Psychiatry/Psychology

Clinical EEG and Neuroscience

2024, Vol. 55(3) 317–328
Proactive and Reactive Inhibitory © EEG and Clinical Neuroscience
Society (ECNS) 2022
Control Strategies: Exploring the Article reuse guidelines:
sagepub.com/journals-permissions

Impact of Interindividual Variables DOI: 10.1177/15500594221145905

journals.sagepub.com/home/eeg

on an ERP Continuous Performance

Task (AX-CPT)

Elisa Schröder1,*, Anais Ingels1,*, Alexandru Dumitrescu2,

Charles Kornreich1, and Salvatore Campanella1

Abstract
According to the Dual Mechanisms of Control (DMC) framework, cognitive control can be divided into two strategies: proactive
cognitive control, which relies mainly on the active maintenance of contextual information relevant to the ongoing task; and
reactive cognitive control, which is a form of transient control triggered by an external cue. Although cognitive control has
been studied extensively, little is known about the specificities of inhibition within the framework of the DMC model and
the influence of interindividual variables on inhibitory control.
Thanks to an inhibitory version of the continuous performance task (CPT), we studied behavioral performances and Event-
Related Potentials (ERPs) related to proactive and reactive inhibition, and their links to psychological profile and cognitive per-
formances. One hundred and five young adults underwent the task, along with a short clinical and cognitive evaluation.
We were able to observe ERPs related to proactive (cue-N1, cue-N2, cue-P3, and the contingent negative variation) and reac-
tive inhibitory control (target-N2 and target-P3). Our results showed that proactive strategies appeared to be linked with impul-
sivity, working memory abilities, dominant response inhibition, gender, and the consumption pattern of nicotine. Reactive
strategies appeared to be linked with attentional and working memories abilities.
Overall, the inhibitory AX-CPT allowed a specific investigation of cognitive control within the framework of the DMC based
on behavioral and ERP variables. This provided us an opportunity to investigate the principal ERP components related to pro-
active and reactive inhibitory control strategies as well as to link them with specific clinical and cognitive variables.

Keywords
ERPs, CPT-AX task, reactive control, proactive control, inhibition

Received January 27, 2022; revised November 25, 2022; accepted November 29, 2022.

Introduction On their own, behavioral analyses have at times failed to

highlight any variation in cognitive control.7,8 In this
Cognitive control has been defined as the ability of the cogni-
tive system to adjust, adapt, and configure itself in order to opti-
1
mally perform an ongoing task.1 Recently Braver and Laboratory of Medical Psychology and Addictology, CHU Brugmann, ULB
Neuroscience Institute (UNI), Université Libre de Bruxelles (ULB), Brussels,
colleagues proposed the dual mechanisms of control (DMC) Belgium
model2,3 that considers cognitive control as a balance 2
Laboratoire de Cartographie Fonctionnelle du Cerveau, Hôpital Erasme, ULB
between proactive and reactive cognitive control mechanisms Neuroscience Institute (UNI), Université Libre de Bruxelles (ULB), Brussels,
(Figure 1). Belgium
Several experimental designs have been described for study- *
ES and AI are contributed equally to this work
ing proactive versus reactive cognitive control strategies, such
Corresponding Author:
as the Stop-Signal Task, the Go/NoGo task,4 and task-switch- Salvatore Campanella, The Belgian Fund for Scientific Research (F.N.R.S.),
ing paradigms.5,6 But, according to Braver and colleagues, a CHU Brugmann, Psychiatry Secretary, 4, Place Vangehuchten - B-1020
continuous performance task, such as the AX version Brussels, Belgium.
(AX-CPT), is the most suitable for studying the differential Emails: [email protected]; [email protected]
use of proactive versus reactive cognitive control.2,3 Full-color figures are available online at journals.sagepub.com/home/eeg
318 Clinical EEG and Neuroscience 55(3)

Figure 1. On the one hand, proactive cognitive control corresponds to the active maintenance of contextual information in order to bias
further processing toward task-relevant information. It is a process of early top-down, goal-directed selection that influences further
perception, attentional orientation, response selection, and action sequences. Proactive control takes place as soon as the intention of action is
conceived, and it has to be actively maintained until the goal is reached. On the other hand, reactive cognitive control refers to a transient
control strategy engaged after the occurrence of the control-demanding event. Contrary to proactive control, it is a form of late bottom-up
re-selection of task goals that needs to be activated by a relevant trigger.3 Both strategies have their advantages and disadvantages. The
proactive strategy is extremely resistant to proactive interference from tasks that are no longer pertinent, although it is also metabolically and
cognitively demanding, which implies that it cannot be exerted for more than a few minutes.3 Moreover, proactive cognitive control depends
on the presence and reliability of contextual cues. Proactive control is less sensitive to changes in the environment, which is both an advantage
and a disadvantage. Indeed, it makes this strategy robust, but also less flexible if any goal-irrelevant information, such as an unpredictable threat,
appears in the environment.2 Reactive cognitive control is a form of minimalist control and is therefore less resistant to interference than
proactive control. It hence allows for habits, procedures, or skills to be learned, developed, and encoded, and to progress toward
automatization. Yet, it depends on the sufficiency of the processing driven by the stimulus.

regard, imaging techniques such as event-related potentials techniques such as ERPs are of great value for gaining an
(ERPs) have displayed a higher sensitivity to variation in cog- understanding of these otherwise largely intangible pro-
nitive control.7,9,10 More specifically, several ERP compo- cesses.18 Altogether, ERPs provide numerous investigative
nents have been associated with proactive control: the advantages in the field of inhibitory control.
cue-N1, associated with early stimulus detection and At the neurofunctional level, functional magnetic resonance
top-down attentional orientation11; the cue-N2, associated imaging (fMRI) studies have suggested that active maintenance
with goal maintenance12; the P3b, related to context updat- of task goals related to proactive control is mainly associated
ing,13,14 and the contingent negative variation (CNV), associ- with sustained activity in the lateral prefrontal cortex (PFC),
ated with response expectation and preparation.13,14 These while reactive control appears to be related to a transient activ-
ERP components have already been observed through use ity in the lateral PFC mediated by conflict-monitoring regions
of an AX-CPT task during the processing of the cue12–14 or such as the anterior cingular cortex (ACC).19
a modified Stop-Signal Task.11,15 The ERP components Overall, the impacts of different variables such as working
related to reactive cognitive control are the target- (or memory, attentional capacities, impulsivity, or mood state on
NoGo-) N2, related to response conflict monitoring,13–16 reactive versus proactive control strategies have already been
and the target- (or NoGo-) P3, related to motor inhibition investigated, sometimes with some inconsistencies. Working
per se.14,15,17 These ERP components have been observed memory abilities can greatly influence cognitive control,20
after the onset of the target in the AX-CPT,13,14 or following and more specifically proactive control, while reactive control
the apparition of a NoGo stimulus in a cued Go/NoGo task17 relies on transient attentional capture and conflict detection
or a Stop-Signal stimulus in a Stop-Signal Task.15 Moreover, mechanisms that will signal the need for reactive control activa-
when the focus is on inhibitory control processes, and as inhi- tion.3 Attentional abilities have also been strongly associated
bition is a covert process that—when successful—produces with proactive control.11,21 Impulsivity on the other hand
little or no overt measurable behavior, cognitive neuroscience appears to have an unclear impact on proactive versus reactive
Schröder et al 319

Table 1. Clinical Characteristics.

Demographic data Age in years (mean ± standard deviation) 23.7 ± 4.75

Gender (M/F) 48/56
Laterality (right-handed/left-handed) 93/11
Smokers (/104) 20
Consumption pattern AUDIT (mean ± standard deviation) 6.28 ± 5.37
DAST-10 (mean ± standard deviation) 0.45 ± 0.954
Fagerström (mean ± standard deviation) 0.52 ± 1.488
Affective and personality factors BDI-II (mean ± standard deviation) 7.2 ± 5.9
STAI-state (mean ± standard deviation) 45.68 ± 7.28
STAI-trait (mean ± standard deviation) 46.11 ± 9.68
UPPS total (mean ± standard deviation) 104.88 ± 13.65
UPPS emergency (mean ± standard deviation) 27.68 ± 5.46
UPPS premeditation (mean ± standard deviation) 21.37 ± 5.11
UPPS perseverance (mean ± standard deviation) 21.72 ± 5.13
UPPS sensation seeking (mean ± standard deviation) 34.08 ± 7.9
BIS (mean ± standard deviation) 19.91 ± 3.99
BAS drive (mean ± standard deviation) 9.36 ± 2.43
BAS fun seeking (mean ± standard deviation) 11.89 ± 2.02
BAS reward responsiveness (mean ± standard deviation) 16.49 ± 2.98
Cognitive evaluation Stroop interference-denomination time (mean ± standard deviation) 33.13 ± 15.4
Stroop interference-denomination errors (mean ± standard deviation) 2.5 ± 2
N0 correct hits (/40) (mean ± standard deviation) 39.89 ± 0.31
N2 correct hits (/40) (mean ± standard deviation) 33.92 ± 2.96
N3 correct hits (/40) (mean ± standard deviation) 26.77 ± 5.38
D2 GZ-F score (mean ± standard deviation) 497.51 ± 71.16
Digit span forward (mean ± standard deviation) 10.19 ± 2.02
Digit span backward (mean ± standard deviation) 6.81 ± 1.95

Abbreviations: AUDIT, alcohol use disorders identification test; BAS, behavioral approach system; BIS, behavioral inhibition system; BDI-II, beck depression
inventory; D2 GZ-F, total number of items processed minus total number of errors on the D2 test; DAST-10, drug abuse screening test; STAI, state and trait
anxiety inventory; UPPS, urgence premeditation perseverance sensation seeking impulsivity scale.

strategies, with high impulsivity sometimes being associated One participant had to be excluded as they had a history of psy-
with poor proactive,22 or reactive control.23 Regarding the chiatric illness. The characteristics of the group are presented in
affective variables, studies have suggested a specific impact Table 1.
of depression on proactive control,7,24 as well as anxiety
which appears to induce a shift from a proactive to a reactive
strategy.25,26 At the personality level, reward sensitivity Procedure
appears to be linked to a higher use of proactive control.27 The local ethics committee of Brugmann hospital provided its
Surprisingly few studies have taken an interest in the impact approval for this study. Participants were accommodated in a
of alcohol consumption28,29 and cigarette consumption on pro- testing room and they were immediately informed of the
active versus reactive control, considering how they affect nature and the duration of the task, as well as what was expected
inhibitory control. Consequently, and for the fresh perspective from them and what type of information would be collected.
that this newly adapted task offered, the main objective of the Informed written consent was obtained from each participant.
present study was to propose an exploratory evaluation of the A short questionnaire assessing their personal history, such
impact of these clinical, cognitive, and demographic variables as their level of education and their neurological and psychiatric
on proactive versus reactive inhibitory control strategies. history was administered first to ensure that the participants met
the inclusion criteria. The AX-CPT task was then administered.
A description of the task is presented in Figure 2. Participants
Materials and Methods had to complete multiple questionnaires in regard to issues
that could have a potential impact on cognitive control and inhi-
Participants bition abilities3,7,11,20–26,28–30: the Alcohol Use Disorder
One hundred and five young healthy participants participated in Identification Test (AUDIT)31; the Drug Abuse Screening
this study. They were recruited through social media. The inclu- Test (DAST-10)32; the Fagerström test for nicotine dependence
sion criteria were as follows: no history of neurological or psy- (Fagerström)33; the Beck Depression Inventory (BDI-II)34; the
chiatric illness, 20/20 vision (if need be, with corrective lenses). State-Trait Anxiety Inventory (STAI-A and STAI-B)35; the
320 Clinical EEG and Neuroscience 55(3)

Figure 2. The task was an adaption of the AX version of the continuous performance task (AX-CPT). Participants were asked to respond to
certain pairs of cue-target stimuli. They had to push on the button of a joystick with their right index as fast and accurately as they could every
time they saw the target “X” but if and only if the X was preceded by the cue “A”. Therefore, participants were confronted with four types of
pairs: valid cue—valid target (“AX” trials, valid pair), valid cue—invalid target (“AY” trials, invalid pair), invalid cue—valid target (“BX” trials,
invalid pair), and invalid cue—invalid target (“BY” trials, invalid pair). AX trials were the most frequent type of stimuli presented (68.83% of all
trials, 530 trials in total), while AY, BX, and BY trials were each presented 10.39% of the time (80 trials in total). The trials were presented in a
pseudorandom order according to the following rules: one block had to begin with at least one valid pair (AX trial); an invalid pair (AY, BX, or
BY) was never followed by an invalid pair; 1 to 4 valid pairs were presented between invalid pairs.
Each letter was presented on the center of a black background for 150 ms with a cue-target interstimulus interval of 1400 ms. The task was
divided into 11 blocks of 77 cue-target pairs each. The presentation of each block lasted approximately 4 min, for a total task duration of
approximately 50 min. Ten blocks were conventional AX-CPT as described above, while the 11th block consisted of only AX trials. This block
allowed us to measure the performances and ERPs related to a context in which participants would never have to rely on proactive or reactive
cognitive control strategies as no inhibition was ever required.38 Participants were warned about the particularity of this “AX-only” block
through specific instructions. This block was randomly presented to the participants as either the first block of the whole task or as the last one
to ensure that no effect from fatigue could influence the results. Response times, the number of correct answers (ie, pushing the button for an
AX trial), and the number of commission errors (ie, pushing the button for a BX, AY, or BY trial) were recorded. This allowed us to have
access to three main clues: (1) the number of hits that should not have been generated on AY trials (valid cue, invalid target), indexing reactive
control capacities; (2) the supplementary time needed to adjust performance in a context requiring inhibition versus a context with no
inhibition at all, assessed by the subtraction “RTs AX trials minus RTs AX-only”38; and (3) the proactive behavioral index (PBI)17([AY−BX] /
[AY + BX]), as a low index value suggests a strategy mainly based on reactive control while a high index value mainly refers to a proactive one.
Schröder et al 321

Urgency Premeditation Perseverance and Sensation seeking the “AX-only” context. The averages CNV was computed for
impulsive behavior scale (UPPS)36; and the Behavioral each participant individually on the following set of electrodes:
Inhibition, Behavioral Activation scales (BIS/BAS).37 The par- Fz, FC1, FC2, C3, Cz, and C4.12
ticipants were allowed to complete the questionnaires after the
first 5 blocks of the AX-CPT task to reduce the mental burden Reaction to Targets. We created epochs from −200 to +800 ms
of task fatigue. around target onset, with a baseline that went from −200 to
At the end of the AX-CPT, the participants underwent a short 0 ms. We averaged the N2 (the largest negative value within
cognitive evaluation comprising conventional neuropsychologi- the 200 to 400 ms interval) and the P3 (the largest positive
cal tasks assessing working memory (WAIS-III39 and N-back value within the 300 to 600 ms interval) components following
task with 3 levels of difficulty40), inhibition (Stroop task41), and an AX trial, an AY trial, a BX trial, a BY trial, and an AX trial
attention (D2 task42), as these cognitive abilities are thought to with the “AX-only” context. The averages N2 and P3 compo-
influence the efficiency of cognitive control strategies.3 nents were computed for each participant individually on fron-
tocentral electrodes F3, Fz, F4, FC1, FC2, C3, Cz, and C4.16
EEG Recordings and Analysis
Electroencephalogram (EEG) activity was recorded with 32 Statistical Analysis
electrodes mounted on a Quik-Cap in standard and intermediate We performed a forward stepwise regression analysis to
positions, following the 10–20 system. The impedance of all of explore the influence of potential predictors (demographic
the electrodes was kept below 10 kΩ. Recordings were made data, cognitive abilities, personality traits, and clinical tenden-
using an online linked-mastoids reference. The signal was cies) on the behavioral performance and ERP components
amplified by battery-operated ANT® amplifiers with a gain related to cognitive control as defined by the DMC model.
of 30,000 and a bandpass filter of 0.01 to 100 Hz. EEGs were The choice of the independent variables among the demo-
recorded continuously at a sampling rate of 1024 Hz with graphic, cognitive, and clinical/personality data was based on
ANT Eeprobe™ software. Only correct answers were consid- correlation analysis and was therefore different for each depen-
ered in the subsequent analysis. Every trial compromising eye dent variable.43 The correlation results and the resulting inde-
movements (such as eye blinks) or muscular artifacts was pendent variables selected for each regression analysis are
removed manually and excluded from further analysis. presented in Table 2, as are the results of the resultant regres-
Consequently, overall, 63.08% of the trials were retained and sion analysis. The performances and neurophysiological data
averaged. A minimum of 10 accepted trials was required to were explored as dependent variables.
average and compute the related components. The data were fil- Moreover, the links between behavioral performances and
tered with a 0.3 to 30 Hz bandpass filter. Different time frames ERP components were explored using a linear regression
were created around cues and targets in order to analyze reac- model including the amplitudes of the cue-N1, cue-N2,
tion to cues, response preparation, and reaction to targets. cue-P3 for B-cues, and the CNV for A-cues, thus requiring pro-
This coding allowed us to compute different averages of ERPs. active inhibitory control, as predictors of the BX commission
error rate, and the target-N2 and the target-P3 amplitudes,
Reaction to Cues. We created epochs from −200 ms before the requiring reactive inhibitory control, as predictors of the AY
onset of the cue to +800 ms after the cue presentation, using the commission error rate.
−200 to +0 ms interval as baseline. This allowed us to average A Holm–Bonferroni correction was applied for each regres-
the (1) cue-N1 (the largest negative value within the 0 to sion analysis.44 The resulting new P-value (p′ ) are reported in
200 ms interval); (2) the cue-N2 (the largest negative value Table 2.
within the 200 to 400 ms interval); and (3) the cue-P3 (the
largest positive value within the 300 to 600 ms interval) follow-
ing the apparition of an A-cue, a B-cue, and an A-cue in the Results
“AX-only” context. The averages cue-N1 and cue-N2 were
computed for each participant individually on the frontocentral
Descriptive Statistics
electrodes: F3, Fz, F4, FC1, FC2, C3, Cz, and C4.10,11 The Behavioral Performances. The participants performed the task
averages cue-P3 was computed for each participant individu- with a reaction time (RT) of 360.38 ± 65.23 ms in the Go/
ally on the parietal electrodes P3, Pz, and P4.12 NoGo condition and 290.43 ± 66.18 ms in the Go-only condi-
tion. The participants were significantly slower in the Go/
Response Preparation. We created epochs from −600 ms before NoGo condition (t(103) = 13.84; P < .001), displaying an effi-
the onset of the target to +200 after the target presentation, cient speed adaptation consequent to a restraint of automatic
using the −600 to −400 ms interval as baseline. We were, responses in order to avoid erroneous responses, and, thereby,
therefore, able to average the contingent negative variation attesting to proactive control activation in a situation of uncer-
(CNV), defined as the largest negative value within the 0 to tainty.45 On average, participants made 18.5 ± 13.23 (/80)
200 ms interval following an A-cue, a B-cue, and an A-cue in commission errors on AY trials, 4.08 ± 4.73 (/80) commission
 Table 2. Links Between Behavioral Performances and ERPs Components with Demographic, Personality, and Cognitive Variables.

322
Dependent Independent variables (P ≤ Model’s Test Value Model’s Δ R2 (Adjusted Β (Adjusted
Variables .05*; P ≤ .01**) Model (Adjusted P-Value) R2 P-Value) Introduced Variables p-Value)
BEHAVIORAL Reaction time 1.Alcohol/week** Model 1 12.873 (p′ = .003) 0.113 0.113 (p′ = .03) N2 reaction time 0.336 (p′ <.001)
PERFORMANCES AX (Go/ 2. N2 back reaction time** Model 2 10.648 (p′ = .003) 0.176 0.063 (p′ = .014) N2 reaction time 0.335 (p′ = .002)
NoGo 3. Audit* Fagerstrom 0.250 (p′ = .007)
condition) 4.Fagerstrom* Model 3 8.823 (p′ = .003) 0.211 0.035 (p′ = .038) N2 reaction time 0.308 (p′ = .003)
Fagerstrom 0.220 (p′ = .034)
Alcohol/week 0.192 (p′ = .038)
Reaction time 1. Fagerstrom* Model 1 4.351 p′ = .039 0.041 0.041 (p′ = .039) Fagerstrom 0.203 (p′ = .039)
AX (Go-only) 2. N2 back reaction time* Model 2 4.305 p′ = .032 0.079 0.038 (p′ = .078) Fagerstrom 0.202 (p′ = .076)
N2 reaction time 0.195 (p′ = .076)
Commission 1. Educational level * Model 1 7.422 p′ = .008 0.068 0.068 (p′ = .024) N2 commission 0.262 (p′ = .008)
errors BX 2. N2 back correct errors
response* Model 2 6.131 p′ = .006 0.109 0.041 (p′ = .062) N2 commission 0.230 (p′ = .036)
3. N2 back commission errors
errors ** N2 correct response −0.204 (p′ = .036)
Model 3 5.830 p′ = .003 0.150 0.041 (p′ = .062) N2 commission 0.221 (p′ = .06)
errors
N2 correct response −0.219 (p′ = .06)
Educational Level −0.203 (p′ = .06)
Commission 1. Age ** Model 1 7.548 p′ = .007 0.070 0.070 (p′ = .014) Age −0.264 (p′ = .007)
error AY 2. Educational level * Model 2 7.240 p′ = .002 0.126 0.057 (p′ = .014) Age −0.266 (p′ = .010)
3. N2 back reaction time* N2 Reaction time −0.239 (p′ = .012)
ERPs Cue-N1 (B-cue) 1. Age* Model 1 6.065 (p′ = .016) 0.058 0.058 (p′ = .018) Gender −0.240 (p′ = .016)
COMPONENTS 2. Gender* Model 2 6.806 (p′ = .004) 0.122 0.064 (p′ = .018) Gender −0.265 (p′ = .012)
3. Audit* 4. UPPS urgency* UPPS Urgency 0.255 (p′ = .012)
Cue-N2 (B-cue) 1. Gender * Model 1 6.121 (p′ = .015) 0.058 0.058 (p′ = .015) Gender −0.241 (p′ = .015)
2. Alcohol consumption
per week *
3. Fagerström *
4. Forward digit span *
Cue-P3 (B-cue) 1.UPPS total * Model 1 4.158 (p′ = .044) 0.04 0.04 (p′ = .044) UPPS total −0.201 (p′ = .044
2. UPPS lack of
premeditation*
CNV (A-cue in 1. Gender** Model 1 9.055 (p′ = .003) 0.082 0.082 (p′ = .009) Gender −0.287 (p′ = .003)
Go/NoGo 2. Audit score** Model 2 8.750 (p′ = .003) 0.149 0.067 (p′ = .01) Gender −0.290 (p′ = .004)
condition) 3. Fagerström score ** Stroop I-D error 0.258 (p′ = .006)
4. UPPS urgency * total
5. D2 score Model 3 8.984 (p′ = .003) 0214 0065 (p′ = .01) Gender −0.257 (p′ = .006)
6. Error total on Stroop Stroop I-D error total 0.287 (p′ = .01)
(interference minus Fagerström score 0.259 (p′ = .01)
denomination) **
CNV (A-cue in 1.Fagerstrom score * Model 1 11.045 (p′ = .001) 0.099 0.099 (p′ = .001) UPPS total 0.315 (p′ = .001)
2. STAI trait **

(continued)
Schröder et al 323

errors on BX trials, and 0.36 ± 0.85 (/80) commission errors on

BY trials. An average of 10.37 ± 15.9 omission errors were

Abbreviations: CNV, contingent negative variation; D2 GZ-F, total number of items processed minus total number of errors on the D2 test; ERP, event-related potentials; STAI, state and trait anxiety
0.299 (p′ = .004)

0.288 (p′ = .006)

Β (Adjusted
p-Value)
made by the participants in the Go/NoGo condition and 6.94
± 9 in the Go-only condition. The participants failed to
respond significantly more in the Go/NoGo condition (t(103)
= 3.109; P < .002), which can be interpreted as a consequence
of the uncertainty of the response production due to the Go/
Introduced Variables

NoGo condition.
Forward Digit span
Neurophysiological Profile of the Components. For the remaining
104 participants, the results for 3 participants had to be removed
D2 Gz-F

The independent variables were chosen based on correlation analysis. All the P-value are adjusted with the Bonferroni–Holm correction for multiple comparisons.
from every analysis due to poor EEG signal quality (<10 aver-
aged trials per condition), leading to a data-loss ratio of 3%. The
analyses and grand average processes were, therefore, per-
formed on 101 participants.
Δ R2 (Adjusted

0.089 (p′ = .004)

0.083 (p′ = .006)

As described in the Introduction section, the 6 components

P-Value)

of interest were the cue-N1,11 cue-N2,12 cue-P3,13,14 and the

CNV13,14 in regard to response preparation, and the
target-N213–16 and target-P314,15,17 as target-components. The
characteristics (amplitude and latency) of each component are
presented in Figure 3.
Model’s
R2

0.089

0.083

Regression Analyses
Model’s Test Value
(Adjusted P-Value)

Only the models that explain the most variance will be pre-
Model 1 8.846 (p′ = .004)

Model 1 8.024 (p′ = .006)

sented here. Please refer to Table 2 fore mode details about

the inferior models.

Behavioral Performances. The reaction times in the Go/NoGo

condition were predicted by “the reaction time of the N2
back task,” “the Fagerström score,” and “the weekly alcohol
inventory; UPPS, urgence premeditation perseverance sensation seeking impulsivity scale.
Model

consumption.” This model explains 21.1% of the variance

(F(3.99) = 8.823, p′ = .003). Individuals with a slower reaction
time in the N2 back task, a higher level of nicotine dependence,
Independent variables (P ≤

and/or higher alcohol consumption per week, were slower to

1. N2 back reaction time*
2. Forward digit span **

realize the task.

2. D2 GZ-F score **
.05*; P ≤ .01**)

3. D2 GZ-F score*

The reaction time in the Go-only condition was predicted by

4. UPPS urgency**

the Fagerström score (F(1.101) = 4.351, p′ = .039). This model

premeditation *
3. UPPS total**

5. UPPS lack of

explains 4.1% of the variance. The more the participants were

1. Gender*

nicotine-dependent, the slower they were in this task in a

context of certainty.
The variance of the BX commission error rates was signifi-
cantly predicted by the “commission error rates in the N2 back
task” (F(1.101) = 7.422; p′ = .08). This model explains 6.8% of
Targets N2 (AY

Target P3 (AY
Dependent
Variables

the variance. Individuals who make more commission errors in

condition)
Go-only

targets)

targets)

the N2 back tended to also make more mistakes related to pro-

active control.
The variation of AY error rates could be predicted by “the
Table 2. (continued)

age” and “the reaction time in the N2 back task.” This model
explains 12.6% of the variance (F(2.100) = 7.240, p′ = .002).
Younger participants and individuals with faster reaction
times in the N2 back task committed more AY errors.
All of the statistical parameters are presented in Table 2.
Regarding the link between the behavioral performances and
the related ERPs, a linear regression computed with a model
324 Clinical EEG and Neuroscience 55(3)

Figure 3. Grand average of the ERP components and the neurophysiological characteristics for (IIIa) the proactive components following the
onset of the cue, (IIIb) the proactive components related to response anticipation, and (IIIc) the reactive components following the onset of the
target. Abbreviations: CNV, contingent negative variation; ERP, event-related potentials.
Schröder et al 325

composed of the amplitudes of the cue-N1, cue-N2, cue-P3 for been described as proactive-related components,12–14 in proac-
B-cues, and the CNV for A-cues, thus requiring proactive inhib- tive inhibitory control as their amplitudes significantly pre-
itory control, significantly predicted the commission error rate dicted the performance in regard to proactive control (ie, the
for BX trials (R2 = 0.126; F = 3.431; P < .05). In regard to reac- error rate on BX trials). Similarly, we were able to analyze
tive control, a model combining the 2 components from AY ERP components thought to reflect reactive control (ie, the
trials, the target-N2 and the target-P3 amplitudes, failed to signif- target-N2 and target-P314). Here, the link between these com-
icantly predict the commission error rate for AY trials (P = .511), ponents and the direct behavioral aspect of reactive control
although it successfully predicted the reaction time with AX (ie, the AY commission error rate) is less straightforward.
trials (R2 = 0.23; F = 13.279; P < .001), which in turn was Indeed, reactive ERPs failed to predict the AY commission
highly predictive of the AY commission error rate (R2 = 0.276; error rate, although they successfully predicted the reaction
F = 38.924; P < .001). However, this is not a direct effect, and times for AX trials, which in turn appear to be predictive of
it should, therefore, be treated with a degree of caution. the AY commission error rate. This may have a double
origin: (1) as the study involved healthy students, we had to
ERPs Component Amplitude. The B-cue N100 amplitude was avoid threshold accuracy performance, and the instructions
predicted by the “Gender” and the “UPPS urgency score.” emphasized the need to be “as fast as possible” when pushing
This model explains 12.2% of the variance (F(2.98) = 6.806, the button for Go trials, as time pressure is known to generate
p′ = .004). Men and individuals with higher scores on the a higher rate of false alarms on NoGo trials. In our study, reac-
UPPS urgency subscale appeared to have smaller N100 ampli- tive ERP components were then directly linked to the process-
tudes for B-cues. ing speed of Go trials, which predicted NoGo errors. (2) The
The amplitude of the B-cue-N2 was predicted by gender establishment of a dominant response relies on the development
(F(1.99) = 6.121; p′ = .015). This model predicted 5.8% of the of an automatic process, which is robust, fast, and efficient, and
variance. Within this model, men appeared to have smaller that is dependent on the efficiency of reactive control.2,3 In this
amplitudes than women. way, speed is directly linked to reactive cognitive control.
The amplitude of the B-cue P3 was predicted by the total UPPS
score (F(1.99) = 4.158; p′ = .044). This model explains 4% of the
Proactive Inhibitory Control
variance. Participants with higher scores on the UPPS total scale
appeared to have smaller amplitudes for the B-cue P3. It is of particular interest to observe in this study that only the
The amplitude of CNV in Go-NoGo condition was predicted proactive side of the inhibitory control was affected by the
by gender, the “total number of errors I-D on the Stroop task” gender of the individual. It has already been shown that
and the “Fagerström score.” This model explains 21.4% of the impaired impulse control associated with substance use disor-
variance (F(3,97) = 8.984, p′ = .003). Men and individuals with ders is more prevalent in males than females46 and that proac-
a higher number of errors on the Stroop test and higher nicotine tive control is greater in women than in males.47 However, to
dependence had a smaller CNV in a context of uncertainty our knowledge, this is the first study to show, at an electrophys-
regarding the motor preparation. iological level, that women allocate more cognitive resources
The amplitude of the CNV in the Go-only condition, thus in than men in proactive cognitive control.
a context of certainty, can be significantly predicted by the In a cognitive perspective, we were able to highlight the
UPPS total score (F(1.99) = 11.045; p′ = .001). This model potential role of working memory in proactive inhibitory
explains 9.9% of the variance. Individuals with higher scores control, since working memory abilities significantly predicted
on the UPPS total scale have smaller amplitudes of the CNV the performance on BX trials, as predicted by Braver.3 This
in a context of certainty. result was expected since proactive control is conceptualized
The amplitude of the NoGo N2 for AY trials, thus requiring as active maintenance of goal representations before and
reaction inhibition, was significantly predicted the score on the throughout a task. Moreover, it has already been demonstrated
forward digit span (F(1.99) = 8.846; P = .004). It explains 8.9% that an increase in working memory load during a task induces
of the variance. Individuals with higher scores for the forward a switch from proactive to reactive control, as proactive control
digit span task have smaller NoGo N2 amplitudes for AY trials. requires a high availability of working memory capacity.48
The amplitude of the NoGo P3 for AY trials was predicted Proactive inhibitory control also appeared to be predicted by
by the GZ-F score (F(1.99) = 8.024; P = .006). It explains the number of errors on the Stroop test (interference-
8.3% of the variance. Individuals with a higher score on the denomination condition), as a poor performance on the
D2 task have a higher NoGo P3 amplitude for AY trials. Stroop test was predictive of a reduced CNV. This result was
All of the analysis results are presented in Table 2. somewhat surprising as the Stroop task is often presented as a
“reactive task” in the literature.26 However, evidence from
studies with patients with executive and frontal deficits suggests
Discussion
a specific role for the frontal executive system in the top-down
This exploratory study allowed us to highlight the role of the selection mechanism at play in the Stroop task, along with the
cue-N1, cue-N2, cue-P3, and CNV, which have previously suppression of automatic responses.49 Several studies have also
326 Clinical EEG and Neuroscience 55(3)

indicated that a higher working memory load leads to a worse “attentional inhibitory control” model, in which proactive and
performance on the Stroop task,50 thereby emphasizing a poten- reactive control are considered to be reflections of sustained
tial role of proactive control strategies when performing the and selective attention in the implementation of inhibitory
Stroop task. control (reviewed in Perri21). Interestingly, our analyses failed
Interestingly, nicotine consumption significantly predicted to highlight a role of sustained and selective attention, as mea-
the amplitude of the CNV (A-cues), suggesting that higher con- sured by the D2 test, on proactive inhibitory control variables.
sumption patterns predict lower amplitudes for this components. Working memory abilities appear to play a significant role in
The effect of tobacco consumption30 on inhibitory control points reactive inhibitory control since, in this study, we observed that
toward a general deficit in inhibitory control in individuals who both the reaction time of the N2 back task and the forward digit
smoke. ERP modulations in these populations have also been span are predictive of reactive inhibitory control (AY error rate
quite well documented: N1 reduction related to alcohol con- and the amplitude of N2 for AY trials). In both cases, we
sumption,50 and a reduced N2 among smokers.51 However, to observed that lower working memory performances are an indi-
date, it is still not clear whether a dissociation can be seen cator of lower performances in reactive inhibition. This is con-
between proactive and reactive inhibitory control in alcohol con- sistent with the fact that both proactive and reactive inhibitory
sumers and/or smokers. Our results suggest that nicotine- control is dependent on the working memory load.3
dependance already has an impact at the very beginning of the
processing chain and, therefore, may disrupt the top-down regu-
lation mechanisms necessary for the establishment of proactive Conclusions
inhibitory control. Indeed, the neural substrates of proactive inhi-
Our results highlight the impact of cognitive processes on cog-
bition include the right inferior frontal gyrus,52 which is known
nitive control, such as inhibition of a dominant response on pro-
to be dysregulated in nicotine dependance.53 The results in the
active control and working memory (as predicted by the
literature in regard to the CNV are surprisingly unclear, although
model3), as well as selective attention on reactive control.
our results point toward a disruption of response expectation pre-
Moreover, our results suggest an early impact of alcohol and
dicted by the nicotine use. These results are of particular interest,
cigarette use on the efficiency of cognitive control, and specif-
as they suggest proactive cognitive control might be specifically
ically on proactive control strategies. Finally, our results high-
impaired in substance use disorders, which is crucial, as proac-
light the importance of taking into consideration demographic
tive cognitive control might be critical in the establishment
and personality factors in future research in light of their influ-
and maintenance of abstinence.54
ence on cognitive control.
Finally, proactive control, evaluated by the N100 and P300
components for B-cues, appeared to be predicted by the UPPS
total score and the urgency subscale score, suggesting that Limitations
people with a high level of impulsivity tend to allocate less cog-
nitive resources to the proactive inhibition of automatized behav- In regard to the regression analyses performed, as our sample
ior. This result is interesting because, to our knowledge, no study was comprised almost exclusively of university students
to date has highlighted the link between impulsivity measured by between their first and final year of studies, it may, therefore,
the UPPS scale and proactive inhibitory control. Since impulsiv- lack variability. Our results should be replicated with a larger
ity is considered to be a vulnerability factor for many psychiatric and more diverse sample, which could highlight influences
disorders, including substance use disorders,55 the fact that this that were not discernible in the present study due to the relative
construct only predicts the proactive side of inhibitory control homogeneity of our sample. Moreover, the independent vari-
makes its study particularly interesting. ables selected in the predictive model were chosen based on
the literature and variability in our sample, although this selec-
tion process was also dictated by the size of our sample. Further
Reactive Inhibitory Control studies should consider the use of alternative variables as
Target-P3, related to reactive control, appeared to be predicted markers of the investigated processes based on larger samples.
by the attentional abilities as measured by the D2 test. This is
consistent with the described mechanism of reactive control, Author Contributions
as it is based on the recruitment of attentional processes as a
late correction mechanism mobilized only when needed.3 It ES is a research fellow at the Fond pour la Recherche en Sciences
Humaines (FRESH) from the Belgian Fund for Scientific Research
is, therefore, crucial for reactive control to rely on efficient
(FRS-FNRS, Belgium). SC is a senior research associate at the
attentional orientation abilities. While the interdependence Belgian Fund for Scientific Research (FRS-FNRS, Belgium).
between attention and inhibitory control has been well docu-
mented, it is, however, worth noting that a differential impact
of attentional abilities on proactive versus reactive control is Ethical Approval
widely debated.21 Moreover, this interdependence of attention The local ethics committee of Brugmann Hospital (“Comité d’Ethique
and inhibition constitutes one of the main arguments for an Hospitalière OM 026”) provided its approval for this study.
Schröder et al 327

Declaration of Conflicting Interests 14. Chaillou AC, Giersch A, Hoonakker M, et al. Evidence of impaired
proactive control under positive affect. Neuropsychologia. 2018;
The author(s) declared no potential conflicts of interest with respect to
114:110‐117.
the research, authorship, and/or publication of this article.
15. Schevernels H, Bombeke K, Van der Borght L, Hopf J-M, Krebs
RM, Boehler CN. Electrophysiological evidence for the involve-
Funding ment of proactive and reactive control in a rewarded stop-signal
The author(s) disclosed receipt of the following financial support for task. NeuroImage. 2015;121:115‐125.
the research, authorship, and/or publication of this article: This work 16. Bailey K, Bartholow BD. Alcohol words elicit reactive cognitive
was supported by the Fonds De La Recherche Scientifique—FNRS, control in low-sensitivity drinkers. Psychophysiology. 2016;53(11):
Brugmann Foundation (grant number Call “FRESH” 2015— 1751‐1759.
FRESH, Grant Call 2019). 17. Kropotov J, Ponomarev V, Tereshchenko EP, Müller A, Jäncke L.
Effect of aging on ERP components of cognitive control. Front
Aging Neurosci. 2016;8:69.
ORCID iD 18. Smith J, Jamadar S, Provost A, Michie P. Motor and non-motor
Salvatore Campanella https://orcid.org/0000-0002-7331-5906 inhibition in the Go/NoGo task: an ERP and fMRI study. Int J
Psychophysiol. 2013;87(3):244‐253.
19. Braver TS, Paxton JL, Locke HS, Barch DM. Flexible neural
References mechanisms of cognitive control within human prefrontal
1. Botvinick MM, Braver TS, Barch DM, Carter CS, Cohen JD. cortex. Proc Natl Acad Sci USA. 2009;106(18):7351‐7356.
Conflict monitoring and cognitive control. Psychol Rev. 20. Chambers CD, Garavan H, Bellgrove MA. Insight into the neural
2001;108(3):624‐652. basis of response inhibition from cognitive and clinical neurosci-
2. Braver TS, Gray JR, Burgess GC. Explaining the many varieties ence. Neurosci Biobehav Rev. 2009;33(5):631‐648.
of working memory variation: dual mechanisms of cognitive 21. Perri RL. Is there a proactive and a reactive mechanism of inhibi-
control. In: Conway ARA, Jarrold C, Kane MJ, Miyake A, tion? Towards an executive account of the attentional inhibitory
Towse JN, (eds) Variation in working memory. New York: control model. Behav Brain Res. 2020;377:112243.
Oxford University Press, 2007, pp. 76–106. 22. Sharif-Razi M, Hodgins DC, Goghari VM. Reactive and proactive
3. Braver TS. The variable nature of cognitive control: a dual mech- control mechanisms of response inhibition in gambling disorder.
anisms framework. Trends Cogn Sci. 2012;16(2):106‐113. Psychiatry Res. 2019;272:114‐121.
4. Verbruggen F, Logan G. Control in response inhibition. In: The 23. Xiang L, Chen Y, Chen A, Zhang F, Xu F, Wang B. The effects of
Wiley handbook of cognitive control. John Wiley & Sons, Ltd; trait impulsivity on proactive and reactive interference control.
2017:97‐110. Brain Res. 2018;1680:93‐104.
5. Karayanidis F, Whitson LR, Heathcote A, Michie PT. Variability 24. Vanderhasselt MA, De Raedt R, De Paepe A, et al. Abnormal pro-
in proactive and reactive cognitive control processes across the active and reactive cognitive control during conflict processing in
adult lifespan. Front Psychol. 2011;2:318. major depression. J Abnorm Psychol. 2014;123(1):68‐80.
6. Czernochowski D. ERPs dissociate proactive and reactive control: 25. Fales CL, Barch DM, Burgess GC, et al. Anxiety and cognitive
evidence from a task-switching paradigm with informative and efficiency: differential modulation of transient and sustained
uninformative cues. Cogn Affect Behav Neurosci. 2014;15(1): neural activity during a working memory task. Cogn Affect
117‐131. Behav Neurosci. 2008;8(3):239‐253.
7. West R, Choi P, Travers S. The influence of negative affect on the 26. Yang Y, Miskovich TA, Larson CL. State anxiety impairs proac-
neural correlates of cognitive control. Int J Psychophysiol. tive but enhances reactive control. Front Psychol. 2018;9:2570.
2010;76(2):107‐117. 27. Jimura K, Locke HS, Braver TS. Prefrontal cortex mediation of
8. Holmes AJ, Pizzagalli DA. Task feedback effects on conflict mon- cognitive enhancement in rewarding motivational contexts. Proc
itoring and executive control: relationship to subclinical measures Natl Acad Sci USA. 2010;107(19):8871‐8876.
of depression. Emotion. 2007;7(1):68. 28. Petit G, Kornreich C, Noël X, Verbanck P, Campanella S.
9. Grane VA, Brunner JF, Endestad T, et al. ERP correlates of pro- Alcohol-related context modulates performance of social drinkers
active and reactive cognitive control in treatment-naïve adult in a visual Go/No-Go task: a preliminary assessment of
ADHD. PLoS ONE. 2016;11(7):e0159833 11. event-related potentials. PLoS ONE. 2012;7(5):e37466.
10. Morales J, Yudes C, Gómez-Ariza CJ, Bajo MT. Bilingualism 29. Franken IH, Luijten M, van der Veen FM, Van Strien JW.
modulates dual mechanisms of cognitive control: evidence from Cognitive control in young heavy drinkers: an ERP study. Drug
ERPs. Neuropsychologia. 2015;66:157‐169. https://doi.org/10. Alcohol Depend. 2017;175:77‐83.
1016/j.neuropsychologia.2014.11.014 30. Detandt S, Bazan A, Schröder E, et al. A smoking-related back-
11. Elchlepp H, Lavric A, Chambers CD, Verbruggen F. Proactive ground helps moderate smokers to focus: an event-related poten-
inhibitory control: a general biasing account. Cognit Psychol. tial study using a Go-NoGo task. Clin Neurophysiol.
2016;86:27‐61. 2017;128(10):1872‐1885.
12. Lu D, Zhang H, Kang C, Guo T. ERPs evidence for the relation- 31. Saunders JB, Aasland OG, Babor TF, de la Fuente JR, Grant M.
ship between fluid intelligence and cognitive control. Development of the alcohol use disorders identification test (AUDIT):
Neuroreport. 2016;27(6):379‐383. WHO collaborative project on early detection of persons with
13. van Wouwe NC, Band GPH, Ridderinkhof KR. Positive affect harmful alcohol consumption—II. Addiction. 1993;88(6):791‐804.
modulates flexibility and evaluative control. J Cogn Neurosci. 32. Skinner HA. The drug abuse screening test. Addict Behav.
2011;23(3):524‐539. 1982;7(4):363‐371.
328 Clinical EEG and Neuroscience 55(3)

33. Fagerström KO. Measuring degree of physical dependence to 45. Criaud M, Wardak C, Ben Hamed S, Ballanger B, Boulinguez P.
tobacco smoking with reference to individualization of treatment. Proactive inhibitory control of response as the default state of
Addict Behav. 1978;3(3-4):235‐241. executive control. Front Psychol. 2012;3:59.
34. Beck AT, Steer RA, Carbin MG. Psychometric properties of the 46. Noble WS. How does multiple testing correction work? Nat
beck depression inventory: twenty-five years of evaluation. Clin Biotechnol. 2009;27(12):1125‐1137.
Psychol Rev. 1988;8(1):77‐100. 47. Smittenaar P, Rutledge RB, Zeidman P, et al. Proactive and reac-
35. Spielberger CD, Gorsuch RL. Manual for the state-trait anxiety tive response inhibition across the lifespan. PLoS ONE. 2015;
inventory (form Y):(“self-evaluation questionnaire”). Consulting 10(10):e0140383.
Psychologists Press, Incorporated; 1983. 48. Speer NK, Jacoby LL, Braver TS. Strategy-dependent changes in
36. Whiteside SP, Lynam DR. Understanding the role of impulsivity and memory: effects on behavior and brain activity. Cogn Affect
externalizing psychopathology in alcohol abuse: application of the Behav Neurosci. 2003;3(3):155‐167.
UPPS impulsive behavior scale. Exp Clin Psychopharmacol. 49. Kalanthroff E, Avnit A, Henik A, Davelaar EJ, Usher M.
2003;11(3):210‐217. Stroop proactive control and task conflict are modulated by
37. Carver CS, White TL. Behavioral inhibition, behavioral activa- concurrent working memory load. Psychon Bull Rev. 2014;
tion, and affective responses to impending reward and punish- 22(3):869‐875.
ment: the BIS/BAS scales. J Pers Soc Psychol. 1994;67(2):319. 50. Oscar-Berman M. Alcohol-related ERP changes in cognition.
38. Dias EC, Foxe JJ, Javitt DC. Changing plans: a high density elec- Alcohol. 1987;4(4):289‐292.
trical mapping study of cortical control. Cereb Cortex. 51. Buzzell GA, Fedota JR, Roberts DM, McDonald CG. The N2
2003:13(7):701‐715. ERP component as an index of impaired cognitive control in
39. Wechsler D. WAIS-III, Wechsler adult intelligence scale: admin- smokers. Neurosci Lett. 2014;563:61‐65.
istration and scoring manual. Psychological Corporation. 1997. 52. Meyer HC, Bucci DJ. Neural and behavioral mechanisms of pro-
40. Baddeley A. Working memory: looking back and looking active and reactive inhibition. Learn Mem. 2016;23(10):504‐514.
forward. Nat Rev Neurosci. 2003;4(10):829‐839. 53. Bell S, Froeglier B. Association between smoking abstinence,
41. Stroop JR. Studies of interference in serial verbal reactions. J Exp inhibitory control, and smoking behavior: an fMRI study. Front
Psychol. 1935;18(6):643. Psychiatry. 2021;1:150.
42. Brickenkamp R. Manuel du test d2. Editest; 1969. 54. Wilcox CE, Dekonenko CJ, Mayer AR, Bogenschutz MP, Turner
43. Costanza A, Rothen S, Achab S, et al. Impulsivity and JA. Cognitive control in alcohol use disorder: deficits and clinical
impulsivity-related endophenotypes in suicidal patients with sub- relevance. Rev Neurosci. 2014;25(1):1‐24.
stance use disorders: an exploratory study. Int J Ment Health 55. Robbins TW, Gillan CM, Smith DG, de Wit S, Ersche KD.
Addict. 2021;19(5):1729‐1744. Neurocognitive endophenotypes of impulsivity and compul-
44. Holm S. A simple sequentially rejective multiple test procedure. sivity: towards dimensional psychiatry. Trends Cogn Sci. 2012;
Scand J Stat. 1979;6(2):65–70. 16(1):81‐91.