Emotion Review

http://emr.sagepub.com/

Emotional States from Affective Dynamics

William A. Cunningham, Kristen A. Dunfield and Paul E. Stillman
Emotion Review 2013 5: 344
DOI: 10.1177/1754073913489749
The online version of this article can be found at:
http://emr.sagepub.com/content/5/4/344

Published by:
http://www.sagepublications.com

On behalf of:

International Society for Research on Emotion

Additional services and information for Emotion Review can be found at:
Email Alerts: http://emr.sagepub.com/cgi/alerts
Subscriptions: http://emr.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav

>> Version of Record - Sep 3, 2013

What is This?

Downloaded from emr.sagepub.com by ancuta anca on October 25, 2014

489749
October 2013

EMR5410.1177/1754073913489749Emotion ReviewCunningham et al. Affective Dynamics

Emotional States from Affective Dynamics

Emotion Review
Vol. 5, No. 4 (October 2013) 344355
The Author(s) 2013
ISSN 1754-0739
DOI: 10.1177/1754073913489749
er.sagepub.com

William A. Cunningham

Department of Psychology, University of Toronto, Canada

Rotman School of Management, University of Toronto, Canada
Department of Psychology, The Ohio State University, USA

Kristen A. Dunfield

Department of Psychology, The Ohio State University, USA

Department of Psychology, Concordia University, Canada

Paul E. Stillman

Department of Psychology, The Ohio State University, USA

Abstract
Psychological constructivist models of emotion propose that emotions arise from the combinations of multiple processes, many
of which are not emotion specific. These models attempt to describe both the homogeneity of instances of an emotional kind
(why are fears similar?) and the heterogeneity of instances (why are different fears quite different?). In this article, we review the
iterative reprocessing model of affect, and suggest that emotions, at least in part, arise from the processing of dynamical unfolding
representations of valence across time. Critical to this model is the hypothesis that affective trajectoriesover timeprovide
important information that helps build emotional states.

Keywords
dynamics, emotion, psychological constructivism

A major debate regarding the psychological processes of emotion has surrounded whether emotions are natural kinds (e.g.,
the basic emotions models) or whether they result from the
interaction of more elemental units combined in various ways
(e.g., appraisal or dimensional models). According to natural
kind models, specific emotions, such as fear or anger, developed
independently to help an organism respond adaptively to specific challenges in the environment (Ekman, 1992). For example, the widening of the nose and eyes in the facial expression of
fear are the result of increasing vigilance and sensitivity to
incoming information when threatened, whereas the narrowed
nose and eyes in the facial expression of disgust comes from the
need to reduce input from a source of potential contamination
(Susskind etal., 2008). Similarly, patterns of autonomic
responses and body posture have been taken to suggest that
emotions, once triggered, activate a unified whole body response
(Levenson, 1988). In contrast, elemental models propose that

emotions are not unitary modular phenomena, but rather reflect

the interaction of more general processes arising from more
basic affective and/or cognitive representations. For example,
some models propose that emotions originate from a core affect
(a two-dimensional space representing valence and arousal;
Russell, 1980; see also Thayer, 2012; Watson & Tellegen, 1985)
that is translated into emotions through cognitive elaboration
(Russell, 2003). Sadness and fear share negative valence, but
differ in arousal (sadness involves negative valence and low
arousal, while fear involves negative valence and high arousal).
Other models state that events are appraised on a number of
goal relevant dimensions (e.g., goal relevance and goal congruence), and the resulting emotion is a function of these interpretations (e.g., Lazarus, 1991; Scherer, 2009). Despite the fact that
these two broad families of models (natural kind and elemental)
have fundamentally different views on the nature of emotion,
how emotions arise, and the consequences of emotion, neither

Author note: The authors thank Alison Duncan Kerr, Julie Huang, and the members of the Social Cognitive Science Lab for helpful comments on an earlier version of this article.
Corresponding author: William A. Cunningham, Department of Psychology, University of Toronto, 100 St. George Street, Toronto ON M5S 3G3, Canada.
Email: [email protected]

Downloaded from emr.sagepub.com by ancuta anca on October 25, 2014

Cunningham et al. Affective Dynamics 345

perspective has been able to completely rule out the other in

order to establish dominance.
Due in part to the fundamental disagreements regarding the
nature of emotion, new perspectives have surfaced that attempt
to understand both the great homogeneity of emotions (different
instances of fear seem more similar to one another, on average,
than they do to different instances of anger) as well as the great
heterogeneity of emotions (not all instances of fear are the same,
and indeed some fears may more resemble anger than other
fears). Under the general rubric of psychological constructivism, these new models attempt to understand the basic elements
of emotion, and how these basic elements combine to result in
the emergent states of emotional experience and behavior
(Barrett, 2009). Although these models agree that emotions
arise from more elemental units, and agree that these units combine to create emergent states, the exact number and nature of
these elements is just beginning to be explored. In this article,
using the iterative reprocessing (IR) model, we propose that
emotions arise, at least in part, from the processing and interpretations of changes in valenced states over time.

The Psychological Construction of Mind

In developing models to articulate the form and function of the
human mind, psychologists and philosophers have worked to
develop and refine numerous shorthand categories to reduce
the overwhelming complexity of the human experience into
elemental units that could be more readily comprehended. For
thousands of years, categories such as cognition, emotion, and
attitude have been used to simplify the challenge of understanding behavior and thought. Yet, however useful these categories have been as starting points and simplifying devices,
we argue that modern behavioral and brain sciences have
made the conceptual mistake of reifying these heuristics into
natural kind categories. In doing so, psychological theory
has imbued these shortcuts with dissociable causal properties,
while neuroscientific research has sought to find localized
neural tissue associated with a particular category to further
objectify it, and to argue for its realness. Indeed, there is a
growing consensus that some brain regions are cognitive,
others emotional, and there has been an ongoing pursuit to
find the neural homes of specific emotions like fear and
disgust, or specific evaluative concepts such as the implicit
attitude or the explicit attitude.
We, in line with other psychological constructivist
approaches, argue that the tactic of searching for natural kind
modules within the brain may not be ideal for understanding
how nature is carved at its jointsproposing instead that greater
attention should be directed toward more elemental and computational aspects of mind. As one important example, we believe
that the distinction between emotion and cognition is a false
dichotomy (Cunningham & Kirkland, 2012). Instead, we take a
broad view of cognition as encompassing any information processing (see Newell, 1990; Oatley & Johnson-Laird, 1987). In
this way, emotion as we experience it is inseparable from
cognition, in that all mental operations require some form of

information processing. Indeed, when considering the subjective experience of emotion, it is probable that the underlying
cognitive processes are complex and multifaceted, and possible
that when more fully articulated may not even correspond to
current linguistic categories. By refocusing our examination to
the information processing elements that underlie emotion, with
an eye toward linking levels of analysis (how the elements combine is as important as knowing what the elements are), we
believe that it will be possible to understand not only the homogeneity of emotional experiences (there are likely similarities
among instances of anger), but also the heterogeneity of emotional experience (not all instances of anger are the same). In
this article we use the IR model (Cunningham & Zelazo, 2007;
Zelazo & Cunningham, 2007) to understand affect and emotion
as the dynamic emergent result of hierarchically organized brain
systems.

The Iterative Reprocessing Model

The IR model is a dynamical systems account of human information processing rooted in the hierarchical organization of
brain function. Importantly, instead of drawing strict distinctions along the lines of traditional dichotomies, IR views all
information processing as an emergent property of more general
processes. Fundamental to the IR model is the observation that
although the brain is organized hierarchically, it allows for bidirectional influences, such that processes that are typically considered automatic or reflexive can influence, and are influenced
by, processes that are typically considered controlled or reflective. When faced with environmental changes, or internal
changes from cognitive processing (e.g., imagery or reconstrual), people quickly process active information to determine
its meaning. Yet this initial processing following a perceived
change does not necessarily provide a final state (or even a state
that lasts for more than a few milliseconds). Rather, the information is continuously reprocessed through iterative cycles
potentially creating richer evaluations of the information and
thereby more nuanced interpretations and thus affect.
Importantly, although reprocessing allows an event to be
understood in a more nuanced manner by situating it in an
ever-broadening array of considerations, additional iterations
do not always lead to more nuanced or complex evaluations.
Rumination, for example, may involve multiple iterations, but
this does not necessitate updating of the information. Rather,
the same information may be gated such that a dominant representation remains rigid, despite the fact that it is not useful
for a goal. Moreover, by considering the reciprocal (feedforward and feedback) nature of these activations, we can see the
continuum of human information processing that extends from
what is traditionally considered relatively simple and automatic, to more complex and reflective.

IR Model and the Hierarchical Brain

As humans interact with their environment they strive to maintain a healthy internal environment while making accurate

Downloaded from emr.sagepub.com by ancuta anca on October 25, 2014

346 Emotion Review Vol. 5 No. 4

predictions about the external world. Therefore, any discrepancy between the expectation and experience of either the
internal states or external world initiates a sequence of evaluative processes in which the information is interpreted and reinterpreted in iterative cycles. Sometimes the processing and
reprocessing is accomplished quickly and effortlessly, and
other times more complex, meaningful, and multifaceted representations need to be constructed. Specifically, whenever any
new information is encountered, be it a stimulus in the world or
an experience generated in our own mind, it is initially evaluated for goal valence (i.e., harmful/beneficial) and relevance
(Sander, Grafman, & Zalla, 2003), resulting in an affective
state including some degree of arousal/relevance (e.g., Russell,
2003; Scherer, 1984, 2009). These initial responses result in
unreflective motivational behaviors such as approach or avoidance, and occur within the first few hundred milliseconds of
perception (Oya, Kawasaki, Howard, & Adolphs, 2002). Given
the rapidity of these initial responses they typically involve
processing in the subcortical brain, specifically the amygdala
and ventral striatum (nucleus accumbens) are likely based on
innate biases (e.g., LeDoux, 1996; hman & Mineka, 2001)
and learning (e.g., Armony & Dolan, 2002; Phelps etal., 2001;
Whalen etal., 1998). Previous functional magnetic resonance
imaging (fMRI) research has demonstrated that the amygdala
quickly and consistently responds to a wide variety of valenced
cues (e.g., Anderson etal., 2003; Canli, Zhou, Brewer, Gabrieli,
& Cahill, 2000; Isenberg etal., 1999; Morris etal., 1996;
Morris, hman, & Dolan, 1998; Small etal., 2003; Whalen
etal., 1998; Williams etal., 2006; Winston, Strange, ODoherty,
& Dolan, 2002). In addition to producing relatively automatic
responses, these subcortical areas probably also play an important role in generating and updating representations in light of
subsequent reflective processes. Importantly, although these
initial undifferentiated responses allow individuals to quickly
prepare and respond to experiences, they rarely take into consideration the full range of motivational implications of any
particular piece of information.
With additional iterations, current experience can be reinterpreted in light of a larger range of more complex considerations
such as active goal states, expected rewards and punishments,
and current context (Beer, Heerey, Keltner, & Scabini, 2003;
Blair, 2004; Frank & Claus, 2006; Rolls, 2000). By reinterpreting the current experience in light of these constraints, it is possible to make more nuanced evaluations that are consistent
with more stable long-standing goals, desires, and intentions.
Indeed, in order to enact an appropriate response, it is important to know what caused the change being processed, how
much power one has in a situation, and critically what behavioral options are available (see Arnold, 1960; Lazarus, 1991;
Ortony, Clore, & Collins, 1988; Roseman, 1984; Scherer, 2009,
for examples). These more nuanced interpretations are aided
through direct reciprocal connections between the orbitofrontal
cortex (OFC) and the amygdala and hypothalamus. Given these
connections, the OFC is in a special, and particularly effective,
position to modulate initial responses in order to fit with a particular context. By integrating input from multiple sensory

modalities the OFC allows for more nuanced stimulus evaluations (posterior medial orbitofrontal cortex) and the integration
of novel information with more long-standing goals and motivations (anterior medial orbitofrontal cortex; Cunningham,
Kesek, & Mowrer, 2009).
Often the reprocessing of information in order to come to a
more nuanced and an appropriate interpretation requires the
integration of complex rules and goals. To that end, additional
regions of the prefrontal cortex (PFC; such as lateral PFC) can
bias representations to reduce residual uncertainty. Mirroring
the hierarchical structure of the whole brain, the PFC is also
organized hierarchically. As information is reprocessed in the
prefrontal cortex, it spreads from ventrolateral, to dorsolateral,
to rostrolateral prefrontal cortex (e.g., Badre & DEsposito,
2007; Botvinick, 2008; Bunge & Zelazo, 2006; Koechlin, Ody,
& Kounelher, 2003). Within the PFC there is a segregation of
the processing of rules at different levels of complexity:
Conditional rules are processed in the ventrolateral prefrontal
cortex (VLPFC) and dorsolateral prefrontal cortex (DLPFC),
whereas the rostrolateral prefrontal cortex (RLPFC) deals with
explicit considerations of task sets (Bunge & Zelazo, 2006).
Importantly, the integration of the lateral prefrontal cortex
allows for the regulation and biasing of activate representations,
not by creating entirely new low-level states, but by affecting
attention to various aspects of the stimulus through the selective
amplification and/or suppression of attention (e.g., Cunningham,
Raye, & Johnson, 2004; Cunningham, van Bavel, Arbuckle,
Packer, & Waggoner, 2012; Ochsner, Bunge, Gross, & Gabrieli,
2002; Ochsner etal., 2004; Todd, Cunningham, Anderson, &
Thompson, 2012). The lateral frontal lobes play an essential
role in the reprocessing of information because of their influence on working memory (allowing relevant aspects of the stimulus to be kept in mind) and inhibitory control (allowing
information to be selectively attended to), two key abilities
required for reflective interpretations (Ochsner, 2004).

The Nature of Change and Its Representation

IR utilizes connectionist approaches to information processing
developed in the area of computational cognitive neuroscience
to understand how representations are organized (e.g., OReilly,
Munakata, Frank, & Hazy, 2012). Connectionist models view
higher order mental activity (such as an evaluation or an affective state) as an emergent property of the patterns and interactions of interconnected networks (e.g., units of neurons). These
frameworks have several properties that help explain how the
dynamic nature of information can be used to generate affective
states, and how affect emerges over time. A fundamental premise of connectionist models is that meaningful information
resides in the variable and dynamic patterns of activation across
multiple units, as opposed to a single specific activation within
a unit (for in-depth reviews of connectionist models in social
psychology, see E. R. Smith, 1996, 2009).
A critical component of connectionist models is that the units
within the network gravitate towards stable patterns of activation
called attractor states. An attractor state is the most probable

Downloaded from emr.sagepub.com by ancuta anca on October 25, 2014

Cunningham et al. Affective Dynamics 347

pattern of activation given a wide array of neuronal activations.

This means that, given different inputs, the network will settle on
a single internal representation. For example, if one were to visit
a pet store each dog has unique characteristics based on things
like breed, age, and sex, yet despite all the variations we can still
settle on a single representationdog. The diverse array of
inputs gravitates toward the same internal representation (e.g.,
categorization). The strength of these attractor states is that they
allow us to quickly come to a stable internal representation even
in cases where inputs are ambiguous, uncertain, or novel. Yet,
this dog representation is not processed in isolation, and contributes to a sum total of experience that includes not only the
percepts related to the dog, but also activations associated with
the context in which the dog is encountered. This is particularly
important in light of the processing goals to build a stable representation of the environment that allows for prediction and
appropriate reactions (Bar, 2009). The process of the network
settling into a given pattern of activation allows us to predict
how the event or stimulus will interact with the environment,
thus giving our representations greater nuance. So when a fluffy
blur comes running at you, you can quickly categorize it as a
dog, and you can predict that and anticipate a future state that
includes some cuddles and fetch. Most of the time, the settling
process is quick and accurate. However, the network can settle
into an incorrect internal representation, resulting in inaccurate
predictions (incorrectly categorizing as dog instead of wolf
and the incorrect predictions that follow).
Connectionist models store information via a distributed pattern of neural activation. When stimuli are detected in the environment, multiple individual neurons fire in response to different
aspects of the stimulus and it is the combined activation of these
distinct neurons that leads to the representation of the current
stimulus (see OReilly etal., 2012, for a review). Because the
recognition of any given stimulus is the function of the probabilistic sum of activation from multiple independent neurons,
distributed representations allow for both stability and flexibility. The active representation of any given piece of information,
be it an attitude, emotion, or self-perception, is the function of
both the preexisting connection weights (that are relatively stable and stored in memory) and the current state of activation
(which is a function of factors such as goals, context, and current hedonic experiences, and serves to make certain attractor
states more or less gravitational). Indeed, stronger connection
weights (which are a function of previous experience) make it
more likely that similar patterns of activation will be generated,
given a perceptual input, whereas foregrounding can change
stimulus construals by reflectively creating distinct patterns of
activation and altering the starting weights by which a stimulus
is processed.
The goal of the mind is to settle into a stable, predictive internal representation of the environment, similar to a system going
from a high entropy state to a low entropy state (see Friston,
2012, for a formal discussion of these principles for biological
systems). Here, what we mean by entropy is the degree of
organization of active representations. Representations that
have not yet settled into an attractor state have a possibility of

settling into a number of different stable internal representations, and thus there is greater uncertainty about how they will
settle. Put another way, unstable, uncategorized representations
have many possible configurationsthe same (unstable) representation can be generated by many different patterns of activation. As the stimulus or event settles into a stable pattern of
activation, the number of probable forms the representation will
take decreases, thereby reducing entropy. Successive iterations
allow for more nuanced representations and therefore less
entropy and greater prescriptive predictions.
Boltzmann defines entropy as the number of possible microstates that can account for a given observable macrostate
(Boltzmann, 1877; Shannon, 1948; see also Hirsch, Mar, &
Peterson, 2012). Applied to representations (i.e., a stable representation that the network has settled into), entropy can be
thought of as the number of possible unique patterns of activation within the neuronal network that could produce that representation. In other words, entropy is the number of possible
patterns of neuronal activation (microstates) that could create
the observed representation (macrostate). A stimulus that has
just been encountered, and has not yet settled into a stable
internal representation, produces a representation that is disorganized and not unique from many other disorganized representations (though it does not usually stay in this form for very
long). This is a high entropy state because many arrangements
of the neuronal network can produce a disorganized representation (just about any set of inputs can result in a disorganized
representation at first). Generally, a representation wont stay
in this state for very long (e.g., far less than a second) before
subsequent iterations refine the representation, for example
going from a jumbled mess of fur, activity, to my dog. Along
the way, the associated affect of the active representation can
change, as an unknown dog may be more threatening than my
own dog.
Many fewer arrangements of the network could produce
these more specific representations, and as such the entropy of
the network is successively lowered as the iterative refinement
process progresses. An equivalent way of saying this is that each
iteration of processing allows for the reduction of the number of
potential higher-order representations that can be implied by a
given set of representations, or in other words, the number of
potential stable representations that the pattern can easily fall
into at that time. With the reduction of entropy comes predictive
power, since more refined representations carry more specific
predictions. Thus, as the mind progresses through iterative
cycles, it simultaneously reduces network entropy and increases
predictive power.
This conceptualization has several implications. First and
foremost is that any change creates entropybe it small
changes, such as encountering a new dog in your neighborhood, or big changes, such as your beloved family dog unexpectedly biting you (a giant error in prediction from what was
anticipated). The increase in entropy will be proportional to the
degree to which ones internal representations shift based on
the event (and will usually correspond to how novel or
expectancy-violating the event is). This is because such events

Downloaded from emr.sagepub.com by ancuta anca on October 25, 2014

348 Emotion Review Vol. 5 No. 4

launch representations out of their stable state and thus require

the neuronal network to resettle. It also suggests that some
sources of entropy (e.g., your dogs appearance on return from
the groomer) are easily reducible because the network can easily settle into a new stable representation, whereas other sources
may create entropy that is much more difficult to reduce (e.g.,
your dog reciting a Shakespearean sonnet). Entropy can be difficult to reduce for a number of reasons, but two important
sources of persistent entropy are either novel stimuli that do not
settle easily into a preset attractor state, or violations of expectation that not only fundamentally alter ones current representation, but can also disrupt other representations in the network
(or at other layers of the network), thus launching them out of
their attractor states as well. Since the brain is trying to reduce
entropy, these persistent entropy sources will warrant further
iterations in order to reduce the entropy by arriving at a stable
internal representation for that event.
One of the crucial implications of a dynamic iterative system
is that the prior state of the system is extremely important.
Because the system is constantly updating its representations
through rerepresentation, and because these rerepresentations
are at least in part dependent on the previous representation,
taking into account the state of the system at Time (n 1) is necessary to understand and predict the state of the system at the time
of interest. In this way, the previous states of the system act as
powerful biasing factors for the further processing of information. Put another way, from an IR perspective, there is no such
thing as time zerothe previous states of the system are
always influencing the representation process in principled and
important ways, for example by altering the attractor state
weights for subsequent processing (see Figure 1). From this perspective, the traditional distinctions between emotional reactivity and emotion regulation somewhat fall away. Because the
prior state of the system influences how information is processed, all information is regulated to some degree.
This formulation is consistent with the proposals of control
theory (Carver & Scheier, 1982) in that a source of affect is the
direct result of discrepancies in representations related to goal

Timen - 1

Timen

Timen + 1

Situaonn - 1

Situaonn

Situaonn + 1

directed action. In their cybernetic model, progress toward

goals is constantly monitored in feedback loops, and to the
extent that a discrepancy (as the result of a comparator process) is found, the direction of the discrepancy generates an
affective state. If the discrepancy is beneficial (e.g., faster than
expected progress toward a goal), the resulting affect is positive, whereas if the discrepancy is harmful (e.g., slower than
expected progress toward a goal), the resulting affect is negative (Carver & Scheier, 1982). In the case of intense emotions,
it is likely that these incompatible representations are not easily resolved, and as such an ongoing emotional episode will be
experienced until resolution can occur. This reduction can
come from cognitive processing (such as reappraisal; Gross,
2008) or by changing ones actions (and ones physiological
state to perform those actions).1
It is important to note that although we have focused nearly
exclusively on central nervous system processes of emotion, the
peripheral nervous system (e.g., the body) provides important
signals and constraints. Autonomic feedback is a key feature of
emotional experience. For example, Barrett and Bliss-Moreau
(2009) suggest that sensory information from the world is represented in somatovisceral, kinesthetic, proprioceptive, and neurochemical fluctuations and that these bodily representations
form a core affective state. This body state is cortically rerepresented in the somatosensory cortex, particularly the insula,
which can then be integrated into subsequent steps of affective
processing through connections to the amygdala and OFC.
Cortical interpretations of these body states can provide information about the state of the individual and, following some
cognitive interpretation, lead to the development of more
nuanced emotional experiences. In addition to using information from the body to inform brain states, changes in affectively
related brain states often lead to changes in body states. For
example, once a potential threat has been detected, the body
may need to organize in preparation for an immediate fight or
flight response. This organization of body into action provides
an important cue for future iterative reprocessing of valenced
information.

Affecve
staten - 1

Affecve
staten

Predicted affecve
staten + 1

Affecve
staten + 1

Predicted affecve
staten + 2

Figure 1. Multiple determinants of emotional state. At any given moment in time, an individuals current affective state is partially determined by
(a) the situation, or what is occurring in the environment and (b) the individuals affective trajectory: comparing the current state of the world with
what the individual had predicted for himself. A current affective state also naturally leads to a prediction for the future: whether things will improve,
worsen, or remain the same. For example at Time (n + 1), the individuals affective state is jointly determined by his representation of the world at Time
(n + 1) and what he had predicted for himself at Time (n). This composite affective state informs a prediction for his affective state at Time (n + 2).

Downloaded from emr.sagepub.com by ancuta anca on October 25, 2014

Cunningham et al. Affective Dynamics 349

From a computational viewpoint, then, the body serves as

an important biasing agent for the generation of representations
and rerepresentations. Previously, we discussed how structures
such as the prefrontal cortex can influence representations both
by allowing for more sophisticated rerepresentation or by altering the attractor state landscape such that stimuli are more likely
to settle into a certain internal representations. The body can
serve a similar rolebody input can alter and constrain the
attractor landscape, therefore biasing subsequent representations
to make certain representations more or less likely. In the early
stages of an emotion (when entropy has just spiked and the mind
is trying to resolve it) the mind sends signals to the body to act in
some way. This action is then reinterpreted by the mind and
serves to constrain or bias the way information can be rerepresented. In this way, the body may make the experience of an
emotion more likely or more rapid via biasing input towards settling into certain attractor sets, which can then be subsequently
refined by further iterations. As an example, if you are frightened
by something and begin running, the act of running is reinterpreted by the mind to further constrain subsequent representations to be consistent with the notion that you are frightened
(somewhat akin to James, 1890).
Conceptualizing an origin of an emotional episode as arising
from persistent unresolvable entropy in the network of the mind
means that the body is not necessary for emotions per se. That
said, the body gives important cues to the mind and contributes
to emotions in meaningful ways, and our experience of emotions would likely be qualitatively different without the body.
Without the preexisting constraints that the body provides, emotions would likely be muted and differ in important ways. For
example, the unfolding of an emotional experience likely takes
longer without bodily feedback to help constrain incoming
inputs to a certain representational space. The extent to which
the body is important for emotions will differ, however, depending both on the specific emotion and the particular instantiation
of a given emotional experience. If the bodily constraints are
less necessary for representations to settle in certain ways, then
the body will be less important for a given emotional experience. This approach also suggests that valence and salience do
not need to be dependent on the body, but rather that they are a
function of innate reinforcement circuitry which, at any given
time, may or may not be receiving input from the body. Although
bodily experiences can influence and constrain the experience
of valence and arousal, they are not obligatory for the affective
experience.
The body is also necessary for action, and different emotional states are correlated to a greater or lesser degree with different action tendencies (Frijda, Kuipers, & ter Schure, 1989).
Once one makes the prediction that something bad is about to
happen, and that this threat is imminent, there are a limited
number of behavioral (and organized action patterns) options
available to the organism. One can hide/freeze, run, or attack (or
perhaps attempt to negotiate). Each of these behavioral
responses has functional utility, and may have clearly differentiated physiology. For example, Susskind etal. (2008) suggest
that the facial expressions associated with fear may enhance

sensory processing, whereas the facial expressions associated

with disgust may enhance sensory rejectionappropriate
responses for many instances of these emotional states. Yet, to
use this example, there may be many pathways to sensory
enhancement or rejection, only one of which may be fear or
disgust, respectively. Indeed, there may be cases when one is
afraid, but the appropriate response may be to reduce intake,
and as such the expression in that situation may more reflect
canonical disgust than fear.
Considering the temporal unfolding of an emotional episode,
it could be that whereas the antecedents of emotional processing
may be varied, and the processes may be numerous, once a specific action is required, different instances may look more similar to the extent that the resultant actions are similar. That there
are more finite behavioral responses (only a certain number of
muscle and autonomic patterns make sense), and more infinite
combinations of cognitive processes that can get one to an emotional state (no two emotional situations are likely identical),
may help reconcile the differences found by researchers in the
basic emotions and the constructivist camps. In his review of
affective neuroscience, Berridge (2003) notes that whereas the
animal work tends to focus on brain stem and limbic regions,
human work tends to focus on cortical regions. The animal work
seems to suggest more basic emotions and motor programs,
whereas the human work emphasizes more domain general processes and flexibility. As he notes, it is quite possible that these
two lines of work converge rather than diverge. The animal
work may highlight the finite behavioral options (correlated
with antecedents that are correlated with emotional states),
whereas the human work highlights the numerous ways in
which one can process the environment. As such, what is typically taken as evidence of basic emotions from the literature
may be better labeled survival circuits (LeDoux, 2012). In other
words, a given emotional state (however it is generated) may be
probabilistically associated with an appropriate response (which
may come prepackaged evolutionarily), but the activation of a
survival circuit is not the emotion, but rather the consequence of
the organism operating within the environment.
Together, this suggests that the heterogeneity of emotional
experience comes from the dynamic nature of the emotional
episode. As shown in Figure 2, an emotional episode can begin
with any affective change, whether it be a discrepancy between
what one was feeling and what one is now feeling, between
what one expected to feel and what one actually feels, or
changes in what one expects to feel in the future. Each of these
discrepancies result in information entropy that needs to be
resolved, and it is in the resolution of this entropy that the emotional episode is experienced. For example, one can cognitively
change ones representations, perhaps through reappraisal processes (Ray, Wilhelm, & Gross, 2008) such as reducing the
importance of an event, changing the affective meaning of an
outcome, or altering ones memory to create less discrepancy.
Alternatively, one can act behaviorally to change the situation,
such as fleeing from a fearful event or attacking a potential
threat. In these cases, discrepancy reduction can be achieved
through situation modification (Gross, 2008). As an emotional

Downloaded from emr.sagepub.com by ancuta anca on October 25, 2014

350 Emotion Review Vol. 5 No. 4

Figure 2. The dynamics of emotion generation as a function of entropy

reduction. Affective changes result in increases in entropy, which can be
resolved either behaviorally or cognitively. The emotional episode lasts
until this entropy is sufficiently reduced.

episode unfolds, the strategies can shift as the situation itself

changes (either internally or externally), with the episode ending once entropy levels are at a lower rate.2

The Emergence of Differentiated Emotion

Categories
From a psychological constructivist perspective, we take as a
starting point the premise that the wide variety of emotional
experiences can be generated through the interactions of a
more limited number of basic mental ingredients (Barrett,
2006a, 2006b, 2009). On our view, however, the affective
ingredients typically proposed are not sufficient to capture the
processes of affect as they too are constructed from more elemental processing units. That is, although affect experientially
may comprise a circumplex structure of two axes capturing
some degree of valence and arousal (Plutchik, 1962; Russell,
1980, 2003; Watson & Tellegen, 1985; Yik, Russell, & Barrett,
1999), this does not suggest that valence and arousal are unitary processing elements. Rather, the processing of valence
occurs dynamically in time, with separable representations of
valence for the present, past, and future (Cunningham & van
Bavel, 2009; Kirkland & Cunningham, 2012). As described
earlier, evaluative states are constructed dynamically through a
series of iterative neural loops occurring multiple times per
second. These multiple mental systems serve as a way of tracking our affective trajectories through time. Incoming information is compared to previous information and the discrepancy
between these two states is computed (see also Scherer, 2009).
This in turn informs interpretation and future prediction. Thus,
our affective space is comprised of our past, our present, and
what we predict for our future. The current affective state is
constructed based on the newest incoming information (what
just happened, including comparisons to previous predictions)
and the information that existed before (past events or feeling

states) as well as any predictions being made about what may

occur next (see also Carver & Scheier, 1990, for a similar argument about goal pursuit and the experience of general positive
and negative states). The emotion categories used by humans are
thus a way to label and differentiate the various affective trajectories we experience as we move continuously through time.
To the extent that affect is dynamic, and reprocessed from
moment to moment, it is likely that the affective states labeled
as emotional also reflect the ongoing dynamics of affective
experience within a temporally sensitive framework. Previous
affective states are composed of memory representations of an
individuals immediate affective past. The current affective
state is an evaluation of ones current state as a function of
outcomes. Predicted affective states are evaluations of what is
likely to happen next. Critically, comparisons can be made
between these time points through communication among the
relevant neural circuits, allowing us to map out our particular
affective place in time (Cunningham & van Bavel, 2009;
Kirkland & Cunningham, 2012). By focusing on affective
dynamics as a starting point for emotional processing, this suggests that emotional states result primarily from changes in the
representation of affect within or between systems. That is, just
as our perceptual systems are sensitive to changes in sensory
input, our affective system is sensitive to changes in valence.
Importantly, change or discrepancy, especially if it cannot be
resolved quickly, leads to an increase in the entropy of the system and a motivation to reduce the entropy either through
behavioral or cognitive modifications. According to this view,
a predicted negative state may be labeled fear, whereas a current negative state may be labeled sadness. Differentiating
between these categories is important because they provide
information as to where in the system the change has occurred.
A negative event that can still be avoided may lead to a different set of behavioral options than one that has already occurred
(Lazarus, 1982).
As an emotional experience unfolds, these processes continuously interact to generate an emotion episode. Because the
episode lasts until entropy is reduced, different emotions can
last quite different periods of time, from an instant of joy quickly
resolved, to a week of anxiety regarding a lost dog. Thus,
although emotions are generated through a dynamic cycle of
processing, some emotions may result from attention to only
parts of the system. For example, the emotion category fear may
simply require labeling the feeling state generated by a negative
prediction signal (e.g., the moment between when you notice
the front door being left open and finding Fido curled up safely
in the hall). Others, such as joy or sadness, may require more
comparison; in these instances the comparison might reveal an
upward or downward affective trajectory, respectively. For
example, when a new dog is first introduced in the family it may
not be immediately clear if this is cause for celebration (because
the animal quickly bonds with the family, thus increasing affective trajectory) or will be cause for concern (because the animal
is unable to be left alone, resulting in a negative affective trajectory). In contrast to models that tacitly begin emotional processing after stimulus presentation (at a figurative time zero), this

Downloaded from emr.sagepub.com by ancuta anca on October 25, 2014

Cunningham et al. Affective Dynamics 351

perspective suggests that emotional states are rarely separate

from the affective and motivational context in which they arise
and may, in fact, necessarily require changes in affective processing from previous to current states (i.e., the addition of a
second dog may improve family relations if the existing family
dog has trouble being alone, but may damage family relations if
it causes a previously relaxed dog to become aggressive and
territorial). Within this frame, the hard distinction between cognition and emotion falls awayeach referring to a different
aspect of a unified dynamic system.
To the extent that these changes in dynamic affect, at least in
part, underlie the construction of emotional experience, we
should expect that the linguistic categories that we use should
mirror our predictions. To test this hypothesis, Kirkland and
Cunningham (2012) presented participants with information
about affective trajectories and recorded which emotion they
thought best fit the scenario. To eliminate any semantic information that could be used to determine emotional categories, participants were simply provided with information about past,
present, and predicted valence. For example, you are feeling
good, and you predict that something bad will happen. Which
emotional label best characterizes this situation? or you expect
something good to happen, but something worse than expected
happens. Which emotional label best characterizes this situation? Importantly, participants were given the option of reporting that they would not experience any emotion. As expected,
each of the basic emotions fits into particular quadrants of the
affective space. For example, when a worse-than-expected outcome follows the prediction that something good will happen,
that situation is labeled as causing anger, while when a worsethan-expected outcome follows the prediction that something
bad will happen, that situation is labeled as causing sadness.
Additionally, emotion categories are differentiated to a greater
extent when participants are required to think categorically than
when participants have the option to consider the possibility of
multiple emotions and degrees of emotions. This work indicates
that information about affective movement through time and
changes in affective trajectory may be a fundamental aspect of
emotion categories. Another important factor of the model is that
as we often experience a particular affective trajectory in conjunction with a suite of behavioral and regulatory strategies,
associated thoughts and interpretations, as well as the labels that
we used to categorize it. Through time, these pairings can
become self-organizing such that a particular trajectory becomes
associated with a specific suite of behavioral strategies and a certain emotional label (see Lewis, 2005). The repeated experience
of similar affective trajectories and their behavioral correlates
eventually alter the attractor-state landscape, resulting in the categorization of a particular suite of prototypical elements as a specific emotional experience. Similar to how bodily feedback
influences the unfolding of emotions by constraining them via
altering the attractor landscape, ones categories constrain ones
experience by making certain representations more likely. This
makes the prediction that different people can have a similar
label for an emotional experience (e.g., fear), but that label corresponds to different attractor-state landscapes. Indeed, depend-

ing on the representations associated with a particular emotional

label, a particular emotion may have vastly different consequences for the two individuals. For example, fear experiences
often differ in the extent to which they suggest fight versus
flight responses. For example, an individual who is typically
in fight-appropriate situations might have their attractor landscape altered in such a way that fight is activated even when
flight is more appropriate, in part because their emotional experience of fear has been developed around the behavioral affordance
of fight, making it harder to activate the flight response. In
this way, our emotional categoriesdeveloped over time and
repeated emotional experiencesconstrain and alter our emotional experiences as well.

Interactions: Cognitive Representations,

Appraisals, and Construction
Although comparisons among the temporal valence representations can provide information for the construction of a particular
emotional experience, iterative reprocessing allows for processing to become more refined with time. That is, although affective responses begin with some combination of physiological
activation and neural evaluations, emotional episodes also
appear to involve some degree of cognitive interpretation.
Indeed, a primary goal of appraisal theorists has been to understand which cognitive interpretations are necessary and/or sufficient for an emotional response. According to appraisal theory,
different emotions arise from the cognitive interpretation of the
environment and the implications that these environments have
for the perceiver. Although the specific dimensions vary from
theory to theory (Frijda et al., 1989; Ortony etal., 1988;
Roseman, 1984; C. A. Smith & Ellsworth, 1985), most involve
a calculation of whether an event is self-relevant, predictable,
consistent with ones goals, caused internally (by the self) or
another, and whether one has the capacity to deal with the
change (Scherer, 1988). If one considers these appraisals in
multidimensional space, different emotions occur in different
quadrants. Thus, our appraisals about affectsuch as who or
what is causing it, how much control we have, whether the state
is consistent with our goals, and so forthhelp us to understand
and even define our emotional experience. The particular emotion that is experienced may be largely dependent on the aspects
of the situation or object to which one attends. The situation
reflects the perceivers unique interpretation of his or her surroundings in terms of personal relevance. Given the computational nature of these appraisals, current models propose an
iterative sequence of appraisal checks that begin with a basic
sense of relevance and valence and build in complexity toward
a differentiated emotional experience (Scherer, 2009).
As with the conceptual act model (Barrett, 2006a, 2006b),
our model of emotion requires the integration of aspects of
valence processing with cognitive categories to fully explain the
full range of emotional behavior and experience. Especially,
although we have articulated how the processing of valence
across time can lead to emotions, we do not believe that these

Downloaded from emr.sagepub.com by ancuta anca on October 25, 2014

352 Emotion Review Vol. 5 No. 4

trajectories alone are sufficient to create all emotional experiences. Rather, we propose that these trajectories are one of the
ingredients that are used in combination with other processes.
On this view, a pattern of valence information regarding the
past, the present, and the anticipated future prime our cognitive
systems toward a particular emotional state. If one is predicting
that something bad will happen, emotional responses typically
associated with fear are more likely. If one also experiences a
change toward negative valence, but this is a downward trajectory, then emotional responses typically associated with sadness
are more likely. Yet the increased probability that the state will
be labeled as sadness is not the same as saying that the particular
pattern of valence representations is sadness. Rather, this information needs to be combined with our interpretations of the
environment (appraisals) and the different behavioral options
that are available at the moment. A predicted bad event that can
have the potential for escape is likely experienced quite differently from one when trapped. Thus, the trajectory model can be
thought of as providing a preappraisal of dynamic valence.
The combination of trajectory information with additional
ingredients allows for multiple expressions of emotional experiences. Not all situations are the same, and not all options are
present given the same cues. As such, the ways in which people
choose to engage with the environment shape the experience,
the body with respect to the environment, and the actual event
itself (Gross, 2008), giving rise to a heterogeneity of emotional
experience and behavior.

The Development of Emotion Ingredients

To the extent that emotions result from the interaction of hierarchically organized neural systems, and that these systems provide important constraints across levels of processing, this
perspective suggests important hypotheses regarding the development of emotional experiences. As discussed previously, the
experience of any particular affective state is the emergent property of the integration and evaluation of multiple representations (e.g., valence, current goals, expected outcomes, etc.).
Moreover, resolution of much uncertainty is thought to occur as
information is processed at increasing higher-order brain areas
(such as the OFC and PFC), thus neuronal maturation plays an
important role in the development and experience of emotions.
Although PFC function is first observed towards the end of
the first year of life (e.g., Chugani & Phelps, 1986; Diamond &
Goldman-Rakic, 1989), the area continues to develop throughout childhood and well into adolescence (e.g., Giedd etal.,
1999; Gogtay etal., 2004). Similarly, infants emotional experiences begin in the first year of life and increase in diversity and
complexity with development. Early research on infant emotion
(e.g., Wolff, 1987) suggests that babies are born with relatively
undifferentiated, simple affective systems that mainly comprise
positive and negative affect. These two broad classes of affective experience can be linked to basic motivational drive states
resulting in a general tendency to approach positive and avoid
negative stimuli. For example, within the first few months of
life infants begin to smile in response to positive stimulation,

including gentle touches, high-pitched voices, and engaging

images (such as static faces; Sroufe, 1995); whereas infants
respond negatively with cries and distress in the face of negative
stimulation like medical injections (Izard, Hembree, & Huebner,
1987), or frustration (Camras, 1992; Hiatt, Campos, & Emde,
1979; Lewis, Alessandri, & Sullivan, 1990). Importantly, as
infants age and gain better control over their self and their environment they can move beyond simple environmental based
reactions.
Indeed, through maturation children start to pair environmental cues with affective expectations (potentially through
activation in the amygdala-striatal circuits; Cunningham &
Zelazo, 2009), resulting in more complex affective experiences. For example, at around 3 months of age infants start to
pair people with the pleasure of social interactions and start
exhibiting social smiles (White, 1985), eliciting reciprocal
delight from others in their environment (Camras, Malatesta,
& Izard, 1991; Huebner & Izard, 1988). At the same time,
these more complex expectations also lead to the experience
of negative emotions such as fear and frustration (e.g.,
Alessandri, Sullivan, & Lewis, 1990). Thus, as the PFC
develops, children can move beyond their immediate hedonic
experience, to have affective experiences associated with
expectations.
As children move into their second year of life, increasing
neuronal maturation allows for the representation of absent
stimuli and the ability to imagine different affective states. This
maturation also allows for better affective predictions and more
complex emotional experiences. A good example of this association between the development and emotion is the relation
between prefrontally maintained working memory (e.g., Baird
etal., 2002; Liston & Kagan, 2002) and the emergence of stranger anxiety (Kagan, 1972, 1981). Specifically, it is the ability to
anticipate a negative trajectory in ones current affective state
that results in the experience of anxiety at the approach of an
unknown individual. Indeed, it is orbitofrontal cortex function
that is thought to be critical for the ability to integrate present,
previous, and predicted hedonic states.
Finally, in the third year of life children can begin to integrate self-reflection with their understanding of the mental
states of others. As children begin to integrate the thoughts,
beliefs, and evaluations of others into their affective predictions, the result is more complex affective experiences that
rely on social comparison, such as shame, guilt, empathy, and
pride (Lewis, Sullivan, Stanger, & Weiss, 1989). Interestingly,
childrens recognition and labeling of others emotional experiences follows a similar, albeit protracted, developmental trajectory with children first making broad, valenced attributions
(i.e., overgeneralizing from happy and sad) that narrow
into more specific emotional states over the preschool years
(Widen & Russell, 2003, 2008). Taken together, the IR model
in general, and the affective trajectories hypothesis in particular, grounded in the hierarchical function of the brain, make
predictions consistent with the emergence of emotions over
time from early undifferentiated, body based affect (e.g., positive vs. negative), to emotions based on environmental

Downloaded from emr.sagepub.com by ancuta anca on October 25, 2014

Cunningham et al. Affective Dynamics 353

predictions (e.g., anxiety and frustration), to more complex

socially based affective experiences (e.g., empathy, shame, guilt).

Conclusion
Neuroscience methodologies and perspectives have been useful
tools in the continuing process of understanding how affect
manifests in the human brain, and the implications of these findings for models of emotion. Much is now known about the neural systems involved in affective processing that was relatively
inaccessible even 20 years ago. In our review of the literature,
we identify four aspects of affective processing that fall out of
the consideration of dynamic cognition (and in particular the IR
model). These include the generation of affective predictions
for the future, the representation of current affective states, the
integration of information from the body, and the engagement
of reflective processing to integrate appraisals, interpretation,
categorization, and meaning. This perspective should allow for
new research aimed at not only understanding the homogeneity
of emotional experience (the similarities among instances of
fear), but also the heterogeneity of emotional experiences (the
differences among the fear episodes). By taking into consideration the role of time in both the short (moment by moment)
and long (life-span development) term, we can better understand how emotions unfold and transform to adapt to changing
environments.

Notes
1

According to this view, a sharp distinction between motivation and

affect/emotion is not necessary. High entropy states entail a set of processes to resolve the entropy and return to a low entropy state. When
considering the full episode it may be labeled an emotional state, but
when considering the time course and actions performed it may be
labeled a motivated state. Thus, emotion and motivation computationally may be, in turns of a core ingredient, two sides of the same coin.
It is possible that different people have different ideal levels of
entropy, and this may lead to different types of trait emotionality.

References
Alessandri, S. M., Sullivan, M. W., & Lewis, M. (1990). Violation of expectancy and frustration in early infancy. Developmental Psychology, 26,
738744.
Anderson, A. K., Christoff, K., Stappen, I., Panitz, D., Ghahremani, D. G.,
Glover, G., & Sobel, N. (2003). Dissociated neural representations
of intensity and valence in human olfaction. Nature Neuroscience, 6,
196202.
Armony, J. L., & Dolan, R. J. (2002). Modulation of spatial attention by fear
conditioned stimuli: An event-related fMRI study. Neuropsychologia,
40, 817826.
Arnold, M. B. (1960). Emotion and personality. New York, NY: Columbia
University Press.
Badre, D., & DEsposito, M. (2007). FMRI evidence for a hierarchical
organization of the prefrontal cortex. Journal of Cognitive Neuro
science, 19, 118.
Baird, A. A., Kagan, J., Gaudette, T., Walz, K., Hershlag, N., & Boas, D.
(2002). Frontal lobe activation during object permanence: Evidence for
near infrared spectroscopy. NeuroImage, 16, 11201126.

Bar, M. (2009). The proactive brain: Memory for predictions. Philosophical Trans
actions of the Royal Society, Series B: Biological Sciences, 364, 12351243.
Barrett, L. F. (2006a). Emotions as natural kinds? Perspectives on Psycho
logical Science, 1, 2858.
Barrett, L. F. (2006b). Solving the emotion paradox: Categorization and
the experience of emotion. Personality and Social Psychology Review,
10, 2046.
Barrett, L. F. (2009). The future of psychology: Connecting mind to brain.
Perspectives in Psychological Science, 4, 326339.
Barrett, L. F., & Bliss-Moreau, E. (2009). Affect as a psychological primitive. Advances in Experimental Social Psychology, 41, 167218.
Beer, J. S., Heerey, E. A., Keltner, D., & Scabini, D. (2003). The regulatory function of self-conscious emotion: Insights from patients with
orbitofrontal damage. Journal of Personality and Social Psychology,
85, 589593.
Berridge, K. C. (2003). Comparing the emotional brain of humans and other
animals. In R. J. Davidson, H. H. Goldsmith, & K. Scherer (Eds.),
Handbook of affective sciences (pp. 2551). New York, NY: Oxford
University Press.
Blair, R. J. R. (2004). The roles of orbital frontal cortex in the modulation of
antisocial behavior. Brain and Cognition, 55, 198208.
Boltzmann, L. (1877). Uber die Beziehung zwischen dem zweiten Hauptsatz
der mechanischen Warmetheorie und der Wahrscheinlichkeitsrechnung
respektive den Satzen uber das wwrmegleichgewicht. [On the relationship between the second law of the mechanical theory of heat and the
probability calculus]. Wiener Berichte, 76, 373435.
Botvinick, M. M. (2008). Hierarchical models of behavior and prefrontal
function. Trends in Cognitive Sciences, 12, 201208.
Bunge, S., & Zelazo, P. D. (2006). A brain-based account of the development of rule use in childhood. Current Directions in Psychological
Science, 15, 118121.
Camras, L. A. (1992). Expressive development and basic emotions. Cognition
& Emotion, 6, 269283.
Camras, L. A., Malatesta, C., & Izard, C. (1991). The development of facial
expressions in infancy. In R. Feldman & B. Rim (Eds.), Fundamen
tals of nonverbal behavior (pp. 73105). New York, NY: Cambridge
University Press.
Canli, T., Zhou, Z., Brewer, J., Gabrieli, J. D. E., & Cahill, L. (2000). Activation in the human amygdala associates event-related arousal with
later memory for individual emotional experience. Journal of Neuro
science, 20, 15.
Carver, C. S., & Sheier, M. F. (1982). Control theory: A useful conceptual
framework for personalitySocial, clinical, and health psychology.
Psychological Bulletin, 92, 111135.
Carver, C. S., & Scheier, M. F. (1990). Origins and functions of positive
and negative affect: A control-process view. Psychological Review, 97,
1935.
Chugani, H., & Phelps, M. (1986). Maturational changes in cerebral function in infants determined by 18FDG positron emission topography.
Science, 231, 840843.
Cunningham, W. A., Kesek, A., & Mowrer, S. M. (2009). Distinct orbito
frontal regions encode stimulus and choice valuation. Journal of Cogni
tive Neuroscience, 21, 19561966.
Cunningham, W. A., & Kirkland, T. (2012). Emotion, cognition, and the
classical elements of mind. Emotion Review, 4, 369370.
Cunningham, W. A., Raye, C. L., & Johnson, M. K. (2004). Implicit and
explicit evaluation: fMRI correlates of valence, emotional intensity, and
control in the processing of attitudes. Journal of Cognitive Neuroscience,
16, 17171729.
Cunningham, W. A., & van Bavel, J. J. (2009). Varieties of emotional
experience: Differences in object or computation? Emotion Review, 1,
5657.
Cunningham, W. A., van Bavel, J. J., Arbuckle, N. L., Packer, D. J., &
Waggoner, A. S. (2012). Rapid social perception is flexible: Approach

Downloaded from emr.sagepub.com by ancuta anca on October 25, 2014

354 Emotion Review Vol. 5 No. 4

and avoidance motivational states shape P100 responses to other-race

faces. Frontiers in Human Neuroscience, 6, 17.
Cunningham, W. A., & Zelazo, P. D. (2007). Attitudes and evaluations:
A social cognitive neuroscience perspective. Trends in Cognitive
Sciences, 11, 97104.
Cunningham, W. A., & Zelazo, P. D. (2009). The development of iterative
reprocessing: Implications for affect and its regulation. In P. D. Zelazo,
M. Chandler, & E. Crone (Eds.), Developmental social cognitive neuro
science (pp. 8198). New York, NY: Taylor and Francis.
Diamond, A., & Goldman-Rakic, P. S. (1989). Comparison of human
infants and rhesus monkeys on Piagets AB task: Evidence for dependence on dorsolateral prefrontal cortex. Experimental Brain Research,
74, 2440.
Ekman, P. (1992). An argument for basic emotions. Cognition & Emotion,
6, 169200.
Frank, M. J., & Claus, E. D. (2006). Anatomy of a decision: Striatoorbitofrontal interactions in reinforcement learning, decision making,
and reversal. Psychological Review, 113, 300326.
Frijda, N. H., Kuipers, P., & ter Schure, E. (1989). Relations among emotion, appraisal, and emotional action readiness. Journal of Personality
and Social Psychology, 57, 212228.
Friston, K. (2012). A free energy principle for biological systems. Entropy,
14, 21002121.
Giedd, J. N., Blumenthal, J., Jeffries, N. O., Castellanos, F. X., Liu, H.,
Zijdenbos, A., Rapoport, J. L. (1999). Brain development during
childhood and adolescence: A longitudinal MRI study. Nature Neuro
science, 2, 861863.
Gogtay, N., Giedd, J. N., Lusk, L., Hayashi, K. M., Greenstein, D., Vaituzis,
A. C., Thompson, P. M. (2004). Dynamic mapping of human cortical development during childhood through early adulthood. Proceed
ings of the National Academy of Sciences, USA, 101, 81748179.
Gross, J. J. (2008). Emotion regulation. In M. Lewis, J. M. Haviland-Jones,
& L. F. Barrett (Eds.), Handbook of emotions (3rd ed., pp. 497512).
New York, NY: Guilford.
Hiatt, S. W., Campos, J. J., & Emde, R. N. (1979). Facial patterning and
infant emotional expression: Happiness, surprise, and fear. Child
Development, 50, 10201035.
Hirsh, J. B., Mar, R. A., & Peterson, J. B. (2012). Psychological entropy:
A framework for understanding uncertainty-related anxiety. Psycho
logical Review, 119, 304320.
Huebner, R. R., & Izard, C. E. (1988). Mothers responses to infants facial
expressions of sadness, anger, and physical distress. Motivation and
Emotion, 12, 185196.
Isenberg, N., Silbersweig, D., Engelien, A., Emmerich, S., Malavade, K.,
Beattie, B., Stern, E. (1999). Linguistic threat activates the human
amygdala. Proceedings of the National Academy of Sciences, USA, 96,
1045610459.
Izard, C. E., Hembree, E. A., & Huebner, R. R. (1987). Infants emotion
expressions to acute pain: Developmental change and stability of
individual differences. Developmental Psychology, 23, 105113.
James, W. (1890). Principles of psychology. London, UK: Macmillan.
Kagan, J. (1972). Motives and development. Journal of Personality and
Social Psychology, 22, 5166.
Kagan, J. (1981). The second year: The emergence of self-awareness.
Cambridge, MA: Harvard University Press.
Kirkland, T., & Cunningham, W. A. (2012). Mapping emotions through
time: How affective trajectories inform the language of emotion.
Emotion, 12, 268282.
Koechlin, E., Ody, C., & Kounelher, F. (2003). The architecture of cognitive control in the human prefrontal cortex. Science, 302, 11811185.
Lazarus, R. S. (1982). Thoughts on the relations between emotions and
cognition. American Physiologist, 37, 10191024.
Lazarus, R. S. (1991). Progress on a cognitive-motivational-relational
theory of emotion. American Psychologist, 46, 819834.

LeDoux, J. (1996). The emotional brain: The mysterious underpinnings of

emotional life. London, UK: Simon & Schuster.
LeDoux, J. (2012). Rethinking the emotional brain. Neuron, 73, 653676.
Levenson, R. W. (1988). Emotion and the autonomic nervous system:
A prospectus for research on autonomic specificity. In N. H. Wagner
(Ed.), Social psychophysiology: Theory and clinical applications
(pp. 1742). London, UK: Wiley.
Lewis, M. D. (2005). Bridging emotion theory and neurobiology through
dynamic systems modeling. Behavioral and Brain Sciences, 28,
169193.
Lewis, M., Alessandri, S. M., & Sullivan, M. W. (1990). Violation of
expectancy, loss of control, and anger expressions in young infants.
Developmental Psychology, 26, 745751.
Lewis, M., Sullivan, M. W., Stanger, C., & Weiss, M. (1989). Self development and self-conscious emotions. Child Development, 60, 146156.
Liston, C., & Kagan, J. (2002). Brain development: Memory enhancement
in early childhood. Nature, 419, 896.
Morris, J. S., Frith, C. D., Perrett, D. I., Rowland, D., Young, A. W.,
Calder, A. J., & Dolan, R. J. (1996). A differential neural response in
the human amygdala to fearful and happy facial expressions. Nature,
383, 812815.
Morris, J. S., hman, A., & Dolan, R. J. (1998). Conscious and unconscious
emotional learning in the amygdala. Nature, 393, 467470.
Newell, A. (1990). Unified theories of cognition. Cambridge, MA: Harvard
University Press.
Oatley, K., & Johnson-Laird, P. N. (1987). Towards a cognitive theory of
emotions. Cognition & Emotion, 1, 2950.
Ochsner, K. N. (2004). Current directions in social cognitive neuroscience.
Current Opinion Neurobiology, 14, 254258.
Ochsner, K. N., Bunge, S. A., Gross, J. J., & Gabrieli, J. D. E. (2002).
Rethinking feelings: An fMRI study of the cognitive regulation of
emotion. Journal of Cognitive Neuroscience, 14, 12151299.
Ochsner, K. N., Ray, R. D., Cooper, J. C., Robertson, E. R., Chopra, S.,
Gabrieli, J. D., & Gross, J. J. (2004). For better or for worse: Neural
systems supporting the cognitive down- and up-regulation of negative
emotion. NeuroImage, 23, 483499.
hman, A., & Mineka, S. (2001). Fear, phobias and preparedness: Toward
an evolved module of fear and fear learning. Psychological Review,
108, 483522.
OReilly, R. C., Munakata, Y., Frank, M. J., & Hazy, T. E. (& contributors).
(2012). Computational cognitive neuroscience [Wiki book, 1st ed.].
Retrieved from http://ccnbook.colorado.edu
Ortony, A., Clore, G. L., & Collins, A. (1988). The cognitive structure of
emotions. New York, NY: Cambridge University Press.
Oya, H., Kawasaki, H., Howard, M. A., & Adolphs, R. (2002). Electrophysiological responses in the human amygdala discriminate emotion
categories of complex visual stimuli. Journal of Neuroscience, 22,
95029512.
Phelps, E. A., OConnor, K. J., Gatenby, J. C., Grillon, C., Gore, J. C., &
Davis, M. (2001). Activation of the human amygdala to a cognitive
representation of fear. Nature Neuroscience, 4, 437441.
Plutchik, R. (1962). The emotions: Facts, theories, and a new model. New
York, NY: Random House.
Ray, R. D., Wilhelm, F. H., & Gross, J. J. (2008). All in the minds eye?
Anger rumination and reappraisal. Journal of Personality and Social
Psychology, 94, 133145.
Rolls, E. T. (2000). The orbitofrontal cortex and reward. Cerebral Cortex,
10, 284294.
Roseman, I. J. (1984). Cognitive determinants of emotion: A structural
theory. In P. Shaver (Ed.), Review of personality & social psychology,
Vol. 5: Emotions, relationships, and health (pp. 1136). Beverly Hills,
CA: SAGE.
Russell, J. A. (1980). A circumplex model of affect. Journal of Personality
and Social Psychology, 39, 11611178.

Downloaded from emr.sagepub.com by ancuta anca on October 25, 2014

Cunningham et al. Affective Dynamics 355

Russell, J. A. (2003). Core affect and the psychological construction of

emotion. Psychological Review, 110, 145172.
Sander, D., Grafman, J., & Zalla, T. (2003). The human amygdala: An
evolved system for relevance detection. Reviews in Neurosciences, 14,
303316.
Scherer, K. R. (1984). Emotion as a multicomponent process: A model and
some cross-cultural data. Review of Personality & Social Psychology,
5, 3763.
Scherer, K. R. (1988). Criteria for emotion-antecedent appraisal: A review.
In V. Hamilton, G. H. Bower, & N. H. Frijda (Eds.), Cognitive per
spectives on emotion and motivation (pp. 89126). Il Ciocco, Italy:
Springer.
Scherer, K. R. (2009). Emotions are emergent processes: They require a
dynamic computational architecture. Philosophical Transactions of the
Royal Society, Series B: Biological Sciences, 1535, 34593474.
Shannon, C. E. (1948). A mathematical theory of communication. Bell
System Technical Journal 27, 379423.
Small, D. M., Gregory, M. D., Mak, Y. E., Gitelman, D., Mesulam,
M. M., & Parrish, T. (2003). Dissociation of neural representation
of intensity and affective valuation in human gestation. Neuron, 39,
701711.
Smith, C. A., & Ellsworth, P. C. (1985). Patterns of cognitive appraisal in
emotions. Journal of Personality and Social Psychology, 48, 813838.
Smith, E. R. (1996). What do connectionism and social psychology
offer each other? Journal of Personality and Social Psychology, 70,
893912.
Smith, E. R. (2009). Distributed connectionist models in social psychology.
Social and Personality Psychology Compass, 3, 6476.
Sroufe, L. A. (1995). Emotional development: The organization of emo
tional life in the early years. Cambridge, UK: Cambridge University
Press.
Susskind, J. M., Lee, D. H., Cusi, A., Feiman, R., Grabski, W., & Anderson,
A. K. (2008). Expressing fear enhances sensory acquisition. Nature
Neuroscience, 11, 843850.

Thayer, R. E. (2012). Moods of energy and tension that motivate. In

R. Ryan (Ed.), The Oxford handbook of motivation (pp. 408420).
New York, NY: Oxford University Press.
Todd, R. M., Cunningham, W. A., Anderson, A. K., & Thompson, E.
(2012). Affect-biased attention as emotion regulation. Trends in
Cognitive Sciences, 16, 365372.
Watson, D., & Tellegen, A. (1985). Toward a consensual structure of mood.
Psychological Bulletin, 98, 219235.
Whalen, P. J., Rauch, S. L., Etcoff, N. L., McInerney, S. C., Lee, M. B., &
Jenike, M. A. (1998). Masked presentations of emotional facial expressions modulate amygdala activity without explicit knowledge. Journal
of Neuroscience, 18, 411418.
White, B. L. (1985). The first three years of life. New York, NY: Prentice Hall.
Widen, S. C., & Russell, J. A. (2003). A closer look at preschoolers freely
produced labels for facial expressions. Developmental Psychology, 39,
114128.
Widen, S. C., & Russell, J. A. (2008). Children acquire emotion categories
gradually. Cognitive Development, 23, 291312.
Williams, L. M., Liddell, B. J., Kemp, A. H., Bryant, R. A., Meares, R. A.,
Peduto, A. S., & Gordon, E. (2006). Amygdalaprefrontal dissociation of subliminal and supraliminal fear. Human Brain Mapping, 27,
652661.
Winston, J. S., Strange, B. A., ODoherty, J., & Dolan, R. J. (2002). Automatic and intentional brain responses during evaluation of trustworthiness of faces. Nature Neuroscience, 5, 277283.
Wolff, P. H. (1987). The development of behavioral states and the expres
sion of emotions in early infancy: New proposals for investigation.
Chicago, IL: University of Chicago Press.
Yik, M. S. M., Russell, J. A., & Barrett, L. F (1999). Integrating four
structures of current mood into a circumplex: Integration and beyond.
Journal of Personality and Social Psychology, 77, 600619.
Zelazo, P. D., & Cunningham, W. (2007). Executive function: Mechanisms
underlying emotion regulation. In J. Gross (Ed.), Handbook of emotion
regulation (pp. 135158). New York, NY: Guilford Press.

Downloaded from emr.sagepub.com by ancuta anca on October 25, 2014