Visual P2 component is related to theta phase-locking
, a, aaaR. Freunberger, W. Klimesch, M. Doppelmayr and Y. Höller
aDepartment of Physiological Psychology, University of Salzburg, Institute of Psychology, Hellbrunnerstr. 34, A-5020 Salzburg, Austria Received 21 June 2007; revised 29 August 2007; accepted 30 August 2007. Available online 5 September 2007.
In this study we investigated the hypothesis whether P2-related differences tested in a visual priming paradigm are associated with theta phase-locking. We recorded the EEG from 31 electrodes and calculated phase-locking index and total power differences for frequencies between 2 and 20 Hz. ERPs (event-related potentials) were analyzed for P1, N1 and P2 components. P2 showed strongest task-related amplitude differences between congruent and incongruent targets. A source analyses was performed for the P2 component using sLoreta that revealed local generators of the P2 in parieto–occipital regions. Phase-locking
analyses showed specific effects in the theta range (4–6 Hz) appearing
in time windows at around the P2 component. We draw the conclusion that phase-locked theta reflect top-down regulation processes mediating information between memory systems and is in part involved in the modulation of the P2 component.
Keywords: Theta; Oscillations; PLI; Total power; Phase-locking
There is increasing evidence that brain oscillations play a fundamental role in the modulation or even generation of event-related potentials
,  and . With the actual study we want to outline the (ERPs)
possible relationship between ERP-P2 component (at around 200 ms post-stimulus) and human theta oscillations (4–6 Hz) in a visual
semantic priming paradigm.
Theta oscillations have been investigated intensively in both humans ,
 and  and animals ,  and  during the last few years.
Animal research showed that theta reflects an important mechanism for the temporal encoding of episodic events . In human EEG studies theta
responses were found to mirror memory retrieval  and ,
memory-capacity demands  and central executive functions such as task switching . Sauseng et al.  found that fronto-parietal theta
coupling is associated with top-down processing and reflects a mediating mechanism between memory systems.
It was proposed that the visual P2 component is involved in cognitive processes (cf. ) such as working memory  and  memory
performance  and semantic processing . P2 has been found mainly in
priming tasks and was interpreted as a correlation of repetition suppression ,  and . For example, in an auditory cuing paradigm it has been shown that validly cued targets were characterized by reduced
. P2 amplitudes as compared to invalidly cued targets
We addressed the question whether theta oscillations are also relevant in conscious semantic priming. Thus, we established a paradigm where
presented items ensured a top-down activation of semantic categories. A second aim of using this paradigm was that we wanted to ensure, to find ‘classical’ P2 effects as found in priming paradigms (e.g. repetition suppression). Using different EEG measures such as ERPs, total power and PLI (phase-locking index, ) we intend to show physiological
correlates of semantic priming.
In our study subjects performed a task involving photographs of real objects. An item (picture of either a living or non-living object) was presented and followed by a retention interval after which a target was shown that had to be classified as living or non-living. Items were
supraliminal so that a conscious processing could be enabled. presented
To summarize our hypotheses, we assume that underlying processes in our task are (i) top-down modulation that is probably reflected in reaction time differences, (ii) repetition suppression characterized by P2 amplitude differences and (iii) a time-dependent relationship between P2 and theta phase-locking.
Twenty subjects (10 females) with mean age = 22.10 years (S.D. = 2.90) participated in our study. All subjects were right handed and none reported mental or neurological disturbances. The written informed consent was given by the subjects and the experiment was conducted in accordance with the Declaration of Helsinki . Subjects had normal or
corrected to normal vision. Two subjects were excluded because of too abundant ocular artifacts which resulted in a sample of 18 subjects for data analyses.
For our paradigm we used photographs of living (e.g. plants and animals) and non-living (e.g. vehicles and buildings) objects already used in a previous picture detection task . The stimuli covered a visual angle
of 4.15? × 6.22?. Each subject was seated at 1.3 m in front of a
computer monitor (refresh rate = 75 Hz) while the electroencephalogram (EEG) was recorded. Items were presented for 66 ms followed by a mask (randomly distorted image of black and white color blobs) that was presented for 200 ms (cf. Fig. 1). Subjects were asked to concentrate on
the presented objects and to try to figure out which kind of object was presented without giving a behavioral response. After a 1744 ms retention interval the targets were presented for 500 ms and the subjects had to respond as fast as possible to the semantic category of the presented targets belonged to (i.e. ‘living’ or ‘non-living’) by pressing a
button, respectively. The targets were either identical to the items or non-identical, the following cases were possible for identical items and targets: (i) living item–living target; (ii) non-living
item–non-living target and for non-identical items; (iii) living item–non-living target (iv) non-living item—living target. For further
analyses we defined targets as congruent (item and target belong to same semantic category) or as incongruent (item and target do not belong to same semantic category).
Display Full Size version of this image (32K)
Fig. 1. Experimental procedure. After the target was presented, the subjects had to indicate whether the image shows a living or a non-living object.
EEG-signals were recorded using a 32-channel biosignal amplifier system (SynAmps, Neuroscan Inc.) with a sampling frequency of 1000 Hz. Signals were referenced to linked earlobe electrodes and acquired within a bandwidth of 0.15–70 Hz with a notch filter at 50 Hz. Thirty-one Ag–AgCl
electrodes were mounted according to the extended international of 10–20
system using an EasyCap. Impedances were kept below 8 kΩ. To control for
vertical eye artifacts an additional bipolar EOG-channel was used. For ERP analyses the data were filtered from 1 to 30 Hz and a baseline from ?100 to 0 ms was subtracted. We investigated the P1, N1 and P2 components by performing a peak-detection, semi-automatically. The time frame for the P1 was set between 70 and 120 ms, for the N1 between 120 and 170 ms and for the P2 between 170 and 270 ms.
ERP source analyses were performed on basis of sLoreta algorithm (for technical details see ). sLoreta calculates the standardized current source density (CSD) of a total of 6430 voxels at 5 mm spatial resolution producing zero localization error. Source analyses were performed for the P2 component only.
PLI describes the variance of a phase-distribution of an oscillation at
a certain time point (for detail see ). A PLI value of 1 indexes maximal
phase-locking (no variance) whereas a value of 0 means no phase-stability (maximal variance). PLI values were estimated on basis of Gabor-wavelet transformation for an epoch length of ?500 to 500 ms around targets and
within 2–20 Hz (frequency resolution = 1 Hz). Frequencies were averaged to reveal 2 Hz bins to reduce data. The post-stimulus interval was split into five consecutive 100 ms time windows.
For total power analyses single trial data was Gabor-filtered and subsequently averaged for frequencies from 2 to 20 Hz in 2 Hz steps. For statistical analyses of the ERP components we used parietal and occipital electrode sites where all components could be clearly detected. We performed two-way ANOVAs with the factor Electrode (P3, P4, Pz, Po3, Po4, Po7, Po8, O1, O2 and Oz) and Condition (congruent, incongruent) for amplitude and latency values separately. For PLI estimates two-way ANOVAs
were calculated for every 2 Hz frequency bin and for each electrode with factors time (0–100 ms, 100–200 ms, 200–300 ms, 300–400 ms and
400–500 ms) and Condition (congruent, incongruent). For total power values the same ANOVAs were performed as for the PLI but with an additional pre-stimulus interval ranging from ?100 to 0 ms. sLoreta comparisons were
performed on basis of paired-sample t-tests, the critical value for a
significance level of 1% was established by a permutation method using 5000 randomizations.
Subjects showed on average 95.3% correct responses for congruent and 95.8% for incongruent targets. Mean reaction times were 531.2 ms for congruent and 554.4 ms for incongruent targets. Reaction times differed significantly between the two conditions (t(17) = ?2.64; p < .05),
indicating faster reaction times for congruent compared to incongruent targets (cf. Fig. 2A).
Display Full Size version of this image (64K)
Fig. 2. (A) Reaction times were significantly slower to incongruent targets compared to congruent. Vertical lines depict standard error of means. (B) ERPs of CT and IT: incongruent targets were characterized by a stronger N1 and P2 amplitude. (C) Left side: the theta (4–6 Hz) filtered
ERP waveforms for the two conditions for electrode Po7. Differences between the conditions are apparent in the N1-P2 time window. Right side: topographical plots of the filtered theta-ERP are depicted for the P2 peak at 208 ms for congruent and 212 ms for incongruent targets. Arrows lead to electrode Po7.
ANOVAs for the P1 amplitude and latency differences showed no significant results. ANOVA for the N1 amplitude resulted in a significant main effect for Electrode (F = 3.81; p < .05) and a significant interaction 9/153
(Condition × Electrode: F = 2.73; p < .05). Post hoc comparisons 9/153
(t-tests) showed effects at electrodes O2, Po4 and Pz, indicating a stronger negative deflection for incongruent targets. The two-way ANOVA for the P2 component resulted in a significant main effect for Electrode (F = 28.26; p < .001), as well as in a significant 9/153
Condition × Electrode interaction ( = 2.95; < .01), indicating Fp9/153
higher amplitudes for incongruent targets compared to congruent targets (cf. Fig. 2A). Post hoc t-tests showed that this effect was apparent at electrode Po7.
As can be seen in Fig. 3 current source density is strongest at
parieto–occipital and temporal areas for both conditions. The performed t-test did not yield significant differences between congruent and incongruent targets.
Display Full Size version of this image (92K)
Fig. 3. Depicted are the sLoreta current source density plots of the grand-average ERPs for congruent and incongruent targets at P2 amplitude peak. Note that congruent targets show a slightly more right hemispherical activation and incongruent more left lateralized activity. Both conditions show CSD-activity at parieto–occipital and temporal regions.
t-Tests did not yield significant differences between the conditions. Results of the two-way ANOVAs for PLI and total power estimates are plotted in Table 1 and Table 2, respectively. PLI shows strong effects on electrode
Po7 for frequency bands of 2–4 Hz and 4–6 Hz in time windows including the P2 component (t2 and t3). Additionally, theta (4–6 Hz) shows a
significant main effect for Condition at electrodes Po7, Po4, Po8 and O1.
In general incongruent targets show enhanced theta PLI in a late time window indicated by the significant Condition × Time interaction for
4–6 Hz. This effect might be due to the fact that PLI values for congruent
targets reach faster at the baseline level again because they show decreased theta PLI in early time windows (t2, t3) that are not significant (Fig. 4).
P3 Pz P4 Po7 Po3 Po4 Po8 O1 Oz O2
Condition2–4 Hz , F(1,17)
Time, 9.09.106.574.034.16 2********F(4,68)
Conditiont3: * * *
× time, IT >(t4: (t2: (t2:
F(4,68) CT; CT >IT >IT >
t4: IT) CT) CT)
Condition5.524.675.039.124–6 Hz , F(1,17) * * * **
Time, 8.316.53 F(4,68) ** **
P3 Pz P4 Po7 Po3 Po4 Po8 O1 Oz O2
5.334.52** 4.504.10* 6.074.373.67(t1: 5.883.23* * (t2, Condition** * * CT >** * (t3, (t4, t3, × time, (t5: (t5: (t5: IT; (t5: (t5: t5: t5: t4, F(4,68) IT >IT >IT >t4, IT >IT >IT >IT >t5: CT) CT) CT) t5: CT) CT) CT) CT) IT >IT > CT) CT)
*p < .05; **p < .01.
P3 Pz P4 Po7 Po3 Po4 Po8 O1 Oz O2
Condition, 4.702–4 Hz F(1,17) *
Time, 26.724.518.919.618.715.622.214.312.117.1 3**8**6**2**9**5**2**7**2**7**F(5,68)
4.383.15* * (t2, (t2: Condition t3: IT >× time, IT > CT; F(5,68) CT; t4: t4: CT >CT > IT) IT)
Condition, 4.68 4–6 Hz F(1,17) *
P3 Pz P4 Po7 Po3 Po4 Po8 O1 Oz O2
F(5,68) 2** 2** 3** 1** ** ** 6** 8** ** 0**
*p < .05; **p < .01.
Display Full Size version of this image (45K)
Fig. 4. PLI and total power estimates are shown for the 4–6 Hz frequency
range across time. Note that PLI shows significant differences in the P2 time window whereas total power did not. Vertical lines represent standard error of means.
Total power estimates show significant interactions in the delta range for Po3 and Po7 only, but not in the theta range.
We found a task-related ERP components peaking at 200 ms, which we considered as P2, showing smaller amplitudes for congruent compared to incongruent targets (cf. Fig. 2A). These ERP results can be interpreted
in terms of a repetition suppression mechanism (cf. ). Repeated
stimuli lead to a reduction of cortical activity and, hence, induce a sharpening process . The P2 component has also been related to several cognitive processes, such as feature detection and retrieval (see ).
In a cued attention task it was shown that invalid compared to valid trials were represented by higher P2 amplitudes , which would accord to the
presented data. Rossell et al.  found in a semantic priming task, using word-pairs of semantically congruent and incongruent words, that in