Machine Learning to Identify Neural Correlates of Music and Emotions - Guide to Brain-Computer Music Interfacing - page 87

so-called discrete emotion theories and dimensional and appraisal theories, among
others, which constitute many theoretical lenses through which results from psy-
chology and neuroscience can be interpreted (see Chap. 6 for an introduction, and
Cornelius 1996 for an accessible and comprehensive review).
Discrete emotion theories or basic emotion theories originated after Charles
Darwin s seminal work on emotional expression. This theoretical tradition con-

ceptualises emotions as differentiated entities that possess unique properties, dis-
tinguishing one emotion from the others. In this lens, each emotion is mediated by
separate and distinct neural substrates, which are the result of evolutionary pressure.
A basic emotion occurs rapidly, and automatically, upon perception and classifi-

'

-

cation of a stimulus as belonging to a speci c class of events (Tomkins 1962 ;

Ekman 2003 ; Izard 2007 ).
Dimensional theories and appraisal theories are cousin traditions, which con-
ceptualise affective phenomena as the result of the integration of a number of
separate processes. Unlike basic emotion theories, these theoretical traditions
attempt the description of mechanisms common to most, if not all emotions.
Whereas appraisal theories focus on the mechanisms antecedent to the emotional
experience (Scherer et al. 2001 ; Roesch et al. 2006 ; Fontaine et al. 2007 ),
dimensional theories place the emphasis on the correlates of this experience, in the
brain and the behaviour, and represent emotion in a low-dimensional space.
In the context of our understanding of the neural correlates underlying the
emotional experience of music, the theoretical tradition that one chooses to depart
from will have strong consequences for the design of empirical work and the
practical decisions that researchers have to make for the analysis of their data. In
particular, this choice will have consequences when attempting to capture the
subjective experience of a participant placed in a speci

c context, as well as when
choosing neural features of interest and implementing analysis of the data.

5.2.2 Self-reporting Felt Emotion

There are a wide variety of methods available for individuals to report their felt
emotional response, all of which aim to accurately identify the emotional state of an
individual. Techniques commonly focus on identifying one property of the induced
affective state such as the intensity of the response (affect intensity measure, AIM)
(Larsen and Diener 1987 ), the valence (pro

le of mood states, POMS) (Albrecht
and Ewing 1989 ), or how aroused the individual is (UWIST mood adjective
checklist) (Matthews 1990 ). Most of these methods rely on participants accurately
interpreting descriptive emotive terms as well as having an adequate level of
emotional intelligence (Salovey and Pizarro 2003 ) to be able to assess their own
affective state.
To avoid any semantic misunderstanding of the adjective descriptions as well as
the dif culty of re ecting them to one ' s emotional state, many researchers use the