133NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
Phenomenology and the Cognitive Sciences 1: 133–167, 2002. © 2002 Kluwer Academic Publishers. Printed in the Netherlands.
Toward a neurophenomenology as an account of generativepassages: A first empirical case study
ANTOINE LUTZLENA – Neurosciences Cognitives et Imagerie Cérébrale, CNRS UPR 640, Hôpital de la Salpêtriére,47, Blvd. de l’Hôpital, 75651 Paris cedex 13, France (E-mail: [emailprotected])
Abstract. This paper analyzes an explicit instantiation of the program of “neurophenomenology”in a neuroscientific protocol. Neurophenomenology takes seriously the importance of link-ing the scientific study of consciousness to the careful examination of experience with a spe-cific first-person methodology. My first claim is that such strategy is a fruitful heuristic becauseit produces new data and illuminates their relation to subjective experience. My second claimis that the approach could open the door to a natural account of the structure of human expe-rience as it is mobilized in itself in such methodology. In this view, generative passages de-fine the type of circulation which explicitly roots the active and disciplined insight the subjecthas about his/her experience in a biological emergent process.
1. Preamble
1.1. Purposes of this paper
This paper analyzes an explicit instantiation of the program of “neuro-phenomenology” (Varela 1996) in a neuroscientific protocol. This recent work(Lutz et al. 2002) studies the correlation between on-going conscious statesand brain coherent dynamics during a simple perceptual task and illustrateshow accounts of experience by trained subjects and experimental data fromthese experiences can share an explicit relation of “mutual or reciprocal con-straints” (Varela 1996). The first claim is that the basic study discussed herealready validates this research program because it produces new data and il-luminates their relation to subjective experience. It can thus be used as a first-step, toy-model for future investigations.
The second claim is epistemological and theoretical dealing with the spe-cific nature of this circulation. I will argue that this study is an attempt to movebeyond a simple “phenomenal isomorphism” and offers a putative exampleof “generative passages” between the phenomenal accounts and their neuro-biological counterparts (Varela 1997). In this new theoretical approach to thestudy of consciousness the neurophenomenological strategy is integrated as
134 ANTOINE LUTZ
an active and necessary component of the epistemological relation. Mutualconstraints define the type of circulation which explicitly roots the active anddisciplined insight the subject has about his/her experience in a biologicalemergent process. Such a relation was already present as an assumption in theneurophenomenology program. What is at stake is the broadening of the natu-ralization of experience into the realm of its direct self-manifestation and self-affection.1 Such a self-referring relation could give immediate insight into thenature of the causal emergent processes that underlie phenomenal appearance.This will be discussed in a third section in light of the operative concepts of“reciprocal causation” as recently introduced by Thompson and Varela (2001).
1.2. The neurophenomenological approach
The neurophenomenological program encourages researchers to pay attentionnot only to neuronal or physiological data but also to the data produced byaccounts of subjective experience. It has been proposed as a methodologicalanswer to the theoretical debate around the status of phenomenal data in na-ture (Roy et al. 1999).2 To summarize this current debate, functionalist orconnectionist cognitive science can provide theories of how the cognitivemind/brain works in itself, but not of how it seems to work for itself.3 Indeed,cognitive events (such as perception, emotion, pain . . .) are associated withfirst-person events: things appear and are perceived or felt by a “self”, a “sub-ject” who can provide an account of it. But, to borrow Joseph Levine’s termfor it, there is still an “explanatory gap” between these subjective attributes4
and their objective scientific counterparts (Levine 1983).One of the virtues of Varela’s proposition has been to shift this discussion
from a strictly abstract and theoretical framework to a pragmatic one that isexplicitly anchored in lived experience and open to scientific inquiry. Thisstrategy can be summarized as follows: In line with many authors (for instance:Chalmers 1996; Flanagan 1992; Jackendoff 1987; Searle 1992), and in con-trast to the eliminativist position (Churchland 1992), this approach acknowl-edges first-person experience as a field of phenomena unto itself, irreducibleto any other. Thus subjective phenomena are recognized as having a charac-ter of immediate givenness that can be explored. At this point, no particularclaims or presumptions need to be made about their epistemic status.
A second principle is that investigation of this field of phenomena requiresa specific, rigorous technique. What is needed here is to overcome the “just-take-a-look” attitude with regard to experience that is pervasive in cogni-tive protocols or the dominant philosophy of mind. Western and Eastern
135NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
phenomenological traditions have been favored, as we will see later, as ap-propriate pragmatical tools5 but other first-person approaches are being ex-plored.6 The phenomenological methodology relies on the cultivation of agesture of reflexive awareness, called phenomenological reduction (PhR).7 Thegoal of this methodological reduction is to attain intuitions of the descriptivestructural invariants of an experience.
This principle was introduced in the context of the following working hy-pothesis: “phenomenological accounts of the structure of experience and theircounter parts in cognitive science relate to each others through reciprocalconstraints” (Varela 1996). The key point is that both forms of evidence aregranted an equal importance and therefore need the same attention. It is easyto understand how scientific data might account for a mental state, but thereciprocal movement, from the experiential to the experimental, is usuallydismissed. Phenomenological accounts are necessary for at least two reasons:first, they can provide direct experiential validation for a neurobiological pro-posal (conditional to a rigorous methodological examination); second, struc-tural aspects of human experience itself can constrain the interpretation ofempirical results. In this paper, for instance, I will discuss the constraints ofphenomenal temporality on neurodynamics. Neurophenomenology thus dif-fers from dualism in that it does not need to assume extra ontological entitiesto allow us to progress toward bridging the explanatory gap.8
1.3. The organization of this paper
Neurophenomenology is thus a pragmatic methodology for combining first-and third-person data. With the notion of mutual constraints, it proposes anapproach that explicitly integrates experiential accounts provided by trainedsubjects with experimental accounts. In the study to be discussed, we haveapplied it to test a theory that naturalizes the flow and unity of time-conscious-ness.9 It is crucial to realize that this theory seeks to give a natural account ofthe capacity of the subject to access his/her own experience, the same par-ticular skill mobilized under the name of phenomenological reduction inneurophenomenology (Varela 1997, p. 368). The two domains of discourse(the natural and the phenomenological) are circularly intertwined. This posi-tion could be seen as contradictory for those preoccupied by a priori argumen-tation, but those who are sensitive to the pragmatic dimension of knowledgewill understand that our main motivation is to gain new insights about thismulti-perspectival dynamics of phenomena. In this specific realm of experi-ence, the mutual constraint viewed as a generative passage can be developed
136 ANTOINE LUTZ
at three levels: the methodological (= in regard to developing a praxis), theo-retical (= providing a natural account) and epistemological ones (= develop-ing a critical perspective on the conditions and the formation of such theory).
With the experimental studies discussed here, we explore these methodologi-cal, theoretical and epistemological claims within a classical neuroscientificprotocol. Again it is merely a first step. These three aspects have been con-jointly evaluated in order to provide the reader with a case study of the gen-eral approach.
My argumentative strategy will be to contrast the neurophenomenologicalaccounts of two cognitive studies involving simple visual protocols: in the firstthe Mooney face (Rodriguez et al. 1999) and in the second the 3D illusion(stereogram) (Lutz et al. 2002). The neurophenomenological strategy was putto work only in the second protocol. Their differences, as we will see, are notdue to the stimuli, but to the role given to the subject and to the way the joint-analysis of first- and third-person data was performed. I will start, in the nextsection, by illustrating the praxis of neurophenomenology. Then, in the sec-tion that follows, I will argue that the first protocol instantiates a “phenomenalisomorphic” relation between these two levels whereas the second protocol,which takes mutual constraints into account, leads potentially to the “genera-tive passages” that instantiate the more complex and demanding relationsbetween them. In the last section I will analyze this theoretical interpretationof mutual constraints with illustrations from preliminary data.
2. Reciprocal constraints as a heuristic strategy guided by a praxis
2.1. Application to the study of brain dynamics
The driving force of neurophenomenology is its pragmatic dimension. Thegoal of this section is consequently to provide the reader with an explicit casestudy implementing this approach. In this example we used refined first-per-son data to validate a scientific proposal about the functional role in cognitionof large-scale integration mechanisms. I will now briefly recall the scientificassumption that is central here.
Neural assemblies: a dynamical framework to cognitionIt is a well-accepted fact that in the brain any cognitive act is characterized bythe simultaneous activity of distributed brain regions that are functionallyspecialized and constantly interacting. Any assumption about the substrate ofa moment of consciousness must therefore account for the self-coordination
137NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
of these different components necessary to produce a global brain activity thatis transiently unified. In neuroscience the notion of neural assemblies (Abeles1982; Hebb 1949) has been introduced to provide a conceptual frameworkfor the integration of distributed neural activity. Neural assemblies can be de-fined as distributed local networks of neurons transiently linked by recipro-cal dynamical connections. A useful analogy might be a Word Wide Websystem such as Napster, in which geographically distant computers punctu-ally exchange data within transient assemblies that are formed on a staticnetwork of hardwired connections (Varela et al. 2001). In the brain, the emer-gence of a specific neuronal assembly is thought to underlie the operation ofevery cognitive act. Neurons that belong to a given assembly are linked byselective interactions; they interact preferentially with a sub-ensemble of otherneurons that are interconnected. These neural assemblies have a transient anddynamical existence encompassing the span of an elementary cognitive act(a fraction of a second). At the same time, however, this dynamical link hasa minimum time span, long enough for this long-distance coordination toemerge. This span necessarily involves several cycles of reciprocal spike(s)exchanges with transmission delays that last tens of milliseconds. So, in boththe brain and the Web analogy, the relevant parameter to describe these as-semblies is not so much the individual activities of the system as the dynami-cal link between them.
Phase-synchrony as a large-scale integrative mechanismFor a majority of neuroscientists this dynamical link between componentscould be mediated by a temporal code constituting a transient functional rela-tion between them (for recent reviews of this see Engel and Singer 2001; Engel,Fries, and Singer 2001; Singer and Gray 1995; Varela et al. 2001). The tran-sient synchrony of oscillating neuronal discharges over multiple frequencybands (4–80 Hz) constitutes a plausible mechanism for this link. It has beenextensively studied in the last decade and still constitutes an active field ofresearch. Neural synchrony is a multiscale phenomenon in the brain: a localand global scale can be distinguished both at the anatomical and functionallevels. Of the two, local synchrony has been the most extensively studied inelectrophysiology and refers to a local dynamical link within a specific areaor modality. For instance, in vision areas short-scale synchrony has been pro-posed to subserve the “perceptual binding”, that is the extraction of Gestaltproperties of a visual stimulus. Long-range synchrony has just started to bereported between distant regions and could mediate the global coordinationof multi-modal areas during a cognitive act. Such recent data on neuronalsynchrony has led several authors to suggest that synchrony could underline
138 ANTOINE LUTZ
the sensory awareness or account for the unity of consciousness. Such propo-sitions are in good agreement with the fundamental phenomenological factthat, in a non-pathological context, the different aspects of an experience(sound, color, movement, feeling) are not fragmented but appear as a coher-ent whole (see for instance the Gestalt theory, (Köhler 1947)).
In this section I am not directly concerned with the theoretical relation oflarge-scale dynamics to first-person data but simply with the more efficientstudy of these brain integrative mechanisms by the use of first-person data.To showcase the originality of neurophenomenology I will first present acase study that uses a technique of analysis and a type of stimuli similar toour own study, but does not explicitly mobilize the neurophenomenologicalapproach.
2.2. Case study 1 – without the use of neurophenomenology
The Mooney face protocol (Rodriguez et al. 1999)In this protocol subjects were shown upright and upside-down Mooney fig-ures (high contrast faces) which are easily perceived as faces when presentedupright but as meaningless black-and-white forms when upside-down (seeFigure 1). Subjects were asked to rapidly press a button indicating whether atfirst glance they perceived a face or not. The electrical activity was recordedat the scalp surface (EEG).
Estimation of the local and large-scale integrationThe synchronization of neural populations can be observed in the EEG at twocomplementary levels: (a) either “locally”, in the signal of a single electrodeor (b) over a longer distance, between the signals of two electrodes.10 Thesetwo measures provide an estimation in extracortical recordings of the short-and long- scale brain synchronies.
ResultsThe upright presentation of the Mooney face first evoked a brain activity phase-locked to the stimulus around 50–100 ms. This activity was followed after aninterval of 100–150 ms by a first global pattern of long-distance synchroni-zation involving parieto-occipital and fronto-temporal electrodes (Figure 1).Such a large-scale pattern of synchrony was not detected when subjects didnot perceive a face (with upside-down Mooney figures). This pattern couldthen be interpreted as the “shadow” of a large-scale coherent assembly sub-serving the Gestalt perception.
139NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
This interval of coherence was interrupted at around 500 ms by a massiveperiod of decrease of synchronization compared to a preceding baseline, fol-lowed at around 700 ms by a second period of synchronization. This sec-ond peak was correlated to the motor response in both the perception andnon-perception. These data suggest that the two dynamical patterns corre-late to the two acts of perception and action. These two patterns were punctu-ated by a pattern of phase-scattering, which can be interpreted as a dynamicaltransition from the neuronal assemblies involved in perception (associativememory, emotion, vision) to the ones involved in motor response.
Opacity of brain activity: its intrinsic variabilityLike most EEG studies, the Mooney face study relies on averaging techniques– across trials and often across subjects. These techniques are very effectivein finding the major components of the neural activity. Yet these techniquescannot take into account the highly variable EEG signals. For instance in thisstudy the two patterns of synchronous oscillations were quite variable inlatencies, frequencies, and spatial distribution from repetition to repetition.The source of this variability is believed to reside mainly in fluctuations ofthe subject’s cognitive “context” defined by his attentive state, his spontane-ous thought-process, his strategy to carry out the task and so on. This illustratesa well-known fact that even during well-calibrated cognitive tasks successivebrain responses to repeated, identical stimuli are highly variable. Such fluc-tuating brain response derives from the active interaction between this cogni-tive background and the stimulation that disturbs it: the neural response isshaped by the ongoing activity (Engel et al. 2001). Although it is common incognitive sciences to control, at least indirectly, some of the factors that con-dition this ongoing state such as attention, vigilance, or motivation, the on-going activity has not yet been analyzed systematically. In practice this typeof qualitative first-person data has been usually omitted from brain imagingstudies and the variability is defined as noise. The reason for this is that ver-bal reports can be biased or untrue making it difficult to collect data aboutinner experience.
This case study offers a perfect example of how the subject could becomean active collaborator in brain analysis and why first-person methodologies arecrucially needed to rigorously collect these data. The neurophenomenologicalprogram specifically addresses these two issues. We implement its methodol-ogy in the recent study described next, which takes into account first-person dataabout the cognitive context in order to make understandable a variability in thebrain neural responses usually defined as noise.
140 ANTOINE LUTZ
Figure 2. The shadow of perception within an ongoing phenomenal and dynamical context.EEG was recorded from electrodes at the scalp surface. Subjects were shown first a back-ground image with random-dot points. They were asked to fuse two little squares at the bottomof the screen and to remain in this position for several seconds. At the end of this preparationperiod, a stereogram (3D illusion) was presented. Subjects were instructed to press a buttonas soon as the shape had completely emerged and to give a brief verbal report of their expe-rience (see Section 2.3). Dynamical neural signatures (DNS) of the pre-stimulation activitiesand the brain responses are presented for one subject during readiness with immediate per-ception (SR) (154 trials) and spontaneous unreadiness with surprise during stimulation (SU)(38 trials). Color-coding indicates scalp distribution of gamma power around 35 Hz normal-ized compared to a distant baseline ([–8200 to 7200 ms], 0 ms corresponds to the presentationof the stereogram) (the normalization technique is different than in the Mooney one). Blackand white lines correspond to significant increase and decrease in synchrony, respectively,compared to the baseline. For a precise description of the Figure see Section 4.3. Modifiedfrom Lutz et al. 2002.
→
1.3. Case study 2 – employing neurophenomenology
The protocol (Lutz et al. 2002)This protocol was based on a well-known illusory depth perception task (Fig-ure 2). We chose this paradigm because the perception of a 3D object arisingfrom an autostereogram (Julesz 1971) triggers a vivid phenomenal experiencewith identified neurobiological mechanisms. It is formally very similar to theMooney protocol: it also uses a visual stimulus and performs the same syn-chrony analysis techniques.
The task began when the subjects fixed a dot-pattern containing no depthinformation. After an auditory signal, the subjects were asked to fuse two lit-tle squares at the bottom of the screen and to remain in this eye position forseven seconds. At the end of this preparation period, the random-dot patternwas changed to a slightly different random-dot pattern with binocular dispari-ties (autostereogram). Subjects were readily able to see a 3D illusory geometricshape (depth illusion). They were instructed to press a button with their righthand as soon as the shape had completely emerged. This ended the trial, afterwhich the subjects gave a brief verbal report of their experience. In their re-ports the subjects used phenomenal invariants (or categories) found and sta-bilized during the training session.
Training session: a broad use of phenomenological reductionThe purpose of this pre-recording session was to invite the subject to care-fully explore the variations in his/her cognitive context during repetitive ex-
141NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
Figure 1. The shadow of perception. Average scalp distribution of gamma activity and phase synchrony. EEGwas recorded from electrodes at the scalp surface. Subjects were shown Mooney figures (high contrast faces),which are easily perceived as faces when presented upright (here Marylyn Monroe’s profile), but usuallyperceived as meaningless black-and-white forms when upside-down. The subjects’ task was a rapid two-choicebutton response of whether or not they perceived a face at first glance. Color coding indicates gamma power(averaged in a 34–40 Hz frequency range) over a given electrode and during a 180 ms time window. Gammaactivity is spatially homogenous and similar between conditions over time. By contrast, phase synchrony ismarkedly regional and differs between conditions. Synchrony between electrode pairs is indicated by blackand green lines, corresponding to a significant increase or decrease in synchrony, respectively. These are shownonly if the synchrony value is beyond the distribution of shuffled data sets. Modified from Rodriguez et al.1999.
142 ANTOINE LUTZ
posure to the task. Subjects were therefore intensively trained to perform thetask during this preliminary session. The precise gesture of phenomenologicalreduction was broadly adapted for this study in order to open up the field ofinvestigation. The main concern was to “come back to things themselves” asHusserl frequently urged. Thus the subjects were asked to ascertain their ownphenomenal categories. This strategy can be seen as an extension of the tra-ditional procedure in cognitive sciences that is based on the use of verbal re-ports and questionnaires (Ericsson and Simon 1984). To gain new descriptiveinsights, phenomenologists cultivated a specific method based on the gestureof reduction. In a nutshell, reduction begins by the bracketing of habitual at-titudes in order to shift the attention from what habitually appears in the world,say the 3D percept, toward the immediate arising of the appearance itself, saythe process of emergence of the percept. Reduction can be described as a par-ticular reflexive act designed for considering experience differently (Deprazet al. 2002). Varela, in his program, underlined four principal phases in thismethod: (1) inducing a suspension of beliefs, (2) gaining intimacy, or intui-tive evidence with the domain of investigation, (3) extracting descriptive in-variants and using intersubjective validations; (4) long-term training to acquireknow-how in 1–3 (for details see Varela 1996).
In this study the gesture of reduction was either self-induced by subjectsfamiliar with it, or induced by the experimentalist through open questions(see Petitmengin, 1999; Vermersch 1994). Open questions posed immedi-ately after the task can help the subject to redirect his/her attention towardsthe implicit know-how he/she implemented to carry it out or towards thetexture of his/her experience during the task.11 Subjects were re-exposed tothe stimuli until they found their own stable experimental invariants to de-scribe the main elements of the cognitive context in which they perceivedthe 3D shapes.
This training session addresses the need to go beyond the “just-take a look”attitude in order to bring to the fore relevant first-person data. In the next stepwe tried to collect these first-person data directly during the EEG recordingin order to maintain the link between the immediate experience and the sci-entific measurements.
Recording session: joint-collection of first- and third-person dataIn a second part, we recorded both subjects’ electrical brain activity (EEG)and, after each trial, their own verbal report about their cognitive context.Subjects briefly and precisely labeled their experiences based on the invari-ants found in the training session. These categories were used to divide theindividual trials into several phenomenological clusters (PhC). We recorded
143NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
two to three sessions to gather at least 40 trials per PhC.12 The degree of prepa-ration felt by the subject and the quality of his perception appear as a com-mon factor through the subjects; we used it to cluster the trials into threecategories: Steady readiness (SR),13 Fragmented readiness (FR),14 and Un-readiness (SU, SIU).15 Sub-categories, describing the unfolding of the percep-tion for instance, were found in individuals.
Further refinement is needed to capture the potential richness of even thissimple perceptual experience. These clusters are clearly just a first step. Yetthey already illustrate the pragmatic dimension of the neurophenomenologicalapproach, namely the need for collecting first-person data from trained sub-jects in a disciplined way.
2.4. Testing the working assumption: Joint-analysis of first- and third-person data
This pragmatic stage allowed us to rigorously collect both relevant phenom-enal data and concomitant scientific data (EEG signal). It is thus possible tosee if a particular PhC concerning preparation is characterized by a specificbrain responses stable through trials. This joint-analysis of first and thirdperson material provides an instantiation of the mutual constraints. The sub-ject can now play an active role, not necessarily in the sense of being morevoluntary during the task (the subject is actually spontaneously distracted insome trials), but because his experience is integrated in the analysis via thephenomenal clustering of trials.
ResultsIn this section I will analyze only the responses to stimulation. This consti-tutes half of the data reported in this study. We found that the preparatory state,as reported by the subjects, modulates both the behavioral performance andthe brain responses that follow. The reaction times were dependent on thedegree of preparation reported by the subjects: they were longer when thesubjects were less prepared. The induced response (the oscillatory activityarriving around 250 ms in the Mooney face study) was modulated in ampli-tude in posterior electrodes (visual areas) in function of the degree of prepa-ration. A separate study of each subject and each cluster reveals differentdynamical trajectories that are stable- specific to each cluster during the brainresponses (for instance see Figure 2). In this particular example of clusterswe can see a similar topographical pattern of large-scale synchrony during themotor response in the prepared versus unprepared pattern, but occurring re-
144 ANTOINE LUTZ
spectively at 300 ms (on average over 100 trials) and 600 ms (on average over38 trials). This later pattern of synchrony correlates in the unreadiness clusterof trials with longer reaction times. Such highly contextual patterns of syn-chrony would have been canceled out by a global averaging like the one inthe Mooney study. By combining this new type of knowledge, variability inthe behavioral and brain responses has became more intelligible compared tothe Mooney study and it has generated new data.
A first evaluationThis simple case study is just a first-step but already illustrates how fertile thisapproach could be to identify biophysical properties and to understand theirrelation to experience. In our case this strategy was very useful to take intoaccount fluctuation of the subject’s emotional state, attention, or mental strat-egy which occur and cannot be fully controlled in a protocol. The objective isto pay more meticulous attention to the intimate and direct knowledge that asubject has about his/her experience and to access this knowledge in a suffi-ciently controlled manner so that it is compatible with the more traditionalmethods for the collection of neural data. We have taken two steps in thatdirection, by (a) keeping a trial-by-trial account of the subject’s report, and(b) using the subject’s own categories to organize the trials into clusters withsimilar experiential features. Further refinement is needed to capture the po-tential richness of even this simple perceptual experience. This depends pri-marily on the possibility of working with subjects trained to discriminate andstabilize their experience.
To summarize, the notion of “mutual constraints” was defined in this firstmethodological section as a heuristic strategy to provide mutual insights be-tween the first-person and scientific accounts. This heuristic method dependson a pragmatics for the cultivation of our capacity for attentive bracketing andintuition (reduction). In effect, it is this pragmatic and disciplined dimensionof the method which could make it, at the epistemological level, more than asimple heuristic. In this more demanding view the disciplined first-personaccount would be an integral element for the validation of neurobiologicalproposals.
This point has been similarly made by Michel Bitbol in this issue of PCS.Yet although I am in complete sympathy with his analysis about the radicalprocedural dimension of neurophenomenology, I do not subscribe to his viewthat it is in itself “sufficient to dissolve the ‘hard problem’ ”.16 In the science ofconsciousness, human phenomenality is an empirical question as much as it isa methodological one. Thus bridging the explanatory gap not only means to finda relevant methodology and epistemology but also, in a complementary way,
145NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
to find an explicit natural theory giving sense to these choices. The fact that Varelaintroduces neurophenomenology as a “working assumption” and not simply asa methodological principle attests that pragmatic tools, like phenomenologicalreduction, possess both an instrumental and a theoretical side (Thompson andVarela 2001; Varela 1997; 1999). Such a more encompassing view presented inthe rest of the paper describes the neurophenomenology project as a triptychwith pragmatic, epistemological and theoretical dimensions.
3. A typology of reciprocal constraints as a naturalizing account ofexperience
This section is more epistemological and attempts to analyze, in the light ofthese two case studies, the nature of the relation established between first-person and third-person data. This topic addresses the possibility of giving anaturalized account of phenomenology. It is well known that this project hadbeen rejected from the start by Husserl who had always argued that somethingin experience escaped the jurisdiction of the natural sciences. In his view itwas the consequence of a fundamental opposition between the exact essencesand the ideal concepts of mathematics and the inexact and the morphologicalessences involved in what we experience immediately (Roy et al. 1999, p. 42).Yet, as claimed by the editors of Naturalizing Phenomenology (Petitot et al.1999), such a position could be challenged. Indeed the contemporary com-putational theory and the theory of complexity make it possible to col-lapse the classical opposition between the body and the mind.17 On this view,Husserl’s anti-naturalism was more a criticism against the sciences of his timethan against the naturalization project per se.
Among the various ways of naturalization overviewed in this book, I amfollowing a typology of mutual constraints which distinguishes between threetypes of circulation: (1) bridge locus as analytic isomorphism; (2) phenom-enal isomorphism; and (3) generative passages (Varela 1997). I will firstanalyze the two former relations and their limits before moving to a closeexamination of the latter.
3.1. Bridge locus as analytical isomorphism
The first relation, very popular among neuroscientists, consists of interpretingspecific neurons or structures like the bridge locus between a phenomenal fea-ture and the neuronal substrate (for a complete discussion see Pessoa et al. 1998).Even if specialized and localized structures can provide the necessary condi-
146 ANTOINE LUTZ
tion for a cognitive act they are not sufficient to account for the whole experi-ence. Indeed as mentioned above cognitive acts involve the distributed andconcurrent activity of a mosaic of brain areas. The limitation of this eliminativistattitude is that it does not fully account for the phenomenality of perception.
3.2. Phenomenal isomorphism
In contrast to “analytical isomorphism,” which focuses on specific structures,“phenomenal isomorphism” tries to broaden the scope of explanation to theglobal neural operations involved during a cognitive act. The Mooney facestudy offers a typical example. It investigates a global process, namely large-scale integrative mechanisms, which is proposed to underline the phenom-enal unity of the perceptual and motor experience (the shape, emotion,posture, movement appear for the subject as a coherent whole). This origi-nal work demonstrated for the first time a logical isomorphism between thefirst-person descriptions (perception versus non-perception of the Mooneyface) and the third-person descriptions involving a large-scale integrativemechanism (increase versus non-increase of the level of coherence in thegamma frequency band). In this study the phenomenal accounts were usedto identify the correct explanatory mechanisms subserving it at the neurallevel.18 More generally in this type of relation first-person data can eithervalidate a link to third-person data or provide constraints invalidating first-person models.
Although it is a fruitful strategy, this isomorphic interpretation nonethelesshas as an implicit premise the separation of phenomenal description and sci-entific explanation. For instance, in the Mooney face study, the subject onlyhas to activate a button when the Gestalt of the face passively emerges to his/her perception. The observer’s only role is to assert the concomitance or theparallelism between the phenomenal and empirical series. Therefore he/shecannot provide any direct insight into the nature of the causal process whichunderlies his/her experience. Consequently only the third-person data arerelevant in the account.
Such an epistemological choice could be understood as ruled by a princi-ple of prudence: phenomenal consciousness is recognized as a scientific ques-tion, but must be reduced to the third-person realm in the final account. Thiscould explain why visual illusions (gestalt or depth illusion, multistable orambiguous figures) are so frequently used. These stimuli have a direct andcontrasted manifestation in the experiential-mental sphere and are still wellcontrolled by the experimentalist. They make possible the discovery of rel-
147NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
evant relations between experiential and scientific levels without changing ourusual scientific style. But to what degree can this strategy be productive? Iwill now successively address this question at the experiential, natural, andformal levels.
Limitations from the experimental realmIndeed, this approach still does not explicitly account for the fact that experi-ence can be manifested to itself or affected by itself. “What it is to be a sub-ject doing a visual protocol” changes from trial to trial while something inthe phenomenal content of the perception remains the same.19 Seeing a visualillusion affects my experience, and the texture of this experience is also af-fected by my ongoing phenomenal state during the stimulation. Husserl re-markably summarized this point:
Every experience is “consciousness” (Bewusstsein) and consciousness is always conscious-ness-of . . . Every experience is itself experienced (selbst erlebt) and to that extent alsointended (bewusst).20
This self-affection and self-manifestation of experience attests to a particularmode of access to our own mental life. It has to do with reflection but shouldnot be merely identified with it. It is important here to diffuse a potentialmisunderstanding. As it has been noted, for instance by Merleau-Ponty (1945/1962), a strict reflexive paradigm, like the early Husserl’s, risks overestimat-ing the constitutive role of the “observer” by assuming his/her constant prec-edence over the appearing object. Among others limitations it can dangerouslyseparate the subject, almost in a dualistic way, from being situated in the worldand in relation to his/her body. The notion of being situated is expressed in anauto-affection paradigm (Depraz 1998) which emphasizes in a complementaryway the passive dimension of consciousness. Such a position can thus preservethe unity of experience while still being engaged into a differentiating attitude.
The class of phenomenological invariants, I have just sketched, is unfortu-nately dismissed a priori by phenomenal isomorphism because the subject’sonly role is to describe what appears to him/her and not how his/her ownexperience appears to him/her as well. The “experience about an experience”could still be present structurally in the neurobiological account, for instanceas a particular class of dynamics, if the role of the trained subject in the ac-count was extended past simply maintaining the link between the first-per-son and third person phenomena. This type of assumption is rejected a prioriin phenomenal isomorphism, and, therefore, constitutes a damageable limi-tation to this type of relation.
148 ANTOINE LUTZ
Limitations from the biophysical accountThis issue has already been introduced in the discussion of the problem ofvariability in the brain responses. The high variability of the synchrony pat-tern in its latency, topography, or frequency suggests that phenomenal isomor-phism can only account for relations between the appearance of an “external”object for the subject and an average pattern of large-scale synchrony. Yet sucha relation cannot fully account for the emergence of a peculiar configurationof the dynamic during a particular trial. The limitation here is that such emer-gence has not been related yet to the genesis of a particular moment of con-sciousness as it is experienced.
Limitations from the dynamical frameworkThis issue is more clear when it is formulated within the dynamical frame-work of neural assemblies. A stimulation like the Mooney face can be inter-preted as an external perturbation of the ongoing state of the system and theinduced patterns of synchrony can be interpreted as the average dynamicallinks common to all the neural assemblies participating in the perception andmotor response. The limitation of phenomenal isomorphism can thus be statedas follows: since “what it is like to be a subject doing the task” is not takeninto account, the ongoing neural assembly which is perturbed by the stimula-tion is not involved in the account but is defined as neutral. In practice this isimplicitly assumed by defining as a neutral baseline the pre-stimulus dynamicalstate. The putative role of the current ongoing neural assembly as a selectiveconstraint is again implicitly rejected.
To summarize, phenomenal isomorphism is not satisfying because it is in-sufficient to account for the emergence of particular phenomenal state and itsneurobiological signature in response to a specific external perturbation. Atthe first-person level the limitation comes from the implicit view of the sub-ject as a passive and disembodied spectator. At the neurobiological level thelimitation comes from the simplification of the cognitive system to a context-free input-output device or, at least to a chain of exogenously constrainedcognitive processes. At the formal level, modeled as a dynamical system,phenomenal isomorphism implicitly assumes that the system state beforestimulation can either be defined as neutral or sufficiently controlled by ex-ternal forces (for instance distractors, priming).
3.3. Generative passages
The relation of generative passages can first be defined negatively as an at-tempt to overcome the limits of phenomenal isomorphism. A positive defi-
149NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
nition requires a further step, that of extending the mutual circulation be-tween phenomenal account and natural science to the entire phenomenalrealm. Therefore, further distinctions are first required to give sense to thisnew epistemological relation. This will be done in the next section. So farone can say that such a circulation could become “operationally generative”(Varela 1997) because reciprocal insights could be grounding explicitly inbiological emergence the disciplined experience one can have about oneown’s experience (this refers in phenomenology to eidetic descriptions underreduction analysis). Therefore, in this specific realm of experience, themutual constraint as generative passage can be defined both at the methodo-logical and theoretical levels. Theoretically, mutual constraints have beendefined in a neurophenomenological account with three threads (Varela1997):
1. phenomenological data and invariant structural features of experience(thread #1)
2. neural and somatic substrates (thread #2)3. formal dynamical models (thread #3)
In the last section, I will attempt to refine the distinctions within each of thesethree levels in order to show, then, how they can be fruitfully articulated intoan original relation of mutual constraints. To illustrate this theoretical approachI will use data from the protocol on 3D illusion.
4. Reciprocal constraints as reciprocal causation?
4.1. The dynamical expression of generative passages
The “enactive” approachThis last expression of reciprocal constraints is more theoretical and partici-pates in the cognitive paradigm of enaction (Varela et al. 1991). In this ap-proach cognition is not understood as abstract, computational processes butas based on situated and embodied agents (see e.g., Clark 1997; Thompsonand Varela 2001). The focus is on (1) the ongoing coupling of the cognitiveagents with an environment during sensori-motor activities and (2) how thiscoupling modulates the ongoing endogenous activity participating in theorganismic regulation of the agent’s life. This style of explanation of cogni-tion naturally joins the (more general) dynamical approach (for inspiring ex-amples see Freeman, 1999; Haken et al. 1996, Kelso 1995) and differs markedlyfrom the computationalist (Fodor 1984; Pylyshyn 1984) or connectionnist para-
150 ANTOINE LUTZ
digms (McClelland 1986).21 Indeed, what is at stake is an understanding of howthe system’s state changes over time through external perturbations withoutspecifying a pre-given and unique relation between an input and an internal state.Thus for the dynamical approach formal concepts and tools inspired by non-linear dynamical system theory are more adequate than symbolic and informa-tion processing models. This is why dynamical concepts have been extensivelyfavored as the formal model (thread #3) in the neurophenomenal account. Thistheoretical question is still actively debated between partisans of these threeparadigms because it directly constrains any account of consciousness.
“Reciprocal” causationsThe dynamical expression of mutual constraints (thread #3) has been proposed(Thompson and Varela 2001; Varela 1997, 1999) from the dynamical analy-sis of neural assemblies within the theory of non-linear self-organization (e.g.,Arbib 1995). Indeed self-organized properties can be modeled with non-lin-ear oscillator networks (e.g., Acebron 2001; Kopell 2000) expressing elegantlysimilar behaviors as the one found in synchronized group of neurons. It isfruitful thus to formalize the collective behavior of a neuronal population asan emergent process. Emergence can be crucially described as a two-direc-tional process. First, there is a local-to-global determination or “upward cau-sation” by which global processes emerge from the collective interaction ofneurons. These global processes are endowed with intrinsic behaviors andspecific dynamical properties (for instance their life span). Second, there isglobal-to-local determination (“downward causation”), whereby the globalfeatures (or order parameter)22 constrain the local activity. The idea is that thiscollective behavior confines or “enslaves” (Haken 1996) the individual com-ponents to have specific dynamical interactions with others individuals. In“Radical Embodiment”, Thompson and Varela suggested calling this recip-rocal (but not symmetrical)23 relationship a “reciprocal causality.”
Theoretical evaluationThe introduction of reciprocal causation might be perceived as a subtle ex-pression of dualism. It may be useful to highlight that it corresponds to opera-tional distinctions introduced to study the relation between conscious activityand the self-organization of the organism. Self-organization is involved atvarious scales in the organism (Thompson and Varela 2001). The relevant levelof description favored here is to define local/global at the spatio-temporal scaleof short-scale/long-scale synchrony processes24 (Section 2.1). In the particu-lar case of our protocol, the relation between local-to-global and global-to-local processes convey useful distinctions to investigate the mutual constraints
151NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
between the genesis of a moment of consciousness belonging to a flow of othersmoments and the formation, maintenance and change of coherent patterns inthe brain. The notions of downward/upward causations are potentially fruit-ful because they express directly a transient directionality in the neuro-dynamical emergence making it possible to root explicitly in such a dynamicalprinciple the intentionality of conscious moments. This formalism is also in-teresting because it accounts for the intrinsic tendencies of the system tomaintain a coordination between neural groups or to spontaneously bifurcate.
Role of the formal level in the relation of generative passagesHow could this formal dimension (thread #3) be related to the two others inthe neurophenomenological account? Briefly, this component constitutes anintermediate and neutral level in which both the experiential and biophysicallevels can be expressed.25 First person data (thread #1) can fully constitutedynamical descriptions (thread #3) as soon as they are structurally preciseenough (see next section). In that sense they are needed not only to maintaina direct link to the experiential realm but also to explicitly constrain the neu-robiological level. Similarly dynamical descriptions can be directly groundedin the biophysical data (thread #2) and illuminate their analysis. Such a for-mal level is thus essential to mediate a continuous passage from experienceto natural processes or, to say this differently, to allow mental and naturalproperties to coexist without contradictions.
At the end of the previous section, generative passages were presented asan epistemological relation attempting to ground explicitly in biological emer-gence the disciplined experience one can have of one own experience. At thetheoretical level, the section justified the need for systemic and dynamicaldescriptions of the organism. Following several authors the paradigm of self-organizational emergence was favored. Emergence was described as the waya global dynamical behavior results from the interaction of local processes.At the epistemological level, this section justified the need for a formal descrip-tion as an intermediate level between first- and third-person data. The dynami-cal distinctions presented here can potentially refine the notion of mutualconstraints as soon as a similar granularity is obtained at both the phenomenaland neurobiological levels. The pragmatic requisite of neurophenomenologyappears again at this stage, but in a more demanding way.
4.2. The phenomenal region of generative passages
For didactic reasons, the formal level has been analyzed first. Yet, to be rigor-ous, this phenomenal section should have been presented first, because expe-
152 ANTOINE LUTZ
riential distinctions precede their theoretical formulations. This section willstart thus from the limits, in the experiential realms, found in the phenomenalisomorphism relation. It will attempt to extent the field of first-person data todistinctions insufficiently considered in this previous relation. Several of thesedistinctions will clearly constitute open questions for further researches. Yetexplicit phenomenal features will be presented as counterparts to the previ-ous dynamical distinctions.
The immanent/transcendent distinctionsThe specificity of generative passages is to extend the first-person investiga-tion to the domain of phenomenological reduction (Husserl 1913). This ques-tion has already been addressed (Section 3.2) by contrasting “what” appearsto the subject to “how” it appears to him/her. More technically, this domain isgrounded in the distinctions in experience between transcendent and imma-nent mode of appearance. (Husserl 1913). I perceive an object as transcend-ent when it appears to me as “real,” existing in itself, “there” in space and time.This refers to our daily perception. On the other hand I can suspend my be-liefs about what is being examined to turn my attention to the way this eventmanifests itself to me as a direct and immediate phenomenon. This notion ofimmanent describes the intimate mode of access by which conscious activitycan perceive itself as being spontaneously changing or being affected. This tran-sient relation to our own experience unfolds a new horizon of meanings whichcan initiate an attention shift to a new intentional moment. In that sense imma-nence describes rather the self-motion of consciousness whereas transcendencedescribes the appearance of something in consciousness (Varela 1999, p. 295).These two modes are intrinsically linked. This immanent access to experienceis a central aspect of the neurophenomenological method. It corresponds to thenotion of intuition or intimacy to experience (Varela 1996, p. 345).
The immanent/transcendent domain is crucial because it is where the rela-tions between the mind, the body and the world fully articulate themselves.26
An exemplary way to investigate such relations is to study their temporal struc-ture and their embodied dynamics. Indeed, objects perceived in the world (tran-scendent mode) exist in consciousness as temporal object-events. Similarly, Ican directly experience the flow of temporality (immanent mode), as nicelyremarked by Merleau-Ponty: “it is not I who take the initiative of temporal-ity (. . .) time is fusing through me, whatever I do.”27
Temporality is furthermore a paradigmatic example at the biophysical levelbecause the brain and the body are intrinsically dynamical processes, as de-scribed in Section 4.1. Therefore, time-consciousness constitutes a fundamen-tal case study (Varela 1999).
153NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
A complete description of time-consciousness under reduction is clearlybeyond the scope of this paper. My purpose here is simply (a) to offer exam-ples of a phenomenal counterpart to the notion of local/global; (b) to suggestthat the description of the immanent constitution of temporality is the com-plementary source of knowledge which is sought for the theorization of gen-erative passages.
Static analysis of time-consciousnessThe notion of “now” is usually employed to describe the temporal side ofexperience; Husserl preferred to use the term “living present” in order to evokethe embodiment of temporality (Husserl 1928). The “now” is not like a pointin physics but has a complex texture and a flexible span. In a first approxima-tion, the living-present possesses a center and a periphery analogous to thevisual field. The center is what appears in the immediacy of “now”, say thesensory impression of a musical note, whereas the periphery is what appearsin the present but as just-past, for example the holding, or retention, of thedirect presence of the sound. In brief, sensing a tone appears as a present phe-nomena, whereas the just-past, even if it is intentionally present, appears ashaving already slipped in the immediate past. It is the aboutness of my listen-ing which constitutes the just-past act as a horizon of the “now” moment. Simi-larly the living-present is open to the next moment. While I am listening to asong for instance, I am already anticipating the next tone of the melody in theimmediate future. This act of anticipation is named protention and introducesa level of intentional direction in temporal consciousness.
This center/periphery structure presents thus a distinction between a shortelementary event (scale 1/10), for example, the immediate impression (in the“pure present”), from a longer one (scale 1), say the span or the horizon ofthe living present. Varela’s original intuition (1999, p. 273) is to suggest thatthese two scales are grounded on different temporal scales according to localand global integrative processes (see Section 2.1). A large-scale process re-quires an incompressive temporal framework to synchronize and coordinatedistributed brain areas. Such a span, required to dynamically glue distributedlocal processes into a larger assembly (scale 1), would account for the “depth”of a moment of “now”, that is, the living present.
In our protocol for instance an elementary event corresponds to the impres-sion of the 3D stimulus. The event is rooted in local processes correspondingto the activity of the brain evoked by the stimulation in the early 100 ms. Ourempirical work is thus an attempt to study how this local perturbation willinterfere with the ongoing large-scale dynamics. The way in which this elemen-tary event is lived within the current now-moment can be directly related to
154 ANTOINE LUTZ
the emergence of a particular large-scale pattern of synchrony. This is the coreidea of generative passages.
Genetic analysis of time-consciousness
This notoriously more difficult level of analysis is dedicated to the descrip-tion of the flow of consciousness, namely, the “primal source-point that fromwhich springs the ‘now’ ” (Husserl 1966, §36). This immanent constitutionof events indicates the passive dimension of consciousness. As Husserl re-marked
the entire life of spirit is traversed by the “blind” efficacity [Wirksamkeit] of associations,of impulses [Trieben], of affects [Gefühlen] as excitation [Reizen] and as the determiningsources of impulses, of tendencies emerging [auftauchenden] from obscurity etc. whichdetermine the further course of consciousness in accordance with “blind” rules. (Husserl1952, cited in Depraz 1999, p. 478).
The role of emotion and affect is fundamental to the self-movement of the flow.As Merleau-Ponty pithily summarizes it: “time is affection of oneself by one-self [temps est affection de soi par soi]” (1945/1962, p. 131). The constitu-tive role played by affect in the dynamics of consciousness is remarked on byDepraz: “Affect is there before being there for me in full consciousness: I amaffected before knowing that I am affected” (1994). Emotions are thus con-stitutive of the next moment because they can interrupt the current “now”moment, shift attention to another content and lead the subject to become awareof a change in his/her own experience. As suggested by this short descriptiona profound link exists between the arising of affect and the triggering of re-flective awareness (for a further development see Depraz 1994, 1998). Be-cause reflective awareness is an essential aspect of the phenomenologicalreduction, the extension of mutual constraints to these types of data couldprovide a way to naturalize the reductive gesture itself.
In our protocol we began gathering such genetic distinctions: for instance,when subjects were unprepared, they frequently reported being “interrupted”and surprised by the image while in the middle of a thought (memory, fan-tasy. . .). This aspect of the experience is genetic because it describes how thespontaneous appearance of the 3D percept affects the ongoing thought proc-esses. This shift of intentional content was usually accompanied by a feelingof surprise. These classes of first-person data could be essential for understand-ing the flow of the dynamic, that is the bifurcation from one cognitive mo-ment to another.
155NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
To further examine the role of affection in the constitution of time, one candescribe it in relation to the context provided by the former moment. The dis-position to action, for instance, plays a key role (Varela 1999). This refers tothe subject’s ongoing active involvement in the world. In our protocol forinstance the subject became accustomed to doing the task. The state of readi-ness then becomes an explicit disposition to quickly press the button. Thismeans the pre-stimulation moment is an active expectation of the next mo-ment, that is the stimulation. Consequently, in clusters of trials with stablereadiness, subjects report a feeling of “continuity” with the pre-stimulationmoment and sometime a feeling of “self-satisfaction”. This suggests that be-cause the stimulation was in accordance with the expectation there was no realemotional shift before the motor response (after it, on the contrary, the feel-ing of satisfaction could induce a change). On the other hand when the sub-ject was not prepared (distraction, tiredness, etc.), the stimulation triggered abreakdown within the ongoing “now”. This is accompanied by a feeling of“surprise” and discontinuity with the former moment as reported by somesubjects.28 There was thus a close relationship between the disposition to ac-tion and the tonality of the next moment, in the subject’s experience.
As I tried to sketch briefly here, genetic analysis of temporality is aboutthe description of the self-movement of consciousness. Emotion, affect, dis-position to action and reflexive awareness are typical distinctions within thisdomain which contains both a spontaneous (passive) and a voluntary side.Their self-referenced features invite us to move beyond the phenomenal iso-morphism and its view of the subject as a passive spectator of his/her cogni-tion. They describe an asymmetrical relation within the phenomenal spherebetween what is experienced as immanent and the experiencing pole. Suchasymmetry could be structurally related to the asymmetrical relationship be-tween local and global processes. In order to reveal such an endogenous rela-tion the collection of these data requires the implication of the subjects as activeand disciplined collaborators describing precisely the “what it is like to be asubject doing the task.”
Dynamical Expression (thread #3) of first-person structural invariants(thread #1)The “translation” of these first-person data into structural dynamical invari-ants (thread #3) is more technical because it borrows concepts from the theoryof non-linear dynamical system (for details see Varela 1999). The central ideais that the dynamical trajectories of global moments are shaping the “dynamicallandscape” of the system into a specific geometry (named phase space).29 Theemergence of the new global state will depend therefore on these specific
156 ANTOINE LUTZ
dynamical constraints (active side of genetic experience) but also of otherboundary conditions such as the ongoing arising of perturbations (either ex-ogenous ones like the contextual setting of the task – i.e., a new stimuli – orendogenous ones – i.e., a spontaneous thought or a somatic event = passiveside of genetic data). It is this instantaneous intertwining of the ongoingperturbations within the transient geometry of the phase space that accountsfor the emergence of a particular large-scale synchrony. The perturbation ofthe system will attract or draw the ongoing dynamic away from a trajectoryor an attractor within this phase space. Kelso and co-workers express a simi-lar view when they propose that the brain is operating in a “metastable” dy-namical regime distinguished by a balance between integrating and segregatinginfluences (for this model and experimental data see Bressler et al. 2001). Thisself-movement of the dynamic system (attraction, bifurcation, repulsion) andthe description of the self-movement of consciousness can thus mutually pro-vide one another with insights. As an illustrative example the disposition toaction (protention), as described during preparation, can be seen as a self-induced shaping of the phase space in order to create dynamical conditionsto coordinate quickly to motor response to the stimulus. The emotional af-fect during surprise could be seen as a reshaping of this space during thesolicitation of the stimulus. In that case the bifurcation of the dynamic andits reorganization to perform the motor response would require a longerrelaxation time. This is in strong accordance with the trajectory of the dy-namical patterns and the longer reaction time that I would like to describenow.
4.3. The neurobiological dimension of generative passages
Finally, the notions of local and global can also have a neurobiological ex-pression (thread #2) based on anatomical and functional distinctions (seeSection 2.1). We therefore chose, in our study, to detect from EEG signals thelocal and long-distance synchrony occurring before and after the task betweenoscillating neural populations (studied from 6–60 Hz). We call this the dynami-cal neural signature (DNS). The heuristic strategy of mutual constraints ledus to study for each subject the DNS of each phenomenological cluster (PhC)of trials. The idea was to seek a common dynamical trajectory stable throughtrials presenting the same preparatory context (Lutz et al. 2002). To instanti-ate further the hypothesis of mutual constraints as generative passages oneneeds first to find a proper dynamical expression to these spatio-temporalpatterns. This part is done now at a speculative level.
157NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
Propositions for dynamical expressions (thread #3) of DNSs (thread #2)I would like to use two of these DNSs as a starting point for a theoretical inter-pretation of the concepts of reciprocal causations. This descriptive analysiscan provide useful intuitions to guide further quantification of these dynami-cal signatures.
In the first PhC (Figure 2, steady readiness cluster) the subject reported thathe was prepared and that he immediately saw the illusion. On the DNS a frontalpattern of synchrony gradually emerged in the gamma band several secondsbefore the stimulus. This contextual activity was still present during percep-tion and motor response and was mixed with fronto-occipital long-distancesynchrony induced by the stimulus. In contrast in the other DNS (Figure 2cluster of unreadiness), when the subject was unprepared and surprised by thearrival of the stimulus, there was no stable pattern in the gamma band (around40 Hz) before the stimulation and transitory neural patterns emerged afterstimulation, in discontinuity with the prestimulation activity. The effect ofsurprise was associated with a different temporal structure in the neural re-sponse, combining phase-scattering30 and an increase in synchrony. The mo-tor response was accompanied by a pattern of synchronies that were spatiallysimilar to that observed during preparation, but were delayed by 300 ms.
If one defines the brain activity evoked by the stimulation in the first 100ms after the stimulation as an elementary event or local event, then this da-tum could be interpreted as follows. In the case of preparation, expectation isan intentional act about the stimulus and is correlated to a large-scale coher-ent activity (partially characterized by this frontal pattern) before stimulation.This large-scale activity could be said to have a downward causation on thelocal, elementary event in the sense that this event appears as integrated, bothdynamically and phenomenally in the pre-existing activity. The pre-exist-ing global pattern could act thus as an order-parameter modulating the lo-cal processes. In contrast, in the case of surprise, no stable patterns are foundon average before stimulation. Therefore, there is a weaker disposition, ei-ther phenomenally or dynamically, to perceive the stimulus. The genetic analy-sis of this cluster suggests that the stimulation induced a transition in theflow of consciousness from one content to another. The neurodynamical andbehavioral correlates offer a striking relation with this description: the effectof surprise was associated with a different temporal structure in the neuralresponse, combining phase-scattering (white-line) and an increase in syn-chrony (black lines).
Phase-scattering thus mediates the necessary transition between two verydistinct neural assemblies, in particular during the adaptive response to a sa-lient change in sensory flow. The motor response was accompanied by a pat-
158 ANTOINE LUTZ
tern of synchronies that were spatially similar to the one observed duringpreparation, but were delayed by 300 ms. In that case it is tempting to say thatthis local event makes the ongoing activity bifurcate onto another global dy-namical trajectory. In that sense it induces a stronger upward causation be-cause it triggers a shift in both the content of the now and the dynamicalpatterns. This was revealed particularly well by the patterns of phase-scatter-ing, which are already found in the transition between two cognitive moments.
4.5. Weaving the three threads of generative passages
This part is conjectural and suggests from the previous analysis a possibleway to explicitly test the assumption of mutual constraints as generativepassages.
In contrast to phenomenal isomorphism this relation expressly mobilizes aformal level (thread #3) and a multi-scale analysis (local/global) within thefirst-person accounts, the biophysical processes, and the dynamical variables.Such relation is triangular and no longer binary. Thus, providing that theneurophenomenological praxis has been implemented and that the relevantmulti-scale dynamical variables have been identified, then the working as-sumption could be tested as follows: trials could be clustered (a) dependingon structural properties expressing a global to local relation as described byphenomenological accounts and (b) depending on dynamical trajectories ex-pressing a reciprocal causation as quantified by the scientific analysis of neu-robiological data. A relation of generative passages could be demonstrated ifthe partitioning of the space of trials in mode (a) significantly overlap thepartitioning of space in mode (b).
In Lutz et al. (2002) we have already demonstrated an overlap betweenphenomenological and behavioral clusters. We then showed significant physi-ological differences in the self-generated activity before stimulation and thebrain responses between the various phenomenological clusters. This resultis very encouraging but is insufficient to constitute, strictly speaking, a gen-erative passage because we did not mobilize explicitly a dynamical level.Further quantification is first needed which may not be possible with thesecurrent data. Indeed, the lack of spatial resolution of the EEG makes it diffi-cult to interpret the synchrony patterns directly in terms of precise neuralnetworks. Also, neural interactions may take multiple forms that are not de-tected by linear measures such as phase-synchrony. Finally the cultivation ofthe reductive gesture and the degree of refinement of first-person accounts mayrequire working with experts. For instance, subjects with long-term training
159NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
in practices rooted in Eastern traditions such as meditation practice could be-come paradigmatic collaborators.
Further improvements are clearly required at these three levels of the neuro-phenomenological accounts. This purpose of this section has been to clarifyand exemplify such strategy.
4.5. Toward a typology of reciprocal causations
This last section attempts to further generalize the distinctions found betweenthese two DNSs.
The first DNS (during steady preparation) expresses an “easy” case ofdownward causation where global to local can be equated to top-down influ-ences on sensory processing. Such data corroborates with several studies whichshow that “expectation” or “anticipation” do indeed affect the temporal struc-ture both of the ongoing activity pattern and the brain responses (for an re-cent review see, Engel, Fries, and Singer 2001). For instance Riehle et al.(1997) showed that, in a delayed reaching task, synchrony occurred particu-larly when the monkey was expecting a GO signal to appear on a screen.Remarkably, in those trials in which the GO signal appeared after prolongedperiods of expectation, the spike synchrony became more precise as the GOcue signal approached. This attests to a relationship between growing stimu-lus expectancy and the synchronization of the neuronal network involved inthe task. Thus, in this study the level of synchrony appears to have predictivepower both on the performance and the reaction times of the animal.
In this first DNS the downward causation can be simply described as fol-lows: the emergence of the act of anticipation corresponds to the constitu-tion of a coherent assembly (upward causation) from which and during whichthe neural activity in sensory regions will be “interpreted”. This notion of“neuronal interpretation” (Chiel 1993; Varela 1997, p. 367) means that thecompetitive process underlying emergence consists either in discarding orintegrating concurrent local activities into the dominant resonant mode. Thisfederation of the neural activity expresses thus a transient “point of view”or a dynamical “background” from which other local activities are “as-sessed”. This “hermeneutics” (Varela 1997, p. 367) and selective dimensionof synchronization offer a striking insight about the processes underlyingthe immanent constitution of temporality. The quality of preparation (steadyor fragmented), for example, can directly provide insights about the selec-tive and competitive process itself of emergence underlying the deploymentof attention.
160 ANTOINE LUTZ
The second DNS (unreadiness with surprise) provides intuitions aboutanother, more sophisticated, class of global/local processes mediating thereflective dimension of experience. As I suggested in the previous section, thisreflexive gesture was induced by an affective reaction following a shift in theintentional content of the experience. This first-person account was related,in the average DNS, to a massive disorganization in synchrony patterns. Thisphase-scattering was concomitant to the emergence after stimulation of syn-chrony patterns between visual and frontal electrodes. This suggests that theemotional shift initiated the emergence of a new large-scale assembly therebyconstituting an original point of view. The experience of one’s own experi-ence could be mediated by the interpretation within this new assembly of lo-cal processes previously implicated in the former dominant assembly. Forexample, if, before the stimulation, the content of my experience was about ameeting I had to attend after the protocol, then some elements of the neuralnetwork involved in this act could be still resonating locally after the pertur-bation generated by the stimulation. After the phase-scattering, these elemen-tary processes might be re-enslaved by the new assembly. This re-enslaving,would mediate, I propose, the experience of reflective awareness. In contrastto the former class of global to local processes the global activity here wouldconstrain and integrate endogenous local processes. This global to local in-teraction is likely to be mediated by the complex interplay between severaloscillatory modes as suggested by the different frequency bands detectedduring top-down and buttom-up interactions (Von Stein, Chiang, and Konig2000). This case of reciprocal causation potentially opens the door to thenaturalizing of the reductive gesture itself, providing that reflection triggeredby an affective disposition contains some of the structural features of the ges-ture of reduction (Depraz et al. 2002).
To summarize, this section has illustrated the difference between the “re-ciprocal causation” theory and a classical top-down/ buttom-up theories(Mumford 1992). The latter emphasize the anatomical organization of cogni-tion whereas the former highlights its dynamical self-coordination. For thelatter, top-down influences correspond to the activity of feedback connectionsin a processing hierarchy, whereas in the former it is the temporal structure oflarge-scale dynamics which influences local processes largely independentlyfrom their location. A similar, central role of collective coherent dynamics hasrecently been presented in Engel, Fries and Singer’s “dynamicist” view of top-down influence (2001) or Kelso and Bressler’s “coordination dynamics” theory(2001). Mutual constraints theory takes this a step further by attempting torelate these third-person concepts to first-person accounts. The first-personaccount of immediate experience produced by a trained subject is more than
161NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
a heuristic: it expresses a hermeneutical and asymetrical process which mustbe fully integrated as part of the neurophenomenological account. In this sensedownward and upward causation are operational concepts for untangling theintertwined relations between local and global processes in order to investi-gate the driving or enslaving effect of an emergent process on others activi-ties and their relations to the phenomenal level.
Conclusion
Neurophenomenology stresses the specific need to start by exploring experi-ence with a rigorous pragmatics before investigating more fully its symbioticrelation to natural entities. A first report from this study is that this strategy isnot a thoughtful but impractical method. On the contrary, it is a doable, fruit-ful and promising approach to the collection of new phenomenal and neuro-dynamical data and the identification of their mutual relationships. Its initialformulation was placed in the lineage of continental phenomenology. This choicecould have scared some readers. Yet what we learnt on the job is that these “tech-nical” distinctions are meaningful and inspiring descriptions as soon as we takethe time to look at them. The phenomenological tradition can thus provide con-ceptual tools, stimulating questions and a style of research. A phenomenologicalreduction was rudimentarily instantiated in our study by training the subject tobe aware of his/her preparatory context and by inducing after each trial a re-flective gesture on his/her just-past experience. Further refinements are obvi-ously needed. A central issue is to get a better instrumental description of thisreductive gesture (see Depraz et al. 2002; Petitmengin-Peugeot 1999).
The core concept of “reciprocal constraints” was defined in this paperthrough its methodological, epistemological and theoretical modes. Recipro-cal constraints were first read as a heuristic approach guided by a praxis. Itwas instantiated (a) by training the subject to find his/her own categories todescribe his/her preparatory context and visual experience and (b) by doing ajoint-analysis of first- and third-person data. This particular approach to mu-tual constraints was implemented through the characterization of the dynamicalneural signatures (DNS) of phenomenological clusters (PhC) of trials.
The second section analyzes and illustrates the limitations of the phenom-enal isomorphism. The risks associated with this static relation are to dismissthe active, embodied and situated side of subjective experience. The relationof “generative passages” is thus an attempt to do justice to the genetic andemergent dimension of experience. It provides a way to better interpenetrateexperience and its natural substrates. Generative passages can be envisagedthanks to (a) a neurophenomenology framework (b) a multi-scale analysis of
162 ANTOINE LUTZ
these phenomena (c) an extension of first-person data to conscious acts andtheir self-manifestation and (d) an explicit introduction of a neural dynami-cal level in the neurophenomenological account.
The question of the relationship between phenomenal isomorphism andgenerative passages requires further thought. On the one hand, generativepassages might simply correspond to a more refined, precise, and accurateversion of phenomenal isomorphism. It would be thus a three-way isomor-phism (i.e., one-one mapping) among generic-genetic (and static) propertiesin the phenomenal, neurally emergent, and formal dynamical domains. On theother hand, generative passages might correspond only to an homomorphicepistemological relation among the phenomenal, the neural, and the dynami-cal (i.e., a reduction of phenomenal structural features when projected ontothe neural-dynamical categories). For instance, it might not be possible to mapall of the structural-phenomenal features in the neural domain, because thenatural domain would need to be broadened beyond the brain to include theorganism as a whole (and the organism’s environment).31 These questions re-quire further empirical investigation.
In Memoriam
This article is dedicated to Francisco Varela in memory of his innovatingwork, generous teaching and inspiring life. For an obituary seehttp://psyche.csse.monash.edu.au/v7/psyche-7-12-thompson.html.
Acknowledgements
For discussion I am grateful to the Neurodynamic Group of the LENA (CNRSUPR 640, Paris), Evan Thompson and the CREA (Ecole Polytechnique, Paris),in particular Jean Petitot, Michel Bitbol, Jean-Michel Roy and Bernard Pachoud.Diego Cosmelli, Shaun Gallagher and, very specially, Amy Cohen-Varelaprovided clarifying remarks and corrections of the text. Very special acknowl-edgments are due to Natalie Depraz for her essential teaching and constantintellectual support through this work.
Notes
1. In Husserlian terms, the issue consists in the possibility of grounding the field of purephenomenology, as accessed by the reduction (Husserl 1913, §75), in a psychologicalphenomenology (Husserl 1925), or even in the natural realm (Husserl 1918–26). The
163NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
idea (Husserl 1913, §57) is that once the constitution of the natural object is sufficientlythematized in the phenomenal domain, it can also be viewed as part of psychologicalconsciousness and therefore as part of an organism. This claim will be elaborated through-out this text suggesting that it does not contradict Husserl’s anti-naturalist arguments.Such a position still expresses a non-reductive way to study experience. It is stronglyinspired by Husserl’s later work on the foundational and constitutive role of the bodyand the world in relation to the world of spirit. See his work on the lived-body (Leib)and the life-world (Lebenswelt) in Husserl (1966). And see Depraz (1999) for furthercontemporary developments of this delicate topic.
2. A phenomenon, as it is used since the Presocratics, describes the fact that somethingappears for someone. Therefore it is something relational. It is a being for in contrast toa being in itself independent of its apprehension by another point of view. In the analyti-cal literature this distinction has been captured by Thomas Nagel’s phrase “what it islike” to experience something (1970).
3. For a detailed review see for instance Roy et al. (1999).4. The term “attribute” subscribes tacitly to a formal description of subjectivity as a third-
person object. It is usually used as such in the analytic tradition. We prefer here the term“phenomenal account” which seems more appropriate to capture the problem of “whatit is like” or “what it is to act as a subject.”
5. Varela (1996), or for a collection of papers see the contemplative and phenomenologicalsection in Petitot et al. (1999) and Varela and Shear (1999).
6. See the introspection section in Varela and Shear (1999).7. For a cross-disciplinary study of this topic see Depraz et al. (2002).8. For a more elaborate philosophical discussion see Hanna and Thompson (in press).9. See Varela (1997) for the move to a dual status of neurophenomenology as a method
and an object of investigation. See Varela (1999) for the naturalization in itself of time-consciousness.
10. (a) Local synchronization (time-frequency power emission) occurs in EEG when neu-rons recorded by a single electrode transiently oscillate at the same frequency with acommon phase: their local electric fields add up to produce a burst of oscillatory powerin the signal reaching the electrode. By averaging such emissions across successiveresponses to repeated stimulations, we can estimate the latencies and frequencies atwhich bursts are likely to occur. (b) Long-distance synchrony can occur when two neu-ral populations recorded by two distant electrodes oscillate with a precise phase-relation-ship that remains constant during a certain number of oscillation cycles. The emergenceof such large-scale neural assemblies is believed to result from long-range interactionsbetween neural populations and may mediate the large-scale integration between func-tionally distinct neural processes. This method estimates for each pair of electrodes thestability in time of their phase difference at a given frequency. This provides a measureof the raw synchrony for a given electrode pair and trial (Lachaux 1999).
11. The experimentalist plays here the role of a mediator who incites the subject to describehis (even pre-reflexive) experience. The formulation of questions is crucial to invite thesubject to come back to his/her direct experience. Open questions are used to preventhim/her from theorizing the experience. For example: Experimenter: “What did you feelbefore and after the image appeared? Subject S1: “I had a growing sense of expectation,but not for a specific object; however when the figure appeared, I had a feeling of con-firmation, no surprise at all”; or subject S4: “it was as if the image appeared in the pe-riphery of my attention, but then my attention was suddenly swallowed up by the shape.”
164 ANTOINE LUTZ
12. It is a sufficient number of trials to test the hypothesis that the studied variable is notrandom.
13. In most trials, subjects reported that they were “ready”, “present”, “here”, “well-pre-pared” when the image appeared on the screen and that they responded “immediately”and “decidedly”. Perception was usually experienced with a feeling of “continuity”, “con-firmation” or “satisfaction”. These trials were grouped into a cluster (SR), characterizedby the subjects being in a state of “steady readiness”.
14. In other trials, subjects reported that they had made a voluntary effort to be ready, butwere prepared either less “sharply” (due to a momentary “tiredness”) or less “focally”(due to small “distractions”, “inner speech” or “discursive thoughts”). The emergenceof the 3D image was experienced with a small feeling of surprise or “discontinuity”. Thesetrials formed a second cluster (FR) corresponding to a state of “fragmented readiness”.
15. In the remaining trials, subjects reported that they were unprepared and that theyonly saw the 3D image because their eyes were correctly positioned. They were sur-prised by it and reported that they were “interrupted” by the image in the middle ofa thought (memories, projects, fantasies etc.). This state of distraction occurred spon-taneously for S1 and S4, whereas S2 and S3 triggered it either by fantasizing or bythinking about plans (subject 3) or by visualizing a mental image (subject 2). Toseparate passive and active distraction, these trials were divided in two differentclusters, Spontaneous Unreadiness (SU) for S1 and S4 and Self-Induced Unreadiness(SIU) for S2 and S3.
16. The notion of “hard problem” is used here simply as a label. A recent questioning of suchphilosophical distinctions can be found in Hanna and Thompson (in press).
17. For an example of a mathematization of perceptual eidetics see Petitot 1999, pp. 330–371.
18. This study however simply showed a correlative link.19. Hanna et al. (forthcoming) developed more technically a similar argument.20. On the Phenomenology of Internal Time Consciousness (Husserl E. p. 291).21. For a critical evaluation of these approaches see e.g., Beer (2000) or Van Gelder (1998).22. An order parameter is a collective variable which describes adequately the behavior of
a population of variables with a much smaller degree of freedom than the one of the wholepopulation if every variable was independent. It is a central concept in the field ofsynergetics (Haken 1996).
23. Local-to-global process manifests through the dynamical interacting variables, whereasthe effect of global-to-local manifests through modifications in boundary conditions andcontrol parameters. For an example see Kelso (1995, p. 6).
24. Local/global scales of analysis are similar to the mesoscopic/macroscopic scales of re-cording used by several authors such as Edelman (1987) or Freeman (1999). Microscopicscale refers to the analysis of single-neuron activity. Mesoscopic scale refers to the re-cording of coordinate behavior of local neuron groups as measured by local field poten-tial. Macroscopic scale refers to extracortical recordings.
25. For a discussion of this point see for instance Varela (1997, p. 375).26. This region is referred to in the phenomenological tradition as the lived body (Leib)
(Husserl 1912–28) which is this “here”, this encompassing point of view from whereeverything which appears takes place in a “there”. What appears as transcendent can eitherbe objects in the world or my own body viewed as a physical entity (Körper). But if naturalevents, upon which science is based, always occur within the purview of the preexistent
165NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
lived body (Leib), this former is ultimately grounded in the Körper, that is my body un-derstood as biological processes. This fundamental duality of appearance is at the heartof the theoretical issue of reciprocal constraints.
27. “Ce n’est pas moi qui prends l’initiative de la temporalité, (. . .) le temps fuse à traversmoi, quoi que je fasse” (Merleau-Ponty 1958, p. 489).
28. A panoply of emotions can occur: disappointment, anxiety or amusement depending onthe degree of commitment to do the task.
29. This trajectory can be modeled as an order parameter. See Freeman (1999) or Varela(1999)
30. Phase-scattering is defined as a significant decrease of the synchrony measure comparedto the average synchrony in a baseline.
31. I am indebted to Evan Thompson for a discussion about this issue.
References
Abeles, M. 1982. Local Cortical Circuits. Berlin: Springer.Acebron, J. A., Perales, A., and Spigler, R. 2001. Bifurcations and global stability of syn-
chronized stationary states in the Kuramoto model for oscillator populations. Physical RevE Stat Phys Plasmas Fluids Relat Interdisciplinary Topics 64: 016218.
Arbib, M. A. 1995. Handbook of Brain Theory and Neural Networks. Cambridge, MA: MITPress.
Bressler, S. L. and Kelso, J. A. 2001. Cortical coordination dynamics and cognition. Trendsin Cognitive Science 5: 26–36.
Chalmers, D. J. 1996. The Conscious Mind: in Search of a Fundamental Theory. Oxford:Oxford University Press.
Churchland, P. S. and Sejnowski, T. 1992. The Computational Brain. Cambridge, MA: MITPress.
Clark, A. 1997. Being There: Putting Brain, Body and World Together Again. Cambridge,MA: MIT Press.
Depraz, N. 1994. Temporalité et affection dans les manuscrits tardifs sur la temporalité (1929–1935) de Husserl. Alter 2: 183–247.
Depraz, N. 1998. Can I anticipate myself? Self-affection and temporality. In: D. Zahavi (ed),Self-Awareness, Temporality and Alterity, pp. 83–99. Dordrecht: Kluwer.
Depraz, N. 1999. When Transcendental Genesis Encounters the Naturalization Project. In: J.Petitot, F. Varela, B. Pachoud and J. M. Roy (eds), Naturalizing Phenomenology, pp. 464–483. Stanford, CA: Stanford University Press.
Depraz, N., Varela, F., and Vermersch, P. 2002. On Becoming Aware: The Pragmatics ofExperiencing. Amsterdam: John Benjamins Press.
Edelman, G. 1987. Neural Darwinism. New York: Basic Books.Engel, A. K. and Singer, W. 2001. Temporal binding and the neural correlates of sensory
awareness. Trends in Cognitive Science 5: 16–25.Engel, A. K., Fries, P., and Singer, W. 2001. Dynamic predictions: oscillations and synchrony
in top-down processing. Nat Rev Neuroscience 2: 704–716.Ericsson, K. A. and Simon, H. A. 1993. Protocol Analysis, Verbal Protocols as Data. Cam-
bridge, CA: MIT Press.Flanagan, O. 1994. Consciousness Reconsidered. Cambridge, MA: MIT Press.Fodor, J. 1984. The Modularity of the Mind. Cambridge, MA: MIT Press.
166 ANTOINE LUTZ
Freeman, W. 1999. How Brains Make Up Their Minds. London: Wiedenfeld & Nicholson.Haken, H. 1996. Principles of Brain Functioning: A Synergetic Approach to Brain Activity,
Behavior and Cognition. Berlin: Springer.Hanna, R. and Thompson, E. (forthcoming). The Spontaneity of Consciousness: Neuro-
phenomenology and Modal Metaphysics.Hebb, D. O. 1949. The Organization of Behavior. New York: Wiley-Interscience.Husserl, E. 1913. In: W. Biemel (ed), Ideen zu einer reinen Phänomenologie und phän-
omenologischen Philosophie, Book 1: Allgemeine Einführung in die reine Phänomenologie.The Hague: M. Nijhoff, 1950.
Husserl, E. 1928. On the Phenomenology of Internal Time Consciousness [Vorlesungenzur Phänomenologie des inneren Zeitbewusstseins 1893–1917], trans. J. Barrett Brough.Dordrecht: Kluwer Academic Publishers, 1991.
Husserl, E. 1952. In: W. Biemel (ed), Ideen zu einer reinen Phänomenologie und phänomen-ologischen Philosophie, Book 2: Phänomenologische Untersuchungen zur Konstitution,1912–28. The Hague: M. Nijhoff.
Husserl, E. 1966. In: M. Fleischer (ed), Analysen zur passiven Synthesis, 1918–26. The Hague:M. Nijhoff.
Husserl, E. 1977. Phenomenological Psychology. Lectures, Summer Semester 1925, trans. J.Scanlon. The Hague: M. Nijhoff.
Husserl, E. 1988. Cartesians Meditations: Introduction to Phenomenology, trans. D. Cairns.Dordrecht: Kluwer Academic Publishers.
Jackendoff, R. 1987. Consciousness and the Computational Mind Cambridge, MA: MIT Press.Julesz, B. 1971. Foundations of Cyclopean Perception Chicago: University of Chicago Press.Kelso, J. A. S. 1995. Dynamical Patterns: The Self-organization of Brain and Behavior.
Cambridge, MA: MIT Press.Köhler, W. 1947. Gestalt Psychology: An Introduction to New Concepts in Modern Psychol-
ogy. New York: Liveright.Kopell, N., Ermentrout, G. B., Whittington, M. A., and Traub, R. D. 2000. Gamma rhythms
and beta rhythms have different synchronization properties. Proceedings of the NationalAcademy of Sciences USA 97: 1867–1872.
Lachaux, J. P., Rodriguez, E., Martinerie, J., and Varela, F. J. 1999. Measuring phase syn-chrony in brain signals. Human Brain Mapping 8: 194–208.
Levine, J. 1983. Materialism and qualia: the explanatory gap. Pacific Philosophical Quar-terly 64: 354–361.
Lutz, A., Lachaux, J.-P., Martinerie, J., and J. F., Varela. 2002. Guiding the study of braindynamics using first-person data: Synchrony patterns correlate with on-going consciousstates during a simple visual task. Proceedings of the National Academy of Sciences U SA: in press.
McClelland, J. L., Rumelhart, D. E., and Group, T. P. R. 1986. Parallel Distributed Process-ing, Vol. 2: Psychological and Biological Models (cited as Parallel Distributed Process-ing, Vol. 2: Psychological and Biological Models). Cambridge, MA: MIT Press.
Merleau-Ponty, M. 1945/1962. Phenomenology of Perception, trans. C. Smith. London:Routledge and Kegan Paul. Original publication: Phenomenologie de la perception. Paris:Gallimard.
Mumford, D. 1992. On the computational architecture of the neocortex. II. The role of cortico-cortical loops. Biological Cybernetics 66: 241–251.
Nagel, T. 1970. What is it like to be a bat? Philosophical Review 79 : 394–403.
167NEUROPHENOMENOLOGY AS AN ACCOUNT OF GENERATIVE PASSAGES
Pessoa, L., Thompson, E., and Noe, A. 1998. Finding out about filling-in: a guide to percep-tual completion for visual science and the philosophy of perception. Behavioral and BrainSciences 21: 723–748; discussion 748–802.
Petitmengin-Peugeot, C. 1999. The Intuitive Experience. In: F. Varela and J. Shear (eds), TheView from Within, pp. 43–77. Exeter: Imprint Academic.
Petitot, J., Varela, F., Pachoud, B., and Roy, J. M. (eds). 1999. Naturalizing Phenomenology.Stanford, CA: Stanford University Press.
Pylyshyn, Z. W. 1984. Computation and Cognition. Cambridge, MA: MIT Press.Riehle A, Grun S, Diesmann M, and Aertsen A. 1997. Spike synchronization and rate modu-
lation differentially involved in motor cortical function. Science 278: 1901–1902.Rodriguez, E., George, N., Lachaux, J.P., Martinerie, J., Renault, B., and Varela, F. J. 1999.
Perception’s shadow: long-distance synchronization of human brain activity. Nature 397:430–433.
Roy, J. M., Petitot, J., Pachoud, B., and Varela, F. 1999. Beyond the gap. An introduction tonaturalizing phenomenology. In: J. Petitot, F. Varela, B. Pachoud, and J. M. Roy (ed),Naturalizing Phenomenology, pp. 1–80. Stanford, CA: Stanford University Press.
Searle, J. R. 1992. The Rediscovery of the Mind. Cambridge, MA: MIT Press.Shear, J., ed. 1997. Explaining Consciousness: The Hard Problem. Cambridge, MA: MIT
Press.Singer, W. and Gray, C. M. 1995. Visual feature integration and the temporal correlation
hypothesis. Annual Review of Neuroscience 18: 555–586.Thompson, E. and Varela, F. J. 2001. Radical embodiment: Neural dynamics and conscious-
ness. Trends in Cognitive Science 5: 418–425.van Gelder, T. 1998. The dynamical hypothesis in cognitive science. Behavioral and Brain
Sciences 21: 615–628; discussion 29–65.Varela, F. J. 1995. Resonant cell assemblies: A new approach to cognitive functions and neu-
ronal synchrony. Biological Research 28: 81–95.Varela, F. 1996. Neurophenomenology: a methodological remedy to the hard problem. Jour-
nal of Consciousness Studies 3: 330–350.Varela, F. 1997. The naturalization of phenomenology as the transcendence of nature. Alter
5: 355–381.Varela, F. and Shear, J. (eds). 1999. The View from Within: First-person Approaches to the
Study of Consciousness. Exeter: Imprint Academic.Varela, F., Thompson, E., and Rosch, E. 1991. The Embodied Mind: Cognitive Science and
Human Experience. Cambridge: MA: MIT Press.Varela, F., Lachaux, J. P., Rodriguez, E., and Martinerie, J. 2001. The brainweb: phase syn-
chronization and large-scale integration. Nat Rev Neuroscience 2: 229–239.Vermersch, P. 1994. L’Entretien d’Explicitation. Paris: ESF.von Stein, A., Chiang, C., and Konig, P. 2000. Top-down processing mediated by interareal
168 ANTOINE LUTZ