Notes on: Schrader, A. (2012) Haunted Measurements: Demonic Work and Time in Experimentation.  Differences. A Journal of Feminist Cultural Studies 23(3) 119--60 Dave Harris [Very long --it took me 5 sessions over 5 days. There are deep assumptions about time buried in scientific culture . Gripping account of the controversies over Maxwell's Demon in thermodynamics and the various issues raised for measurement and observer effects, a clear prelude and parallel to Böhr --and still going in the 1990s. Longer discussion of 'quantum erasure' than you find in Barad, which still structures the argument -- but more difficult to grasp] The subjective experience of time is also important for scientists. They are also well aware of the tendency of human labour to disappear in the final product — like a research proposal. Feminist science studies has questioned a number of binaries and dualisms, but paid little attention to time. Rethinking time is crucial, though, to changing 'oppositional hierarchies' [as she argued in the Pfiesteria piece]. Scientific representations omit human subjectivity and history, and so they imply 'an ahistorical nature' (120). Latour has apparently argued this, and how reducing nature to a series of objects leaves science as a kind of miracle, but again without human labour. So objective notions of time are 'woven into the very fabric of the theories'. It is supported by the notion of scientific progress. However, it is not entirely successful [hence the view that we have never been modern] and we are aware of 'natural – cultural hybrids' (121, so much so that the idea of progress must be changed — it now heads towards complexity rather than modernist simplicity. The quantum mechanics notion of entanglement raises the whole issue of observers and their relation to the universe, but this issue has a long history, for example in thermodynamics. Stengers and Prigogine have argued that some direction for time has to be presupposed for any experiment, but what if the observer is no longer external to this system — we enter the problems of '"metaphysics of representationalism"' for Barad, and '" metaphysics of presence"' for Derrida. For him, time is a series of moving presences, but only the present itself seems real or actual, and time can be understood in terms of becoming. It is sometimes thought of as irreversible [in classical mechanics as we shall see]. It leaves us with a paradox that the transition from present to past is 'both ontologically real and merely an appearance' (122). Sorting out the contributions of objective event and subjective intervention is the problem. Physicists have long discussed the problems in terms of the activity of Demons, often some playful superhuman being who can see everything that's happening. For LaPlace, an all seeing Demon would confirm Newtonian mechanics — a 'spectator theory of knowledge'. Classical mechanics assumes that if time could be reversed, the laws would still hold as we moved back to initial conditions. Quantum mechanics questions these assumptions, initially through examining the influence of observers, then through the influence of measurement itself which provokes 'an irreversible transition from… preparations… to permanent marks', (123) or from potential to actual. Thermodynamics is 'in between', still affected by both classical and quantum mechanics. It was founded on a paradox — that heat '(a microscopic notion related to the kinetic energies of molecules)' can turn into work, a macroscopic type of energy. In particular the Second Law 'explicitly prescribes an "arrow of time". Maxwell's Demon was constructed to imagine what would happen  if humans could acquire knowledge about the microscopic world without leaving a distorting macroscopic trace. Schrader wants to transform Demons into ghosts, which can 'reconfigure the very being of time'. There is a "virtual space of spectrality" for Derrida, and the notion of inheritance does away with a given past and replaces it with a task [to honour inheritance?] . However, 'we cannot just choose our ghosts, that "to be haunted is to be tied to historical and social effects" (Gordon 190)' (124) [that's encouraging!]. The discussion of experiments, including thought experiments in physics, shows the potential of natural sciences to make a difference [in the conception of time], as Kirby saw with her 'Derridean provocation'. It is central to feminist approaches. We need to deconstruct the arrow of time and arrive instead at a "ghostly" conception. If we can do this we can radically challenge dominant terms and concepts, not just modify them [a feminist project really?]. Rejecting a concept of anticipation based on irreversibility, also leaves open 'spacetime for ethical concerns', in the form of new material-discursive practice. Maxwell's concept was extended by Szilard, who connected the debate with information theory. Feminists see this as critical in reading living organisms as a matter of genetic code, and in developing the possibility of artificial lifeforms [both bad]. But we can reread it as an example of how 'a different history is always possible, at any time, here and now' (125). Maxwell's Demon held open the possibility for subjective inputs. The problem was how human knowledge affected the efficiency of heat engines. Maxwell wanted to deny that the Second Law was universal. It assumes a clear irreversible process where heat flows from warmer to cooler systems. Earlier means hotter. More technically, 'entropy always increases in isolated systems'. The term entropy itself has an interesting history, coined [for practical reasons] in 1865, to grasp the '"transformation–content" of work. The technical problem was how to improve the efficiency of the engines, and the finding was that any transformation of temperature into work was always accompanied by losses, some energy was unavailable for useful work [produced by a flow of heat] — entropy. Early formulations applied particularly to cyclic processes, where, for example, heat was absorbed from a reservoir [I thought of early steam engines where the reservoir of cool water used to condense the steam gradually gets hotter itself, but this is another implication]. Heat can never be completely transformed into work, so the production of work implies a [potentially irreversibly exhaustible] reservoir of resources, although the cycle of work itself can be reversed. A later physicist argued that this indicated a cosmically valid law. Nature was seen as a reservoir of energy, but one that could be exhausted until "heat death". No reservoir would eventually mean that no work could be extracted. This was generalised to become a universal 'tendency towards homogeneity and death'. However, there was a problem — was the heat lost in entropy just lost to human beings, wasted, or actually annihilated? The problem shows the clear links between notions of time and 'what was considered useful' (127). Maxwell introduced probability theory into thermodynamics, increasing a possible role for subjective experience. He also saw heat and work as 'fundamentally distinct forms of energy'. The development of the kinetic theory saw heat as a matter of interaction of molecules, proportional to kinetic energy, something 'radically different from our tangible macroscopic world'. Those processes [were still understood in classical terms but?] could only be grasped through probabilities, never directly observed, so human knowledge became an important factor, in effect underpinning the Second Law — we could see the increasing entropy while claiming it to be an objective property. However, this leaves 'a "degree of ignorance"'. Nevertheless, [probabilistic] agreement between scientists could produce a kind of objectivity, at least in the sense that they did '"not depend on anybody's personality"'. However, the irreversibility of entropy was now dependent on human knowledge, and it was still paradoxical — the microscopic processes should still be fully reversible, in the Newtonian sense, but in our experience, heat was lost. Maxwell began by thinking what might happen if we could manipulate individual molecules as if they were macroscopic objects. We could then possibly stop entropy. Maxwell had to conceive of a Demon to do this, given the limits of human beings — in other words he assumed [in 1871] unlimited knowledge to explain the paradox. The Demons could open and shut valves [frictionless ones], intervening in physical processes on the basis of information. The idea would be to determine the speed of the gas molecules and then sort them [not average them] into two compartments, allowing the faster ones to pass through. This would produce order out of disorder, and reverse the arrow of time. The assumption was no work would be involved. The Second Law would no longer be a law but a probability statement. There are wider implications though for the very boundary between microscopic nature and macroscopic human associated processes, including irreversible ones. The Demon was investigated much further. Some actually attempted to see the molecules using light or magnetism. Problems emerged, such as the apparently crucial role of the intelligence of the Demon. A lot of literature was produced. The Demon is now seen as ambiguous — horribly human in its desire to overcome limits of knowledge, but also material, a factor in producing knowledge [by actually changing things]. There is also a problem of Brownian motion — the Demon would be subject to it and would thus itself heat up and eventually be unable to operate — a limit for all '"automatic devices" producing work. However if human beings could be '"continuously and exactly informed of the existing state of nature"' (131), they would not have to expend work because there would be no need to actually direct molecules but just use their senses — although again this might involve a dissipation of energy. The problem was postponed [in 1914] on the grounds that we did not know how cognitive processes actually worked. Enter Szilard, who focused on what humans did when they measured things, and whether intervention could be done without generating additional work. Thermal fluctuations might be able to be exploited in principle to avoid loss of energy. However the issue is whether this is possible continuously. He proposed an apparatus, a machine which could lift weight without using a reservoir of heat, at least in the long run. Human intervention would consist not of manipulation of thermal fluctuations but rather the ability to measure them. However, measurements themselves produce entropy especially if it necessitates memory [which will require work]. In more detail, Szilard proposed a 'one molecule heat engine' (133). A gas made of one molecule is confined in a cylinder surrounded by a constant temperature [so an ideal condition?] heat bath. If a partition is placed in the cylinder which confines the molecule to one side, we will get gas pressure on the partition which might be used to do work. The Demon would decide which half of the cylinder contains the molecule and would then record the result. If we replace the partition with a piston and use our recorded results to couple it to a weight, we will be able to get the piston to lift the weight [because we will know which way the piston will move as a result of gas pressure]. The molecule has transferred its heat energy to the workload. As it loses energy it replaces it from the surrounding bath. As the piston is pushed all the way to one end, 'the gas once again occupies the entire volume; its entropy has not changed', although the heat bath has lower [?] entropy. Nevertheless, we have extracted work from heat alone [from the flow of heat surely?] The Demon will continue to measure and to guide useful work so that we have 'a perfect perpetual motion machine', which continually extracts work from heat [I still don't really get it — Schrader goes on to say this happens without a temperature gradient, but surely there is one between the bath and the cylinder? And overall, the system has surely lost heat?] [There is a diagram and further explanation on page 134]. There is still a problem because all the Demon does is to measure, leaving who actually inserts the piston as unclear — his measurements are crucial in order to correctly connect the weight and the piston, and the engine itself provides 'the memory of the binary decision process'. So the human experimenter seems to have been eliminated. However, Szilard agrees that the steps involved 'belong to one and the same act of measurement', including the human bits where cylinders are separated and partition is inserted. Translations have rendered this as a sequence rather than the same act of measurement happening at the same time. For Szilard, however, any measurement shows a '"memory faculty"' — measurement involves 'coupling [of values provided by instruments with positions of objects] accompanied by memory'. In this case, the movement of the piston embodies the memory of the location of the molecule, and Szilard assumes that the apparatus is then decoupled [and the memory lost?]. There is still a problem deciding when measurement is actually completed, and decoupling takes place, so that the values measured enter memory[?] If we are using the movement of the piston to measure the location of the molecule, this means we can do the measurement only when the piston starts to move [in her philosophical way, this means 'there is no time before the piston moves' (136)]. It is no longer clear when the piston is actually coupled and so it is equally unclear what the period of measurement is [I think she's saying]. 'The very notion of memory assumes the past of the coupling, but that which was coupled only exists [only reveals a complete effect?] upon decoupling' (137). For Szilard, we complete the measurement when we can't draw any conclusions from the initial values we observe [maybe — two parameters are decoupled is how it is put], and where there is only a memory. But this memory can both precede and follow a coupling if the engine is run continuously, so humans [who store the memory] must be intervening constantly. Further, Szilard does not see any difference between the first coupling and subsequent ones [but there probably is one, because the subsequent ones are based on memory?] In general, humans do contribute to measurement, by setting up the apparatus — in this case inserting the partition. This is not a preparation for a measurement, but an intervention that actually establishes what is to be measured. The Demon is situated within [part of] the system. There is a flow of time involved in design, with no precise notion of how it all started, or when exactly the present observation becomes a memory of the past.A further development had an ironic result. If the Demon can see the molecules, this is 'information gathering with a light signal' and must lead to entropy increase. Generalising, any information acquisition must be. It fits with Szilard by identifying his memory with information. One implication was that information could be '"divorced from human intelligence"'. Later still, another information theorist suggested that there might be computational processes that would not produce entropy. Szilard got it wrong because there was a true entropy cost not so much in measurement, but in the erasure of memory. 'Finite information-processing Demons had to regularly clear their memory registers, it was argued, and that takes work' (138). Controversy still persisted as to whether entropy is produced during information acquisition or during this process of memory erasure necessary in any cyclic process. We need to remember that measurement was always accompanied by memory for Szilard. 'In plain language', taking a piston and inserting it into a cylinder does not take work, but removing it and then reinserting it does — the latter has a history [which has to be taken into account? Overcome?]. As information theory developed, Szilard's measurement became the detection of the value, and the preparation of an experiment was no longer considered to be measurement. Information retrieval and storage were also seen as independent and separate. This was appropriate for information theory, but the meaning of Szilard's experiment was changed and the central problem displaced into trying to establish exactly which bit of the measurement process is irreversible. Incidentally, Szilard 'quantitatively specified the minimum amount of entropy the demonic measurement work would produce, which became the measure now known as "bits" in computer science' (139). [Apparently, this meant that] where entropy is actually produced in accordance with the Second Law 'remains ambiguous'. The earlier accounts assumed that irreversibility arose only once interactions left a trace in memory. This became a problem of drawing a boundary between human interventions and the measurement device that was already in the system. In effect, the Demon is not part of the thermodynamic system, nor properly external to it. Measurement can still record or create, or do both. We are left with undecidability — what is in the system and what is outside, what counts as a reading and writing of memory, and what a memory trace actually is. Developing this as an information theory problem diverts from the problem of human intervention. The conventional arrow of time works still if coupling happens before measurement and memory only after it. The externality of an observer is also necessary. The reversible system cannot be closed to the macroscopic world, because external influences have to penetrate it including 'an additional measurement on the Demon's memory' (140). As soon as the Demon relates to the external observer, we find all the usual problems about subjectivity and objectivity, matter and mind and so on. The irreversibility of the Second law 'becomes a consequence of the fact that the Observer does not measure but merely observes'. [Very puzzling stuff]. We can now come to a similar measurement problem in quantum mechanics. Indeed the early quantum theorist Von Neumann began with Szilard's Demon, hoping to develop a quantum version of entropy where the quantum state will also be irreversible in a measurement. Early quantum mechanics assumed irreversibility, conventional relations between past and future, preparation and test — entanglement between them produces the actual measurement. However mathematical formalism developed which predicted probabilities for measurements but did not 'account for the measurement itself or its specific outcome' (141). For Schrader, 'measurement implies a discontinuous and irreversible "selection" of a specific value'. When applied to quantum processes, we find a reversible bit before the measurement, the irreversible measurement and therefore a strange boundary between them. The transition from reversible preparation providing a set of possibilities to irreversible marks on bodies is 'the central quantum mystery'. It is sometimes called the collapse of the wave function upon measurement [the quantum state collapses into the familiar macro one] , where probabilities become a definite value [waves produce measurable bands? Particles go through various stages and actually produce marks on recording film?]. There have been a 'large variety' of attempts to explain collapse. In general, two options exist — one sees measurement as a physical disturbance that interrupts correlation between quantum systems; the other sees human intervention as important, a matter of selection, sometimes associated with consciousness '(and by extension, to God as guarantor of determinism enabling the apparent human "choice")'. The wave particle duality is an example [summarised 141 – 2]. Matter also has a wavelike character seen by experiments with electrons rather than photons. As Barad has noted, we get an interference pattern even if electrons are sent through a double slit one at a time, implying that electrons remember the path of earlier ones. Attempts to see which slit is involved brings the disappearance of the interference pattern. Adding a which-path detector produces particle behaviour. All this can be explained in terms of Böhr and complementarity — if we measure the position of electrons one way  we define them as particles, and there can be no interference. If there is no which-path detector, the apparatus 'defines the electrons as waves' with interference. We are not talking about intrinsic wave or particle properties[ so Barad has to go beyond Böhr here]  'Rather, a measurement defines these properties that simply do not exist independently of the measurement apparatus' (142). This in turn relied upon Heisenberg's uncertainty relation — the position and momentum of a particle cannot be determined ['arbitrarily, exactly'] at the same time. A measurement with a positron, like a photon 'necessarily involves a momentum transfer that would smear out the interference pattern', so measurements disturbed the path of the particle physically. This means it is impossible to know how things really are. Later experiments tried to rule out this physical disturbance [I think this is the ingenious erasure  experiment described best in this Barad]. They used atoms instead of electrons and developed a which-path detector that would not influence the movement of atoms. A laser beam excites the electron within the atoms and this means a photon is emitted when travelling through one of the two cavities [placed in front of the slits]. This helps us trace the path of the atom without disturbing its movement. The interference pattern still disappeared, implying that it is the measuring cavity itself which is responsible for the disappearance. For Barad it shows the very 'experimental possibility of distinguishing between the paths of the atoms' (144 — lots of diagrams on the intervening pages] which destroys the interference pattern, and removing the distinguishing apparatus produces another interference pattern. This is the quantum erasure effect, where it looks like we can erase the which-path information. Further work involved adding another photodetector between the two cavities, so that the apparatus 'cannot discern from which cavity the detected photon originates' [and apparently, the detector has an equal 50% chance of receiving a photon from each path]. The hits on the detector are correlated with the marks on the screen to produce 'another interference pattern'. What we have seen here is the effects of 'measurement upon measurement upon measurement, in which the measurement apparatus becomes the object of measurement of another', with each measurement producing a new correlation between atoms and photons. For the physicists concerned [Scully, Englert and Walther], this explains the disappearance and reappearance of interference patterns, but exactly how these correlations are produced remains a question, especially whether subsequent measurements 'somehow undo the previous one', or whether it is a new interference pattern. The first explanation suggests that measurements are reversible, and the second that correlations are irreversibly extended [and added]. This still preserves the conventional arrow of time, and it is that that we will have to critique. Some physicists [Greenberger  and YaSin] have introduced the notion of '"haunted measurement"' (145). If correlations just add up irreversibly, endless regress threatens, '"how do we know that no [measurement] will come along at some future time"' to rearrange existing findings? [This must be a problem for Barads' account too?]. Those advocates insist that there must be genuine destruction of measurements — 'correlations must be physically destroyed' (146), and that measurements that leave a macroscopic trace must retain 'a "latent order"' so that the measurements can actually be undone in the future. Their quantum erasure experiment, using similar kit, shows that subsequent measures can indeed make a measurement disappear — so it is a 'macroscopic "ghost"' 'that disappears if one does not look at it'. Measurement is therefore haunted. 'Looking at' here does not mean an observer, but 'physical interaction that detects the evidence of the passage directly and destroys the coherence [pattern produced before the which-path detectors] for good'. It is again down to photons used in observation hitting objects and therefore changing them, making measurement reversible again. So we can have both 'true' measurements, which are irreversible, and haunted ones which are reversible. How can we distinguish between them? Until we do, the very idea of quantum measurement must remain '"subjective"' and also historically contexted, awaiting more sophisticated measurement in the future. Heisenberg's uncertainty principle is also upheld. But Scully et al could extend the notion of haunted measurements to include an effect by experimenters here and now, for example in manipulating the which-path detectors, even after the atoms of pass through them, meaning that there is a choice between which-way information and quantum erasure '"at any time"' (147), a '"delayed choice" mode of the quantum erasure effect'. Schrader finds this misleading and prefers to replace 'choice' by 'material "definition"'. Measurements do seem to be affected ['haunted'] by human choice. For Scully et al human measurements do not destroy quantum coherence, but create a series of correlations that produce the interference patterns, and the experimenter can choose which one to use. The actual path of the atom is not affected: it is future possibilities that are the basis of choice. Even so, the choice must remain potential [apparently an actual measurement implies that the wave function permanently or irreversibly physically collapses upon measurement, which they reject]. In more familiar terms, 'it only appears to us as if the atom took a particular path'. They go on to say that all physics is about '"as–if–realities"', and get quite phenomenological about what physics deals with. The measurement effects are 'mere "mental processes"'. Nevertheless, the correlations have some objective reality, because they can be demonstrated experimentally — but again this is only the appearance of objective reality. This helps solve the problem of the order of time, because an objective past cannot possibly be influenced by future possibilities, although this can appear to be the case within 'immaterial' 'human knowledge production'(148). Schrader points out that this involves 'the same insurmountable contradictions' as we saw earlier with Szilard — first we eliminate human activities, and then we reintroduce them as a determining factor [deep inconsistency at the heart of experimental practice]. We can make progress with this by rethinking time. The arguments so far remain with conventional definitions, [even with Böhr, she says — the past might be questioned by arguing the individual objects do not exist before measurement, but the future remains conventional] -- potential futures is a matter of human choice. Just as Böhr challenge the idea that the world consists of discrete objects, so we must challenge the assumption, with Derrida, that time is made up of successive linking of presents [but this is a daft conception in the first place]. The older conception implies a human subject autonomously deciding to choose an experimental setup before investigation, but, as Kirby argues, the human is not separate from the world it experiments on, not autonomous from Derridean writing. [For me, these two are not the same. If people are not autonomous with respect to the operation of writing, that means not that they are therefore a part of nature, but rather a part of a linguistic system — the autonomous individual is replaced by the individual in language]. We must account for 'the materiality of human contributions to measurement', via Barad and agential realism. We can combine agential realism with Szilard's notion of measurement as a coupling process with memory. This will help us solve some problems such as material traces existing before the effects, as in quantum erasure, and the 'mysterious movements of a piston that nobody caused'. First we have to agree with Szilard that the coupling of measurement apparatus an object is not a preparation, but actually part of a measurement — this resolves the issue of reversible preparation and an irreversible recording process. We have also to replace the idea of individual human choice and the point about future generations and doing the results. We need a new Demon of our own, 'to do some real work'. Agential realism replaces the awkward position of human observers either as external, or as the only thing that does knowledge production. These 'are just different sides of the same coin' (149). For Barad, there is no external observer nor an outside boundary of the experimental apparatus [which raises its own problems, because the apparatus now expands to include the whole world?]. The subject is one agent of observation emerging through specific cuts, as is the boundary between subject and object. Measurements do not discover what is already been there nor do they create something entirely new — 'they rather materially re(con)figure the world', as with Haraway again. If Szilard is right to say that the preparation of an experiment includes measuring the value, 'there is no temporal or spatial distinction between the taking place of a measurement and the detection of the value' [but wasn't this distinction between preparation and experiment criticised just above?]. This means that 'the production of human knowledge cannot be regarded as ontologically distinct from the setting up of an experiment'. If we go back to the one molecule engine and see the piston as a measurement device, embodying the memory of an encounter with the molecule, the piston is not detecting the pre-existing location of the molecule, nor can it create it retrospectively after being decoupled. Instead it actually 'contributes to the definition of the past location of the molecule' [maybe -- the 'left' location means' once the piston has moved'?]. The intra-action is between inserting a movable piston, and then tracking its movement with the molecule [which helps us] 'give meaning to the statement that the molecule was located [in a particular place]'. The molecule did not have a location before the piston was inserted because 'there were not two sides yet' [it still had a location, of course, although we were not in a position to describe it in terms of sides?]. Only when the piston is moved to the right can we say that the molecule was on the left. And we can't distinguish between left and right before the piston moves [again highly debatable]. Therefore being on the left side is not 'an intrinsic feature of the molecule that has to be somehow detected' (150) [but its location within the cylinder is an intrinsic feature of the molecule?]. Saying that it is on the left side is meaningless 'neither true nor false, and, most important, not potentially true' until the piston moves to the right. 'It is not a future possibility that awaits human determination or decision' [except humans have constructed the possibility for the piston to move]. The molecule has to leave a material trace first, so that memory gets embodied in the piston, but once this happens, 'the molecule is no longer confined to the left side of the cylinder' [once the movement of the piston is completed, that is, and if we have some absolute notion of left side rather than one relative to the piston]. As a result [really stretched conclusion], the existence of memory presupposes 'the decoupling of the molecule's current position from its "past"', a kind of 'delayed defining event (150)'. If the molecule is actually to push the piston, we have to affirm that it '"really" was on one side of the partition, and this is not just a retrospective appearance. But what does this 'really happened' mean? The point is that measurement 'is a material – discursive practice to which not only humans contribute' [the piston and molecule are contributing?] . Definitions are not just human. What we see is 'an intra-activity to which the entire experimental apparatus contributes' — pistons, cylinders, molecules, and even 'the history of thermodynamics' [trivially true?]. Definition itself is a 'material–discursive practice producing a record'[so definition has become measurement here?]. The whole material arrangement which includes the experimenter is what enacts the agential cut so that 'the correlation between measurement apparatus and object yields a definite value'. There is no absolute separation between any of the components. Together, they engender a trace. It is not therefore strictly speaking a decoupling [maybe], more a specification of a correlation, a boundary making practice. These have now finality for Barad but are 'nevertheless ("retroactively") causal' [a whole string of musts here, I suspect to preserve coherence]. Measurement practices should be seen as both causal and open-ended at the same time. Retroactive causality does not mean that the piston somehow causes the location of the molecule — we can never affirm that the molecule just is somewhere before the piston moves. The agential cut that decouples 'defines the past as memory that has never been present [before the cut?]' (151) [looking suspiciously definitional here]. It is like the Derrida trace, which, he says can never just be reappropriated and turned into a simple present. It 'must be thought as an "originary repetition"' [God knows what that means — that repetition is inherent somehow? If we don't exactly know, this is only argument by authority, of course]. We can still affirm that a molecule really was on one side, but we should not confuse causality with this Derridean link between consecutively present moments, time as an externally directed flow where only the present is real [defines no known version of human consciousness or scientific concept?] We might need to change metaphors [sic] and talk about writing and reading instead of coupling and decoupling. Reading is not something that only human subjects do. Larger experimental apparatuses can act so as to produce a cut between object and measurement agency. This 'reading defines the writing as that which precedes it' [so we have deconstructed the apparent self-sufficiency of writing in our reading? Infinite regress here too if that reading also becomes a writing for subsequent readings etc]. Reading localises writing, seeing it as 'a material trace associated with a particular apparatus'. Reading does not cause writing retrospectively. 'There is no memory before it is read' . Causality is also an effect of a reading [we discover it, or construct it by reading?] [Somehow, this is not just a linguistic operation, but] 'an expression of the materiality of traces' and we must account for them. 'Causality itself is not caused' [an old paradox here]. We define it in a delayed way, after reading has enacted a boundary 'and a temporal order'. It is inevitably 'internally haunted' because the past we specify is also defined 'in reference to an open-ended future' [because we recognise the agential nature of reading and its local effects?]. To make things worse, readings do not just precede or follow writings, and every reading is a rewriting [writing goes all the way down — at least until it stops with Derrida's philosophy]. This does not erase memories of past intra- actions but produces it [the intra-action] as something past, 'locating and dislocating it at the same time'. A new record does not just add to earlier ones because earlier ones are no longer 'equally' [weasel] accessible. The world does not become increasingly tangled but rather 'differently entangled' (152), and different entanglements have different specifics. Reading and writing requires work, it is 'material practice'. There is no external drive to greater complexity [so what explains scientific choice or intellectual development?]. Whenever we measure something we are specifying the writing that determines the meaning of the correlation and how it becomes (de)coupled [with the usual caution, 'there is no coupling before a decoupling']. Everything will depend on how we think of an event, as a moment in time, or as something 'always already spectral'. Latour is right to say that the past can be changed in the here and now, but Schrader wants to go further and say 'the [conventionally objective and independent] past has never been present nor will it ever be present in terms of a future present'.  All the terms have to be produced by work, and '"all work produces spectrality"' say Derrida and Steigler [because this is still linguistic work though?]. Scully et al argue that all measurements are potentially given by an experimental setup and that the human experimenter chooses the relevant one. This is a matter of choice, and so measurements are reversible. Greenberger and YaSin assume that the outcome of a measurement is still determined by physical disturbance, giving a deterministic notion of reading — but that becomes a problem when new knowledge retrospectively changes the potentials. Both groups see the 'immateriality of human knowledge'. Both assume ' the givenness of the past, that is, its presence' [is she deliberately trying to confuse us with homonyms and puns?]. We must abandon all these assumptions. Measurements are haunted, so, to use the old language, 'Demons contribute to work'. There is no potential future that follows automatically from the past. Instead, measurements are 'internally haunted such that work creates time' [agential cuts create boundaries and temporal sequences?]. Measurements start with memories, material traces, and they offer 'material re-writings/readings'. Objects and measurement devices 'do not exist independently before a measurement', nor are they objectively correlated 'before the action begins'. All interactions leave traces 'But the trace by itself is not' [has no independent existence]. The same goes for memories which have to be read in order to exist. This all requires work, which in turn means 'material determinations'. It is only when agential cuts localise that we have a notion of 'proper space and time' for memories. We can reconfigure the past, 'but that does not render it less real [real to us that is] ' [but the reality has to have this demonic quality? Or is she saying that once we have made an agential cut, the event we have created is real, even when it passes into the past, and subject to rereading?]. Any rereading 'relies on the materiality of traces' [only in empirical science? You could extend the notion of 'material' of course] The old arrow of time idea assumes a closed system, disturbed only by something external, producing reversible effects. This notion even extends to quantum measurements, although here, the assumption is that these are irreversible. In both cases, a bounded system is assumed [she says there is no substantive difference between relatively open and relatively closed systems]. A closed system is often assumed in defining what is 'natural' — something that operates by itself, often heading towards equilibrium. But 'there are no systems by themselves' (153). There can be no automatic entropy either [lots of implications here for the heat death of the universe?] — 'entropy is a feature not of a system but of an experiment' [quoting somebody called Jaynes] and experiments themselves are 'open-ended material discursive practices'. We still need some notion of the Demon even though we can actually see the molecules now. However, the molecules do not occupy a preestablished system. There are indeed limits to the ability of observers, who can never move outside finite part of the universe. Nor can we 'simultaneously be a part of nature and have an external view of nature's activity', because we actually enact system boundaries that we become part of [so further twist to Kirby's assumption that there is nothing outside nature?]. Everything depends on how boundaries are constructed. The idea of a working Demon will 'incorporate the spectre of the "past" and memories of the future'. Measurement is work, intra-activity, omnipresent, and irreversible in the sense that it is irreducible. It doesn't just accumulate in time 'rather, it constitutes time'. Processes are irreversible because material traces and boundaries are irreducible [presumably meaning not outside work? We are close to assuming an arrow of time with work here though? When we revisit the past, we don't go back to a less complex reading?]. Transformative work is just productive work, intra-activity. It is not just something that is opposed by wasteful heat [which was originally defined as transformation work and therefore as something that doesn't really count as work] Haunted measurements show that the work of scientists is not just a matter of operating on components that are already there — it is always transformative not just additive. It follows that the history of science is no longer purely productive. Those judgements also depend on what we mean by time [because there is always an assumption of productive work taking place without wasting time?]. Nevertheless, every intra-action produces 'possibilities of changing its "direction"' [really idealistic about the potential of science here? And the direction of science is always towards complexity? Ignores the social constraints of maintaining research programmes etc --acknowledged in the Pfisteria piece] ]. So we don't invalidate the Second Law, and its meaning remains contested, despite all the effort to exercise Demons. [As usual, the notes are quite gripping setting out the relation with Latour, for example or describing the various attempts to pin down Demons — note 6 tells us that the Demon is nearly always masculine so his gender is relevant in understanding his activities! Some of the puzzling remarks about memory erasure requiring work are explained in note 15 — memory erasure is always logically irreversible. Note 19 says that she herself has read Böhr in a particular way, influenced by Scully et al — they have added the bit about the potential future existence of objects.]