DOC

Are event-related potential components generated by phase resetting of brain oscillations

By Alma Roberts,2014-01-20 10:05
13 views 0
Are event-related potential components generated by phase resetting of brain oscillations

    Are event-related potential components generated by phase resetting of brain oscillations? A critical discussion

    a, , aab, W. Klimesch, W.R. Gruber, S. Hanslmayr, R. P. Sauseng

    aaFreunberger and M. Doppelmayr

    aDepartment of Physiological Psychology, University of Salzburg, Hellbrunnerstrasse 34, A-5020 Salzburg, Austria

    bDepartment of Psychology, University of Regensburg, Universitätsstrasse 31, D-93053 Regensburg, Germany

    Accepted 11 March 2007. Available online 24 April 2007. Abstract

    The event-related potential (ERP) is one of the most popular measures in human cognitive neuroscience. During the last few years there has been a debate about the neural fundamentals of ERPs. Two models have been proposed: The evoked model states that additive evoked responses which are completely independent of ongoing background electroencephalogram generate the ERP. On the other hand the phase reset model suggests a resetting of ongoing brain oscillations to be the neural generator of ERPs. Here, evidence for either of the two models is presented and validated, and their possible impact on cognitive neuroscience is discussed. In addition, future prospects on this field of research are presented.

    Key words: alpha; electroencephalogram; EEG; cognitive neuroscience; evoked model; phase synchronization

    Abbreviations: EEG, electroencephalography/electroencephalographic; ERP, event-related potential; MEG,

    magnetoencephalography/magnetoencephalographic

Article Outline

    Is current evidence for phase resetting valid?

    Preconditions

    (p1) An oscillation has to be ongoing before it resets phase

    (p2) If reset of an ongoing oscillation generates the ERP, the ERP must show the respective frequency characteristics

    (p3) If the ERP is generated by phase reset of ongoing oscillations, the neural sources of the ERP and those of ongoing oscillations generating the ERP have to be the same

    Theoretical criteria

    (c1) If the ERP is generated by phase resetting, in response to a stimulus phase concentration over trials has to be obtained in contrast to a pre-stimulus period where a uniform distribution of phases over trials should be given

    (c2) If ERPs are solely generated by phase reset there is no power increase in single trials from pre-stimulus to post-stimulus

    (c3) If the ERP is generated by ongoing oscillations, amplitude and/or phase of ongoing oscillations at stimulus onset should influence the ERP

    (c4) If phase resetting of oscillatory brain activity generates ERP, individual and dominant EEG frequency should negatively correlate with ERP latency

    (c5) If there is non-linearity of post-stimulus power modulations the ERP cannot be generated by a simple additive evoked response independent of background EEG activity (as proposed e.g. by Arieli et al., 1996)

How valid is evidence against the phase reset model?

    Evoked response versus phase reset model: Where do we stand, and where should we go?

    Why is it important to know whether ERP components are generated by evoked response or phase resetting?

    Which are the main topics to be answered by further research?

    Which technical developments will be necessary in the future?

    Acknowledgements

    References

By averaging over a large number of single sweeps of

    electroencephalographic (EEG) activity the event-related potential (ERP) can be obtained. It is informative about the exact timing of cognitive processes in the human brain. In the last few years there has been a renaissance of the debate about the neural basis of ERP generation (first discussed by Sayers et al., 1974) initiated by work of (Basar 1999),

    (Brandt 1997) and (Tass 2000) and recently Makeig et al. (2002). The

    authors argue that the ERP does not simply emerge from evoked, fixed latencyfixed polarity responses that are additive to and independent of ongoing EEG. Instead, evidence is presented that early ERP components are generated by a superposition of ongoing EEG oscillations that reset their phases in response to sensory input. The principle of this idea is depicted in Fig. 1. When single EEG traces that are not phase-locked are averaged no evoked activity should remain (Fig. 1a). When a constant

    evoked response is linearly added onto unrelated single trials the average of single traces will resemble the evoked response (Fig. 1b). The

    alternative explanation for generation of ERPs is shown in Fig. 1c. When

    the ongoing oscillation of single trials undergoes phase resetting in

    response to a stimulus an ERP is obtained without any additive evoked response which is added to the EEG traces.

    Display Full Size version of this image (17K)

    Fig. 1. General idea of evoked and phase rest model. (a) When a number of unrelated single trials which are not phase-locked to a stimulus are averaged a flat line will ideally be the result. (b) The evoked model assumes that in each single trial a constant evoked response is added onto the ongoing EEG. The background EEG is considered as noise completely unrelated to the ERP. When the single trials composed of the background EEG and an additive evoked response are averaged the ERP which exactly reflects the original evoked response will result. (c) The phase reset model suggests that the ERP is generated by a phase resetting of oscillatory background EEG. Without any additive evoked response an ERP will arise when single trials are averaged.

    In the last few years a large number of papers have been published either arguing for an additive, evoked model of ERP generation or for phase resetting as the neural mechanism on which ERPs are based (e.g. [Barry

    et al 2003], [David et al 2005], [Düzel et al 2005], [Fell et al 2004],

    [Fuentemilla et al 2006], [Gruber et al 2005], [Hamada 2005], [Hanslmayr

    et al 2007], [Jansen et al 2003], [Klimesch et al 2004a], [Klimesch et

    al 2004b], [Kruglikov and Schiff 2003], [Makeig et al 2002], [Mäkinen et

    al 2005], [Mazaheri and Picton 2005], [Mazaheri and Jensen 2006], [Naruse

    et al 2006], [Penny et al 2002], [Rizzuto et al 2003], [Schack and Klimesch

    2002], [Shah et al 2004] and [Yamagishi et al 2003]). However, many of

    the arguments and methods for either of the models (evoked model vs. phase reset model) seem to be inappropriate and thus (as outlined below) cannot really dissociate between these two hypothesized mechanisms. This is the

    reason why many contradictory results are communicated in this field of research.

    In this short review we critically discuss the validity of arguments for as well as against phase resetting as mechanism for ERP generation. This is followed by an outlook concerning a possible future of research addressing the question of the neuronal origin of ERPs. Finally, we present some important issues which will have to be addressed by future investigations in the field.

    Is current evidence for phase resetting valid?

    Most of the current evidence for phase resetting as a mechanism which generates at least in part the ERP is based on a few theoretical criteria (c1c5). In the following these criteria are discussed. Prior to that, a number of preconditions (p1p3) are presented that have to be given

    before the theoretical criteria can be applied.

    Preconditions

    (p1) An oscillation has to be ongoing before it resets phase

    The rationale of this precondition is that a resetting of EEG phase in response to a stimulus can only occur if the oscillation does already exist prior to the reset. If the oscillation is not ongoing (i.e. if it does not exist in a pre-stimulus interval) one could argue that in response to a stimulus the phase of a particular oscillation does not reset but that this oscillation does arise completely new. This rationale is logically sound. However, there is a methodological problem with defining whether an oscillation is ongoing. Usually, only rhythmic brain activity that shows a dominant peak in the power spectrum is considered as ongoing. But it is questionable whether the dominance of an oscillation (meaning

    that no other frequency shows higher amplitudes) is an adequate measure to prove that an oscillation is ongoing. This means, that from the argument that a dominant peak in the power spectrum indicates that an oscillation is ongoing there follows not automatically the opposite, namely that the absence of dominance in the power spectrum proves the non-existence of this oscillation. Therefore, if this sensible precondition is used, a new way of operationalization is needed for it.

    (p2) If reset of an ongoing oscillation generates the ERP, the ERP must show the respective frequency characteristics

    The logic of this precondition is that some frequencies might undergo phase resetting after stimulus onset and that this effect must be contained in the ERP. However, there might be frequencies with so small power (i.e. amplitude) so that it is unlikely that they generate an ERP component. For instance, it was shown that around 90 ms after presentation of visual stimuli high frequency oscillatory brain activity in the gamma band (beyond 30 Hz) shows phase reset (see e.g. Herrmann et al., 2004).

    But amplitude at gamma frequency usually is so small that a phase reset at this band can be doubted to generate ERP components. On the other hand, the P1N1 ERP complex has the frequency characteristics of alpha (and partly theta) oscillations (e.g. Klimesch et al., 2004a). And late

    endogenous ERP components reflect theta and delta oscillations (Fell et

    al., 2004). Therefore, it seems legitimate to further investigate whether the latter frequencies contribute to the generation of these ERP components. The frequency characteristics of the ERP can thus be used as constraints for further analysis of the relation between phase reset and ERP generation.

    (p3) If the ERP is generated by phase reset of ongoing oscillations, the neural sources of the ERP and those of ongoing oscillations generating the ERP have to be the same

    The logic of this precondition is that when an ongoing oscillation is present in a cortical patch A but not in a cortical patch B then only an ERP arising from cortical patch A can be generated by this ongoing oscillation. An ERP which originates from cortical patch B cannot be generated by the oscillation at patch A. The neural sources of rhythmic activity and the ERP have to show some overlap at least. However, the methodological problem here is that the scale of scalp EEG might be too large to precisely determine the neural sources of the ERP and rhythmic brain activity. Even if source analysis is run with EEG data and identical sources in the sense of cortical areas for ongoing oscillations and ERPs are demonstrated, the question remains whether it is exactly the same neural population that generates both phenomena as nevertheless the spatial scale of EEG will be on a too large level because even the resolution of source analysis methods is limited.

    The above presented preconditions are useful and necessary. However, it is difficult to operationalize them. The question is how to determine, whether there is an ongoing oscillation in a pre-stimulus interval or whether the ERP reflects certain frequency characteristics. In terms of the first precondition one possible way would be to relate pre-stimulus power of a certain frequency to task-relevant cognitive parameters. For instance, although theta oscillations do not show up as a dominant peak in the power spectrum in the healthy human, theta power in a pre-stimulus interval or a resting period is related to memory performance ([Klimesch

    1996], [Klimesch 1999] and [Klimesch et al 2001]). The second precondition

    can be operationalized by calculating a power spectrum for the averaged ERP. Only those frequencies which show a peak in the ERP-spectrum (power spectrum derived from the averaged ERP) should be further investigated concerning phase reset. The third precondition is the most difficult to operationalize in the human EEG. One would have to obtain single or

    multiunit activity to measure whether exactly the same sample of neurons reflects ongoing oscillatory activity and the ERP. However, it is advisable to run source localization analyses with EEG data to first investigate whether the brain regions exhibiting oscillatory activity and the ERP overlap at all. If they overlap there is still the uncertainty described above. But if they do not overlap it is evident that there is no need for further analysis regarding phase resetting, as then the ERP cannot be generated by the ongoing EEG.

    Theoretical criteria

    (c1) If the ERP is generated by phase resetting, in response to a stimulus phase concentration over trials has to be obtained in contrast to a pre-stimulus period where a uniform distribution of phases over trials should be given

    This criterion is most often applied in studies dealing with the topic of phase resetting and ERP (see Table 1). The criterion is valid and

    necessary. If there is phase resetting this must show up by phase concentration. However, meeting this criterion does not prove phase resetting. If evoked responses with fixed latency and fixed polarity are added on trials of ongoing EEG activity which are completely independent of the evoked responses, then also a phase concentration will occur artificially. This is because filtering (which is necessary for determining an oscillation’s instantaneous phase) will turn the evoked response into an artificial oscillatory response which influences the estimated phase of the signal (see Fig. 2). Therefore, a measure of phase

    concentration alone will not dissociate between phase reset and evoked model.

    Table 1.

    Arguments and premises for phase resetting from recent literature

Study c1 c2 c3 c4 c5

    Barry et al. (2003)

    Barry et al. (2004)

    Barry et al. (2006)

    Düzel et al. (2005)

    Fell et al. (2004)

    Fuentemilla et al. (2006)

    Gruber et al. (2005)

    Hamada (2005)

    Hanslmayr et al. (2007)

    Jansen et al. (2003)

    Klimesch et al. (2004a)

    Klimesch et al. (2004b)

    Kruglikov and Schiff (2003)

    Makeig et al. (2002)

    Mazaheri and Picton (2005)

    Naruse et al. (2006)

    Rizzuto et al. (2003)

    Yamagishi et al. (2003)

    c1: In response to a stimulus phase concentration over trials has to be obtained. c2: If ERPs are generated by phase reset there is no power increase in single trials from pre-stimulus to post-stimulus. c3: If the ERP is generated by ongoing oscillation, amplitude and/or phase of ongoing oscillations at stimulus onset should influence the event-related potential. c4: If phase resetting of oscillatory brain activity generates ERP individual and dominant EEG frequency should negatively correlate with ERP latency. c5: If there is non-linearity of post-stimulus power

    modulations the ERP cannot be generated by a simple additive evoked response independent of background EEG activity.

     Display Full Size version of this image (141K)

    Fig. 2. Additive evoked response may give the impression of phase resetting. (a) One hundred single trials of resting EEG where no stimulus has been presented are depicted. The traces reflect background EEG completely unrelated to each other and non-phase-locked. When the single trials from (a) are averaged nearly a flat line is the result. (c) When a phase-locking index (PLI; Schack and Klimesch, 2002) which is a measure

    for phase concentration over trials (comparable to inter-trial coherence, Makeig et al., 2002) is calculated it indicates that there is no

    phase-locking at any frequency and at any time frame. When a constant evoked response (d) is added onto the unrelated single trials [single trials from (a) plus evoked response from (d) are depicted in (e)] a signal nearly identical to the original evoked response is obtained after averaging (f). Estimation of phase-locking using the single trials from (e) will result in strong phase concentration at the time frames where the evoked response occurs (g). This shows that using phase concentration estimates alone without any further information will not be able to dissociate between phase resetting and additive evoked response. (c2) If ERPs are solely generated by phase reset there is no power increase in single trials from pre-stimulus to post-stimulus

    Shah et al. (2004) argued that an additive evoked response would result in post-stimulus power increase, whereas phase resetting should not lead to any enhancement of amplitude. Demonstrating phase concentration in the

Report this document

For any questions or suggestions please email
cust-service@docsford.com