Assume that we as theorists have a system which allows us to categorise
static pictures of faces and other stimuli according to which emotion we
think they are expressing: some faces are happy, others fearful, and so on.
From five months of age,
or possibly much earlier \citep{field:1982_discrimination},
through to adulthood, humans are better at distinguishing faces when they
differ with respect to these categories than when they do not
\citep{Etcoff:1992zd,Gelder:1997bf,Bornstein:2003vq,Kotsoni:2001ph,cheal:2011_categorical,hoonhorst:2011_categoricala}.
To illustrate, consider these pictures of faces.
The idea is this.
With respect to all features apart from the expression of emotion, each face picture differs from its neighbours no more than any other picture differs from its neighbours.
Most neighbouring pairs of face pictures would be relatively hard to distinguish,
especially if they were not presented side-by-side.
But most people find one pair of neighbouring face pictures relatively easy to distinguish---you may notice this yourself.
What underlies these patterns of discrimination?
Several possibilities that would render them uninteresting for our purposes can be ruled out.
The patterns of discrimination do not appear to be an artefact of linguistic labels (\citealp{sauter:2011_categorical}; see also \citealp{laukka:2005_categorical}, p.\ 291),%
%
\footnote{
Puzzlingly, experiments by \citet{fugate:2010_reading} using photos of chimpanzee faces with human subjects are sometimes cited as evidence that categorical perception of expressions of emotion depends on, or can be modulated by, the use of verbal labels for stimuli (e.g.\ \citealp[p.\ 288]{barrett:2011_context}; \citealp[p.\ 315]{gendron:2012_emotion}).
Caution is needed in interpreting these findings
given that there may be differences in the ways humans process human and chimpanzee faces.
In fact, what \citeauthor{fugate:2010_reading}'s findings show may be simply that `human viewers do not show [categorical perception] for the chimpanzee facial configurations used in their study' \citep[p.\ 1482]{sauter:2011_categorical}.
}
%
nor of the particular choices subjects in these experiments are presented with \citep{bimler:2001_categorical,fujimura:2011_categorical}.
Nor are the patterns of discrimination due to narrowly visual features of the stimuli used \citep{sato:2009_detection}.
We can be confident, then, that the patterns of discrimination probably reflect one or more processes which categorises stimuli by expression of emotion.
Don't have details, this is just to stress it's early (around 200ms) and plausibly automatic.
Batty & Taylor (2003, figure 2)
At least some of the processes underpinning categorical perception of facial expressions
of emotion are rapid (occurring within roughly 200 milliseconds of a stimulus' appearance), pre-attentive \citep{vuilleumier:2001_emotional} and automatic in the sense that whether they occur is to a significant degree independent of subjects' tasks and motivations \citep{batty:2003_early}.%