Francisco Varela's Home Page
Articles on Neurophenomenology and First person Methods
The Specious Present : A Neurophenomenology of Time Consciousness
Francisco J. Varela
To appear in:
J.Petitot, F.J.Varela, J.-M. Roy, B.Pachoud and (Eds.),
Naturalizing Phenomenology: Issues in Contemporary Phenomenology and Cognitive Science,
Stanford University Press, Stanford (1997, in press)
My purpose in this article is to propose an explicitly naturalized account of the experience of present nowness on the basis of two complementary sources: phenomenological analysis and cognitive neuroscience. What I mean by naturalization, and the role cognitive neuroscience plays will become clear as the paper unfolds. It would be foolish to claim that one can tackle this topic and expect to be satisfied. The experience of temporality addresses head-on the fundamental fact that we exist only within a transparent web of time. Its elucidation occupies a central place in the history of thought altogether, and most certainly in the Phenomenological tradition. Many have referred to St.Augustine's Confessions , (book XI) on the paradoxes of a past-containing nowness (distentio animi), or more recently to the pragmatic psychology of William James as found in the famous Principles of Psychology where the apt expression 'specious present' that gives this article its title is coined. Edmund Husserl considered temporality a foundational axis of his phenomenological research: all other forms of mental activity depend upon temporality, but it does not depend upon them. He worked on these questions until his death (see Appendix I). Unlike his illustrious predecessors who never picked their descriptions to the bones, Husserl did manage to bring about essential progress in the formulation of the basic structures of intimate time. I will start by drawing from the phenomenological account developed by Husserl, fully aware that I am stepping right into a complex debate within phenomenological research. The reason for taking as a base his texts on time is not some kind of Husserlian scholastic obsession. I cannot overemphasize that my use of Husserl's use of the phenomenology of time is not concerned with a close textual reading in order to prove or disprove a point in the author's thought. I prefer to take my cues from Husserl's style as an eternal beginner, always willing to start anew; this is the hallmark of phenomenology itself, (but it has not always been the case in practice). What interests me most is the unfinished motion of his writings: only in their turns and jumps we get a glimpse of how his descriptions contains, among many other ideas in germ, a dynamical bent which is key, I will argue, for a naturalization project. This will provide the bridges to cognitive neuroscience discussed here in relation to recent results from neural dynamics and the emergence of large-scale assemblies that provide a counterpoint to the constitution of temporality. I take quite literally the importance of seeing experience as a first-hand description. As B. Besnier has recently remarked, we should not "...neglect this essential guiding principle of phenomenology, namely, that it advances by description ...of an experience that one must therefore re-do" . It is on the basis of such active involvement in the phenomenology of time that I introduce some plausible additions and interpretations to the current discussion. In brief, my approach to temporality is a study case of a general research direction I have called neurophenomenology , in which lived experience and its natural biological basis are linked by mutual constraints provided by their respective descriptions (Varela, 1996). This also means that I will not hesitate to mix in the same mode of discourse both partners in this dance as if they belonged to a common language, for that is the aspiration, and only in action can we see it happening. Given the importance of the topic of the experience of temporality, let it be clear that I consider this as an acid test of the entire neuro phenomenological enterprise.
As is common to any true phenomenological study, the exploration of time involves the gesture of reduction, and the identification of descriptive invariants. Husserl made no exception in his analysis. Phenomenological reduction is an entire topic that I cannot expand here, but what I have in mind more than the overall philosophical justification and context for reduction, is its actual practice, its pragmatics . For the reader unfamiliar with this terminology let us at least say that reduction starts by a disciplined suspension of one's habitual attitudes, a bracketing of what we seems to know. This bracketing provides the opportunity for a fresh look at phenomena, in this case temporality as it appears directly to our flesh and bones selves. As soon as we enter a study with this kind of authentic philosophical attitude it becomes apparent that the familiar account of time inherited from our modern western cultural background is inadequate. In fact, we have inherited from classical physics a notion of time as an arrow of infinitesimal moments, which flows in a constant stream. It is based on sequences of finite or infinitesimal elements, which are even reversible for a large part of physics. This view of time is entirely homologous to that developed by the modern theory of computation. As a refined expression of general computation, a Turing machine and its writing head inscribes symbols one by one in an infinite string, giving rise to time as a sequence-stream, exactly as in classical mechanics. As computational views entered into cognitive science in the form of the computationalist (or cognitivist) viewpoint, computational time was unquestioningly used in the cognitive study of time. Some research continues to base experimental studies on an 'internal timer' giving rise to duration at various scales. A (hypothetical) clock emits pulses which translate into behavior; judgments on duration depend on pulse counting, and are reflected on memory and decision (Church and Broadbent, 1990). This strict adherence to a computational scheme will be, in fact, one of the research frameworks that needs to be abandoned as a result of the neuro-phenomenological examination proposed here. But I will return to this conclusion later, after presenting my main argument.
Even under a cursory reduction, already provided by the reflections of Augustin and James, time in experience is quite a different story from a clock in linear time. To start with, it does present itself as a linear sequence but as having a complex texture (whence specious, it is not a "knife-edge" present), and its fullness is so outstanding that it dominates our existence to an important degree. In a first approximation this texture can be described as follows: There is always a center, the now moment with a focused intentional content (say, this room with my computer in front of me on which the letters I am typing are highlighted). This center is bounded by a horizon or fringe that is already past (I still hold the beginning of the sentence I just wrote), and it projects towards an intended next moment (this writing session is still unfinished). These horizons are mobile: this very moment which was present (and hence was not merely described, but lived as such) slips towards an immediately past present. Then it plunges further out of view: I do not hold it just as immediately, and I need an added depth to keep it at hand. This basic texture is the raw basis of what I will be discussing in extenso below. In its basic outline, we shall refer to it as the three-part structure of temporality. It represents one of the most remarkable results of Husserl's research as a result of phenomenological reduction. Another important complementary aspect of temporality as it appears under reduction, is that consciousness does not contain time as a constituted psychological category. Instead, temporal consciousness itself constitutes an ultimate substrate of consciousness where no further reduction can be accomplished, a"universal medium of access to whatever exists...Constitutive phenomenology can well be characterized as the consistent and radical development of this privilege of consciousness into its last ramifications and consequences" (Gurwitsch, 1966, p xix).
We find a converging conclusion in James concerning the apparent paradox of human temporal experience: on the one hand there is the unity of the present, an aggregate we can describe where we reside in basic consciousness, and on the other hand this moment of consciousness is inseparable from a flow, a stream (Chapter IX of Principles). These two complementary aspects of temporal consciousness are the main axes of my presentation.
This rough preliminary analysis of time consciousness leads, then, to distinguish three levels of temporality which will guide my argument:
(1) A first level proper to temporal objects and events in the world. Thus it is close to the ordinary notions of temporality in human experience which grounds the one currently used in physics or computation .
(2) The phenomenologist starts from this level, but reduction makes apparent the second level, that of acts of consciousness that constitute objects-events . This is the "immanent" or "internal time" of acts of consciousness. Their structure forms the main body of the phenomenological analysis in Husserl's Lectures. (3) Finally, (and this is the more subtle levels of analysis), these first two levels are constituted from another level where no internal-external distinction is possible, and which Husserl calls the "absolute time constituting flow of consciousness" (PZB 73) .
Time never appears detached but as temporal object-events which are the correlates or the intentional focus of the temporal consciousness: temporal object-events are what these acts are about . In contrast to what might be of interest for the psychologist or the neuroscientist, for the phenomenologist the content of the object is not as important as the manner of its appearance .
At various points in his research Husserl returns to the basic observation that what is proper to temporal objects is their double aspect of duration and unity (PZB 23, 113-4). Duration is correlative to the intentional direction: this house I am walking by, that bird that flew from here to there,.... These are actual durations, and refers to the object having a location in time. Unity is correlative to the individuality of the object-events in question, standing out as a distinct whole against the background of other events. Thus a temporal object-event covers a certain span T, a complete act. However the entire act is a continuous process in the course of which moments of nowness are articulated, not as a finished unity but in a succession. It is this mode of constitution of object-events, whatever their duration T and their content, that interests me here. It must be the case that consciousness of succession derives from structural features of the acts of consciousness. Our problem is the characterization of these structures.
In his writings Husserl is characteristically sparse with examples. In PZB he uses one recurrent illustration: listening to a melody (a choice most later commentators have followed as well). I think it is important to examine this issue more carefully for various reasons. First, because when Husserl brings his example to hand, it is, strangely enough, as an un-situated subject, in an abstract mode: we don't know in what circumstances the music is being listened to (is he alone, in a concert hall?), nor whether this is background listening or an intense emotional concentration (is it a moving piece, is he familiar with it?). All this is not merely anecdotal, since without these particularities the mode of access to the experience itself which is lost.
We are still missing a phenomenology of internal time consciousness
where the reductive gestures and the textural base of the experience figures
explicitly and fully. The entire phenomenology of internal time consciousness
should be rewritten with the precision of a mode of access to experience
that serves as support for reduction; what transpires in Husserl's writing
is often far from such texture . In what follows I depart from usage by
inviting the reader to actually engage in a specific experience of multi-stable
visual perception that will provide reader and author with an explicit
common ground. Multi-stable perceptions are a good case study: they are
precise and complex enough for the analysis, while still providing corresponding
neurocognitive correlates. Indeed, multi-stable visual perception phenomena
are deceptively clear.
A TASK IN VISUAL PERCEPTION´ Consider the following image:
Out of context, most western subjects naturally see either a "pyramid" or a "hallway":
´ Suspend habitual interpretations and consider multiple variations together. We may infer that since ??depicts neither pyramid nor hallway, it is an ambiguous or multi-stable perception. ´ These variations can be practiced by the following strategy: 1. Fixate on the center of the figure, 2. Blink your eyes, 3. "Aim" at its alternative. These instructions turn the subject into an active agent, as the condition of possibility of the perception, the constitution of meaning itself illustrated here in a minimal case. There is a lot more to explore in this experiment concerning the perceptual horizon used for changing perception, as well as in regards to the strategy of language (the naming of the alternative, the 'aiming' at the other possibilities) . But let me stay with the temporal aspects of this perception. We notice, as it has been known since the Gestaltists, that as we acquire proficiency, the reversal can be accomplished with little effort and upon request. It is also clear that the reversal has, in itself a very complex dynamics that takes on a "life" of its own (more on this later). We also notice that the gesture of reversal is accompanied by a "depth" in time, an incompressible duration that makes the transition perceptible as a sudden shift from one aspect to the other, and not as a progressive sequence of linear changes. Granted, this well-know phenomenon is not common in ordinary life. A more 'ecological' case would be when I open a door, and as I move across the threshold I run into somebody who is directly in front of me. Both through body motion and visual orientation the person's face comes spontaneously into focus: I recognize a colleague, and thrust out my hand for a greeting. In both the multi-stable visual perception and the vignette just presented, I want to highlight what happens as one moves away from object-events dominated by passive attitude as in listening to music. In the example of multi-stable perceptions we do not need to actively move our entire body. However an important layer of motion is actively present, whether in head adjustement, frowning and blinking, and, surely, in eye movements of various kinds. This is important, for little can be concluded from cases where motricity is absent. As phenomenological research itself has repeatedly emphasized it is in the active sensori-motor interdependence that perception is based on. Independently, several traditions in cognitive research have, in their own way underlined perception-action as a key . It is this active side of perception which gives temporality its roots in living itself . Within this general framework, I will concentrate more precisely on the structural basis and consequences of this sensori-motor integration for our understanding of temporality.
My overall approach to cognition is based on situated, embodied agents. I have introduced the name enactive to designate this approach more precisely. It is comprised of two complementary aspects.
(1) On the one hand, the ongoing coupling of the cognitive agent, a permanent coping that is fundamentally mediated by sensori-motor activities.
(2) On the other hand, the autonomous activities of the agent whose identity is based on emerging, endogenous configurations (or self-organizing patterns) of neuronal activity.
Enaction implies that sensori-motor coupling modulates, but does not determine, an ongoing endogenous activity that it configures into meaningful world items in an unceasing flow.
I cannot expand this overall framework more extensively , but it is the background of my discussion of temporality as a neurocognitive process. Enaction is naturally framed in the tools derived from dynamical systems, in stark contrast to the cognitivist tradition that finds its natural expression in syntactic information-processing models. The debate pitting embodied-dynamics vs. abstract-computational as the basis for cognitive science is still much alive. For some time I have argued for the first and against the second, and this choice justifies the extensive use of dynamical tools in this paper (cf. Appendix II).From an enactive viewpoint, any mental act is characterized by the concurrent participation of several functionally distinct and topographically distributed regions of the brain and their sensori-motor embodiment. It is the complex task of relating and integrating these different components that is at the root of temporality from the point of view of the neuroscientist. A central idea pursued here is that these various components require a frame or window of simultaneity which corresponds to the duration of lived present . In this view, the constant stream of sensory activation and motor consequence is incorporated within the framework of an endogenous dynamics, (not informational-computational one), which gives it its depth or incompressibility. This idea is not merely a theoretical abstraction: it is essential for the understanding of a vast array of evidence and experimental predictions . These endogenously constituted integrative frameworks account for perceived time as discretized and not linear, since the nature of this discreteness is a horizon of integration rather a string of temporal "quanta". At this point it is important to introduce three scales of duration to understand the temporal horizon as just introduced: (1) basic or elementary events, (the '1/10' scale);(2) relaxation time for large-scale integration, (the '1' scale);(3) descriptive-narrative assessments, (the '10' scale).This recursive structuring of temporal scales composes a unified whole, and it only makes sense in relation to object-events. It addresses the question of how something temporally extended can show up as present but also reach far into my temporal horizon. The importance this tri-level recursive hierarchy will be apparent all through this paper.
The first level is already evident in the so-called fusion interval
of various sensory systems: the minimum distance needed for two stimuli
to be perceived as non-simultaneous, a threshold which varies with each
sensory modality. These elementary events can be grounded in the intrinsic
cellular rhythms of neuronal discharges, and in the temporal summation
capacities of synaptic integration. These events fall within a range of
10 milliseconds (e.g. the rhythms of bursting interneurons) to 100 msec
(e.g. the duration of an EPSP/IPSP sequence in a cortical pyramidal neuron).
These values are the basis for the 1/10-scale. Behaviorally these elementary
events give rise to micro-cognitive phenomena variously studied as perceptual
moments, central oscillations, iconic memory, excitability cycles and subjective
time quanta. For instance, under minimum stationary conditions reaction
time or oculo-motor behavior displays a multimodal distribution with a
30-40 msec distance between peaks; in average daylight, apparent motion
(or 'psi-phenomenon') requires 100 msecs. This leads us naturally to the
second scale, that of long-range integration. Component processes already
have a short duration, on the order of 30-100 msec; how can such experimental
psychological and neurobiological results be understood at the level of
a fully constituted, normal cognitive operation? A long-standing tradition
in neuroscience looks at the brain basis of cognitive acts (perception-action,
memory, motivation and the like) in terms of cell assemblies or,
synonymously, neuronal ensembles. A Cell Assembly (CA) is a distributed
subset of neurons with strong reciprocal connections. CAs comprise distributed
and actively connected neuronal populations (including neo-cortical pyramidal
neurons, but not limited to them). Because of their pressumed strong interconnections,
a CA can be activated or ignited from any of its smaller subsets, sensori-motor,
or internal. The term reciprocal is crucial here: one of the main
results of modern neuroscience is to have recognized that brain regions
are indeed interconnected in a reciprocal fashion (what I like to refer
to as the Law of Reciprocity). Thus, whatever the neural basis for cognitive
tasks turns out to be, it necessarily engages vast and geographically separated
regions of the brain. These distinct regions cannot be seen as organized
in some sequential arrangement: a cognitive act emerges from the gradual
convergence of various sensory modalities into association or multimodal
regions and into higher frontal areas for active decision and planning
of behavioral acts. The traditional sequentialistic idea is anchored in
a framework in which the computer metaphor is central, with its associated
idea that information flows up-stream . Here, in contrast, I emphasize
a strong dominance of dynamical network properties where sequentiality
is replaced by reciprocal determination and relaxation time . The genesis
and determination of CAs can be seen as having three distinct causal and
temporal levels of emergence: an onto-genetic level which sets the anatomical
architecture of a given brain into circuits and subcircuits; a second,
developmental-learning level: sets of neurons that are frequently co-active
strengthen their synaptic efficacies; and a third level of determination
for the CAs' constitution. The third level is the faster time scale of
the experience of immediate daily coping, which manifests at the perception-action level, a duration of the order of seconds. In the language of the dynamicist,
the CA must have a relaxation time followed by a bifurcation or
phase transition, that is, a time of emergence within which it arises,
flourishes, and subsides, only to begin another cycle. This holding time
is bound by two simultaneous constraints: (1) it must be longer than the
time of elementary events (the 1/10 scale); (2) it must be comparable to
the time it takes for a cognitive act to be completed, i.e. on the order
of a few seconds, the 1-scale) (e.g. Varela et al. 1981; Pöppel, 1988).
In brief, as we said before, (the relevant brain processes for ongoing
cognitive activity are not only distributed in space, but they are also
distributed in an expanse of time that cannot be compressed beyond a certain
fraction of a second, the duration of integration of elementary events.
In view of the above, I will now introduce three interlinked, but logically
independent, working hypotheses:
Hypothesis I: For every cognitive act, there is a singular, specific cell assembly that underlies its emergence and operation .The emergence of a cognitive act demands the coordination of many different regions allowing for different capacities: perception, memory, motivation, and so on. They must be bound together in specific grouping appropriate to the specifics of the current situation the animal is engaged in (and are thus necessarily transient), in order to constitute meaningful contents in meaningful contexts for perception and action. Notice that Hypothesis I is strong in the sense that it predicts that only one dominant or major CA will be present during a cognitive act. What kind of evidence is there to postulate that every cognitive act, from perceptuo-motor behavior to human reasoning, arises from coherent activity of a sub-population of neurons at multiple locations? And further, how are such assemblies transiently self-selected for each specific task? Since this will be a recurrent topic for the remainder of this article, I'd like to formulate it as the second part of my working Hypothesis. The basic intuition that comes from this problem is that a specific CA emerges through a kind of temporal resonance or "glue". More specifically, the neural coherency-generating process can be understood as follows:
Hypothesis II: A specific CA is selected through the fast, transient phase locking of activated neurons belonging to sub-threshold ,competing CAs. The key idea here is that ensembles arise because neural activity forms transient aggregates of phase-locked signals coming from multiple regions. Synchrony (via phase-locking) must per force occur at a rate sufficiently high so that there is enough time for the ensemble to "hold" together within the constraints of transmission times and cognitive frames of a fraction of a second. However, if, at a given moment, several competing CAs are ignited, different spatio-temporal patterns will become manifest and hence the dynamics of synchrony may be reflected in several frequency bands. The neuronal synchronization hypothesis postulates that it is the precise coincidence of the firing of the cells that brings about unity in mental-cognitive experience. If oscillatory activity promotes this conjunction mechanism, it has to be relatively fast to allow at least a few cycles before a perceptual process is completed (e.g. recognition of a face and head orientation). Recently this view has been supported by widespread findings of oscillations and sychronies in the gamma-range (30-70 Hz) in neuronal groups during perceptual tasks. The experimental evidence now includes recordings during behavioral tasks at various levels, from various brain locations both cortical and sub-cortical, from animals ranging from birds to humans, and from signals spanning broad-band coherence from single units, local field potentials and surface evoked potentials (electric and magnetic) (see Singer, 1993; Varela, 1995 for history and summary of this literature). Figure 1 provides a sketch to help clarify these ideas. This notion of synchronous coupling of neuronal assemblies is of great importance for our interpretation of temporality, and we will return to it repeatedly below (see also Appendix III). Here is the point where things get really interesting in our development of a view cognition which is truly dynamic, making use of both recent advances in non-linear-mathematics and of neuroscientific observations.
A diagram depicting the three main Hypothesis I-III utilized here. A cognitive activity (such as head turning) takes place within a relatively incompressible duration, a 'cognitive present'. The basis for this emergent behavior is the recruitment of widely distributed neuronal ensembles through increased frequency coherence in the gamma (30-80 Hz) band. Thus, the corresponding neural correlates of a cognitive act can be depicted as a synchronous neural hypergraph of brain regions undergoing bifurcations of phase transitions from one cognitive present content to another.
Thus, we have neuronal-level constitutive events that have a duration
on the 1/10-scale, forming aggregates that manifest as incompressible but
complete cognitive acts on the 1-scale . This completion time is dynamically
dependent on a number of dispersed assemblies and not a fixed integration
period, in other words it is the basis of the origin of duration without
an external or internally ticking clock. These new views about cognitive-mental
functions based on a large-scale integrating brain mechanisms have been
emerging slowly but with increasing plausibility. Nowness, in this perspective,
is therefore pre-semantic in that it does not require a rememoration, (or
as Husserl says a presentification, see below) in order to emerge. The
evidence for this important conclusion comes, again, from many sources.
For instance, subjects can estimate durations of up to 2-3 secs quite precisely,
but their performance decreases considerably for longer times; spontaneous
speech in many languages is organized such that utterances last 2-3 secs;
short intentional movements (such as self-initiated arm motion) are embedded
within windows of this same duration. This brings to the fore the third
duration, the 10- scale, proper to descriptive-narrative assessments. In
fact, it is quite evident that these endogenous, dynamic of nowness horizons
can be, in turn, linked together to form a broader temporal horizon. This
temporal scale is inseparable from our descriptive-narrative assessments,
and linked to our linguistic capacities. It constitutes the "narrative
center of gravity" in Dennett's metaphor (Dennett, 1991), the flow
of time related to personal identity (Kirby, 1991). It is the continuity
of a self that breaks down under intoxication or in pathologies such as
schizophrenia or Korsakoff's syndrome. I am now ready to advance the last
step I need to complete this part of my analysis:
Hypothesis III:The integration-relaxation processes at the 1-scale are strict correlates of present-time consciousness. We are thus referred back to the experiential domain, and the nature of this link is what we need to explore carefully. Distinctions between ongoing integration in moments of nowness, and how their intergation gives rise to broader temporal horizons in re-membrance and imagination are at the core of the Husserlian analysis of intimate time, to which we now return.
Temporal objects appear to us as such only because of the correlative acts of consciousness which have specific modes of appearance that are at the very heart of the issue of immediate temporality. Normally we designate these modes by the terms present now, past and future. Beyond this cursory designation however, reduction clearly points to the mode of 'now' as having a unique or privileged status (PZB 35). Two lines of analysis lead to this. First, the texture of now which James calls specious. In effect, now is not just a mere temporal location, it has a lived quality as well: it is a space we dwell in rather than a point where an object passes transitorily. Second, it is in relation to the rich structure of present nowness that all other modes of temporality take form. Both of these lines of analysis will be explored here.
To start, in spite of Husserl 's own descriptions and geometric depictions, nowness does not correspond to a point but rather to a location. It is not an object, but a field with a structure analogous to the center and periphery structuring of the visual field. Husserl himself speaks of nowness as a "temporal fringe" (PZB 35). In other words, the very mode of appearance of nowness is in the form of extension, and to speak of a now-point obscures this fact:`
"...present here signifies no mere now-point but an extended objectivity which modified phenomenally has its now, its before and after" (PZB 201).
Husserl is grappling here with one of the antinomies of time: is always changing and yet in some sense always the same. A temporal object-event such as my identifying the figure as a pyramid has a unity which first appears as present nowness. It then slips away when it appears anew as a hallway. The previous recognition (and its given-ness) has now sunk into the past, as when an object moves from center to periphery in space. This marks the beginning of the resolution of the apparent contradiction between sameness and difference, constancy and flow, which we will elaborate more below. To do so, I must now move to another further level of detail in the study of the structure of consciousness, one which is constitutional, (in Husserlian jargon), insofar as it provides the temporal features of mental acts that unify them into a single flow of consciousness.
The key issue in this stage of my examination of temporality is the contrast between the mode of appearance of now and of the just-passed, the act which reaches beyond the now. As Husserl points out, commenting on similar reasoning in Brentano:
"We could not speak of a temporal succession of tones if...what is earlier would have vanished without a trace and only what is momentarily sensed would be given to our apprehension" (PZB 397).
But how can this structure of the time perception be constituted? What is preserved is also modified. If when I see a pyramid I could still hold unchanged the nowness of when I saw the hallway, all temporal structure would disappear. The relation of now to just- past is one of slippage organized by very strict principles:
"...new presentations each of which reproduces the contents of those preceding attach themselves to the perceptual presentation and in so doing append the continuous moment of the past" (PZB 171, my emphasis)
This phrase expresses the intuitions behind the analysis of dynamics intrinsic to these slippages of appearance, as developed in the following section. It is key for my search for bridges between naturalization the texture of temporal experience.
What form takes this slippage from nowness to immediate past? Is it another emergent aggregate of temporal horizon, similar to the one we assume (Hypothesis III) underlies the experience of nowness? This was Brentano's postulate: the constitution of the past is re-presentation or memory presentification of what preceded. In several remarks dispersed over the years, Husserl actually gets close to a demonstration that the now to-just-past slippage is not the same as immediate memory retrieval or presentification . To the appearance of the just-now one correlates two modes of understanding and examination (i.e. valid forms of donation in the phenomenological sense): (1) remembrance or evocative memory and (2) mental imagery and fantasy.There are at least two main arguments for Husserl's observation. (1) First, the nature of memory retrieval is one that is created now, and there is surely a nowness to the act of remembering . Thus we cannot account for the past with an act that is supposedly happening in the now. (2) Second, when I remember having seen the hallway ( as opposed to being in the embodied situation seeing it now) the past and the successive past that receded into oblivion have an immediacy, an evidence to them. In the present I "see" what just passed; in memory I can only hold it in a representation as if through a veil. Thus memory and evocation have a mode of appearance that is qualitatively different from nowness.Let us return back to the visual task, and resume our examination. As I look at the pyramid, I experience the near side at that moment of now, and as well there is the unity of the object as a durable unity (the pyramid is namable). This reveals the play between the primal impression of the near side of visual perception-action and the constituting unity that make an identifiable object-event appear. It is the manifold of retentions, as illustrated by the excersice, that makes sense of the entire event: the now is experienced in an "original" way."For only in primary remembrance do we see what is past; only in it is the past constituted, i.e. not in a representative but in a presentative way. ...It is the essence of primary remembrance to bring this new and unique moment to primary, direct intuition, just as it is the essence of the perception of the now to bring the now directly to intuition" (PZB 41).
Accordingly Husserl makes a disciplined distinction between impressional as opposed to representational consciousness. In impression an object is originally constituted, and is thus given as present (I am now looking into the page and see the pyramid). Representation represents an object-event already given to impression, (I evoke seeing the pyramid a little while ago for the first time). The Ur-impression is the proper mode of the now, or in other words is where the new appears; impression intends the new (I was not counting on seeing the pyramid as I looked). Briefly: impression is always presentational, while memory or evocation is re-presentational. Besnier remarks: " This is what is usually expressed by saying that retention belongs to a 'living present' (lebenhaftig or lebendig). But there is something that retention de-presents (démomentanéise) which one could translate by ent-gegenwärtigung,, but this choice might create confusions...". Similar conclusions can be drawn from neurocognitive evidence. For a long time cognitive psychology has distinguished between evocative memory as implied above, and other forms of immediate retention, for example: short-term, working or iconic memory . I am taking these as correlates of the just-past. Interestingly, the chronometric studies of brain correlates of memory draw a clear distinction between impressional perceptual events and perceptual events requiring the mobilization of memory capacities. For instance, comparing rote vs. elaborate mnemonic recollection of items, yields a substantial (200-400 msec) shift in the corresponding ERPs (see e.g Rugg, 1995). More recently, brain imaging methods have begun to establish that quite different structures are mobilized during active presentational tasks whether they are a form of memory (episodic, operational), or imagination. All these presentational tasks mobilize a whole new set of capacities which add to the flow of time but are obviously not required for it.
Husserl introduces the terms of retention and protention to designate this dynamics of impression. Retention is the attribute of a mental act which retains phases of the same perceptual act in a way that is distinguishable from the experience of the present, (but that is not a re-presentation, as we just saw). The key feature for it is retention that is its direct contact with earlier perceptions making perception at any given instant contain entities that show up as temporally extended. As we discussed, under reduction duration has a speciousness, it creates the space within which mental acts display their temporality. Similarly (but not symmetrically) another distinction seeks future threads or protentions. This is the three-part structure that transforms an intentional content into a temporal extension. Figure 2 reproduces a sketch, one of the 'figures of time' taken from the 1905 lectures, a geometrical depiction of the three-part structure of temporality. The now and the degrees of pastness are what Husserl refers to as Ablaufsmodi, slippages or elapsing modes, and which are depicted in reference to a source-point, whence the discreteness in this diagram. The use of lines and points is unfortunate, since it distracts from Husserl's remarkable insight; one might guess this is an echo of his training as a mathematician. The fate of these figures of time has been curious. Since they are easily grasped they have been used extensive as summaries of Husserl's position. Yet these 'figures of time' hardly stands for the scope of his explorations; at best they illustrate an earlier and static view (Fig.2). The appeal of the diagrams came through in Merleau-Ponty's Phenomenology of Perception (p.477) where the author comments approvingly on these depictions. However Merleau-Ponty was well aware of the shortcoming as he added next to the diagram: "Time is not a line but a network of intentionalities" (p. 477; see Fig. 2 bottom for a sketch). Further on:"The emergence of a present now does not provoke a piling behind of a past, and a pulling of the future. Present now is the slippage of a future toto the present, and the just-past to the past: it is in one single movement that time sets in motion in its entirety" (p.479)
However, many writers have retaken uncritically this and several other diagram in Husserl's writings.
The 'figures of time' as introduced by Husserl. (a) The diagram chosen by Stein published in the Lessons (PZB 28); (b) one of several variants kept by Husserl posterior to 1905 (PZB 230); (c) the rendition by Merleau-Ponty (Phénomenologie de la Perception 477), which prolongs the lines into retentional time, as suggested by Husserl himself (cf. PZB 331); my own rendition of Merleau-Ponty's later critique of this diagram (Phénomenologie de la Perception 479).
In brief then:
"primal impression intends the new and actual phase of the temporal objects-events, and presents it as the privileged now. Retention presents just elapsed phases and presents them as just past in various degrees" (Brough, 1989, p. 274).
Retention is then a specific intentional act intending the slipping object constituting it as just past. Retention is not a kind of holding on to the now by its edge; it is an active presentation of an absence that arises from the modifications and dynamic apprehension of the now. Metaphorically it is more like moving from center to periphery than the after-effect of an image, which is like the present modified only in intensity. But in the temporal realm, it is a curious structure indeed: it is present-living past. Or, as Husserl quipped in a hand-written marginal scribbling on his working notes: "But 'perceived past' doesn't that sound like a 'wooden iron'? (hölzernes Eisen)" (PZB 415).
It is indeed a wooden iron, unless we take a dynamical view of how the origin of the now can be formulated on the basis of our three working Hypotheses I-III. The use of 'append' and 'slippage' in the phenomenological description already ovokes this view but it needs to be developed fully. Besides the fact that there is substantial experimental support for these hypotheses, it is essential to recognize that we are dealing with a bona fide candidate for the synthesis of a temporal space where cognitive events unfold. The discussion that follows requires at least some understanding of dynamics of non-linear phenomena as applied to cognitive events. The reader not familiar with basic notions can turn to Appendix II for a sketch.
Let us be more specific. Large-scale phenomena seen in the nervous system as long-range integration via synchronization of ensembles cannot be dissociated from intrinsic cellular properties of constituent neurons. Intracellular recordings studied in vivo and in vitro slices of various brain regions, of both vertebrate and invertebrates, have shown the pervasive presence of intrinsic, slow rhythms mediated by specific ionic gating mechanism (Llinás, 1988; Nuñez et al., 1993). In other words what we have are a large arrays of neural groups which, due to their intrinsic cellular properties, qualify as complex non-linear oscillators. In Appendix III an explicit dynamical account of fast-slow non-linear oscillators is presented to make the discussion concrete. The reader not interested in such technical details may skip it and stay with the general discussion.
Why is this of importance here? Because it leads us directly to an explicit view of the particular kinds of self-organization underlying the emergence of neural assemblies. These arise from collections of a particular class of coupled non-linear oscillator, a very active field of research (e.g. Mirollo and Strogartz, 199?; Winfree, 1980; Mackey and Glass, 1988; Kelso, 1994). These dynamical processes, in turn, illuminate the mechanims whereby neuronal assemblies 'now' posses a three-part structure.
The key points to keep in mind for our discussion are the following. Self-organization arises from a component level, which in our case has already been identified as the 1/10- scale of duration, and re-appears here as single or groups of non-linear oscillators. Second we need to consider how these oscillators enter into synchrony (see Hypothesis II) as detected by a collective indicator or variable, in our case relative phase. Third, we need to explain how such a collective variable level manifest itself at a global level as a cognitive action and behavior, which in our case corresponds to the emergence of a percept in multistability. This global level is not an abstract computation, but an embodied behavior subject to initial conditions (e.g. what I' aimed' at, what was the preceding percept), and non-specific parameters (e.g. changes in viewing conditions, attentional modulation). The local-global interdependence is therefore quite explicit: the emerging behavior cannot be understood independently of the elementary components; the components attain relevance through their relation with their global correlate.
What have we gained? Very simply that these kinds of emergent processes there can be a natural account for the apparent discrepancy between what emerges and the presence of the past. In effect, the fact that an assembly of coupled oscillators attains a transient synchrony and that it takes a certain time for doing so is the explicit correlate of the origin of nowness (Hypothesis III). As the model, (and the data, see below) shows, the synchronization is dynamically unstable and thus will constantly and successively give rise to new assemblies. We may refer to these continuous jumps as the trajectories of the system. Each emergence bifurcates from the previous ones from its initial and boundary conditions. Thus the preceding emergence is still present in the succeeding one.
To bring this idea closer to first-hand experience let us perform a second task in visual perception:
A SECOND TASK IN VISUAL PERCEPTION
´ Consider the series of images below, (Fischer, 1967):
´ Practice the following task:
1 Starting on the upper left (the 'man' position 1) slide your eyes sequentially over the figure reaching the bottom right (the 'girl' position 17).
2 At some point along this trajectory the perception switches from man to girl. Mark the point at which this happens.
3 Start again at the 'girl' end, and do the reverse perusal. Mark the place where perception switches.
After following the above instructions, observer reports that the positions at which the perceptual switches occur do not coincide, but are lengthened in either direction of the image sequence.
When this is analyzed from a dynamical point of view, the following account is plausible: Looking at the extremes of the series ('girl', 'man') the resonant assemblies are closer to a stable attractor basin corresponding to a single percept. As the parameter of ambiguity is increased (the place on the series), suddenly the emergence of a new percept is possible, that is, we have passed through a bifurcation or phase transition. However the system tends to stay close to the fixed point of origin, as if this point had an active residue (hysteresis in technical jargon) for the trajectory, a remnant which is appended (to retake Husserl's term). The order parameters constrain the trajectories, the initial percepts hover around the stable percepts, but these may wander to different positions in phase space.
This embodies the important role of order parameters in dynamical accounts. Order parameters can be described under two main aspects:
(1) the current state of the oscillators and their coupling, or initial conditions;
(2) the boundary conditions that shape the action at the global level: the contextual setting of the task performed, and the independent modulations arising from the contextual setting where the action occurs (i.e. new stimuli or endogenous changes in motivation).
This second visual task makes it is clear that we are not dealing with an abstract purely syntactic description. Order parameters are defined by their embodiment, and are unique to each case. The trajectories of this dynamics, then, enfold both the current arising and its sources of origin in one synthetic whole as they appear phenomenally. A wooden iron indeed.
The kinds of specific dynamics we have brought to bear to the understanding of retention and the just-past are not simple. In particular arrays of coupled oscillators are interesting because they do not in general behave according to the classical notion of stability that derives from a mechanical picture of the world. Stability here means that initial and boundary conditions lead to trajectories concentrated into a small region of phase space wherein the system remains, a point attractor or a limit cycle. In contrast, biological systems demonstrates instability is the basis of normal functioning rather than a disturbance than needs to be compensated, as in a mechanical picture of causality.
Lets us return once again to our experiential ground of visual multistability. As we saw in the example of Fisher's figures, the origin of multistability is due to properties that are generic to coupled oscillators and their phase relations. In other words their mode of appearance is an invariant under certain conditions and reporting subjects . This is further clarified by recent experiments performed in order to study the dynamics of multistability in visual perception. As shown in Fig. 3, Kelso et al. (1994) presented observers with variant perspectives of the classical Necker cube, a close relative of our first visual task. By asking the observer to push a button when the perceptual reversal occurs, one obtains a time series of reversal which obeys a stochastic distribution (a gamma function with a mean at around 1 sec). Kelso et al., however, requested observers to perform the same task of time measurement while noting separately the time series as a function of perspective, which is thus used as an order parameter.
Multistability of perception of the Necker cube from different perspectives, used here as an order parameter. When the cube is reversed from projection (*), the time intervals between spontaneous inversion are shown in the lower diagram, with an upward trace each time a reversal is perceived. This interswitch reversal time series has a probabilistic distribution shown below, a distribution that is considerably spread when the viewing projection is changed to (**). In this latter case the switch is more difficult, and the observer is "stuck" with one interpretation for longer periods of time. In the insets, the results of simulation of phase transition of coupled non-linear oscillators (cf. Appendix II) under one order parameter (here the viewing perceptive) followed through phase as a global variable. From Kelso (1994, p. 219-21).
The interesting observation is that, again, at the extremes of the images, (the 'hexagon' mode, and the 'square' mode), the distribution of reversal intervals is considerably flattened. The subject is more likely to have sporadic figure reversal, or to be 'fixed' on one mode for a longer duration (Fig.3). At these extremes subjects report getting 'blocked' by an interpretation. As before, one can think of these results as the way the coordination of a wide arrays of oscillators appears via a common variable of phase. By introducing perspectival variants, the location in phase space is accordingly modified, and new dynamical modes appear, in this case revealing a saddle instability (Fig. 4).
That this dynamical interpretation is actually linked to neuronal ensembles as we have been assuming here is shown by recent experiments (Leopold and Logothetis, 1996). A monkey was rigorously trained to voluntarily 'aim' at reversing a set of ambiguous figures, (binocular rivalry, a visual task known to be similar to Necker cube or Fisher figure reversal), and then to indicate the moment at which this reversal appeared for his perception. At the same time, individual neurons were recorded from a number of its visual cortical areas. The authors report that in motion sensitive area MT, a percentage of neurons correlate with reversal and can be modulated by the perceptual requirements of the tasks; this percentage is diminished in primary regions V1/V2. This kind of evidence strongly supports the notion that multistability arises through the large-scale collaborations between neurons at many different places in the visual cortex and elsewhere in the brain, a concrete example of an emerging CA for a specific task which has perceptual and phenomenal correlates.
Elsewhere in this volume, van Gelder also discusses the three-part structure of time from a dynamical cognitive science viewpoint. It is quite reassuring that we have independently reached converging conclusions concerning the key role of dynamical interpretation in a research program of naturalized phenomenology, and more particularly concerning the re-understanding of the three-part structure of time. However, our views differ in some significant aspects, which are worth underlining here for the reader.
In his discussion van Gelder grounds his analysis on "Lexin", a connectionist network developed by Anderson and Port for auditory pattern recognition. As in a large class of connectionist work, Lexin changes its connection according to a training period, following which a sound input triggers a dynamical trajectory that approaches an attractor, a unique location in state space. Lexin's architecture has its roots in work of the early 80s, which made popular the standard view of neural networks as dissipative dynamical processes governed by point attractors and limit cycles, or, as they are usually called, as attractor neural networks (ANNs; Amitt, 1991).
In contrast, in more biologically oriented dynamical studies, the generic role of instabilities has been stressed more recently, a needed expansion of results in neural networks and connectionism. ANNs have played a role in our understanding of biological circuits and, more importantly, in the developments of artificial intelligence. In the hands of the engineer, motion and perceptual devices that mimic some living behavior have provided impressive performances (see Arbib, 1995 for a recent overview). Lexin in particular is meant to retrieve a learned set of auditory patterns. However, only with a dynamical description that incorporates instability as having a role can we satisfactorily account for enactive cognition such as the three-part structure of time. In this class of dynamical systems the geometry of phase space needs to be characterized by an infinity of unstable regions, and the system flows between them even in the abscence of external input. There are no attractor regions in phase space, but rather ongoing sequences of transient visits in a complex pattern of motion, only modulated by external coupling.
Auditory perception particularly in mammals is so rapid (for example making individual sound discriminations in the order of 2-3 msecs on the 1/10-scale), that it is far from clear how a slowly converging ANN could be a good model. More important, any model based on point attractors will surely lack the intrinsic interest of endogenous instabilities proper to brain dynamics. This is, again, one of the reasons we stay close to biologically-inspired, coupled non-linear oscillations and frequency binding as a source of dynamical correlates proper to complete cognitive acts.
Van Gelder's analysis and mine coincide, however, on the intuition of the pertinence of dynamical views for the illumination of retention. Unlike him I have my reasons (discussed in Section VII) to see protention as quite asymmetrical from retention, and not as having the same geometric property of retention of a constrained path in phase space.
I need to turn now to the last step in my analysis, a step that is taken on far less trodden ground than the intentionality of object-events and retentional dynamics, two "classical" topics in the phenomenology of time after Husserl. Since the discussion so far has been concentrated on a particular kind of intentional acts, retentional dynamics still belongs to the type of constitutional analysis referred to as 'static', the classic level at which most commentators of Husserl's time analysis have remained. It is called static since it is intentionally directed to object-events and to what appears (either as external processes or immanent duration, for instance the completion of a movement). In other words so far we have considered so far the more accessible levels of temporarily, the appearance of temporality object-events, and the acts of consciousness which constitute them.
Husserl brings to the fore a third and last level of analysis which has been less explored by later work. This is "the absolute time-constituting flow of consciousness" (PZB 73). In the context of the present use of dynamical ideas it is interesting that Husserl's choice to refer to this level of study as the 'flow' is suggestive:
"It is absolute subjectivity and has the absolute properties of something to be denoted metaphorically as 'flow' ('Fluss') as a point of actuality, primal source-point, that from which springs the 'now', and so on" (PZB §36; also t. No.54, p.368).
The idea of flow opens up to two less "classical" issues that I will dwell on here: (1) the genetic constitutional analysis of time, just introduced above as the flow of absolute time, and (2) its close relative the affective dimension (cf. Section VII). Husserl established the distinction between static and constitutional analyses of time during in the years 1917-23. These topics are notoriously more difficult to explore not only because they are grounded on Husserl's late production (see Appendix I), but also because they touch on subtle areas, which is what makes them more attractive. Accordingly the reader is asked to consider what I propose in the remainder of this text as a sketch of future work more than anything else. Return to the first visual task (Do it!). Aim to a change in percept, and return to the initial one. It is clear that we have been given two distinct experiences with a similar content. The link joining both as two-of-the-same demonstrates the basic fact that there is an underlying temporalization which has a relative independence of the particular content of the views. As Husserl concisely remarks, "Every experience is 'consciousness' ('Bewusstsein') and consciousness is always consciousness-of...Every experience is itself experienced (selbst erlebt) and to that extent also intended ('bewusst'). This being intended (Bewussst-sein) is consciousness of the experience. (Erlebnis).." (PZB 291).
The link is a reflection, which is not always present but that may always be put into action, it accompanies all my acts. This reflection is temporal since experiences are (immanent) object-events with duration: they appear slipping into the past and gradually disappear into the fringes of time. It is against this background of the flow of experiencing that the duration of object-events and the experience of temporality is constituted. This underlying flow raises then a new apparent paradox: it can be detached from the temporal object-events but at the same time it appears inseparable from them, since a flow without object-events does not manifest itself.
The nature of this immanence is retaken by Husserl in the following (remarkable) passage which summarizes his previous analysis:
"I may express the situation thus: What is perceived, what manifests (selbstgegeben ist) as an individual object, is always given in unity (Einheit) with an absolutely non-manifest domain (nicht gegeben Mannigfaltigkeit)" (PZB 284).
Proper immanent temporality is that of lived experiences themselves. Here we reach what might rightly be perceived as the second aporia of temporality: the coexistence of permanence and change. Consciousness is an unceasing background where distinct temporal acts and events with their own duration appear . In around 1911 Husserl introduces the name of "double intentionality" (PZB 80, 379) for this articulation, since not only is there a retention (of the object event) but also a retention of retention (a reflective sawareness of that experience). These two sides of intentionalities work together and are inseparably interwoven into the unitary flow of consciousness. Consciousness could not exist apart from the acts which it intends or experiences, but it remains distinct from it, a pregnant unity of appearance and non-appearance.
We may now ask what avenues of access or exploration are available to study the immanent flow? As discussed before, one main source is the reflexive act, the becoming-aware of experience as temporal. This is quite immediate and the most convincing argument to establish that flow is an essential phenomenon. However, although accessible, immediate reflection is a task made difficult by the simple fact that to provide description we need to keep up with the shiftiness of changing experience. As we have said,shifts concerns mere fractions of a second, and even diligent observation makes it hard to provide useful distinctions. The second royal avenue of access to the flow is remembrance, if we understand that such presentification must be done in a 'pure' manner, with a view to its nature and not its specific content. I can very clearly re-live the last visual percept in the task. But this evocation is complete only when it pulls with it also the embodied context in which that image arose (my posture, the car passing in the background, the concurrent fragments of ideation as I was doing the task). In other words, although an evocation intends an object which it brings to presence in specific mode (cf. Section IV) it does so as a field: the intended object is a center, but it is also a periphery full of the context of the embodied experience. This fringe, although not intended is, nevertheless, brought to life by remembrance: "In order to examine properly (in its 'genetic' constitution) the temporality of immanent experience, it is necessary to direct one's reflexive focus during an experience of Erinnerung on the experiences that are re-produced by themselves. This is difficult" (Besnier, p. 347).
Indeed it is, but at least it is feasible enough so that we can concur that there is such a thing as the bringing into life the fringes of a memory. Where one could have found an isolated item sought by remembrance, we find, without specifically looking for it, the appended re-lived threads of the experience itself. Or if I may speak figuratively: in re-membering an intended object, it comes out bristling with retention threads of the original experience. Stated positively: what makes remembrance have such a retentional fringe must be the manner in which these very retentions have been constituted.
In our attempt to naturalize retentional acts, we attempted to solve the wooden iron nature of the past-always-present (cf. Section V). Further examination of the mode of appearance of immanent temporality confront us again with a new a new apparent paradox, not unlike that of the 'wooden iron'. The process of slippage itself (i.e. de-presentation, Entgegenwärtigung, dé-momentanéisation) has the marks of being an active or even self-generated process. Can there be a process that is a cause of itself? In this case the paradox takes the more classical form of a regressum in infinitum. Is there an angle to illuminate this second apparent contradiction?
An answer has already been sketched in passing and it now needs to be unfolded more fully. The neurodynamicas of time we have been pursuing is essentially based on non-linear coupled oscillators. As we saw, this class of dynamical systems finds its wealth of behavior in that constitutional instabilities are the norm and not a nuisance to be avoided. The case of multistability makes this quite evident experientially: the percepts flip from one to another (depending on order parameters, Fig. 3) by the very nature of the geometry of the phase space and the trajectories.
This is a widespread characterization not only applicable to this study case. Complex, non-linear, or chaotic systems in general provide a self-movement that does not depend (within a range of parameters) on where the systems are. In other words whether the content of my visual percept is a man/girl or a pyramid/ hallway, the intrinsic or immanent motion is generically the same. If the specific place in phase space is a correlate of the intentional content of an object-event, the system never dwells on it, but approaches, touches and slips away in perpetual, self-propelled motion. Cognitively this corresponds to the observation that in brain and behavior there is never a stopping or dwelling cognitive state, but permanent change punctuated by transient aggregates underlying a momentary act (1-scale of duration; Hypothesis I-III above). Formally this is expressed by the pervasive presence of stable/unstable regions, so that any slight change in initial and boundary conditions makes the system move to a nearby stable/unstable region.
This notion has been and is still explored by dynamicist under various guises. In the specific case of multistability of the Necker cube (see Section V.2), we assume the coordination of a wide array of oscillators via a common variable of phase. By introducing perspectival variants, the dynamical landscape is accordingly shifted to new dynamical modes appearing in the phase portrait. This saddle instability implies that at that point there is mixture of tendencies to be attracted into that position (i.e. perceptual content) or to move away from it, and the slightest perturbation will push and pull the trajectories along (Figure 4).
(a) The synchrony between coupled non-linear oscillators can be well followed through their phase relations ?, as contained in a function V(???that expresses space-time symmetry, bistability and experimental behavior in various cases such as perceptual switches. One may write: V(?+2?? = V(???? V(-??? -acos? - bcos2???An elegant representation plots ? against d?/dt, a vector flow parametrized by the ratio a/b. In the picture the solid and lighter lines correspond to attractive and repelling trajectories of regions over phase space. At their intersection, one finds a saddle instability on either side of the nullcline (where d?/dt=0). Adapted from Kelso (1995, p. 56).
(b) A highly schematic diagram of the current view that a complex, chaotic dynamics should be regarded as having a geometry in phase space where multiple instabilities are found locally (grays lines). A system's trajectory shifts (in black lines) constantly from one local instability to another, in an unceasing flow, under the modulation of boundary conditions and initial conditions.
(c) Experimental evidence for local saddle instabilities in the time series from the cortex of an implanted epileptic patients. The peaks of the local discharge in the temporo-occipital cortex are followed in their return map or Poincare section Tn vs Tn+1 indicated by the background points. We may investigate further the structure of the trajectories by examining individual short sequences of points around the diagonal. A typical example of recurrent observed trajectories is shown, displaying the sequence of points numbered 351-355. From points 351-354, the state of the system is drawn toward the diagonal (that is, a strictly periodic regime) but after point 354 it diverges away from the diagonal with a different linear slope. The slopes for the approach and divergence are regular and thus an invariant of the system's dynamical geometry. From Le van Quyen et al., (1997b).
More generally, many apparently "noisy'' natural events (such as the transition times between Necker cube reversala) have recently yielded unexpected deterministic patterns under nonlinear dynamical analysis beyond the reach of traditional linear analysis (cf. Appendix II). The main feature of these methods is to give us a view of dynamics not only based on trajectories, but in the more encompassing frame of the geometry of the phase space landscape.
This is not merely a formal description. These geometrical patterns can be studied even in a highly localized grouping of neuronal activity such as that measured by an electrode on the surface of the brain (a few cubic millimeters of cortex). For instance, in a local temporal epileptic focus which seems like a noisy oscillation, we have found evidence for such multiple determinism and instabilities (Le van Quyen et al., 1997 a,b). While recording from a subject's brain he was asked to perform simple visual and auditory discrimination. We then studied the interval between these discharges much like the time series from Necker percept reversal. This provided consistent indications these temporal dynamics cannot be characterized as a simple "noisy'' periodicity. Even a simple first-return map (the value of one time interval plotted against the value of the following one), reveals detailed changing geometrical patterns depending on experimental conditions, but shows consistent evidence of a saddle instability with stable and unstable manifolds. This observation suggests that the phase space landscape can be characterized by departures from strict periodicity in a non-random manner (Auerbach et al., 1987). In particular the different perceptual discriminations "pull" this local dynamics towards a distinct unstable periodicity. Although the positions of the periodic points are shifted between the behavioral conditions studied, the related slopes of approach to the instabilities appear as invariant features of the dynamics for all our experimental conditions.
In this study case we stress the relevance of local nonlinear properties in brain events which are often lost when global, averaging methods (like spectra, or even dimension estimators) are applied. There is a surprising degree of detail displayed by the trajectories or orbits of the epileptic dynamics as modulated by perceptual tasks. A paradigmatic manifestation of this is the fact that a chaotic trajectory typically includes an infinite number of unstable periodic orbits. These orbits are unstable in the sense that the smallest deviation moves the state away from the periodic orbit. Thus, a non-linear deterministic (or chaotic) system never remains long in any of these unstable motions but continually switches from one periodic motion to another, thereby giving an appearance of randomness (Artuso et al., 1990) (Fig.4).
We have gained, then, a renewed intuition to resolve the riddle of this second wooden iron of the mixture of passivity and activity, of invariance and change of double intentionality. "The self-appearance (Selbsterscheinung) of the flow does not require a second flow, rather it constitutes itself as phenomenon itself (in 'sich' selbst) " (PZB 381). Merleau-Ponty refer to this paradoxical aspect of reductive description by saying that to exist time "must be already in me (fuse en moi )", and, as it arises as a flow of retentions, it also self-manifests (Selbsterscheinung') . In fact: "I have not chosen to be born, but once born, time must already permeate me (le temps fuse a travers moi), whatever I do" Husserl develops his descriptive account of this paradoxical appearance of 'double intentionality' in the notions of transversal and longitudinal intentionality (Quer- and Längsintentionalität). The first is retentional dynamics, the static constitution. Longitudinal intentionality, in contrast, is the genetic constitution of the temporalization of experiences themselves, their self-manifestation. These are necessarily interdependent, but their mode of dependence and the root of their difference is what is difficult to express : "For all of that we have no names." (PZB 371). Now, longitudinal intentionality acts by an integration from within the now itself, it provides an unchanging substrate from which the flow emerges. As Besnier remarks (p.350), there is a great temptation to transpose the analysis of perceptual intentionality by envisaging this as a "pure" substrate or Ur-hyle. I do not have to enter into the thorny technical debate this mode of analysis has produced since Husserl (Depraz, 1996). My contribution is to bring to the fore the intuition derived from the generic non-linear flows, following our Hypotheses. Self-manifestation appears in our analysis as self-motion or generic instability, which is not a mere artifact of description, but an invariant formal description for self-organization. Thus its relevance to temporality is appropiate. The flow, in the neurodynamical sense, is precisely a wooden iron that exists as flow only to the extent that it is constituted in individual trajectories (not an inert geometrical magma) as ongoing self-propelled transient trajectories visit various regions in phase space (corresponding to an intended object-events, an appearance). The inseparability of these two intentionalities here is not only descriptively accurate, but part of the intrinsic logic of complex non-linear dynamics. It would be inconsistent to qualify the self-motion as a "deeper layer" of the dynamical process, and to describe trajectories as mere appearance (Gallagher, 1979). Mutatis mutandi, it seems illusory to isolate a "deeper" layer of genetic constitution where experience would be constituted from an absolute time, and only then made manifest in consciousness intentionality. What is deep is the link between self-motion (immanence) and trajectories (appearance).Enough has been said about the immanent or absolute flow to suggest its importance and perspicacity. It is surely a topic that needs exploration, and it is the natural ground for bridges into other varieties of experience and towards other traditions concerned with human consciousness . Indeed, this level of analysis touches more than any other on the ground of self, pure ego, or basic consciousness. Brough (1989) summarizes:"And thanks to the infinite horizon opened up by the absolute flow, we can be sure that we can go right on changing and accumulating a past, while still remaining the same. There is a fissure, then, in consciousness. thanks to this fissure, any one of my acts, asserting itself and holding sway for a more or less brief time...is able to 'slide off' without taking my whole self with it. If my internal awareness were glued without gap to my fleeting experiences, the passage of time would rip my ego to shreds" (p.288).
But we must now leave these considerations and turn to our last topic which is the closely related issue of the appearance of this self-motion from the perspective of affect.
Return again to the first task and re-examine more closely the nature of the switching, as it happens. One essential component of the experience is that the shift is sudden and accompanied by a (more or less distinct) emotional change when the visual perception shifts abruptly. Thus, the mode of slippage of nowness into just-past, the retentional trajectory, appears as presence of the past not only in a way that is distinct from representational memory. It also gives us the cue that the emotional tone is an integral part of the phenomenon. What is the role of emotion or affect in the self-movement of the flow? And what is its role, if any, in the anticipation of what is to come, protention?
In Husserl's published texts, protention is not extensively analyzed and I have the impression that he implicitly assumes a certain symmetry with retention, as if the same structure of invariance for the past could be flipped towards the future. But protention intends the new prior to an impression and thus can only be a pre-figuration. Husserl speaks of " empty constitution" (PZB 52), but it is not expectation or anticipation in the sense of containing a representation of what the next now will bear. To see why this is so, one need only to apply the same arguments as used for distinguishing retention and evocative memory. Indeed protention has a mode of openness: "...the only thing definite is that without exception something will come"; in listening to a melody (his example) there is a predictable side to protention since it intends further phrases of the music (PZB 106, 84). These and similar remark from the Lessons seem to have provided the basis for the symmetrical view of the three-part structure of time, and as indicated in the oft-used figure of time discussed above (Figure 2).
This analysis can be substantially enriched. There is at least two main sources of evidence to conclude that protention is generically not symmetrical to retention. The first is, precisely, that the new is always suffused with affect and emotional tone that accompanies the flow. In fact, protention is not a kind of expectation that we can understand as "predictable", but an openness which is capable of self-movement, an indeterminate that is about to manifest. In this quality it provides the natural link into affection, or, more aptly with some form of self-affectedness (see below). The second is that retention has the structure of a continuum, but portention can only be a bounded domain, since we cannot anticipate and anticipation that is yet to come. While the threads of retention set the stage for protention, it cannot modify the retentional threads retroactively .Time and affect were never systematically treated by Husserl. A substantial part of his notes are still unpublished or are accessible only from secondary sources (see Appendix I). However, a deepening of late analysis of time makes it possible to trace some important fragments of a view of affection as initiating the drive of the lived flow itself, as Depraz (1993) points out, (linking it to the parallel question of the constitution of space , p. 70). Thus my propositions in the remainder of this Section are not entirely at variant with Husserl's later research and thus in continuity with what has been already developed here. As he says:"How is the self (Ich) the center of this life it experiences? How is it experienced? It is affected by that which consciousness is conscious of (Buwusstsein bewusst ist), it follows affect, or still it is attracted, held, taken in by that which affects it» (Ms. C III/1).
Husserl notes contain recurrent references to this primordial aspect with regards to the child's early life where one finds an "instinctive" intentionality (Triebsintentionalität). As Depraz remarks:
"Affect is thus this non-form which makes constitution of the self by itself, that affects it in the strict sense of structure, that of constitutive temporality...Affect is there before being there for me in full consciousness: I am affected before knowing that I am affected..It is in that sense that affect can be said to be primordial" (p.73, 75).
How is this pertinent for the three-part structure of time? Husserl remarks that during a melody the sounds affect me differently as it creates its retentional threads, an attentional tendency (eine Tendenz der Zwendung). Or we may say it provides a disposition which is marked by a gradual intensities. This temporalizing effect puts protention at the center stage: ".. it is not only the impressions from hyle that affect, but already 'hyletical anticipations of data'." (Depraz, p.79).
Time as a royal road to the study of affect continued after Husserl, first by Heidegger and Merleau-Ponty, where the discussion about self-affectedness is considerably enriched. The main innovation in their philosophies is the treatment of time as self-affectedness. As Merleau-Ponty says: "Le temps est 'affection de soi par soi'", and notes that the expression derives from Kant as modified by Heidegger in his Kantbuch . Self-affectedness becomes a key insight into the nature of consciousness: "...even the most precise consciousness of which we are capable is affected by itself or given to itself. The very word consciousness has no meaning apart from this duality" (p. 488)
With Levinas (1988), a further sphere of affection, hetero-affection, is brought to the fore: an Other, alterity is the primary clue for time's constitution. We are not only affected by representations and immanent affection ( "affection de soi par soi"), but alterity as inseparable from the sphere of an ego-self. In this move the very distinction between auto- and hetero- ceases to be relevant, since in all cases it all is brought down to the same manifestation: it is a question of 'something other', the experience of an alterity, a difference in the identity of the present, whether by the inevitable slippages to retention, or by the anticipations in protention.
But these philosophical discussions are not central for us here in all their detail (Depraz, 1997). We are seeking to move beyond the apparent paradox between an original impression in time that would be colored by affection, or, conversely, the primacy of affection that would underlie temporality. We seek a non-dual synthesis whereby affect is both constitutive of the self, and the same time contains a radical openness or unexpectedness with regards to its occurring.
In order to move further in our analysis we need to have a concrete base on which to base our examination, much as the visual tasks provided the basis for the static analysis. Many alternatives exist, but I have chosen to explore the role of affection in the constitution of time in the context of active involvement in the world and through the dispositional quality of affect and its gradations. From the point of view of enactive cognitive neuroscience coping plays a central role (cf. Section II.2). It should be kept in ind, however, that focussing on this region of affect is by no means exhaustive.
From the perspective of phenomenology, we are referring to the phenomenology of immediate coping which finds its first expression in the study of affectedness (Befindlichkeit) as one of the constitutive moments of Dasein that Heidegger described in Sein und Zeit (Division I, especially Chapter III). As it is well known, Heidegger studied most particularly our involvment with the surrounding network of tools and equipment, what kind of being humans might be that such encountering are possible.
I am not going to re-discuss this key points of the constitution of Dasein, but examine more closely the observation that both equipment and the user show up as transparent in immediate coping. It is at this point that the Heideggerian analysis is much more helpful than Husserl's. Heidegger calls this stance Umsicht, a kind of seeing that is not reflective or thematized in any way . How is the user's transparency shifted to a conscious mode that brings nowness to the fore? As he says:"Being in the world, according to our interpretation hitherto, amounts to a nonthematic, absorbed transparency (unthematische, umsichtige Aufgehen), in activities that are constitutive for the network of equipment at hand. Our concerns (Besorgen) appear the way they do because of the familiarity with the world. In this familiarity Dasein can lose itself in what it encounters within the world...[H]ow can the worldliness of this familiarity be illuminated (erleuchten)? How can entities at hand be thrust to the fore by the possible breaks in the network of equipment being handled in which transparency 'operates' (die Umsicht sich 'bewegt')?" Heidegger points to how intentionality, in its traditional sense, can make irruption within a flow of action. If, as I write this, I hit a control key, and I am shown a message saying "Do you really wish to erase this text?", I find myself deliberately avoiding pressing the 'OK' button, in an emotional tone of hope and tension. The awareness of a fatal mistake breaks into the present trigerring a (more or less marked) shattering of transparency. In prallel a new stance in ongoing coping emerges: I deliberately click on the 'Cancel' button. Thus, Dasein is absorbed in the world, but malfunctions and breakdowns of various kinds can shift him out of it. I use the word transparency here to indicate precisely this unreflective absorption. Transparency is but a description of that which can be broken in the flow of experience, whence its interest for our purposes here . When the carpenter hammers the hammering action and its objects are transparent. Transparency is what it is because it is eventually broken, as when the hammer slips and lands on the finger. Such breakdown brings the transparent equipment into view and a new set of action-assessments begins. This standard Heideggerian vignette can be extended to all embodied actions, that is, actions in a fluid context where there is always a mixture of immediate coping and concurrent secondary activities of language and mental life. It is a disposition for involvement. Transparency is therefore always in the context of a series of ongoing action-behaviors. It is related neither to content nor to efficacy (cf. my faulty typing, or the carpenter's mistake). However transparency is not obligatory since an attentional tendency can produce its disappearance at various degrees of abruptness. In contrast to breakdown in equipment, attention can also be endogenously motivated in a reflective gesture motivated by a critical review and design. The carpenter may stop to evaluate the quality of the wood before using it, or his relationship to the employer. Notice, in passing, that the gesture of carrying out phenomenological reduction is a loss of transparency by self-motivation. This can be seen as the form reduction takes in the Heideggerean lineage of phenomenology. Transparency, then, is a readiness or dispositional tendency for action in a larger field of a specific ontological readiness, that is, an expectation about the way things in general will show up. For this very same reason transparency has everything to do with habitus, the recurrence of our lives. Learning a skill is a prototypical example of transparency acquisition. However, the temporal constitution of transparency is not limited to individual actions: it extends to the full span of human life, into history altogether. In the Heideggerian jargon, transparency and its dispositions extend into historicality .The loss of transparency is never distant from a dispositional affective-tone, as we have seen. But we can now see that different degrees of breakdown in transparency and the mutiple manners in which it happen opens a panoply of affective tonalities: fear, jealousy, anger, anxiety, self-assurance and so on. Accordingly, the word emotions is used here in tis specific sense: the tonality of the affect that accompanies a shift in transparency. Affect, on the other hand is a broader the dispositional orientation which will pre-condition the emotional tone that may appear. This is quite essential.As I write now, I have a dispositional attitude that engages me in a anticipation of writing and shaping my thought into sentences. As I write this word now, the disposition is colored by an emotional charge, a moderate resentment for not finding the proper expression. But that emotional tone appears against a background of the exalted mood of a productive day devoted to finishing this text. More explicitly, I want to distinguish three scales for affect, homologous (but not isomorphic ) to the three scales of temporality used above. (1) The fist scale is emotions: the awareness of a tonal shift that is constitutive of the living present. (2) The second is affect, a dispositional trend proper to a coherent sequence of embodied actions. (3) Finally mood, the scale of narrative description over more or less long duration.
Examined from this perspective, emotions cannot be separated from recurrent constitution since its transparency is not deliberate, but part of unthematic coping. The embarrassment that accompanies a loud burp in a French restaurant is entirely transparent to a european; it seems the 'natural' reaction to have. This un-reflective appereance of emabrrasement as affective tone is only perceived when, say in a trip to the Middle East. The entire social phenomenon of eating and manners is cultural-historical, but this does not contradict the visible signs of bodily emotions that go with it-- redness of the face , change of breathing and tightness of facial muscles.
For the ethologist, affect and emotions are a relatively small repertoire of immediate dispositions that are physiologically inscribed in a species inheritance, although in most mammals habit and sustained learning may shape it significantly . Neurobiologically, can be associated to a relatively stable set of neural correlates (e.g.. Damasio, 1995). Studies of human emotional responses, even in relatively artificial situations, reveal the extent to which the biological endowment of 'basic' emotional patterns is enfolded in the historical recurrence of an individual, its historicity and language. Individual habits, historicity and language, constitute the palette of human emotional life incorporating the biological make-up to an end which is but which is historically and individually unique .Homologously, we can say that the experience of time has a biological base in elementary events (1/10-scale), but this basis is enfolded with other structures of temporalization into the specious present that is our theme. To deny that such a deeply rooted biological basis plays a role in the appearance of temporality is fruitless. Similarly, I am not reducing emotions to their empirical correlate in a reductionistic move. As considered here, emotions are an integral part of an ontological readiness. However, this should not obscure the fact that such an ontological constitution has roots in basic emotional dispositions inseparable from our history as living being and minute events in brain physiology. When I break induce a brake transparency following reduction when looking at the visual image of our task, I bring to it an emotional disposition which pre-figures the change in my perception. In saying 'I expect to see' I also provide exogenous, additional order parameters that alter the geometry of the phase space. This process of "sculpting" a dynamical landscape is intrinsically distinct from the trajectories that move within it, but form an inseparable unity. In fact, it has been known for some time that the intention to carry out a movement is coupled with a change in emotional tone that varies in degree. As a global variable, induced changes in dynamical landscape can be detected. One well known case is the readiness potential. For a finger movement, a large slow electrical potential can be measure over the entire scalp which precedes by a fraction of a second the beginning of the motion, and the subject can reports that he has decided to initiate the movement . This is not a correlate of intention (as it is sometimes said), but it does give a concrete idea of how vast a reconfiguration of a dynamical landscape is involved at the origin of a fully constituted now (moving the fingers). In the results of Leopold and Logothetis (1996) already mentioned, a similar reconfiguration of the disposition for firing of individual neuronal responses is visible some 100-200 msecs before the monkey indicates that it has switched to a new percept. Why is all this relevant here? Because it is direct evidence of the manner in which emotional tonality plays into the dynamics of flow. Emotional tonality is, by its very action, a major boundary and initial condition for neurodynamics. This diffuse, constitutive effect is in accord with the mechanism of action via neurotransmitters that have been know for some time to condition the modes of response at the neuronal level, as the body of knowledge of psycho-pharmacological agents attests. This sketch of the nature of protentions via affective-tonality has taken us to a third and final step of what seems to be formally a genetic constitution of temporality. I have introduced a last dynamical principle that applies to neurocognitive dynamics as well. I refer to the mutual bootstrap between the phase space landscape and the specific trajectories that move in it, and the fact that the very same trajectories provide the very conditions for an embodied coupling, since through their coupling they shape their dynamical landscape. Metaphorically, the walker and the path are intrinsically linked. This bootstrap principle seems to be present in a variety of natural systems, and has recently been referred to as 'operating at the edge of chaos', or 'self-organized criticality'. For example, this idea provides a renewed view of evolution since it provide an answer to the old nature (genetic expression) vs. nurture (environmental coupling conditions) dilemma. In this synthetic view (cf. Kauffman, 1993) the relation between natural forms, (the Baupl┼E4Đne of organisms), and the selection process in their ecological embeddedness, is not one of contradiction but precisely a mutual imbrication when seen through dynamical glasses . This built-in shiftiness, enfolding trajectories and geometry, gives a natural system a possibility of always staying close to regions in phase space that have multiple resources (e.g. at least two in bistable visual perception). To conclude: The generic structure of double intentionality proposed by Husserl is, I submit, of this class of dynamical boorstrap, and the analysis of affect and emotional tonality can give the evidence for it.
To gather all the threads that have been developed here, and to echo the tradition started by Husserl himself, I would like to propose a new figure of time, the fourfold structure of nowness (Fig.5).
This is not so bold or farfetched. By now it seems clear that the point-by-point, linear time depiction at the base of the figures of time is insufficient. One major improvement is to introduce, then, not lines but flows, dynamical trends. A second major improvement is to take into explicit account what surfaced in the later work of Husserl himself, the central role of double intentionality, static and genetic constitution. This final ingredient gives to the homologies between the constitution of space and time the preeminence they deserve. I do this by taking the center/fringe structure as the very core of a new figure of time. Once these three basic aspects have been incorporated a new representation falls into place quite naturally.
1. To start let us consider the role of dynamics that has been central in our development, and hence move away from the use of discrete points in a line as in the traditional diagram. Not only do we leave behind a line geometrical figure, but we must introduce an asymmetry into it. With regards to static constitution, we have discussed two different kinds of dynamical ideas: retentional trajectories (the past) and order parameters for anticipation (the future). In the new diagram the dynamical quality is displayed by arrows on lines, as is traditional in mathematics, but I distinguish trajectories from anticipatory landscaping by dropping the arrowheads in Fig.5-I. With regards to genetic constitution, we also have two distinct ideas: the immanent temporalization of self-motion, and the directed intentionality relative to a position in phase space.
2. Next we consider the spatial ingredients, that is, the role of a center-periphery configuration at the core of temporalization. With regard to static constitution, we of course recover the retention-protention axis. Again, these are asymmetrical, since retentions recede into past, but the protentional fringe is an open horizon of anticipation. Thus we again find that the center-periphery structure does not provide all the necessary distinctions, since the fringes as they move away from center, become qualitatively different. As to genetic constitution, the fringe re-appears in the pre-conscious, affective substrate (permanence) on one direction, and the conscious, embodied ego, aware of emotional change on the other (change).
We thus arrive at ingredients that come in two sets of four, a suggestive fourfold. The center/periphery configurations are analogous in the spatial and dynamical schemes, since they underscore the paradoxical nature of past in the present, and of change within permanence that puzzled Husserl (and many others) throughout his work. These pairs are, then, the wooden irons of the figure of time. I do not presume to have resolve them, only to have given new insights into them.
Fundamentally, the added insight comes from the fact that the component ingredients have a generic link between them, an internal interdependence that has been explored throughout this paper. In other words, the new figures of time are not only graphical combinations of items, but they display effective links which are not only descriptive. Phenomenologically, I have stressed the full interdependence of both intentionalities, the inseparability of the static and genetic analysis, and the mutual determination of instinctive and cognitive constitution of self. In parallel, trajectories and the landscape of their phase space are a unity in a complex non-linear system. Correspondingly, we have examined the many aspects under which determinism, trajectories, regions in phase space, adaptive geometrical landscapes are complementary. I consider these mutual interdependencies and their role in the constitution of temporality the most immediate insight that naturalization can provide.
With these elements in position, the figure of time, as I said, falls naturally into place. Figure 5 places these ingredients in relation to each other, in regards to both the role of dynamics and space, and the circular causality for the longitudinal but not for the transversal intentionality. Since the order parameters for protention do not influence retroactively the system's history, we arrive to an asymmetrical four-fold figure of time . We are now in a position to stand back, and re-consider our analysis of temporality. Has neurophenomenology passed the acid test? Let us return to the neurophenomenological research program: the circulation between the external and the experiential. In Varela (1996) the neurophenomenological working hypothesis was stated as follows:Phenomenological accounts of the structure of experience and their counter parts in cognitive science relate to each through reciprocal constraints. On the one hand, we are concerned with a process of external emergence with well defined neurobiological attributes; on the other, with a phenomenological description which stays close to our lived experience. The nature of the sought-after circulation one seeks is no less that of mutual constrains between both accounts, including both the potential bridges and contradictions between them. What are is the specific nature of the passages between these two accounts in the case at hand? What have we learned in the specific case of immediate temporality?One thing is clear: the specific nature of the mutual constraints is far from a simple empirical correspondence or a categorical isomorphism. Three ingredients have turned out to play an equally important role: (1) the neuro-biological basis, (2) the formal descriptive tools mostly derived from non-linear dynamics, and (3) the nature of lived temporal experience studied under reduction. What needs to be examined carefully is the way in which these three ingredients are braided together in a constitutive manner. What we find is much more than a juxtaposition of items side by side. It is an active link, where effects of constraint and modification can circulate effectively, modifying both partners in a fruitful complementary. A proper analysis of this complementary cannot be done here, and I must close by defer ring the reader elsewhere for its presentation (Varela, 1997). But at least it could be said that the neurophenomenological research program emerges seen from this study beyond a hopeful declaration, as an open road for exploration.
This paper is the result of many months of gestation in the productive atmosphere provided by several colleagues in Paris in the form of two working groups on Phenomenology and Cognitive Science, and Phenomenological Psychology who have generously provided teachings, ideas and encouragement. My special thanks go to Natalie Depraz, Bernard Pachoud, Jean Petitot, Jean-Michel Roy, and Pierre Vermersch. The work carried on with my research group on neurodynamics has provided me much of the empirical substrate for this discussion; I am specially grateful to Jacques Martinerie and Michel Le Van Quyen. The joint work with Evan Thompson has continued to provide insights and resources at every stage. Finally, the incisive commentaries to the text by Amy Cohen-Varela and Shaun Gallagher were very helpful. To all of these friends my heartfelt thanks; I am fully responsible for the remaining errors and shortcomings.
Husserl's work on time has been a difficult and active field of investigation for phenomenologists for half a century. Since the avatars of the publication of Husserl's work and the secondary literature it is of some significance for us here, I will briefly outline it. The saga unfolds in four acts.
In 1917 Husserl requested his assistant Edith Stein to prepare his work on temporality for publication. He provided her with texts from his 1905 seminar, but included some from as far back as 1901 and as late as 1917. Stein carried out her task during 1917, putting aside more than half of the sources and rewriting the texts extensively, including titles and section headings. She returned her work back to her teacher in September 1917. But, as was often the case with Husserl's requests to his assistants, he then decided to introduce new, significant revisions and postponed its publication. It was only in 1927 after discussing with Heidegger the imminent publication of Sein und Zeit, that Husserl apparently went back to Stein's work and sent it to Heidegger. Although Heidegger added only minor changes, the book appeared under his nominal editorship with the title: Vorlesungen über das innere Zeitbewusstsein aus dem Jahre 1905 (Halle, 1928) (The English translation by J.S.Churchill: The Phenomenology of Internal Time Consciousness, appeared in 1966) . For modern presentations based on this early material see MacInerney (1991) and Miller (1984).
The Lessons can adequately be described as Stein's very personal rendition of the original texts. Shortly after publication, several readers realized to what extent her influence had marked the only available material of Husserl's research. The French translator of the Lessons, Henri Dussort, had already raised this issue but could not complete a new edition closer to the originals. In 1966 Rudolf Boehm produced a critical edition, available as Husserlianna X, Zur Phänomenologie des inneren Zeitbewusstsein (1983-1917) (abreviated here as PZB). Boehm's work provided the public with a wider selection of writings, a careful chronological analysis of the 1928 addition, and several unpublished appendices and sketches. Rudolf Bernet has also provided some comments on the complementary texts to the Boehm edition (Barnet, 1985), with some new indications about dating and an Introduction. This edition provided for the first time a less constrained reading of his work, and it has been the main source of virtually all published work since. It is also the main source for this article, along with the excellent introduction by J.Brough (1989) who has provided an English translation of PZB (Kluwer, 1991).
Interestingly, Husserl wrote a significant amount of work during 1917-1918 in Bernau. Sometime later ( in 1928?), he requested his assistant Eugen Fink to work on them. Fink embarked in what to become a long adventure (1928-1935) of revisions of these Bernauermanuskripts, a saga that can now be traced in detail thanks to the meticulous work of R. Bruzina (1993, 1994), and constitutes in itself a rich source of information about the research carried out jointly by Husserl and Fink. Fink kept these manuscripts until his death, and they were not available at Louvain for Boehm's edition in 1966. A criticial edition is not yet available (but is in preparation by I.Kern and D.Lohmar); some texts also appear in Analysen zur passiven Synthesis (1918-1926), Husserliana XI, 1996 . 4. The "späten Zeitmanuskripten" (1929-1935)
The so-called late manuscripts are the last addition to the sagawe are tracing. They comprise Husserl's last corpus of work on time, space and intersubjectivity. They are as yet entirely unpublished, and corresponds to Group C of manuscripts in Husserl's Archives in Louvain. Nevertheless important extracts have been made available in books by G.Brand (1955) and K.Held (1966). For this paper, I have been relied on N.Depraz's (1994) discussion on temporality and affect in these late manuscripts.
This is a bird's-eye view of non-linear dynamics from the point of view of its relevance to neuroscience. It is intended only as a complement for the readers of this papers for whom this is an entirely foreign domain of modern science. For further reading Port and van Gelder (1995) and Arbib (1994) are excellent recent sources; Ott et al. (1993) is more technical but of great value.
Probably the first to see the difficulties and possibilities of chaotic dynamical systems was Henri Poincaré at the turn of the century when he was working on the problem of three celestial bodies under mutual gravitational attraction. Poincaré analyzed orbits resulting from sets of initial points and found very complicated behavior, which we would call chaos today. Despite theoretical work on chaotic dynamical systems by several mathematicians, (like S. Smale or the Soviet school of A.N.Kolmogorov in the 50s), it took a long time for researchers in other fields to recognize the "reality'' of chaos in nature and to turn their attention to the strange behavior of some non-linear dynamical systems. Milestones on the way towards public recognition of the new phenomena in the 70s were articles by N. Lorenz (on the weather), R.May (on animal populations), or O.Rössler (on chemical reactions) who announced their findings with titles like "Simple Mathematical Models with Very Complicated Dynamics" . The accent was on the fact that these systems produced irregular and almost random looking behavior though they are perfectly deterministic. Immediately the question arose, how many of the observed irregularities found in nature were due to such non-linear motion, that is, follow deterministic rules despite their random appearance.
Non-linear dynamics (=chaos theory = complexity theory) is a recent field of research rooted in mathematics but also in physics and other natural sciences. Rigorous definitions have only evolved with the discovery of more and more surprising dynamical effects.
It is perhaps is best to start with the notion of a state or phase space : a domain of variables or measurements which attempts to completely specify a given process. Such specification is a law or a rule, and these system are therefore deterministic, in contrast to a random dynamical systems. The sequence of subsequent states evolving according to the dynamical rule describes a trajectory in state space. In the case of continuous time, the system is defined as a flow,
Smooth flows can be differentiated with respect to time, giving a differential equation
The number of first order differential equations, or equivalently the dimension of the state space, is referred to as the order of the system, sometime as its number of degrees of freedom. A periodic point is one to which the flow or map returns after some time. Experimentalists are mostly interested in dynamical systems defined on a state space ( n real-valued measures) or subsets thereof, since these state spaces have been proven useful for the description of natural phenomena. We distinguish finite dimensional systems, which may be defined as maps or ordinary differential equations and infinite dimensional systems, defined for example by partial differential equations or time-delay differential equations.
We speak of linear dynamical systems if the function F is linear , that is, if
holds. Stated in ordinary language: for linear systems the whole is the sum of its parts. Solutions which move away from the origin exponentially in time are called unstable, while solutions converging exponentially toward the origin are called stable.
All dynamical systems which are not linear, are called non-linear. This name by exclusion might seem surprising. As S. Ulam is said to have remarked: "Calling a science 'non-linear' is like calling zoology 'the study of non-human animals' '' Yet the distinction has been necessary in the natural sciences, which is so accustomed to linear systems. The discovery of new forms of dynamical behavior unrestricted by linearity falls is one of the greatest achievements of this century.
There is not yet a universally accepted definition of chaos. Typical characterizations of chaos are based on low-dimensional (i.e. not higher than 5-6), bounded dynamics with sensitive dependence on initial conditions. The bounded region where the dynamics turns may not be decomposable into separate parts. Note that the first two parts of this description are common. Linear systems with stable dynamics remain bounded in phase space; linear systems with unstable dynamics show sensitivity to initial conditions. What makes a system chaotic is the combination of the two. This combination can only be obtained in non-linear systems.
The mathematically rigorous definition of chaotic behavior in deterministic systems still raises problems, particularly since not all kinds of behavior which would intuitively be called ``chaotic'' are known yet. For example, the older definition of chaos requiring a dissipative system with a strange attractor is out of date. As more and more of the unexpected behavior of non-linear dynamical systems is found, the mathematical concepts eventually will converge. In fact, the science of non-linearity resembles a sea of ignorance with some small islands where results are known and applicable, as shown in the sketch below (which I owe to Thomas Schereiber).
Apart from the basic ideas sketched above, in the study of neural systems some tentative ideas, (some indicated in the chart below), have proven quite important for neuroscience, and are used in the present study as well. In fact, when considering a large ensemble of complex interacting components such as the brain, one is quickly confronted with the issue of self-producing pattern formation of self-organization. These are distinguished, salient states of motions in phase space, which arise from the reciprocal or collective cooperation between components, quite independently from outside inputs. These patterns can be modeled by the same tools as other dynamical systems, and parametrized by order parameters (such as initial and boundary conditions). An explicit example is found in Appendix III, and a less formal description in the way in which visual multistability was presented above. These large collective ensembles are typically quite unstable, as we saw before (cf. Fig.4). This instability leads to more or less frequent shifts from one pattern to another, and these transitions are a topic of intense interest today. These are referred to as bifurcations (from a given state), or as phase transitions.
The use of dynamical tools in cognitive neuroscience has come from several relatively independent lineages of research.
The early tradition of Gestalt psychology founded by Kohler and Koffka is explicitly dynamical although they did not couch their results in that language. But their influence has echoes even today. One of the influential representative is R.Thom with his neo-gestaltist ideas issued from his work in bifurcations, and elaboraed extensively by J.Petitot.
A second lineage comes from the interest of physiologists working in motion and behavior. In the early' 50s in Russia A.Bernstein produced the first fully dynamical account of coordination on neural terms, and at the same time in Germany, E.von Holst provided pithy detailed examples that were amongst the first to use phase portraits and instabilities. This tradition is still active today in the work of J.J.Gibson and J.S.Kelso for example.
A third trend issued from the early cybernetics movement in the 50s in the USA, especially from the influential work of W.Mculloch, A. Rosenblatt (who is the pioneer of modern connectionist views), and H. von Foerster. From this lineage we owe the inspiration of W.Freeman who was the first to explicitly introduce the notion of ensemble synchrony and phase coherence as a key mechanism in neuro-cognitive pattern formation.
To make my ideas quite precise, it is important to work on the basis of an explicit dynamical model for the emergence of synchronous neuronal assemblies. The model introduced here focuses on a subset of middle-layer pyramidal neurons known to play a role in cortical coherences: the intrinsically bursting cells (IB) . The IBs are distinguished by their ability to generate clustered bursts of several spikes riding upon a slow depolarization wave, followed by a large hyperpolarization. Since the IBs possess an extensive tree of divergent axons extending far into the horizontal dimension of layers IV-V, they are likely to contribute to an excitatory interconnected network responsible for the organization of large numbers of neurons into macroscopic space-time patterns of transiently coherent activity. Furthermore, many of the cells in layer V can generate rhythmic bursts in the range of 4-10 Hz and are believed to play a significant role in cortical rhythmogenesis. Relatively little attention has been paid to the possible significance of such slow bursting components for temporal binding. We suggest that populations of interacting bursting neurons with slow rhythms, such as IBs, are a natural candidate for the generation of synchronous assemblies. Our analyses of synchrony among a network model of intrinsically bursting neurons demonstrate that cellular discharges can aggregate into separate phases and display regular time lags between their rhythmic spike trains. Global feedback inhibitions associated with the tonic depolarization of the coactivated bursts play a leading role in the phase aggregation by the sculpturing of the activities with hyperpolarization. If our conjecture is relevant, various brain oscillations (including the gamma-range, 30-80 Hz) may be considered as time sequences of more elementary components, as discussed in Section II.2. Two basic questions are raised by our view. First, how could the slow rhythmic bursting neurons become mutually synchronized fast enough, and, second, how could coexisting synchronous aggregates be distinguished or labeled for further integrative processes? IB cells are characterized by two distinct times scales of the membrane potential: a slow dynamic corresponding to the smoother phases (bursting and the quiescent repolarization), and a fast one corresponding to the sharp transition between them. Generically such behavior can be characterized as relaxation oscillators. The Fitzhugh-Nagumo equations (FN) are a representative of this class, and provide a mathematical formulation of IB intrinsic rhythmicity on the basis of ionic gatings. Basically, these two-dimensional equations describe the repetitive interplay between fast and slow time scales in biological membranes: where V denotes the mean memebrane potential, u(t) is a slow component of the membrane currents, I(t) represents the sum of synaptic and external currents entering the cell, ? is a time constant specifying the scale difference between fast and slow time scales, and is a non-linear function which mimics the N-shaped relationship between the fast membrane currents and the cell potential V. As shown by the phase-plane (V,u) in Fig. AIII-1, V increases during the upper solution of u (depolarization) and decreases during the lower stable solution (hyperpolarization). When either end of the bistable region is reached, V abruptly jumps to the other branch, the transitional points being given by the nullclines (indicated by the thin lines in Fig.AIII-1).
A realistic feature of this model is that when the oscillator is perturbed by an arriving spike, the oscillation may be rapidly re-set via the fast jump onto a new phase, depending on the moment of arrival of the perturbation. A phase resetting (shown in the Figure by a dotted line in both time and phase plane) can be induced by briefly making I negative during the depolarization: the key consequence of this inhibitory pulse is to displace the nullcline (dashed line) in the space (V,u) of the oscillator, and therefore modify the position of its transitional regions, leading to a premature hyperpolarization. Because of their sudden-transition possibilities and because the rate of change of the slow variable u(t) before the sudden transition is less than after the jump, a spike shared by two oscillators can readily induce a quick phase-locking of their activities (Somers and Kopell, 1993). The Figure shows that the new phase difference after a brief inhibitory current (1 msec) from a common source, applied to two (noncoupled) oscillators, is strongly dependent on the moment of the cycle during which it arrives.
In brief, this dynamical behavior means that neuronal relaxation oscillators (such as cortical IBs) naturally afford a mechanism for synchronization as a consequence of various excitatory and inhibitory feedbacks mediated by recurrent local circuitry. In fact, fast GABA-mediated inhibitory interneurons are known to play an important role in the organization of local coherence in the neocortex. This synchronization of ensembles is validated by our simulations concerning populations of IB-like neurons randomly connected with varying degrees of strength in connectivity, and all under a common GABA-like common inhibitory control. As the bottom simulation shows, increased comon inhibition increases the degree of spatial synchronicity, that is otherwise patchtly distributed over the array of simulated neurons. Recent results from hippocampal slices confirm this idea quite elegantly (Traub et al., 1996).
Arbib, M.A. (1995), Handbook of Brain Theory and Neural Networks, MIT Press, Cambridge.
Amitt, D. (1991), Attractor Neural Networks, Cambridge UP, Cambridge
Artuso R, Aurell E, Cvitanovic P. (1990), Recycling of strange sets. I. Cycle expansions, Nonlinearity 3: 325-359.
Auerbach D, Cvitanovic P, Eckmann JP, Gunaratne G, Procaccia I. (1987), Exploring chaotic motion through periodic orbits, Phys. Rev. Lett. 58: 2387-1380.
Bernet, R. (1985), Texte zur Phänomenologie des Inneren Zeitbewusstseins: Texte nach Husserliana X,, Felix Meiner, Hamburg.
Bernet, R. (1994), La Vie du Sujet, Presses Universitaire de France, Paris.
Besnier, B. (1994), Remarques sur les Leçons sur la Conscience Intime du Temps de Husserl, Alter No. 1, p. 319-356.
Brand, G. (1955), Welt, Ich und Zeit, M.Nijhoff, The Hague.
Brough, J. (1989), Husserl's phenomenology of time consciousness, in: J.N.Mohanty and W.R.McKenna (Eds.) Husserl's Phenomenology: A textbook, Univ. Press of America, Washington DC.
Bruzina, R. (1993), The revision of the Bernau time-consciousness manuscripts: Status questionis - Freiburg 1928-1930, Alter No.1, pp.357-383.
Bruzina, R. (1994), The revision of the Bernau time-consciosuness manuscripts: New questions - 1930-1933, Alter No.2, pp.367-395.
Church, R.M. and H.Broadbent (1990), Alternative representation of time, number and rate, Cognition 37: 55-81.
Damasio, A. (1994), Descartes' Error, Grossman-Putnam, New York
Deecke, L, P.Scheid, and H.Kornhuber (1969), Distribution of readiness potentials preceding voluntary finger movements, Exp.Brain. Res. 7: 158-168.
Dennett,D. (1991), Consciousness Explained, Little Brown, Boston.
Dennett, D. and M.Kinsbourne (1991), Time and the observer: The where and when of time in the brain, Beh.Brain Sciences 15:183-247.
Depraz, N. (1992), Les figures de l'intersubjectivité. Etudes de Husserliana XIII-XV, Arch.Philosophie 55:479-498.
Depraz, N. (1994), Temporalité et affection dans les manuscrits tardifs sur la temporalité (1929-1935) de Husserl, Alter No.2, pp.63-86.
Depraz, N. (1995), Incarnation et Transcendence, J.Vrin, Paris.
Depraz, N. (1996), Qu'est-ce que la hylè transcendentale?, Analecta Husserliana vol.L, pp.115-123.
Depraz, N. (1997), Can I anticipate myself?, D. Zahavi (Ed.), ???, Kluwer, Dordrecht (in press).
Depraz, N. Varela,F. and Vermersch,P. (1997), On Becoming Aware: Steps to a phenomenological pragmatics (forthcoming).
der Waals, F. (1995), Good Natured, Harvard Univ. Press, Cambridge.
Dreyfus, H. (1991), Being in the World, MIT Press, Cambridge
Ekman, P. and R.Davidson (1994) (eds.), The Structure of Emotions, Oxford Univ. Press, New York.
Fisher, G.H. (1967), Measuring ambiguity, Am.J.Psychol. 80:541-547.
Gallagher, S. (1979), Suggestions towards a revision of Husserl's Phenomenology of time-consciousness, Man and World 12:445-464.
Glass, L. and M.C.Mackey (1988), From Clocks to Chaos, Princeton University Press, Princenton.
Granel, G. (1968), Le Sens du Temps et de la Perception chez E.Husserl, Gallimard, Paris.
Gurwitsch, A. (1966), Studies in Phenomenology and Psychology, Northwestern Univ. Press, 1966
Held, K. (1966), Lebendige Gegenwart, M.Nijhoff, The Hague.
Heidegger, M. (1928/67), Sein und Zeit, Max Niemeyer, Hambourg
Husserl, E. (1966), Zur Phänomenologie des Inneren Zeitbewusstseins (1893-1917), Husserliana X , R.Bohm (Ed.), M.Nijhoff, The Hague. (The Phenomenology of Internal Time Consciousness, Indiana Univ.Press, Blomington, 1964 (Translated by James S. Churchill).
Ihde, D. (1977), Experimental Phenomenology: An Introduction, Putnam & Sons, New York.
James, W. (1898/1988), Principles of Psychology, 2 vols. Dover, New York
Kelso, J.A.S. (1995), Dynamic Patterns: The self-organization of brain and behavior, MIT Press, Cambridge.
Kelso, J.A.S., P.Case, T.Holroyd, E.Horvath, E.Raczaszek, B.Tuller, and M.Ding (1994), Multistability and metastability in perceptual and brain dynamics, In: M.Staedler and P.Kruse (Eds.), Multistability in Cognition, Springer Series on Synergetics, Berlin.
Kerby, A. P. (1991),Narrative and the Self, Indiana University Press, Bloomington.
Larrabee, M. (1994), Inside time-consciousness: Diagramming the flux, Husserl Studies 10:181-210.
Leopold, D. and N. Logothetis (1996), Activity changes in early visual cortex reflect monkey's percepts during binocular rivalry, Nature 379:549-553.
Le Van Quyen, M., H.Schuster and F. Varela (1996), Fast rhythms can emerge from slow neuronal oscillators, Int.J.Bifurcation Chaos, 6: 1807-1816.
Le Van Quyen, M., J.Martinerie, C. Adam, J-Ph.Lachaux, M.Baulac, B.Renault, and F.Varela (1997a), Le Van Quyen, M., J.Martinerie, C. Adam, J-Ph.Lachaux, M.Baulac, B.Renault, and F.Varela (1997) , Temporal patterns in human epileptic activity are modulated by perceptual discriminations, Neuroreport 8: 1703-1710.
Le Van Quyen, M., J.Martinerie, C. Adam, H.Schuster, and F.Varela (1997b)Unstable periodic orbits in human epileptic acivity,Physica E (in press).
Levinas, E. (1988), En découvrant l'existence avec Husserl et Heidegger, J.Vrin, Paris.
Llinás, R. (1988), The intrinsic electrophysiological properties of mammalian neurons, Science 242:1654-1664.
Lorenz, K. (1972), Gesammelte Abhandlungen, 2 vols, Piper Verlag, Münich.
Mackey, D. and L. Glass (1988), From Clocks to Chaos, Princenton Univ. Press, Princenton.
McInerney, P. (1991), Time and Experience, Temple Univ.Press: Philadelphia.
Merleau-Ponty, M. (1945), Phénomenologie de la Perception, Gallimard, Paris
Merleau-Ponty, M. (1964), Le Visible et l'Invisible, Gallimard, Paris.
Miskiewicz, W. (1985), La Phénoménologie du Temps chez Husserl, Ph.D. Thesis, University of Paris Sorbonne.
Miller, I. (1984), Husserl, Perception and Temporal Awareness, MIT Press: Cambridge.
Mirollo, R. and S.Strogartz (1990), Synchronisation of pulse-coupled biological oscillators SIAM J.Appl.Math. 50:1645-1662.
Nuñez, A., Amzica, F., Steriade, M. (1993), Electrophysiology of cat association cortical cells in vivo: intrinsic properties and synaptic responses, J.Neurophysiol. 70:418-430.
Ott, E. T.Sauer, and J.Yorke (1993) (Eds.), Coping with Chaos, Wiley, New York
Petitot, J. (1992), Physique du Sens, Editions du CNRS, Paris.
Pöppel, E. (1988), Mindworks: Time and conscious experience, Harcourt Brace Jovanovich: Boston.
Pöppel, E. and K.Schill (1995), Time perception: Problems of representation and processing, In: M.A.Arbib (Ed.) Handbook of Brain Theory and Neural Networks, MIT Press, Cambridge, pp. 987-990.
Port, R. and T. van Gelder (1995), Mind as Motion: Explorations in the Dynamics of Cognition, MIT Press, Cambridge.
Rorty, E. (1980) (Ed.), Explaining Emotions, University of California Press, Berkeley.
Rugg, M. (1995), Event-related potential studies of human memory, In: Gazzaniga, M (Ed.), The Cognitive neurosciences, MIT Press, Cambridge, Chapter 50.
Singer, W. (1993) Synchronization of cortical activity and its putative role in information processing and learning. Ann. Rev. Physiol., 55: 349-374.
Thompson, E. and F. Varela (1997), Why the Mind is not in the Head, Harvard University Press, Cambridge (forthcoming)
Traub, R.D., M. Whittington, I.Stanford, and J.Jeffreys (1996), A mechanism for generation of long-range synchronous fast oscillations in the cortex, Nature 383:621-624.
van Gelder, T. (1997), Wooden iron: contemporary cognitive science meets Husserlian phenomenology, In: Petitot et al. (1997) (this volume)
Varela,F., A.Toro. E.R.John, and E.Schwartz (1981), Perceptual framing and cortical alpha rhythms, Neuropsychologia 19:675-686.
Varela, F. (1979, 1995), Invitation aux Sciences Cognitives, Seuil, Paris.
Varela, F. (1995), Resonant cell assemblies: A new approach to cognitive functioning and neuronal synchrony, Biol.Research 28:81-95.
Varela, F. (1996), Neurophenomenology: A methodological remedy for the hard problem, J.Consc.Studies 3:330-350.
Varela, F. (1997), The Naturalization of Phenomenology as the Transcendence of Nature: Searching for generative mutual constraints, Alter: Revue de Phénoménologie, No.5 (in press)
Varela, F. (1997), Laying Down a Path in Walking: Essays on enactive cognition, Zone Books/MIT Press, New York (in press)
Varela, F. E.Thompson, and E.Rosch (1991), The Embodied Mind: Cognitive Science and Human Experience, MIT Press, Cambridge.
Winfree, A. (1980), The Geometry of Biological Time, Springer Verlag, New York.
Zahavi, D. (1994), Husserl's phenomenology of the body, Etudes Phénoménologiques No.19, pp. 63-84.