Theta oscillations and the ERP old/new effect: independent phenomena?
Introduction
The aim of the present study is to test the hypothesis whether those ERP components which are related to (recognition) memory performance are selectively associated with an increase in theta band power. The investigation of this hypothesis is motivated by 3 different lines of evidence, stemming from animal research about hippocampal theta, findings from our laboratory (with human subjects) about a selective memory-related theta synchronization and ERP research about a memory-related old/new effect (remembered ‘old’ items or targets elicit a significantly larger positive component than ‘new’ items or distractors). Findings from these different research areas which will be briefly reviewed in the following 3 paragraphs suggest a direct physiological relationship between the ERP old/new effect and theta activity.
In animal research, convincing evidence for the view that theta synchronization is related to the encoding of new information comes from the fact that long-term potential (LTP), as the most important electrophysiological phenomenon reflecting the encoding of new information, is closely linked to the synchronous activity of the hippocampal theta rhythm. (i) LTP can be best induced with stimulation patterns that mimic theta rhythm (e.g. Larson et al., 1986). (ii) LTP has been demonstrated in several brain regions, but it is most robust and, thus, has been studied most extensively in the hippocampus (Maren et al., 1994). (iii) The induction of LTP occurs primarily during the positive phase of the theta rhythm (Pavlides et al., 1988). (iv) The strength of the induced LTP increases linearly with increasing theta power (Maren et al., 1994). (v) Pharmacological manipulations demonstrate that drugs which decrease theta activity also block learning (Givens and Olton, 1995), whereas drugs that promote the theta rhythm (and, thus, enhance the induction of LTP) also facilitate learning (Stäubli et al., 1994, Stäubli and Fang, 1995). These findings support the view that synchronous hippocampal theta activity is important for the induction of LTP and is related to the encoding of new information in a similar way as LTP is. In addition, studies focusing on the hippocampal theta rhythm in animals have provided good evidence that theta power is related to memory processes (Miller, 1991, Buzsaki et al., 1994).
In a series of experiments, we were able to demonstrate that even in the human scalp EEG, theta band power responds selectively to the encoding and retrieval of new information. This effect was first demonstrated by Klimesch et al. (1994) in a study that consisted of a semantic and episodic memory retrieval task. Subjects first performed a semantic task in which they had to judge whether the sequentially presented words of concept-feature pairs (such as ‘eagle-claws’ or ‘pea-huge’) are congruent. Then, without prior warning, they were asked to perform an episodic recognition task. This was done in an attempt to prevent subjects from using semantic encoding strategies and to increase episodic memory demands. In the episodic task, the same word pairs were presented together with new distractors (generated by repairing known concept-feature words). Now, subjects had to indicate whether or not a particular concept-feature pair was already presented during the semantic task. Because distractors were semantically similar and generated by repairing already presented words, subjects were able to give a correct response only if new episodic information (represented by a specific combination of a concept and feature word) was actually stored in memory. Because pairs of items were presented, the episodic and semantic task can be performed only after the second item of a pair (i.e. the feature) is presented. Thus, the critical issue was to compare the amount of theta synchronization during the presentation of the concept and feature word in the episodic and semantic tasks. The results show that a significant increase in theta power developed only in the episodic task and only when the correct feature word could be retrieved. During the semantic task, however, a significant decrease in upper alpha power was observed. There is, thus, a dissociation between theta synchronization which is maximal during the processing of new (episodic) information and upper alpha desynchronization which is maximal during retrieval and processing of semantic information.
In a series of more recent studies these findings could be replicated and extended by focussing on encoding processes and on memory-related inter-individual differences in band power (Klimesch et al., 1996, Klimesch et al., 1997a, Klimesch et al., 1997b, Klimesch et al., 1997c, Burgess and Gruzelier, 1997, Gevins et al., 1997, Tesche, 1997). With respect to encoding processes we found that during a study phase, those words that could later be remembered in a recall (Klimesch et al., 1996) or recognition task (Klimesch et al., 1997c) exhibited a significantly larger increase in theta power than words that could not be remembered. With respect to inter-individual differences in memory performance, our findings indicate that the extent of upper alpha desynchronization is significantly correlated with semantic memory performance, whereas the extent of theta synchronization is significantly correlated with episodic memory performance (Klimesch et al., 1997a; for related results see also Vogt et al., 1998, Klimesch et al., 2000).
Taken together, the important conclusion is that theta band power of the human scalp EEG increases in response to increasing encoding or retrieval demands in a small frequency window (see Section 2 for the definition of frequency bands), just as hippocampal theta (recorded from animals) does (e.g. Givens, 1996). We assume that our results reflect theta activity that is induced into the cortex via cortico-hippocampal re-entrant loops (cf. Miller, 1991 for a comprehensive review on this topic).
ERP research indicates that correctly remembered items (targets) elicit a significantly larger positive component than distractors (cf. the review in Rugg and Doyle, 1994). This enhancement of a memory-sensitive late positive component (Van Petten and Senkfor, 1996) can be observed during the recognition phase of explicit and implicit memory tasks (although showing a different time course and topography in the latter case, e.g. Rugg et al., 1998) as well as during the encoding of items that will be subsequently remembered (cf. the Dm-effect reported, e.g. Sanquist et al., 1980, Karis et al., 1982, Neville et al., 1986, Fabiani et al., 1986, Fabiani et al., 1990, Karis et al., 1984, Paller, 1993). Many researchers have identified this positive going component as the P300 (e.g. Paller, 1993, Mecklinger, 1998). However, more recent studies have shown that this memory-sensitive late positive component may be composed of or influenced by the N400, P3a, P3b as well as other components (Nielsen-Bohlman and Knight, 1994, Fernández et al., 1998, Rugg et al., 1998). There is good evidence that the medial temporal lobe (with the hippocampal formation as a crucial component of the distributed limbic-cortical network), above all of the left hemisphere, contributes to the generation of the ERP old/new effect, both from intracranial recordings made in epileptic patients (Grunwald et al., 1995, Grunwald et al., 1998, Elger et al., 1997) and from studies in patients with brain lesions (Smith and Halgren, 1989).
In order to test the hypothesis whether there is a relationship between the ERP old/new effect and a memory-related increase in theta band power (larger theta for targets than distractors), a standard recognition paradigm (consisting of 96 target words and distractors) was used in the present study. ERPs were recorded and changes in induced band power (IBP) in individually adjusted frequency bands were calculated by using a method suggested by Klimesch et al. (1998). In contrast to traditional measures of event-related band power which are composed of evoked (phase-locked) and non-evoked (non-phase-locked) EEG components, IBP is devoid of and, thus, lacks evoked EEG activity. Thus, we will be able to test the additional hypothesis whether an increase in theta band power during successful retrieval is contaminated by ERP components.
The basic idea for calculating IBP is to subtract evoked band power from the band pass filtered EEG. As an example, let us consider the theta band and let μ(i)theta represent evoked theta activity at each sample point (i). Evoked activity, μ(i)theta is obtained by averaging the band pass filtered data over the number (n) of trials (j): μ(i)theta=[Σx(i,j)]/n. The procedure for calculating theta IBP is to subtract μ(i)theta from each sample point (i) and each trial (j) of the theta band pass filtered EEG, x(i,j). Then, the differences between x(i,j) and μ(i)theta are squared (in order to obtain a simple measure for band power) and averaged over the j=1, …, n epochs. This squared difference is also termed ‘inter-trial variance’ (Kaufman et al., 1989, Kalcher and Pfurtscheller, 1995).
It is important to note that μ(i)theta is calculated from single trials and represents the event-related potential of the theta band (theta ERP or μtheta). Under normal circumstances, the sum of the band pass filtered ERPs should add up to the standard ERP which is obtained by averaging the raw EEG within certain frequency limits. Let us assume that the ERP is calculated within the same frequency limits as IBP and let us assume (as is the case for the analysis in the present study) that the frequency limits are 0.5 Hz for the lower boundary and 2 Hz above the individual alpha frequency (IAF) for the upper boundary. As explained in Section 2, we use 6 different frequency bands, which are termed ‘subdelta’, ‘delta’, ‘theta’, ‘lower-1 alpha’, ‘lower-2 alpha’ and ‘upper alpha’. These frequency bands cover the entire frequency range from 0.5 Hz to IAF+2 Hz. Within these frequency boundaries, the standard ERP is expected to equal the sum of the band pass filtered ERPs, which is: ERP=μsum=μsubdelta+μdelta+μtheta+μlower-1 alpha+μlower-2 alpha+μupper alpha.
In the unlikely case, however, that single trial evoked responses take the form of half waves with varying latency, the standard ERP will not equal μsum. When calculating standard ERPs, half waves in the theta frequency range may add (average) up to a very slow wave with a ‘frequency’ in the subdelta band. Band pass filtering of these half waves will now lead to different results because filtering distorts half waves (i.e. they are turned into sinus-like ‘wavelets’) which, when averaged, will show a μtheta that is different from the respective component of the regular ERP. However, this discrepancy may occur only in the rather unlikely case that the single trial responses are ‘half wave-like’. This discrepancy will further diminish in relation to how rhythmic (sinus-like) the original single trial evoked response is. If single trial evoked responses already have a rhythmic and approximately sinus-like form, the ERP will, in statistical terms, equal μsum. The reason for this is that averaging of a sinus wave with frequency f will always result in a sinus with the same frequency. Thus, the averaging procedure underlying the calculation of ERPs and that of band pass filtered signals should give (statistically) identical results. Accordingly, the critical test is whether μsum equals the standard ERP. If both are identical in statistical terms (some minor differences are to be expected due to inaccuracies of filters), then the conclusions are that (i) none of the (standard) ERP components are generated by half waves and (ii) the ERP does not contain components which cannot be effectively eliminated by IBP. In such a case we could support the hypothesis that ERPs can be understood and described in terms of a superposition of several event-related oscillations of different frequency domains. This has been suggested by Basar and coworkers for a long time (cf. the overview in Basar, 1998) and was investigated by a series of recent studies (e.g. Yordanova and Kolev, 1998, Spencer and Polich, 1999).
Section snippets
Subjects
Subjects were 25 right-handed students (4 males, 21 females) who participated voluntarily in the experiment. Their mean age was 22.8 years with a range of 18–34 years. Handedness was controlled by asking the subjects about the hand they use in different tasks such as handwriting, throwing a ball, etc. A subject was considered right-handed if he/she indicated use of the right hand for all of these different tasks.
Material
As in several earlier studies, a set of 96 words was used as targets. Subsamples of
Behavioral data
The percentage of hits (i.e. remembered targets) and incorrectly rejected (i.e. not remembered) targets was 78.2 and 21.8%, respectively. From the distractors 82.6% were correctly identified. The false alarm rate (yes response to a distractor) was 17.4%. The mean reaction time for hits was 966 ms (SD 164 ms). For correct rejections subjects needed 1107 ms on average (SD 174 ms).
ERPs
In the left column of Fig. 1, the ERPs for hits and correct rejections recorded from the midline positions, Fz, Cz and
Discussion
The results show 5 basic findings. (i) Remembered words (hits) display a significant ERP old/new effect. They elicit a significantly larger positive component (which we term P300) than correctly rejected words (cf. Fig. 1). (ii) Together with the old/new effect, the P300 is generated by very slow (subdelta) frequencies which lie below the individually determined delta band. (iii) Induced theta and delta but not alpha band power display a highly significant oscillatory old/new effect (cf. Fig. 2
Acknowledgements
This research was supported by the Austrian ‘Fonds zur Förderung der wissenschaftlichen Forschung’, Project P-13047.
References (50)
- et al.
Human temporal lobe potentials in verbal learning and memory processes
Neuropsychologia
(1997) - et al.
Effects of mnemonic strategy manipulation in a Von Restorff paradigm
Electroenceph clin Neurophysiol
(1990) - et al.
Bidirectional modulation of scopolamine-induced working memory impairments by muscarinic activation of the medial septal area
Neurobiol Learn Mem
(1995) - et al.
Alterations of intrahippocampal cognitive potentials in temporal lobe epilepsy
Electroenceph clin Neurophysiol
(1995) - et al.
Discrimination between phase-locked and non-phase locked event-related EEG activity
Electroenceph clin Neurophysiol
(1995) - et al.
“P300” and memory: individual differences in the von Restorff effect
Cognit Psychol
(1984) EEG alpha and theta oscillations reflect cognitive and memory performance: a review and analysis
Brain Res Rev
(1999)- et al.
Episodic and semantic memory: an analysis in the EEG-theta and alpha band
Electroenceph clin Neurophysiol
(1994) - et al.
Brain oscillations and human memory performance: EEG correlates in the upper alpha and theta bands
Neurosci Lett
(1997) - et al.
Event-related desynchronization in the alpha band and the processing of semantic information
Cognit Brain Res
(1997)