NeuroImage 143 (2016) 223–234

Contents lists available at ScienceDirect

NeuroImage
journal homepage: www.elsevier.com/locate/neuroimage

Is P3 a strategic or a tactical component? Relationships of P3

sub-components to response times in oddball tasks with go, no-go
and choice responses
Rolf Verleger a,b,n, Nils Grauhan a, Kamila Śmigasiewicz a
a
Department of Neurology, University of Lübeck, Germany
b
Institute of Psychology II, University of Lübeck, Germany

art ic l e i nf o a b s t r a c t

Article history: P3 (viz. P300) is a most prominent component of event-related EEG potentials recorded during task
Received 4 July 2016 performance. There has been long-standing debate about whether the process reflected by P3 is tactical
Accepted 22 August 2016 or strategic, i.e., required for making the present response or constituting some overarching process.
Available online 26 August 2016
Here, we used residue iteration decomposition (RIDE) to delineate P3 subcomponents time-locked to
Keywords: responses and tested for the temporal relations between P3 components and response times (RTs). Data
RIDE were obtained in oddball tasks (i.e., tasks presenting two stimuli, one rarely and one frequently) with
Decision rare and frequent go, no-go, or choice responses (CRs). As usual, rare-go P3s were large at Pz and rare no-
Context updating go P3s at FCz. Notably, P3s evoked with rare CRs were large at either site, probably comprising both go
P300
and no-go P3. Throughout, with frequent and rare responses, P3 latencies coincided with RTs. RIDE
decomposed P3 complexes into a large CPz-focused C-P3 and an earlier Pz-focused response-locked R-P3.
R-P3 had an additional large fronto-central focus with rare CRs, modeling the no-go-P3 part, suggesting
that the process reflected by no-go P3 is tightly time-locked to making the alternative response. R-P3
coincided with the fast RTs to frequent stimuli and C-P3 coincided with the slower RTs to rare stimuli.
Thus, the process reflected by C-P3 might be required for responding to rare events, but not to frequent
ones. We argue that these results are nevertheless compatible with a tactical role of P3 because responses
may not be contingent on stimulus analysis with frequent stimuli.
& 2016 Elsevier Inc. All rights reserved.

1. Introduction et al., 2010), reduced motor-related activation (Kopp et al., 1996;

Event-related potentials (ERPs) are voltage changes recorded
from the scalp in temporal coincidence to events (Luck and Kap-
penman, 2012). When some task has to be performed with these
events, a prominent component of the ERP is the broad positive
potential termed P3 (Ritter et al., 1968) or P300 (Donchin and
Cohen, 1969). P3 amplitudes are larger after rarely than after fre-
quently occurring stimuli presented in random series. This oddball
effect (Duncan-Johnson and Donchin, 1977; Squires et al., 1975) is
focused at parietal midline (Pz) when rare stimuli require re-
sponses (go P3) and at more anterior sites (FCz) when rare (or
even equiprobable) stimuli require that responses be withheld
(no-go P3) (Pfefferbaum et al., 1985; Pfefferbaum and Ford, 1988).
No-go P3 has often been interpreted as indicating inhibition
(Pfefferbaum et al., 1985) or, alternatively, response conflict (Smith
n
Corresponding author at: Klinik für Neurologie, Universität Lübeck, D 23538
Lübeck, Germany.
E-mail address: rolf.verleger@neuro.uni-luebeck.de (R. Verleger).
Verleger et al., 2006) or attention-modulated monitoring (Polich,
2007). Go-P3, considered to be P3 proper, has been the object of
numerous hypotheses, of which the context-updating hypothesis
(Donchin, 1981) is most wide-spread (cf. Polich, 2007) which
states that P3 reflects updating of models about the environment.
The nature of the relationship between go-P3 and overt re-
sponses has remained controversial, and the precise relationship
of no-go P3 to inhibitory processes is still a matter of study. The
present study was designed to contribute to these issues.
A major cornerstone of context-updating hypothesis is the as-
sumption that (go-)P3 reflects a strategic rather than a tactical
process. In using these terms, Donchin and Coles (1988) argued
that the process underlying P3 is not involved in deciding on re-
sponses (e.g., by key-press) to the present stimulus but rather in
some overarching process occurring in parallel or afterwards. In
Donchin and Coles (1988) context-updating hypothesis, this
overarching process is proposed to be an active process of chan-
ging the metacontrol setting of priorities, biases, probabilities,
which may be prompted by a conflict between new information
and expectations derived from these settings (cf. Kamp et al.

http://dx.doi.org/10.1016/j.neuroimage.2016.08.049
1053-8119/& 2016 Elsevier Inc. All rights reserved.
224 R. Verleger et al. / NeuroImage 143 (2016) 223–234
(2013) for a recent account of the hypothesis). Some others have
maintained that P3 does relate to the decision on how to respond
to a given stimulus (Hillyard and Kutas, 1983). Pertinent evidence
is the temporal relation between P3 and RTs. Two aspects are re-
levant here. First, it is at issue when P3 reaches its peak. If this
occurs before or at overt responses then it is reasonable to assume
that P3 reflects some process leading to responses. In contrast, the
strategic interpretation is supported if P3 reaches its peak mark-
edly after overt responses. This latter result was indeed reported
by Kutas et al. (1977) for a subset of trials. However, since this
subset consisted of trials with incorrect responses, it may be ar-
gued that these positive potentials were not stimulus-evoked P3s
but rather error-evoked positivities, evoked by the incorrect re-
sponse (Pe; Falkenstein et al., 2000; Leuthold and Sommer, 1999;
though see Shalgi et al. (2009), for a divergent interpretation of
Pe). In support of the tactical interpretation, a number of recent
studies have found that, when P3 is time-locked to RT and aver-
aged across trials, the peak of P3 occurs precisely at the moment of
responding (Saville et al., 2011; Verleger et al., 2005) at least in
young participants (Cid-Fernández et al., 2016; though see Verle-
ger et al. (2013)) closely resembling the time-course of random-
walk processes leading to decisions (Kelly and O'Connell, 2013;
O'Connell et al., 2012; Twomey et al., 2015). On the other hand,
when all stimuli require the same responses, and responding,
therefore, is very fast, P3 may be clearly later than RTs (Berchicci
et al., 2016). In the present study, P3 latencies and RTs were
compared for frequent and rare responses in the oddball task.
According to the tactical view, RTs and P3 latencies were expected
to coincide, both occurring early for frequent stimuli and late for
rare stimuli.
A second aspect is whether the P3 complex contains compo-
nents that are time-locked to overt responses. If present, and if
occurring before RTs, such components would conform to the
tactical interpretation. The existence of such subcomponents has
been postulated since long, based on common latency shifts of RTs
and the entire P3 complex or of parts of it (Falkenstein et al., 1994;
Verleger, 1997) and based on comparisons between averages
across trials that were time-locked to stimuli, as usual, and that
were time-locked to responses: When containing response-related
components, average waveforms should have a clearer or a dif-
ferent structure in response-locked than in stimulus-locked wa-
veforms. There have been some hints in favor of this assumption
(Berchicci et al., 2016; Cid-Fernández et al., 2016) but in most
publications that compared stimulus- and response locked P3s
waveforms appeared quite similar to each other (Johnson et al.,
2003; Kelly and O'Connell, 2013; O'Connell et al., 2012; Saville
et al., 2011; Twomey et al., 2015; Verleger et al., 2005) except for
pathological cases (Verleger et al., 2013).
Reasons for this similarity might be that stimulus-locked wa-
veforms still contain response-related processes, and response-
locked waveforms still contain stimulus-related processes and,
moreover, that components specific to stimulus or response pro-
cessing might be covered by the presence of parts of the P3
complex that are neither time-locked to stimuli nor to responses.
In order to disentangle stimulus-locked, response-locked, and re-
latively independent components from each other, residue itera-
tion decomposition (RIDE) has been developed (Ouyang et al.,
2015a). RIDE aims at resolving superposition of stimulus-related
(S), central (C), and response-related (R) component clusters. RIDE
achieves this goal by iteratively estimating these three clusters
across single trials, starting from a crude model of the C cluster
and refining this model during the following iterative procedure
by distinguishing it from the stimulus-locked and response-locked
portions. In previous applications, the major part of P3b was
modeled by the C cluster while the R cluster often consisted of a
phasic positivity overlying the C cluster, and the S cluster might
include an early P3a-type component overlapping the C cluster
(Ouyang et al., 2011; Ouyang et al., 2013; Ouyang et al., 2015a;
Verleger et al., 2014). In the present study, RIDE was applied to
specifically study the timing of the C and R clusters with respect to
RTs, for frequent and rare responses in oddball tasks. For the C
cluster, which we may regard as P3b proper, the tactical view on
P3 predicts that it will culminate at the moment of responding,
whereas the strategic view allows for more flexible temporal re-
lations between C clusters and RTs. The R cluster, being strictly
time-locked to responding by definition, may be expected to occur
at the very time of responding, where it might consist of motor-
cortex activation for key-pressing or of somatosensory reafference
from key-pressing. In the strategic view, the R cluster is acciden-
tally overlapping with P3 proper, whereas the tactical view might
consider the R cluster as genuine part of the P3 complex.
The oddball task usually consists of frequent no-go and rare go
stimuli (termed no-go/go in the following). Comparisons between
frequent and rare stimuli are therefore, confounded by the dif-
ference between no-go and go. To disentangle the two factors, we
additionally applied a go/no-go task that consists of frequent go
and rare no-go stimuli. Thereby we had frequent go stimuli from
the go/no-go task and rare go stimuli from the no-go/go task, both
of which could be decomposed by RIDE. Additionally, the go/no-go
task, with its rare no-go stimuli, would yield no-go P3s with their
fronto-central maximum (cf. above). We wished to apply the RIDE
rationale to no-go P3s, too, asking whether the process reflected
by fronto-central no-go P3s is time-locked to the stimuli or is a
central process, or is time-locked to non-responding. But the time-
point of non-responding is not given, making answers to this
question impossible without making further assumptions. The
further assumption made here was that rare choice responses
(CRs) might serve as a proxy for rare no-go responses. Only few
studies have compared CR tasks with go/no-go tasks. The available
evidence (Smith et al., 2010) suggests that the fronto-central part
of CR-P3 is as large as no-go P3. This hypothesis was tested in the
present study by applying an oddball task with CRs. Thereby, there
were three tasks (no-go/go, go/no-go, CR) with frequent and rare
stimuli, resulting in six conditions (Table 1). In the absence of overt
responses, two of these six conditions (frequent no-go and rare
no-go) were not available for RIDE decomposition into S, C, and R
clusters. The four conditions that were available for RIDE decom-
position were: frequent go, rare go, frequent CR, and rare CR. A
natural consequence of this design was that comparisons were
also made between CRs and go responses. Since the RIDE com-
putations were done independently for CR and go responses, this
comparison would provide an assessment of the stability of the
RIDE solutions.
2. Material and methods
2.1. Participants
Twelve young adults participated. They were 6 women and
6 men, aged 20 to 26 years (mean¼24 y). 10 were right-handed,
1 was ambidextrous, and 1 was left-handed. Informed written
consent was obtained and 15 € were paid. Participants reported
Table 1
The six conditions of the present study: frequent and rare stimuli were presented in
three tasks.
no-go/go go/no-go CR
Frequent no-go go choice1
Rare go no-go choice2
 R. Verleger et al. / NeuroImage 143 (2016) 223–234
normal or corrected-to-normal vision and no history of neurolo-
gical disorders.
2.2. Stimuli and procedure
Participants were seated in a comfortable armchair in a dar-
kened room, with about 1.2 m viewing distance from the com-
puter screen. A computer keyboard was put on their lap. Response
keys were left ctrl and right ctrl, to be pressed with the index
fingers. Controlled by a Presentation 14.5 program (www.neurobs.
com) this computer presented the stimuli, recorded responses,
and sent stimulus and response codes to another computer that
recorded EEG.
One of the two black letters X and U (Helvetica, 35 points)
appeared at the center of a light grey 17″ screen for 200 ms. X and
U were presented in random order with probabilities of 80/20%,
but without immediate repetitions of the rare letter. The next
stimulus followed 900 ms after the response when a response was
required, otherwise 1.7 s after stimulus onset. There were 250
stimuli in each block. The experiment consisted of two parts, key-
press and count. The count part always followed the key-press
part and will not be reported in this paper. In the key-press part,
the left key was always assigned to X and the right key to U. There
were three tasks (Table 1): no-go/go (press to rare), go/no-go (press
to frequent), CR (choice-response). One of the six possible orders of
these three tasks was presented to two participants each. This task
triplet was presented twice, once with frequent X and rare U (re-
quiring frequent right and/or rare left key-presses) and once with
frequent U and rare X (requiring frequent left and/or rare right
key-presses). Order of these two triplets was counterbalanced
across participants.
2.3. EEG recording and processing
EEG was recorded with Ag/AgCl electrodes (Easycap, www.
easycap.de) from 60 scalp sites, including eight midline positions
from AFz to Oz and 26 pairs of symmetric left and right sites.
Further electrodes were placed at the nose-tip for off-line re-
ference and at Fpz as connection to ground. On-line reference was
Fz. For artifact control, EOG was recorded, vertically (vEOG) from
above vs. below the right eye and horizontally (hEOG) from po-
sitions next to the outer tails of the eyes. Voltages were amplified
from DC to 250 Hz by a BrainAmp MR plus, A–D converted and
stored at 500 Hz per channel. Off-line processing was done with
Brain–Vision Analyzer software (version 2.03). Data were re-re-
ferenced to the nose-tip, low-pass filtered at 25 Hz (Butterworth
zero phase filters, attenuation of 12 dB/octave), and segmented
from 200 ms before stimulus onset to 1000 ms afterwards. In
editing for artifacts, trials were scrutinized with a three-step
procedure. First, trials with gross artifacts were rejected from
analysis. Second, artifacts of ocular origin (blinks above all) were
subtracted out. Third, trials with artifacts were rejected from
analysis. The first editing for gross artifacts was done lest the en-
suing ocular correction would be distorted by such artifacts. (In
our experience, this risk applies in particular to the regression
calculated for non-blink trials). To this end, trials were rejected
when neighboring data points differed by more than 50 mV or
when minimum and maximum points in any EEG channel differed
by more than 250 mV (except EOG and AF3, AFz, AF4, lest trials
would be rejected for blinks). Then, ocular artifacts were corrected
by linear regression, separately for trials with and without blinks
(Gratton et al., 1983) as implemented in the Brain–Vision Analyzer
software. Finally, with the large voltages produced by blink arti-
facts being subtracted out, artifact-checking proper was per-
formed. To this end, data were referred to mean amplitudes of the
100 ms before stimulus onset as baseline in each channel, and
225
trials were rejected when voltages exceeded 7150 mV in any EEG
channel. The remaining trials were included if the first key-press
response was correct (in go and CR trials) or no key was pressed
(in no-go trials).
Trials were pooled from blocks with left and right-hand re-
sponses, separately for frequent and rare stimuli in each of the
three tasks. The mean number of included trials across partici-
pants was at least (lowest numbers of the three tasks) 366 for
frequent stimuli and 63 for rare stimuli, with minimum numbers
in single participants of 286 and 37. For conventional analysis,
trials were averaged separately for rare and frequent stimuli in
each task in each participant. For RIDE analysis of the four con-
ditions with overt responses, single-trial data were exported from
Brain Analyzer, decomposed into S, C, and R clusters (see Section
2.4.3 for details) separately for each condition, averaged across
trials for each cluster and condition, and re-imported to Brain
Analyzer.
2.4. Data analysis
2.4.1. Behavior
Percentages of wrong responses were determined, which were
pressing the incorrect or no key (within 150–1000 ms) in go-trials
and any key in no-go trials. Response times (RTs) of key-pressing
with respect to stimulus onset were analyzed for correct re-
sponses. Left and right responses in the same condition were
pooled and RTs were averaged across trials in each task, separately
for frequent and rare targets. Likewise, RTs of incorrect responses
to rare stimuli were analyzed, but median rather than mean RT
was taken as measure, due to single outlying values.
2.4.2. Conventional ERP analysis
Forming distinct peaks in all conditions, P3 components were
quantified as most positive peak 250–650 ms after stimulus onset
at midline sites Fz, FCz, Cz, CPz, Pz, POz, Oz. Latencies and am-
plitudes were measured.
2.4.3. Residue iteration decomposition
RIDE analysis (using Matlab R2014.b) closely followed the
methods described in Ouyang et al. (2015b) as applied by us before
(Verleger et al., 2014) using the toolbox available on http://cns.
hkbu.edu.hk/RIDE.htm. RIDE decomposes ERPs into a stimulus-
locked S cluster, an RT-locked R cluster, and a central cluster C,
defined as being neither fully time-locked to stimulus onsets nor
to RTs. The temporal reference points (latencies) for deriving S and
R (LS and LR) are stimulus onsets and RTs, respectively, but vary
over trials for C. These LCs are initially estimated in each single trial
as reflecting some global waveform and then are iteratively im-
proved. Based on the information given by the estimated LC and by
the reference points LS and LR, the contents of S, C, and R get
iteratively dissociated from each other. As mentioned above, de-
composition was done in conditions with overt responses only,
separately for each condition, which were frequent go (from the
go/no-go task), rare go (from the no-go/go task), frequent CR, and
rare CR.
(1) Initial estimation of C latency. Initial estimation of LC used
each participants' average from 250 ms to 650 ms after sti-
mulus onset as a template. LC was then estimated in each
single trial by the shift needed to maximize the cross-corre-
lation of the single-trial data to the template. Cross-correlation
time courses were calculated for each electrode and were
averaged across all electrodes. The time-point of the max-
imum in this averaged cross-correlation was defined as initial
estimate of LC in each single trial.
(2) Iterative decomposition. Next, the three component clusters S,
226 R. Verleger et al. / NeuroImage 143 (2016) 223–234

C, and R were dissociated from each other in an iterative way,

starting from the initial setting S(t)¼ C(t)¼ R(t) ¼0 (i.e., the
template function used in step 1 is no longer used here). In
each step, RIDE obtains an estimate of S, C, and R respectively.
To estimate S, RIDE subtracts C and R from each single trial and
aligns the residues to LS in order to obtain S as the median
waveform (to safeguard against outliers) of the single-trial
waveforms. The same procedure is applied to obtain C and R,
using the time-markers LC and LR, respectively. The whole
procedure is iterated until convergence. The waveforms are
restored from median to mean after convergence. Decom-
position is performed for data from each electrode separately,
but the same time markers are used for all electrodes.
(3) Iteratively improving the latency estimation of C. So far, LC had
been obtained by rough estimation with the cross-correlation
method (step 1). After having obtained a new estimate of C in
step (2), this estimate was improved by the following outer
loop of iteration: S and R were subtracted from each single
trial and cross-correlations were computed between the re-
sidue and the C component shifted in time. The time point
with the highest cross-correlation (averaged across the scalp
and low-pass filtered at 5 Hz to reduce high-frequency noise)
was used as new estimate for LC. This new estimate of LC was
then again fed into step (2) to decompose S, C and R until
convergence. These new estimates of S, C, and R were then
again fed into step (3), iterating the loops of steps (2) and
(3) until convergence of both the latency LC and the compo-
nent clusters S, C, and R. During the iterations, evolution of the
LC estimates for each single trial was constrained to be
monotonic, i.e., iterations were terminated once LC changed
direction relative to the preceding iterations. The template of C
used for the cross-correlation calculations spanned 400 ms,
from 250 ms to 650 ms after onset of the letter. The midpoints
of the 400 ms single-trial waveforms to be cross-correlated to
C varied from a minimum of 250 ms (50–450 ms) to a max-
imum of 650 ms (450–850 ms).

P3 sub-components were determined in each participants'
averaged C and R clusters in each condition by measuring latencies
and amplitudes of the most positive peaks 250–650 ms after sti-
mulus onset at each midline site (Fz, FCz, Cz, CPz, Pz, POz, Oz) in
the C and R clusters.
Fig. 1. Behavioral results. Percentages of error trials are displayed in the upper panel
and response times (RTs) in the lower panel. Bold black lines are data from go trials,
thin black lines from no-go trials (for error percentages only) and grey lines from
choice responses (CRs). Errors in go trials are missing responses, in no-go trials false
alarms, and in CR trials wrong and missing responses.

2.5. Statistical analysis 3. Results

Analyses of variance (ANOVAs) were used for statistical ana-
lyses. For analysis of error rates and RTs, factors were Frequency
(frequent vs. rare stimulus) and Response (CR vs. go for RTs; go vs.
no-go and no-go vs. CR for error rates). Note that the go and no-go
levels of the Response factor by necessity come from different
tasks: frequent go responses are from the go/no-go task, frequent
no-go responses from the no-go/go task, correspondingly for no-
go responses. For P3 latencies and amplitudes, factors were Fre-
quency, Recording Site (Fz, FCz, Cz, CPz, Pz, POz, Oz), and Response
(CR vs. go, or CR vs. no-go). For P3 peaks of RIDE's C and R clusters,
P3 subcomponent (C vs. R cluster) was used as additional factor.
For comparing P3 latencies to RTs, factors were Frequency and
Response, and Measure (P3 vs. RT). This was done for P3 latencies
from Pz and from FCz (where go and no-go P3 were largest, re-
spectively) for conventionally measured P3 and for RIDE's C and R
peaks.
Partial eta-squared will not be explicitly reported, being easily
derived from the reported F-values by the formula ηp2 ¼(F/df)/
(1 þF/df). Greenhouse-Geisser corrected p-values and, for trans-
parency, uncorrected degrees of freedom will be reported for ef-
fects of Recording Site.
3.1. Behavior
3.1.1. Error rates
Key-press error rates are depicted in the upper panel of Fig. 1.
There were almost no errors with go responses (0.7% on average)
but very many false alarms with rare no-go stimuli (17.2%) and
even more errors with rare stimuli in the CR task (26.8%). There-
fore, main and interaction effects of Frequency and Response were
significant when comparing go and no-go responses (all
F1,11 Z26.4, p o.001) and when comparing no-go to CR responses
(all F1,11 Z 10.0, p r.009). In the latter comparison, more wrong
responses were committed in the CR task than with no-go stimuli
not only with rare stimuli (26.8% vs. 17.2%; F1,11 ¼12.6, p ¼.005) but
also consistently, though on a low level, with frequent stimuli
(2.0% vs. 0.5%; F1,11 ¼17.8, p ¼.004).
3.1.2. Response times
Mean key-press RTs are presented in the lower panel of Fig. 1.
RTs were about 100 ms faster with frequent than with rare re-
sponses (Frequency: F1,11 ¼66.2, p o.001) and were about 30 ms
faster in the CR task than with go responses (Response: F1,11 ¼6.1,
 R. Verleger et al. / NeuroImage 143 (2016) 223–234 227

Fig. 2. ERP grand means: waveforms and P3 topography. Waveforms depict voltages recorded from FCz and Pz (vs. nose reference). Unit of x axis is ms. Time-point zero is letter
onset. Unit of y-axis is mV, with positive polarity plotted downwards. Tickmarks denote 10 mV. Dashed lines denote frequent stimuli, solid lines rare stimuli. Bold black lines
are data from go trials, thin black lines from no-go trials and grey lines from choice responses (CRs). Mean amplitudes of P3 peaks at midline are depicted in the upper right
panel. Like with the waveforms, unit of y-axis is mV, with positive polarity plotted downwards, line-style is the same as with the waveforms. Distributions of P3 peak
amplitudes across the entire scalp are shown by the maps. View is from above, with Cz in the center and ear level ( ¼120°) at the outer rim, nose is above. Red is positive,
white is zero. Color ranges are individually scaled in each map to range symmetrically between 7 125% of absolute maximum amplitude (e.g., with maximum at þ 8 mV, the
range would be 7 10 mV; 125% was chosen rather than 100% to avoid oversaturated coloring). (For interpretation of the references to color in this figure legend, the reader is
referred to the web version of this article.)

p ¼.03). These two factors did not interact (F1,11 ¼0.3, n.s.).
3.2. ERPs: conventional analysis
Fig. 2 displays grand-mean waveforms from FCz and Pz. The
major features to be presented in the following analyses are the
topography of P3 in the CR task, compared to go-P3 and no-go P3
Table 2
ANOVA results (F-values and their p-values) on P3 amplitudes in the key-press
tasks. Comparison of choice responses (CR) with go and no-go responses are
compiled in the left and right halves, respectively. Effects including Response (CR
vs. go, or CR vs. no-go) are entered in the respective columns. p-values were en-
tered when pr .10. Values are printed in bold when p r .050.
CR vs. go CR vs. no-go

(Section 3.2.1) and the common analysis on P3 latencies and RTs Response Response
(Section 3.2.2).
P3 AMPLITUDES
3.2.1. Analysis of amplitudes 7.2 8.7
.02 .01
Topographic information on the distribution of P3 peak am- Midline site 11.1 13.8 14.3 4.6
.001 .001 <.001 .03
plitudes is displayed in Fig. 2 as maps and as profiles of midline Frequency 0.3
129.0 29.6 133.2
amplitudes. ANOVA results for P3 amplitudes recorded from the <.001 .001 <.001
Freq.  Midl. site 3.9 8.2 11.0 5.8
midline are compiled in Table 2. .06 .007 .003 .02
CR vs. go: P3 was larger with rare than frequent stimuli (Fre-
quency F1,11 ¼129.0, p o.001) and largest at CPz, Cz, and Pz (Re-
cording Site, F6,66 ¼11.1, p¼ .001). Rare CR-P3s differed from rare
go-P3s by their larger fronto-central amplitudes: P3 was larger significant for frequent stimuli, F6,66 ¼ 4.1, p ¼.04).
with CRs than with go responses for rare stimuli (Response: CR vs. no-go: P3 (being likewise larger with rare than frequent
F1,11 ¼7.2, p¼ .02; Response  Frequency F1,11 ¼ 29.6, p ¼.001; effect stimuli, Frequency F1,11 ¼ 133.2, p o.001) was largest at FCz, Cz, CPz
of Response for rare stimuli F1,11 ¼14.1, p¼ .003; for frequent sti- (Recording Site, F6,66 ¼ 14.3, p o.001), particularly for rare stimuli
muli F1,11 ¼0.0, n.s.) with a more anterior topography, focused at (Frequency  Recording Site F6,66 ¼11.0, p ¼.003). P3 was larger in
FCz, for CR than for go (Response  Recording Site: F6,66 ¼13.8, CR than with no-go stimuli (Response F1,11 ¼8.5, p ¼.01) though
p ¼.001) particularly for rare stimuli (Response  Recording Si- with different topographies for frequent and rare stimuli (Re-
te  Frequency: F6,66 ¼8.2, p ¼.007): Effects of Response were sig- sponse  Recording Site: F6,66 ¼4.6, p ¼.03; Response  Recording
nificant at Fz, FCz, Cz, CPz, Pz for rare stimuli (F1,11 Z8.0, p r.02) Site  Frequency: F6,66 ¼5.8, p ¼ .01). For the large P3 with rare
and were not significant at any single site for frequent stimuli stimuli, P3 was larger in CR than with no-go responses at posterior
(although the interaction of Response  Recording Site did become sites (effects of Response separately at CPz, Pz, POz F1,11 Z 6.4,
228 R. Verleger et al. / NeuroImage 143 (2016) 223–234
Fig. 3. Peak latencies of P3 in conventionally measured waveforms and in RIDE's C and
R clusters. Mean latencies, averaged over participants are shown from FCz and Pz.
Unit of y-axis is ms. The horizontal dashed lines denote mean RTs (taken from
Fig. 1). Thin lines denote frequent stimuli ("fq"), bold lines rare stimuli. Black lines
are data from go trials, grey lines from choice responses (CRs).
p r.03) and did not differ at anterior sites Fz and FCz (F1,11 r0.7, n.
s.). For frequent stimuli, P3 was larger in CR than with no-go re-
sponses throughout, except for the most posterior sites (effects of
Response at Fz, FCz, Cz, CPz, Pz F1,11 Z13.0, p o.004).
To summarize, P3s evoked by rare stimuli in the CR task were
as large as with go responses at posterior sites (POz, Oz) and as
with no-go responses at FCz: CR-P3 apparently combined parietal
go-P3 and fronto-central no-go P3.
3.2.2. Analysis of latencies
3.2.2.1. CR vs. go. The two left-most positions of Fig. 3 display
mean peak latencies at FCz and Pz for CR and go responses. In
analysis on all seven midline sites, P3 peaked earlier in CR than
with go responses (Response: F1,11 ¼23.4, p ¼.001) and peaked
earlier with frequent than with rare stimuli, except at the most
anterior sites Fz and FCz, with largest mean difference at Pz
(Frequency F1,11 ¼19.0, p ¼ .001; Recording Site  Frequency:
F6,66 ¼6.1, p ¼.008). Other effects were not significant at p o.05.
(The interaction of Frequency and Response suggested by Fig. 3
reached p¼ .08 only).
We compared effects on P3 peak latencies at FCz and Pz with
the above-reported effects on RTs (dashed horizontal lines in
Fig. 3). Effects are compiled in Table 3. P3 latencies at Pz did not
differ in any respect from RTs (all effects of Measure F1,11 ¼1.3, n.s.).
This was in contrast to FCz latencies which, by their being un-
affected by Frequency, unlike RTs (Frequency  Measure
F1,11 ¼12.8, p ¼.004; effect of Frequency on FCz latencies F1,11 ¼ 0.0,
n.s.) were later than RTs for frequent stimuli and tended to be
earlier than RTs for rare stimuli (effects of Measure separately for
frequent stimuli F1,11 ¼6.6, p ¼.03; separately for rare stimuli
F1,11 ¼3.6, p ¼.09).
3.2.2.2. CR vs. no-go. Similarly to the CR vs. go analysis, P3 peaked
earlier in CR than with no-go responses (Response: F1,11 ¼17.7,
Table 3
ANOVA results (F-values and their p-values) on comparisons between RTs and P3 latencies in conditions with overt key-presses. Factors are Measure (RT vs. P3), Response
(CR vs. go), and frequency (frequent vs. rare). C-P3 and R-P3 mean P3 latencies of RIDE's C and R clusters. p-values were entered when p r .10. Values are printed in bold
when p r .050.
P3-FCz P3-Pz
Measure (M) 1.0 0.2
Response 0.9 1.3
xM
Frequency x M 12.8 1.2
.004
RxFxM 2.1 1.1
p¼ .001) and peaked earlier with frequent than with rare stimuli at
posterior sites (CPz and beyond), with largest mean difference at
Pz (Recording Site  Frequency: F6,66 ¼ 4.2, p ¼.04). Other effects
were not significant at p o.05. (Of course, P3 latencies from no-go
conditions had no corresponding RTs to be directly compared
with).
3.3. RIDE analysis of Go vs. choice responses
Fig. 4 displays grand-mean waveforms of S, C, and R clusters at
FCz and Pz as well as midline amplitudes of P3-type components
in the C and R clusters and maps of their topographic distributions.
(No such components were evident in the S-cluster waveforms).
The major features to be presented are that the major part of the
P3 complex is modeled by the C cluster, as expected (Section 3.3.1)
and that, with frequent but not with rare stimuli, this C-cluster-P3
reaches its peak markedly after overt responses (Section 3.3.2),
that the no-go type fronto-central part of P3 in the CR task is
tightly response-locked, thereby being part of the R cluster (Sec-
tion 3.3.1), and that R-cluster P3 coincides with, or precedes, RTs
with frequent and rare stimuli (Section 3.2.2).
3.3.1. Analysis of amplitudes
ANOVA results for R and C cluster amplitudes at midline sites
are compiled in Table 4. Amplitudes were larger in C than R
clusters (F1,11 ¼54.2, p o.001). Interactions of the Cluster factor
with the other factors (s. Table 4) were resolved by separate ana-
lyses of either cluster.
Main characteristics of the C cluster were its being larger for
rare than frequent stimuli (Frequency F1,11 ¼ 50.0, po .001) and
having a distinct maximum at Cz and CPz (Recording Site
F6,66 ¼ 31.9, p o.001) with CPz being the focus for rare stimuli and
Cz for frequent ones (Recording Site  Frequency F6,66 ¼15.6,
po .001). There was no main effect of Response (F1,11 ¼0.1, n.s.).
The interactions of Response  Recording Site (F6,66 ¼ 3.6, p ¼.045)
and Response  Recording Site  Frequency (F6,66 ¼4.1, p¼ .03) in-
dicated that amplitudes tended to be more positive in CR than
with go responses for rare stimuli at anterior sites, in contrast to
frequent responses. However, this effect was weak: effects of Re-
sponse  Frequency or of Response did not become significant at
any single recording site.
The R cluster was generally larger for rare than frequent stimuli
(Frequency F1,11 ¼ 61.0, p o.001) and was strongly affected by the
Response factor. Amplitudes were larger with CR than go re-
sponses, except at posterior sites (Response F1,11 ¼28.6, p o.001;
Response  Recording Site F6,66 ¼23.2, p o.001; effects of Re-
sponse were significant at Fz, FCz, Cz, CPz, Pz, F1,11 Z10.1, p r.009).
This effect was more distinct with rare than frequent stimuli
(Response  Frequency: F1,11 ¼ 16.2, p ¼ .002; effect of Response for
rare stimuli: F1,11 ¼26.7, p o.001; for frequent stimuli: F1,11 ¼3.8,
p¼ .08). Also the topographical differences between responses
were more distinct for rare stimuli (Response  Recording Si-
te  Frequency F6,66 ¼11.9, po .001) where go responses had their
RT vs. C-P3-FCz C-P3-Pz R-P3-FCz R-P3-Pz
0.6 0.6 3.4 3.9
.09 .07
2.4 5.0 0.5 4.5
.047 .058
43.3 15.7 6.4 1.6
<.001 .002 .03
0.6 2.9 1.5 0.9
 R. Verleger et al. / NeuroImage 143 (2016) 223–234 229

Fig. 4. RIDE cluster grand means: waveforms and P3 topography. Like in Fig. 2, waveforms depict voltages recorded from FCz and Pz (vs. nose reference). Unit of x axis is ms.
Time-point zero is letter onset. Unit of y-axis is mV, with positive polarity plotted downwards. Tickmarks denote 10 mV. Dashed lines denote frequent stimuli, solid lines rare
stimuli. Bold black lines are data from go trials and grey lines from choice responses (CRs). No-go trials were not analyzed because R clusters cannot be estimated. Mean
amplitudes of P3 peaks at midline od the C and R clusters are depicted in the lower left panel. Like with the waveforms, unit of y-axis is mV, with positive polarity plotted
downwards, line-style is the same as with the waveforms. Distributions of P3 peak amplitudes across the entire scalp are shown by the maps, separately for the C and R
clusters (beneath the respective waveforms). See legend of Fig. 2 for details on map scales.

Table 4
ANOVA results (F-values and their p-values) on P3-like amplitudes in the R and C RIDE clusters estimated in conditions with overt key-presses. Results of the overall ANOVA
on both clusters are compiled in the top half of the table, and separate analyses on R and C clusters are compiled in the bottom half. Effects including Response (choice
response vs. go), RIDE cluster (R vs. C) and their interaction are entered in columns in the upper half, and of Response only in columns of the two parts of the lower half.
Effects including Midline Site (Fz, FCz, Cz, CPz, Pz, POz, Oz), Frequency (frequent vs. rare) and their interaction are entered in rows. p-values were entered when p r .10.
Values are printed in bold when p r .050.

CR vs. go

Effects of row labels Response RIDE cluster Response x RIDE cluster

Cluster amplitudes
Effects of column labels 12.0 54.2 5.5
.005 <.001 .04
Midline site 15.2 20.0 20.5 4.8
<.001 <.001 <.001 .02
Frequency 80.4 22.4 3.2 1.1
<.001 .001 .10
Frequency x Midl. Site 4.0 15.3 13.7 2.7
.053 <.001 <.001 .07

R cluster C cluster

Effects of row labels Response Effects of row labels Response

Effects of column label 28.6 0.1

<.001
Midline site 2.5 23.2 31.9 3.6
<.001 <.001 .045
Frequency 61.0 16.2 50.0 1.0
<.001 .002 <.001
Frequency  Midl. Site 2.1 11.9 15.6 4.1
<.001 <.001 .025

maximum at Pz and POz whereas CR amplitudes were largest were largest at anterior sites (effects of Response at Fz, FCz, Cz,
anteriorly, at FCz (Responses  Recording Site for rare stimuli CPz, Pz F1,11 Z8.4, p r.02). Topographical differences between re-
F6,66 ¼19.3, p o.001) such that differences between responses sponses were less pronounced for frequent stimuli
230 R. Verleger et al. / NeuroImage 143 (2016) 223–234
Fig. 5. Illustration of temporal relations between waveforms and RTs. ERP data (left panel) are the same as in Fig. 2, except for no-go trials which are left out here and for the
insertion of RTs (same as in Figs. 1 and 3). C-cluster and R-cluster data are the same as in Fig. 4, except for the different arrangement (C and R clusters are here shown
overlapping, separately for frequent stimuli, middle panel, and rare stimuli, right panel) and, again, for the insertion of RTs. C cluster waveforms are plotted as thin lines, R
cluster waveforms as bold lines.
(Response  Recording Site F6,66 ¼7.4, p¼ .003) but amplitudes
were still larger in CR than with go responses at Fz, FCz, Cz,
(F1,11 Z 7.0, pr .02). Thus, a major characteristic of the R cluster
was the task dependency of its topographical distribution, with
posterior positivity for the go responses in the go/no-go tasks and
additional anterior positivity in the CR task, obviously modeling
the no-go-P3-like fronto-central P3 in this task.
3.3.2. Analysis of latencies
Mean peak latencies of P3-type components in the C and R
clusters are displayed in Fig. 3, and latency relationships may also
be appreciated in Fig. 5 where waveforms of R and C clusters are
presented in the same graphs. Positive peaks were reached earlier
in R clusters than in C clusters (Cluster: F1,11 ¼16.1, p ¼.002) and
earlier with frequent than with rare stimuli (Frequency: F1,11 ¼31.2,
p o.001). No other effects reached the p o.05 threshold.
Like with P3 peak latencies (Section 3.2.2), the P3-type peak
latencies of the C and R clusters at FCz and Pz were compared with
RTs (cf. Fig. 3). ANOVA effects are compiled in Table 3.
C-P3 latencies were less affected by Frequency than RTs (Fre-
quency  Measure for FCz: F1,11 ¼43.3, p o.001; for Pz: F1,11 ¼15.7,
p ¼.002). Thereby, C-P3 latencies for frequent stimuli were ap-
preciably later than RTs, both at FCz (F1,11 ¼11.4, p ¼.006) and at Pz
(F1,11 ¼ 7.2, p¼ .02). For rare stimuli, Pz latencies did not differ from
RTs (F1,11 ¼0.5, n.s.) while FCz latencies tended to occur earlier than
RTs (F1,11 ¼4.6, p ¼.056). Additionally, unlike RTs, C-P3 latencies
were not earlier for CR than go responses (Response  Measure
F1,11 ¼5.0, p ¼.047).
R-cluster-P3: By their early latencies, R-P3 peaks at FCz were as
early as RTs to frequent stimuli, and earlier than RTs for rare sti-
muli (Frequency  Measure F1,11 ¼ 6.4, p ¼.03; effect of Frequency
on FCz latencies F1,11 ¼0.7, n.s.; effect of Measure with frequent
stimuli F1,11 ¼ 0.0, n.s.; with rare stimuli F1,11 ¼11.7, p ¼.006). In
contrast, R-P3 peaks at Pz, by their being early for frequent and
late for rare stimuli, did not differ from RTs in terms of the fre-
quency effect. By not differing between CR and go responses, R-P3
peaks at Pz tended to be earlier than RTs with go responses (Re-
sponse  Measure F1,11 ¼4.5, p ¼.058; effect of Measure with go
responses F1,11 ¼4.5, p¼ .058; with CR F1,11 ¼ 1.4, n.s.).
4. Discussion
4.1. P3 in choice response vs. P3 in go/no-go
In several of our recent papers on P3 with oddball series we
had used choice-response (CR) tasks (e.g., Verleger et al., 2014;
Verleger and Śmigasiewicz, 2016) assuming that P3 evoked by rare
and frequent stimuli would not differ between CR tasks and the
mostly used no-go/go tasks. The present results suggest that P3s
obtained in these two tasks do resemble each other but are also
different: P3 evoked by rare stimuli in the CR task indeed re-
sembled P3 evoked by rare go stimuli at sites posterior to Pz but
turned out to be much larger than this P3 at FCz, resembling the
P3 evoked by rare no-go stimuli. We conclude that CR-P3 is a
combination of parietally focused go-P3 and fronto-centrally fo-
cused no-go P3. This does make sense, because participants must
abandon their preferred response with rare stimuli in the CR task
in the same way as in the go/no-go task, which may be reflected in
no-go P3, and must activate the rare response in the same way as
in the no-go/go task, which may be reflected in go P3.
We would have liked to corroborate this conclusion by re-
ference to previous evidence but, much to our surprise, we have
not become aware of previous studies in which CR-P3s were
compared to go-P3 and no-go P3 in oddball tasks. Smith et al.'s
(2010) work quoted in the Introduction did show the similarity of
CR-P3 to no-go P3 but had its focus on sequence effects and,
therefore, used equiprobable stimuli rather than frequent and rare
ones.
4.2. Behavioral results
Responses were faster with frequent than rare stimuli, in line
with hundreds of studies (e.g., Miller, 1998) most probably re-
flecting participants' prevailing tendencies of executing the fre-
quent response. Moreover, many errors were committed with rare
stimuli in the CR task and in the go/no-go task where participants
simply continued executing the frequent response rather than
switching to the other response (in the CR task) or withholding the
response (in the go/no-go task). Thus, obviously, participants had
strong tendencies of simply repeating the frequent response. Mean
error rates of 17% (for rare no-go stimuli) and 27% (for rare CRs) are
remarkable indeed, in such simple tasks. Similarly high and even
higher false-alarm rates have been occasionally reported for rare
 R. Verleger et al. / NeuroImage 143 (2016) 223–234
no-go stimuli in similar tasks (Harper et al., 2014; Inzlicht and Al-
Khindi, 2012; Nieuwenhuis et al., 2004), so the present results do
not seem exceptional. Tendencies for stereotyped responding may
have been somewhat boosted in the present study by the fixed
response-stimulus intervals (of 900 ms) which may have pro-
moted repetitive, rhythmic behavior.
RTs were slightly but reliably smaller with CRs than with go
responses, equally for frequent and rare responses. This runs
counter to results obtained by Donders (1868/1969) where "Task
C" (go/no-go) had faster RTs than "Task B" (CR), likewise, e.g.,
Gomez et al. (2007), Ulrich et al. (1999). Gomez et al. (2007)
suggested that the main difference between the two tasks is in the
setting of response criteria. In line with this suggestion, it does not
seem far-fetched to assume that motor activation might have been
on a constantly higher level in the present CR task, where some
response had to be made in 100% of trials, than in the go/no-go
tasks where responses had to be given in 80% or 20% of trials only.
4.3. Lags between RTs and P3 latencies
There were no reliable differences between RTs and con-
ventionally measured P3 latencies at Pz: Key-press responses and
P3 peaks occurred early for frequent stimuli and late for rare ones,
and both measures occurred somewhat earlier with CR than with
go responses. Nor was the constant lag between the two measures
reliably different from zero. In other words, P3 always reached its
peak at the time of the overt key-press. This speaks for the tactical
view on P3, according to which P3 reflects some process leading to
the response and needed for responding.
However, the RIDE decomposition provided a more differ-
entiated picture, to be discussed next.
4.4. RIDE decomposition of ERPs
RIDE decomposed P3 complexes into large C (central) and
somewhat smaller R (response-related) clusters. Contributions of
the S (stimulus) cluster to the P3 complex were negligible. Notably,
the S cluster included the N2 component that was evoked when
rare stimuli required no-go or CR responses. Thereby, RIDE re-
solved overlap of N2 with the simultaneously ongoing P3 complex,
like in Ouyang et al. (2013). Focusing on the P3 complex, the
present discussion will deal with the C and R clusters only. In go
trials, the C-P3 (P3-portion of C clusters) had its topographical
focus at Cz and CPz, and the R-P3 at Pz and POz, thereby ac-
counting in their combination for the Pz focus of conventionally
averaged go-P3. In rare CR trials, the R clusters had an additional
focus at FCz, thereby accounting for the additional nogo-P3-like
focus at FCz in conventionally averaged CR-P3.
4.4.1. C cluster results
C-P3s were much larger for rare than frequent stimuli, cap-
turing a major feature of the P3 complex (replicating Verleger
et al., 2014). C-P3s also reached their peaks later with rare than
frequent stimuli, similarly to conventional P3s though not to equal
extent. Rather, C-P3 latencies were later than P3 latencies for
frequent stimuli, reducing the latency difference between frequent
and rare stimuli in C-P3s to about half the size of the latency
difference in conventionally measured P3. Correspondingly, with
frequent stimuli these peak latencies of C-P3 occurred significantly
later than the overt responses. In contrast, there was no reliable
difference between RTs and C-P3 latencies for rare stimuli. Thus,
the tactical view on P3 may be applied to C-P3 latencies in rare
trials but not in frequent trials: C-P3s might reflect some process
needed for responding with rare stimuli but not with frequent
stimuli because there the C clusters occurred too late. This de-
viation from the pattern predicted by the tactical view is in
231
accordance with the strategic view on P3 (Donchin and Coles,
1988). Moreover, C-P3s were not significantly affected by response
mode (CR vs. go), neither in their latencies nor in their amplitudes,
in contrast to conventionally measured P3s. This conforms to the
notion expressed, among others, by Kutas et al. (1977) stating that
P3 latencies are pure measures of the time needed for stimulus
processing, independent of response requirements. Any exceptions
from this rule might be attributed to overlapping response-related
processes not belonging to P3 proper. Such processes are classified
by RIDE as belonging to the R cluster, to be discussed next.
4.4.2. R cluster results
The R cluster appeared in the present data as a cluster in its
true sense, consisting of different components, all of which closely
correlated with RTs (by definition of the R cluster). On the one
hand, there was Pz-focused positivity, with its peak principally not
differing from RTs, both with frequent and rare responses. The one
exception were rare go stimuli where the peak was reliably earlier
than its corresponding RT. We do not have a good explanation for
this exception. Possibly there was some overlap from the frontally
focused activation to be discussed below. In principle, then, these
Pz-focused activations might reflect some process closely asso-
ciated with responding, like somatosensory reafference from
moving the fingers or from feeling the key (e.g., Bötzel et al., 1997).
Lacking further information about these processes, these con-
jectures must remain speculations. To learn more about con-
tributing processes and their timing, we are currently replicating
the present experiment and recording EMG from the finger-mov-
ing muscles. An alternative interpretation of the Pz-focused part of
the R cluster is that this is the tactical portion of the P3 complex,
reflecting decision and activation of response alternatives leading
to the key-press response. This will be discussed below.
Additionally to the Pz-focused activation, with rare CRs, the R
cluster included the fronto-centrally focused no-go-type P3. Thus,
this component which might reflect inhibition of frequent re-
sponses, both in go/no-go and CR tasks, was most closely time-
locked to the process of executing the alternative rare response,
otherwise it would not have been included in the R cluster. We
conclude that not performing the frequent response and per-
forming the rare response are rigidly connected in the present
data. We came to a similar conclusion by probing the activation of
the two motor cortices with transcranial magnetic stimulation
during the time-period of the lateralized readiness potential: in-
creasing motor activation for one response was firmly linked to
decreasing motor activation for the alternative response (Verleger
et al., 2009; cf. the ERP study by Praamstra and Seiss, 2005).
It was not a matter of course that the no-go type P3 would be
such closely correlated to RTs, as revealed by RIDE: The fronto-
central no-go-type P3 might have been as independent from re-
sponding as is the large portion of P3 assigned to the C cluster, and
thus might have become part of the C cluster. But this did not
happen. We conclude that no-go P3 is not simply the no-go
equivalent of the usual response-associated P3 but adds to this
latter P3.
4.4.3. Methodological considerations
Like any method, RIDE has its prerequisites. Ideally, LC and LR,
the latencies of the C and R clusters, should be uncorrelated across
trials. The more the true latencies of any two clusters are corre-
lated across trials, the less separable will these clusters be (Fig. 4 in
Ouyang et al., 2015a). Therefore, it may well be that the small
differences between response modes (go vs. CR) that were ob-
tained on the C cluster, or the effects of frequency on the Pz-fo-
cused part of the R cluster are due to such incomplete separation
of clusters, with some variance of the R cluster being "mis-
allocated" (Wood and McCarthy, 1984) to the C cluster.
232 R. Verleger et al. / NeuroImage 143 (2016) 223–234
It is unrealistic to assume that a distinct C component will
occur in every single trial. Thus, RIDE's estimate of LC will always
be noisy. To reduce noise, it may be suggested that trials be dis-
carded where the crosscorrelation with the template remains be-
low some threshold (cf. Ford et al., 1994; Gasser et al., 1983;
Woestenburg et al., 1981). However, by clipping one segment of
the noise distribution, this procedure would introduce some bias.
Moreover, because the RIDE algorithm is iterative, the result will
be biased more and more to some direction or, on the contrary,
will be prevented from leaving some incorrect initial position.
Similar considerations apply to focusing the estimate of LC on
some recordings where the signal is clearest. Again, this might
improve results but also entails the risk of biasing the results of
the iterative algorithm. This is not to say that the RIDE algorithm
cannot be improved. But this would transcend the scope of the
present paper.
We may note that RIDE stands alone in finding such a de-
composition of ERPs, compared to other methods of multivariate
decomposition used in ERP research. For example, decomposition
by oscillatory frequency, which may be a straightforward method
in some circumstances (e.g, Harper, et al., 2014) will not approx-
imate the present solution because the large C and R clusters with
rare stimuli have similar oscillatory frequencies (Fig. 5). Principal
Component Analysis of time-points (e.g., Dien et al., 2005; Möcks
and Verleger, 1991) will not come close to the present solution
either, because the R cluster will not be identified as a coherent
component, by its latencies of occurrence differing between tasks,
recording sites, and stimulus frequencies (Figs. 3 and 4; cf. Möcks,
1986). Furthermore, simple-structure rotations of PCA solutions,
e.g., Varimax and Promax, will have difficulties in assigning the
overlapping time-courses of the C and R clusters to different
components (cf. results for P1 and N1 in Dien (2010); further
Möcks and Verleger (1986); Wood and McCarthy (1984)). Nor will
Independent Component Analysis of recording sites (Makeig et al.,
2002) arrive at solutions similar to the present one, because scalp
topographies of the R cluster vary widely across conditions (Fig. 4)
whereas ICA components are defined by their topography (cf.
Fig. 4 of Ouyang et al. (2015a)). This is not to say that the solutions
achieved by the mentioned methods will be less "true" than the
RIDE solution. But what can be said is that the solution obtained
by RIDE is special and, in our view, appropriate to the present
questions.
4.5. P3: a strategic or a tactical component?
We have asked whether the process reflected by P3 is strategic
or tactical, i.e., independent of present responding or leading to the
present response. A major argument in favor of the strategic view
has been that P3 waveforms may sometimes reach their peak after
responding only. The present conventional measures of P3's peak
did not conform to this argument: In all conditions, P3s did not
peak later than at the time of overt responses. In particular, this
also applied to the fast responses made with frequent stimuli:
Even here, peak latencies at Pz were as early as these RTs, around
300 ms (Fig. 5, left panel). This evidence supports the tactical view.
However, RIDE decomposition provided a more nuanced pic-
ture: R-P3 (i.e., the Pz-focused part of the R cluster) peaked at the
time of overt responses (with frequent stimuli) or slightly before
(with rare stimuli) thereby accounting for the early peak with
frequent stimuli in conventionally measured P3 which coincided
with RT. C-P3 (i.e., the P3-type peak of the C cluster) did coincide
with responses to rare stimuli but was reliably later than re-
sponses to frequent stimuli. Since C-P3 may be considered the
representation of P3 proper (Ouyang et al., 2011; 2013; 2015a;
Verleger et al., 2014) we may conclude that the overlap of R-P3
provided a biased picture of the temporal relationships between
Fig. 6. A model of behavior in the oddball task. See text for detailed explanation.
P3 and RTs in conventionally measured EEG. C-P3's behavior
suggests that, rather than always being needed for responding, as
conventional measurement suggested, the process reflected by P3
may be needed for rare responses but not for frequent ones. More
parsimoniously, one may agree with Donchin and Coles (1988)
suggestion that, being not needed for frequent responses, the P3
process is generally not needed for responding and, therefore, is a
purely strategic process.
4.5.1. A model of processing in the oddball task
However, this parsimonious model of equal strategic functions
of P3 with frequent and rare stimuli is justified only if processing
indeed does not differ between frequent and rare stimuli. Yet, the
very large error rates with rare stimuli, where frequent responses
were committed in spite of clear evidence to the contrary, suggest
that processing for making the correct response in the present
oddball tasks did differ between frequent and rare stimuli. As il-
lustrated in Fig. 6, these large error rates may be taken to suggest
that participants' behavior went along two paths of processing,
termed in Fig. 6 the fast and the ultra-fast path. We will argue that
the process reflected by P3 is sidestepped with the ultra-fast path
but forms the core of the fast path.
(a) The ultra-fast path develops from priming by preceding trials
where the frequent response had been made. This path then
consists of executing this frequent response in response to
stimulus onset, with less weight put on identity of the sti-
mulus, i.e., as if this was a simple-response task. This will be a
correct response with frequent stimuli and a wrong response
with rare stimuli.
(b) Since these are not simple-response tasks and stimulus iden-
tity does matter, participants will identify the stimulus and
then proceed along the fast path, which is slower than the
ultra-fast path. The fast path is based on stimulus-response (S–
R) links, e.g., the associations X - 4 left and U -4 right es-
tablished by instruction and practice at task onset. Identifying
one of the two alternative stimuli will activate their respective
S–R links (cf. Hommel et al., 2001; Hommel et al., 2014). The
utility of these S–R links will differ between frequent and rare
stimuli. With frequent stimuli, where the ultra-fast route may
have been taken, this activation of S–R links will not be needed
for responding any more because the primed response is
 R. Verleger et al. / NeuroImage 143 (2016) 223–234
already being executed. In contrast, with rare stimuli, where
the rare response is not the primed default response, S–R link
activation is required for making the response. We assume
that P3 in general, and C-P3 in particular, reflects this tactical
process of reactivation of S–R links. Therefore, responses may
be executed before C-P3 with frequent stimuli but need the P3
process and, therefore, are executed when C-P3 reaches its
peak with rare stimuli.
(c) For completeness, Fig. 6 also depicts the controlled, non-
automatic slow route of responding. This route has to be taken
when no S–R link is available, e.g., because of complex, rarely
used S–R mapping rules (Verleger et al., 2014). Then, response
selection has to take place instead of simple associative S–R
link activation, and will be reflected by fronto-central nega-
tivity (Verleger et al., 2014, 2015).
5. Conclusion
The present data from oddball tasks with different response
instructions suggest: (1) With rare choice responses P3 consists of
a combination of parietal go-P3 and fronto-central no-go P3.
(2) This no-go P3 part of choice responses is tightly time-locked to
making the other (correct) response. (3) P3 reaches its peak at the
moment of overt responding with both the slower responses to
rare stimuli and the fast responses to frequent stimuli. (4) This
effect with frequent stimuli may be due to overlap of response-
related positivity with P3 proper: In fact, the central component of
P3 (as measured by the C cluster) reaches its peak with frequent
stimuli after responding only. We propose that this is because
responses may not be contingent on stimulus analysis with fre-
quent stimuli.
Acknowledgments
We thank Christopher Schulte for writing a batch procedure for
RIDE analysis.
This work was supported by funding granted to R.V. by the
Deutsche Forschungsgemeinschaft (Ve110/17-1).
References
Berchicci, M., Spinelli, D., Di Russo, F., 2016. New insights into old waves. Matching
stimulus- and response-locked ERPs on the same time-window. Biol. Psychol.
117, 202–215. http://dx.doi.org/10.1016/j.biopsycho.2016.04.007.
Bötzel, K., Ecker, C., Schulze, S., 1997. Topography and dipole analysis of reafferent
brain activity following the Bereitschaftspotential. Exp. Brain Res. 114, 352–361.
http://dx.doi.org/10.1007/PL00005643.
Cid-Fernández, S., Lindín, M., Díaz, F., 2016. Information processing becomes slower
and predominantly serial in aging: characterization of response-related brain
potentials in an auditory-visual distraction-attention task. Biol. Psychol. 113,
12–23. http://dx.doi.org/10.1016/j.biopsycho.2015.11.002.
Dien, J., 2010. Evaluating two-step PCA of ERP data with Geomin, Infomax, Oblimin,
Promax, and Varimax rotations. Psychophysiology 47, 170–183. http://dx.doi.
org/10.1111/j.1469-8986.2009.00885.x.
Dien, J., Beal, D.J., Berg, P., 2005. Optimizing principal components analysis of event-
related potentials: matrix type, factor loading weighting, extraction, and rota-
tions. Clin. Neurophysiol. 116, 1808–1826. http://dx.doi.org/10.1016/j.
clinph.2004.11.025.
Donchin, E., 1981. Surprise!…surprise? Psychophysiology 18, 493–513. http://dx.
doi.org/10.1111/j.1469-8986.1981.tb01815.x.
Donchin, E., Cohen, L., 1969. Anticipation of relevant stimuli and evoked potentials:
a reply to Näätänen. Percept. Motor Skills 29, 115–117. http://dx.doi.org/
10.2466/pms.1969.29.1.115.
Donchin, E., Coles, M.G.H., 1988. Is the P300 component a manifestation of context
updating? Behav. Brain Sci. 11, 357–374. http://dx.doi.org/10.1017/
S0140525X00058027.
Donders, F.C., 1868/1969. Over de snelheid van psychische processen. (On the speed
of mental processes) (W. Koster, Trans.). In: W.G. Koster (Ed.), Attention and
Performance II. Amsterdam: North Holland, pp. 412-431. http://dx.doi.org/10.
233
1016/0001-6918(69)90065-1.
Duncan-Johnson, C.C., Donchin, E., 1977. On quantifying surprise: the variation of
event-related potentials with subjective probability. Psychophysiology 14,
456–467. http://dx.doi.org/10.1111/j.1469-8986.1977.tb01312.x.
Falkenstein, M., Hohnsbein, J., Hoormann, J., 1994. Effects of choice complexity on
different subcomponents of the late positive complex of the event-related
potential. Electroencephalogr. Clin. Neurophysiol. 92, 148–160. http://dx.doi.
org/10.1016/0168 5597(94)90055-8.
Falkenstein, M., Hoormann, J., Christ, S., Hohnsbein, J., 2000. ERP components on
reaction errors and their functional significance: a tutorial. Biol. Psychol. 51,
87–107. http://dx.doi.org/10.1016/S0301-0511(99)00031-9.
Ford, J.M., White, P., Lim, K.O., Pfefferbaum, A., 1994. Schizophrenics have fewer and
smaller P300s: a single-trial analysis. Biol. Psychiatry 35, 96–103. http://dx.doi.
org/10.1016/0006-3223(94)91198-3.
Gasser, T., Möcks, J., Verleger, R., 1983. Selavco: a method to deal with trial-to-trial
variability of evoked potentials. Electroencephalogr. Clin. Neurophysiol. 55,
717–723. http://dx.doi.org/10.1016/0013-4694(83)90283-3.
Gomez, P., Ratcliff, R., Perea, M., 2007. A model of the go/no-go task. J. Exp. Psychol.:
Gen. 136, 389–413. http://dx.doi.org/10.1037/0096-3445.136.3.389.
Gratton, G., Coles, M.G.H., Donchin, E., 1983. A new method for off-line removal of
ocular artifact. Electroencephalogr. Clin. Neurophysiol. 55, 468–484. http://dx.
doi.org/10.1016/0013-4694(83)90135-9.
Harper, J., Malone, S.M., Bernat, E.M., 2014. Theta and delta band activity explain N2
and P3 ERP component activity in a go/no-go task. Clin. Neurophysiol. 125,
124–132. http://dx.doi.org/10.1016/j.clinph.2013.06.025.
Hillyard, S.A., Kutas, M., 1983. Electrophysiology of cognitive processing. Annu. Rev.
Psychol. 34, 33–61. http://dx.doi.org/10.1146/annurev.ps.34.020183.000341.
Hommel, B., Müsseler, J., Aschersleben, G., Prinz, W., 2001. The theory of event
coding (TEC): a framework for perception and action planning. Behav. Brain Sci.
24, 849–878. http://dx.doi.org/10.1017/S0140525X01000103.
Hommel, B., Memelink, J., Zmigrod, S., Colzato, L.S., 2014. Attentional control of the
creation and retrieval of stimulus-response bindings. Psychol. Res. 78, 520–538.
http://dx.doi.org/10.1007/s00426-013-0503-y.
Inzlicht, M., Al-Khindi, T., 2012. ERN and the placebo: a misattribution approach to
studying the arousal properties of the error-related negativity. J. Exp. Psychol.:
Gen. 141, 799–807. http://dx.doi.org/10.1037/a0027586.
Johnson Jr., R., Barnhardt, J., Zhu, J., 2003. The deceptive response: effects of re-
sponse conflict and strategic monitoring on the late positive component and
episodic memory-related brain activity. Biol. Psychol. 64, 217–253. http://dx.
doi.org/10.1016/j.biopsycho.2003.07.006.
Kamp, S.-M., Brumback, T., Donchin, E., 2013. The component structure of ERP
subsequent memory effects in the Von Restorff paradigm and the word fre-
quency effect in recall. Psychophysiology 50, 1079–1093. http://dx.doi.org/
10.1111/psyp.12090.
Kelly, S.P., O'Connell, R.G., 2013. Internal and external influences on the rate of
sensory evidence accumulation in the human brain. J. Neurosci. 33,
19434–19441. http://dx.doi.org/10.1523/JNEUROSCI.3355-13.2013.
Kopp, B., Mattler, U., Goertz, R., Rist, F., 1996. N2, P3 and the lateralized readiness
potential in a nogo task involving selective response priming. Electro-
encephalogr. Clin. Neurophysiol. 99, 19–27. http://dx.doi.org/10.1016/
0921-884X(96)95617-9.
Kutas, M., McCarthy, G., Donchin, E., 1977. Augmenting mental chronometry: the
P300 as a measure of stimulus evaluation time. Science 197, 792–795. http://dx.
doi.org/10.1126/science.887923.
Leuthold, H., Sommer, W., 1999. ERP correlates of error processing in spatial S-R
compatibility tasks. Clin. Neurophysiol. 110, 342–357. http://dx.doi.org/10.1016/
S1388-2457(98)00058-3.
Luck, S.J., Kappenman, E.S., 2012. The Oxford Handbook of Event-related Potential
Components. Oxford University Press, New York. http://dx.doi.org/10.1093/ox-
fordhb/9780195374148.001.0001.
Makeig, S., Westerfield, M., Jung, T.-P., Enghoff, S., Townsend, J., Courchesne, E.,
Sejnowski, T.J., 2002. Dynamic brain sources of visual evoked responses. Sci-
ence 295, 690–694. http://dx.doi.org/10.1126/science.1066168.
Miller, J., 1998. Effects of stimulus-response probability on choice reaction time:
evidence from the Lateralized Readiness Potential. J. Exp. Psychol.: Hum. Per-
cept. Perform. 24, 1521–1534. http://dx.doi.org/10.1037/0096-1523.24.5.1521.
Möcks, J., 1986. The influence of latency jitter in principal component analysis of
event-related potentials. Psychophysiology 23, 480–484. http://dx.doi.org/
10.1111/j.1469-8986.1986.tb00659.x.
Möcks, J., Verleger, R., 1986. Principal component analysis of event-related poten-
tials: a note on misallocation of variance. Electroencephalogr. Clin. Neurophy-
siol. 65, 393–398. http://dx.doi.org/10.1016/0168-5597(86)90018-3.
Möcks, J., Verleger, R., 1991. Multivariate methods in biosignal analysis: application
of principal component analysis to event-related potentials. In: Weitkunat, R.
(Ed.), Digital Biosignal Processing. Elsevier, Amsterdam, pp. 399–458.
Nieuwenhuis, S., Yeung, N., Cohen, J.D., 2004. Stimulus modality, perceptual over-
lap, and the go/no-go N2. Psychophysiology 41, 157–160. http://dx.doi.org/
10.1046/j.1469-8986.2003.00128.x.
O'Connell, R.G., Dockree, P.M., Kelly, S.P., 2012. A supramodal accumulation-to-
bound signal that determines perceptual decisions in humans. Nat. Neurosci.
15, 1729–1735. http://dx.doi.org/10.1038/nn.3248.
Ouyang, G., Sommer, W., Zhou, C., 2015a. Updating and validating a new framework
for restoring and analyzing latency-variable ERP components from single trials
with residue iteration decomposition (RIDE). Psychophysiology 52, 839–856.
http://dx.doi.org/10.1111/psyp.12411.
Ouyang, G., Sommer, W., Zhou, C., 2015b. A toolbox for residue iteration
234 R. Verleger et al. / NeuroImage 143 (2016) 223–234
decomposition (RIDE) – a method for the decomposition, reconstruction, and
single-trial analysis of event related potentials. J. Neurosci. Methods 250, 7–21.
http://dx.doi.org/10.1016/j.jneumeth.2014.10.009.
Ouyang, G., Herzmann, G., Zhou, C., Sommer, W., 2011. Residue iteration decom-
position (RIDE): a new method to separate ERP components on the basis of
latency variability in single trials. Psychophysiology 48, 1631–1647. http://dx.
doi.org/10.1111/j.1469-8986.2011.01269.x.
Ouyang, G., Schacht, A., Zhou, C., Sommer, W., 2013. Overcoming limitations of the
ERP method with residue iteration decomposition (RIDE): a demonstration in
go/no-go experiments. Psychophysiology 50, 253–265. http://dx.doi.org/
10.1111/psyp.12004.
Pfefferbaum, A., Ford, J.M., 1988. ERPs to stimuli requiring response production and
inhibition: effects of age, probability and visual noise. Electroencephalogr. Clin.
Neurophysiol. 71, 55–63. http://dx.doi.org/10.1016/0168-5597(88)90019-6.
Pfefferbaum, A., Ford, J.M., Weller, B.J., Kopell, B.S., 1985. ERPs to response pro-
duction and inhibition. Electroencephalogr. Clin. Neurophysiol. 60, 423–434.
http://dx.doi.org/10.1016/0013-4694(85)91017-X.
Polich, J., 2007. Updating P300: an integrative theory of P3a and P3b. Clin. Neu-
rophysiol. 118, 2128–2148. http://dx.doi.org/10.1016/j.clinph.2007.04.019.
Praamstra, P., Seiss, E., 2005. The neurophysiology of response competition: motor
cortex activation and inhibition following subliminal response priming. J.
Cognit. Neurosci. 17, 483–493. http://dx.doi.org/10.1162/0898929053279513.
Ritter, W., Vaughan Jr., H.G., Costa, L.D., 1968. Orienting and habituation to auditory
stimuli: a study of short term changes in averaged evoked responses. Electro-
encephalogr. Clin. Neurophysiol. 25, 550–556. http://dx.doi.org/10.1016/
0013-4694(68)90234-4.
Saville, C.W.N., Dean, R.O., Daley, D., Intriligator, J., Boehm, S., Feige, B., Klein, C.,
2011. Electrocortical correlates of intra-subject variability in reaction times:
average and single-trial analyses. Biol. Psychol. 87, 74–83. http://dx.doi.org/
10.1016/j.biopsycho.2011.02.005.
Shalgi, S., Barkan, I., Deouell, L.Y., 2009. On the positive side of error processing:
error awareness positivity revisited. Eur. J. Neurosci. 29, 1522–1532. http://dx.
doi.org/10.1111/j.1460-9568.2009.06763.x.
Smith, J.L., Smith, E.A., Provost, A.L., Heathcote, A., 2010. Sequence effects support
the conflict theory of N2 and P3 in the Go/NoGo task. Int. J. Psychophysiol. 75,
217–226. http://dx.doi.org/10.1016/j.ijpsycho.2009.11.002.
Squires, N.K., Squires, K.C., Hillyard, S.A., 1975. Two varieties of long-latency posi-
tive waves evoked by unpredictable auditory stimuli in man. Electro-
encephalogr. Clin. Neurophysiol. 38, 387–401. http://dx.doi.org/10.1016/
0013-4694(75)90263-1.
Twomey, D.M., Murphy, P.R., Kelly, S.P., O'Connell, R.G.O., 2015. The classic P300
encodes a build-to-threshold decision variable. Eur. J. Neurosci. 42, 1636–1643.
http://dx.doi.org/10.1111/ejn.12936.
Ulrich, R., Mattes, S., Miller, J., 1999. Donders's assumption of pure insertion: an
evaluation on the basis of response dynamics. Acta Psychol. 102, 43–75. http:
//dx.doi.org/10.1016/S0001  6918(99)00019-0.
Verleger, R., 1997. On the utility of P3 latency as an index of mental chronometry.
Psychophysiology 34, 131–156. http://dx.doi.org/10.1111/j.1469-8986.1997.
tb02125.x.
Verleger, R., Śmigasiewicz, K., 2016. Do rare stimuli evoke large P3s by being un-
expected? A comparison of the oddball effects between standard-oddball and
prediction-oddball tasks. Adv. Cognit. Psychol. 12, 88–104. http://dx.doi.org/
10.5709/acp-0189-9.
Verleger, R., Jaśkowski, P., Wascher, E., 2005. Evidence for an integrative role of P3b
in linking reaction to perception. J. Psychophysiol. 19, 165–181. http://dx.doi.
org/10.1027/0269-8803.19.3.165.
Verleger, R., Schroll, H., Hamker, F., 2013. The unstable bridge from stimulus pro-
cessing to correct responding in Parkinson's disease. Neuropsychologia 51,
2512–2525. http://dx.doi.org/10.1016/j.neuropsychologia.2013.09.017.
Verleger, R., Baur, N., Metzner, M.F., Śmigasiewicz, K., 2014. The hard oddball: ef-
fects of difficult response selection on stimulus-related P3 and on response-
related negative potentials. Psychophysiology 51, 1089–1100. http://dx.doi.org/
10.1111/psyp.12262.
Verleger, R., Hamann, L.M., Asanowicz, D., Śmigasiewicz, K., 2015. Testing the S-R
link hypothesis of P3b: The oddball effect on S1-evoked P3 gets reduced by
increased task relevance of S2. Biol. Psychol. 108, 25–35. http://dx.doi.org/
10.1016/j.biopsycho.2015.02.010.
Verleger, R., Paehge, T., Kolev, V., Yordanova, J., Jaśkowski, P., 2006. On the relation
of movement-related potentials to the go/no-go effect on P3. Biol. Psychol. 73,
298–313. http://dx.doi.org/10.1016/j.biopsycho.2006.05.005.
Verleger, R., Kuniecki, M., Möller, F., Fritzmannova, M., Siebner, H.R., 2009. How do
the motor cortices resolve an inter-hemispheric response conflict? An ERP-
guided TMS study of the flankers task. Eur. J. Neurosci. 30, 318–326. http://dx.
doi.org/10.1111/j.1460-9568.2009.06817.x.
Verleger, R., Metzner, M.F., Ouyang, G., Śmigasiewicz, K., Zhou, C., 2014. Testing the
stimulus-to-response bridging function of the oddball-P3 by delayed response
signals and residue iteration decomposition (RIDE). Neuroimage 100, 271–280.
http://dx.doi.org/10.1016/j.neuroimage.2014.06.036.
Woestenburg, J.C., Verbaten, M.N., Sjouw, W.P.B., Slangen, J.L., 1981. A statistical
Wiener filter using complex analysis of variance. Biol. Psychol. 13, 215–226.
http://dx.doi.org/10.1111/j.1460-9568.2009.06817.x.
Wood, C.C., McCarthy, G., 1984. Principal component analysis of event-related po-
tentials: simulation studies demonstrate misallocation of variance across
components. Electroencephalogr. Clin. Neurophysiol. 59, 249–260. http://dx.
doi.org/10.1016/0168-5597(84)90064-9.