Relevance-ArticleHow does the human brain decide what information to attend to and what information to ignore?

Relevance detection refers to neural processes determining the salience of a stimulus.

According to the Appraisal Theory of Emotion, relevance detection is mainly modulated by three variables:

1) Novelty (in terms of familiarity and/or predictability)
2) Pleasantness (in terms of neutral, positive or negative valence)
3) Goal / Need Relevance (in terms of social versus nonsocial or communicative relevance, etc.)

In a number of research projects, I tried to disentangle the effects of these three variables on brain activity, both on a temporal (EEG) as well as a functional anatomical (fMRI) level. I furthermore assessed how the human brain represents an additional relevance category, namely social (versus nonsocial) information content. Although social information is known to be or high intrinsic relevance to the human organism, and previous fMRI investigations have shown that it likely is processed as an independent and unique stimulus dimension, it remains largely unknown how relevance detection in terms of novelty, pleasantness, and goal/need relevance may interact with the processing of social relevance.
Main region of interest for fMRI analyses was the amygdala (see references by Sander, Ousdal, as well as Pessoa and Adolphs). I was also interested in individual differences influencing such processes.


Social versus Emotional Relevance – fMRI
In one study published in Frontiers in Human Neuroscience (2013), we investigated the neural substrates of social versus emotional relevance. In order to do so, we presented images depicting complex visual scenes to healthy adult participants that could either have a social or nonsocial content, and be of either positive, negative or “neutral” valence.

relevanceWe found that in several brain areas, comprising the bilateral amygdala, fusiform gyrus, anterior superior temporal gyrus and medial orbitofrontal cortex, brain activity reflected interactive processing of social content and emotional valence. Firstly, activity was always higher to emotional versus neutral images. Second, activity was always higher to social versus nonsocial images. And third, activity to social images did not significantly differ as a function of positive versus negative valence, whereas for nonsocial images, activity was higher for negative versus positive images (see Figure above illustrating such pattern in bilateral amygdala).

Our findings suggest that the human brain, or at least the regions mentioned above, attribute(s) stronger relevance to social (versus nonsocial) information, regardless of valence. This reflects the notion that social information is highly salient for human behavior per se. In turn, probably because of the higher biological salience in terms of survival, nonsocial negative (versus positive) information is preferentially processed, allowing for a fast detection of potential threats.


Social versus Emotional Relevance – EEG

To build upon the above data, Annekathrin Schacht (University of Göttingen, Germany) and I took the same complex visual scenes as used in the fMRI study published in 2013, and embedded them into an EEG task to investigate the more precise temporal unfolding of social and emotional relevance processing. Our main questions were which of these two stimulus dimensions was processed first – i.e. which relevance check occurred first -, and when in time social and emotional relevance processing was interactive (as opposed to separate and/or additive), as suggested by our fMRI data.

The corresponding paper is currently under review. More information will be posted here as soon as it becomes available.


Emotional versus Social Relevance versus Novelty
In another fmRI study published in Emotion (2013), we examined brain responses to dynamic emotional facial expressions and comparable nonsocial stimuli (“lottery game”) as a function of stimulus novelty and valence. Novelty was conveyed by the presence versus absence of surprise, while valence was reflected by positive versus negative versus neutral stimulus content. We were especially interested in how the amygdala processes such different kinds of stimulus relevance.

Our findings revealed an intriguing activation pattern, for both social and nonsocial stimulus processing. The amygdala did not seem to differentiate between valence and novelty as such, but was more strongly activated to dynamic information containing positive versus negative surprise. According to the available literature, and particularly a previous study by Yacubian et al. (2006), such findings suggest that the human amygdala – at least under conditions of uncertainty involving surprise conveyed by dynamic facial expressions – (i) encodes a negative expected value (anticipation of a negative outcome), and subsequently (ii) updates such expectation by means of a negative expected value prediction error if a positive outcome occurs instead.

Our data propose that relevance processing in the human amygdala may not only be modulated by the static versus dynamic nature of stimulus presentation, but also depend on the degree of uncertainty / predictability of information.


For more detailed information regarding my research on relevance detection, please refer to the Publications page or contact me with any questions.