CIHA London 2000.
Thirtieth International Congress of the History of Art
Art History for the Millenium: Time.
Section 23
Digital Art History Time
London, 3-8 September 2000

Claus Pias <>, Bauhaus-University Weimar, Faculty of Media, History and Theory of Artificial Worlds, Bauhausstrasse 11, 99423 Weimar


Überlegungen zum "Wissen" von Bildern


reflexions upon the "knowledge" of images

© each author has full responsibility in owning copyright on the texts and on the images they publish on this website


Ladies and gentlemen, time is restricted to 20 minutes and for that reason my lecture will be a somehow incompatible blend of highly specialized examples and very generalized implications - a mixture of far-stretched theory and myopic detail. I would beg you to understand this as a kind of small, aphoristic form which considers extensive questions confronting the discipline of art history and media studies today.

My title implies that the lecture deals with the "knowledge of images", and perhaps you ask yourself if images contain knowledge at all. One would suggest that images have no knowledge themselves but are objects that enable the production and organization of knowledge by an academic discipline called art history. Images don't know anything - they just fill in a certain gap in the construction of this science. And this is the place of the silent, the dead, the naïve object that doesn't speak for itself but makes the art historian speak.

Now there is no knowledge without media, and there is nothing in our knowledge that we don't know via media. By media I mean all those instruments, institutions, techniques and technologies that make it possible to know something without themselves being an object of knowledge. A library saves and organizes data without being the data itself. If we look at a drawing of a building that follows the laws of central perspective we don't see the rules of central perspective but a building. If we listen to a telephone call we hear someone calling but not the telephone itself. There is a strong tendency for the media to become invisible, to extinct itself in favor of its "subject". Media allows something to appear without becoming an appearance itself. Media becomes apparent only when it doesn't function, if there's noise in the telephone line, if a computer crashes, if the perspective isn't right. Media is the dark, the anaesthetic side of aesthetics. But everything that is stored, submitted or processed by media has to follow the laws of a specific medium. Media always formats what it pretends to mediate in an unconcerned way. In Heideggers' terms, media is always "guilty", and it tries to eradicate the traces of its deeds.

We all know that every change in media history has a deep impact on what "knowledge" means. Remember the new definition of scientific knowledge emerging from the printing press. Remember the new definition of scientific fact by the experiment in natural sciences. Remember the new status of the event attained through statistics.

In a similar sense, the media of art history defines what an image is. An image is something that has [an address] its place: in a museum, a library, an image bank, in a frame or in a text, or more generally: in a designated space that is reserved for objects that we call images. And the image is always something different than the medium that makes it appear as an image. When images went digital, we followed this tradition according to McLuhans' statement that new media always contain older media. Image files are strictly separated from program files on our hard disks. Programs appear as menu bars on the upper border of our screens - images appear in separate frames called windows. In databases we sort images according to the traditional criteria of image banks, museums and archives: by author, by date, by technique, by size or topography. But the computer as a medium determines in a new sense what an image is: an image is something not only with a designation as a complete entity but one which is addressable in every single part without exclusion. The image consists of a number of discrete symbols that are arranged in a standardized format for images (TIF, GIF, JPG etc.).

Well, data is something that wants to be processed. But that is exactly what we - as art historians - don't usually do when we deal with images. The logic of the museum, of the image bank, and of the copyright laws that we transcribe to the computer tells us that the image is something intangible. That's why we don't work with data but with images. Art history doesn't become digital, it just becomes digitized.

Some art historians now suggest that the computer could 'make sense' by integrating it into their every-day work. Others strictly refuse to work with things like computers. According to the sociologist Karl Mannheim we might call these positions 'liberal' and 'conservative'. The conservative position believes in a progress of differentiation of today's structures as the best form that we can realistically achieve. The liberal position believes in a better future but does not encourage quick or radical attempts to realize it. Although those two positions seem to be contradictory, they have something in common because they don't go beyond the media-historically dated epistomology of art history. Both positions don't usually ask how the type of art history has emerged that allows something within it to 'make sense'. So the biggest problem with the liberal position is, that the category of logical or sensible improvement is derived from disciplinary conditions which have historically emerged from media which are totally different from the computer. As I understand it, the computer as a medium demands analysis of the discourse of art history in itself. The computer should not be regarded merely as a useful instrument or tool for research.

In reference to Heidegger's dictum - that the quality of a science is determined by the ability of this same science to analyze and criticize its fundamental terms or concepts - let me make the following, somewhat anarchistic suggestion.

How does art history work?

Art historians (and I am referring to the academic field) deal with texts and images. The image is an object of knowledge constructed by discursive agreements, and the history of art history might be described by questions like: what could be an object of knowledge at a certain time, what could be said about it, who is allowed to talk about it and from which position, etc. The image is - in Foucaults terms - determined by the 'archive' as the "law of what could be said", as the "system that regulates the appearence of statements as certain statements."

On their material side, the images of academic art history are photographic, technically reproducible images. Photography could become the material basis of art history because its apparatus works like a sense ("Sinnesorgan") but doesn't differentiate and records more detail in less time than the eye is capable of doing. Photography seems to provide a reliable tool for establishing a database for art history because it seems to be able store image data without processing it. Art history is based on an exchange, on an archive of substitutes. In its institutionalized gloominess it confirms Oliver Wendell Holmes' dictum that one could burn things down after taking photographs of them.

Art historical rhetoric organizes itself as a narrative in a kind of feedback loop with irregular substitutes (like Hayden White has eloquently demonstrated for history and its documents). After a poetic act of pre-structuring, the historian builds up lexical, grammatical, and syntactical protocols by which the field of research could be described in the historian's language and not in the language of the documents or the images. Presentation and explanation, etc. are poetical efforts which generate the object of research as well as setting parameters in language. One could say that art history is word processing plus an appendix of images. The materialisation of knowledge acquired through images is text. And in reverse, text conveys the conceptualization of certain images as carriers of knowledge. It is assumed/imputed that knowledge is contained within the image. (Lacan) The art historical economy of knowledge is completely based on text, however, but it requires the image as a kind of gold standard.

I think it was Aby Warburg who raised the question if it might be possible to save the knowledge about images in an image format and not in a text format. Could images organize themselves as images and not through text? Is there a discourse between images? Aristotle stated that knowledge could only appear in the medium of language, because the basic element of every knowledge is negation. The opposition of affirmation and negation in spoken language is the fundamental opposition that anticipates all other oppositions. In his Metaphysics Aristotle writes: "Everything reason can think is expressed by affirmation or negation". So contradictory statements are as impossible as contradictory knowledge. The proposition of contradiction ("Satz vom Widerspruch") is the foundation upon which the entire structure of aristotelian logic rests.

But the image has no negation, and therefore it has no opposition of affirmation and negation. It can't assert or state anything and therefore it is no medium or form of knowledge. An image is able to state that there is something special, but it is not able to state that there isn't something special. So a 'statement' of an image is always affirmative. More exactly: affirmation can only be defined as an antithesis related to negation, and that means that an image is neither affirmative nor negative but always contradictory, and - according to Aristotle - not suited to be a medium of knowledge.

The point I would like to make is, that under the conditions of digital computers, images are strings of symbolic data that can be processed by the basic logical or boolean operators AND, OR and NOT. The computer-image is a text - not just in any metaphorical sense but as a linear series of discrete symbols on a hard disk. And this is the crucial point where two ways of image processing split: the 'narrative' or 'human' way of image processing by writing art historians and the 'discrete' or 'machinable' way of image processing by computers.

This corresponds to the question about digital or digitized art history. The conservative position is interested especially in the discourse-elements and techniques that are not able to be implemented on computers, the liberal position is interested in those that are adaptable to computers to make the traditional discourse more efficient, and the anarchistic position is interested in those that are only able to be implemented on digital computers. The interesting question is: what happens if images and texts become objects of scanners and algorithms and not of the eyes of beholders and readers. What happens if not meaning is to be processed but brute quantities of resolution, chrominance, luminance or alphanumerical strings.

Digital art history can perhaps be described by the use of certain machines which generate an amount of knowledge that art history never asked for. Perhaps art history has no question to which this knowledge might be the answer. And the reason why there is no question might be that the image never before had any knowledge but was just a distant reference to a purely textual production of knowledge.

Fractal image compression and the order of images

As a small example I like to focus your interest on the 'transhuman' labor of perception done by methods of fractal image compression as it is used i.e. for the storage of satellite data.

'Fractal' means a structure that consists of similar forms linked in a recursive way. Imagine a photograph of a series of potato chips (Abb. 1): The similarity of the chips could be described in a finite set of instructions and data. By using some image-areas of the first chip, by rotating, resizing, and stretching them, it is possible to generate the images of other chips. Fractal algorithms are suited best to describe objects that are irreducible to simple euclidian forms, like well-known coast lines, mountains, clouds and skins - exactly those things that satellites love to look at.

Fractal image compression methods reduce the pixel-flurry of the real to iterable (?) basic elements from which the image could be reconstructed independent of scale. The difference to other ways of compression is that neither each pixel is spelled out (LZW) nor are some pixels forgotten (JPG), but nearly all pixels are forgotten. Fractal image compression derives instructions from raster images, deletes the pixels, and is able to construct raster images from those instructions that look like the images deleted. It's a 'mimesis without reason' to use Horst Bredekamps words. In this new similarity there is no such remembrance like in a scanner or a photographic camera. *There is no recognition. The image appearing on the screen is a pure product of visualized self-reference.

Let me tell you some more details. Rotating, resizing, and stretching of an object is done by a so-called 'affine transformation'. The parameters for this operation are written down in form of a matrix like this (Abb. 2):

Abb. 3 shows the transformation W applied to form F. The Transformation ("W of F" or W(F)) is smaller than F. If the transformation is applied repeatedly like W(W(F)), W(W(W(F))) etc., then W converges in a certain point - the so-called attractor of W or AW (Abb. 4).

An image compressed in this way contains a minimal area plus a transformation matrix that contains all required information to reconstruct something similar to the original image by rotating, resizing and stretching this area. Therefore the task of fractal encoding algorithms is to find redundant areas of an image and to reduce these areas to the necessary information about attractor and transformation matrix. To perform this task, the image is divided into 'domains' in coarse or fine resolution depending on the redundancy of the image (Abb. 5). Then the routine searches the image for 'ranges' that have a similarity to the domain that can be described by an affine matrix. Abb. 6 shows a range block R' for domain D, Abb. 7 shows the operations necessary to transform R' into D. After this is done for the complete image, the resulting file consists of data concerning just domains, chrominance- and luminance-distributions and transformation-matrices. With this data it is possible to construct images that all look the same, independent of the original resolution. The image has become a self-referential, independent system.

What does this mean to digital art history? I think it would be interesting to use fractal image compression not only for compressing images but for comparing images. The functionality of art historical archives and image banks are ordinarily not based on images but on the description of images. Instructed by texts, art historians look at certain differences in images and write signifiers ("Signifikanten") on filing cards (whether they be analog or digital) - things like name of the artist, date, iconography etc. What is organized afterwards are not the images but merely the signifiers - and their sortability (*their "stupidity" as Lacan would have said) is their great advantage. As I mentioned before, I think that Aby Warburg's idea was to overcome the stupidity of the signifiers by allowing them to escape from the archive and letting that which is signified interact. I think that this concept is only realizable when computers process images as data.

There is no time left to speculate about attractors in multidimensional space, but there is time for a few words about the importance of image processing to the archive of art history. I am not talking about aesthetics based on information science as did Abraham Moles or Max Bense. I am not talking about expectable results like the fact that Malevich's Black Square has high redundancy with a minimal set of transformation instructions in comparison to a gothic cathedral, which has high redundancy along with a maximum set of transformation instructions. But I think that several methods or models of structuring the historical field are 'fractal'. Remember not only Morelli's ears and fingernails but also terms like "Kunstwollen" or "style" or "age", which are all based on a kind of recursive similarity. It's information and redundancy that stabilize those concepts, and exactly these terms become calculable. What might be said about venetian colore if we generate transformation instructions for color space instead of geometrical space? Which images look similar to a computer program that doesn't compare historical dates or 'Gestalten' but transformation matrices or domain blocks? I think such a program would bring together things no art historian would have brought together. Fractal image compression algorithms look at differences and repetitions but they don't imitate human recognition. They don't use the 'Gestalt'-concept and they are completely ahistorical. I think they are completely apart from traditional art historical concepts because they are strictly computational in the sense of algorithmic recursiveness.


I don't want to question all the helpful applications of computers for art history. I would only like to be suprised and not bored by new media. If we compare the recursiveness of images by a computer, it might find similarities which are imperceptible to the human eyes and too complex for human brains. This other way of 'seeing' might be a critique of the anthropocentric theory of the human beholder as the privileged producer of knowledge. When machines like scanners and computers process images, they deal with perceptions below or above our borders of perception ("Wahrnehmungsschwellen"), and they deal with a complexity beyond our capability for complexity. This media-conscious process of seeing and letting see belongs to everyday life in the natural sciences: researchers in this field let machines look for them and they use well-calculated constructions to see what can't be seen with human eyes or is to complex for human brains. But humanities since the age of Dilthey have restricted themselves to what could be 'experienced' ("erlebbar"). And experience implies that something is big enough, is slow enough and is less complex enough to be recognized by our own built-in and limited senses. And this restriction was perhaps nothing more than a reaction to media technology an experimental psychology in 19th century that told our senses where their limits are. So I would be in favor of a use of media technology that gives it a chance to awaken us from an anthropocentric sleep, and to surprise us in a way that is totally intelligible and totally incommensurable at the same time.

Dr. Claus Pias
Bauhaus University Weimar
Faculty of Media
Bauhausstr. 11
99423 Weimar




Title (of no more than 20 words)


Überlegungen zum "Wissen" von Bildern


considerations on the "knowledge" of images

The aims and objectives of the proposed paper (between 3-5 lines)

Ziel des Vortrags ist es, zu zeigen, daß der Gründungsakt der Kunstgeschichte besonderen Form der Blindheit beruht. Wenn Kunsthistoriker schreiben, ist das Bild abwesend. Das Wissen um das Bild ist kein Wissen des Bildes, sondern ein Wissen der Texte, die das Bild als kunsthistorischen Gegenstand konstituieren. Digitale Medien machen diese Verschiebung offensichtlich.

The aim of the proposed paper is to show that the founding of art history as a discipline is based on a certain blindness. When art historians write, the image disappears. The knowledge about the image is not the knowledge of the image, but rather the knwoledge of the text which constitutes the image as an object of art history. Digital media make this displacement obvious.

A summary description of the primary material to be discussed in the proposed paper

Der Vortrag wird zeigen, daß die Teilung von Geistes- und Naturwissenschaften eine Reaktion auf die technischen Medien und die Experimentalpsychologie des 19. Jhd. ist. Seit Dilthey betrifft Hermeneutik das, was langsam und groß genug für menschliche Sinne ist. Zweitens wird der Vortrag mit Foucaults Archäologie zeigen, welche Funktion Bilder im kunsthistorischen Diskurs haben. Drittens werde ich fraktale Bildkompressionsverfahren als kunsthistorische Werkzeuge vorstellen, die die tradierten Ordnungsbegriffe von Ikonographie, Gestalt und Geschichte revidieren. Viertens werde ich (mit Bezug auf Aby Warburg) zeigen, daß diese digitalen Verfahren ein Wissen aus den Bildern selbst produzieren, zu dem die Kunstgeschichte keine Frage formulieren kann.

In a first step the paper tries to show that the separation of "exact" and "human" sciences is a result of the development of technical media and experimental psychology in the nineteenth century. According to Dilthey, hermeneutics are concerned with what is big and slow enough for human senses. In a second step (and referring to Foucault's concept of "archaeology") I will show which function images possess in art historical discourse. The third (and main) part will present fractal image-compression algorithms as art historic tools. Fractal image compression makes the traditional criteria of classification (iconography, gestalt, history) seem obsolete. In a conclusion I will stress Aby Warburg to show that some digital procedures are able to produce a knowledge out of the images themselves to which art history is not able to formulate a question.

The central argument or arguments of the proposed paper

Computer können an Bildern sehen, was (a) für menschliche Sinnesorgane zu komplex ist und (b) für etablierte Methodologien unsichtbar bleibt. Digitale Verfahren ermöglichen eine Kritik der grundlegenden Annahmen der Disziplin, die plötzlich wieder sichtbar werden.

Computers are able to "see" (a) what is too complex for human senses, and (b) what remains invisible for the established methodology of art history. Digital media obtain the chance for a critique of the fundamental suggestions of our discipline which suddenly become obvious.

How the proposal paper will relate to the main themes of the preferred section.

Die Tagung zeigt anscheinend auf vielfältige Weise, wie Computer die die kunsthistorische Praxis vereinfachen. Das Problem einer der Kunstgeschichte kann es aber nicht nur sein, ihre institutionalisierten Verfahren von analog auf digital umzustellen, da dies nur eine digitalisierte Kunstgeschichte bedeuten würde. Der Vortrag versucht, einen technisch-medialen Mehrwert aufzuzeigen, der über bloß quantitative Steigerung von Zugriffs- und Distributionsgeschwindigkeit und Datenvolumen hinausgeht und die Grundlagen der Disziplin befragen.

In numerous ways the section tries to show how computers make the art historian's life easier. But shifting from analog to digital while relying on the institutionalized Operations just means a 'digitalized' art history and misses the chance of a truly "digital" one. The proposed paper tries to reveal a techno-medial surplus that exceeds the quantitative increasing of data-amount, speed, and distribution, and questions the fundamentals of the discipline itself.

The official proposals which might be a bit provocative, but perhaps it fits into the sections 23.3 (Web Sites and Other Digital Sources: Research) or 23.5 (Digital Images).



Art History Webmasters ASSOCIATION des webmestres en histoire de l'art
Research and Communication Tools in Art History
Outils de recherche et de communication en histoire de l'art
Since November 14, 1997. Depuis le 14 novembre 1997.

haut de page