Research report
Phase-locked alpha and theta oscillations generate the P1–N1 complex and are related to memory performance

https://doi.org/10.1016/j.cogbrainres.2003.11.016Get rights and content

Abstract

An oscillatory phase resetting model is presented and data are reported which indicate that early components of the event-related potential are due to the superposition of evoked oscillations. The following hypotheses were tested and could be confirmed: (i) theta and alpha show a significant increase in phase locking during the time window of the P1 and N1 as compared to a prestimulus reference, (ii) the dynamics of event-related changes in evoked theta and alpha power obey the same principles as are known from event-related de-/synchronization research, and (iii) latency measures of the P1–N1 complex are negatively correlated with individual alpha frequency. In addition, we have found that theta phase locking is larger during encoding than recognition and that good memory performers show a larger increase in theta and alpha phase locking during recognition in the time window of the N1. Our general conclusion is that the P1–N1 complex is generated primarily by evoked alpha and theta oscillations reflecting the synchronous activation of a working- and semantic memory system.

Introduction

The aim of the present study is to test the hypothesis whether the P1–N1 complex of event-related potentials (ERPs) is generated by a superposition of evoked oscillations [3], [4] in the theta and alpha frequency range. We develop this hypothesis within the framework of an oscillatory phase resetting model for ERP generation and then demonstrate that a superposition of evoked theta and alpha waves synchronized in absolute phase during the N1 latency window can nicely describe the typical P1–N1 waveform. In considering findings from event-related de/synchronization (ERD/ERS) research [22], [24], [31], the proposed hypothesis allows us to predict that evoked theta and alpha amplitudes show the same principles of event-related changes that are known from traditional bandpower analyses [22].

The typical latencies of the P1–N1 wave already suggest a superposition of evoked theta and alpha. The latency range of the P1 (at posterior sites) for visual stimuli is about 90–120 ms and about 150–190 ms for the N1 [25]. In young, healthy adults alpha has a frequency of about 10 Hz (with a range of about 7–13 Hz). Accordingly, mean latency between two alpha peaks of the same polarity is about 100 ms. Thus, mean latency of the P1 (of about 105 ms) lies in the average alpha frequency range, whereas the interpeak latency (between the P1 at about 105 ms and the N1 at about 170 ms) of about 65 ms (representing a half cycle because the P1 and N1 have different polarities) lies in the slow alpha frequency range, close to the ‘transition’ of theta (a cycle of 2×65=130 ms represents a frequency of 7.7 Hz). Particularly at posterior sites, the early ERP components consist of sharp positive and negative peaks that form the P1–N1 complex. They appear to reflect modality-specific sensory processes that are sensitive to the direction of attention (for a comprehensive review, see Ref. [25]) and are well investigated in the visual [15], [19], [21], [26], [28], [29] and auditory information processing domain [18], [20], [43], [44], [45], [46]. The fact that the P1 and N1 are sharp waves with alternating polarities clearly demonstrates that neural synchrony occurs in narrow time windows and probably alternates between inhibitory and excitatory processes. There is some evidence that the P1 amplitude is associated with the suppression of irrelevant, and the N1 with the processing of the attended, relevant information. Thus, it was suggested that the P1 reflects inhibitory, and the N1 excitatory processes (cf. Ref. [21], p. 23). It is important to emphasize that EEG waves with alternating polarities are due to changes in the (relative) level of depolarization in the (dendritic and somatic) membrane potential of masses of neurons and, thus, reflect phases of low vs. high excitability. This is exactly the way oscillatory processes are operating as was shown by a variety of different investigators analyzing local field potentials and multi-unit activity in animal studies and the EEG in humans (for summaries see, e.g. Ref. [13] and [3], [4], respectively). At the neuronal level, the basic mechanism thereby is that variations in the strength of the local field potential are related to periods of enhanced or suppressed multi-unit activity which reflect bursts of action potentials [8], [12].

Although the similarities between the P1–N1 wave and evoked theta and alpha are obvious, it may be objected that this similarity is superficial and that the P1 and N1 are generated as evoked components that are independent from ‘background activity’ such as ongoing oscillations. Thus, the critical issue is what type of evidence allows us to distinguish between an oscillatory phase resetting and evoked model of ERP generation. The oscillatory phase resetting model (depicted in Fig. 1A) assumes that an ongoing oscillation (of a given frequency domain) undergoes an (i) event-related modulation of phase (i.e. exhibits phase resetting) and, in addition, may also undergo an (ii) event-related modulation in amplitude. The evoked model (shown in Fig. 1B) assumes that superimposed on ‘background activity’ (depicted as an ongoing oscillation in the frequency domain of interest with random phase between trials and no resetting) a fixed latency and polarity evoked component is generated in response to a stimulus and/or task. It is important to note that phase resetting per se (which is well documented [9], [27], [37], [41], [42]) does not allow us to distinguish between the oscillatory and evoked model. A fixed polarity fixed latency component superimposed on a (random) oscillation would also lead to a transient reduction in the intertrial phase variability and, thus, mimic phase resetting. Nonetheless, the two models can be distinguished when considering—in addition to phase resetting—the type of event-related amplitude modulation which is different for theta and alpha. Alpha is characterized by an event-related decrease in amplitude (i.e. by ERD) whereas theta by an increase in amplitude (ERS). Because in the evoked model, ‘phase resetting’ can occur only together with an increase in (evoked) amplitude and only during that time window where the evoked component appears, in the oscillatory model, phase resetting can occur independent of the type of amplitude modulation (i.e. increase, decrease or no change in amplitude). Furthermore, in the oscillatory model, the time windows for phase resetting and amplitude changes may vary independent of each other. It should be noted (as depicted in Fig. 1A) that even in a case when amplitudes do not change (and hence there is no single trial evoked component), an ERP will be obtained due to phase resetting.

When evaluating the two models, three different measures (explained in Materials and methods below), one for phase resetting (termed phase locking index or PLI in the following), one for amplitude changes in single trials (termed whole power in the following) and another for the amplitudes of the evoked potential (termed evoked power in the following) will be used. PLI reflects the degree of phase variation between trials (for very similar measures see, e.g. the ‘intertrial coherence’ suggested by Ref. [27] and the ‘phase locking factor’ used by Ref. [41]). PLI is a normalized measure with a PLI=0 reflecting maximal and a PLI=1 reflecting no phase variability [37]. Whole power (a measure similar but not identical to traditional band power) and evoked power will be determined for each frequency band (in the theta and alpha range) on the basis of Gabor wavelet analyses in order to achieve adequate time-frequency resolution and direct amplitude estimates. Evoked power will be calculated on the basis of individual ERPs. It reflects amplitude estimates of evoked theta and alpha oscillations, respectively. For whole power, single trial power estimates will be obtained which then will be averaged over trials.

Under both models we expect an increase in PLI (due to phase resetting) and an increase in evoked power. But only the oscillatory model is capable of predicting a case where (in addition to an increase in PLI) whole power does not change or decrease, whereas evoked power increases. Or, in other words, in cases of either no change or an ERD, a significant increase in PLI constitutes evidence for an oscillatory phase resetting model. We expect this latter case for the alpha band because it is so well-established that alpha power decreases (whereas theta increases) in response to a stimulus and/or task demands.

The similarity of the P1–N1 waveform with an evoked alpha/theta wave can easily be demonstrated by adding (superimposing) an alpha and a theta wave. The critical point here is that in order to generate a typical P1–N1 waveform and latency pattern, the alpha and theta wave must not be randomly superimposed with respect to their phase. We assume that theta and alpha oscillations reflect the activity in different neural network systems (cf. Ref. [22] for a review) and that the two systems become synchronized with respect to the polarity of their evoked components within a narrow time window that is represented by the N1 component. An example is shown in Fig. 2. Two different evoked oscillatory frequencies, one in the theta range with 6 Hz and a second in the mean alpha frequency range with 10 Hz are considered. The latency of the N1 was assumed with 150 ms poststimulus. From this it follows that in the alpha frequency range, the P1 occurs at exactly 100 ms (cf. Fig. 2A), whereas in the theta range, the P1 occurs earlier at 66.6 ms (cf. Fig. 2B). When the two waves are shown superimposed, it becomes clear that in the resulting ERP the N1 latency will not change, but the P1 will occur earlier (somewhere between 66.6 and 100 ms, depending on the amplitude of the evoked alpha and theta oscillation; cf. Fig. 2C). Furthermore, it should be noted that the first zero crossing of the two waves occurs at the same time, i.e. 25 ms poststimulus. It should be noted that, if there would be no phase resetting, the peaks of the alpha and theta waves would be randomly distributed and would not add up to generate the P1–N1 complex. The result of a simple simulation is shown in Fig. 2D, which depicts the predicted waveform of the P1–N1 complex if two sinus waves with different frequencies (6 and 10 Hz) but the same amplitude are added (bold line in Fig. 2D). Although differences in amplitude modulations for theta and alpha are not considered in the simulation, the resulting waveform already depicts the most important features of the P1–N1 complex.

The proposed model, as outlined in Fig. 2, is focusing on the superposition of evoked oscillations but does not make specific assumptions about the possible neural sources of the P1–N1 complex. The general view—based on findings from animal research in particular—is that multiple structures contribute to any surface ERP (cf. the work by Schroeder et al. [39], [40]).

Our main hypotheses refers to the modulation of phase and amplitude of theta and alpha oscillations by the presentation of a stimulus in a memory task during the time window of the P1–N1 complex. We can make the following predictions:

  • (1)

    Theta and alpha will show a significant increase in PLI during the time window of the P1 and N1 as compared to a prestimulus reference. During this time window, the event-related modulation of theta and alpha amplitudes will differ. Theta amplitudes are expected to increase (i.e. exhibit ERS) but alpha amplitudes are expected to decrease (i.e. show ERD).

  • (2)

    Due to the influence of ERS, we expect that the increase in evoked theta—from the P1 to the N1 time window—will be larger than the increase in evoked alpha. Within this time window we do not assume differential effects of the PLI for alpha and theta.

Because we have found repeatedly that theta and alpha play an important role for memory [22], we will in addition focus on the question whether interindividual differences in memory performances are reflected by differences in PLI, evoked or whole power.

If the P1–N1 complex is generated at least in part by alpha and theta oscillations, the amplitudes are expected to behave in a similar way as is known for ERD/ERS. Thus, when estimating P1 and N1 amplitudes by evoked theta and alpha power, we assume that the dynamics of changes in ERD/ERS with respect to reference power are similar for evoked power. The relevant observation here is that the extent of ERS and ERD depends on prestimulus (reference or resting) power but in different ways for theta and alpha [14], [22]. Whereas small theta resting power enhances ERS, large alpha resting power enhances ERD. Thus, we predict a similar relationship between pre- and poststimulus power also for evoked theta and alpha oscillations. In particular we predict the following hypotheses:

  • (3)

    Small theta resting power will enhance evoked theta (when calculated in a similar way as for ERS, i.e. as a percentage of a power change; cf. Fig. 3A).

  • (4)

    Large resting alpha will attenuate evoked alpha (due to power suppression, cf. Fig. 3B).

The experimental evaluation of the latter two hypotheses requires some methodological clarifications with respect to the type of power measurements and the relationship between reference (ref.) power and event-related changes in power. ERD/ERS is the percentage of a change in whole (band) power during a ‘test’ interval (i.e. some selected interval during the actual processing of a stimulus and/or task) with respect to a reference interval (i.e. a time period preceding the processing of a stimulus and/or task) according to the formula: ERD/ERS %=((ref. power−test power)/(ref. power))×100. When relating ERD/ERS to absolute ref. power, we have to consider three different cases for the alpha and theta band as depicted in Fig. 3. For theta the typical observation is that for a group of subjects with small ref. power (cf. the white bars in Fig. 3A) the percentage of an event-related increase in power (ERS) is larger than for subjects with large reference power (cf. cases 1 and 2 in Fig. 3A). For evoked theta, we expect the same relationship. We will test this relationship by analyses of variance (ANOVA's) with one grouping factor (termed factor POWER: subjects with small vs. large theta resting power) and one within subject factor (termed TIME: ref. power vs. test power). According to hypothesis 3 (and referring to Fig. 3A) we predict either a significant interaction between factor POWER and TIME (in the direction as predicted by case 1) or no interaction (as indicated by case 2). We predict this relationship for theta whole power as well as for evoked theta. For alpha the typical observation is that for a group of subjects with large ref. power (cf. the gray bars in Fig. 3B) the percentage of an event-related decrease in power (ERD) is larger than for subjects with small reference power (cf. case 1 Fig. 3B). Thus, for alpha power we predict—according to hypothesis 4—a significant interaction as indicated by case 1 in Fig. 3B. The rationale underlying this hypothesis is that a suppression of alpha reduces evoked power, despite the expected significant increase in PLI. Thus, for alpha we have to consider opposite effects between PLI and power suppression on evoked power. Due to phase locking, evoked alpha is expected to increase from reference to test but the extent of evoked alpha will be reduced by a suppression of power. Because desynchronization depends on resting power (case 1 in Fig. 3B), the prediction is that large reference power should lead to only a small increase in evoked power from reference to test.

Finally, according to our basic assumption that alpha is involved in the generation of the P1–N1 complex and that this involvement is particularly clear for the P1 (because during an earlier time window theta is comparatively less important than for the N1), we predict a fifth hypothesis.

  • (5)

    Individual alpha frequency (IAF) and individual P1 latency will exhibit a significant covariation.

Section snippets

Subjects

After informed consent, a sample of 36 right-handed students (10 males, mean age=25.3; SD=2.8; 26 females, mean age=22.1; SD=3.6) participated voluntarily in the present experiment.

Materials and design

A set of 180 pictures was taken randomly from a large sample of line drawings [36]. Stimuli were presented at the center of a computer monitor, placed 1.2 m in front of the subjects. The line drawings cover a visual angle of 9.5×5° and were presented in white on a black background.

The experimental design consisted of

Behavioral data

On the average, subjects correctly identified 45.7 targets (76% correct). The number of correctly identified distractors was 107.4 (89% correct). The number of incorrect responses to targets (misses) was 12.1 and the number of incorrect responses to distractors (false alarms) was 12.3. Average d′ was 2.14 (SD=0.60).

Whole power

The results of a four-way ANOVA shows significant main effects for FREQ (F(3, 105)=5.96; p<0.01), LOC (F(5, 175)=10.04; p<0.01) and TIME (F(2, 70)=5.50; p<0.05) but not for RECO.

Discussion

The findings of the present study support the hypotheses derived from our model. Most importantly, the finding that for alpha a significant PLI is accompanied by a decrease in amplitudes (as measured by whole power) clearly suggests the validity of the oscillatory phase resetting model (cf. Fig. 1A). More specifically, as predicted, (i) theta and alpha show a significant increase in PLI during the time window of the P1 and N1 as compared to a prestimulus reference (cf. hypothesis 1), (ii) the

Acknowledgements

This research was supported by the Austrian Science Fund, P-13047.

References (46)

  • M. Steinschneider et al.

    Cellular generators of the cortical auditory evoked potential initial component

    Electroencephalogr. Clin. Neurophysiol.

    (1992)
  • M.G. Woldorff et al.

    Modulation of early auditory processing during selective listening to rapidly presented tones

    Electroencephalogr. Clin. Neurophysiol.

    (1991)
  • E. Basar

    EEG Brain Dynamics. Relation Between EEG and Brain Evoked Potentials

    (1980)
  • E. Basar

    Brain function and oscillations

    (1999)
  • E. Basar

    Brain function and oscillations

    (1999)
  • E. Basar et al.

    Important associations among EEG-dynamics, event-related potentials, short-term memory and learning

    Int. J. Neurosci.

    (1985)
  • E. Basar et al.

    The EEG is a quasi-deterministic signal anticipating sensory-cognitive tasks

  • B.H. Bland

    The physiology and pharmacology of hippocampal formation theta rhythms

    Prog. Neurobiol.

    (1996)
  • M.E. Brandt et al.

    The relationship between prestimulus alpha amplitude and visual evoked potential amplitude

    Int. J. Neurosci.

    (1991)
  • G. Buzsaki et al.

    High frequency network oscillation in the hippocampus

    Science

    (1992)
  • G. Buzsaki et al.

    Oscillatory and intermittent synchrony in the hippocampus: relevance to memory trace formation

  • M. Doppelmayr et al.

    Individual differences in brain dynamics: important implications for the calculation of event-related band power measures

    Biol. Cybern.

    (1998)
  • R.G. Eason

    Visual evoked potential correlates of early neural filtering during selective attention

    Bull. Psychon. Soc.

    (1981)
  • Cited by (275)

    • Theories of brain functioning and the quantitative electroencephalography

      2023, Introduction to Quantitative EEG and Neurofeedback: Third Edition
    • Deciphering the code: Identifying true gamma neural oscillations

      2022, Experimental Neurology
      Citation Excerpt :

      Importantly, this may be relevant to reports of auditory evoked gamma power deficits in cognitive disorders, such as schizophrenia (Leicht et al., 2010), as suppression of the auditory P50 response has also been reported in this disorder (Boutros et al., 1993). While substantial literature indicates that some ERP features may arise from changes in dynamics of oscillatory activity (Klimesch et al., 2004), simply demonstrating an increase in gamma band power associated with these responses is insufficient to demonstrate increased power of oscillatory events. This distinction is important as elevated evoked gamma power due to transients or single-cycle waveform components may arise due to distinctly different neurophysiological mechanisms compared with sustained gamma oscillatory rhythms.

    View all citing articles on Scopus
    View full text