Nogo tasks

Nogo tasks

ELSEVIER Electroencephalography and clinical Neurophysiology 96 (1995136-43 Late ERP components in visual and auditory Go/ Nogo tasks M. Falkenstei...

949KB Sizes 13 Downloads 155 Views

ELSEVIER

Electroencephalography

and clinical Neurophysiology 96 (1995136-43

Late ERP components in visual and auditory Go/ Nogo tasks M. Falkenstein

*, N.A. Koshlykova a, V.N. Kiroj a, J. Hoormann and J. Hohnsbein

Institut fiir Arbeitsphysiologie, Universitiit Dortmund, Ardeystr. 67, D-44139 Dortmund (Germany), and ’ Institute for Neurocybernetics, University of Rostou-on-Don, 194/l Stachka ALL, 344104 Rostou-on-Don (Russia)

(Accepted for publication: 20 June 19941

In an audio-visual Go/Nogo paradigm we studied whether the Go/Nogo difference, usually found in the time range of the visual Summary N2, is also present after auditory stimuli, which bears on the common response inhibition hypothesis of this N2 effect. Moreover the possible presence and variation of P300 subcomponents were studied with the goal of clarifying the reasons for the commonly observed P300 topography changes between Go and Nogo trials. To disentangle possible P300 subcomponents we applied a crossmodal divided attention (DA) condition, in which the subcomponents are known to be separated after auditory stimuli in choice tasks. An N2 effect was found after visual but not after auditory stimuli, which is evidence against the response-inhibition hypothesis. After visual stimuli a positive complex (P400) was seen, whereas after auditory stimuli two dissociated components (P400 and P507) were found instead. The P507 had a parietal maximum for both Go and Nogo trials. It was larger and it peaked later in Go than in Nogo trials. The P400 showed topographic differences between Go and Nogo trials, which could be explained by the overlap of the two subcomponents. We assume that (8 both subcomponents have a stable topography across response type, and (ii) the first subcomponent is invariant with response type, whereas the second (which overlaps the first one) is larger and peaks later on Go than on Nogo trials. Keywords: Event-related

potential; N2; Late positive complex; P300; Go/Nogo

Only few experiments have explicitly studied ERP differences between Go and Nogo trials with equally probable targets and non-targets. One difference between Go and Nogo ERPs is a negative displacement in the 200-400 msec range after visual Nogo stimuli, as compared to Go stimuli (Pfefferbaum et al. 1985; Kok 1986; Jodo and Kayama 1992; Eimer 1993). This “N2 effect” has a frontal maximum and is usually interpreted as a real-time correlate of response inhibition which is active on Nogo, but not on Go, trials. If the N2 effect is related to inhibition, it should be present on Nogo trials regardless of stimulus modality. However, the effect appears to be absent or even reversed after auditory stimuli. In the figures of Karlin et al. (1969), as well as of Hillyard et al. (19761, who used auditory stimuli, a larger (i.e., more negative) N2 was seen on Go rather than on Nogo trials. However, the authors did not comment on this. Banquet et al. (1981), who also used auditory stimuli, also reported a more negative N2 for Go than for Nogo trials. In the present study we used both visual and auditory stimuli within the same experiment in order to resolve the contradic-

* Corresponding author. Tel.: + 49-231-1084-277; Fax: +49-2311084-401; E-mail: [email protected].

task

tions concerning the N2 effect in the two modalities. An absence of an auditory N2 effect under this condition would strongly argue against the response-inhibition hypothesis. Additionally we measured the lateralized readiness potential (LRP; Coles and Gratton 1986; Smid et al. 1987) for both response conditions, since a possible inhibition might also influence central response tendencies. Most of the cited Go/Nogo studies report only a single-peaked “P300” which exhibits topographic differences across response types: on Nogo trials the P300 has a more central maximum than on Go trials, on which it has a parietal maximum (Karlin et al. 1969; Hillyard et al. 1976; Simson et al. 1977; Pfefferbaum et al. 1985; Pfefferbaum and Ford 1988; Kok 1986; Jodo and Inoue 1990). The central enhancement of the P300 on Nogo trials has been explained as being due to an inhibition mechanism (Karlin et al. 1969) or to the absence of a negative motor potential (Kok 1986). However, Pfefferbaum et al. (1985) showed that the effect was also present in Count/Nocount paradigms, which suggests that overt motor inhibition or motor potentials cannot fully account for the topography shift. More recent hypotheses assume that the P300 is related to different processes or arises from different

0168-5597/95/$09.50 0 1995 Elsevier Science Ireland Ltd. All rights reserved SSDI 0013-4694(94)00182-K

EEP 93691

M. Falkenstein et al. / Electroencephalography

and clinical Neurophysiology 96 (1995) 36-43

37

by different generators of the “P300” in Go and Nogo ERPs. In choice tasks we found that the P-CR, but not the P-SR, was delayed after auditory stimuli, when the stimulus modality varied at random within a block (crossmodal divided attention paradigm (DA); Hohnsbein et al. 1991), which led to a dissociation of the subcomponents (Fig. 1~). Therefore we also used this paradigm in the present study. This implies a presentation of visual and auditory stimuli in close temporal proximity, which is also important for the N2 issue.

Method Subjects

Nine right-handed subjects (5 males and 4 females) with a mean age of 22.6 years (range: 18-33) were paid for participating in the study. The subjects had normal visual acuity and no known auditory deficits.

P-SR

(DA1 Stimuli

200

300

100

500

600

ms

Fig. 1. Model of the P300 in choice tasks (heavy lines) as a result of the overlap of two subcomponents, P-SR (thin lines) and P-CR (dashed lines). In the ERPs of simple reaction only the P-SR is present (thin lines). The overlap is usually large, especially after visual stimuli (a), resulting in a large P300 complex. The P-SR peaks earlier after auditory stimuli, which leads to a smaller overlap and a partial dissociation of the subcomponents (b). In certain conditions (such as crossmodal divided attention (DA), as used in the present study) the P-CR is delayed after auditory stimuli, and the subcomponents are even better dissociated (c).

generators in Go and Nogo trials (Pfefferbaum and Ford 1988; Jodo and Kayama 1990; Eimer 1993). In our previous work (Hohnsbein et al. 1991; Falkenstein et al. 1993, 1994a) we showed that the “P300” in choice tasks consists of two subcomponents (Fig. 1). The first (called P-SR, i.e., Positivity in Simple Reactions, because it is seen uncontaminated on simple reactions) has a central maximum. The second (called P-CR, i.e., Positivity in Choice Reactions) has a parieta1 maximum and is absent in simple reactions. The P-SR peaks earlier and has a more frontal topography after auditory than after visual stimuli. Usually both subcomponents overlap strongly and, particularly after visual stimuli (Fig. la), merge to a single large positive compound, the “P300.” In the present study we analysed the component structure of the P300 complex in Go and Nogo trials. We hypothesized that, if both components were also present in the ERPs of Go and Nogo trials, the topographic differences of the “P300” could be explained by different patterns of overlap of the subcomponents on Go and Nogo trials, rather than

The stimuli were single letters (“F” and “J”> that were presented visually or acoustically. The visual letters (0.5” high at a viewing distance of 57 cm) were presented for 200 msec just below a fixation spot in the middle of a visual display unit (VDU). The auditory letters were spoken (German pronounciation) and digitized and presented diotically by Sennheiser HD 425 headphones. (The intensity of the “F” was 55 dB SPL, the J was adjusted to seem equally loud to the subjects.) The auditory presentation of a letter took 300 x msec. The luminance of the visual stimuli was so adjusted (to 50 cd/m*) that their brightness was judged to be equivalent to the loudness of the auditory stimuli (Stevens 1975). Procedure

The subjects were trained in two sessions before the two main sessions. In each block visual and auditory “F’s and “J’s were presented in random order, with equal probability for letters and modalities. The IS1 was randomized with a mean of 1500 msec and a range from 1050 to 1950 msec (equal distribution). Two hundred stimuli (50 “J’s and 50 “F”s for each modality) were presented in each of 2 blocks. In block 1 the subjects had to respond to each “F” stimulus by pressing the (left-hand) “F” key with the left index finger, and to withhold the response to each “J” stimulus, regardless of modality. In block 2, the “J” was designated the target (which had to be responded to by the (right-hand) “J” key with the right index finger), and the “F” the non-target. The same two blocks were run in both experimental sessions. (In 3 further blocks with the same stimulus set, the subjects had to perform 2-way choice responses and simple responses with the left or the right index finger to each of the letters.

38

M. Falkenstein et al. / Electroencephalography and clinical Neurophysiology 96 (1995) 36-43

These data have already been published elsewhere (Falkenstein et al. 19931.) A moderate time pressure was imposed by an acoustic signal (a 300 msec, 1 kHz, 60 dB SPL tone burst), which was presented at 900 msec after letter onset if the subject failed to respond to the letter within 500 msec. (The feedback was delayed to prevent its ERP interfering with the stimulusrelated ERP.) The subjects were instructed to avoid the feedback tone at the risk of committing errors. (The 500 msec RT deadline was chosen since all subjects were able to keep within it in nearly all trials after the two practice sessions.) The sequence of blocks was randomized and hence different for each subject and in each session. The subjects had to keep their eyes open and directed towards the VDU in all conditions. EEG recording

During the two experimental sessions the EEG was recorded with Ag/AgCl electrodes from Fz, Cz, C3, C4, Pz and Oz against linked mastoids. The forehead was grounded. The vertical EOG was recorded from the upper and lower canthus of the left eye. The recorded activity was amplified (EEG: gain = 100,000; EOG: gain = 20,000) with a l-pole high-pass (3 dB: 0.03 Hz) and a lo-pole Butterworth low-pass as antialiasing filter (3 dB: 60 Hz). EEG and EOG were digitized with 200 samples/set and stored on a hard disk of an IBM-AT computer for off-line analysis. Artifact rejection and averaging

EEG epochs of 950 msec length were averaged using stimulus onset as trigger. The pre-stimulus epoch was 50 msec long, the post-stimulus epoch 900 msec. Correct (hit) and incorrect (false alarm) trials were averaged separately. For hit and false alarm trials the key press was also used as trigger. Epochs with samples exceeding the preamplifier range in any channel (including EOG) were rejected for all channels. By this procedure large blink artifacts were avoided. To compensate for the influence of small blinks, which did not cause overflow, propagation factors between the EOG and the EEG channels were computed (Verleger et al. ‘1982). The EOG was weighted by the propagation factors and subtracted from the EEG channels.

msec, which was referred to as P400. In the (rare) case of two peaks in this window, the first peak was taken as the P400. (A peak was defined as a local maximum with a minimum amplitude difference to the preceding and following trough of 0.5 pV.) The auditory ERPs also revealed a peak in the range 300-500 msec, called P400, and a second peak in the latency range 400-700 msec, called P507. In the (rare) case of two peaks in this window, the second peak was taken as the P507. The peak latency and amplitude of the N300 and the P400 were subjected to 4-way repeated measures ANOVAs (BMDP4V; Dixon 1990). The factors were electrode (E) (levels: Fz, Cz, Pz, Oz), condition (C) (Go, Nogo), stimulus modality (M) (visual, auditory), and letter (L) (J, F). The peak latency and amplitude values of the early components (visual: P170, N200; auditory: N140, P230) as well as of the auditory P507 were subjected to 3-way ANOVAs with the factors E, C and L. To test whether possible interactions of electrode with other factors reflect true topography differences across factors, these data were normalized (McCarthy and Wood 1985), and new ANOVAs were computed. In all ANOVAs the degrees of freedom were corrected by using the Greenhouse-Geisser procedure. In the error trials a negativity with Cz maximum (called error negativity, N,) was present. Its latency and amplitude were evaluated in the response-triggered averages (cf., Falkenstein et al. 1991) and subjected to ANOVAs with factors M and L. The lateralized readiness potential (LRP) was computed according to Coles (1988). For both hands stimulus-locked ERPs from C3 and C4 were subtracted (right-hand task, when “J” was the target: C3 minus C4; left-hand task, when “F” was the target: C4 minus C3) and averaged. The mean amplitude of the LRP in the time window beginning 100 msec before and ending at the mean Go reaction time was computed for all conditions and subjected to an ANOVA with the factors M, C and L. Mean reaction times were computed separately for correct (hits) and incorrect (false alarm) responses and subjected to ANOVAs with the factors M and L. The error rates were evaluated with the same design.

Data analysis

The averaged ERP data were digitally low-pass filtered with a cut-off at 17 Hz and pooled across sessions. The amplitudes were referred to electrical zero, since they were not correlated with the baseline (the mean amplitude of the 50 msec pre-stimulus period; cf., Riisler 1979). In all conditions the stimulus-triggered visual ERPs contained the early components P170, N200 and N300, and the auditory ERPs N140, P230 and N300. The visual ERPs showed a large positive complex peaking in the latency range 300-500

Results Petformance

The reaction times on correct Go trials (hits) were nearly the same after visual (354 msec) and auditory (362 msec) stimuli, and for both letters and hands (F/left hand: 359 msec; J/right hand: 358 msec). No significant main effect or interaction was found. The RTs for errors (false alarms) were shorter than for hits, and more so for visual (293 msec) than for auditory

M. Falkenstein et al. / Electroencephalography

stimuli (342 msec), which was reflected in a significant modality main effect (F (1, 8) = 11.0, P = 0.0127). The error rates were 10.1% for visual stimuli, and 8.7% for auditory stimuli (F (1, 8) = 4.4, P = 0.0701).

P230

Fz

Fig. 2 shows the grand means of the ERPs and the LRP, pooled across letters (and hence response hands). After visual stimuli the components P170, N200 (with parieto-occipital maximum) and N300 (with frontal maximum), and after auditory stimuli the components N140, P230 and N300 are visible. After visual stimuli these waves are followed by a large positive complex peaking at about 400 msec (“P400”). For Go trials the complex is broader, particularly at Cz. After auditory stimuli there were clearly two components, the P400 and a later positivity peaking at about 500 msec (“P507”).

PLO0

cz

Table II presents the means and S.D.s for the latency and amplitude of the P400 and the auditory P507, and Fig. 4 shows the scalp distribution of the amplitude data for the two components. (a) P400.. There was an electrode main effect (F (2.0, 15.8) = 6.8, P = 0.0074) on P400 amplitude, but no

clear maximum of the P400, particularly after auditory stimuli. For auditory stimuli the topography of the P400 was more frontal than for visual stimuli (Fig. 41, as reflected in a E X M interaction (F (2.3, 18.4) = 15.76, P = 0.0001) that remained significant after scaling for M. The latency of the P400 was clearly longer after visual (432 msec> than after auditory (370 msec) stimuli (F (1, 8) = 146.5, P < 0.0001). Also, its amplitude was

N300

*

Pz N-200

Early components No effects were found for the visual P170 and N200,

Late components

auditory

visual

ERP data

and the auditory N140 and P230, with the exception of electrode main effects for the auditory components, which reflect their central (N140) and fronto-central (P230) maximum. The N300 (Table I and Fig. 3) was more negative for visual than for auditory stimuli (I; (1, 8) = 8.3, P = 0.0205). Also, it was larger for Nogo than for Go trials (F (1, 8) = 6.0, P = 0.0398). However, as revealed in an M X C interaction (F (1, 8) = 8.59, P = 0.01901, this was true only for visual stimuli (F (1, 8) = 9.3, P = O.OlSS>,while for auditory stimuli there was no N300 amplitude difference between Go and Nogo trials (F (1, 8) = 1.5, P = 0.2546; see also Fig. 2). The visual Go/Nogo effect was largest at Fz (3.8 WV). The LRPs are also shown in Fig. 2. Their interindividual variance was so large, that only very small deflections were seen preceding the reaction in the grand means, and hence no significant effect was found.

39

and clinical Neurophysiology 96 (1995) 36-43

~

--ZZ

LRP -

EOG-

-

-

4,

S

,

,

,

4,

,

-Go

,

,

,

,

S 200 LOO600 600 ms

200 LOO600 800 ms

-

Nogo

Fig. 2. Grand means across 9 subjects of the ERPs for Go (heavy lines) and Nogo trials (thin lines) after visual (left panel) and auditory stimuli (right panel). LRP: lateralized readiness potential (C3 minus C4 for left-hand, C4 minus C3 for right-hand responses). EOG: vertical EOG. The associated Go reaction times are shown by vertical bars in the LRP panel. S = stimulus onset. Positivity is plotted upwards.

larger after visual (11.6 PV) than after auditory (6.7 PV) stimuli (F (1, 8) = 29.1, P = 0.0007). For both modalities the P400 was smaller on Go than on Nogo trials at Fz and Cz, while it was larger on Go than on Nogo trials at Pz and Oz. This was reflected in an E X C interaction (F (1.2, 9.9) = 11.7, P = 0.0050) which remained significant after scaling for TABLE I Means and standard deviations (in parentheses) for amplitude (in /JV) and latency (in msec) of the auditory and visual N300 for Go and Nogo trials. Auditory

Fz Cz Pz Oz

Go Nogo Go Nogo Go Nogo Go Nogo

Visual

Amplitude

Latencv

Arnolitude

Latencv

0.5 (3.5) 0.2 (4.8) 0.3 (5.3) 0.4 (5.9) 1.9 (5.0) 0.7 (4.9) 0.2 (2.6) 0.9 (3.6)

301 (26) 291 (18) 285 (21) 287 (23) 278 (19) 280 (29) 270 (28) 280 (32)

-

302 306 288 292 285 288 279 284

-

-

0.9 (3.2) 4.7 (4.9) 3.0 (6.1) 5.6 (6.0) 0.0 (4.8) - 1.8 (4.6) 0.3 (4.0) - 1.3 (3.7)

(34) (27) (24) (22) (21) (19) (22) (23)

40

M. Falkenstein et al. / Electroencephalography and clinical Neurophysiology 96 (1995) 36-43

;

I

-___

-

P400

GO NOGO

N300

y$-pry

----

GO NOGO

P507

I

F

P

C

0

0

I

I

1

Fig. 3. Amplitude distribution of the N300 component in Go (solid lines) and Nogo trials (dashed lines) for auditory (filled circles) and visual stimuli (open circles). The data are from Table I.

F Fig. 4. Amplitude complex (P400 and lines) for auditory

C P 0 F C P 0 distribution of the subcomponents of the P300 P507) in Go (solid lines) and Nogo trials (dashed (filled circles) and visual stimuli (open circles). The data are from Table II.

C. Both effects were quite similar for both modalities, but about twice as large after visual than after auditory stimuli (cf., Table I), as reflected in a E X C x M interaction (I; (2.1, 16.6) = 3.76, P = 0.0438), which vanished after scaling for M. The simple effects at the single electrodes showed that for visual stimuli there were significant Go/Nogo effects at Cz, Pz and Oz, whereas for auditory stimuli there was a significant Go/Nogo effect only at Oz, and a tendency at Cz.

tion (F (1.3, 10.1) = 4.9, P = 0.0447). This interaction vanished after scaling for C. The P507 peaked later on Go (526 msec) than on Nogo trials (489 msec) (F (1, 8) = 11.59, P = 0.0093). In 7 of the 9 subjects the P507 was also seen as a separate peak on Go trials after visual stimuli. For these subjects and this condition its mean amplitude (averaged across leads) was 8.6 PV and its mean latency 516 msec.

(b) P507.. The (auditory) P507 had a parietal maximum, as reflected in an electrode main effect (F (2.0, 16.3) = 12.8, P = 0.0004). It was larger for Go (6.4 pV> than for Nogo (4.3 PV) trials (F (1, 8) = 8.91, P = 0.0175). This amplitude difference was strongest for Pz and Oz (cf., Table II), as revealed in a E X C interac-

ERPs of error (false alarm) trials

Fig. 5 presents the data for the false alarms and the correct Go and Nogo trials at Cz. In the stimulus-triggered averages (upper panels) a strong negative deflection (called NJ peaking after the incorrect reaction is seen for both modalities, which is followed by a later

TABLE II Means and standard deviations (in parentheses) for amplitude (in PV) and latency (in msec) of the auditory and visual P400 and of the auditory P507 (the visual P507 was only detectable for a subset of subjects in the Go condition). P400

Auditory

FZ CZ PZ 02

Visual

Fz CZ Pz 02

P507

Amplitude

Latency

Amplitude

Latency

Go Nogo Go Nogo Go Nogo Go Nogo

7.3 (4.5) 7.8 (4.0) 6.6 (4.3) 7.9 (4.0) 7.7 (4.4) 6.8 (3.4) 5.5 (3.1) 3.9 (2.3)

363 (19) 370 (19) 367 (24) 374 (25) 374 (291 369 (21) 379 (28) 367 (21)

3.0 (3.0) 2.6 (2.8) 7.1 (3.6) 5.4 (3.1) 9.0 (3.7) 5.9 (2.8) 6.4 (2.6) 3.5 (2.3)

521 (37) 486 (29) 536 (41) 490 (33) 525 (34) 484 (32) 521 (43) 494 (41)

Go Nogo Go Nogo Go Nogo Go Nogo

9.0 (3.1) 10.6 (3.6) 12.8 (4.5) 15.2 (5.1) 14.2 (5.1) 12.2 (4.7) 10.9 (4.8) 8.1 (4.2)

416 (25) 433 (29) 431 (27) 437 (31) 436 (29) 439 (31) 433 (52) 433 (29)

M. Falkenstein et al. / Electroencephalography

,\\ G

STA

4,

S

s 200 LOO600 600 ms

,Ne ,

4,

,

,

200 LOO600 800 ms

auditory

The “N2” effect ,

.

,

RTA

I’I

-200

-

4,

1'1'1

I’,

R 200 LOO600 ms

Nogo

-Go

41

in general the errors were fast guesses. However, after auditory stimuli the RT shortening on error trials was much less than after visual stimuli. Hence after auditory stimuli some errors appear to have been committed because of the above mentioned slowing of response selection.

cz

visual

and clinical Neurophysiology 96 (1995) 36-43

-200

4,

1’1’1

R 200 LOO600 ms

-

FA

Fig. 5. Grand means at Cz of the ERPs in Nogo trials (thin lines), Go trials (medium lines) and error (false alarm, FA) trials (heavy lines) after visual (left panel) and auditory stimuli (right panel). Upper panels: stimulus-triggered averages (STA); lower panels: responsetriggered averages (RTA). S = stimulus onset; R = key press. The associated RTs are given as vertical thin (hit) and heavy (false alarm) bars in the STA panels.

positive complex. The N, could even be detected in ERPs with few sweeps. For visual data the N, peaked at about 340 msec, i.e., 47 msec later than the mean RT of the false alarm, whereas for auditory stimuli the N, peaked at about 440 msec, i.e., 98 msec after the mean false alarm RT. This asymmetry was confirmed and evaluated in the response-triggered averages (lower panels), which show that the N, peaked later with respect to the overt response (R) after auditory (+ 101 msec> than after visual stimuli (+ 70 msec> (F (1, 8) = 20.14, P = 0.0020). The N, amplitude was large (about - 13 PV in the RTAs) and constant across conditions.

Discussion Performance

For correct trials, the Go RTs were not different for the modalities. Hence the frequent finding of shorter auditory than visual RTs was not seen in our study. This can be attributed to our divided attention paradigm,, which slows down late cognitive processes (most likely response selection) after auditory stimuli (Hohnsbein et al. 1991). The error (false alarm) RTs were shorter than the correct RTs, which suggests that

After visual stimuli a fronto-central negative enhancement at about 300 msec was found on Nogo compared to Go trials, which agrees with the literature. However, a comparable N300 enhancement was absent after auditory stimuli. This agrees with the data of Karlin et al. (19691, Hillyard et al. (1976) and Banquet et al. (1981) and strongly argues against the hypothesis that the N2 effect reflects a response inhibition on Nogo trials. In our study response inhibition (if there is any) appears to be even more effective after auditory stimuli, as reflected in the tendency toward fewer false alarms after auditory than after visual stimuli. Eimer (1993) found a smaller N2 effect when his (visual) stimuli were less attended. Since we have shown earlier that in our divided attention condition attention appears to be biased towards the visual channel (Hohnsbein et al. 1991), our auditory stimuli may be less closely attended to, which may have reduced the auditory N2 effect. However, in a recent control experiment (Falkenstein et al. 1994b) we have shown that even when only the auditory stimuli were attended to, the auditory N2 effect was absent. In sum, our results argue against an inhibition hypothesis for the N2 effect. Late positive complex

The P300 complex in the Go/ Nogo ERPs was found as a single P400 after visual stimuli and was dissociated in two positive peaks (P400 and P507) after auditory stimuli. This extends our earlier findings, showing that two P300 subcomponents (P-SR and P-CR) are not only present in choice reaction tasks (Fig. l), but also in Go/Nogo tasks. Our Go/Nogo data are fully compatible with the model we developed for choice tasks: the P-SR peaks later, and the P-CR is larger, after visual stimuli, which leads to a larger compound (the P400) in the visual than in the auditory modality. The amplitude of the P-SR itself, as seen in the simple reaction ERPs of the same subjects (Falkenstein et al. 1993), was not different across modalities (cf. also Fig. 1). Its latency suggests that the P400 is what is usually named “P3” or “P300” in visual paradigms (e.g., Pfefferbaum et al. 198.5; Eimer 1993). The fact that the P400 contains two subcomponents (P-SR and P-CR) explains also the flat and unusual topography of the auditory P400 (cf. Table I). The P400 shows topographic differences between Go and Nogo trials: for Go trials it was larger at

42

M. Falkenstein et al. / Electroencephalography

parieto-occipital leads, and smaller at fronto-central leads, than for Nogo trials. The fact that the overlap of P-SR and P-CR was only marginal after auditory stimuli, together with the finding that both effects were very small after auditory stimuli (Fig. 4), suggests that the topographic differences are caused by the overlap, rather than by intrinsic topography changes of the subcomponents. In fact, the smallness of the topography effects on the auditory P400 (which consists mainly of the P-SRI suggests that the topography of the P-SR itself is unaffected by response type. The auditory P507 (which consists mainly of the P-CR) showed also no Go/ Nogo topography difference. However, the auditory P507 was larger for Go than for Nogo trials, and mainly so at Pz and Oz (Figs. 2 and 4). This appears to be also valid for the visual P507 (cf. also Fig. 21, which was only detectable as separate peak in Go trials. Hence the parieto-occipital enhancement of the P400 on Go vs. Nogo trials can be explained by the influence of the P-CR, which is larger on Go than on Nogo trials. Contrary to the suggestion of Kok (1986), the central enhancement of the P400 on Nogo vs. Go trials cannot be due to the presence of a negative movement-related potential (MRP) on Go as opposed to Nogo trials. The negative MRP is known to persist only until shortly after the key press (e.g., Banquet et al. 1991). Since the P400 peaks shortly after the Go response to auditory stimuli, and about 70 msec after the Go response to visual stimuli, the influence of a negative MRP on the P400 should be larger after auditory than after visual stimuli, while we observed just the opposite. Rather, we propose that the Cz effect is also caused by the overlap of P-SR and P-CR. First, the Cz effect was only tendentiously present after auditory stimuli, where the overlap is also small. Second, the P507 peaked earlier in Nogo than in Go trials, which should make the overlap stronger and hence increase the P400 amplitude. At Pz the P507 latency effect is cancelled (auditory stimuli) or overcompensated (visual stimuli) by the amplitude effect. Our finding that the P507 had a shorter latency on Nogo than on Go trials, which is the opposite of what is usually observed (e.g., Pfefferbaum et al. 19851, can be explained by the extended training of our subjects, which appears to shorten the latency of the Nogo“P300,” as shown by Jodo and Inoue (1990). Since we recently found that the P-CR in choice tasks depends on the complexity of the stimulus-response mapping (cognitive response selection, RS) (Falkenstein et al. 1994a) we proposed that the P-CR is related to RS. Hence our present result of a smaller and earlier P-CR on Nogo trials suggests that for trained subjects the cognitive choice not to respond is faster (and needs less resources) than the choice to press the button. In summary, we explain the apparent topographic

and clinical Neurophysiology 96 (1995) 36-43

differences of the P400 between Go and Nogo trials as due to the overlap of two subcomponents (P-SR and P-CR), which are present in Go as well as in Nogo trials with constant topographies. This model can explain the often reported Go/Nogo effects on the topography of the visual P300, which is a compound of P-SR and P-CR, as also seen for our visual data. Hence our results also offer evidence against the hypothesis of different P300 generators in Go and Nogo trials. Error trials

In error trials, similar phenomena (a central negativity, called N,, and a subsequent positivity) are seen, as we have described for 2-way choice tasks (Falkenstein et al. 1991). In those studies we proposed that the N, reflects an error-detection mechanism, in the sense of a mismatch between the overt response with the outcome of the stimulus-response mapping, i.e., the cognitive response selection or decision stage, which is also conducted on error trials. The presence of the N, on false alarms shows that the correct cognitive choice is also carried out on false alarm trials. Since the false alarms were based on premature guesses, the choice stage is the later of the two processes to be matched and hence determines N, latency. The delay of the N, after auditory stimuli confirms and extends our earlier finding that the choice stage is slowed down after auditory stimuli under divided attention (Hohnsbein et al. 1991). Conclusions

The absence of an enhancement of the N2 component in Nogo vs. Go trials for auditory stimuli argues against the common hypothesis that this effect reflects a response inhibition mechanism. The frequently observed (apparent) topographic differences of the visual P300 complex between Go and Nogo trials were replicated and explained by the differential overlap of two P300 subcomponents. Hence we replace the hypothesis of different “P300” generators in Go and Nogo trials by a model that assumes two P300 generators in both Go and Nogo trials. The first generator is stable, whereas the activity of the second generator is larger on Go than on Nogo responses. We would like to thank Prof. CR. Cavonius, Dortmund, for valuable comments on earlier versions of the manuscript. We are indebted to Ludger Blanke and Christiane Westedt for their committed technical assistance. This research was supported by the DFG (Grant 438 17-157-92).

References Banquet, J.P., Renault, B. and Leskre, stimulus probability on evoked potentials. 203-214.

N. Effect of task and Biol. Psychol., 1981, 13:

M. Falkenstein et al. / Electroencephalography

Coles, M.G.H. Modern mind-brain reading: psychophysiology, physiology, and cognition. Psychophysiology, 1988, 26: 251-269. Coles, M.G.H. and Gratton, G. Cognitive psychophysiology and the study of states and processes. In: G.R.J. Hockey, A.W.K. Gaillard and M.G.H. Coles (Eds.), Energetics and Human Information Processing. Nijhoff, Dordrecht, 1986: 409-424. Dixon, W.J. (Ed.). BMDP Statistical Software. University of California Press, Berkeley, CA, 1990. Eimer, M. Effects of attention and stimulus probability on ERPs in a go/nogo task. Biol. Psychol., 1993, 35: 123-138. Falkenstein, M., Hohnsbein, J., Hoormann, J. and Blanke, L. Effects of crossmodal divided attention on late ERP components. II. Error processing in choice reaction tasks. Electroenceph. clin. Neurophysiol., 1991, 78: 447-455. Falkenstein, M., Hohnsbein, J. and Hoormann, J. Late visual and auditory ERP components and choice reaction time. Biol. Psychol., 1993, 35: 201-224. Falkenstein, M., Hohnsbein, J. and Hoormann, J. Effects of choice complexity on different subcomponents of the late positive complex of the event-related potential. Electroenceph. clin. Neurophysiol., 1994a, 92: 148-160. Falkenstein, M., Hohnsbein, J. and Hoormann, J. The “N2-effect” in the ERP of audio-visual Go/Nogo tasks (abstract). 1994b, in press. Hillyard, S.A., Courchesne, E., Krausz, H.T. and Picton, T.W. Scalp topography of the P3 wave in different auditory decision tasks. In: W.C. McCallum and J.R. Knott (Eds.), The Responsive Brain. Wright, Bristol, 1976: 81-87. Hohnsbein, J., Falkenstein, M. and Hoormann, J. Effects of crossmodal divided attention on late ERP components. I. Simple and choice reaction tasks. Electroenceph. clin. Neurophysiol., 1991, 78: 438-446. Jodo, E. and Inoue, K. Effects of practice on the P300 in a Go/Nogo task. Electroenceph. clin. Neurophysiol., 1990, 76: 249-257.

and clinical Neurophysiology 96 (1995) 36-43

43

Jodo, E. and Kayama, Y. Relation of a negative ERP component to response inhibition in a Go/Nogo task. Electroenceph. clin. Neurophysiol., 1992, 82: 477-482. Karlin, L., Martz, M.J. and Mordkoff, A.M. Motor performance and sensory-evoked potentials. Electroenceph. clin. Neurophysiol., 1969, 28: 307-313. Kok, A. Effects of degradation of visual stimuli on components of the event-related potentials (ERP) in Go/Nogo reaction tasks. Biol. Psychol., 1986, 23: 21-38. McCarthy, G. and Wood, C.C. Scalp distributions of event-related potentials: an ambiguity associated with analysis of variance models. Electroenceph. clin. Neurophysiol., 1985, 62: 203-208. Pfefferbaum, A. and Ford, J.M. ERPs to stimuli requiring response production and inhibition: effects of age, probability and visual noise. Electroenceph. clin. Neurophysiol., 1988, 71: 55-63. Pfefferbaum, A., Ford, J.M., Weller, B.J. and Kopell, B.S. ERPs to response production and inhibition. Electroenceph. clin. Neurophysiol., 1985, 60: 423-434. Rosier, F. Zur psychologischen Bedeutung evozierter Hirnrindenpotentiale: methodische Probleme. Psychol. Beitr., 1979, 21: 1-21. Simson, R., Vaughan, H.G. and Ritter, W. The scalp topography of potentials in auditory and visual Go/NoGo tasks. Electroenceph. clin. Neurophysiol., 1977, 43: 864-875. Smid, H.G.O.M., Mulder, G. and Mulder, L.J.M. The continuous flow model revisited: perceptual and central motor aspects. In: R. Johnson, Jr., J.W. Rohrbaugh and R. Parasuraman (Eds.), Current Trends in Event-Related Potential Research (EEG Suppl. 40). Elsevier, Amsterdam, 1987: 270-278. Stevens, S.S. Psychophysics. Wiley, New York, 1975. Verleger, R., Gasser, T. and Mocks, J. Correction of EOG artifacts in event-related potentials of the EEG: aspects of reliability and validity. Psychophysiology, 1982, 19: 472-480.