ERP and behavioral evidence of increased sensory attenuation for fear-related action outcomes

Highlights

• Increased sensory attenuation for fearful compared to neutral faces.

• Neurophysiological attenuation for predicted fearful faces only.

• Enhanced subjective fear ratings for congruent fearful faces.

Abstract

Voluntary action selection entails the representation of the expected consequences of the action. Previous evidence suggests that accurate action–effect prediction modulates both ERP and behavioral markers of sensory processing—a phenomenon know as sensory attenuation. This may play an important role in monitoring the success or failure of our actions, or attributing agency. Nonetheless, the vast majority of studies in this domain focus on simplistic visual and auditory stimuli. Given that we rarely perform voluntary actions with the aim of generating such stimuli in social contexts, this provides little indication of the extent to which sensory attenuation operates in everyday behavior. The present study investigated ERP and behavioral measures of sensory attenuation for fearful and neutral facial expressions. Participants were trained to associate one voluntary action with the presentation of a fearful face, and another action with a neutral face. We measured both ERP responses and behavioral ratings following presentation of faces whose emotional content was either consistent or inconsistent with the action prediction. We observed significant modulation for fearful outcomes only, suggesting that sensory attenuation is heightened to stimuli of high social relevance. The N170 response was significantly attenuated for congruent fearful faces, but not for congruent neutral faces (in comparison to incongruent faces). Similarly, behavioral ratings were modulated only for fearful faces but not neutral faces. This provides new insight into how social and affective outcomes modulate sensory attenuation and may have implications for implicit sense of agency for socially relevant stimuli.

Introduction

Sensory attenuation refers to the fact that the effects of one’s voluntary actions exhibit a reduced neural response. This may play an important role in monitoring the success or failure of our actions, or attributing agency. The present study investigated whether the processing of predicted sensory action effects is modulated by their affective content. This provides important new insight in to the role of motor prediction in perception for socially relevant stimuli.

The majority of recent research on sensory attenuation has focused on processing of simple auditory, and, to a lesser degree, visual stimuli that follow voluntary action effects (for recent reviews see Hughes et al., 2013b, Waszak et al., 2012). These studies have shown that self-triggered tones exhibit reduced auditory (Baess, Jacobsen, & Schroger, 2008) and visual (Hughes & Waszak, 2011) ERP responses compared to externally triggered stimuli. Other researchers have manipulated the degree to which the observed action effect is congruent with a particular action (Cardoso-Leite et al., 2010, Hughes et al., 2013a, Roussel et al., 2013, Roussel et al., 2014), or whether actions are consistent with unconscious primes (Stenner et al., 2014). These studies have shown reduced early visual (Roussel et al., 2014) and auditory (Hughes et al., 2013a) ERP responses to congruent action effects, as well as reduced luminance discrimination (Roussel et al., 2013, Roussel et al., 2014) and target detection (Cardoso-Leite et al., 2010).

Taken together, these findings suggest that sensory events that are predicted by one’s voluntary actions are processed very differently to unpredicted stimuli. The present study investigated both neurophysiological and behavioral responses to fearful and neutral faces, to provide new insight in to the degree to which sensory attenuation might be modulated by these social and affective stimuli. The present study assessed whether sensory attenuation might differ dependent on the emotional content of voluntary action effects. Since faces are crucial for communication and social interaction, they are an ideal stimulus to measure the role of emotion on the sense of agency. One previous study (Hughes & Waszak, 2014) provided preliminary evidence of the efficacy of such ecologically valid stimuli in generating neurophysiological attenuation. However, that study focused only on neutral faces and houses, and did not incorporate a behavioral measure of sensory processing. In the current study, participants were trained to associate one of two buttons with the presentation of a fearful face, and the other button with a neutral face. Occasional violation of this pattern allowed for the assessment of the influence of action prediction on emotion processing in faces. Following the presentation of the face, participants were asked to rate the amount of fear presented in the face. This behavioral index differs significantly from those previously employed, which either use subjective ratings of intensity (Blakemore et al., 1999, Roussel et al., 2013, Roussel et al., 2014) or psychophysical measures such as stimulus detection (Cardoso-Leite et al., 2010) or point of subjective equality (Sato, 2008, Stenner et al., 2014). In contrast, participants were asked to provide a rating based on the emotional content of the visual stimulus rather than it’s physical strength. This provides novel insight into how action prediction modulates the processing of the content of action effects.

The absence of sensory suppression in patients suffering from hallucinations (Blakemore, Wolpert, & Frith, 2000) has been taken as evidence for the role of this phenomenon in self-monitoring and sense of agency (Frith, 2012, Frith et al., 2000). Recent evidence from the intentional binding paradigm (Takahata et al., 2012, Yoshie and Haggard, 2013) point to the possibility that pre-reflective agency is increased for positively valenced stimuli. Therefore, the current study also aims to provide further evidence about the role of emotion in sense of agency.

Previous research on sensory attenuation with simple visual and auditory stimuli has observed reduced auditory N1 amplitude (Baess et al., 2008, Hughes et al., 2013a), or primary visual responses (Roussel et al., 2014). Therefore, we might expect to observe attenuation in the visual P1 response. However, one previous study using neutral faces and houses (Hughes & Waszak, 2014) observed attenuation only in later components, suggesting that accurate prediction of higher-level visual features results in modulation of later stages of sensory processing. Of particular interest for the present study are the N170 component, and the later P2 component. The N170 ERP component is reliably observed following face stimuli (Bentin, Allison, Puce, Perez, & McCarthy, 1996) and has been seen to be modulated by emotional content of the face (Batty and Taylor, 2003, Blau et al., 2007), such that a greater N170 is observed for fearful faces. Other studies have shown that later components are modulated by both emotion (Eimer & Holmes, 2002) and facial recognition (Gosling & Eimer, 2011). Hughes and Waszak (2014) observed modulation only in the positive peak (P2) immediately following the N170 as a function of motor prediction.

Given that the P1 component is not typically modulated by the emotional content of a face (Eimer & Holmes, 2002) and that previous research observed no attenuated of P1 for predicted faces (Hughes & Waszak, 2014), modulation of this component in the present study would be unlikely. Rather, we predicted that we would observe modulation the N170 component for fearful versus neutral faces in this present study, with a larger N170 for fearful faces. Importantly, this should also be modulated as a function of action–effect congruency, such that the N170 should be more typically fearful (greater) for incongruent fearful faces, compared to congruent fearful faces. This would reflect attenuation of the N170 response to a fearful face. Precise predictions in the P2 time range are more difficult to establish, with previous research reporting either sustained positive (Eimer & Holmes, 2002) or negative components (Gosling & Eimer, 2011), or more focal modulations (Hughes & Waszak, 2014). Nonetheless, given previous reports of prediction related differences in this time range using a similar paradigm (Hughes & Waszak, 2014), we predicted some modulation of the P2 component.