The Natural Poesis of Remembering and Forgetting
Why do we remember the past but don’t remember the future?
What’s the difference between memory and physical continuity?
What’s the difference between cause and logical inference?
If the present state of something is all that has survived of it from its past, how can we say that its past state was different from its present state? (If some aspect of it has survived, it is not in the past but the present; if it has not survived, then it does not exist and we cannot know it, can we?)
What is the function of forgetting?
Can the same cause have two different effects? If so, how do the branched or bifurcated timelines that result relate to each other?
How far down the evolutionary scale does the capacity for memory go? How far down does forgetting?
These questions answer each other in a rather peculiar fashion, requiring an upending of our familiar way of making definitions, an exchange of figure for ground. In doing so we will be following the example of our founder J. T. Fraser, whose profound insight it was that objects don’t exist in time, but time exists in objects. If we exchange figure and ground we will find that we cannot define remembering in terms of the past and future, but must define past and future in terms of remembering–the past is what can only be remembered, the future is what can only be made. We must see memory not as a small subset of what physically survives, but rather as the choice of what physically survives. We must see cause as the opposite of logical inference–a multitude of premises gives rise to a single logical inference, while a cause gives rise to multiple effect-worlds, which must be either reconciled with one another or rejected and forgotten. We must see the main function of memory as forgetting. We must extend the concepts of both memory and forgetting to be characteristics (weaker as we descend our evolutionary tree into the past, but never entirely absent) of the whole of the physical universe.
Time, goes the old joke, is nature’s way of making sure that everything doesn’t happen at once. Like many jokes, this one has a fairly large grain of truth. If two states of the same object are allowed by the universe, such as being red and being green, or being in one place and being in another–or if two objects, such as two fundamental particles of the same mass, spin, and charge, can occupy exactly the same state and place–then major problems arise. Either, in the first case, the principle of identity is violated–is it one object or two?–or, in the second case, there is not room in space for both objects at once (in physics, this problem is known as the Pauli exclusion principle). In a universe of pure space, without time, either only one distinct state and place could be allowed for each object (which is empirically false) or the laws of science could not exist because identity and location could not be reliably established.
Time gives the universe a way of connecting different states of the same object (first it was red, then it turned green; first it was in one place, then it was in another) and of spacing out events and objects so that they do not get in each other’s way (first one atom was there, then another). Further, if one of the states or locations of an object has to exist if the other is to exist, time provides an order of events. The tree cannot exist unless its seedling has earlier existed, nor the seedling without the seed; whereas the seedling could exist without the tree but not without the seed. An object with velocity has to have been somewhere else before it was there. Time is whatever space a logical scheduling problem requires for its solution.
We can actually study situations where time almost doesn’t exist. In the tiny and always minutely brief world of quantum mechanics there is so little time that identity and location do indeed lose a good deal of their clarity and indeed their distinction from one another: a particle can exist in a state of superposition, in which two different things are true of the same object, and it can exist very tenuously in two places at once. A clump of a few dozen beryllium atoms in a very cold Bose-Einstein condensate has been made to do just that. But for objects with more solidity and persistence, time is necessary not just tautologically for them to exist “in” but also as the only noncontradictory way of resolving paradoxes of being and location.
The Enlightenment description of time, which is familiar to us all–and which even after ninety years has not been replaced in our intuitive imagination by relativity theory, let alone other modifications of it–sounds like the simplest way of providing the “spacing-out” and scheduling function that is so important to the universe. What it says is that time is very like a spatial dimension–say, length–extending out in a straight line, and that everything in the universe at any given moment is at the same point on that line; what is on one side of us is the past, what is on the other is the future, and where we all are is the present.
But wait: all is not well here. When it comes to the “passing” of time, the familiar model begins to get complicated. In the received account the present moment moves along the line in a futureward direction providing each point in it with an infinitesimal moment of reality–or, in the opinion of some physicists and philosophers, the whole line is always already real, and our consciousness moves along the line, like a spotlight, giving us the illusion of encountering a new future and leaving behind a dead past. But the space analogy has already begun to break down. Our experience of space doesn’t necessarily include a “passing of space”; there’s no necessary point of maximal existence along a line, or maximal human attention to it, and there’s no place on a spatial line that all of the universe is at. And if either reality or our awareness moves along the timeline, in what time is it moving? How fast? How many minutes per what? Can it accelerate? How could we tell the difference between slow and fast?–and if we can’t, how could we tell if it had or hadn’t stopped altogether? What does “move” mean if there’s no discernible distinction? Is there a second time in which reality or awareness moves along the line of the first time? Then why not a third time in which the reality of the reality or the awareness of the awareness moves along the second? And so on.
If the whole timeline is already there, moreover, then the future is merely awaiting its actualization or our attention to it, and cannot be changed. So we are bound to the rails of fate and such fundamental values as morality, freedom, responsibility and creativity are illusions. This reflection might be bearable for a philosopher who preferred truth to moralistic wishful thinking, if it did not also imply that the philosopher’s own cogitations are part of the same clockwork, and the feeling that something must be logically true is too, so truth is also an illusion. And there is no way of checking whether such an automaton is correctly calibrated, so as to verify that the illusion of truth coincides with its reality.
Equally problematic is the direction of time. Space doesn’t have a preferred direction. In space one can get from London to Paris and from Paris to London, but whereas one can get from A. D. 1980 to A. D. 2004 in time, one can’t, as one could with space, get a return ticket. Nineteenth century thermodynamics showed that thermal and energetic events in the universe always went one way–toward the increase of entropy. You can burn a log but not unburn it, you can let perfume diffuse out of an open bottle but not suck it back in again, you can turn work into heat and heat into work but only if you pay a tax or interest of work energy on the exchange each time. But then the study of biological metabolism and evolution seemed to show that living systems can feed upon the flow of the increase of entropy, like paddle-wheels in a torrent. Without violating the Second Law of thermodynamics, a tree can turn ash (soil) and smoke (atmospheric CO2) and heat (sunlight) back into a log, and a rose can suck chemicals out of air and soil and make perfume. So not only can time possess at least two directions, different kinds of organized systems can take different directions.
In the twentieth century, relativity theory showed that the universe is not all at the same point in the line; a present moment is not something simply given to the universe, but rather something rather fuzzily earned by two-way communication among objects and events that puts them in synch with one another. An “in-synch” region is called an inertial frame. Some parts of the universe are in different inertial frames from others and our present knowledge of them is necessarily of their past, while other parts of the universe are over our event horizon and we can never know them. And if we cannot know them, the proposition that they exist and share our present moment is a purely metaphysical and unscientific notion–unless we improve our model of time.
More recently still, quantum theory showed that the state of knowledge that exists about a particle partly determines its nature and identity–disturbing enough, but more so if we reflect that one can only know about something after it has happened, since even the fastest messenger, light, has a finite speed. This means that knowledge must somehow retroactively affect what it is knowledge of; so the present-point on the line can be neither the spotlight of awareness (we are aware only of past events), nor the place where reality momentarily condenses (since it is busily condensing previous realities).
A new view of time is emerging from a variety of disciplines. However, this new conception is as yet fragmentary, divided among schools and disciplines often at odds with each other. I shall here treat it as a unity rather than as a collection of competing hypotheses, because I am convinced that in this case to demand the kind of elegance we expect of a scientific theory, that complex details and phenomena can be reduced to a single simple principle, would be a mistake. If time is, as I would argue, precisely the condition of complexification, we should expect any true theory of it to be a messy Rube Goldberg contraption rather than a graceful piece of abstract geometry.
We can summarize the new time-conception under six propositions:
1. Time is complex and concentric.
2. Time is generated by objects and organisms and its local properties vary accordingly.
3. Time is evolving and emergent, in a breakneck struggle to avoid paradoxes that spring up at every turn.
4. Time is branchy and thus free.
5. Time is looped and nonlinear.
6. Time’s branchiness is self-pruning.
Given these characteristics, we can now return to the questions with which we began. We cannot remember the future because we cannot forget it. That is, when we look at the past, we are looking back down the timeline we are on, in which all other possible timelines have either been pruned or incorporated–that is, they have been forgotten. But when we look at the future, we cannot know which of the immense number of unpruned branchpoints out there we are going to take. The problem is not that there is no future to remember, but that there are too many futures with no indication which one is going to win out. We can’t forget any of them, so we can’t remember any one of them in particular. Certainly we can act to bring one timeline about and eliminate and forget the others–but by that time we would have made that future into the present and would confront a new future, a new dendrification of branchpoints.
It seems likely that the evolution of brains in higher animals was driven by the huge adaptive premium on being able to predict what all the other objects in the world are going to do. Lower animals rely on simply being their own memory of what has happened to them and what might happen to them in the future. They produce cheaply a huge number of clones and near-clones, allow forgetting to take place by means of the death of maladapted individuals, and remember collectively in terms of the ones that survive. But the problem with this strategy is that it has no “save and reboot” button: a sufficiently bad experience is terminal, and brings both the organism and the memory to an end together, since they are the same thing. This method of remembering, epigenetically, is slow and imprecise. For more expensive and adaptable organisms it is much better to have a second, virtual memory, in which bad things can happen before they happen in reality, thus without damaging the organism itself; and this is largely what brains are for. Fear turns out to be the motivational system that works in tandem with the capacity to model bad scenarios and avoid them. But for it to be effective, fear, with its costly suite of neurotransmitters and physiological reflexes, must be selective; all the unlikely scenarios must be forgotten. The inability to forget them–obsessive anxiety–is a sign of damage to our neurohumor system.
But then, what should be remembered? The most obvious plan is to be able to remember all the likely futures and have contingency plans ready–that is, to make all the futures safely into the past where they have already been virtually survived. Like computer chess teams buying fast computers with huge memories, higher species were engaged in an evolutionary arms race to be able to see one more move into the future than their rivals. Some species bet on trying to remember everything–the monotremes, which include the spiny echidna and the platypus, and have huge brains for their body size, seem to have adopted this strategy. However, the monotremes are relatively stupid, because, as with Borges’ character Funes the Memorious, there is not enough room in any conceivable memory space to remember everything, and certainly not enough room to make fancy contingency plans. Much of what can be remembered can’t be changed: that’s what we call “the past”. In terms of preparedness for the future, the only past memories we need are the ones that provide trends and policies for choosing likely futures to remember. The fewer the likely futures, the more the future resembles the past, and the more the future can be usefully remembered (or, as we say, anticipated).
Let us look at this from a more philosophically exact point of view. A single cause can have multiple effects; the laws of nature (which are themselves the historical result of earlier cause-effect events) constrain those possible effects but do not reduce them to one. This is what I mean by “branchiness”. All the effects that do not completely neutralize each other take place, distributed in number and intensity according to their probability over a bell-shaped curve. The difference between the number of bytes required to describe the “cause state” and the number required to describe the later “effects state” is the measure of the increase of its entropy and its thermodynamic disorder, but also of the amount of new information that has entered the world, and thus of the real time that has been generated. That difference also describes the degrees of freedom of the initial state (if there is only one “branch”, a single possible effect, the system has no freedom: if there are many, the system has free play to that extent).
However, a cause-effect event can bring about a set of effects that are either mutually reinforcing or dissonant; the mutual reinforcement of those effects is the measure of its informational (as opposed to its thermodynamic) order. If that state of mutual reinforcement limits the number of new effects, we may call it a barren order; if it makes possible a further set of mutually-reinforcing but productive effects, we may call it a rich order. The definition of evolutionary reproductive success in biology is very similar: as many offspring as is consistent with the survival-to-reproduce of the offspring’s offspring, and of theirs in turn. Thus we have a theoretical basis for choosing one set of actions over another: we should act to create the greatest freedom consistent with the richest order in the effect-state it produces.
If there is only one possible future, however, it is indistinguishable from the past and can in theory be remembered perfectly; but since it is going to happen anyway, a memory of it would be useless. Thus all practically useful futures–futures we can do something about and change–are branchy, but not too branchy. Memory needs to be pruned, the bad branches lopped off, the good ones kept, but the airspace (the phase-space of action) nicely covered. The same thing later occurred to chess programmers: a large part of their programming is devoted to throwing out unlikely and unprofitable lines of play–that is, to forgetting. Many higher organisms, especially among the mammals, devised sophisticated systems for forgetting. One of the most important of these seems, according to some very interesting recent research, to be the memory potentiation filter of the hippocampus during rapid-eye-movement sleep. We experience this process as dreaming; in some way or other, however bizarre, we remember what bits of the recent past stick together with the tried-and-tested and thus remembered distant past, and forget the rest.
What does it mean to say that the future is “branchy”? To answer this question we need to understand the nature of cause, which is best done by distinguishing it from logical inference. Cause is different from logical inference precisely because a cause has many possible effects, and is branchy, while inference–which is what classical Turing computers do–can have only one right answer. The branchiness of causes is called “entropy”–the diffusion of simple uniform energy/information states into more differentiated ones. Claude Shannon showed that the increase of entropy was simply the increase of the probability of a system: it is more likely that all its possible states of being are occupied than that only a few are, and time is the increase of probability. What is a simple, unlikely, and thus memorable state now will tend to branch into many finely distinguished and hard-to-remember states later. We can’t remember the future because there is not one future but many, and choosing one is not memory, but action. We prune the branchpoints that the present offers us by action, and memory is what remains. (We may observe parenthetically that this is why attempts by public policy to guard against all possible negative outcomes are doomed to failure. Such attempts are equivalent to being afraid of every conceivable outcome and thus being paralyzed with terror. There is no avoiding the tragic nature of time; and courage thus turns out to be the most effective time-management strategy. )
Thus the difference between memory and physical continuity is that mere physical continuity does not forget anything. All the particles in the universe are entangled through their quantum “memory” of the big bang. But since all virtual universes are also entangled, quantum memory is indefinite unless higher forms of memory collapse, epitomize and prune off the tails of the wave functions of the past. Space and the speed of light constitute another kind of physical “memory”. A bundle of information in the form of a set of light waves passing through light-years of vacuum remains virtually unchanged as other parts of the universe alter. Information preserved in light patterns expanding through space can be observed later; but it is only when it is observed, its progress halted, its immortality violated, and its uncertainties rudely dismissed and forgotten, that it can be remembered in the higher sense of being able to alter the future.
The crucial function of memory is to forget, in the same sense that the crucial function of evolution is extinction. The present suite of species on the Earth is defined by all the intermediate species that have died out, and are biologically forgotten. The fetal baby’s hand distinguishes itself from its original flipper by the death–apoptosis–of all the cells between the fingers. The memory-trace in the brain is constituted by the disabling of all the synaptic junctions that would specify alternative memory traces. Forgetting is the chisel that sculpts the memory out of the rock of mere physical continuity. Borges’ story “Funes the Memorious” is a thought-experiment that shows us that to remember everything is to be nothing in particular; as Michael Polanyi says, to attend to something is to attend away from everything else. Or as Nietzsche says, to act is to forget, to draw around oneself a horizon of self-limitation. To act is to murder all other opportunities: in economics any purchase has an opportunity cost. Forgetting, then, is acting backwards in time, acting is forgetting forwards in time; and forgetting is the function by which more temporally evolved organisms transcend mere physical continuity.
In poetry, the use of rhyme and meter radically prunes the number of possible sentences and expressions that the poet can use and the reader remember. The perfect poem is one in which there are no alternatives to the one chosen. It cannot be forgotten because no other version can be remembered; all the others are forgotten. Samuel Taylor Coleridge said that for a true poem it was as hard to remove a single word as it would be to push a brick out of a wall with your finger.
Poetry to be poetry must be metrically organized. My own research shows that all human poetry, with the exception of some modernist experimental free verse, is constrained by a line that is about three seconds long, is regularly repeated in terms of the length, stress, tone, rhyme, and/or number of syllables. The regularity of the repetition sets in motion an iterative process that drives a brain rhythm–a repeated burst of neuronal firings around complex circuits. This rhythm corresponds in period to the brain’s own three-second present moment and short-term memory. We forget every three seconds most of the content of the previous three seconds, preserving only the gist of it that is relevant to our immediate desire and focus of attention; it is as if we dreamed the rest of it into forgetfulness in little microsleeps. Poetry, by repeating the form in which its content is embodied, preserves the memory of it across those little sleeps; in A Midsummer Night’s Dream Shakespeare comments that poetry is a kind of coherent dreaming that “bodies forth/ The forms of things unknown”, “Turns them to shapes, and gives to airy nothing/ A local habitation and a name.” Homer said that the Muses were the daughters of Memory, and this is what he meant. Memorability is not just a pleasant characteristic of poetry: it is the defining characteristic. And poetry achieves this feat by doing our forgetting for us, by making metrically impossible any other idea than the ones in the poem.
To summarize the history of memory, memory begins with the almost perfect continuous identity of an elementary particle not interacting with anything else. Such a particle does not change because all of its possible states throughout the universe already exist, but do so at a vanishingly small probability for each. Forgetting begins when such a probability function is collapsed, for instance when a photon collides with a piece of matter; the tails of the probability function are lopped off and the probability of what remains increases. To be something in particular it has sacrificed all its might-have-beens. It joins the world of somethings, but it is now a hostage to the universe of causes with multiple effects. Instead of being safely distributed into a multitude of superposed states, it must now face a multitude of possible futures. Many of them spell destruction. Only very stable low-energy conglomerations of matter and energy, like crystals, survive. They remember their former states by being their former states. But survival can be enhanced by developing a second, virtual memory, using recording devices such as immune system T-cells or neural networks. The limits of such virtual memory systems require a second kind of forgetting, to flush out memories irrelevant to survival, retaining useful knowledge, but clearing the decks for action. Higher animals do so by techniques such as REM sleep and hippocampal potentiation. But now our efficient forgetting system, aided by the death of each generation of individuals, runs up against the limits of what it can imagine if only immediate and practical information is preserved. Human recording devices, such as poetry, painting, sculpture, and later writing and books and compact disks, introduce a new, almost unlimited, form of remembering that does not, however, burden the individual memory. We remember our dead and thus transcend the traditional forgetting system.
So we stand at the end of a long dialectic between remembering and forgetting–quantum remembering giving way to classical-physical forgetting, classical-physical remembering giving way to thermodynamical forgetting, thermodynamical forgetting overcome by cellular and neural remembering, neural memory clutter cleaned up by REM sleep forgetting, cultural memory defying individual oblivion. J. T. Fraser showed us that time evolves. Perhaps one way of understanding how it did so is that the to-and-fro whipsaw of remembering against forgetting generated the higher capacities for experiencing time and necessitated the emergence of the higher temporalities. In the poem that follows, with its repetitive meter and rhyme, about seventeen cultural memories are organized by the dream-logic of poetry around the achievements of our founder, so as to establish a memory of him, to place him in the company of the dead or mythical ancestors, and celebrate his place in our future.
The Man Who Lives Many Times:
Greetings to J.T. Fraser on his Eightieth Birthday
I think that Julius is a multitude,
Myriad-minded, Coleridge would have said;
I think he’s got the secret of the food
On which Wells’ brood of baby giants fed.
I think that Julius is a kind of god:
A spirit of the woods, a wandering man;
He’s in disguise; he’s got the rather odd
And sly and wise and tragic face of Pan.
For Julius taught us, if we could but hear it,
That everything must feel, so it survive;
That everything has got a sort of spirit,
And that it must be locally alive;
That everything’s a knower and a known,
A clock that ticks in its own proper time,
A tuning-fork to catch the ambient tone,
Predicting what is coming, by its rhyme.
I think that Julius is a sort of poet:
He took with Alighieri that dark way
True Thomas took, who traveled it to know it,
The path that winds about the ferny brae:
The path to fair elfland, elysium,
The place where time is still and never-ending,
Rimbaud’s blue isles, Yeats’s Byzantium,
Where there’s no more ascending or descending;
But ordinary heroes can’t return
(Hic labor est, said Virgil, who had been);
Julius got back with just a trace of sunburn,
And exact photographs of what he’d seen…
I think that Julius is old Li Bai,
Who danced once with his shadow and the moon;
I think he’s Odin, with his one bright eye,
And his oak staff carved with a great rune;
I think he rides the jump-seat of the photon,
I think he’s trunked Ganesha, lapis blue,
I think he’s Basho with his hat and coat on,
I think that Julius is the Wandering Jew.
I think he’s Gyula, from the Hunnish Puzsta,
I think he’s Jacob, come home in the end,
I think that he’s Coyote, the old trickster,
I think that he’s our wise and funny friend.