Summary: Russel (2003) – Core Affect and the Psychological Construction of Emotion

This is my summary of: Russell, J. A. (2003). Core affect and the psychological construction of emotion. Psychological review, 110(1), 145. In it Russel proposes a framework to talk about emotion at the psychological level, and aims at avoiding folk psychological concepts in his approach.

Basically, he argues that many traditional approaches have been founded on everyday concepts and categories of emotion, e.g. fear, anger and happiness (he primarily hits on any form of categorical approaches). These terms — as well as “emotion” itself — form the language that is used to talk about emotional phenomena, but they may not be the best starting point for scientific investigations. He argues to test them for their usefulness of accurately describing the topic, and hints at the disagreement on a definition for emotions: Are they biologically given modules or socially constructed roles?

Research often focuses on prototypical cases („blue ribbon“ emotions), but these may not be representative. In many cases the same emotional labels refer to very different manifestations and behaviors (e.g. afraid from the bear results in running, being afraid of the spider results in squishing it, etc).

Russel suggests using the word „emotion“ to cover the general topic, but not giving it any scientific or definitional meaning, and argues against trying to account for all emotional phenomena under a specific emotion label in the same way. He proposes a framework that combines traditional theories of dimensional and categorical views on emotions. William James‘ suggestion, that self-perception of automatic processes are vital in emotions, plays a key role in his synthesis. He focuses on accounting for prototypical emotional episodes.

Every framework consists of proposed primitives that account for more complex phenomena, and Russel describes basic emotions as traditionally important in this role. However, for him they imply a form of intentionality (they have a content): fear is the fear OF something, anger is anger ABOUT something, etc. He argues that this makes them not suitable as primitives, because they already are complex phenomena due to this cognitive structure. Also, he argues, that they cannot really account for non-directed emotional phenomena, e.g. moods. Russel proposes two different kind of primitives instead: A form of free-floating Core Affect (CA), consisting itself of the two dimensions of pleasure-displeasure (Valence) and activation-deactivation (Arousal), together with the perception of affective qualities (PAQ) of objects.

Core concepts

  • Core Affect (CA): A neurophysiological state that is conciously accessible as a simple, nonreflective feeling that is an integral blend of hedonic and arousal values (pleasure and arousal). It is an evaluation of ones current condition. CA is principally not directed at an object, but can become bound through attribution, It is an irreducible primitive and can always be expressed as a point in PA space.It is similar in perception to temperature: Simple, non-cognitive (not directed at an object), experience, but complex on a biological level, just like body temperature. Some feelings add a cognitive component to CA (e.g. Pride). There is always only one CA present in a subject, and it can manifest itself in conscious differently under varying circumstances (sometimes in relation to an object, sometimes as free floating mood). It varies in intensity and depending on this also presence in consciousness. It may even recede entirely from conscious thought, but forms an ever present background.CA is best envisioned as having a baseline defined by subconscious internal factors (hormones, immune reactions, etc) and external causes operating on it. Thus, many factors play together in shaping CA — some accessible by conscious, others not –, with occasionally a single strong external factor dominating the picture (He refers to the bear, that changed CA from tranquility to distress). Changes in CA can be brought about by any object in the mental life, independent of whether it is a perception or an imagination (Virtual Reality hypothesis). Russel stresses the importance of the complexity of causation of CA and the limited ability of subjects to keep track of what causes a it to change.He identifies the function of CA in guiding cognitive processes according to mood congruency: A change in CA triggers a search for the reasons behind it, and in doing so locates attention to „like-valenced material“. Positive CA makes a situation remembered/encountered/envisioned seem also positive and has therefor an impact on decision making through the formation of preferences and attitudes (important in relation to AR).
  • Affective Quality (AQ): The ability to cause a change in a subjects CA, and are described in the same PA factors.The world is perceived and envisioned with affective qualities attached (boring movie, nice day, interesting people, etc). All of these qualities accessible to conscious at a given moment have an impact on subsequent reactions. AQ is a property of an object that relates to its possibility to change CA.
  • Perception of Affective Quality (PAQ) is a perceptual process that is bound to a specific stimuli and assesses its AQ. This process may impact CA but it does not have to: It can change on its own without any external cause, and AQs can be perceived without changing CA itself (Russel gives an example of a depressed patient that admits the beauty of a sunset, without being shaken out of his mood).
  • Attributed Affect (AA): Change in CA that is attributed to a specific object. Is usually the result of an automatic process, but can also happen on a conscious level. Russel describes it despite it complex structure as „phenomenologically simple“ (a feeling of being afraid of the bear, or of liking a new tune) and gives three necessary and jointly-sufficient features for AA: (1) change in CA, (2) presence of an object, and (3) attribution of CA to the object. This process is not impacted by the reality of the object in question (Virtual Reality hypothesis). Attribution is the creation of causal chains in events and is impacted by individual and cultural idiosyncrasies. At the subjective level attributions usually seem correct. AA is in Russel‘s view responsible for many different topics such as „affective reactions, liking, displeasure motives, and empathy”. AA follows two functions beyond CA: (1) guides attention and behavior towards the object, and (2) AA is the main route to the affective quality of the object
  • Affect Regulation (AR): Actions with the aim of changing a subjects own CA. Russel uses examples of self-rewarding behavior (morning coffee, evening brandy. AR seems to be generally related in choosing actions such as to increase pleasure, and minimize displeasure. Explicitly not in relation to a specific object.
  • Object: Can be persons, condition, situations („psychological representation“), that may be fictional or situated in the future/past.

Secondary concepts

  • Mood: Ongoing undirected CA, possesses no clear definition of duration
  • Empathy: Attributed affect caused my mental simulation of the experience of another
  • Prototype: A cognitive structure that specifies typical structure, order and pattern of an emotional concept
  • Prototypical emotional episode (PPE): Emotional episode that counts as an excellent member of a category, because it is close to the prototype
  • Emotional episode (EE): Occurring event that relates to an emotional prototype and counts as an instance of it.
  • Emotional meta-experience: Self-perception of CA in terms of emotional concepts
  • Emotion regulation (ER): Contrasts with AR. Tries to alter the Emotional Category oneself is in (requires already an emotional meta-experience, in contrast to affect regulation), and is impacted a larger order of knowledge, such as social norms and roles (e.g. a man should not feel fearful.
  • Mood-congruent priming: CA makes available more information related to the same valence, such as one sees more positive qualities in an object when one is happy, thus overestimating the pleasure related to it.
  • Misattribution: A change in CA is attributed to a source that is not actually causing the change, e.g. when happy mistakenly attributing the pleasant feeling to an object that is not actually causing it, leading also to an overestimation of its positive aspects.

The Structure of an Emotional Episode

Russel argues that CA alone can account for emotions in the same way that a normal dimensional theory of emotion can (e.g. his own pleasure-arousal dimensional model), but on themselves this still excludes a great many of EEs. In his framework he describes the components of a prototypical EE in the following way:

  • Antecedent Event: An obvious external causal event creates a complex psychological representation
  • Affective Quality: The AQs of the event are perceived
  • Core Affect: As a reaction CA changes drastically from its initial state, at times before the event itself has been registered on a conscious level. CA changes throughout the development of EE and impacts other components.
  • Attribution: The change in CA is attributed to the event, so that it becomes the object, providing the subject with the experience of feeling a certain way BECAUSE of the object.
  • Appraisal: Perceptual-Cognitive processing of the objects relation to the perceiving subject‘s goals, its place in a causal chain, etc. Mood congruent information is more accessible than other.
  • Instrumental Action: Behavior is displayed, if it is perceived as a problem or opportunity that requires action. Pleasure and displeasure act as quantification mechanism and may cause approach or withdrawal preparations. Specific actions depend on the evaluation of the circumstances involved, and the relations to plans and goals. No specific emotional actions need to be recruited for a specific emotion.
  • Physiological and Expressive Changes: Facial expressions, changes in prosody, and autonomic reactions manifest themselves, and are explainable in terms of CA or as preparation/recovery of instrumental actions. Russel argues that there is no unique pattern specific for every discrete emotion.
  • Subjective Conscious Experiences: Additionally to the conscious experiences of CA and AQ there are also „metacognitive“ aspects involved, such as judgment of urgency, feelings of indecision and confusion, rendering the EE as seemingly beyond deliberate control, and rendered emotional through CA.
  • Emotional Meta-Experience: Subjective conscious experience of classifying the EE as a specific emotion (e.g. the person notices she is happy, or afraid or angry). This meta-experience is a self-perception in the form of a categorization of ones inner state, based on the other features of the EE and labeled in folk concepts (in concepts of language). These categories are to be envisioned as being structured according to prototype theory.

Russel explicitly states that this list only describes a prototypical case, and that many EEs can vary significantly in the way the different components are ordered and what role they play. He describes cases of misattribution (an object is not the antecedent), atypical appraisals (e.g. a fear of something known to be harmless), or thrill seeking (extreme CA before the antecedent appears), etc.

He sees traditional theories as based on emotions as entities in their own right, that are caused by something and then produce the different components as manifestations. In his view these theories are created in the very purpose so they create a coherence among features in the form of discrete categories. Emotions in this view serve as explanations for the observable features, e.g. “happiness” causes smiling. In Russel‘s proposed framework, emotional episodes consist of two steps: (1) manifested components occur without being accounted for by an emotion entity, and (2) components are categorized by an observer (either the subject having the EE at the meta-experience stage, or a third person observing the episode. Both need not come to the same categorization).

Russel uses the term psychological construction to hint at the complexity of emotional episodes. He stresses that components are not placed in a fixed package, neither due to biological (basic emotion theorists view) or social (social constructionist theorists view) causes. The categories involved are not fixed and static, i.e. defined by necessary and sufficient criteria that are somehow given, but happen in a gradual comparison of the displayed features to a prototype — „Resemblance is an external fact about the components, not an internal mechanism that joins them into a package“.

He further compares this loosely based connection between emotion-labels and the display of features with the traditional view in which they are highly correlated across all instances and especially inter-correlated among one-another. He argues that in the latter, e.g. happiness needs to be always accompanied by all necessary and sufficient features to be explained. This makes it difficult to account for odd combinations of features, because any non-prototypical cases lack good explanations. A prototypical view does not place harsh boundaries along the lines of strictly emotional or non-emotional on the classification, but allows for a gradual resemblance, that is based on many different connections to a prototype. Additionally, in his understanding there is no fixed causal order to an emotional episode, as many traditional views would have it. Instead, the different components often overlap and are generally ongoing. Russel draws from this the conclusion that there is no need for postulating a mechanism that accounts for all components of an emotional episode at once, because there is non. Instead their coherence stems from the classification and selection based on the folk concept: What makes a certain emotional episode fear, is its resemblance to a prototypical case of fear.


Please note that this summary represents my own interpretation of the original article, and as such might leave out important aspects or even contain errors. It does not attempt to replace reading the original article in any way and is primarily intended as a memory aid for myself.