Short Communication
Gamma flicker elicits positive affect without awareness

https://doi.org/10.1016/j.concog.2012.07.001Get rights and content

Abstract

High-frequency oscillations emerged as a neural code for both positive affect and fluent attentional processing from evolutionary simulations with artificial neural networks. Visual 50 Hz flicker, which entrains neural oscillations in the gamma band, has been shown to foster attentional switching, but can it also elicit positive affect? A three-faces display (2-female/1-male or 2-male/1-female) was preceded by a 50, 25, or 0 Hz flicker on the position of the odd-one-out (i.e., the target). Participants decided on the gender (Block 1) or on the subjective valence (Block 2) of this neutral target in an approach-avoidance task, which served as an implicit affective measure. Only the detection of 25 Hz flicker, but not of 50 Hz flicker, was above chance (Block 3). Faces primed by invisible 50 Hz flicker were explicitly evaluated more positively than with 25 Hz or 0 Hz. This gamma flicker also facilitated approach reactions, and inhibited avoidance reactions relative to 25 Hz and 0 Hz flicker in Blocks 1 and 2. Attentional switching was, moreover, enhanced by the 50 Hz flicker. According to the Affect–Gamma hypothesis, also in biological neural networks, high-frequency gamma oscillations may code for positive affect.

Highlights

► Visual 50 Hz (gamma) flicker elicited positive affect. ► Gamma flicker primed the explicit evaluation of neutral faces. ► Gamma flicker implicitly modulated the speed of approach and avoidance movements. ► Gamma flicker facilitated attentional shifts. ► Gamma oscillations form a neural code for nonconscious positive affect.

Introduction

The scientific analysis of affect, the most basic component of emotions, should not stop at an intuitive understanding only in terms of the subjective experience of good and bad (Phaf & Rotteveel, 2012). Johnston (2003) provided an evolutionary grounding by proposing that positive affect constitutes the neural code of those conditions that enhance evolutionary fitness, whereas negative affect codes for fitness-reducing circumstances. From the evolutionary simulations of Heerebout and Phaf, 2010a, Heerebout and Phaf, 2010b the novel hypothesis emerged that positive affect is coded by high-frequency oscillations (e.g., in the gamma band: 40–70 Hz) and negative affect by lower-frequency oscillations. Such a mechanistic account of affect implies that affect does not need to be conscious, or open to introspection (see also Berridge, 2003, Berridge and Winkielman, 2003). In the simulations, the oscillations also modulated attentional flexibility, with higher frequencies facilitating switching by enabling the faster resetting of competition in the troughs (Heerebout & Phaf, 2010b; see also Jensen, Bonnefond, & VanRullen, 2012). Our AffectGamma hypothesis thus supplements the AttentionGamma hypothesis (Bauer et al., 2009, Herrmann, 2001, Womelsdorf and Fries, 2006, Womelsdorf and Fries, 2007). We extended the findings of Bauer et al. (2009) that invisible gamma flicker, which elicits neural gamma firing (Williams, Mechler, Gordon, Shapley, & Hawken, 2004), facilitates shifting to the flicker position by showing that 50 Hz flicker also acts as a positive prime.

In the computer simulations, which can only be outlined here (for more detail see Heerebout and Phaf, 2010a, Heerebout and Phaf, 2010b), evolution was mimicked by applying genetic algorithms (Holland, 1975) to agents reacting to food and predators (see Fig. 1). In these algorithms adaptation to an environment develops through the repeated application across generations of random variation and subsequent selection of the fittest genes. The fitness of all agents in a generation was tested in a simulated environment with food and predators, which emitted different artificial scents that were registered by the agent. Agents were controlled by simple artificial neural networks with input (i.e., receptor), hidden, and output (i.e., actuator) nodes. Predators had a fixed neural network that made them chase the scent of the agent. The movements of the agents and the predators were governed by the laws of classical mechanics. When a food patch collided with the agent, it disappeared from the environment and a new food patch appeared at a random location. Similarly, a collision between predator and agent resulted in the removal of the agent from the environment. The odds of transferring genes to the next generation for an agent increased with the agents’ success in approaching food (i.e., energy level) and avoiding predators (i.e., time spent in the environment), which was expressed in a single fitness variable. The connection weights between nodes served as ‘genes’ in the simulated evolution. The genes were subjected to random mutations and crossovers (i.e., reproduction by combining the genes of both parents) across generations. Oscillations in the activations emerged serendipitously, when recurrent connections between hidden nodes were enabled (Heerebout & Phaf, 2010a). The emergence coincided with a near doubling of the agents’ fitness, demonstrating their adaptive function. Oscillations facilitated attentional switching between approach and avoidance actions.

The fittest agents could switch quickly from approach to avoidance when collecting food. If a predator suddenly appeared, fast switching contributed to successful escape. However, if the agent was evading predators, it should persist in its action mode and not be distracted by food. Higher oscillation frequencies developed more often with fitness-enhancing (i.e., good) than with fitness-reducing (bad) stimuli (Heerebout & Phaf, 2010b), because switching from food to predator is more adaptive than vice versa. The simulated evolution implemented attentional selection between response options by the resolution of network competition (cf. Duncan, 1996, Phaf et al., 1990), which evolution implemented as mutual inhibition between nodes. The oscillation troughs, representing the lowest level of mutual inhibition, provide opportunities to reset competition and switch ‘winners’. The number of these switching opportunities per time unit increases with oscillation frequency. The well-established link in human research between affect and attentional flexibility (e.g., see Fredrickson, 2004, Heerebout et al., in press, Olivers and Nieuwenhuis, 2006, Tan et al., 2009) thus directly emerged from the simulations. In addition, evolutionary computation generated the novel insight that high-frequency oscillations are associated with positive affect.

To test the Affect–Gamma hypothesis, we adapted Bauer et al.’s (2009) experiment, which supported the Attention–Gamma hypothesis. These authors demonstrated that a high-frequency invisible flicker in the gamma band (40–70 Hz) modulated attentional shifts to target locations. Bauer and colleagues presented three Gabor patches, located equidistantly on a virtual circle. One of the patches flickered at 50, 25, or 0 Hz (i.e., no flicker) in a short preview phase, after which one patch showed a subtle change in spatial frequency and participants had to report its location as fast as possible. The flicker was set to a low contrast gradient so that participants would not consciously detect it. Because average contrast of the patches was held constant across the preview and the experimental phases, however, there could be a “transition flash” on the location of the primed patch (van Diepen, Born, Souto, Gauch, & Kerzel, 2010; but see Cheadle, Parton, Müller, & Usher, 2011). Detection of the subtle change in spatial frequency was facilitated by the 50 Hz flicker, and to a lesser degree by the 25 Hz flicker, at the target location. The larger effect with 50 Hz than with 25 Hz is probably not due to a higher saliency, because the 50 Hz could not be consciously detected, whereas the 25 Hz could. Instead, we argue that it is due to a speeding up of switching by gamma. The affect–gamma hypothesis predicts, moreover, that the 50 Hz flicker should simultaneously induce more positive affect than the 25 Hz flicker.

To investigate affective priming, we replaced the Gabor patches by gray-scaled images of male and female, emotionally neutral, faces. The stimulus display contained either two males and one female, or two females and one male. To eliminate the transition flash, a 0, 25, or 50 Hz flicker in the background gray preceded the diverging target (i.e., the odd-one-out), while the other locations remained at background gray in the preview phase. Participants had to decide on the target’s gender (Block 1), while affect was measured implicitly in the speed of performing approach and avoidance movements, or whether the face made a positive or negative impression on them (i.e., explicit affective evaluation of the target, Block 2). To check for awareness, a flicker-detection task was performed in a third block.

Affective modulation of attention by oscillations as it emerged from our simulations required only simple networks with a few levels of nodes, and recurrent connections between at least two levels. Processing capacities of these evolutionary-early networks are limited and certainly do not involve conscious processing. We, therefore, expected that these effects could be measured implicitly (i.e., without explicit reference to affect or consciousness for the affect-inducing stimulus; e.g., Block 1). As in the simulations, also in humans positive affect has been linked to approach, and negative affect to avoidance, even when there is no awareness of affect in the stimuli or action tendencies (e.g., Phaf & Rotteveel, 2009). Implicit approach and avoidance tendencies (Block 1) were assessed in a gender judgment task with the same approach-avoidance button stand as was used by Rotteveel and Phaf (2004). The explicit evaluations in Block 2 were also performed with the upper and lower buttons of this stand. We expected gamma flicker to result in more positive face evaluations, and speed up approach and slow down avoidance relative to the 0 Hz control condition, which should be superimposed on a general acceleration due to attentional facilitation.

Section snippets

Participants

Forty-six right-handed university students (average age = 22.2 ± 3.1 yrs; 16 men) participated for course credit and signed informed consent. Persons who indicated first or second degree epilepsy or migraine were excluded from participation. The results of two participants were excluded from the analyses due to not following instructions.

Material and apparatus

All stimuli were presented on a 17″ LG Flatron 795FT+ Cathode-Ray-Tube (CRT) monitor set at a 100 Hz refresh rate and resolution of 1024 * 768 pixels. Participants were

Affective priming

With explicit evaluation in Block 2, 50 Hz flicker resulted in more positive ratings of target faces than 0 Hz and 25 Hz (F(2, 42) = 6.16, p = .0031, MSE = 0.051; see Fig. 3A). This was mirrored in the implicit AAT. Facilitation of flexion initiation and inhibition of extension due to 50 Hz flicker relative to 0 Hz and 25 Hz occurred in Block 1 (Frequency × Movement interaction: F(2, 86) = 3.20, p = .049, MSE = 4103, see Fig. 3B), and Block 2 (F(2, 86) = 2.90, p = .0606, MSE = 6510, see Fig. 3C). With the more standard

Discussion

Invisible priming by visual gamma flicker raised positive affect towards neutral faces, both implicitly and explicitly. The amplification of attentional shifting by gamma flicker was replicated (Bauer et al., 2009). The emergence from evolutionary simulations of the Affect–Gamma hypothesis, as well as of the Attention–Gamma hypothesis, illustrates the productive capacity of this computational method. The relationship between gamma and attention has been investigated extensively (Bauer et al.,

Acknowledgments

We thank Jasper Wijnen and Marcus Spaan for technical support and Matthijs Feenstra, Marius Usher, Henk van Steenbergen, and Gezinus Wolters for helpful comments.

References (33)

  • Duncan, J. (1996). Cooperating brain systems in selective perception and action. In T. Inui, & J. L. McClelland (Eds.),...
  • B.L. Fredrickson

    The broaden-and-build theory of positive emotions

    Philosophical Transactions of the Royal Society B: Biological Sciences

    (2004)
  • G.G. Gregoriou et al.

    High-frequency, long-range coupling between prefrontal and visual cortex during attention

    Science

    (2009)
  • Heerebout, B. T., Todorović, A., Smedinga, H. E., & Phaf, R. H. (in press). Affective modulation of attentional...
  • B.T. Heerebout et al.

    Emergent oscillations in evolutionary simulations

    Journal of Cognitive Neuroscience

    (2010)
  • B.T. Heerebout et al.

    Good vibrations switch attention: An affective function for network oscillations in evolutionary simulations

    Cognitive, Affective, & Behavioral Neuroscience

    (2010)
  • Cited by (6)

    • Cutaneous retinal activation and neural entrainment in transcranial alternating current stimulation: A systematic review

      2016, NeuroImage
      Citation Excerpt :

      Interestingly, several studies suggest that photic stimulation can influence perceptual and cognitive performance, even when subjects are not consciously aware of the flickering stimuli (Elliott and Muller, 1998; Knez, 2014). Finally, prior research has shown that task-irrelevant stimulation in the peripheral visual field can alter emotional states (von Gizycki et al., 1998; Heerebout et al., 2013). Observations of increased positive mood in response to (subliminal) IPS hint at a moderating influence of mood in cognitive ability (Heerebout et al., 2013; Knez, 2014).

    • Neural correlates of apparent motion perception of impoverished facial stimuli: A comparison of ERP and ERSP activity

      2014, NeuroImage
      Citation Excerpt :

      Specifically, increments in beta oscillations in the right parietal scalp were suggested as reflecting an index of avoidance (Schutter et al., 2001). Additionally, induced high-frequency oscillations in the gamma-band have been linked to enhanced processing of positive affect in static face stimuli (Heerebout et al., 2013). It is difficult to relate our results to these studies, as in our study there was no affective component to the stimuli, nor was the task one that required a social judgment.

    • Affective modulation of attentional switching

      2013, American Journal of Psychology
    View full text