it is the absolute duplicative replica of many other signs of the same class. As an object, it is simply a double, but inasmuch as it is a sign it is one in which the fidelity of token to type must be absolute; the type prescribes form, size, painted lines and colors, material smoothness of surface, weight, etc. without permitting free variations. Free variations might well allow one to recognize the sign as such, but would induce a sharp observer (such as a policeman) to suspect a fake.
A phoneme does not have to be so faithful to its expressive type: or rather, it has to respect the dictates of its type, but its type does not dictate every material nuance of its occurrences. The type prescribing the form of the image of the King of Spades in playing cards offers many more possibilities for free variation (indeed this sort of stylization will be considered as something half way between replica and invention in 3.6.5.) (10).
Every replica is a token accorded to a type. Thus every replica is governed by a type/token-ratio. But in order to understand many other procedures in sign production let me outline, at this point, a distinction between two different sorts of type/token-ratio; I shall call them ratio facilis and ratio difficilis. There is a case of ratio facilis when an expression-token is accorded to an expression-type, duly recorded by an expression-system and, as such, foreseen by a given code.
There is a case of ratio difficilis when an expression-token is directly accorded to its content, whether because the corresponding expression-type does not exist as yet or because the expression type is identical with the content-type. In other words, there is a ratio difficilis when the expression-type coincides with the sememe conveyed by the expression-token. Using a formulation that will be partly criticized in the following sections, one could say that in cases of ratio difficilis the nature of the expression is motivated by the nature of the content.
It is not difficult to isolate and to understand the cases of ratio facilis; they are all those described in 3.4.8.; the sign is made up of a fairly simple expression-unit corresponding to a fairly precise content-unit. It is the case of words and of various visual entities (for example road signals as well as strongly stylized pictorial entities, such as occur in handicrafts and primitive painting); in order to produce a sign-vehicle meaning so and so, one must produce an object constructed in such and such a way (according to the model provided by the expression-system).
This kind of ratio facilis does not govern a double but a replica, since the expression-type establishes some features as pertinent, and some others as variable and inessential for the isolation of a given unit (11).
Even many texts are replicable according to a ratio facilis; suppose that in a certain primitive civilization there exists a given ritual dance or liturgical ceremony which conveys a vast portion of social content; this civilization may have undercoded some basic movements (event though permitting a lot of free variations) that allow one to isolate a given behavior as the replicable sign-vehicle of a given social content. A type/token-ratio can be facilis even when the type is very imprecise, provided it has been socially recorded. It would seem more difficult to isolate examples of ratio difficilis. As a matter of fact they depend on two different situations of sign production.
First situation: the expression is a precise unit correlated to a precise content, such as occurs with kinesic pointers, and nevertheless the design of the expression in some way depends on the corresponding sememe. These signs are easily replicable and have the curious property of being submitted both to a ratio facilis and to a ratio difficilis (see 3.4.10.).
Second situation: the expression is a textual cluster that should convey imprecise portions of content or a content-nebula. Such is the case of many text-oriented cultures (see 2.14.6.) which have not elaborated a highly differentiated content-system to which a highly elaborated expression-system corresponds. But this is also the case of many undercoded sign-functions in a grammar-oriented culture. In such a situation the expression must be elaborated according to a ratio difficilis and frequently cannot be replicated since the content, even expressed, cannot completely be analyzed and recorded by its interpreters. In this case the instances of ratio difficilis concern activities of code-making (see 3.1.2.). In the following two paragraphs I shall deal with some preliminary instances of these two different situations in which ratio difficilis is required in order to produce expressions.
A propos of kinesic pointers (see 2.11.5.), we have seen that there is no need to have something close to a pointing finger for that finger to acquire a meaning. The pointing finger has a seme of «closeness» and this semantic marker is grasped even if one points into empty air. The presence of the actual thing is not necessary in order to understand the pointer as a sign, even though it is necessary that something be there when the pointer is used to mention. But even when indicating nothing, the pointing finger is nevertheless a physical phenomenon whose nature is different from that of the verbal pointer /that/. It is just this physical nature which must be now analyzed inasmuch as it is the effect of a complex act of sign production.
In a pointing finger the expressive continuum (or matter) is given by a part of the human body. In this continuum pertinent features have been selected according to an expression-form system. Therefore in this sense a pointing finger is subject to a ratio facilis and can be indefinitely replicated.
Nevertheless when it has been said that a pointing finger has four syntactic markers (latitude, apicality, movement toward and dynamic stress) which convey certain semantic markers (such as closeness, direction, distance), it has been noted that the semantic unit «direction» is not independent of the syntactic feature /movement toward/; the force or the weakness of the stress directly conveys markers of distance or closeness. This does not occur with verbal pointers like /this/ whereby the content is arbitrarily correlated to the sign-vehicle.
In kinesic pointers the seme of «closeness» is independent of the presence of the indicated thing (as noted in 2.11.5.) but the movement of the finger must be toward that point in the space where the actual or supposed thing is or should be. It is true that the notion of «a thing in that place» is not a ‘thing’ but a portion of the content, yet one of the features of this supposed content is precisely a spatial situation. The pointing finger means a spatial situation and that spatial situation is intensionally analyzable even if extensionally null; in intensional terms it has certain semantic properties, one of them being that of having spatial co-ordinates. Now, these spatial co-ordinates of the conveyed content to some extent determine the spatial properties of the expression, i.e. the physical properties of the signal, thus introducing a ratio difficilis within a productive process apparently obeying a ratio facilis. Thus a non-verbal index has the same sign-function structure as has a verbal one, the same capacity to be analyzed into syntactic and semantic markers, but some of its syntactic markers seem to be motivated by its content.
Thus the attempt to subsume every kind of sign within the same semiotic categories gives rise to a new category which is not common to every kind of sign. One might over-hastily conclude from this that, even if not dependent upon its proximity to the referent, a pointing finger is nonetheless ‘similar’ to its possible referent and is therefore ‘iconic’.
One of the aims of the following pages will be that of demonstrating that one cannot so easily equate motivation and similarity. But the problem does nonetheless exist, and the theory of codes outlined in chapter 2 cannot eliminate it.
But there are other reasons to render the pointing finger different from a verbal index.
Buyssens (1943) stated that a directional arrow, in itself, does not mean anything; but it may, for example, assume the meaning «turn left» if placed in a particular urban context (or external circumstance). This is not true.
Suppose we find a «turn» signal and a «stop» signal in a city’s traffic department store, and thus view them without relation to any specific urban context. We are nonetheless able to recognize and to distinguish «stop» from «turn». This means that there exists a precise convention whereby those graphic sign-vehicles do have a meaning and thus do convey a precise portion of content. But, while the «stop» signal has the same meaning to everyone everywhere, the «turn to . . . » signal acquires the fullness of its meaning only when certain circumstantial features (relating to its placement in this rather than that place) add additional meanings such as «left» or «right».
It could be said that the position in which the road signal is placed is simply a circumstantial selection that awaits interpretation, or that the fact that the sign is located in that place is nothing more than a mention («this is the place where you have to turn»).
But this situation recalls another, involving verbal signs. These latter always occur before and after other signs, within the context of the phrase. Thus in the expression /John beats Mary/ it is their relative positions that make Mary a victim and John unduly violent; if /John/ were in the place of Mary and vice versa, things would run better for Mary. Contextual position (the order of words in the phrase marker) changes the meaning of an expression to such a degree that Morris (1946) proposed to list syntactical positions among a category of signs to be called “formators”. The syntactical position would thus be a particular syncategorematic sign.
This being so, one can also single out certain ‘formators’ which are in some way topo-sensitive (through owing their meaning to their spatial or temporal co-ordinates, as happens with the direction of movement in the case of the pointing finger); moreover (as with the pointing finger), the nature of these co-ordinates on the expressive plane is motivated by the nature of the co-ordinates on the content plane. In other words /Mary/ is placed after /John/ because first «John» beats and then «Mary» is beaten.
All the examined examples have the feature of vectorialization in common, whether a real movement realizing a direction (the finger) or a virtual direction and movement realized by a feature of order (the phrase). The same happens with the road arrow situated at a given point: the entire sign («turn left») is topo-sensitive because one of its expressive features consists in pointing at the left rather than at the right of the addressee (12).
One might say that features of vectorialization make a sign ‘similar’ to its referent. In this case it would be no longer necessary to elaborate such a category as ratio difficilis and it would suffice to say that certain signs do not have an expression type but directly imitate the object for which they stand. However, the category of ratio difficilis has just been established in order to avoid such a naive interpretation (which will be criticized in 3.5) and a different theorization of vectorializations will be given in 3.6.5.
Let us now consider some examples in which the motivation exercised by the content on the expression seems to be so strong as to challenge, along with the possibility of replicas, the very notion of coded correlation (and therefore of conventional sign-function). We shall first of all examine those cases in which one must express a large number of content-units whose aggregation has not been previously coded and therefore constitutes a discourse. Let us define a discourse as the equivalent of a text on the expression plane.
One may encounter two types of discourse for which no pre-established text exists.
The first is that of factual statements concerning unheard-of events, these events constituting a new combination of cultural units that the content-system has already recognized and classified. To take the problem of verbally describing or visually representing a golden mountain or a new chemical compound, since these entities are the result of a combination of previously recorded semantic units and since the code already provides the corresponding expression-units (both ‘categorematic’ and ‘syncategorematic’), the format of the expression will be established according to the requirements of the content – but not according to its form! Therefore this is not a case of ratio difficilis: the combination of the words /golden/ and /mountain/ has nothing to do with (it is not ‘similar’ to) the orographic structure of the imagined phenomenon. In other words, if an astronomer discovers that small red elephants may be observed living on the moon, every time Capricorn enters into the orbit of Saturn, then his content-system will undoubtedly be upset (and he will have to restructure his world-view) but his expression-system will not be disturbed at all, for the laws of the code allow him to ‘map out’ such a new state of the world (as well as to produce new words for new definable content-units, since the redundance of the expression-system allows him to articulate new lexical items).
But one encounters a quite different set of problems in the case of a new and undefinable content-unit (if in such circumstances one can still talk of a ‘unit’) or rather in that of a content-nebula which cannot be analyzed into recognizable and definable content units. One might speak of a new discourse which has no satisfactory interpretants. Suppose that one had to express the following situation: «Solomon meets the Queen of Sheba, each leading a procession of ladies and gentlemen dressed in Renaissance style, and bathed in a pure and still morning light that gives bodies the air of mysterious statues, etc., etc.». Everyone would recognize in this ‘Verbal’ discourse something vaguely similar to a well-known pictorial ‘text’ by Piero della Francesca: but the verbal expression does not ‘interpret’ the pictorial one. At most, the former suggests the latter only because the latter has already been expressed and recorded by our culture. And even in this case only certain of the verbal expressions refer to recognizable content units (Solomon, to meet, Queen of Sheba, etc.), while many others by no means convey the sort of content that one might receive when looking at the painting (it goes without saying that even such an expression as /Solomon/ represents a rather imprecise interpretant of the corresponding image painted by Piero). When the painter begins work, the content (in its nebula-like structure) is neither coded nor divided into precise units. It has to be invented.
But the expression, too, has to be invented: as noted in 2.14.6., only when a highly differentiated content-system has evolved will a culture dispose of the corresponding expression-system. So we have a paradoxical situation, in which expression must be established according to a content model which does not yet exist as such.
The sign producer has a fairly clear idea of what he would like to ‘say’, but he does not know how to say it; and he cannot know how to do so until he has discovered precisely what to say. The lack of a definite content-type makes it impossible to find an expression-type, while the lack of an appropriate expression device makes the content vague and inarticulable. The difference between mapping into the expression a new but foreseeable content and mapping into the expression a content-nebula is that between a rule-governed creativity and a rule-changing creativity. Thus the painter has to invent a sign-function, and since every sign-function is based on a code, he has to propose a new way of coding.
To propose a code is to propose a correlation. Usually correlations are fixed by convention. But in this case the convention does not exist and the correlation must therefore be founded on something else. In order to make it acceptable, the producer must base his correlation on some evident motivation, for instance a stimulus. If the expression as stimulus is able to direct attention toward certain items of the suggested content, the correlation is then posited (and apres coup could even by recognized as a new convention).
Thus, given a content-type that is in some way cognizable, its pertinent features must be ‘projected’ into a given expression continuum by means of certain transformational rules. This does not mean that the expression must ‘imitate’ the form of the object; a critique of this naive approach to the problem will be proposed in 3.5. If the content-type is complex, then the transformational rules will be equally complex, and will sometimes escape detection, being rooted in the signal’s microscopic texture. In this way the sign (or text) becomes dense.
The more the content-type is new and uncoded or is the result of unfamiliar acts of mentioning, the more the producer must elicit in the addressee perceptual reactions that are to some extent equivalent to the ones he might have in the presence of the actual event. It is this extreme mode of stimulation that has permitted the formation of the notion of an ‘iconic’ sign as the naturally motivated and analogical result of an ‘imprint’ of the object itself on a given material continuum.
When examining kinesic pointers we have discovered that signs can exist that are at once replicable and motivated. As a matter of fact, phenomena like replicability or motivation are not features by which one sign may be distinguished from another; they are modes of production that play differing parts in the constitution of various sorts of sign-functions. This is also the case with an opposition such as ‘arbitrary vs. motivated’. Yet for many centuries this opposition seemed to be so evidently a matter of experience that the whole history of philosophy of language has been dominated by this question, beginning with Plato’s Cratylus, which opposed Nomos (law, convention, arbitrariness) to Physis (nature, motivation, iconic relationship between sign and things).
One should not undervalue these positions, but the problem that they express must be re-thought in a more rigorous way – if only because in recent times the opposition ‘arbitrary vs. motivated’ (already associated with ‘conventional vs. natural’) has finally been coupled with a third opposition, ‘digital vs. analogical’. Insofar as the term ‘analogical’ is understood in a double sense (see 3.5.4.) – as something concerning rules of proportionality and as something connected with an immediate ‘unspeakable’ similarity – and insofar as, in the first sense, analogical is opposed to digital, arbitrary signs are roughly equated with digitally analyzable ones. The same happens in the case of the third opposition and the whole system usually takes the following (apparently logical) form:
digital |
vs. |
analogical |
arbitrary |
vs. |
motivated |
conventional |
vs. |
natural |
in which the vertical columns are supposed to list synonymous categories.
Even a superficial glance at many sign phenomena tells us that the equation is not true and that therefore the oppositions are not synonymous; a photograph is perhaps ‘motivated’ (the traces on the paper are produced by the disposition of the matter in the supposed referent) but it is digitally analyzable, as happens when it is printed through a raster; the smoke revealing the presence of a fire is motivated by the fire but is not analogous to it; a painting representing the Virgin Mary is ‘analogous’ to a woman, but it is recognized as the Virgin Mary because of a conventional rule; a certain type of fever is naturally motivated by TBC but it is due to a convention that it is recognized as a reliable medical symptom. The movement of the pointing finger toward the supposed object is maybe motivated by the spatial co-ordinates of the object but the choice of the pointing finger as index is highly arbitrary; in fact the Cuna Indians use an entirely different device, the ‘pointing lips gesture’ (Sherzer, 1972). A cat’s paw print is motivated by the form of a given cat’s paw but it is by convention that a hunter assigns to that expressive shape the content (the abstract notion of) «cat».
One must, at this point, face the problem of the so-called iconic signs, in order to discover how many semiotic phenomena are commonly covered by this all-embracing term. So-called iconism in fact covers many semiotic procedures, many ways of producing signals ordered to a sign-function, and we will see that, even though there is something different between the word /dog/ and the image of a dog, this difference is not the trivial one between iconic and arbitrary (or “symbolic”) signs. It is rather a matter of a complex and continuously gradated array of different modes of producing signs and texts, every sign-function (sign-unit or text) being in turn the result of many of these modes of production.
It was said in 2.1. that a sign-function is the correlation between an expression and a content based on a conventionally established code (a system of correlational rules), and that codes provide the rules that generate sign-functions. If there exist signs that are to some degree motivated by, similar to, analogous to, naturally linked with their object, then the definition given in 2.1. should no longer be tenable.
The only way to maintain it is to demonstrate that even in these types of signs a correlational convention is in operation. The core of the problem is obviously the notion of convention, which is not co-extensive with that of arbitrary link, but which is co-extensive with that of cultural link. If one examines the mode of production of signs one must not only analyze the mode of production of the signal in itself but also its mode of correlation to its content, the correlating operation being part of the production. To produce a signal such that it may be correlated to a content is to produce a sign-function; the modes whereby either a word or an image are correlated with their respective contents are not the same. The problem is to find out whether the former is a cultural correlation (and therefore a conventional one) and the latter is not; or whether, on the contrary, both involve some sort of cultural correlation even though these correlations are operationally different (ratio facilis vs. ratio difficilis). In order to prove that the image of a dog also signifies a dog by means of a cultural mode of correlation, one must first of all challenge some naive notions. These notions are:
(i) that the so-called iconic sign has the same properties as its object;
(ii) that the so-called iconic sign is similar to its object;
(iii) that the so-called iconic sign is analogous to its object;
(iv) that the so-called iconic sign is motivated by its object. Permeating the critique of these assumptions is a contrasting one, which risks attaining an equal dogmatism, i.e.:
(v) That the so-called iconic signs are arbitrarily coded.
We shall see that it is possible to assert that they are culturally coded without saying that they are totally arbitrary, thereby restoring to the category of conventionality a more flexible sense. But when one has solved these problems one is faced with a last possible assumption:
(vi) that the so-called iconic signs, whether arbitrary or not, are analyzable into pertinent coded units and may be subject to a multiple articulation, as are verbal signs.
We shall see that, if one accepts (v) without reservations, one is also forced to accept (vi), which could lead to a lot of difficulties. But if one views (v) in the flexible and prudent way outlined above, (vi) is no longer strictly and directly dependent upon (v). One could thus assume that so-called iconic signs are culturally coded without necessarily implying that they are arbitrarily correlated to their content and that their expression is discretely analyzable.
According to Morris (1946) a sign is iconic “to the extent to which it itself has the properties of its denotata”. At first glance common sense might mislead one into agreeing with this definition. But a more thorough examination in the light of that same common sense forces one to realize that the definition is more or less tautological and in any case rather naive.
What does it mean to say that the portrait of Queen Elizabeth, painted by Annigoni, has the same properties as Queen Elizabeth? Morris (1946:1.7.) knows very well that “the portrait of a person is to a considerable extent iconic, but is not completely so since the painted canvas does not have the texture of the skin, or the capacities for speech and motion, which the person portrayed has. The motion picture is more iconic but again not completely so”.
Such an approach, when pushed to its limit, would persuade both Morris and common sense to destroy the notion of iconism; “a completely iconic sign would always denote since it would be itself a denotatum”, which is the same as saying that the true and complete iconic sign of Queen Elizabeth is not Annigoni’s portrait but the Queen herself76 (or a possible science fiction doppelganger). Morris himself, in the following pages, corrects the rigidity of the notion and states: “An iconic sign, it will be recalled, is any sign which is similar in some respects to what it denotes. Iconicity is thus a matter of degree” (1946:7.2.). And since, going on to deal with non-visual iconic signs, he even speaks of onomatopoeia, clearly the question of degree appears to be extremely elastic, since the iconic relationship between a /cock-a-doodle-doo/ and the crowing of a cock is very weak; so much so that the French onomatopoeic sign is /cocquerico/, and the Italian one / chicchiricchi/: the problem lies first of all in the meaning given to the expression “in some respects”: if an iconic sign is similar to the thing denoted in some respects, then we arrive at a definition which satisfies common sense, but not semiotics.
Secondly there are certain perplexities surrounding the notion of ‘similarity to objects’. Is one really sure that iconic signs are ‘similar’ to the objects they stand for? Indeed, is one sure that they stand for objects at all? Let us examine an advertisement. An outstretched hand offers me a glass foaming over with freshly poured beer, while over the outside of the glass extends a thin layer of vapor which immediately conveys a sensation of coldness. It would be interesting to see which of the properties of the object this picture contains. There is neither beer nor glass on the page, nor is there a damp and icy film. I feel certain visual stimuli, colors, spacial relationships, incidences of light and I coordinate them into a given perceptual structure. The same thing happens when I look at an actual glass of beer; I connect together some stimuli coming from an as yet unstructured field and I produce a perceptum based on a previously acquired experience.
Thus I can only assume that in the iconic experience certain perceptual mechanisms function which are of the same type as the one involved in the perception of an actual object, but the stimuli that I am concerned with in the first case are not the same as those that I am concerned with in the second. At most, a theory of perception will tell me that there are previous expectations, or models, or codes, that rule both perceptual coordinations. The solution would then be to propose that iconic signs do not possess the ‘same’ physical properties as do their objects but they rely on the ‘same’ perceptual ‘structure’, or on the same system of relations (one could say that they possess the same perceptual sense but not the same perceptual physical support). Thus when looking at the actual glass of beer I perceive on a given surface the presence of a uniform layer of transparent material which, when struck by light, gives off silver reflections, thereby producing that perceptum which I call “ icy film on the glass”; on the other hand in the drawing I perceive on a given paper surface a film of non-transparent material composed of two or more different shades of color that, by their mutual contrast, create the impression of incident luminosity. What kind of structural relationship remains unchanged between film and light, on the one hand, and two different colors on the other? And does the result of the two procedures produce the ‘same’ perceptual effect? Is it not better to assume that, on the basis of previous learning, I view as one and the same perceptual result what are in fact two different perceptual results? Suppose now that I draw the outline of a horse on a sheet of paper by one continuous and elementary line. The sole property that my horse possesses (one continuous black line) is precisely the property that a real horse does not possess. My drawing has defined by that line the space inside the horse separating it from the space outside the horse, whereas the actual horse is in fact a body within or against a space.
Admittedly, if I see the profile of a horse against the background of the sky, the contrast between the boundaries of that body and the background can appear under some circumstances as a continuous line at whose limits the light is absorbed into the dark body. But the process is more complex, the boundaries are not so clear and therefore the black line iconically rendering this perceptual experience is decidedly a simplifying and selective one. Thus a graphic convention allows one to transform, on paper, the elements of a schematic conceptual or perceptual convention which has motivated the sign.
Maltese (1970:VIII) suggests that the continuous line produced by the imprint of a body upon a malleable substance is matter not of visual but of tactile experience. The visual stimulus, which is in itself very poor from an informational point of view, suggests (by means of synesthesia) a tactile experience. That kind of stimulus is not a sign at all. It is only one among the various features of an expressive device that contributes to establishing the correspondence between that expression and a given content(«human hand», or better «a human hand pressed on that point there»). Thus the profile or the global imprint of the hand are not iconic signs which possess some of the properties of the hand; they are surrogate stimuli that, within the framework of a given representational convention, contribute to the signification; they are sheer material configurations that simulate perceptive conditions or components of iconic signs (Kalkofen, 1973, commenting on previous statements of mine – see Eco, 1968).
A certain type of naive theorizing will quickly identify the production of surrogate stimuli with iconism, therefore speaking of iconism à propos of phenomena that can only metaphorically be defined as such.
Let’s take an example. Everyday experience tells us that saccharine ‘resembles’ or has some properties of sugar. Chemical analysis demonstrates that no common objective property exists between the two compounds, even if we examine the structural formulae as well as the chemical components. Sugar (saccharose) is a disaccharid with the basic formula C12 H22 O11, while saccharine is a derivative of o-sulphamidebenzoic acid. Nor do we consider the relationship of visual resemblance, because in that case sugar would be much more similar to salt. So we can say that what we define as equal properties, shared by the two compounds, concerns not their form but their effect. Saccharine simulates perceptive conditions similar to those of sugar (even though in many respects it produces a different taste experience). Both are ‘sweet’. ‘Sweetness’ is not a property of the two compounds, but the result of their interaction with our taste buds. But this result is emically pertinent within a culinary culture that has opposed all that is sweet to all that is bitter on the one hand, and all that is salt on the other – isolating emergent features within the two phenomena that are not common to the two compounds alone, but to their interactive relation with the subjects’ palates. It goes without saying that for a refined cook ‘sweetness’ as such does not exist, but various types of sweetness do, so that saccharine moves into a relationship of diversity (rather than resemblance) with sugar.
So one can see how many problems have come to take the place of the presumed resemblance between the two compounds. Let us simply attempt to list them: a) formal (chemical) structure of the compounds, depending upon the point of view selected as pertinent by the analysis; b) structure of the perceptive process (interaction between compound and taste buds), which may be defined as equal or unequal according to the point of view selected as pertinent (according to the parameter ‘sweet vs. bitter’ it is “equal”, but according to others, for example ‘granular vs. soft’ it is “unequal”); c) structure of the culinary-semantic field, which determines the paradigmatic choice of pertinent elements in the recognition of equality and inequality. Within these three types of phenomena the presumed ‘naturalness’ of resemblances dissolves itself into a network of cultural stipulations that determine and direct ingenuous experience.
There is another and far subtler definition: that proposed by Peirce. According to him a sign is a icon when it “may represent its object mainly by its similarity” (2.276).
To say that a sign is similar to its object is not the same as saying that it possesses some of its properties. In any case this definition relies on the notion of ‘similitude’, which has a scientific status and is less imprecise than that of ‘sharing properties’.
Popular handbooks on geometry define similitude, roughly speaking, as the property shared by two figures that are alike in all respects except in size. Granted that difference in size is not a negligible feature (the difference in size between a lizard and a crocodile could matter quite a lot in everyday life) to decide to disregard it does not sound quite ‘natural’ and seems on the contrary to be culturally or conventionally founded; one decides to recognize as similar two things because one chooses certain elements as pertinent and disregards certain others. This kind of decision asks for a certain training; if I ask a child to compare a miniaturized school model of a pyramid and the Cheops Pyramid, asking if they are the same or ‘similar’, the most obvious answer would be “no”; only after a certain amount of training would my naive interlocutor be able to realize that I am looking for a ‘geometric’ similarity. The only unchallengeable impression of resemblance is due rather to phenomena of congruence, where two figures of equal size can be made to coincide or to fit on top of one another. But they must be two geometrical figures; a death-mask is congruent with the real face of a dead man by making abstraction from matter, color and differences in texture; and it is doubtful whether a non-trained informant would be able to recognize the former as the congruent reproduction of the latter.
But similitude can be even more exactly defined as the property shared by two figures that have equal angles and sides that are proportionally equivalent. Once again the criterion for similitude is based on precise rules that select some parameters as pertinent and disregard some others as irrelevant. Since the rule is proposed and accepted, there is undoubtedly a ‘motivation’ that links two equivalent sides (they are not related one to another by means of a mere convention); but in order to make the motivation detectable a conventional rule is needed. Optical illusions teach us that in many cases there are ‘motivated’ reasons for judging two figures as equivalent – and the motivation is rooted on psychological factors – but only when the geometrical rule is recognized, the parameters applied and the proportions checked, can the correct judgment of similitude be pronounced.
Geometrical similitude is based on spatial parameters chosen as pertinent elements; but in the theory of graphs one finds other forms of so-called similarity that are not based on spatial parameters; thus certain topological relationships, or relations of order, are chosen that cannot be translated into spatial relations if not by means of another cultural decision. According to the theory of graphs the three representations in Table 36 express the same relationships even if they are not spatially (and therefore geometrically) ‘similar’.
The three graphs convey the same information (for example about types of interdisciplinary correlation between six departments of the same university) but they do not realize the same geometrical properties. This happens because a convention has isolated not the spatial disposition of the buildings of the six departments, but the type of scientific collaboration that may take place between them. Suppose that F is a Physics Department, A a Philosophy Department, D a Mathematics Department and C a Theology Department: one can easily see that mathematics and physics have many topics in common and so have physics, mathematics and philosophy, while theology has topics in common with philosophy, but could hardly have much to do with mathematics and physics. From the point of view of ‘common scientific services’ the three graphs are said to be ‘isomorph’.
This kind of isomorphism may be called a form of similarity but it would be very difficult to assert that it is a geometric similitude. To call such a relationship ‘iconic’ is a mere metaphor.
Unfortunately this is exactly the kind of metaphor used by Peirce in his otherwise masterful little treatise on Existential Graphs (4.347-573), in which he studied the properties of logical diagrams. An existential graph is one in which the relations entailed by a syllogism like
All saints are men
Therefore all saints are passionate
are expressed by the geometrical form
while the syllogism
No man is perfect
But any saint is a man
Hence, no saint is perfect
is expressed by the geometrical form:
A propos of a diagram of this sort Peirce says that its beauty springs “from its being veridically iconic, naturally analogous to the thing represented, and not a creation of conventions” (4.368). This assumption sounds rather whimsical if one is accustomed to associate the idea of iconism with a visual relationship between similar spatial properties. It is true that the diagrams show spatial relationships but these spatial properties do not stand for other spatial relationships! To be or not to be passionate is not a matter of spatial distribution. In terms of classical logic it is a matter of possessing or not possessing a given property. The inherence of a property to a subject (the relationship subjectum-praedictum) is, however, a naively realistic concept; to have passions is not an accident that belongs to a subjectum, except in Aristotelian and medieval metaphysics, and even if it were so, the representation furnished by the first graph would have to be reversed. If it is not so, this is because the graph does not translate the classical notion of inherence of a predicate to a subject, but the modern notion of class appartenance. Anyway to make up part of a class is not a spatial property (except if one belongs to the class of all the men meeting in a given place); it is a purely abstract relation. How, in the graphic representation, does the appartenance to a class become appartenance to a given space? By a mere convention (even if based on certain mental mechanisms, used to thinking or to imagining either by temporal succession or by spatial proximity) that establishes that certain abstract relations can be expressed by spatial situations. Naturally the convention follows a proportional criterion of the type ‘space a : space b = entity a1 : entity b2’ just as in a geometrical similitude there is a criterion of proportionality between sides of apparently different size. Therefore there is a convention establishing the way in which a proportion (a particular kind of non-arbitrary motivation) has to be posited and interpreted. To call this complex kind of isomorphism ‘iconism’ is a mere metaphorical license.
Peirce takes a lot of similar licences a propos of iconism, and in some way does so pour cause; he is in fact trying to define that particular kind of relation between an expression and its content-type that we have called ratio difficilis. But Peirce does not abandon the reference to objects, and ‘iconism’ thus remains for him an umbrella-term that covers many different phenomena such as a mental image, a graph, a painting. A graph does in fact exhibit a certain proportionality between a given expression and a given content, that content not being an object but a logical relation. It represents an example of correlation between an expressive item and a content-schema taken as its type without passing through a process of verification with the object. It reinforces the opinion I have expressed in 3.4.9. that in cases of ratio difficilis what matters is not the relationship between an image and its object but rather that between an image and its content. The content, in this case, is the result of a convention, as is the proportional correlation. The elements of motivation exist, but they can only work when they have been conventionally accepted and coded.
Both geometrical similitude and topological isomorphism are a sort of transformation by which a point in the effective space of the expression is made to correspond to a point in the virtual space of a content model. What marks the difference between different sorts of transformation is both the mode of correspondence and the class of elements made pertinent by the conventionalizing procedure, so that only these must be retained as invariant while the others are varied. Thus some procedures aim to preserve metric properties, others topological properties, and so on. In any case there is a transformation, in the technical sense of the term. Every biunivocal correspondence of points in space (and let us consider as space even the virtual one of abstract relational content-models, as in the case of translation into spatial terms of the relations of appartenance to different class) is a transformation. A transformation does not suggest the idea of natural correspondence; it is rather the consequence of rules and artifice. Thus even the continuous line tracing the profile of the horse (see 3.5.2.) may be considered as the institution of a relation of similitude by a transformed correspondence point to a point between the abstract visual content model of a horse and an image drawn on a given surface. The image is motivated by the abstract representation of the horse, but it is nevertheless the effect of a cultural decision and as such requires a trained eye in order to be detected as a horse’s profile. Similitude is produced and must be learned (Gibson 1966).
In the light of what has been said up to now, is it possible to call so-called iconic signs analogous?
If analogy is a sort of native and mysterious parenthood between things or between images and portrayed things, one is obliged to reject so unverifiable a category. But if analogy is understood in the only sense that permits verification, then the notion can be accepted. Except that in this sense analogy and similarity appear to be purely synonymous terms.
Let us see whether one can explain what analogy is by observing the functioning of a so-called ‘analogical’ computer. This for example establishes that a certain intensity of current x denotes a physical size y, and that the denotative relation is based on a proportional one. Proportion may be correctly defined as a type of analogy, but not all definitions of analogy reduce themselves to that of proportion. (13) In any case, in order that there be a proportion there must be at least three terms. We cannot say “Intensity x is to size y” if we don’t complete it with “as size y is to . . . ”. We then realize that a computer is called analogical not because it establishes a constant relation between two entities, but because it establishes a constant proportionality between two series of entities, of which one is assumed as the sign-vehicle of the other. In other words the proportion depends upon the fact that if size 10 corresponds to intensity 1, size 20 will correspond to intensity 2, and so on. The relationship may only be defined as analogical because a correlation between a given intensity of current and a given physical size has been arbitrarily established from the outset, but the computer could make equally exact calculations if it had been established that the intensity 3 corresponded to size 9, intensity 6 to size 18, and so forth. As may be seen, it is not the analogy that institutes the relationship but the proportional relationship that renders legitimate the analogy. But why was it established that size y corresponds to intensity x? If one replies “arbitrarily” or “for economic reasons” then the problem no longer exists. But let us suppose that the reply is that it was done because there was an analogy between intensity x and size y. This analogy will not be a proportion, and other terms are lacking; so that it can be no better defined than as a “resemblance”. But to say of two entities that they resemble each other means that they have a reciprocal iconic relation. So this is why in order to define a form of analogy that is not a geometric proportion one resorts to the notion of iconism. We are thus faced with the absurdity of having semiotics resort to analogy to explain iconism, while invoking iconism to explain analogy. We have a petitio principii.
Thus even analogy, like similarity, does not exclude cultural convention; on the contrary it requires it as an operational starting-point. When on the other hand ‘analogous’ is used as a synonym for ‘unspeakable’ the term is being taken in its most vague and imprecise metaphysical sense. Once this halo of inappropriate connotations is eliminated, analogy proves to be what operationally it is: a procedure instituting the basic conditions for a transformation.
Since transformation seems to be, as yet, the best operational explanation of the impression of iconism, let us eliminate some embarrassing phenomena that might otherwise be supposed to come under the heading of ‘similarity’. I refer to (a) specular reflections; (b) doubles and replicas depending on the ratio facilis; (c) so-called ‘expressive’ signs.
Specular reflection could be called a sort of congruence, insofar as congruences are a type of equality, thus establishing a bi-univocal relation founded on the properties of being reflexive, symmetrical and transitive. In this sense specular reflection is equality and not similitude.
But the first thing to make clear is that a specular reflection cannot be taken as a sign if one follows the definition given in this book. Not only can it not be properly called an image (since it is a virtual image, and therefore not a material expression (14)) but even granted the existence of the image it must be admitted that it does not stand for something else; on the contrary it stands in front of something else, it exists not instead of but because of the presence of that something; when that something disappears the pseudo-image in the mirror disappears too (15)). Even admitting that what happens in a camera obscura is something ‘similar’ to the phenomenon of specular reflection (which is not questionable), what changes is the fact that an image remains traced somewhere, and any successive discussion about its iconic properties deals with the imprinted image and not with the process of projection itself.
The singularity of specular reflections is demonstrated by the fact that if one tries to apply to them the schema of a communicational process many puzzling conclusions arise: source and addressee coincide (at least in cases where a human being looks at him or herself in the mirror); receiver and transmitter coincide; expression and content coincide since the content of the reflected image is just the image of a body, not the body itself; as a matter of fact the referent of a mirror image is pure visual matter.
The image in a mirror is not a sign for it and cannot be used in order to lie (if not by producing a false object to be reflected, but in this case what stands for the supposed object is the false body, not its reflection).
The second phenomenon that we shall not consider as a case of iconism is the existence of doubles (see 3.4.7.). A double is not the icon of its model-object except in a very specific case: i.e. when an object is chosen as an ostensive sign in order to visually describe the character of every object of the same class (see 3.6.3.). (16)
The third exclusion concerns replicas ruled by a ratio facilis. When governed by a ratio facilis, replicas are such insofar as they reproduce certain pertinent features established by their type; and particularly when dealing with signs whose expression is based on spatial parameters, it would seem that such features ought to be reproduced by means of a certain degree of similarity. A road signal is the same as its type just as an uttered phoneme is the same as its emic model, because in both cases a relation of similarity is established. Why would we not assume that the actual recognition of token signs is governed by a principle of similarity and is therefore an example of iconism?
First of all, the type of a phoneme or of a replicable visual item also prescribes the material continuum in which the token should be realized. This does not happen in the case of so-called iconic signs (governed by the principles of ratio difficilis), and it is precisely because of this that transformational rules are needed; two similar triangles remain so even if one is drawn on a sheet of paper and the other is cut from linen.
Secondly the presumed ‘iconism’ that should govern the correspondence of a token to its type is not a theorem that semiotics could demonstrate; it is one of its postulates. The very notion of a sign and of its replicability (and thus of its social nature) depends on postulating that such a recognition is possible. The rules of this recognition are deeply rooted in the mechanisms of human perception and must be assumed as already given in any semiotic enquiry. Thus a token is not the sign of its type (even if under some circumstances it can be assumed as such, thus becoming an ostensive sign; see 3.6.3.).
The partial replica (like the absolute one) does not concern the expression as a functive of a sign-function; it concerns the expression as a signal, and the conditions for a good partial replica are a matter for information engineering (or phonetics, or some other science).
The problem changes when the conditions of replicability concern the expression as a functive, i.e. when the signal’s production procedures affect not only its recognizability as a signal, but also the recognizability of the expressed content. These are cases of ratio difficilis in which the model for the replica is a content-type.
Finally I propose not to consider as ‘iconic’ the so-called ‘expressive’ properties of certain signals, by which they are supposed to ‘induce’ a feeling of similarity between the signal and a given emotion.
Many artists (see for instance Kandinskij) have theorized extensively around the fact that a certain line can suggest (or ‘express’) a feeling of force, another a feeling of weakness, another a feeling of stability, another a feeling of imbalance, and so on. The psychology of empathy (Einfühlung) has studied such phenomena, which do undoubtedly take place in our perception of many so-called ‘signs’ traced by human hands or found in the natural environment.
We may consider all these cases of empathy as mere stimulations that should be studied by the physiology of the nervous system. In a semiotic framework it does not make sense to assert or to deny that these ‘expressive’ properties of signals are based on the ‘universal’ structure of the human mind or that they can vary according to many biological or cultural variables.
Nevertheless such phenomena could be taken into account by semiotics in two cases: (i) when the precise effect usually elicited by a stimulus is culturally recorded so that the stimulus acts as a conventional sign for the sender, though not for the addressee, as the former possesses a sort of code of stimuli (see 3.6.6.); (ii) when a given effect is clearly due to a cultural association and a certain signal does not suggest, let us say, feelings of ‘grace’ because of a ‘natural’ and ‘universal’ structure of the mind but because of a conventional and coded link between that signal and that feeling; in this second case we are dealing with a sign-process, but it is not an iconic one.
In these two cases one can speak of programmed stimulation. Insofar as the expression relies on more or less coded units and the reaction of the addressee cannot be completely foreseen, one must consider programmed stimulation as the invention of a complex text; as such it will be examined in 3.6.7.
With these potential mistakes eliminated, we must now give a clearer definition of real similarity. In this paragraph I shall stress that to say that a certain image is similar to something else does not eliminate the fact that similarity is also a matter of cultural convention; in the following one I shall try to demonstrate that similarity does not concern the relationship between the image and its object but that between the image and a previously culturalized content. Both points are strictly related to each other. As regards the first point, there are many proofs to support this assumption.
Ernest Gombrich has emphasized the conventionality of imitative codes in his Art and Illusion (1956), where for example he explains what happened to Constable when he elaborated a new technique for portraying the presence of light in a landscape. Constable’s painting Wivenhoe Park was inspired by a poetics of the scientific rendering of reality and to us seems decidedly ‘photographic’, with its detailed portrayal of trees, animals, water and the luminosity of a patch of field caught by the sun. And yet we know that when his works appeared for the first time no one felt that his technique of contrasting tones was some sort of imitation of the ‘actual’ effects of light, but rather that he was taking a strange liberty. Constable therefore had invented a new way of coding our perception of light, and of transcribing it onto canvas (17).
Naturally we succeed in understanding a given technical solution as the representation of a natural experience because there has been formed in us a codified system of expectations, which allows us to enter into the semantic world of the artist. Maybe an ‘iconic’ solution is not conventional when it is proposed, but it becomes so step by step, the more its addressee becomes acquainted with it. At a certain point the iconic representation, however stylized it may be, appears to be more true than the real experience, and people begin to look at things through the glasses of iconic convention. Gombrich gives many astounding examples of this sort of perceptual cramp caused by overwhelming cultural habits.
Villard de Honnecourt, the architect and artist of the thirteenth century, claimed to be copying a real lion, and yet reproduced it according to the most obvious heraldic conventions of the time. His perception of the lion was conditioned by current iconic codes; or else his codes of iconic transformation prevented him from transcribing his perception in any other way; and probably he was so used to his own codes that he thought he was transcribing his perceptions in the most suitable possible way. Diirer portrayed a rhinoceros covered with scales and imbricated plates; as a result this image of the rhinoceros remained constant for at least two centuries and reappeared in the books of explorers and zoologists; and although these latter had seen actual rhinos and knew that they do not have imbricated plates, they were unable to portray the roughness of their skin except by imbricated plates, because they knew that only these conventionalized graphic signs could denote «rhinoceros» to the person interpreting the iconic sign.
But it is also true that Diirer and his imitators had tried to reproduce via these means certain perceptual conditions that photographic reproduction does not convey. In Gombrich’s book Diirer’s drawing certainly appears ridiculous in comparison with the photo of an actual rhinoceros, which seems to have an almost smooth and uniform skin; but if we were to examine the skin of a rhinoceros close to, we would notice such roughness that, from a certain point of view (in the case, for example, of a parallel between human skin and that of the rhinoceros), Diirer’s graphic exaggeration, which pays excessive and stylized attention to that roughness, would be rather more realistic than the image in the photograph which by convention portrays only the great masses of color and makes the opaque surfaces uniform, distinguishing them by differences of tone. Thus one could say that Diirer’s rhinoceros is more successful in portraying, if not actual rhinoceroses, at best our cultural conception of a rhinoceros. Maybe it does not portray our visual experience, but it certainly does portray our semantic knowledge or at any rate that shared by its addressees.
This goes to show that signs ruled by a ratio difficilis are motivated, but r iainly by a content-form. If one considers the most common procedures in visual representation with some degree of phenomenological detachment, the above assumption can be made more acceptable.
Suppose that, after having drawn the horse quoted in 3.5.2., one wants to transform it into a zebra. Provided that image is a very schematic one, one need only draw some black sinusoidal stripes on the back of the horse and the transformation will be complete. It may not even be necessary to draw sinusoidal stripes; the mere presence of stripes may be quite sufficient.
This means that we have selected the fundamental aspects of the percept on the basis of ‘recognition codes’; if at the zoo we see a zebra in the distance, the elements which we recognize immediately (and remember) are its stripes, and not its profile, which vaguely resembles that of an ass or mule. And so when drawing a zebra we try to make its stripes recognizable even though the outline of the animal is very similar to that of a horse, and without the stripes would probably be interchangeable. But let us suppose that there exists an African community where the only known quadrupeds are the zebra and the hyena, and where horses, asses and mules are unknown. It is no longer necessary to see stripes in order to recognize the zebra (it could even be identified in the dark, without being able to distinguish the skin) and when drawing a zebra, it would be more important to emphasize the shape of its muzzle and the length of its legs in order to distinguish it from the hyena (which also has stripes, so that the stripes no longer constitute a differentiating factor). Presuming that they exist, recognition codes (like any others) make provision for conveying pertinent features of the content. The recognizability of the iconic sign depends on the selection of these features. But the pertinent features must be expressed. Therefore there must exist an iconic code which establishes the equivalence between a certain graphic device and a pertinent feature of the recognition code.
Here I might introduce an experience that I had with my own son when he was four years old. I once found him lying on his stomach on top of a table, and pivoting on it he began to spin round like the needle of a compass, with his arms and legs stretched out. He said: “I’m a helicopter”. On the basis of his own recognition codes he was extracting from the complex structure of the helicopter the fundamental feature by which it is distinguished from other machines – its rotary blades; of the three rotary blades he had retained only the image of two blades opposite each other – the elementary structure through whose transformation one arrives at various groupings of blades; he had retained the basic geometric relation between the two blades, a straight line pivoting upon its center and rotating through 360 degrees. And having grasped the basis of this relationship, he was reproducing it in and with his own body. At this point I asked him to draw a helicopter, thinking that, since he had grasped its elementary structure, he would have been able to reproduce it in his drawing. Instead he drew a clumsy central structure around which he stuck in any order an indefinite number of parallelepipedal forms, as if the object were a porcupine, explaining: “And here there are lots and lots of wings”. When he used his own body, he reduced the experience to an extremely simple structure, but when he used a pencil he made the object into a fairly complex one.
Now clearly with his body he was also imitating the movement of the blades, while in his drawing he had to suggest this movement through the addition of more wings; but he could have portrayed the movement as an adult would have, for example by numerous straight lines intersecting at the center, so as to form a star. The fact is that he was not yet capable of putting into graphic code the type of structure which he had so well succeeded in portraying with his body. He had perceived the helicopter, and worked out models of recognition, but he was not able to fix the equivalence between a conventional graphic device and a pertinent feature of the recognition code. The definition of an iconic sign as possessing certain properties of the denoted object becomes even more problematic at this point.
Are the properties which it has in common with the object seen or known? During the same period my child drew the outline of an automobile with all four wheels in sight; he identified and reproduced the properties which he knew. Later he learned to code his graphic procedures and portray the automobile with two wheels (the other two, he explained, were behind); now he was reproducing only the properties which he saw. The Renaissance artist reproduces the properties which he sees, the cubist painter those which he knows (but the general public is on the whole used to recognizing only those which it sees and does not recognize on canvas those which it knows). Thus the iconic sign may possess: (a) optic (visible), (b) ontological (supposed), and (c) conventionalized properties of the object. By conventionalized properties I mean those depending on an iconographic convention which has catachresized the previous creative rendering of an actual perceptual experience. A typical example is provided by the iconographic representation of the sun as a circle from which many short lines radiate. The original experience of the sun was gained by looking at it through partly closed eyes. In this case it looked like a shining point from which emanated uneven rays. If we accept a certain graphic convention these rays can be represented by so many straight black lines and the shining point by a white circle. Consequently the iconographic convention (which codifies the original experience) also seems adapted to our sophisticated and scientific experience of the sun, which is understood to be an incandescent sphere from which emanate ‘rays’ of light. But the scientific notion of a luminous ray is an abstraction which has been influenced by precisely the classical iconography referred to above and by the Euclidean geometry which went with it. Whether light is understood as quanta or waves, it has nothing to do with the conventional feature which denotes a ray. Yet this schematic representation of the sun seems to imitate the scientific idea of the sun rather well.
The fact is that relationships do indeed exist but not between the image and the sun as object, but between the image and the abstract model of the sun as a scientific entity. Thus a schematic representation reproduces some of the properties of another schematic representation.
The ‘iconic’ code thus established either a correlation between a graphic sign-vehicle and an already coded perceptual unit, or one between a pertinent unit of the graphic system and a pertinent unit of a semantic system depending on a previous codification of perceptual experience.
We have seen that the umbrella-term ‘iconism’ covers many different phenomena: some of them having nothing to do with signification (specular reflection, duplicative replicas, mere stimulation), others being strung out along a sort of gradated continuum from minimal convention (congruences, such as the death mask) to maximal convention as in stylization or characterization (the heraldic-like representation of the sun as circle).
In section 3.6.7. we shall have to return to this gradated continuum of possibilities. But let us first consider certain other phenomena that are commonly considered iconic and that might be classified in a different way, since they involve different sign production procedures and only give an impression of similarity, as opposed to realizing the type of similitude defined in 3.5.3. People usually say that icons are similar to their object because they embody some of its features. They know that such signs do not imitate (or reproduce) all the features of the object, but assert that, provided certain emergent features are present, these will suffice to establish the impression of similarity. The most prudent attitude in this case is to propose that the iconic device may possess certain elementary iconic markers and that sometimes a minimal resemblance is due to the fact that the iconic sign, even though different in shape from its object (the above-mentioned elementary markers apart), performs the same function.
What I shall try to demonstrate is that both the elementary markers and the acknowledged identity in function are not the result of, but the constitutional operation which gives rise to, the impression of iconism. Gombrich (1951) in his essay on the hobby horse (which, in its most elementary form, is simply a broom handle that the child uses as if it were a horse), notes that the relation of presumed iconism is not based on any formal similitude, other than in the rather vague sense that the horse possesses as a pertinent feature a linear dimension that is also to be found in a stick.
In fact the only aspect that the stick has in common with a real horse is that it can be straddled; hence the child has rendered emergent or pertinent – on the basis of his own physical and psychological motivations – one of the functions permitted by the horse. He has found an object that functions us Ersatz of a horse not because it ‘resembles’ it, but because it serves an analogous function.
The example of the stick given by Gombrich is revealing. It can become the icon of a horse, a sceptre, or a sword. The element that recurs in all these objects (and that the stick seems to reproduce) is the feature of linearity (vertical or horizontal). But it cannot be said that the stick’s vertical quality ‘imitates’ that of the spade; insofar as this vertical quality is a feature of both objects it is the same verticality. We are thus confronted by a classic intrinsically coded act, or rather by signification through the use of a part of the referent.
The most recent studies of kinesics (cf. for example Ekman and Friesen, 1969) bring into evidence gestural signs which are not entirely arbitrary (as are the conventional signs for yes and no) but are based on a certain similarity to the object represented, thus constituting kinesic iconic signs.
One example could be that of a child who points the finger of his right hand and pretends it is a pistol and that his thumb is the hammer (accompanying the gesture with an onomatopoeic sign which represents shooting). But there are other signs which are not directly iconic, and which Ekman and Friesen call intrinsically coded acts. Let us use the example of the pistol again; the boy may realize the same fantasy by moving his bent finger as if it were actually pressing on the trigger, while the other fingers are closed in a fist as if they were clutching the butt. In this case the gun is not being imitated. Nor is the act of shooting being imitated. The act of shooting is being denoted by means of a gestural sign-vehicle which is otherwise only physically present as part of the supposed referent (there is no pistol but there is a hand gripping it; and the significant gesture is precisely that of a “hand grasping a pistol”).
So that here we have a part of the referent used as a sign-vehicle, or a part of the object, metonymically used for the whole (see for a further analysis Veron, 1970; 1973; Farassino, 1973; a different theory of these gestures will be proposed in 3.6.3.).
Signs of this kind are common enough in daily use. If a barber can display a cylinder with red, white and blue stripes to denote his presence we have a symbol in Peirce’s sense, an arbitrary sign; if he displays a placard on which is drawn a razor we have an icon; but if, as some barbers used to do, he displays a bowl used to soap the customer (Mambrino’s helmet in Don Quixote), part of the complex of objects denoted by the sign-vehicle becomes – by metonymy – the sign-vehicle itself. A part of the referent is semioticized and arbitrarily taken as symbolic of the whole complex to which it refers. In this way many so-called iconic signs become reclassified as examples of intrinsically coded acts. The red that appears in the drawing of a red flag is not ‘similar’ to the red of the real flag: it is the same red. If this is true, one could again accept Morris’s definition, according to which the iconic sign possesses some of the properties of its denotata, or Peirce’s, according to which it “refers to the object. . . by virtue of characters of its own”. However, to arrive at the icon of a red flag a red-colored blob is not enough; the red must be contained within a square, or a rectangle, or a parallelogram with undulating sides, etc. This geometrical feature is not something that belongs to the object flag in the same way that the color red belongs to it: the square in the drawing is only ‘similar’ to that of the piece of cloth from which the flag is made. The difficulty in defining the characteristics of an iconic sign is not only caused by this multiplicity of relationships (in the case of the flag we have on the one hand the presence of the referent itself, and on the other a ‘similar reproduction’ of a characteristic of the referent); it is also caused by the fact that certain relationships that seem to belong to the same category in fact belong to different ones. For example, it has been stated that the pertinent property ‘vertically’ is the same in the stick and the horse; so why is the ‘squared’ nature of the flag not the same as that of the drawing of the flag? We are here confronted by two differing levels of abstraction: linearity (verticality or horizontality) is a spatial dimension, it represents a mode of perceiving and choosing space, while a square is already a figure constructed in space.
A reference to ‘old’ philosophical questions would not be out of place in order to clarify this difference; in Kant’s first Kritik space, like time, is a pure intuition, the elementary form that we confer upon experiential data in order to be able to perceive them and place them within the categories. Notions such as vertically and horizontality are not therefore intellectual abstractions, but the intuitive mode within which we frame our perceptions. According to Kant, they are to be studied by Transcendental Aesthetics, whereas geometrical figures ought to come under the aegis of Transcendental Logic, and more precisely that of the Analytics of the Pure Principles of the Intellect (being based upon the Axioms of Intuition), and thus a priori constructions that make possible the application of formal categories to sensory data. Cassirer (1906:20) observes that “space and time are nearer to empirical material than are the categories”: and this would explain why a spatial determination such as verticality can give rise to an intrinsically coded act (can, that is, be a sort of concrete experience capable of being used as the sign of itself); whereas the notion of a square – an intellectual construct, and thus an abstraction that never exists de facto – cannot constitute a referent used as the sign of itself, and must thus give rise to a reproduction of itself.
It is merely the limitations of linguistic usage that make us understand “vertical properties” and “squared properties” as abstractions on the same level. In reality, we are here concerned with differing levels. The spatial dimensions are not an intellectual construction, but the constructive conditions for a possible object, and as conditions they may be reproduced, equal to themselves, in varied circumstances. On the other hand the idea of a square is already an object constructed within the framework of such conditions, and it cannot be reproduced as equal to itself, but only as an abstraction similar to previous constructions.
This doesn’t stop the stick, which reproduces within itself the condition ‘Verticality’, from standing for the horse, nor the square, which reproduces a geometrical construct that is ‘isomorphic’ to the form of the flag, from being a sign that stands for that flag. On a primary semiotic level both are signs. It is simply that the first does not pose problems of iconicity, and the second does. Iconism is not a homogenous phenomenon, but a label that covers differing phenomena that have yet to be completely classified and analyzed.
That the linearity of the stick is not a construct but a condition of every other possible construction (and thus an intuitive artifice able to determine a space) is also demonstrated by another fact. What constitutes the reliability of the substitution stick-horse and stick-sword is not the mere presence of a vertical object, but also the presence (in the case of the horse) of a body astride it and (in the case of the sword) of a hand grasped round the presumed hilt. So much is this the case that we even have the imitation of a horse when the child caracoles without anything between his legs, and the imitation of a sword when he moves his closed fist (with nothing in it) and pretends to fence. Verticality plus the presence of the gesture (which is not an ‘imitation’ of the gesture but the real gesture that would be made if the real object were present) constitute the imitation not of a single object but of an entire mode of behavior. Throughout this process, within which use is made of intrinsically coded acts or contiguous signs, iconism (in the classic sense of the term) makes no appearance, and to have talked about it in this way constitutes a curious example of optical illusion.
The only object that appeared to have an iconic function, instead, serves to impose the necessary spatial conditions for the realization of an intrinsically coded act.
If in the gesture of the child caracoling on the hobby horse there is something that can be called iconic, this is so because: (a) a linear dimension has been used as an expressive feature in order to substitute for the linear dimension that equally, but to a very limited degree, characterizes a horse as such; (b) one part of an entire behavioral pattern, functioning as an intrinsically coded act, has been used as an expressive device in order to convey the idea that the stick is a horse. But at this point one would do well to isolate features of expression from those of content; if the same feature appears to be both conveying and conveyed, how does one set about analyzing this sign? Since one can hardly deny that the hobby horse is a sign, the only solution is to better distinguish the imitans from the imitatum, that which stands for something from that for which something else stands.
A final ambiguity linked to the idea of resemblance is caused by the fact that – on the level of very elementary formal phenomena such as high-low, right-left, or long-wide – everything resembles everything else. Which means that there exist certain formal characteristics that are so generic as to belong to almost all phenomena, and that they may be considered iconic in relation to all phenomena. Jakobson (1970a) points out some differing cultural conventions in which, for instance: a) yes is expressed with a downward movement of the head and no with a movement from right to left; b) yes is expressed with a downward movement and no with an upward movement; c) yes is expressed with a lateral movement and no with an upward movement. One is obliged to conclude that signs of consent and dissent are arbitrary, but Jakobson finds iconic motivations for some of them: bending the head forward to say yes recalls movements of submission; moving it laterally to say no expresses the desire to turn the face away from the interlocutor; lifting the head up to say no expresses a desire to separate oneself from the interlocutor and place him at a distance. However, this does not explain why someone who says no by throwing back his head will then say yes by turning it sideways; in the absence of an iconic explanation Jakobson resorts to a systematic one, and points out that (given a previously established iconic form) its antonym arises by purely formal opposition. It is in fact possible to explain the yes expressed by moving the head sideways iconically as well: it could signify the desire to offer one’s face to the interlocutor repeatedly, rather than the desire to turn it away a number of times (as occurs when the same gesture is used to express no). The truth of the matter is that right and left or forward and backward are such universal features that they are able to become iconic reproductions of every phenomenon. So that if it is legitimate to trace back a number of arbitrary gestures to underlying iconicity it is equally legitimate to trace apparently iconic relationships back to arbitrary underlying codifications.
In the case examined above people had mistaken for an icon what was in fact a constitutive condition for the impression of iconism. In the cases examined in 3.5.6., we discovered elements of conventionality at the heart of ‘iconic’ procedures. One might thus be inclined to an opposite and equally dogmatic conclusion, i.e. that iconic signs are entirely a matter of convention (just as verbal signs are) and that they ought to be susceptible to multiple articulation and complete digitalization. So – in opposition to the usual equation ‘iconic=analogical=motivated=natural’ – one might reverse the situation identifying conventional (and cultural) with arbitrary, and arbitrary with digital (though this was demonstrated as incorrect in 3.5.1.).
But since the possibility of some degree of articulation in visual signs has featured notably in recent research, I should like to return to this problem once more. Particularly so because the denial of such forced identifications and the assumption that, beyond a certain limit, so-called iconic signs are not articulable into smaller units may give rise to the possibility of a new classification, based on the difference between grammar-oriented and text-oriented procedures (see 2.14.6).
The most naive way of formulating the problem is: are there iconic sentences and phonemes? Such a formulation undoubtedly stems from a sort of verbocentric dogmatism, but in its ingenuousness it conceals a serious problem.
Everyone accepts that images convey a certain content. If one tries to verbalize that content one finds some easily recognizable semantic units; for instance, a meadow with a house, two horses, a dog, a tree, a girl. Are there within the image expression-units corresponding to these content units? If the answer is yes, the next question is: are these units coded, and if they are not, how can they be recognized? Supposing that such units exist, are they open to analytical subdivision into smaller and meaningless units (by combining a limited number of which, infinite major units can be generated)?
We have seen that in order to realize iconic equivalents for perception only certain pertinent aspects are selected. Children under four do not acknowledge the human torso as a pertinent feature and portray man with only head and limbs. But if on the level of large-scale units of recognition it would be possible to isolate pertinent features, on the ‘microscopic’ levels the problem is more confused. The presence of discrete units in verbal language is found on all levels: from lexical units to phonemes, and from phonemes to distinctive features, everything would seem open to analysis. On the level of the supposed iconic codes, however, we are confronted with a more confused panorama. The universe of visual communications reminds us that we communicate both on the basis of strong codes (such as language) and indeed very strong ones (such as the Morse code) and on the basis of weak codes, which are barely defined and continuously changing, and in which the free variants prevail over the pertinent features.
In the English language there are various ways of pronouncing the syntagm [gud] with different intonations and accents; but nevertheless certain distinctive features remain which are not redundant and which thus define the limits within which a given ‘etic’ emission may be recognized either as /good/ or as /god/.
The borderline between [u] and [o] is strongly coded. But in the realm of graphic representation, I can make use of an infinite number of ways of portraying a horse. I can evoke it with a play of light and shadow, I can symbolize it with painstaking brushwork or define it with extreme realism (at the same time I can show a horse standing, galloping, three-quarters, rampant, with his head bent to eat or drink, etc.). Admittedly I can also say / horse/verbally in a hundred different languages and dialects; but as long as I use languages and dialects, no matter how many, they can be codified and listed, whereas the thousand different ways of drawing a horse are not foreseeable. On the other hand, verbal expressions are only comprehensible to those who know the languages concerned, whereas the hundreds of different ways of drawing a horse can be understood even by whose who are not acquainted with visual conventions (except in cases of a high degree of schematization). Therefore we find ourselves faced with the fact that there exist large-scale blocks (texts) whose articulatory elements are hard to discern.
One can make use of a series of commutations in order to find out – given, for example, the outline of a horse – what features would it be necessary to change in order to affect its recognizability; but this operation only allows for the coding of an infinitesimal sector of the process. In other words, although one gets the impression of understanding a text, one does not know how to decode it. In an iconic text such complex contextual relations are involved that it seems difficult to separate pertinent units from free variants.
One can isolate pertinent discrete units within iconic continuum, but as soon as they are detected, they seem to dissolve again. Sometimes they are large conventionally recognizable configurations, sometimes merely small segments of line, dots, black areas (as in a drawing of the human face, where a dot represents the eye and a semicircle the lips; yet we know that in a different context the same type of dot and the same semicircle would instead represent, say, a banana and a grape pip). Thus iconic figurae do not correspond to linguistic phonemes because they do not have positional and oppositional value.
These pseudo-features can sometimes assume contextual meanings (dot=eye, when placed in an almond-shaped form) but they are not organized into a system of rigid differences so that the value of a dot is determined by presence or the absence of a straight line or of a circle. Their positional value varies according to the convention instituted by the context. Thus we find a mass of idiolects, some of which are recognized by many, some by a select few; free variants far outweigh pertinent features, or rather free variants become pertinent features and vice versa according to the context. So iconic codes – if they exist – seem to last l’espace d’un matin, and as such look very similar to the aesthetic idiolects defined in 3.7.6 (18).
This also helps us to understand why a person who speaks does not seem to be born with any special ability, but if someone can draw, he already seems ‘different’ from others, because we recognize in him the ability to articulate elements of a code which does not belong to the whole group; and we recognize in him an autonomy in relation to normal systems which we do not recognize in any speaker except a poet (see Metz, 1964:84).
So that, at this juncture, anything taken as an iconic sign must be viewed as: (a) a visual text which is (b) not further analyzable either into signs or into figurae.
An iconic sign is indeed a text, for its verbal equivalent (except in rare cases of considerable schematization) is not a word but a phrase or indeed a whole story; the iconic representation of a horse does not correspond to the word /horse/ but rather to a description (a black horse, standing up, or jumping, etc.), to a mention (this horse is galloping) or to some other different speech act (look, what a beautiful horse!). If inserted in a scientific text, an iconic sign can correspond to a statement of the type /all horses have four legs and such visual properties . . . /.
The units composing an iconic text are established – if at all – by the context. Out of context these so-called ‘signs’ are not signs at all, because they are neither coded nor possess any resemblance to anything. Thus insofar as it establishes the coded value of a sign, the iconic text is an act of code-making.
Thus iconic signs are partially ruled by convention but are at the same time motivated; some of them refer to an established stylistic rule, while others appear to propose a new rule. In certain texts only large-scale coding is permitted, i.e. a prudent undercoding. In other cases the constitution of similitude, although ruled by operational conventions, seems to be more firmly linked to the basic mechanisms of perception than to explicit cultural habits. Some phenomena commonly called iconic turn out not to be so. At the furthest boundary of our enquiry we have encountered texts that seem to be innovatory rules ‘promising’ future semiotic possibilities, rather than signs.
One and only one conclusion seems possible at this point: iconism is not a single phenomenon, nor indeed a uniquely semiotic one. It is a collection of phenomena bundled together under an all-purpose label (just as in the Dark Ages the word “plague” probably covered a lot of different diseases). As we have seen, some of these hidden phenomena can in no sense be viewed as semiotic, and indeed others are not iconic at all. But granted this conclusion, a methodological principle is brought sharply into focus: it is the very notion of sign which is untenable and which makes the derived notion of ‘iconic sign’ so puzzling.
The notion of sign is untenable when confused with those of significant elementary units and fixed correlations; there are on the contrary ‘signs’ that result from the correlation of an imprecise expressive texture and convey a vast and unanalyzable portion of content; and there are expressive devices that convey different content according to different contexts, thus demonstrating what had already been assumed in 2.1., i.e. that sign-functions are the frequently transitory result of processual and circumstantially based stipulations. But it is not only iconic signs that are circumstantially sensitive. They cannot be classified as a unique category since, as we have seen, some of the procedures that rule so-called iconic signs can circumscribe other kinds of signs, while many of the procedures that govern other kinds of signs enter into the definition of the so-called iconic ones.
What we have succeeded in isolating up to now have been modes of producing sign-functions, not types of signs. There is a radical fallacy in the project of drawing up a typology of signs. But if instead one classifies modes of sign production, one can include both grammatically isolated sign-functions and more global textual units which assume the role of large-scale (undercoded) sign functions, such as the so-called ‘iconic signs’ considered in this section: macro-units which undoubtedly have a significant function but in which it is impossible to isolate signs as grammatical units (19).
A problem naturally arises: when dealing with these macro-units can one speak of codes? Are there non-coded macro-significant units (which would bring us back to the difference between analogy and arbitrariness)? All these problems will have to be discussed in the next section, where I shall deal with a typology (not of modi significandi but) of modi faciendi signa.