AG Nonlinear Dynamics, University of Potsdam, Germany
We present a general method to obtain a set of ordinary differential equations from time series. A full nonlinear analysis of the so obtained dynamical system is possible. The power of the approach is demonstrated by numerical and experimental data.
Neurobiology and Biophysics, Albert-Ludwigs-University Freiburg, Germany
Studies of cortical network function on the basis of multiple single-neuron recordings have revealed neuronal interactions which depend on stimulus and behavioral context. These interactions exhibit dynamics on several different time scales, with time constants down to the millisecond range. Mechanisms underlying such dynamic network organization are investigated by experimental and theoretical approaches. Our current research focuses on two interrelated aspects: precision and variability of cortical network activity 1,2 . Extending previous model work 3 in which we investigated conditions for the occurrence of precise joint-spiking events in cortical network activity, I will present recent findings from ongoing experimental and theoretical work in our laboratory 4-8 , undertaken to test and expand some of the model predictions. Specifically, I will discuss new findings regarding the feasibility and constraints of precise synchronization dynamics in cortical networks, resulting from a critical evaluation of biological constraints from cortical connectivity and in vivo physiology, and dynamical constraints from large-scale network simulations.
References:
1. Riehle A, Grün S, Diesmann M, Aertsen A (1997) Spike
synchronization and rate modulation differentially involved in
motor cortical function. Science 278:1950 1953
2. Arieli A, Sterkin A, Grinvald A, Aertsen A (1996) Dynamics of
ongoing activity: explanation of the large variability in evoked
cortical responses. Science 273:1868 1871
3. Diesmann M, Gewaltig MO, Aertsen A 1999 Stable propagation of
synchronous spiking in cortical neural networks. Nature 402:529
533
4. Kuhn A, Rotter S, Aertsen A (2003) Higher-order statistics of
input ensembles and the response of simple model neurons. Neural
Computation 15: 67-101
5. Mehring C, Hehl U, Kubo M, Diesmann M, Aertsen A (2003)
Activity dynamics and propagation of synchronous spiking in
locally connected random networks. Biol Cybern 88: 395-408
6. Kuhn A, Aertsen A, Rotter S (2004) Neuronal integration of
synaptic input in the fluctuation-driven regime. J Neurosci 24:
2345-2356
7. Léger J-F, Stern EA, Aertsen A, Heck D (2005) Synaptic
integration in rat frontal cortex shaped by network activity. J
Neurophysiol 93: 281-293
8. Morrison A, Mehring C, Geisel T, Aertsen A, Diesmann M (2005)
Advancing the boundaries of high connectivity network simulation
with distributed computing. Neural Computation (in press)
Research Unit "Conflicting Rules" and AG Nonlinear Dynamics, University of Potsdam, Germany
An approach to multivariate phase synchronization analysis introduced earlier (Allefeld & Kurths 2004) is applied to event-related potentials obtained with a cognitive paradigm. The experimental design (following Kutas & Hillyard 1980) combines a semantic incongruity in a sentence context with a physical mismatch (color change). In the ERP average, these result in an N400 component and a P300-like positivity, respectively. The synchronization analysis shows an effect of global desynchronization for the semantic incongruity, while the physical mismatch elicits an increase of global synchronization. Both of these effects clearly precede those in the ERP average. A comparable measure of global coherence reveals only part of the effects visible via the synchronization measure. For the physical mismatch, the global synchronization effect can be traced back to an interaction between left and right occipito-parietal areas.
Commission for Scientific Visualization, Austrian
Academy of Sciences, Vienna, Austria
*lecturer
During cognitive functioning, multiple cortical areas may not only become coactive but also functionally connected and this is manifested in EEG signals. These functional couplings are transient (duration in the order of hundreds of ms), dynamic (time varying nature of the strength of association between two cortical regions), non-random (selective participation of any cortical region at any moment of time), and directed (asymmetric association). These features cannot easily be detected by traditional measures of synchronization, such as correlation or magnitude squared coherence nor by nonlinear measures which are based on generalized synchronization. In this study, we consider the partial directed coherence (PDC). The PDC reflects a frequency-domain representation of Granger causality (Granger 1969), which is a more fundamental concept of finding directional association. This measure is able to detect asymmetrical coupling from multivariate time series with limited number of time samples. We initially evaluated the performances of PDC measure by simulated examples. We then applied this measure on recorded multivariate EEG signals during an alternative forced choice task involving preferential or selective decision. The process leading to such a decision is an active, short-term process in which the brain uses a circuit that includes orienting behavior (Shimojo, Simion et al. 2003), particularly when the considered task involves attractiveness. Distinct, long-range posterior-anterior associations were formed prior to the preferential decision but such associations were significantly weaker for the selective decision.
References:
Granger, C. W. J. (1969). "Investigating causal relations by
econometric models and cross spectral methods." Econometrica 37:
424-438.
Shimojo, S., C. Simion, et al. (2003). "Gaze bias both reflects
and influences preference." Nat Neurosci 6(12): 1317-22.
AG Nonlinear Dynamics, University of Potsdam, Germany
Advances in the nonlinear dynamics and coupled oscillators theory have created a solid basis for the study of different nonlinear dynamical aspects of interacting neuronal oscillators. Here we present a novel approach for joint estimation of time delays and directionality of coupling in noisy limit cycle oscillators and/or chaotic oscillators from time series analysis. The theoretical description is complemented by tests on biophysically motivated neural oscillator models.
University of Potsdam, Germany
* Catholic University of Eichstaett
+ Free University of Berlin
Event-related brain potentials (ERPs) were recorded as 48 subjects read silently 144 German sentences, which were presented word by word. The influence of frequency and predictability of words on ERPs was investigated. The word frequency effect is considered an upper limit for the time range of lexical access. Results of recent studies indicate that lexical access takes place within the first 190 ms after a word has been presented (e.g. Hauk & Pulvermüller, 2004). Thus, modulations of the ERPs evoked by word frequency within this time range are taken as a marker for lexical access in the present study. Word predictability reflects the probability with which the next word in a sentence can be predicted correctly. This measure is similar to the cloze probability utilized in several studies investigating the N400 component. The N400 is assumed to reflect integration processes of a stimulus into the context of a sentence (Kutas & Hillyard, 1980). In addition to the cloze probability also word frequency was shown to affect the N400 amplitude (Van Petten & Kutas, 1990). The role and interplay of both frequency and predictability in the N400 component were investigated in the present study.
References:
Hauk, O., & Pulvermüller, F. (2004). Effects of word length
and frequency on the human ERP. Clinical Neurophysiology,
115,1090-1103.
Kutas, M., & Hillyard, S. A. (1980). Event-related brain
potentials to semantically inappropriate and surprisingly large
words. Biological Psychology, 11, 99-116.
Van Petten, C., & Kutas, M. (1990). Interactions between
sentence context and word frequency in event related brain
potentials. Memory & Cognition, 18, 380-393.
Research Unit "Conflicting Rules", Institute of Linguistics, University of Potsdam, Germany
Using event related brain potentials (ERPs), we investigated in two studies the failure to license a negative polarity item (NPI) in German. Saddy et al. (2004) reported an N400 component when the NPI was not accurately licensed by negation. Drenhaus et al. (2004) considered additionally the influence of constituency of the licensor in NPI constructions. A biphasic N400-P600 response was found for the two induced violations (the lack of licensor and the inaccessibility of negation in a relative clause). In this paper, we investigate the presence and absence of the P600 components in both studies by using a recently developed method on ERP data analysis (beim Graben & Kurths 2003), the symbolic resonance analysis (SRA). The SRA revealed an effect in the P600 time window for the data in Saddy et al. which was not found by using the average technique. The SRA of the ERPs in Drenhaus et al. showed that the P600 components are distinguishable concerning the amplitude. It was smaller in the condition where the licensor is inaccessible, compared to the condition without negation in the string. Our findings suggest that the failure in licensing NPIs is not exclusively related to semantic integration costs (N400). The elicited P600 components reflect differences in syntactic repair. Our results confirm and replicate the effects of the traditional voltage average analysis and shows that the SRA is a useful tool to reveal and pull apart ERP differences which are not evident using the traditional voltage average analysis.
NeuroPhysics Group, Physics Department, Philipps-University Marburg, Germany
I will present a review of our work on multiple microelectrode recordings from the visual cortex of monkeys and subdural recordings from humans - related to the potential underlying neural mechanisms [1]. The former hypothesis of object representation by synchronization in visual cortex (or more generally: of flexible associative processing) has been supported by our recent experiments in monkeys. They demonstrated local synchrony among rhythmic or stochastic γ-activities (30-90Hz) and perceptual modulation according to the rules of figure-ground segregation [2]. However, gamma-synchrony in primary visual cortex is restricted to few millimeters, challenging the synchronization hypothesis for larger cortical object representations. We found that the spatial restriction is due to gamma-waves, traveling in random directions across the object representations [3]. It will be argued that phase continuity of these waves can support the coding of object continuity. Interactions across still larger distances, among cortical areas in human data, involved amplitude envelopes of gamma-signals [4]. Potentially underlying neural mechanisms are proposed, based on models with spiking neurons: (i) Fast inhibitory feedback loops can generate locally synchronized gamma-activities. (ii) Hebbian learning of lateral and feed forward connections with distance-dependent delays can explain the stabilization of cortical retinotopy, the limited cortical extent of gamma-synchronization, the occurrence of gamma-waves, and the larger receptive fields at successive levels of visual cortical processing [5]. (iii) Slow inhibitory feedback can support figure-ground segregation [1]. (iv) Temporal dispersion in far projections destroys coherence of high frequency signals but preserves slow amplitude modulations [1,4]. In conclusion, it is proposed that the hypothesis of flexible associative processing by gamma-synchronization, including coherent representations of visual objects, has to be extended to more general forms of signal coupling.
References:
[1] Eckhorn R, Gail A, Bruns B, Gabriel A, Al-Shaikhli B,
Saam M (2004) Different types of signal coupling in the visual
cortex - related to neural mechanisms of associative processing
and perception. IEEE-Trans Neur Networks (in press)
[2] Gail A, Brinksmeyer HJ, Eckhorn R (2000) Contour
decouples gamma activity across texture representation in monkey
striate cortex. Cerebral Cortex 10: 840-50
[3] Gabriel A, Eckhorn R (2003) A multi-channel correlation
method detects traveling γ-waves in monkey visual cortex. J
Neurosci Meth 131: 171-184
[4] Bruns A, Eckhorn R (2004) Task-related coupling from
high- to low-frequency signals among visual cortical areas in
human subdural recordings. Int J Psychophysiol 51: 97-116
[5] Saam M, Eckhorn R (2000) Lateral spike conduction
velocity in visual cortex affects spatial range of
synchronization and receptive field size: A learning model with
spiking neurons. Biol Cybernetics 83: L1-L9
The Wellcome Trust Laboratory for MEG Studies,
Neurosciences Research Institute,
Aston University, Birmingham, United Kingdom
Mean field gamma oscillations have been suggested as a mechanism for binding together information from specialized neurons that code for specific stimulus features (Gray, Koenig, Engel and Singer, 1989, Koenig, Engel, Singer 1995). However, the relationship between oscillations occurring in macroscopic mean field signals such as the EEG and MEG and the neuronal synchronization process occurring at microscopic and mesoscopic scales is not clear. The anatomy of the cortex at mesoscopic scales suggests local coupling (Hellwig, 2000) of dynamically diverse functional units, the cortical columns (Mountcastle, 1997). We present an abstract model of nonidentical, locally coupled chaotic oscillators, which predicts emergence of phase-synchronized clusters at characteristic frequencies. These give rise to frequency specific power changes in the macroscopic mean field. We hypothesize that spatial stimuli, which require integration of activity across distributed cortical columns in primary visual cortex, would be accompanied by the emergence of frequency-specific mean field oscillations. We present a methodology to assess stimulus related changes in the frequency content of the MEG signal arising from a source located in primary visual cortex. We show that different spatial stimuli are accompanied by different temporal frequencies of induced gamma oscillations. Further, we address the bursting character of these oscillations. For the most salient stimuli the distribution of the interburst intervals conforms to a power law-scaling characteristic for dynamic on-off intermittency (Heagy, Platt and Hammel, 1994) and is significantly different to the one obtained in surrogate data, where possible nonlinear structure is destroyed (Theiler et al, 1992). The results suggest that spatial stimuli may be coded in terms of overlapping receptive fields, implemented by interacting synchronized clusters within the gamma band.
References:
Gray, C.M., Koenig, P., Engel, A.K. and Singer,W. (1989).
Oscillatory responses in cat visual cortex exhibit intercolumnar
synchronization which reflects global stimulus properties.
Nature, 338, 334-337
Heagy JF, Platt N, Hammel SM. (1994). Characterization of on-off
intermittency. Phys Rev E. 49(2):1140-1150
Hellwig B. (2000). A quantitative analysis of the local
connectivity between pyramidal neurons in layer 2/3 of the rat
visual cortex, Biol. Cybern., 82:111-121
Konig P, Engel AK, Singer W. (1995). Relation between oscillatory
activity and long-range synchronization in cat visual cortex.
PNAS USA. 92(1), 290-4
Mountcastle VB. (1997). The columnar organization of the
neocortex. Brain. 120 (4): 701-22
Theiler J., Eubank S., Longtin A., Galdrikan B., and Farmer J. D.
(1992). Testing for nonlinearity in time series: The method of
surrogate data. Physica D, 58:77-94
Chair Biological Psychology, Otto-von-Guericke-University Magdeburg, Germany
Electrophysiological activity in the gamma-band (approx. 30-80 Hz) has been claimed to represent cognitive processes in animals as well as humans. We investigated how exogenous stimulus parameters and cognitive processes modulate this type of oscillations by computing time-frequency analyses based on wavelet decompositions. We were able to show that larger visual stimuli evoked gamma-band responses of higher amplitude than smaller stimuli. This result probably reflects the fact that the activity originates from a retinotopic brain region as it is the case for human V1 and V2. At the same time the responses to large stimuli revealed a lower frequency. This makes sense, since activity in response to large stimuli needs to be integrated across a more distributed cortical area and the lower the frequency the less likely it is that neural activity cancels out due to temporal imprecision. In addition, stimuli that were shown in the periphery evoked responses of lower amplitude than those presented foveally. This might be taken as vidence for a representation where neurons coding peripheral space are located more distant from recording electrodes. This is known to be true for human V1 and V2 along the calcarine fissure. Varying the duration of visual stimuli revealed that both onset and offset of the stimuli evoked gamma responses while induced gamma activity was only present after the offset when task-related processing was already finished. In addition to these exogenous modulations we were able to show that certain cognitive factors modulate evoked gamma responses. We believe memory access to be the main reason for such modulations and will present a neural model which explains amplitude changes by means of feedback projections within visual cortex.
Applied Stochastic Processes, Humboldt University at Berlin
The present work introduces an analysis framework for the detection of metastable signal segments in multivariate time series. It is shown that in case of linear data these segments represent transient mutual synchronization, while metastable segments in circular data reflect transient mutual phase synchronization. We propose a single segmentation approach for both types of data considering the space-time structure of the data. Applications to single evoked potentials reveal latency jitters and allow a comparison of mutual synchronization and mutual phase-synchronization.
Institute for Mathematics, University of Lübeck, Germany
Dynamical systems can be considered on different levels of precision. For studying the qualitative behaviour of a system, it can be useful to choose a very coarse-grained description: The state space is divided into a small number of subregions, each coded by a symbol. In a discrete-time dynamical system, sequences of successive states of the system are turned into symbol sequences. Symbolic dynamics studies dynamical systems on the base of the symbol sequences obtained for a suitable partition of the state space. We apply the ideas of symbolic dynamics to the qualitative analysis of EEG time series. The way we go is to consider the delay embedding of a time series for some given delay and dimension and to choose a division of the embedding space focussing to the ordinal structure of a the time series. Here we follow an idea of Bandt and Pompe, who have introduced the ordinal viewpoint into time series analysis. In particular, we present new methods for detecting, visualizing and analyzing temporal and spatial qualitative changes of brain states related to epileptic activity.
Fraunhofer FIRST, Berlin
Electroencephalogram (EEG) recordings give information about electrical activity in the brain. However, EEG is often distorted by other signal sources of the head. A common type of these interferences are DC shifts, which are caused by eye movements and blinks. Electrical fields originating from the eyes propagate across the scalp, deteriorating the quality of, e.g., event-related potentials (ERP). Various approaches have been suggested to correct the EEG signal, e.g. Principal Component Analysis and Independent Component Analysis. Since the Electrooculogram (EOG) measures the movement of the eyes, the measured EOG (mEOG) can be used to correct the measured EEG signal (mEEG) by regression. If, however, mEOG is corrupted by additive noise, regression approaches are prone to mixing this noise into the EEG. Kalman filters are a well-established form of Linear Gaussian state-space equations. In the method presented here, we use the Kalman Filter equations to separate EOG and EEG from the measured mixture of these signals. Additionally, the Kalman Filter explicitly models the additive noise in mEOG and mEEG, thus it has the potential to be more robust as compared to simple regression approaches.
Institut Fédératif des Neurosciences de Lyon, France
While recording the intracranial EEG of epileptic patients involved in a reading task, we found that the analysis of each word was simultaneous with specific energy increases in the 50-150 Hz EEG. Those high-frequency (HF) energy increases were short lasting and distributed selectively in the fusiform gyrus, the superior temporal gyrus, the Broca area and possibly the Wernicke area . In addition, the amplitudes of those HF increases proved to be exquisitely dependent on the level of attention of the patients. We will discuss the anatomical organization of those energy increases and their relative timing to propose a dynamical model of the (some) neural processes underlying reading.
Laboratoire de Neurosciences Cognitives et
Imagerie Cérébrale, LENA,
CNRS UPR 640, Hôpital de la Pitié-Salpêtrière, Paris, France
Epilepsy is an intermittent brain disorder characterized by the spontaneous occurrence of seizures. Despite considerable progress in the pathophysiology of epilepsy, the mechanisms involved in triggering epileptic seizures are still poorly understood. Since the synchronization of neuronal activity is an essential feature of epilepsy, the quantification of transient phase-locking between different cortical areas, ubiquitous during normal physiological conditions such as sensory-motor processing, can give a better insight into subtle dynamic interactions between epileptic and normal brain regions. We have recently examined the degree of phase synchronization in long-term intracranial EEG recordings of 5 patients. These data were part of an international database (1st International Conference on Seizure Anticipation, Bonn 2002), containing continuous multi-channel intracranial EEG signals. With a statistical mesure of phase synchrony, we examined the synchronies between all possible combinations of electrode pairs in 15 frequency bands (2 Hz steps between 0 and 30 Hz) and detected quantifiable spatial or temporal shifts in preictal synchronization far in advance of seizure onset detectable on the EEG. Because normal EEG synchronization is enormously varied, manifesting qualitative changes depending on behavioral state, we paid close attention to potentially confounding factors for seizure anticipation such as normal behavioral state changes. Toward this end, a classification technique was used to identify if a synchronization pattern is likely or not to be a member of a library of synchronization reference patterns representing long periods of interictal activity. In most of the cases (36/52 of all seizures, i.e., 70%), a specific state of brain synchronization was observed several hours before the actual seizures. These changes in synchrony preferentially concerned cortical areas close to the epileptic focus. In accordance with other observations, these findings demonstrated that phase synchronization is capable of extracting information from the EEG that allow the definition of a preictal state. This opens new perspectives for the anticipation of epileptic seizure, as well as for the study of the underlying neurophysiology.
References:
M. Le Van Quyen, J. Soss, V. Navarro, R. Robertson, M. Chavez, M.
Baulac, J. Martinerie, Preictal state identification by
synchronization changes in long-term intracranial EEG recordings,
Clin Neurophysiol. (2005) In press.
AG Nonlinear Dynamics, University of Potsdam,
Germany
*Department of Mathematics and Informatics, Ernst-Moritz-Arndt
Univeristy of Greifswald, Germany
Recent applications of recurrence quantification analysis to EEG data have highlighted its potential for the investigation of event related potentials on a single trial basis. With an innovative modification of recurrence plots, based on rank order structures in the data, recurrence quantification analysis can be further improved. We present new results using order pattern recurrence plots applied to ERP data, demonstrating the improvement over common recurrence plots.
Institute of Computer Science AS CR, Prague, Czech Republic
Several asymmetric measures have been proposed for applications to bivariate time series of amplitudes or phases of coupled dynamical systems in order to establish possible asymmetry of coupling and thus the causality in evolution of the interacting (sub)systems. Such measures, especially those operating with the instantaneous phases of interacting oscillators, are able to identify and quantify the coupling asymmetry in many numerical and experimental examples. In typical real-world systems, however, the data are not only contaminated by noise, but the components of the bivariate series can have very different distributions and/or dominant frequencies. It has been shown that under such conditions so-called directionality indices can be severely biased and falsely indicate an asymmetry in coupling even if the coupling is symmetric, or the indicated coupling direction can be incorrect. Therefore it is important to statistically test the significance of the estimated directionality indices. A possible way is to compute the directionality indices from a set of so-called surrogate data, numerically generated data which mimic most of the statistical and dynamical properties of the tested data, but the coupling asymmetry; in order to estimate bias and variance of the directionality indices. In practice, however, it might be problematic to construct surrogate data with nonlinear symmetric coupling. In an extensive numerical study we compare the size and power of asymmetry tests using surrogate data with various types of mutual dependence. The proposed tests are then applied to electrophysiological data, such as EEG and ECG.
University of Leicester, United Kingdom
Increasing advances in acquisition systems allow the recording of several channels simultaneously. However, these advances have not been followed up by advances in the processing of the data. In fact, manual spike sorting of as many as e.g. 64 channels seems a very time consuming task and is nearly impossible to perform during an experiment. We describe a recently proposed method for detecting and sorting spikes from multiple neuron recordings. The method combines the wavelet transform, which localizes distinctive spike features, with super-paramagnetic clustering (SPC), which allows automatic classification of the data without assumptions such as well-defined cluster means, low variance, or normality. The algorithm has three principal stages: I) Spikes are detected automatically via amplitude thresholding. II) The wavelet transform is calculated for each of the spikes and the optimal coefficients for separating the spike classes are automatically selected. III) The selected wavelet coefficients serve as the input to the SPC algorithm and clustering is performed after automatic selection of the temperature corresponding to the super-paramagnetic phase. We developed several criteria that render the algorithm unsupervised and fast, making it suitable for on-line applications. The method gave an optimal performance both for simulated and real data, outperforming conventional approaches. Finally, we briefly describe the importance of unsupervised and optimal spike detection and sorting of multiple neuron recordings.
References:
R. Quian Quiroga, Z. Nadasdy and Y. Ben-Shaul. Unsupervised spike
sorting with wavelets and superparamagnetic clustering. Neural
Computation, 16: 1661-1687; 2004.
Junior Research Group Neurolinguistics, Phillips-University Marburg, Germany
We show that the uncertainty associated with the interpretation of different language-related ERP components (specifically the N400/P600) can be resolved by means of frequency-analytical dissociations. To this end, we introduce an analysis technique for EEG research on human language comprehension, which supplements ERP measures with corresponding frequency-based analyses. Moreover, we argue that this method not only allows for a differentation of ERP components on the basis of activity in distinct frequency bands and underlying dynamic behaviour (in terms of power changes and/or phase consistency), but also provides further insights with regard to the functional organisation of the language comprehension system and its inherent complexity.
Institute of Physics, AG Nonlinear Dynamics, University of Potsdam, Germany
We analyze the delayed feedback approach to suppression of collective synchrony in a population of globally or randomly coupled neurons. In particular, we consider the main factors of imperfection of the control scheme and their influence on the suppression efficiency. Next, with the help of a realistic model of synaptically coupled population of inhibitory and excitatory neurons we demonstrate the potential of the suppression scheme for neurophysiological applications.
Funktionelle Neurochirurgie, University Hospital Zürich, Switzerland
Neurophysiological recordings in the thalamus - single cell activity and local field potentials (LFP) - as well as scalp EEG recordings provide converging evidence for a thalamocortical dysregulation at the source of chronic neurogenic pain. At the cellular level, low-threshold calcium spike (LTS) bursts have been found in the thalamus of patients suffering from neurogenic pain. About half of the recorded neurons presented LTS bursting activity, and LTS bursts displayed rhythmicity with mean interburst discharge rate of 4 Hz. Thalamic LFP recordings showed the presence of high theta (4-9 Hz) power, coinciding with the theta rhythmicity of LTS bursts. LFP activity was phase-coupled between the theta and beta (14-30 Hz) bands as measured by bicoherence. High coherence between scalp EEG and thalamic theta activities underscores the high level of functional coupling between thalamus and cortex. Scalp EEG showed increased EEG power and a slowing of the dominant peak compared to healthy controls, confirming the existence of a disease-related increase of theta power at the cortical level. In a period of several months after the therapeutic lesion in the thalamus, EEG spectra of patients approach those of healthy controls. These results suggest the increase of theta thalamocortical rhythmicity to originate from disfacilitation of thalamic relay neurons. Asymmetries of corticocortical inhibition lead to cortical over-activation relevant for pain sensation in the beta band. Recurrent thalamoreticulothalamic and corticoreticulothalamic feedback inhibition may develop this into a self-sustained and thus chronic process. The small thalamic CLT lesion initiates a progressive normalization in the affected thalamocortical system, which is reflected in both decrease of EEG power and pain relief.
Department of Mathematical Sciences, Loughborough University, Leicestershire, Great Britain
Nonlinear bifurcations are ubiquitous in complex phenomena and may play an important role in brain dynamics. The aim of this paper is to map out the structure of nonlinear instabilities in a model of the brain's mean field dynamics and use this to explain critical features of the generalized epilepsies. The model treats the cortex as a medium for the propagation of waves of electrical activity and incorporates key neurophysiological processes such as propagation delays, nonlinear membrane physiology and corticothalamic feedback. Previous analyses in the linearly stable regime demonstrated its descriptive validity in a wide range of healthy awake and sleep stages. We show that mapping the structure of the nonlinear bifurcation set predicts a number of key dynamic processes, including the onset of periodic and chaotic dynamics as well as multistability. Quantitative study of the electrophysiological data supports the validity of these predictions and reveals processes unique to the global bifurcation set. Specifically, we argue that the core electrophysiological and cognitive differences between Grand and Petit Mal seizures are predicted by the global bifurcation diagram. The present study is the first to present a unifying explanation of generalized seizures using bifurcation theory and underlines the importance of mapping the brain in the dynamic as well as the spatial domain.
Neural Network Research Centre, Helsinki University of Technology, Finland
Event related potentials and fields are measures of the brain responses to external stimuli, and have been a crucial source of information on the early functioning of the human brain. They are time-locked to the stimuli and are assumed to present a characteristic behaviour over time that is roughly constant for each subject, with a particular configuration of the stimulus. Hence, averaging over several tens of stimuli is commonly used to overcome their very weak signal-to-noise ratio. However, if the brain response changes over time, even slightly, the averaging process may overlook important features associated with those changes. Here, we use the recently introduced denoising source separation (DSS) to achieve a significant increase in signal-to-noise ratio. DSS can be seen as a general framework for exploratory source separation, in which prior or acquired information can be used. Algorithms built under this framework range from almost blind approaches, such as independent component analysis, to highly specialised and problem-tuned algorithms. In particular, these methods have been successfully applied to the analysis of cardiac subspaces in magnetoencephalograms, a problem not so different from that of detecting and analysing single-trial event-related activity.
Department of Empirical and Analytical
Psychophysics,
Institute for Frontier Areas of Psychology and Mental Health,
Freiburg, Germany
We focus on two approaches to the analysis of brain electrical activity, the topographic, space-domain based approach developed since 1970s by Lehmann et al., and the global approach based on state space representation of brain electrical data, applied in our research. In spite of different motivations and developmental paths, basic notions and concepts of the two approaches are related by one-to-one correspondences. Of particular interest are relationships between measures of micro-state diversity (discrete state space dynamics) and those of spatial complexity and its development in time. We address the question whether, or to what extent, these approaches are compatible, or whether they should be seen as two aspects of a unified methodology.
AG Nonlinear Dynamics, University of Potsdam, Germany
Recent research has revealed a rich complexity in the cortical connectivity of the animal brain. It was shown that the structure displays similar statistical properties like other complex networks. The outstanding important questions are: What is the impact of connectivity structure on the dynamics? How is a signal spreading in complex neuronal networks? What important role does synchronization play in these systems? We study these problems with models of coupled neurons. Each brain cortical area is represented by a small-world network of Fitz-Hugh Nagumo neurons. These sub-networks are coupled together according to the connectivity pattern of a cat brain. With this model we investigate the relationship between the structural and functional connectivity of cortical networks. We apply an external input and pursue information processing within the brain areas and in the whole brain. These simulations can lead to a better under- standing of dynamical processes in the brain.
Department of Theory and Data Analysis,
Institute for Frontier Areas of Psychology and Mental Health,
Freiburg, Germany
Contextual emergence is a well-defined non-reductive concept for interlevel relations. It will be briefly illustrated for the relation between statistical mechanics and thermodynamical properties such as temperature. Stability conditions are crucial for a rigorous implementation of contingent contexts as they are required to understand temperature as an emergent property. Such stability conditions can be formulated and utilized for contextual emergence beyond physics as well. The problem of deriving cognitive symbols from a neurobiological level of description is analyzed in terms of partitions of the relevant state spaces. It will be shown that compatible cognitive models emerge if the symbolic partitions of their underlying neurobiological state space are generating and, thus, provide stable cognitive states. If this is not the case, cognitive models are incompatible, complementary, or incommensurable. Consequences of this result for an integration or unification of cognitive models will be indicated.
FGW/ILLC Leerstoelgroep Logica en Cogn.
Weetenschap,
University of Amsterdam, Netherlands
To date, progress in cognitive neuroscience has been hindered by the enormity of the gap between our understanding of some low-level properties of the brain on the one hand, and of some very high-level properties of the mind on the other hand. Research on parallel distributed processing and neural networks (connectionist paradigm) has tried to reduce this gap but was successful in parts only. Recently, Smolensky and Legendre (to appear) have proposed an integrated connectionist/symbolic cognitive architecture (ICS) that promises to overcome this gap. I will comment on some parts of this monumental oeuvre; especially I will refer to the following three aspects of Embodied Cognition:
While ICS is mainly concentrated on the first aspect, I will argue this is not enough - in particular by criticizing the ICS approach to the innateness question.
References:
Smolensky, P. and G. Legendre (to appear). The
Harmonic Mind: From neural computation to
optimality-theoretic grammar. Cambridge, Mass., MIT Press.
Department of Psychology, Cornell University, Ithaca, NY, USA
Most generative approaches to linguistic structure suggest that language is recursive, that recursion is a property of the grammar, and that independent performance constraints limit human abilities on certain kinds of recursive constructions. This paper presents an alternative view, in which quasi-recursive ability is an acquired skill, and in which limitations on the processing of recursive constructions stem from intrinsic constraints on learning and processing. A connectionist model embodying this alternative theory is presented along with corroborating data from behavioral experiments testing predictions derived from the model. Together, the simulations and human data suggest that acquired intrinsic constraints are relevant not only to the processing of complex recursive constructions, such as center-embedding, but also to the processing of the simpler, right- and left-recursive structures.
Institute for Psychology, University of Potsdam,
Germany
+Institute for Psychology, Free University at Berlin,
Germany
*Institute for Psychology, Humboldt University at Berlin,
Germany
Information processing during reading is often studied with EEG experiments that use a sequential, word-by-word presentation of stimuli. Although this RSVP paradigm is somewhat artificial with regard to the natural reading process, it is usually chosen to avoid at least two problems involved with left-to-right reading: First, natural reading requires eye movements that cause large measuring artifacts in the EEG. Second, it may be difficult to disentangle the temporally overlapping EEG contributions of neighboring words and other cognitive processes during natural reading. In the present study, we recorded the EEG of 21 subjects, that were silently reading 144 German sentences in a natural, left-to-right fashion. Fixation position was measured concurrently with a high resolution eye tracker. For artifact correction, we applied the MSEC spatial filter by Berg & Scherg [1]. Previous experience with this method suggests that it is suitable to compensate even for strong eye artifacts during experimental trials [2]. Event-related potentials (ERPs) were averaged time-locked to the onset of each fixation. Various approaches are currently used to control for some of the problems of ERP overlap, such as a multiple regression model and the comparison with a parallel RSVP dataset [3]. Here, we present our basic approach and some preliminary results for the effect of word predictability.
References:
[1] Berg, P. & Scherg, M. (1994). A multiple source approach
to the correction of eye artifacts. Electroenceph Clin Neurophys,
90, 229-241.
[2] Dimigen,O., Schildt,U., Hohlfeld, A. & Sommer, W. The
effect of saccadic eye movements on spoken language
comprehension. Manuscript in preparation.
[3] Kliegl, R., Dambacher, M., Hofmann, M., Jacobs, A. &
Engbert, R. Frequency and predictability effects in event-related
potentials and fixation durations in reading. Manuscript in
preparation.
Research Unit "Conflicting Rules" and AG Nonlinear Dynamics, University of Potsdam, Germany
On the occasion of their remarkable controversy with Smolensky, Fodor and Pylyshyn (1988) proposed two desiderata that should be met by any cognitive symbol processing system. Such a system should have firstly a "combinatorial syntax and semantics for mental representations" comprising the properties of productivity, systematicity, and compositionality, and secondly the "structure sensitivity of processes" which means that constituents have causal roles in computations. In order to fulfill these requirements, Smolensky has developed the tensor product representations of symbolic structures for activation vectors in neural network models (1991). This approach allows indeed for compositional and causally efficacious representations. However, it is not infinitely productive and does therefore not allow for recursion. I shall allude to further shortcomings of the tensor product representations and argue that Smolensky's "coffee story" (1991) entails a possible solution of the problem to ground symbols in dynamical systems, namely by means of equivalence classes of points in the phase space (macrostates) instead of these points itself (microstates). Using an example of language processing (beim Graben et al. 2004), I shall demonstrate that nonlinear dynamical automata (NDA) implement symbol processors as dynamical systems where basic requirements on compositionality and structure sensitivity are met by the symbolic dynamics resulting from a particular partition of the phase space into equivalence classes.
References:
J. Fodor and Z. W. Pylyshyn. Connectionism and cognitive
architecture: A critical analysis. Cognition, 28:3-71, 1988.
P. Smolensky. Connectionism, constituency, and the language of
thought. In B. Loewer and G. Rey, editors, Meaning in Mind. Fodor
and his Critics, pages 201-227. Blackwell, Oxford, 1991.
P. beim Graben, B. Jurish, D. Saddy, and S. Frisch. Language
processing by dynamical systems. Int. J. Bifurcation Chaos,
14(2):599-621, 2004.
Philosophy Department, King's College London, Great Britain
Study of dialogue has been proposed by Pickering and Garrod (2004) as the major new challenge facing both linguistic and psycho-linguistic theory. Two of the phenomena which they highlight as common in dialogue, but posing a significant challenge to theoretical linguists, are alignment between conversational participants, and shared utterances, with conversational participants mirroring each other's patterns at many levels including lexical, syntactic and semantic choices. Shared utterances are those in which participants exchange parser and producer roles mid-sentence. This talk will introduce Dynamic Syntax (Kempson et al 2001, Cann et al forthcoming), a grammar formalism which directly reflects the way interpretation of a sentences is built up by hearers in real time, and show how (i) the framework allows a psycholinguistically plausible model of incremental context-dependent generation, (ii) the resulting tightly coordinated systems of generation and parsing directly reflect dialogue alignment patterns and shared-utterance phenomena. The concept of tree growth is central to Dynamic Syntax; and the system is essentially procedural. Parsing is defined in terms of actions on semantic tree structures, in which structures are incrementally and monotonically built, as dictated by the serial order of words in a string. Generation reflecting word-by-word incrementality can be defined in this framework in terms of the very same actions as manipulated by the parser (following preliminary results of Otsuka and Purver 2003). In particular, both parsing and generation can be defined in context-dependent terms (Purver and Kempson 2004), defining utterance processing and dialogue context alike in terms of (partial) trees, the context comprising (partial) trees that have been constructed prior to the current utterance and the associated tree-update actions used in creating such trees. Alignment can then be seen to result directly from the use of such context as a generation strategy for minimizing general-lexicon search, and switch of speaker-hearer roles in shared utterances can be straightforwardly modelled in virtue of both speaker and hearer constructing tree-structures representing content and then adopting them as context for the next update.
References:
Cann, R., Kempson, R., and Marten, L. (forthcoming). The Dynamics
of Language. Elsevier.
Kempson, R., Meyer-Viol, W. and Gabbay, D. (2001). Dynamic
Syntax. Blackwell.
Otsuka, M. and Purver, M. (2003) Incremental generation by
incremental parsing. Proceedings of 6th CLUK Colloquium,
93-100.
Pickering, M. and Garrod, S. (2004) Towards a mechanistic
psychology of dialogue. Behavioral and Brain Sciences 27,
169-226.
Purver, M. and Kempson, R. (2004). Incremental parsing, or
incremental grammar? In Keller,F., et al. (ed.) Proceedings of
the ACL Workshop on Incremental Prasing: Bringing Engineering and
Cognition Together, pp.74-81. Barcelona.
Konrad-Zuse-Center for Information Technology,
Division Scientific Computing, Department Visualization, Berlin,
Germany
Using two different models of oscillatory modulated activity in the primary visual cortex, we anaylzed the synchronization properties of the networks by an Eigenmode decomposition. Both models used clusters of feature-sensitive neurons representing local object properties like color and orientaton. Whereas in the mean-field model oscillators communicated via their current amplitude, in the phase model oscillator interaction was controlled by phase difference. In both cases, Eigenmode analysis decomposed the complex synchronization patterns into a time-invariant, spatial component, the Eigenmodes, and characteristic functions describing their weight in network state over time. We found that characteristic functions can be associated with representations of objects in a visual scene, and Eigenmodes represent different epistemic possibilities. We interprete this as a realization of lexical and non-lexical complex concepts in the cortex.
Institute of Psychiatry, Catholic University of
Rome, Italy
Department of Psychology, University College London, Great
Britain
Claude E. Shannon posed the foundations of modern studies on the informational structure of texts and speech. His less famous work on the prediction and entropy of printed English is a source for inspiring new research. While we could place Shannon's contributions at the micro-semiotic level George K. Zipf was mostly working at the meso-semiotic level (words). Recent studies used different approaches: Schenkel, for example, studied long range correlation in various human writings by mapping them into a simple 1-D random walk model. Amit studied the long-range correlations in various translations of the same text. We investigated the informational structure of written texts (also in the form of speech transcriptions) using Recurrence Quantification Analysis (RQA). RQA technique provides a quantitative description of text sequences at the orthographic level in terms of structuring, and may be useful for a variety of linguistics-related studies. We used RQA to measure differences in linguistic samples from different subjects. They were divided in subgroups based on personality and culture differences. We used RQA and KRQA (Cross Recurrence) to measure the coupling and synchronization during the conversation (semiotic interaction) of different subjects. We discuss results both for the improvement of methodology and some general implications for neuro-cognitive science.
Center for the Study of Complex Systems, University of Michigan, Ann Arbor, MI, USA
Many systems in statistical physics admit multiple levels of description, from microscopic molecular detail up through very broad macroscopic features. The higher-level descriptions are "coarse-grainings" of the lower levels, and the higher-level variables are generally collective properties of many lower-level objects. Not every coarse-graining leads to a "good" set of macroscopic variables; those that do have certain statistical properties. These properties, in turn, have important information-theoretic implications, and, when the coarse-graining is discrete ("symbolic dynamics"), the system can be modeled by stochastic automata. After sketching these ideas, I suggest some ways they might help cognitive scientists relate symbolic or computational descriptions to neural, dynamical ones.
Empirical Methods in Syntax Research, Institute for Linguistics, University of Potsdam, Germany
I present a parsing model embedded within a pre-existing symbolic cognitive architecture, ACT-R. The model's behavior and predictions are discussed with reference to several key phenomena in human sentence processing research.
Department of Philosophy, Heinrich-Heine-University Düsseldorf, Germany
The neuronal structure of the cortex is usually perceived as radically different from the semantic structure on which cognition and language understanding are defined. This paper argues that, in spite of these prima facie differences, semantic structure can be reduced to neural structure if one assumes that oscillatory networks, which implement the mechanism of neuronal synchrony, be an appropriate model of neural reality. Several neurobiological data on neural synchronization support this assumption (Singer & Gray 1995). In the context of this paper, semantic structures are regarded as structures of mental concepts (Fodor, 1998) and are assumed to subserve two purposes. First, its elements evaluate linguistic expressions semantically and can thus be regarded as the meanings of those expressions. Second, its elements are themselves semantically evaluable with respect to external content. The paper provides a model for the realization of lexical and non-lexical complex concepts in the cortex. The network discussed has both perceptual and semantic capabilities. Three adequacy conditions, the compositionality of meaning, the compositionality of content, and the co-variation with content, are satisfied. Coherency chains and hierarchical mechanisms of binding are postulated.