How does our brain process vocal emotions: in categories or dimensions?
When we hear an emotionally charged voice, does our brain analyse the emotion as a separate
category or as continuous dimensions? An international team of researchers has come up with
a surprising answer to this question using a combination of neuroimaging techniques and
computational models of the voice: it’s not one or the other, but both. Voice emotions are
initially processed by the brain as separate categories, before being refined into dimensions.
We all know how effective the voice is at conveying emotions. What we know less about is
whether our brain treats vocal emotions as separate categories (such as pleasure or fear) or along
a continuum of emotional dimensions (such as negative to positive).
A new study puts an end to this long-standing debate by showing that it is neither categories nor
dimensions, but both, with quite different brain dynamics.
An international consortium of researchers from France, Scotland, Germany, Holland, Canada,
and the United States used brain imaging to measure the brain activity of healthy adult volunteers
as they listened to brief affective vocalizations representing different emotions. The affective
properties of the vocalizations were controlled using a computational model to produce a wide
range of more or less intense and identifiable emotions. The participants each underwent several
sessions of functional magnetic resonance imaging and magnetoencephalography to maximize
the spatial and temporal resolutions of the brain activity measurement as well as statistical
power. Sophisticated analyses indicate that both categories and dimensions explain much of the
brain’s activity, but differently: while early brain activity within 2/10ths of a second reflects
distinct emotional categories, later neural responses (after half a second) are increasingly
consistent with a finer, graduated representation of emotional dimensions.
These new data make it possible to reconcile two long opposing views by describing with a high
degree of spatio-temporal detail the representational dynamics of the cerebral processing of
emotions in the voice.
This study was supported by funding from the British BBSRC and the French ANR; the manuscript
and data are freely available.
Analysis of cerebral representations of vocal emotion. Top left: brief affective vocal expressions
are generated by morphing between recordings corresponding to 4 emotions (anger, fear, disgust,
pleasure) and a neutral expression. Top right: matrices representing the differences in brain
activity between each pair of stimuli, each brain region and each millisecond are compared to
matrices based on emotional judgments reflecting either categories or emotional dimensions.
Bottom: representation of brain regions showing an association with emotional categories at an
Giordano, B.L., Whiting, C., Kriegeskorte, N., Kotz, S.A., Gross, J., Belin, P.
The Representational Dynamics of Perceived Voice Emotions Evolve from Categories to
Dimensions, accepted for publication in Nature Human Behaviour.