These results challenge the assumption that facial identity and
expression are processed independently (Bruce & Young, 1986)
and support the hypothesis that computation of expression is integrated
with identity (e.g., Ganel & Goshen-Gottstein, 2004). This
latter perspective is also consistent with Bruce and Young’s updated
view that analysis of identity and expression is to some extent
interdependent (Young & Bruce, 2011). A number of recent
adaptation studies have shown that emotion aftereffects in individual
expressions (when one of the target expressions matches
the adapting face) are also modulated by identity (Campbell &
Burke, 2009; Ellamil, Susskind, & Anderson, 2008; Fox & Barton,
2007; Vida & Mondloch, 2009). In each of these studies aftereffects
for the same-identity conditions were larger than aftereffects for
different-identity conditions. This was interpreted as evidence that
there are at least two visual representations of emotion expression;
one that is identity-dependent and one that is identityinvariant
(Fox & Barton, 2007; Fox, Oruç, & Barton, 2008).
Preliminary evidence has also recently emerged that the
overlapping visual representations of emotional expressions are
characterized by identity-dependent and identity-independent
components (Skinner & Benton, 2010, 2012). Using a crossemotion
adaptation paradigm that examines the direct effect of
one expression on another, we found further support for this representational
structure. Our results also build on previous evidence
that there is considerable overlap of expression representations
(Cook, Matei, & Johnston, 2011; Hsu & Young, 2004; Rutherford
et al., 2008) and indicate that overlapping expression representations
reconfigured by cross-emotion adaptation are the part of
the same computational space as individual expression representations.
This is latter point is important because previous crossemotion
adaptation research has revealed asymmetric interactions
between disgust and anger that are not accounted for in influential
models of facial expression (e.g., Russell, 1980). It was also notable
that when the data were visualized across the entire morph continuum,
the strongest aftereffects were observed for the most ambiguous
images in the middle of the continuum, consistent with
within-emotion aftereffects.