www.sciencedirect.com

c o r t e x 1 0 6 ( 2 0 1 8 ) 1 2 0e1 3 1

Available online at

ScienceDirect

Journal homepage: www.elsevier.com/locate/cortex

Research report

Hypervigilance during anxiety and selectiveattention during fear: Using steady-state visualevoked potentials (ssVEPs) to disentangle attention

mechanisms during predictable and unpredictable threat

Anna K. Kastner-Dorn a, Marta Andreatta a, Paul Pauli a andMatthias J. Wieser a,b,*

a Department of Psychology (Biological Psychology, Clinical Psychology, and Psychotherapy), University of

Wurzburg, Wurzburg, Germanyb Department of Psychology, Education, and Child Studies, Erasmus University Rotterdam, Rotterdam, The

Netherlands

a r t i c l e i n f o

Article history:

Received 8 January 2018

Reviewed 14 February 2018

Revised 25 April 2018

Accepted 9 May 2018

Action editor Peter Kirsch

Published online 21 May 2018

Keywords:

Fear

Anxiety

Attention

Defensive responding

EEG

ssVEP

Heart rate

* Corresponding author. Erasmus UniversityE-mail address: wieser@essb.eur.nl (M.J. W

https://doi.org/10.1016/j.cortex.2018.05.0080010-9452/© 2018 Elsevier Ltd. All rights rese

a b s t r a c t

Anxiety is induced by unpredictable threat, and presumably characterized by enhanced vig-

ilance. In contrast, fear is elicited by imminent threat, and leads to phasic responses with

selective attention. In order to investigate attention mechanisms and defensive responding

during fear and anxiety, we employed an adaptation of the NPU-threat test and measured

cortical (steady-state visual evoked potentials, ssVEPs), physiological (heart rate, HR), and

subjective responses (ratings) topredictable (fear-related) andunpredictable (anxiety-related)

threat in 42 healthy participants. An aversive unconditioned stimulus (US, loud noise) was

100% predicted by a cue (predictable P-cue) in one context (predictable P-context), but

appeared unpredictably within a different context (unpredictable U-context, U-cue), while it

was never delivered in a neutral safe context (N-cue, N- context). In response to predictable

threat (P-cue), increased ssVEP amplitudes and accelerated HR were found. Both predictable

andunpredictablecontextsyielded increasedssVEPamplitudes compared to thesafecontext.

Interestingly, in the unpredictable context participants showed longer-lasting visuocortical

activation than in the predictable context, supporting the notion of heightened vigilance

during anxiety. In parallel, HR decelerated to both threat contexts indicating fear bradycardia

to these threatening contexts as compared to the safe context. These results support the idea

of hypervigilance in anxiety-like situations reflected in a long-lasting facilitated processing of

sensory information, in contrast to increased selective attention to specific imminent threat

during fear. Thus, this study further supports the defense-cascade model with vigilance and

orienting in the post-encounter phase of threat (anxiety), while selective attention and

defensive mobilization in the circa-strike phase of threat (fear).

© 2018 Elsevier Ltd. All rights reserved.

Rotterdam, Institute of Pieser).

rved.

sychology, P.O. Box 1738, 3000 DR, Rotterdam, The Netherlands.

c o r t e x 1 0 6 ( 2 0 1 8 ) 1 2 0e1 3 1 121

1. Introduction

According to DSM-5, fear is oriented towards specific immi-

nent threat, while anxiety is defined as the anticipation of

unspecific future threat (American Psychiatric Association,

2013). In the same vein, experimental studies consider fear

as a phasic response to concrete threat that dissolves quickly

when the respective threat is removed, while anxiety is

regarded as a sustained affective response (Davis, Walker,

Miles, & Grillon, 2010; Perusini & Fanselow, 2015; Tovote,

Fadok, & Luthi, 2015). In consequence, fear is expected to

activate selective attention mechanisms towards a specific

threat stimulus, whereas anxiety should lead to a sustained

heightened vigilance, whose broadening attention aims at

detecting potential but yet to be detected threat in the

environment.

In experimental research, fear can be modeled using cue

conditioning, where a cue becomes a reliable predictable

signal of an aversive unconditioned stimulus and conse-

quently elicits fear responses. Anxiety can be modeled using

context conditioning, where the US is associated with a long-

lasting stimulus, i.e., the context, in the absence of a pre-

dictable cue. As a consequence, a sustained fear response (i.e.,

anxiety) is elicited throughout the duration of the context

(Davis et al., 2010). Thus, the predictability of the aversive

event plays a crucial role during these conditioning paradigms

(Bouton, 1994; Maren, Phan, & Liberzon, 2013). Fear responses

reflecting defensive activation such as increased startle re-

sponses, can be evoked by a cue predicting an aversive event

(Andreatta & Pauli, 2015; Bradley, Moulder, & Lang, 2005). A

context which is learned to be associated with aversive con-

sequences without predicting the exact time of their occur-

rence elicits anxiety-related responses (Andreatta, Neueder,

Glotzbach-Schoon, Muhlberger, & Pauli, 2017; Baas, Nugent,

Lissek, Pine, & Grillon, 2004; Grillon, Baas, Lissek, Smith, &

Milstein, 2004) and avoidance of this context (Glotzbach,

Ewald, Andreatta, Pauli, & Muhlberger, 2012; Grillon, Baas,

Cornwell, & Johnson, 2006).

The neuronal systems underlying phasic fear and tonic

anxiety are seen as two overlapping, but nevertheless distinct

networks (Davis et al., 2010). While the central nucleus of the

amygdala plays a central role in predictable threat (Boll,

Gamer, Gluth, Finsterbusch, & Buchel, 2013), the (bed nu-

cleus of the stria terminalis BNST) is involved in processing

unpredictable threat (Alvarez, Chen, Bodurka, Kaplan, &

Grillon, 2011). In addition, several studies reported an

increased and sustained activation of extended visual areas,

(anterior cingulate cortex ACC), insula and parietal cortex

(Alvarez et al., 2011; Andreatta et al., 2015; Hasler et al., 2007;

Maren et al., 2013). These latter areas are also part of the

attention network, and hence, these findings support the idea

of sustained anxiety being associated with enhanced

vigilance.

Fear and anxiety can also be classified according to

different stages of the defense cascade model (Blanchard,

Yudko, Rodgers, & Blanchard, 1993; Bradley, Codispoti,

Cuthbert, & Lang, 2001; Fanselow, 1994). In this model, the

proximity of the source of threat determines the type of

defensive behavior. During the pre-encounter phase, where

threat is looming, animals as well as humans show an ori-

enting response with an interplay of the parasympathetic

and sympathetic nervous system (Bradley et al., 2001;

Cacioppo & Berntson, 1994). Together with the post-

encounter phase, when threat is identified but not close,

this stage is associated with anxiety responses since the

threat has not clearly been identified yet. A shrinking dis-

tance to the source of threat, however, leads to mobilization

of the organism's resources, which is regulated by activity of

the sympathetic nervous system (Blanchard et al., 1993; L€ow,

Lang, Smith, & Bradley, 2008). This transition from post-

encounter to circa-strike phase is associated with the state

of fear, since the organism is directly threatened by the

predator. These two stages come along with different (heart

rate HR) response patterns, and thereby differentiate be-

tween fear and anxiety with a HR deceleration during the

post-encounter stage, and an HR acceleration during immi-

nent threat, i.e., circa-strike stage (Bradley et al., 2001; L€ow

et al., 2008). In addition, recent studies showed that heart

rate accelerates or decelerates not only as a function of

threat imminence but also as a function of behavioral op-

tions. For instance, it was shown that when participants

could not actively avoid the threat, stronger heart rate

deceleration was observed (L€ow, Weymar, & Hamm, 2015;

Wendt, L€ow, Weymar, Lotze, & Hamm, 2017).

In order to directly compare anxiety and fear responses,

Schmitz and Grillon developed a paradigm, which manipu-

lates the predictability of threat, the so-called NPU-threat

test (2012). The NPU-threat test includes three different

conditions (neutral, predictable, and unpredictable threat).

In each, a neutral or threat context is determined by an in-

struction presented for a long duration (>30 sec). During

these periods, a cue occurs with a relatively short duration

(i.e., 8 sec) repeatedly. In the (neutral condition N), no

aversive stimulus (e.g., electric shock) is presented, while in

the (predictable condition P) the aversive stimulus only oc-

curs during the presentation of the respective cue. In the

(unpredictable condition U), the aversive stimulus occurs

unpredictably within the context-only parts of the trial.

Startle amplitudes measured during the cue presentation

revealed a fear-potentiated startle during the cue in the P-

condition compared to all other conditions. For the context-

only parts (i.e., periods without any cue presentations),

startle amplitudes were increased during the (unpredictable

context U-context) compared to the two other conditions,

which is defined as anxiety-potentiated startle (Grillon et al.,

2004). As indicated by this startle modulation to predictable

and unpredictable threat, the motivational significance

changes for the context and cue stimuli and thereby pre-

sumably facilitates sensory processing of these stimuli

(Lang, Bradley, & Cuthbert, 1990). However, these results do

not allow any conclusions yet whether anxiety and fear

differently affect sensory processing and whether any tem-

poral dynamics in this processing can differentiate between

fear and anxiety.

The (steady-state visually evoked potential ssVEP) as an

oscillatory field potential measured by (electro-encephalo-

gram EEG), perfectly meets these requirements as

c o r t e x 1 0 6 ( 2 0 1 8 ) 1 2 0e1 3 1122

continuous measure of sensory processing (Norcia,

Appelbaum, Ales, Cotterau, & Rossion, 2015; Vialatte,

Maurice, Dauwels, & Cichocki, 2010; Wieser, Miskovic, &

Keil, 2016a). The ssVEP signal is elicited by the driving fre-

quency of a flickering stimulus and reflects activity of the

primary and extended visual cortex as well contributions of

higher order cortices (Di Russo et al., 2007; Muller, Teder, &

Hillyard, 1997; Wieser & Keil, 2011). As the frequency of the

ssVEP signal is known, it allows to obtain a time-frequency

sensitive measure of visuocortical activity in response to the

respective stimulus (Muller, Teder-Salejarvi, & Hillyard,

1998; Norcia et al., 2015). Augmented ssVEP amplitudes

reveal increased attentional resources towards a stimulus

(Keil, Gruber, & Muller, 2001). Furthermore, an increased

motivational significance of a stimulus evokes heightened

oscillatory cortical responses (e.g., Keil et al., 2010; Keil

et al., 2003; Wieser, McTeague, & Keil, 2011, 2012). Simi-

larly, a stimulus which was learned to be of significance as

during fear conditioning, elicits higher ssVEP amplitudes in

visual cortices (Miskovic & Keil, 2012, 2013b; Moratti, Keil, &

Miller, 2006; Wieser et al., 2008, 2014a, 2014b). This does not

only apply to cue-fear conditioning, but could be recently

shown in context-anxiety conditioning, with sustained

ssVEP amplitudes during contextual anxiety. Importantly,

such higher motivational attention was maintained

throughout the duration of the context (Kastner, Pauli, &

Wieser, 2015). Moreover, the ssVEP approach is perfectly

suited to address questions regarding the perception of so-

cial stimuli within contextual information (Wieser & Keil,

2014; Wieser et al., 2016a).

In a previous study, we successfully adapted the NPU-

threat test in order to compare attention processes during

predictable and unpredictable threat via ssVEPs using

simple Gabor patches and surrounding arrays of geometric

shapes (Wieser, Reicherts, Juravle, & Von Leupoldt, 2016b).

In brief, these results showed that attention and sensory

processing are selectively enhanced for the predictable cue

and the onset of the unpredictable context. The present

study aimed at extending these findings by using more

complex, photorealistic stimuli. The continuous visuocort-

ical changes in response to predictable and unpredictable

threat were assessed in order to observe the high-

resolution changes in time. It was assumed that height-

ened ssVEP amplitudes elicited by the context of the un-

predictable condition should exceed both the activation

during the neutral and predictable contexts and be main-

tained throughout the context. During the cue pre-

sentations, the predictable cue is expected to capture

attentional resources compared to the neutral and unpre-

dictable cue, as could be observed in previous studies on

cue conditioning (Miskovic & Keil, 2012, 2013b; Moratti

et al., 2006; Wieser et al., 2008, 2014a, 2014b) as well as in

our previous NPU-threat study (Wieser et al., 2016b).

Moreover, we expected that this differential response

would be also reflected in cardiovascular responses, with

HR deceleration in the unpredictable context as an index of

anxiety, and HR acceleration during the cue in the pre-

dictable condition as an index of defensive mobilization

(Bradley et al., 2001; L€ow et al., 2008).

2. Methods

2.1. Participants

Participants were 45 female students, who were recruited via

a local online platform. Exclusion criteriawere any psychiatric

or neurological disorders (self-report). Participants received 9

Euros as study reimbursement. Due to a failed differentiation

between the three contexts, three participants had to be

excluded from analysis. Resulting participants' age ranged

from 18 to 34 (M ¼ 23.14, SD ¼ 3.35). Mean trait and state

anxiety scores as measured with the Spielberger State-Trait

Anxiety Inventory were 35.19 (SD ¼ 8.49) and 33.95

(SD ¼ 7.52), respectively.

2.2. Stimulus materials and procedure

Three context stimuli and three cue stimuli were used. The

context stimuli were screenshots from virtual offices created

with the Source Engine from the Valve Corporation (Belle-

vue, USA), used in previous studies (Andreatta et al., 2015;

Glotzbach et al., 2012). The office differed in layout and

furniture, but pictures were balanced for luminance and

complexity by controlling the luminance values returned by

the Image Manipulation Program GIMP 2.8.14 (159, 160, 160,

for the context stimuli) and a quantitative measure of en-

tropy, calculated with the respective MATLAB function (6.68,

6.80 and 7.20, respectively). The cue stimuli were taken from

the (Radboud Faces database RaFD; Langner et al., 2010).

Three female faces with frontal orientation and neutral ex-

pressions were chosen (Rafd090_02, Rafd090_32, Rafd090_58).

As the RaFD pictures only depict faces, the facial stimuli and

a neutral female body (also created with the Source Engine)

were merged in order to create a more realistic situation

with a person standing in an office room (see Fig. 1a). Stimuli

subtended a horizontal visual angle of 33.38� and 2.86� and a

vertical visual angle of 23.90� and 5.32� for context stimuli

and cue stimuli, respectively. In order to evoke ssVEPs,

stimuli were presented in flickering mode at either 12 or

15 Hz. If the context-stimuli flickered at 12 Hz, the cue

stimuli were superimposed at the other frequency (i.e.,

15 Hz), and vice versa (counterbalanced across participants).

The stimuli were presented on a computer screen with a

refresh rate of 60 Hz, which was positioned at a distance of

100 cm from the participants' eyes. The combination of

context and cue stimuli and the assignment to the three

conditions was counterbalanced across participants. A loud

startle probe (white noise), presented via headphones at

95 dB for 500 msec served as aversive unconditioned stim-

ulus (US).

All stimuli were rated before and after the experimental

session for valence, arousal, and anxiety on 9-point visual

analog scales (VAS). These ranged from 1 “very unpleasant”,

“very calm” and “not threatening” to 9 “very pleasant”, “very

arousing” and “very threatening”, respectively. Context and

cue stimuli were rated separately, as well as cue-context

compounds with the cue in the foreground, similar to how

they were appearing in the experimental session.

Fig. 1 e Stimuli and design. (A): Examples of the context stimuli and cues within contexts. (B): Schematic design of the NPU-

threat test, adapted from Schmitz and Grillon (2012). (N ¼ neutral condition, P ¼ predictable condition, U ¼ unpredictable

condition). Each trial lasted for 60 sec and was followed by an inter-trial interval (ITI) between 1750 and 2500 msec. In the N

condition, no US was presented. During the P condition, an US occurred at the last 500 msec of one or two cues, and one or

two US were presented randomly in the context-only phases during U. One possible run is depicted on top, another order

started with the UNPNPNU block, followed by PNUNUNP.

c o r t e x 1 0 6 ( 2 0 1 8 ) 1 2 0e1 3 1 123

After the first ratings, participants received instructions

regarding the three different conditions. Participants were

explained that they were about to learn how to predict the

occurrence of the aversive noise and under which condition

it would not be predictable. They were informed about the

three conditions called “no noise”, “predictable noise” and

“unpredictable noise” and whether to expect the loud noise.

Then the corresponding picture of the room and the asso-

ciated person were displayed. Afterwards, participants were

asked to assign each context and cue to the right condition

in order to check their comprehension of instructions. A

similar test was performed after the experimental session.

As mentioned above, three participants had to be excluded

afterwards, as they mixed up the conditions.

The design of the study followed the procedure suggested

by Schmitz and Grillon (2012) for the NPU paradigm (Fig. 1b).

One trial comprised the presentation of one context for 60 sec

and three occurrences of the associated cue for 4 sec each,with

context-only intervals of 11e13 sec. An instruction screen

preceded each trial for 2,000 msec containing the information

about the following condition. An inter-trial interval showing a

blank screen (ITI) of 1,750e2,000 msec followed each trial. The

aversive noise occurred once or twice within each trial for the

predictable and unpredictable condition (50% 1x occurrence,

50% 2x occurrences within a trial), but was never presented in

the neutral condition. In the predictable condition, the aver-

sive noise was presented during the last 500 msec of cue pre-

sentation, with the cue and aversive noise co-terminating. In

the unpredictable condition, the aversive noise was presented

randomly in the context-only intervals. The order of trials was

determined following the protocol of Schmitz and Grillon

(2012) with blocks of trials with PNUNUNP or UNPNPNU

order. The experimental session consisted of four of these

blocks with alternating order. After half of the session, par-

ticipants had a short break. Altogether, 8 predictable, 8 un-

predictable and 12 neutral conditions were presented.

2.3. Data recording and analysis

2.3.1. Rating dataThe ratings of valence, arousal, and anxiety for the contexts,

cues, and for the cue-context compounds before and after the

task were analyzed with repeated-measures ANOVAs

including the within-subject factor condition (N, P, U) (Grillon

et al., 2004).

2.3.2. EEG dataEEG was continuously recorded with a HydroCel Geodesic

Sensor Net (Electrical Geodesics, Inc., Eugene, OR) with 128

sensors (sampling rate of 250 Hz and online band pass filter of

.1e100 Hz). The vertex electrode (Cz) served as online-

reference electrode. The impedance of each sensor was kept

below 50 kU, as recommended for the Electrical Geodesics

high-impedance amplifiers.

The raw ssVEP signal averaged across all participants and

conditions for a representative electrode (sensor 75, Oz)

together with the Fast Fourier Transformation and the

resulting frequency topographies of this ssVEP signal are

shown in Fig. 2. EEG data preprocessing was performed with

the Matlab based software emegs 2.4 (Peyk, De Cesarei, &

Jungh€ofer, 2011) for contexts and cues separately. For the

contexts, datawas low-pass filtered at 40 Hz and event-related

epochswere extracted from 600msec before until 60,600msec

after context stimulus onset. Due to the unusual long segment

length of 60 sec for the context stimuli, no artifact detection

was performed, as it would have resulted in a rejection of

more than 50% of the trials due to eye movements and eye

blinks. Thus, all 8 trials per condition were used for this

analysis. Due to the robustness of the ssVEP signal, this

approach seems feasible and has been used in other para-

digms (Rossion & Boremanse, 2011). In order to assess cortical

activation during cue processing separately, data was low-

pass filtered at 40 Hz and discrete epochs were extracted,

Fig. 2 e Grand mean ssVEP amplitudes across all participants and all conditions and the respective frequency spectra (Fast

Fourier Transformation). The ssVEP signal was averaged across all participants and all conditions for the first 4 sec of each

trial. The raw signal is shown (A) and the respective Fast Fourier Transformation of this ssVEP signal (B) for sensor 75 (Oz).

The topographies of the two driving frequencies of 12 and 15 Hz are also given (C). The electrodes used for spatial averaging

of the ssVEP amplitudes are given in (D). Dark grey are electrodes used for cue analysis, light grey are electrodes used for

context analysis.

c o r t e x 1 0 6 ( 2 0 1 8 ) 1 2 0e1 3 1124

from 600 msec before to 4,600 msec after cue onset. Artifact

detection on the preprocessed data was performed in a two-

step method as described by (Jungh€ofer, Elbert, Tucker, &

Rockstroh, 2000). All epochs were averaged separately for

every condition, and contexts and cues. A Hilbert trans-

formation using a MATLAB script (see Miskovic & Keil, 2013a,

for a detailed description) was performed on the raw signal in

order to gain time varying ssVEP amplitudes in the respective

stimulation frequencies.

Statistical analysis was performed using IBM SPSS Statis-

tics 22. For context-related responses of the whole trial length

(60 sec), a 3 � 5 repeated measures ANOVA including the

within-subject factors condition (N, P, U) and five time win-

dows (�a 12,000 msec) was calculated, using the ssVEP signal

averages across two occipito-lateral sensor clusters with 5

sensors each (Fig. 2d). The clusters were chosen according to a

previous study from our lab (Kastner et al., 2015), in which

context effects were found to be maximal. In a second step,

four segments with equal length of 7,000 msec were cut out of

the context-only interval between the cue presentations

(0e7,000, 17,000e24,000, 34,000e41,000, 51,000e58,000 msec

from context onset) to exclude any possible effects driven by

the cue presentation. A repeated-measures ANOVA including

the within-subject factors laterality (left vs right electrodes

cluster), condition (N, P, U) and time (4 segments) was

conducted.

For the cue-related responses, repeated-measures

ANOVAs including the factors condition (N, P, U) and time

window (4 time windows of 1,000 msec) were conducted for

the ssVEP signal elicited by the cue and separately for the

ssVEP signal elicited by the context during cue presentation. A

medial-occipital sensor cluster with six sensors around Oz

(Fig. 2d) was used for extracting the cue-evoked ssVEP signal

(Kastner et al., 2015; Wieser et al., 2016b).

2.3.3. Heart rate (HR)HR was assessed by recording an electrocardiogram with two

adhesive Ag/AgCl electrodes. To calculate continuous heart

rate, R-spikes in the electrocardiogram were counted semi-

automatically using VisionAnalyzer2.0 software and the

inter-beat-interval (IBI) was determined. Then, continuous HR

estimateswere determined per sampling point (sampling rate:

1000 Hz). Due to artifacts which made R-peak scoring impos-

sible, only 39 participants could be included in the analysis for

context effects and 36 regarding cue effects. Segments of 4 sec

during cue presentation were averaged for each condition as

were the segments of the first 10 sec of the trial during which

only the context was presented. We decided to analyze the

first 10 sec after context onset only, since this was the only

time where no cue was presented across all trials. Then,

change scores were calculated by subtracting a baseline of

1 sec before context or cue onset, respectively. HR response

c o r t e x 1 0 6 ( 2 0 1 8 ) 1 2 0e1 3 1 125

was scored as mean activity in consecutive 1 sec bins. A

repeated-measures ANOVA including the within-subject fac-

tors condition (N, P, U) and four time windows of 1 sec was

conducted for the cue segments, while the analysis for the

context effects included 10 time windows of 1 msec.

In all analyses, the alpha level was set at p < .05. If the

assumption of sphericity was violated, Greenhouse-Geisser

correction was applied and the Greenhouse-Geisser Epsilon

(GG-ε) is reported. Effect sizes are reported as partial eta (h2p).

3. Results

3.1. Ratings

As expected, the pre-ratings did not reveal any differences in

valence, arousal or anxiety for the context and cue stimuli (all

ps > .482). Regarding the post-ratings (Table 1), the analysis of

the contexts ratings revealed a main effect of condition [F (2,

82) ¼ 26.05, p ¼ <.001, h2p ¼ .39, GG-ε ¼ .84], with more negative

valence ratings for the predictable [t (41) ¼ 5.88, p < .001] and

unpredictable context [t (41) ¼ 5.60, p < .001] compared to the

neutral context, while they did not differ from each other

(p ¼ .348). Similarly, a main effect of condition was found for

arousal [F (2, 82) ¼ 37.67, p ¼ <.001, h2p ¼ .479] and anxiety [F (2,

82) ¼ 36.47, p ¼ <.001, h2p ¼ .47, GG-ε ¼ .87], revealing increased

arousal and anxiety ratings for the predictable [t (41) ¼ 7.37,

p < .001; t (41) ¼ 8.13, p < .001] and unpredictable context [t

(41) ¼ 6.96, p < .001; t (41) ¼ 6.65, p < .001] compared to neutral

context. Again, the predictable and unpredictable context did

not show differences in arousal (p ¼ .297) or anxiety ratings

(p ¼ .372).

Table 1 e Mean ratings (SD) Cue-Context compounds,Cues, and Contexts after the experimental session, in thethree experimental conditions (N: No Shock; P: PredictableShock; U: Unpredictable Shock).

Cue þ Context (n ¼ 42)

N P U

M SD M SD M SD

Valence 6.50 1.52 3.76 1.38 3.79 1.55

Arousal 1.90 1.34 4.64 2.30 4.64 2.30

Threat 1.50 1.13 4.07 2.30 3.83 2.37

Context only (n ¼ 42)

N P U

M SD M SD M SD

Valence 6.48 2.10 4.07 1.68 3.76 1.57

Arousal 2.02 1.33 4.50 2.07 4.83 2.21

Threat 1.52 1.23 3.83 1.96 4.12 2.14

Cue only (n ¼ 35)

N P U

M SD M SD M SD

Valence 6.17 1.52 3.76 1.37 3.79 1.55

Arousal 1.90 1.34 4.69 2.14 4.42 2.17

Threat 1.50 1.13 4.07 2.30 3.83 2.37

Note:Due to saving failure, the cue ratingswere only available for 35

participants.

Regarding the cues-in-contexts ratings, a similar pattern

evolved. A main effect of condition for the valence ratings [F

(2, 82)¼ 45.90, p < .001, h2p ¼ .53] showedmore negative valence

ratings for the predictable [t (41) ¼ 7.95, p < .001] and unpre-

dictable cue [t (41)¼ 7.88, p < .001] compared to the neutral cue

but no difference between the two former (p ¼ .936). Similarly,

amain effect for arousal [F (2, 82)¼ 38.87, p < .001, h2p ¼ .49] and

anxiety ratings [F (2, 82) ¼ 38.40, p < .001, h2p ¼ .48] revealed

increased arousal and anxiety for the predictable [arousal: t

(41) ¼ 7.33, p < .001; anxiety: t (41) ¼ 7.66, p < .001] and un-

predictable cue [arousal: t (41) ¼ 7.05, p < .001; anxiety: t

(41) ¼ 6.51, p < .001] compared to the neutral cue; the pre-

dictable and the unpredictable conditions did not differ in

arousal nor anxiety (all ps > .387).

The cue-ratings (only available from 35 participants due to

saving errors), revealed a similar pattern. A main effect of

condition for the valence ratings [F (2, 68) ¼ 27.55, p < .001,

h2p ¼ .45] showed more negative valence ratings for the pre-

dictable [t (34) ¼ 6.05, p < .001] and unpredictable cue [t

(34) ¼ 5.86, p < .001] compared to the neutral cue, but no dif-

ference between the former two (p ¼ .299). Similarly, a main

effect for arousal [F (2, 68) ¼ 27.92, p < .001, h2p ¼ .45] and

anxiety ratings [F (2, 68) ¼ 30.93, p < .001, h2p ¼ .48] revealed

increased arousal and anxiety for the predictable [arousal: t

(34) ¼ 5.90, p < .001; anxiety: t (34) ¼ 6.71, p < .001] and un-

predictable cue [arousal: t (34) ¼ 5.75, p < .001; anxiety: t

(41) ¼ 5.54, p < .001] compared to the neutral cue; again, no

difference was found between predictable and unpredictable

cues.

3.2. EEG data

3.2.1. Cortical response to contextsThe analysis of the whole 60 sec context segments (including

cue presentations) showed a main effect of condition [F (2,

82) ¼ 9.64, p < .001, h2p ¼ .42] indicating increased ssVEP am-

plitudes during the predictable [t (41) ¼ 7.02, p < .001] and

unpredictable [t (41) ¼ 5.65, p < .001] compared to the neutral

context, while there was no significant difference between

predictable and unpredictable contexts (p ¼ .633) (see Fig. 3). A

main effect of time [F (4, 164) ¼ 53.51, p < .001, h2p ¼ .566, GG-

ε ¼ .30] showed a significant decrease of the ssVEP signal from

the first (0e12 sec) to the second time window (12e24 sec)

across all three conditions [t (41) ¼ 7.49, p < .001] and to all

other following time windows (ts > 6.62, ps < .001). As later-

ality did not return any significant main effect nor interaction

(all ps > .520), this factor was disregarded in any further ana-

lyses and the two sensor clusters were combined.

The analysis of the epochs without any cue presentation

revealed a main effect of condition [F (2, 82) ¼ 23.04, p < .001,

h2p ¼ .36] and phase [F (3, 123) ¼ 54.93, p < .001, hp2 ¼ .57, GG-

ε ¼ .43] (Fig. 4). Furthermore, a significant interaction of con-

dition and time was found [F (6, 246) ¼ 2.26, p ¼ .039, h2p ¼ .05,

GG-ε ¼ .62]. The conditions did not show any differential

cortical activation in the first time interval (0e7 sec) [F (2,

82) ¼ 2.18, p ¼ .119, h2p ¼ .051], while a significant difference

between conditions was found in the second (17e24 sec) [F (2,

82) ¼ 29.53, p < .001, h2p ¼ .42] and third time interval

(34e41 sec) [F (2, 82) ¼ 22.25, p < .001, h2p ¼ .35], with increased

cortical activation for the predictable [second: t (41) ¼ 6.00,

Fig. 3 e Grand average scalp topographies of ssVEP amplitudes in response to the neutral, predicable, and unpredictable

contexts, averaged across the whole trial duration (60,000 msec).

c o r t e x 1 0 6 ( 2 0 1 8 ) 1 2 0e1 3 1126

p < .001; third: t (41) ¼ 5.88, p < .001] and the unpredictable

condition [second: t (41) ¼ 7.41, p < .001; third: t (41) ¼ 5.49,

p < .001] compared to the neutral context. The predictable and

unpredictable context did not differ from each other during

these first three time intervals (all ps > 269). Interestingly, in

the fourth time interval (51e58 sec) the main effect of condi-

tion [F (2, 82) ¼ 28.17, p < .001, h2p ¼ .407] indicated besides

significant higher cortical activity during both the predictable

[t (41) ¼ 5.26, p < .001] and unpredictable context [t (41) ¼ 6.88,

p < .001] compared to the neutral context, a heightened ssVEP

amplitude in response to the unpredictable compared to the

predictable context [t (41) ¼ 2.73, p ¼ .009].

3.2.2. Cortical response to cuesAnalysis of the cue segments (4,000msec) revealedmarginally

significant different cortical responses [F (2, 82) ¼ 2.86,

p¼ .063, h2p ¼ .065]. As thismain effect was predicted by amain

hypothesis, planned paired sampled t-tests were performed.

These showed an increased ssVEP signal elicited by the pre-

dictable compared to the neutral cue [t (41) ¼ 2.17, p ¼ .036],

while neither the comparison between neutral and unpre-

dictable cue nor between predictable and unpredictable cue

returned significant (all ps > .196).

A similar analysis was performed for the ssVEP signal eli-

cited by the contexts during cue presentation. This revealed a

main effect of condition [F (2, 82) ¼ 3.68, p ¼ .029, h2p ¼ .08].

During cue presentation, the predictable context elicited

increased ssVEP amplitudes as compared to the neutral

context [t (41) ¼ 3.21, p ¼ .003], while in the unpredictable cue

condition, the context did not show any differential cortical

activation compared to the neutral and the predictable

context (all ps > .153) (see Fig. 5).

3.3. Heart rate

3.3.1. Response to contextThe heart rate deceleration for the predictable and unpre-

dictable context compared to the neutral context is depicted

in Fig. 6A. The repeated measures ANOVA returned a signifi-

cant main effect of condition [F (2, 76) ¼ 5.62, GG-ε ¼ .91,

p ¼ .005, hp2 ¼ .13] and time [F (9, 342) ¼ 14.37, GG-ε ¼ .35,

p< .001, h2p ¼ .27], and a significant interaction condition x time

[F (18, 684) ¼ 2.80, GG-ε ¼ .36, p ¼ .010, h2p ¼ .07]. Except for the

first two time intervals, both predictable and unpredictable

contexts elicited a significantly larger HR decrease compared

to the neutral context (ts > 2.20, ps < .034). The predictable and

unpredictable contexts did not elicit any differential heart rate

in all time windows (all ps > .104).

3.3.2. Response to cuesThe repeated measures ANOVA concerning the HR changes

during cue presentation returned a significant main effect of

condition [F (2, 70) ¼ 15.01, GG-ε ¼ .78, p < .001, h2p ¼ .30] and

time [F (3, 105)¼ 26.07,GG-ε¼ .52, p < .001, h2p ¼ .427], as well as

a significant interaction between those factors [F (6,

210)¼ 10.44,GG-ε¼ .49, p< .001, h2p ¼ .23]. As seen in Fig. 6B, the

predictable cue elicited a significant HR acceleration between

1 and 4 sec compared to the neutral (ts > .2.51, ps < .017) and

the unpredictable condition (ts > .2.98, ps < .005).

4. Discussion

The present study investigated the differential influence of

predictability of threat on attention processes, with the goal to

demarcate the state of fear as a phasic response with selective

attention during imminent threat and the state of anxiety as a

sustained response with sustained vigilance. A strength of

this study is the use of steady-state visual evoked potentials

within the NPU-test allowing a continuous assessment of

attention-modulated electrocortical responses, thereby taking

into account the temporal distinctiveness between fear and

anxiety as one key characteristic differentiating between

these two affective states.

As expected and in line with previous findings (e.g.,

Grillon et al., 2004; Wieser et al., 2016b), the contexts asso-

ciated with the aversive event (i.e., P- and U-context) were

rated more arousing and anxiogenic as well as more negative

than the control context (N-context). Contrary to previous

studies, we found no differences between predictable and

unpredictable contexts (Grillon et al., 2009). A possible

explanation for this lack of differentiation may be that in our

paradigm, the participants actually had to learn and

remember in which context they were (different office

Fig. 4 e Grand average scalp topographies of ssVEP amplitudes in response to the neutral, predicable, and unpredictable

contexts. The ssVEP signal was averaged and analyzed separately for four selected time intervals (7 sec) of the trial during

which no cues were presented (0e7 sec, 17e24 sec, 34e41 sec, and 51e58 sec) after context onset.

c o r t e x 1 0 6 ( 2 0 1 8 ) 1 2 0e1 3 1 127

rooms), whereas in the studies by Grillon and colleagues

participants continuously received written information about

the specific context on the screen. Possibly, the differentia-

tion between U and P contexts in our protocol might have

been more complex, which consequently led to a less clear

discrimination between the two conditions. The ratings for

the cues embedded in the respective context revealed a

similar pattern. The cues presented in both the unpredictable

and predictable condition were rated as more unpleasant,

more arousing and more anxiogenic compared to the cue

presented in the neutral context. The same was found for the

ratings of predictable and unpredictable cues alone (Alvarez

et al., 2011; Haaker, Lonsdorf, Thanellou, & Kalisch, 2013).

Based on these ratings we conclude that all stimuli

Fig. 5 e Grand mean ssVEP amplitudes in response the cues, and in response to the contexts during cue presentation. The

upper row shows the grand mean scalp topographies of the ssVEP amplitudes elicited by the cues within each condition,

averaged across 4 sec. The lower row shows the mean ssVEP amplitudes elicited by the contexts during the cue

presentation averaged across 4 sec.

c o r t e x 1 0 6 ( 2 0 1 8 ) 1 2 0e1 3 1128

associated with the US, even when not in an explicit manner,

elicited strong negative affective responses.

During both the predictable and unpredictable condition,

increased cortical activation in visual cortex was observed

throughout the whole 60 sec of context presentation. Inter-

estingly, when only considering time windows without cue

presentation, the ssVEP amplitudes showed a differential

activation in response to the unpredictable compared to the

predictable threat. A facilitated sensory processing of the

Fig. 6 e Averaged heart rate responses for the neutral, predictabl

a -1 sec pre-stimulus baseline) for the first 10 sec of each trial d

predictable threat elicited larger HR decelerations compared to

(Asterisks indicate significant effects at p < .05, for the post-hoc

a -1sec pre-stimulus baseline) during the 4 sec of cue presenta

acceleration between 1 and 4 sec compared to the neutral and

effects at p < .05, for the post-hoc comparisons P vs N and P vs

unpredictable context only evolved in the last window of the

trial. This means that during the predictable context, sus-

tained attentional resources were reduced as the threat of an

US occurrence dissolved with the last cue presentation. In

contrast, the threat was present continuously in the unpre-

dictable context and therefore increased attentional resources

were continuously maintained. This is in contrast to our

previous study, which only found effects for the unpredictable

context restricted to the very onset of the condition (Wieser

e, and unpredictable condition. (A) Heart rate changes (from

uring context presentation; Unpredictable as well as

the neutral condition in all but the first two time intervals.

comparisons P vs N and U vs N). (B) Heart rate changes (from

tions. The predictable cue elicited a significant larger HR

the unpredictable condition (Asterisks indicate significant

U).

c o r t e x 1 0 6 ( 2 0 1 8 ) 1 2 0e1 3 1 129

et al., 2016b). Possibly, the more complex stimuli used here

required more time to be identified and hence more atten-

tional resources at a later time window as compared to the

previous study, which only used arrays of very simple

geometrical shapes as contexts. We conclude that the present

findings strongly support the notion of anxiety as a state of

sustained, longer lasting response with continuous height-

ened vigilance during unpredictable threat.

As expected, an increased cortical activation was observed

in response to the predictable cue compared to the neutral

cue, which corroborates findings of enhanced startle re-

sponses (Grillon et al., 2009) and ssVEPs (Wieser et al., 2016b)

for the cues in the predictable condition. Interestingly, the

context surrounding the predictable cue during its 4 sec pre-

sentation elicited also increased ssVEP amplitudes. It seems

that not only the cue for the aversive event, but also the sur-

rounding context gains motivational significance and threat

value. This finding corroborates results from our previous

study (Wieser et al., 2016b). Similar results were observed for

visual contexts, which contained a fearful face; the aversive

cues led to an amplification of the contexts, indicated by

enhanced ssVEP amplitudes (Wieser & Keil, 2014). Together,

these results support the notion that during specific threat,

processing of surrounding contexts is also enhanced.

Our results suggest that unpredictable and predictable

threat similarly draw attentional resources and lead to

increased electrocortical activation when no crucial cue is

present. One explanation could be that the time between cues,

when only the context is present, serves for both conditions as

a time of uncertainty about the occurrence of the aversive

event. While in the unpredictable condition, the occurrence is

not predictable in any way, even in the predicable condition

the context signals some uncertainty about when the actual

cue, and consequently the aversive event appear.

Both the predictable and unpredictable condition were

associatedwith pronounced decelerations in heart rate during

the first 10 sec of the trials, in which only the context was

present. Such HR decelerations (often labeled fear brady-

cardia) are known from the animal literature to reflect

an orienting response during the post-encounter phase of

the defense cascade model and indicate hyper-alertness

(Fanselow, 1994; L€ow et al., 2008, 2015). As soon as the cue

appeared, HR of the participants significantly increased in the

predictable condition only compared to both the neutral and

the unpredictable condition. This response resembles the

circa-strike phase, in which the threat becomes proximal and

physiological arousal increases as indicated by heart rate ac-

celeration and increased skin conductance (Fanselow, 1994;

Lang, Davis, & €Ohman, 2000; Richter et al., 2012). As soon as

the threat e that is the cue in the predictable condition e is

present, heart rate significantly increases as index for sym-

pathetic activation necessary for the defensive response.

The defense cascade model is ideally suited to explain the

similarities and differences in the electrocortical and physio-

logical activation between predictable and unpredictable

threat observed in this study. During the post-encounter

phase, a general facilitated sensory processing reflected by

increased electrocortical activation in occipital areas (unspe-

cific vigilance) has to be expected; as the anticipated threat is

not as imminent yet, the organism needs to evaluate the

situation in order to prepare any suitable responses. The

observed increased cortical responses during the unpredict-

able condition, reflecting increasedmobilization of attentional

resources, are in line with this explanation. Even during the

predictable condition, a preparation for the approaching cue

and subsequently also for the occurrence of the aversive event

has to take place, as seen with increased ssVEP amplitudes in

response to the context of the predictable condition. However,

a differentiation between these two conditions is observed

during the cue presentation. The occurrence of the predictable

cue can be classified as the circa-strike phase during defense.

The proximity of the unpleasant event leads to increased

physiological activations in order to prepare immediate re-

sponses together with selective attention for the threat-

specific cue. These processes are reflected in our findings of

increased ssVEP amplitudes aswell as a heart rateacceleration

in response to the cue in the predictable condition.

Some limitations of the present study should be

mentioned. First, due to practical reasons we only included

female participants, which naturally hampers the generaliz-

ability of our findings. Second, in order to analyze pure

context timewindows,we had to remove the cue periods from

both the P and the U condition. This asymmetrically removes

US-presentations from the P but not the U-condition. Thus,

the U-condition in that set of analyses may be contaminated

with time periods of presumably increased arousal due to the

noise burst presentations, while these periods are removed

from the P-condition. As a consequence, the comparison of N

and U in those analyses does not necessarily inform about the

difference between fear and anxiety. However, if arousal

would be the inly driving factor, we should have seen this

effect also at earlier time windows.

In sum, the present study aimed at differentiating the

attention mechanisms and defensive responding during fear

and anxiety using ssVEP, HR, and ratings data. Increased

activation in occipital areas reflecting facilitated sensory gain

was found for both predictable and unpredictable threat

conditions. Without the availability of a cue, unpredictable

and predictable threat showed similar cortical activation, but

importantly the activation was longer lasting in the unpre-

dictable threat condition, which is in line with the predictions

of the defense cascade model. Heightened vigilance during

post-encounter and selective attention during circa-strike

were disentangled using electrocortical measures and

distinctive heart rate responses. Overall, this study confirms

that anxiety and fear are indeed different affective states with

distinctive patterns of attention and physiological reactivity.

In a next step, research should consider the interplay of

cue and context conditioning, perhaps including extinction

processes, to advance our understanding of attentional

mechanisms in anxiety disorders and treatment processes.

For example, it should be interesting to see if anxiety sensi-

tivity as a potential risk factor for panic disorder is also

associated with selective enhanced visuocortical responding

during the unpredictable threat condition, which was

observed recently (Stevens, Weinberg, Nelson, Meissel, &

Shankman, 2018). In the same vein, possible plasticity of

these neurocognitive mechanisms in response to treatments

targeting unpredictability of threat (Gorka et al., 2017) may

inform us about successful treatment components. With

c o r t e x 1 0 6 ( 2 0 1 8 ) 1 2 0e1 3 1130

regard to extinction processes, it will be important to check if

participants with risk factors for anxiety disorders (e.g.,

enhanced trait anxiety, anxiety sensitivity or a risk allele such

as of the genes coding for the 5-HT transporter [5HTT] and the

NPS receptor [NPSR1] show slower extinction learning for

contexts, cues, or both (Glotzbach-Schoon et al., 2013a;

Glotzbach-Schoon et al., 2013b). Last but not least, in terms

of the RDoC initiative it may also be important to test if this

paradigm may contribute to specifically refine underlying

processes contributing to the psychopathology of anxiety and

stress-related disorders (Lang, McTeague, & Bradley, 2016;

Lonsdorf & Richter, 2017).

Acknowledgements

This workwas supported by the German Research Foundation

(SFB/TRR-58, projects B01 and B05).

r e f e r e n c e s

Alvarez, R. P., Chen, G., Bodurka, J., Kaplan, R., & Grillon, C. (2011).Phasic and sustained fear in humans elicits distinct patternsof brain activity. Neuroimage, 55, 389e400.

American Psychiatric Association. (2013). The diagnostic andstatistical manual of mental disorders (5th ed.). Washington, DC:American Psychiatric Publishing.

Andreatta, M., Glotzbach-Schoon, E., Muhlberger, A.,Schulz, S. M., Wiemer, J., & Pauli, P. (2015). Initial andsustained brain responses to contextual conditioned anxietyin humans. Cortex, 63, 352e363.

Andreatta, M., Neueder, D., Glotzbach-Schoon, E., Muhlberger, A.,& Pauli, P. (2017). Effects of context preexposure and delayuntil anxiety retrieval on generalization of contextual anxiety.Learning and Memory, 24(1), 43e54.

Andreatta, M., & Pauli, P. (2015). Appetitive versus Aversiveconditioning in humans. Frontier Behaviour Neuroscience, 9, 128.

Baas, J. M., Nugent, M., Lissek, S., Pine, D. S., & Grillon, C. (2004).Fear conditioning in virtual reality contexts: A new tool for thestudy of anxiety. Biological Psychiatry, 55(11), 1056e1060.

Blanchard, R. J., Yudko, E. B., Rodgers, R. J., & Blanchard, D. C.(1993). Defense system psychopharmacology: An ethologicalapproach to the pharmacology of fear and anxiety. BehaviouralBrain Research, 58(1e2), 155e165.

Boll, S., Gamer, M., Gluth, S., Finsterbusch, J., & Buchel, C. (2013).Separate amygdala subregions signal surprise andpredictiveness during associative fear learning in humans. TheEuropean Journal of Neuroscience, 37(5), 758e767.

Bouton, M. E. (1994). Context, ambiguity, and classical-conditioning.Current Directions in Psychological Science, 3(2), 49e53.

Bradley, M. M., Codispoti, M., Cuthbert, B. N., & Lang, P. J. (2001).Emotion and motivation I: Defensive and appetitive reactionsin picture processing. Emotion, 1(3), 276e298.

Bradley, M. M., Moulder, B., & Lang, P. J. (2005). When good thingsgo Bad: The reflex physiology of defense. Psychological Science,16(6), 468e473.

Cacioppo, J. T., & Berntson, G. G. (1994). Relationship betweenattitudes and evaluative space - a critical-review, withemphasis on the separability of positive and negativesubstrates. Psychological Bulletin, 115(3), 401e423.

Davis, M., Walker, D. L., Miles, L., & Grillon, C. (2010). Phasic vssustained fear in rats and humans: Role of the extended

amygdala in fear vs anxiety. Neuropsychopharmacology, 35(1),105e135.

Di Russo, F., Pitzalis, S., Aprile, T., Spitoni, G., Patria, F., Stella, A.,et al. (2007). Spatiotemporal analysis of the cortical sources ofthe steady-state visual evoked potential. Human Brain Mapping,28(4), 323e334.

Fanselow, M. S. (1994). Neural organization of the defensivebehavior system responsible for fear. Psychonomic Bulletin andReview, 1(4), 429e438.

Glotzbach-Schoon, E., Andreatta, M., Reif, A., Ewald, H., Troger, C.,Baumann, C., et al. (2013a). Contextual fear conditioning invirtual reality is affected by 5HTTLPR and NPSR1polymorphisms: Effects on fear-potentiated startle. FrontierBehaviour Neuroscience, 7, 31.

Glotzbach-Schoon, E., Tadda, R., Andreatta, M., Troger, C.,Ewald, H., Grillon, C., et al. (2013b). Enhanced discriminationbetween threatening and safe contexts in high-anxiousindividuals. Biological Psychology, 93(1), 159e166.

Glotzbach, E., Ewald, H., Andreatta, M., Pauli, P., & Muhlberger, A.(2012). Contextual fear conditioning predicts subsequentavoidance behaviour in a virtual reality environment.Cognition and Emotion, 26(7), 1256e1272.

Gorka, S. M., Lieberman, L., Klumpp, H., Kinney, K. L.,Kennedy, A. E., Ajilore, O., et al. (2017). Reactivity tounpredictable threat as a treatment target for fear-basedanxiety disorders. Psychological Medicine, 47(14), 2450e2460.

Grillon, C., Baas, J. M., Cornwell, B., & Johnson, L. (2006). Contextconditioning and behavioral avoidance in a virtual realityenvironment: Effect of predictability. Biological Psychiatry,60(7), 752e759.

Grillon, C., Baas, J. P., Lissek, S., Smith, K., & Milstein, J. (2004).Anxious responses to predictable and unpredictable aversiveevents. Behavioral Neuroscience, 118(5), 916e924.

Grillon, C., Pine, D. S., Lissek, S., Rabin, S., Bonne, O., &Vythilingam, M. (2009). Increased anxiety during anticipationof unpredictable aversive stimuli in posttraumatic stressdisorder but not in generalized anxiety disorder. BiologicalPsychiatry, 66(1), 47e53.

Haaker, J., Lonsdorf, T. B., Thanellou, A., & Kalisch, R. (2013).Multimodal assessment of long-term memory recall andreinstatement in a combined cue and context fearconditioning and extinction paradigm in humans. PLoS One,8(10), e76179. https://doi.org/10.1371/journal.pone.0076179.

Hasler, G., Fromm, S., Alvarez, R. P., Luckenbaugh, D. A.,Drevets, W. C., & Grillon, C. (2007). Cerebral blood flow inimmediate and sustained anxiety. The Journal of Neuroscience,27(23), 6313.

Jungh€ofer, M., Elbert, T., Tucker, D. M., & Rockstroh, B. (2000).Statistical control of artifacts in dense array EEG/MEG studies.Psychophysiology, 37(4), 523e532.

Kastner, A. K., Pauli, P., & Wieser, M. J. (2015). Sustained attentionin context conditioning: Evidence from steady-state VEPs.International Journal of Psychophysiology, 98, 546e556.

Keil, A., Bradley, M. M., Ihssen, N., Heim, S., Vila, J., Guerra, P.,et al. (2010). Defensive engagement and perceptualenhancement. Neuropsychologia, 48(12), 3580e3584.

Keil, A., Gruber, T., & Muller, M. M. (2001). Functional correlates ofmacroscopic high-frequency brain activity in the humanvisual system. Neuroscience and Biobehavioral Reviews, 25(6),527e534.

Keil, A., Gruber, T., Muller, M. M., Moratti, S., Stolarova, M.,Bradley, M. M., et al. (2003). Early modulation of visualperception by emotional arousal: Evidence from steady-statevisual evoked brain potentials. Cognitive Affective and BehavioralNeuroscience, 3(3), 195e206.

Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (1990). Emotion,attention, and the startle reflex. Psychological Review, 97(3),377e395.

c o r t e x 1 0 6 ( 2 0 1 8 ) 1 2 0e1 3 1 131

Lang, P. J., Davis, M., & €Ohman, A. (2000). Fear and anxiety:Animal models and human cognitive psychophysiology.Journal of Affective Disorders, 61(3), 137e159.

Lang, P. J., McTeague, L. M., & Bradley, M. M. (2016). RDoC, DSM,and the reflex physiology of fear: A biodimensional analysis ofthe anxiety disorders spectrum. Psychophysiology, 53(3),336e347.

Langner, O., Dotsch, R., Bijlstra, G., Wigboldus, D. H. J.,Hawk, S. T., & van Knippenberg, A. (2010). Presentation andvalidation of the Radboud faces database. Cognition andEmotion, 24(8), 1377e1388.

Lonsdorf, T. B., & Richter, J. (2017). Challenges of fear conditioningresearch in the age of RDoC. Zeitschrift fur Psychologie, 225,189e199.

L€ow, A., Lang, P. J., Smith, J. C., & Bradley, M. M. (2008). Bothpredator and prey emotional arousal in threat and reward.Psychological Science, 19(9), 865e873.

L€ow, A., Weymar, M., & Hamm, A. O. (2015). When threat is near,get out of here: Dynamics of defensive behavior duringfreezing and active avoidance. Psychological Science, 26(11),1706e1716.

Maren, S., Phan, K. L., & Liberzon, I. (2013). The contextual brain:Implications for fear conditioning, extinction andpsychopathology. Nature Reviews Neuroscience, 14(6), 417e428.

Miskovic, V., & Keil, A. (2012). Acquired fears reflected in corticalsensory processing: A review of electrophysiological studies ofhumanclassical conditioning.Psychophysiology, 49(9), 1230e1241.

Miskovic, V., & Keil, A. (2013a). Perceiving threat in the face ofSafety: Excitation and inhibition of conditioned fear in humanvisual cortex. The Journal of Neuroscience, 33(1), 72e78.

Miskovic, V., & Keil, A. (2013b). Visuocortical changes duringdelay and trace aversive conditioning: Evidence from steady-state visual evoked potentials. Emotion, 13(3), 554.

Moratti, S., Keil, A., & Miller, G. A. (2006). Fear but not awarenesspredicts enhanced sensory processing in fear conditioning.Psychophysiology, 43(2), 216e226.

Muller, M. M., Teder-Salejarvi, W., & Hillyard, S. A. (1998). Thetime course of cortical facilitation during cued shifts of spatialattention. Nature Neuroscience, 1(7), 631e634.

Muller, M. M., Teder, W., & Hillyard, S. A. (1997).Magnetoencephalographic recording of steady-state visualevoked cortical activity. Brain Topography, 9(3), 163e168.

Norcia, A. M., Appelbaum, L. G., Ales, J., Cotterau, B. R., &Rossion, B. (2015). The steady-state visual evoked potential invision research: A review. Journal of Vision, 15.

Perusini, J. N., & Fanselow, M. S. (2015). Neurobehavioralperspectives on the distinction between fear and anxiety.Learning and Memory, 22(9), 417e425.

Peyk, P., De Cesarei, A., & Jungh€ofer, M. (2011). Electro magnetoencephalography software: Overview and integration withother EEG/MEG toolboxes. Computational Intelligence andNeuroscience, 2011. Article ID 861705.

Richter, J., Hamm, A. O., Pane-Farre, C. A., Gerlach, A. L.,Gloster, A. T., Wittchen, H. U., et al. (2012). Dynamics of

defensive reactivity in patients with panic disorder andAgoraphobia: Implications for the etiology of panic disorder.Biological Psychiatry, 72(6), 512e520.

Rossion, B., & Boremanse, A. (2011). Robust sensitivity to facialidentity in the righthumanoccipito-temporal cortexas revealedby steady-state visual-evoked potentials. Journal of Vision, 11(2).

Schmitz, A., & Grillon, C. (2012). Assessing fear and anxiety inhumans using the threat of predictable and unpredictableaversive events (the NPU-threat test). Nature Protocols, 7(3),527e532.

Stevens, E. S., Weinberg, A., Nelson, B. D., Meissel, E. E. E., &Shankman, S. A. (2018). The effect of panic disorder versusanxiety sensitivity on event-related potentials duringanticipation of threat. Journal of Anxiety Disorders, 54, 1e10.

Tovote, P., Fadok, J. P., & Luthi, A. (2015). Neuronal circuits for fearand anxiety. Nature Reviews Neuroscience, 16(6), 317e331.

Vialatte, F. B., Maurice, M., Dauwels, J., & Cichocki, A. (2010).Steady-state visually evoked potentials: Focus on essentialparadigms and future perspectives. Progress in Neurobiology,90(4), 418e438.

Wendt, J., L€ow, A., Weymar, M., Lotze, M., & Hamm, A. O. (2017).Active avoidance and attentive freezing in the face ofapproaching threat. Neuroimage, 158, 196e204.

Wieser, M. J., Flaisch, T., & Pauli, P. (2014a). Raised middle-finger:Electrocortical correlates of social conditioning withnonverbal affective gestures. Plos One, 9(7), e102937.

Wieser, M. J., & Keil, A. (2011). Temporal trade-off effects insustained Attention: Dynamics in visual cortex predict thetarget detection performance during distraction. The Journal ofNeuroscience, 31(21), 7784.

Wieser, M. J., & Keil, A. (2014). Fearful faces heighten the corticalrepresentation of contextual threat. Neuroimage, 86, 317e325.

Wieser, M. J., McTeague, L. M., & Keil, A. (2011). Sustainedpreferential processing of social threat Cues: Bias withoutcompetition? Journal of Cognitive Neuroscience, 23(8), 1973e1986.

Wieser, M. J., McTeague, L. M., & Keil, A. (2012). Competitioneffects of threatening faces in social anxiety. Emotion, 12,1050e1060.

Wieser, M. J., Miskovic, V., & Keil, A. (2016a). Steady-state visualevoked potentials as a research tool in social affectiveneuroscience. Psychophysiology, 53(12).

Wieser, M. J., Miskovic, V., Rausch, S., & Keil, A. (2014b).Different time course of visuocortical signal changes tofear-conditioned faces with direct or averted gaze: A ssVEPstudy with single-trial analysis. Neuropsychologia, 62,101e110.

Wieser, M. J., Pauli, P., Reicherts, P., Gerdes, A. B. M.,Hermann, M. J., & Muhlberger, A. (2008). Fear of publicspeaking and neurophysiological correlates of face processing.Cognitive Neuroscience Society, 15.

Wieser, M. J., Reicherts, P., Juravle, G., & Von Leupoldt, A. (2016b).Attention mechanisms during predictable and unpredictablethreat e a steady-state visual evoked potential approach.Neuroimage, 139, 167e175.