Emotion processes and perceptual control of action choice

ABSTRACT

This editorial introduces an invited article by Andreas Eder on a new perceptual control theory of action choice, based on the comparison of real and simulated interoceptive signals generated by action alternatives. Eder extends the cognitive action-control framework, postulating a bi-directional connection between outcomes and actions by introducing “emotional feelings”, defined as valued interoceptive signals from the body. An invited commentary by Agnes Moors compares this theory with her own goal-directed theory of action control. While agreeing on the central role of a control cycle and the goal-directed nature of emotional actions, Moors disagrees on the content of the representations involved in the control cycle and the nature of the feelings involved. A second commentary by Bob Bramson and Karin Roelofs discusses the issues of the distinction between perception control vs. action control, the need for biologically plausible implementation alternatives, and potential implications for psychopathology and clinical intervention. Finally, the potential relevance of predictive coding theory and the role of appraisal processes in emotion generation with respect to their bearing on action comparison and choice are discussed.

KEYWORDS:

• Action control
• perceptual control theory
• interoceptive signals
• action simulation
• appraisal processes

Getting into an emotional state will often affect people’s behaviour. Indeed, emotion scientists have long recognised the importance of the interface between emotion and action (Darwin, 1872; Frijda, 1987; James, 1884). In recent years, the emotion-action interface has gained new actuality, as testified by the large number of theoretical proposals and empirical investigations, often centered around the notion of “action control” (see the major review by Hommel & Wiers, 2017). This topic is taken up with renewed vigour in this Theory section in Cognition and Emotion. The invited article by Andreas Eder (A Perceptual Control Theory of Emotional Action, this issue) proposes that “emotional feelings”, based on the evaluation of interoceptive signals of bodily reactions to events, are of paramount importance for action control. Specifically, Eder suggests that these feelings are integrated with the perceptual representation of actual events and of simulated outcomes of alternative response actions (based on the experience during earlier events). In the framework of perceptual control theory, as originally proposed by Powers (1973), these imagined outcomes are compared as to their desirability, with the most desirable option chosen for execution.

Eder's theoretical model of the role of emotional feeling in action control, based on the perceptual control paradigm, provides an important and valuable contribution to the increasingly popular research area of action control. The innovative character of the model is that it builds on a purely mechanistic model of action control from cognitive psychology, and extends this framework by introducing emotional feelings. Basically, modern mechanistic ideomotor approaches of action control (e.g. Hommel et al., 2001) propose a bi-directional connection between outcomes and actions that produced these outputs in the past; anticipating the outcome at a later point in time can trigger the associated action (e.g. Elsner & Hommel, 2001; but see Sun et al., 2020, 2022). In this framework, emotional feelings serve the function of selectively activating desired outcomes (compared to non-desired outcomes). This activation of desired outcomes then leads to a selective activation of those actions that will produce these outcomes.

Eder’s model should be considered as a groundbreaking innovation since it extends cognitive models of action control by adding an emotional selection mechanism that can help to understand affective, goal-directed action control (for an earlier model incorporating effects of reward into ideomotor models, see Marien et al., 2013), Furthermore, by offering a full-fledged mechanistic account of how emotional feelings can select and control behaviour, Eder’s model goes beyond existing models of emotional action control. Previous models of emotional action control typically postulated connections between specific emotional states and abstract action tendencies (e.g. fear – fight/flight, anger – aggression, gratitude – helping; Frijda et al., 1989; Plutchik, 1980), but did not specify the exact process that translates emotion into action. By combining cognitive models of ideomotor action control with Powers’ perceptual control theory, emotional feelings can be incorporated into this basic mechanistic framework.

In her invited commentary, Agnes Moors (Varieties of Instrumental Theories of Emotional Action, this issue), compares Eder’s proposal with her own goal-directed theory of emotional action. Stressing fundamental convergence between the two theories, she identifies some points of divergence concerning the nature of representations and the nature and functions of feelings, especially with regard to consciousness. Specifically, she addresses the issue of the comparator system and the types of feelings concerned – the specific role of feelings involved in action causation, the nature of instrumental actions, the degree of flexibility of the postulated mechanism, and the difference between emotional and non-emotional actions.

In their invited commentary, Bob Bramson and Karin Roelofs (Perceptual control or action-selection?, this issue) applaud Eder’s general approach but point to three non-trivial challenges that should be addressed in the future, in particular (1) the central object of control – perception or action –, (2) the underlying neural and behavioural processes, and (3) the implications for emotional disorders and potential interventions. They specifically insist on the need to specify the concrete neural, computational, and behavioural hypotheses involved as that exercise may put important constraints on theory development and could help the formulation of testable hypotheses.

As always, a new theory proposal also faces a number of challenges, concerning theoretical refinement as well as empirical validation. I will briefly review some of these. In general, Eder seems to limit the intrinsic coding of the bodily signals to (intrinsic?) pleasantness. In consequence, the account of emotional action regulation seems to mostly rely on the influence of valence, and does not explicitly consider other, maybe higher-order components of emotional appraisals that are also highly relevant for action control. However, it would seem that valence alone is not sufficient to choose an appropriate action for many situations. Appraisal theorists suggest coping potential, including control, power, and adjustment potential, as central appraisal criteria for the emotional response – and potentially successful action. Roseman et al. (1996) have reviewed the convergent predictions of several major appraisal theorists to the effect that appraisals of controllability, coping potential and power have a major effect on the emergence of emotional action tendencies. They conclude:

Having an appraisal of high vs. low control potential that determines whether one experiences a contending emotion vs. an accommodating emotion makes sense within a functionalist perspective, in which appraisals serve to elicit the particular emotion whose response tendencies are relatively likely to help a person adapt successfully to the type of situation he or she is facing. (Roseman et al., 1996, p. 262)

This is particularly important for the case of emotion disturbances. Thus, the model seems not to capture well-known inhibitory effects of low-control emotions on action (e.g. sadness and depression fostering listlessness and inactivity, or extreme anxiety leading to immobility and freezing). Although one might assume that at least some aspects of (low) control could be explained on the basis of the Eder model, based on the assumption that outcomes that have not been produced by actions of the individual in the past (i.e. outcomes that were not yet experienced as being under the control of the individual) cannot activate corresponding actions, simply due to a lack of matching action-effect episodes, this does not explain why even well-established actions are no longer activated, when an individual becomes depressed. Furthermore, the model is also mute on other aspects of emotional action control that are based on higher-order appraisals which go beyond valence (e.g. feasibility, likelihood of success, norm-conformity).

Eder suggests,

that actions are selected, instigated, and controlled by cognitive anticipations of sensory effects that were contiguously produced by the action in the past. A sensory action effect becomes selected as a goal for action if the anticipated state represents a significant improvement in comparison to the existing situation. (p. 8 in the ms)

This seems to imply that the choice of action alternatives is restricted to actions one has already performed in the past. If this were the case, the choice might be rather limited, particularly with respect to the “mental simulation of expected events” referred to earlier. This suggests that one’s action potential is restricted to actions one has performed before, which would be rather limiting, especially for children and adolescents, and prevent exploring novel activities. Another issue concerns the definition and the causal role of emotional feeling. If emotional feeling is based only on valenced interoceptive bodily signals following the event, before actions are selected, the latter cannot include interoceptive signals of action preparation (e.g. increasing muscle tension), which most emotion theorists would consider a major part of emotional feeling. For example, there are a large number of changes in the central, peripheral, and somatic nervous systems to be expected as the result of different appraisal checks (see Scherer, 2001, pp. 109–112), which can be considered as “interoceptive signals” for action control.

The conceptualization of “emotional feelings”, as “valued bodily reactions to a personally significant event”, may need some further elaboration and specification. How does the body evaluate the significance of an event for the person? Then it is suggested that the interoceptive signals are integrated with ongoing perceptual activities from other senses (vision, audition, touch, etc.), “which code for” (i.e. “evaluate”?) whether a perceptual event feels good or bad at a given instant in time. This seems to mean that valence appraisal is performed by the senses. The underlying mechanism would need further elaboration. Also, especially in the case of mentally simulated actions, it would need to be described how potentially relevant action tendencies are chosen and where the respective interoceptive signals are stored.

Eder’s action control loop seems to consider a momentary assessment following an event, to achieve action control. However, emotions (and the respective feelings) are often processes, given that situations and evaluations can change fairly rapidly, requiring frequent reappraisal. This may be of major consequence for action control theory. For example, it might be necessary to examine in which form such sequential changes in interoceptive signals linked to an earlier action are stored and retrieved.

Finally, a number of definitional issues need to be raised. Eder opposes a binary classification of actions into instrumental versus emotional as “the action-generating machinery is the same for generated actions during emotional and non-emotional episodes and only the mode of operation will differ between these episodes”. It is of course reasonable to expect gradual differences, especially as instrumental actions can be produced with underlying emotional feelings (except for cold-blooded murder), but it might still be useful to provide definitions of the endpoints of the continuum. This is particularly true for the emotional end. Given that there are substantial differences between the emotion definitions suggested by major theorists, it would be valuable to have the author’s working definition, given the frequent use of “emotional”.

Another issue concerns the use of “perceptual”. Given that Eder’s theory is based on Powers’ (1973) Perceptual Control Theory (PCT), which highlights perception as the phenomenon to be controlled, it is understandable that the term will be used frequently. However, for readers not familiar with Powers’ approach, there is a danger of misunderstanding. The common definition of “perception”, in the APA Dictionary of Psychology (VandenBos, 2015), runs as follows:

the process or result of becoming aware of objects, relationships, and events by means of the senses, which includes such activities as recognizing, observing, and discriminating. These activities enable organisms to organize and interpret the stimuli received into meaningful knowledge and to act in a coordinated manner.

In other words, presumably perception is generally considered as the basis for cognition, which adds interpretation and meaning.

It will be of interest to compare Eder’s PCT to other emotion and action control (PC) theories, in particular the Predictive Coding tradition (Friston, 2008, 2012; Rao & Ballard, 1999). Whereas PCT attempts to account for controlling perceptions by comparing current interoceptive feedback with estimated feedback from mentally simulated events, PC focusses on prediction error reduction. Specifically, the assumption is that the brain continually generates models of the world based on context and information from memory to predict sensory input. A predictive model is created in higher cortical areas and communicated through feedback connections to lower sensory areas. In contrast, feedforward connections process and project an error signal, i.e. the mismatch between the predicted information and the actual sensory input (Rao & Ballard, 1999). The predictive model is constantly updated according to this error signal (see https://www.frontiersin.org/research-topics/599/predictive-coding).

A PC framework for emotional feeling has been developed by Anil Seth (2013) who proposes a predictive, inferential perspective on interoception:

interoceptive inference conceives of subjective feeling states (emotions) as arising from actively inferred generative (predictive) models of the causes of interoceptive afferents. The model generalizes ‘appraisal’ theories that view emotions as emerging from cognitive evaluations of physiological changes, and it sheds new light on the neurocognitive mechanisms that underlie the experience of body ownership and conscious selfhood in health and in neuropsychiatric illness. (Seth, 2013, p. 565)

The issue of action control has recently also been addressed by Marshall et al. (2018):

… we implicate interoceptive predictions in the generation of subjective motor-related feeling states. Furthermore, we propose a distinction between reflexive and pre-reflexive modes of agentic action control and suggest that interoceptive input may affect each differently. Finally, we advocate the necessity of continuous interoceptive input for conscious forms of agentic action control. (Marshall et al., 2018, abstract)

In their commentary, Bramson and Roelofs provide some very pertinent examples for the role of outcome predictions, referring to the PC literature. They also highlight the neglected issue of individual differences and psychopathology in the domain of action control. Recently, Smith et al. (2021) have described advances in the application of predictive coding and active inference models within clinical neuroscience. They review the neural circuit architectures that can accomplish these predictive functions and show how the associated processes may break down or become aberrant within clinical conditions (Smith et al., 2021).

While many emotion theories include action tendencies as a major component of emotion, they rarely discuss how these action tendencies are elicited. Eder’s PCT remains silent about the choice of actions, current or simulated, that are controlled. In my own appraisal theory of emotion (Component Process Theory; Scherer, 2001, 2009) emotions are conceptualized as recursive process of an evaluation of a series of appraisal criteria in a given sequence (as determined by the need of preceding input). The results of each of the individual criteria appraisals will produce changes in other components of the process, such as action tendencies (including facial, vocal, and bodily expressions and physiological changes). As to action tendencies, the following list illustrates the nature of the expected patterning (see also Scherer & Moors, 2019, pp. 724–729):

• Novelty: novel and goal relevant: orienting, focussing/ alerting

• Valence: Pleasant: Incorporation/Recommending // Unpleasant: Rejection/Warning

• Conduciveness: High: Relaxation/Stability // Obstructive: Activation/Reactivity

• Control/Power: None or low: Withdrawal, Submission // High: Assertion/Dominance

• Normative conformity: High: Relaxation, Bolstering // Low: Activation, Self-consciousness

As mentioned above, Eder focusses mostly on valence – pleasantness and desirability whereas it seems that to decide on potentially successful actions, other criteria are even more important, in particular control and power. In fact, Eder also acknowledges the possibility of striving “for unpleasant emotions if instrumental for the task at hand, for example, when preparing for a confrontation” (ms p. 23).

In consequence, it may be useful to extend action control theories by using a larger set of evaluation criteria. It seems possible to expand the idea of comparing alternative actions that are mentally simulated with the help of predicted profiles of appraisal results and thus determine the choice of an optimal outcome for the action decision. The difficulty might be to maintain the assumption that these processes are exclusively mediated by bodily signals. While appraisal theories predict action tendencies as well as physiological reactions as a result of specific checks, the underlying processes might be too complex to expect action-specific patterns. This is all the more the case, as some of these bodily changes are likely to occur only during the execution of an action which would mean that all action candidates would need to have already been performed before.

As mentioned by Bramson & Roelofs, it is important to also consider the reappraisal mechanism, especially with regard to individual differences and potential dysfunctions of the action control. A central issue is to what extent the mentally generated predictions are accurate (see Sterzer et al., 2018). In the context of appraisal research, we have been able to show that there are potent appraisal biases (such as abnormally low control beliefs) that can create stable emotion dispositions (such as frequently recurring sadness or anxiety), which present important risk factors for depression and generalized anxiety (Scherer, 2021; Scherer et al., 2022). In general, the importance of beliefs and values for the action control computations requires more attention in this domain.

In conclusion, Eder’s theoretical proposal serves the important function to remind emotion theorists to consider the effect of emotion on action (and vice versa) in greater detail. However, to allow useful comparisons between different theoretical approaches and agree on appropriate experimental procedures, more stringent working definitions of terms such as emotion and feeling, and probably also perception, cognition, and action, will be required.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.