Datamoshing and the emergence of digital complexity from digital chaos

The myth of perfect reproducibility

Walter Benjamin (2004 [1935]) argues that a work of art loses its ‘aura’ when it is or can be mechanically reproduced. The focus of Benjamin’s famous essay is the politicisation of art/the aestheticisation of politics in the 1930s – with war being the inevitable consequence of this process (2004: 810). Cinema in particular has the power to shape experience, leading audiences away from the contemplation of art towards a state of distraction as the film keeps cutting to a new shot before the viewer has had time to think upon it (2004: 808). As a result, film viewers absorb art as opposed to being absorbed by the work of art, a position that, as the late Miriam Hansen has pointed out, is not necessarily a lament as one defined by ambiguity (1987: 187).

In a later essay, Hansen points out that Benjamin’s work continues to be relevant in the digital age, where the problems Benjamin confronted ‘persist, albeit in different configurations and at an exponentially vaster scale’ (2004: 19). She argues that the digital era, with its video games, computers, digital effects, and music clips, is more so now than in the 1930s a period of play, a concept that Benjamin introduced into his 1936 version of the ‘Mechanical reproduction’ essay, and which Hansen defines as being a ‘political ecology of the senses’ (2004: 20). That is, in the same way that Benjamin’s essay develops an argument concerning the politicisation of art/the aestheticisation of politics, and in the same way that cinema in particular can shape our senses (Benjamin’s famous example is to compare the camera to the surgeon, who enters into us), then so too, is life in the digital age defined by negotiating political artworks/aestheticised politics that change the mode of functioning of societal strata.

The term ‘digital age’ is perhaps inappropriate, since to speak of the contemporary world as an ‘age’ surely means something very different to the ‘age’ that Benjamin evokes in his essay – as WJT Mitchell points out in his consideration of ‘the work of art in the age of biocybernetic reproduction’ (2003: 489). Whereas Benjamin was writing between two world wars in which technology had very visible and sinister effects on the contemporary world, especially in the realm of armed conflict, Mitchell argues provocatively that we now live in an era in which technology is ubiquitous, wars are ongoing and invisible, and yet nothing seems to happen or change – that is to say, it is not an ‘age’ defined by societal shifts, as per Benjamin’s, but a ‘static’ period in which nothing much seems to shift at all. Furthermore, in the ‘biocybernetic age’ (in which Mitchell says that computers and biology are combined at all levels of life, from the fitness club to cloning), the original artwork may have lost its ‘aura,’ but the copy is quite probably an improvement upon it, possessing ‘even more aura than the original’ (2003: 487). Adobe Photoshop can remove photographic flaws, while genetic engineering can carry out a similar job with regard to physical ‘flaws’ on and/or in the body. If the cinema functioned like a surgeon, Mitchell points out, then now surgery is like cinema, or rather virtual reality, since surgeons can and do now work remotely and in a virtual environment that is only indirectly connected to the real but absent body undergoing the surgery (2003: 488).

Mitchell goes on to argue that certain artists can ‘compete’ with cinema in terms of what its (political and, in Hansen’s terms, ecological) effects are, or what work the art can do (2003: 492–498). And in some ways this essay will argue with Mitchell that ‘alternative’ audiovisual media production practices, namely datamoshing, can do work that competes with the effects of cinema. But it shall do so by arguing that a digital copy often does not have more aura than the original, but that in some senses it deliberately has less. In another article that evokes the title of Benjamin’s essay, astrophysicist and Leonardo editor Roger F Malina has written that software is ‘ideally suited’ for what he terms ‘post-mechanical, or generative, reproduction’ (1990: 37). In other words, when software is used in conjunction with an artwork to ‘copy’ it, something new can be generated (it functions as a ‘genotype’) – and, we shall add, in the case of the datamosh, and glitch art more generally, this paradoxically takes place through the deliberate degeneration of the original.

Without wishing too much to twist Benjamin’s words away from his original intention, then, in the context of datamoshing and glitch art it is worth remembering that computer technologies and their appropriation by artists for creative purposes do not bring about the ability perfectly to reproduce originals in a mechanical/digital fashion – even though this belief persists in some quarters. Holly Willis (2005: 7), for example, argues that:

in terms of reproduction, digital tools allow for the theoretically infinite reproduction of material with no loss of quality, affording new modes of production based on appropriation and sampling, and indeed, contributing to a culture of sharing and recycling.

Contra Willis, storing data on a hard drive does not mean that they are kept safe forever, nor are those data perfectly reproducible. In part, this can be related to technological innovation, which in turn renders ‘old’ systems of information storage – both in terms of hardware and software – obsolete, such that the data become inaccessible because ‘illegible.’ However, this is also because data simply decay. With regard to films, nowhere is this made more apparent than when Disney and Pixar decided to release Toy Story (John Lasseter, USA, 1995) on DVD in 2000. As Paolo Cherchi Usai (2001: 100) has explained:

twelve percent of the digital masters had already vanished. For three months, Pixar Animation Studios staff scoured the system for the toys’ missing parts – salvaging all but one percent of what had been lost in the computers. The remaining scenes were reassembled. For subsequent Pixar movies, Lasseter said, ‘we have a better backup system.’

Data compression

Furthermore, not only can data be simply corrupted and lost, but so too are data commonly compressed (typically for the sake of storage space). Stephen Keane (2007: 37) points out that the original data of a digital film remain the same even though they might be converted into a different and new file according to whichever output format the filmmaker wishes to use. That is to say, a filmmaker may wish to convert an .mp4 file of an audiovisual text into a .dv file, but, after Keane, it is not the ‘original’ (and invisible) data that are converted, so much as the format in which those data are outputted. Keane notwithstanding, there is a loss of data that takes place when they are compressed to fit a new form, such that – in an extreme and admittedly non-Benjaminian sense – rarely do we see the ‘original’ of a film. To create a DVD of a 90-minute film shot on HDV format, for example, requires the compression of the data that comprise the film, because an uncompressed 90-minute film might typically be 20gb in size (if not significantly more), while the average single-layer DVD can only hold 4.4gb of information.

Although the fact of compression is common knowledge and therefore to discuss it runs the risk of being simplistic, it is important that we make clear what happens during data compression, because it helps to set up a definition of datamoshing. When the data that make up a film are compressed to fit onto a DVD, it is not that the original 20gb go on to the 4.4gb disc in a perfectly miniaturised fashion, but that enormous amounts of those original 20gb are discarded. Typically this involves keeping all of the data from prominent, or key, frames (hence the term ‘keyframes’, which can also be referred to as i- or image-frames – e.g. the first frame after a cut or when there has been a large amount of change through movement between frames). However, for the frames between keyframes, commonly referred to as p-frames, only the aspects of the image that have changed (e.g. pixels whose colour value has shifted) are kept, the unchanged aspects/pixels being made simply to refer back to the same colour value in the keyframe. The result of compression is not that all of the ‘original’ information sits hidden on the DVD, but that it is disconnected from it. ¹

In this sense, and against Mitchell, a DVD is typically not an ‘improvement’ on the original film, even if that film was shot digitally. Indeed, if we can argue that the cinephile’s desire to see films theatrically is in some ways an attempt to confer on to cinema an ‘aura’ that Benjamin considered already to be missing, then the digital file stored on a DVD has undergone an even greater loss of aura, which is not to mention other, older big-to-small screen compression procedures such as pan-and-scanning. Incidentally, in an epoch when Film Studies could just as easily be called DVD Studies, the study of compressed cinema that is the viewing of feature films on DVD becomes the legitimisation of studying simulacra, since, if we allow the foregoing logic to stand, to study DVDs is to study a less auratic version of a form that Benjamin already saw as involving a loss of aura.

Be that as it may, we shall turn our attention now to those artworks that use the loss of data via compression and corruption as a means of creativity and expression. To this end, we shall discuss the practice of datamoshing and glitch art more generally, and we shall consider, after Mitchell, what work this art form can do.

Datamoshing: Monster Movie

The technique of datamoshing is said to have first been used by artists Takeshi Murata, Sven König and Paul B Davis in collaboration with American art collective Paper Rad (see Pfeiffer, 2009). The technique (which is sometimes referred to as ‘bleeding pixel effect’) involves using compression artefacts as a visual style. That is, the artists use the changing elements of the p-frames that arise when video files are compressed and they add these to i-frames from different digital moving images, with the result that the i-frame of one image, typically paused momentarily on screen, suddenly seems to dematerialise as the moving aspects of the p-frames from another moving image begin to manifest themselves on, within or from behind it. As a result, blotches or ‘bleeding pixels’ appear, which might initially be considered glitch-like errors by the viewer, but which the artists use expressively so as to create new meanings via the ‘bleeding’ of one image into the other. For example, Murata’s Monster Movie (2005) is a 4-minute video that consists of monsters intermittently emerging from a swirl of pixellated colours set to a heavy percussion track by Plate Tectonics. ² The film, which visually recalls early colour-swirling artworks such as Arabesque (1975) by John Whitney and, latterly, Malcolm Le Grice’s Digital Aberration (2004), can be considered important in several different ways.

Firstly, Monster Movie literally realises Lev Manovich’s (2001: 302) dictum that, from the perspective of the computer, a digital film is simply ‘colours changing in time’ – since only at certain moments do recognisable shapes and figures emerge from what is otherwise a swirl of colours. As Manovich goes on to argue, this emphasis on colours changing in time means that films become structured less around actors, shots and the conventional vocabulary used to define narrative cinema; instead viewers of datamoshes, Monster Movie in particular, are presented with colours from which emerge temporarily recognisable shapes that disperse as quickly as they arise.

Secondly, Miriam Hansen reminds us through a reference to Manovich that within cinema there has been a shift in the relationship between animation and what she terms ‘life-action’ (2004: 19): the digital era sees live-action cinema emerge as a subset of animation, as opposed to the other way round. Hansen then relates this shift to Benjamin’s interest in Mickey Mouse as a creature capable of making us re-question the dividing line between self, technology and world. In terms that recall Mitchell’s ‘biocybernetic’ reproduction, Hansen says:

Benjamin’s Mickey Mouse points toward the general imbrication of physiological impulses with cybernetic structures which, no longer limited to the imaginative domain of cyber-fiction, has become common in science and medicine, architecture and design, and a host of other areas. (2004: 19)

Thirdly, it is also apt that Murata chooses to use compressed images of monsters in his film, because the datamosh, not being structured so overtly around character and action, lends itself more to being understood as a spectacle of colour with little or no narrative. This in turn recalls the early cinematic mode of production and exhibition defined as monstration by André Gaudreault (1990). According to Gaudreault (inter alia), early films showed rather than told; that is, they were a form of monstration structured around simply depicting movement. The pre–1907 films that he discusses were not made up of separate shots as narrative cinema came to be through the development of montage, but instead (predominantly) featured single shots, as might best be characterised by Louis and Auguste Lumière’s first films, including L’arrivée d’un train à la Ciotat/Arrival of a Train at La Ciotat (1896). Since the datamosh is not visibly made up of separate shots, but instead of various changes in colour only gradually from which does a recognisable shape emerge, then, so too, might monstration be a suitable framework through which to understand it, not least because it shows colours punctuated by monstrous shapes, but also because using this concept of monstration allows us to consider datamoshing within a longer history of audiovisual media production and artistic practice.

Datamoshing goes mainstream?

Other examples of datamoshing include König’s Download Finished! made in collaboration with Bitnik, which is a website that, in the artists’ own words, ‘transforms and re-publishes films from p2p networks and online archives’, ³ and David O’Reilly’s Compression Reel (2008), which incorporates the artist’s own earlier animation work such as Octocat Adventure (also 2008), and which involves a vertiginous onslaught of almost incomprehensible ‘moshed’ images (including words) that lend themselves also to the embodied experience that is moshing at rock concerts. ⁴

More recently, however, datamoshing has been incorporated into commercial music videos for Kanye West and electro-pop group Chairlift. Directed by Ray Tintori, Chairlift’s music video for ‘Evident Utensil’ (2009) uses datamoshing to create a psychedelic, mind-bending trip with the members of the band, while the video for Kanye West’s ‘Welcome to Heartbreak’ (also 2009) involves a grittier take on the technique. The video, directed by Nabil Elderkin with contributions from Ghost Town Media, involves West and contributing singer Kid Cudi ‘melting’ into each other as an otherwise photorealistic image becomes pixellated/moshed before reforming as another photorealistic or nearly photorealistic image of the other singer. ⁵ Given that the song is a (banal) lament about the difficulties of being a multi-million dollar-earning rapper, the technique as used in the video does lend itself to existential interpretations: as West melts into, becomes dominated by and then re-emerges from the background colours that surround him, viewers might get a sense of the difficulty that West has in asserting his individuality in the world, because this seems literally to be what we see as West is swallowed up by and intermittently emerges from the backdrop of pixels around him.

Finally, datamoshing has also found its way into the mainstream film industry, although still through the music video form, since the technique is used in the video to Linkin Park’s song, ‘New Divide’. ⁶ ‘New Divide’ is the title song from Transformers: Revenge of the Fallen (Michael Bay, 2009), and the video features the band and the disguised robots from the film morphing into each other in a now-familiar swirl of colour. Despite the vapidity of the film, again the technique here seems appropriate in that the constant transformations that come about as a result of the datamoshing reflect the unsteady nature of the robots in the film, an unsteady nature that challenges the conventional notions of cinematic subjectivity as being fixed or stable, and which instead recall Mitchell’s Benjamin-inspired ‘biocybernetic’ age in which the dividing line between the machinic and the living, between the organic and the inorganic, is being challenged. Furthermore, to borrow a concept from Gilles Deleuze and Félix Guattari (1984), this datamosh (and perhaps the film more generally) might suggest a subjectivity that constantly is in the process of becoming, to which we shall return later, particularly as it relates to the concept of emergence. ⁷

Datamoshing, glitch art and time

First, however, we shall place datamoshing within the wider and more established practice of glitch art. In an argument reminiscent of Malina’s discussed earlier, Olga Goriunova and Alexei Shulgin (2008: 111–114) propose that glitches are ‘computing’s aesthetic core’, and that by disrupting the ways in which computer-generated art conforms to the conventions of human spatial organisation, glitches allow us to rethink computer art and, we shall add, space and time themselves. This ‘rethinking’ may not, as Goriunova and Shulgin themselves admit, be ‘one-hundred percent compatible with customary human logic, visual, sound, or behavioural conventions of organizing and acting in space’. This indeed seems to be true when we look at Monster Movie, ‘Welcome to Heartbreak’ and ‘New Divide’. Their contents are hard to describe in terms of where and when the different shots take place and how they are related to each other, that is to say, using the conventional criteria of spatiotemporal measurement – in much the same way that, after Manovich, they are hard to define using the conventional terms of shots, actors, and narrative, which in turn means that a concept from early silent cinema, monstration, seems a more appropriate theoretical framework through which to understand them.

Demanding as these works are of an ‘unconventional’ reading, we shall also propose that, in addition to the Benjaminian and film historical readings offered earlier (via Hansen and Mitchell, and Gaudreault respectively), a Deleuze-Guattarian reading is also appropriate. As per the earlier reference to becoming, the contents of datamoshes suggest the ‘schizophrenic’ subjectivity that lies at the heart of Deleuze and Guattari’s work on Capitalism and Schizphrenia (1984, 2002). They argue that through our interactions with other people and objects (or the ‘assemblages’ with them that we co-constitute), we are constantly changing/becoming. As a result, we do not have fixed subjectivities, nor are we reducible to a single ‘essence’, but instead we are defined by complexity and plurality; in effect we are all of our relations, not one or some of them more than any others. Since the ‘characters’ or ‘monsters’ in datamoshes are similarly without a fixed subjectivity, instead being able at any moment to melt into or form an assemblage with the background/another image, what Deleuze and Guattari would term a ‘schizoanalytic’ interpretation, whereby no one image or (part of the) frame is more ‘accurate’ than the others, seems most appropriate.

What the ‘schizoanalytic’ approach in turn suggests is that the time of the datamosh is not a linear time dominated by human action. If, within film studies at least, audiovisual media are traditionally understood as involving a chronological temporality, the rhythm and tempo of which is dominated by human action/movement (see, for example, Deleuze’s Cinema 1: The Movement-Image, 2005a), then it is not that human movement disappears from the datamosh, since we at times do see human figures moving (for example, we see musicians performing). Rather, it is that the human temporality comes into contact with a different temporality, that of the ‘background’ into which the humans and monsters ‘melt’ and from which they subsequently emerge. If earlier, after Mitchell and Hansen, we described this human–background relationship as an ‘ecological’ one, this is not because we are seeing humans interacting with nature when we watch a datamosh. However, this relationship is ecological in the sense that we are seeing humans interact with/assemble with, or from, pixels; that is, within the frame of the datamosh, the human and monstrous figures have an ecological relationship/form an ecology/form a network with computers (for more on such ‘media ecologies’, see Matthew Fuller, 2005). And the computer has its own temporality.

In a comparison of computer ‘time’ and human ‘time’, Steve Goodman (2008: 256–259) explains that, after Henri Bergson (1944 [1911]), humans do not experience time in discrete units, a conception that led Bergson so vehemently to attack the cinematograph for its breaking up of duration into the units that are frames, but that the computer does. Indeed, if each moment in a human’s life passes from one into the other in a chronological fashion (but in such a way that we cannot easily distinguish when one moment ends and another begins), the computer organises time in such a way that different moments can be accessed not in a uniquely linear fashion, but in a random manner, akin to a database (see Manovich, 2001: 219–221). The datamosh sees these two temporalities interpenetrate each other, not least as the human and the ‘background’ seem to do literally that. Datamoshing involves a rearrangement of the human–computer assemblage as the (digital) environment comes to the fore, engulfing and swallowing up those human (or, in Monster Movie, monstrous) figures that typically we might think of as agents. Furthermore, this rearrangement takes place as the films dart back and forth between ‘characters’ and ‘perspectives’ in a swirl of colours changing in time. As a result of this visual expression of two interpenetrating or assembling temporalities, we feel tempted to argue that the datamosh constitutes, after Gilles Deleuze (2005b), a sort of digital time image.

The emergence of digital complexity from digital chaos

Although he is not referring to Holly Willis’ work, Adrian Mackenzie reminds us of Willis’ assumption concerning the perfect reproducibility of digital data when he says that ‘technology is often framed as neutralising contingency by increasing technological mediation’ (2002: 218). Since the datamosh takes p-frames from one image and inserts them into i-frames from other images, the practice reflects on how compression leads to precisely a contingent change in the otherwise perfectly reproducible image, which the artist subsequently reworks into a novel form. If the degradation of artworks through compression (and glitches) can be equated to entropy, then this degradation can be linked to chaos – the decay that digital information undergoes over time, as mentioned earlier via Paolo Cherchi Usai in the context of the source files for Toy Story. However, if a new and artistic form is created out of these supposedly degraded images, then, after Ilya Prigogine and Isabelle Stengers (1984), datamoshing can be seen to constitute a form of digital ‘order out of chaos.’

Claire Evans’ film, Digital Decay III (2007), was created in response to artist Douglas Davis’ pronouncement that digital bits ‘can be endlessly reproduced, without degradation, always the same, always perfect’ ⁸ – a pronouncement that featured in his 1995, Walter Benjamin-inspired essay ‘The work of art in the age of digital reproduction’ (Davis, 1995). Evans’ film features this sentence shown in hundreds of repeatedly more compressed versions until it becomes illegible, suggesting that, contra Davis, entropy is as much a part of the digital lifecycle as it is a part of reality.

The second law of thermodynamics suggests that entropy, or increased levels of disorganisation or chaos, is inevitable within a closed system and that this process is irreversible (for an overview, see Gleick, 1998: 9–31). The typical example given is that two liquids are held in a container in two separate compartments. This initial condition is deemed to be a highly ordered state: the liquids are perfectly separated. When the partition that separates the liquids is removed, the liquids mix and will continue to do so until there is an even spread between them, such that they are considered to be in the most disorganised state possible (in spite of the fact that the word ‘even’ in the phrase ‘even spread’ lends itself to notions of order, and not disorder). This process is irreversible because it is highly improbable that the liquids will at any point spontaneously separate out again and reachieve an ordered state. In a similar fashion, then, it is highly unlikely that a compressed or corrupted digital image file will spontaneously reorganise itself into the ordered and recognisable image that once it was.

However, while according to physicists entropy is seemingly inevitable in closed systems (water within the container; the digital file on a computer), the biological world, in which complex organisms and, on a grander scale, complex ensembles of complex organisms emerge, would seem to disprove this (Prigogine and Stengers, 1984: xxix). After Humberto R Maturana and Francisco J Varela (1991), self-organisation or autopoiesis occurs; or, as biologist Jack Cohen and mathematician Ian Stewart argue, chaos ‘collapses’ and complexity emerges (Cohen and Stewart, 1994: 251–252).

What is it that marks this shift from chaos to complexity? Firstly, if chaos theory applies only to closed systems, then perhaps complexity theory applies more rightly to open systems. What is allowed to happen in an open system is the possibility for order to emerge out of chaos, with emergence proving a key term in this process. And yet what is it that defines a system as ‘open’ as opposed to closed? With regard to biological complexity, life is of course a key ingredient, in that life involves what we might term a ‘will to organisation’ that is not often attributed to inert matter. But to name ‘life’ as this ingredient is not enough, for this is in some respects as vague a term as ‘openness’ – and it simply transposes the question from ‘what is openness?’ to ‘what is life?’ Rather, what seems to differentiate the closed from the open system is time, the element that Prigogine and Stengers feel must be ‘rediscovered’ in order for physics to meet the challenge that biology sets before it (1984: 213–232). Time here is not simply ‘clock’ time, but it is the differing temporalities that are at the base of human perception and experience, and which lead Bergson to his concepts of duration and creative evolution. In the spirit of Prigogine and Stengers, but drawing also upon contemporary string theory, according to which the smallest particles of which the universe is made, quanta, do not so much differentiate in kind but in the frequency with which they oscillate (i.e. they differentiate according to their temporalities; see Greene, 2000), everything can be said to have its own duration, its own temporality. From this perspective, complexity, order out of chaos, or life, is what emerges from the interpenetration not of matter itself, but of the differing temporalities, the differing tempos in which that matter oscillates. The simultaneous coexistence of different temporalities, then, is what keeps a system open and which allows for emergence to take place.

If everything has its own temporality, then complexity, or the emergence of order out of chaos, which here we are also equating to life, does not apply simply to ‘living’ organisms. Mackenzie’s definition of transduction is here useful, in that the dissolution of the boundary between the living and the non-living (2002: 173–176) is necessary if we are fully to understand what it is that technology is and does. Mackenzie, like Mitchell with his theory of Benjaminian biocybernetic reproduction, sees the human/biological and the technological as inseparable, and life is what emerges from the interaction between the two, not as something brought to the assemblage between them by the biological alone. Given the foregoing discussion, particularly at the level of quanta, we see the interpenetration of temporalities that are the conditions for life not as the interpenetration of specifically biological and technological temporalities, but as the interpenetration of temporalities tout court.

With regard to datamoshing, it seems that we are confronted with a visual expression again of a certain complexity theory. Digital images may undergo entropy when left alone (i.e. when in a closed system, as per the Toy Story files), but this does not mean that they are not susceptible to interventions – here, on the part of an artist. What the artist does is to reappropriate the ‘chaos’ of the corrupted/compressed file and to turn it into an artwork, or what we shall argue here is a new ‘order’. The same seems to happen visually in the films: from the digital soup and swirling colours changing in time, patterns seem spontaneously to emerge, such that a new aesthetic meaning occurs. In the same way that the process of order out of chaos seems in many respects to defy ‘common sense’ logic, which is predicated upon stability and an absence of change, so, too, do the datamoshes visualise a non-common sense ‘logic’ of changing subjectivity, swallowing up and emergence, or what in short we shall call, after Manovich (2001) and Deleuze and Guattari (1984) combined, colours becoming in time. What to some observers is an error or a thing of ugliness, precisely because it appears to be disordered, is to others a thing of beauty, a new type of ‘order’ that challenges and allows our conceptions of ‘beauty’ to evolve.

Glitch art and datamoshing may both lie at the margins of computer science, art and filmmaking, but it is precisely on the margins or boundaries that an otherwise closed system is open to and will first of all experience the effects of change, that is, the effects of the future, of time itself. It may be here – on the margins – that the illogical predominates, but it is the mixing of the different/illogical temporality of the margin with the ‘logical’ temporality of the mainstream that enables creativity, or evolution, to take place. In this sense, datamoshing as a practice is a necessary outlier that belongs to a different temporality, perhaps what we can call the future. The entropy of digital data may from a certain point of view be inevitable and irreversible (and cause for dismay/pessimism), but from another perspective, it is an opportunity to redefine art, to create something new, and, to reintroduce a word that has in this essay been problematised via Walter Benjamin, and which we do not wish now to deproblematise, to create something original. Counter-intuitive to many though this may seem, therefore, we put forward the notion that datamoshes constitute a new type of time-image, one that will inevitably fade, but which for the time being remains an image of the future in our present. They are, as argued at the outset of this essay, an art form that deliberately embraces a loss of aura in order, paradoxically, to have, after Mitchell, ‘even more’ aura.