UvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl)
UvA-DARE (Digital Academic Repository)
Avoidance: From threat encounter to action execution
Arnaudova, I.
Publication date
2015
Document Version
Final published version
Link to publication
Citation for published version (APA):
Arnaudova, I. (2015). Avoidance: From threat encounter to action execution. Boxpress.
General rights
It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).
Disclaimer/Complaints regulations
If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible.
Chapter 2
Threat avoidance response
selection and execution (TeARS):
A comprehensive model of
avoidance
A version of this chapter is in preparation as:
Arnaudova, I., Craske, M., Fanselow, M., Kindt, M., & Beckers, T. (2015).
Threat avoidance response selection and execution (TeARS): A comprehensive model of avoidance.
Abstract
Dysfunctional avoidance lies at the core of anxiety pathology. Here, we pro-pose a model for threat avoidance response selection and execution (the TeARS model), which describes the processes involved in avoidance from threat encounter to final behavioral output. In our model, which incorporates empirical research and theories from animal, social, clinical and experimental psychology research, avoidance is an umbrella term encompassing a large variety of defense behaviors observed in human threat responding. Our framework centers on the importance of threat imminence appraisal and automatic avoidance tendencies and the no-tion of two behavioral control systems, one that is reflective and another that is reflexive. Further, we propose five pathways through which adaptive avoidance can become dysfunctional in clinically severe anxiety.
“Never confuse movement
with action“
Ernest Hemingway (quoted in Hotchner, 1966)
Since Confucius (475 BC) and Aristotle (350 BC), philosophers and scientists have been fascinated by the behaviors of individuals, the motivational forces be-hind them, their purpose and (ir)rationality. However, action or lack thereof in emotional situations remains poorly understood despite being one of the primary contributors to functional impairment in individuals suffering from a number of psychiatric disorders. In particular, behavioral avoidance can lead to substan-tial limitations in daily life for people with clinically severe anxiety (Rachman, Craske, Tallman, & Solyom, 1986). The link between threat and avoidance behav-ior has been addressed in a number of theories (e.g., Bolles, 1970; Lovibond, 2006; Mowrer, 1939), but empirical assessment of the relationship is limited (Grillon, Baas, Cornwell, & Johnson, 2006). A recent spur of research on the topic has highlighted the need to enhance dialogue among researchers from various disci-plines, who often define avoidance in fundamentally different ways (Eder, Elliot, & Harmon-Jones, 2013; Lovibond, 2006). A common conceptual framework for understanding avoidance in the context of threat is needed. The purpose of this review is to evaluate existing theoretical assumptions and empirical data on the relationship between threat and avoidance and to propose an evidence-based inte-grative model for threat-avoidance responding. Our Threat Avoidance Response Selection and Execution (TeARS) model addresses the selection and execution of avoidance responses based on threat imminence appraisal. We also consider the ways in which avoidance response selection and execution is dysregulated in the context of clinically severe anxiety. We believe that the TeARS model will gener-ate ideas for basic and applied research into the mechanisms underlying avoidance responding and will eventually help to improve interventions for emotional disor-ders.
Our paper begins with an operationalization of avoidance responding based on human and non-human animal research and theorizing to date. Next, we examine animal models of behavior, which elucidate the guiding forces behind response selection in the context of threat. We review evidence for a threat imminence model of response selection and the findings from human studies that support a role of threat imminence in human avoidance response selection. We then turn to empirical data from human laboratory studies regarding automatic action tenden-cies, including avoidance tendentenden-cies, as well as overt avoidance behavior, spanning from withdrawal and flight to threat endurance with or without the aid of safety behaviors. We introduce dual process models of behavior, where action control is governed by the interaction between two systems, a reflexive (automatic) and a reflective (controlled) one. Within the TeARS model, we incorporate elements of threat imminence appraisal, automatic action tendencies (as part of a reflexive mode of responding) and reflective control of overt behavior under conditions of less imminent threat. Finally, we examine the effects of clinically severe anxiety on avoidance response selection and execution and suggest five pathways through which avoidance might become pathological. Throughout the paper, we provide indications about further research and highlight the gaps in our current
under-standing of avoidance.
2.1 Definition of avoidance
The wide-ranging usage of the term avoidance among social psychologists, clin-icians, learning theorists and laymen impedes scientific progress and conceptual advancement. Therefore, before embarking upon a detailed analysis of threat avoidance response selection and execution, a clear definition of avoidance is war-ranted (Lang & Bradley, 2013; Lovibond, 2006). In this section, we evaluate common elements of definitions of avoidance from different theoretical perspec-tives. We then derive a new definition of avoidance behavior in response to threat that applies to healthy as well as anxious individuals.
Dictionary definitions state that to avoid is “to prevent something bad from happening” and that to avoid is “to stay away from someone or something, or not use something” (“Avoid”, 2014). In psychology, the notion of avoidance plays a central role in theories of fear learning. There, avoidance is defined as a response that serves to prevent an unpleasant event (i.e., an unconditioned aversive stimu-lus, US, such as an electric shock) from occurring in the future (Lovibond, 2006). The overlap with the first meaning of the dictionary definition is clear. Early theorists proposed that the avoidance response was reinforced by termination of a warning signal (i.e., conditioned stimulus, CS, such as a neutral tone) that preceded the aversive event in an aversive Pavlovian conditioning situation (e.g., Mowrer, 1939). The term escape referred to any response that resulted in the ter-mination of the US following its onset (Lovibond, 2006; Mowrer, 1939). However, subsequent research by Bolles (1970) showed that termination contingencies play a minimal role in motivating any defensive behavior in animals, including avoidance, thus undermining the idea that instrumental learning (acquiring the knowledge that a specific action has a particular outcome) is crucial for the acquisition of avoidance responding.
Instead, Bolles (1970) suggested that animals have a range of pre-wired species-specific defense responses, including freezing, fleeing, and fighting1, which are
activated in the presence of threat, because they have a phylogenetic association with survival. These behaviors were judged to have persevered from generation to generation due to their functionality, but the specific behaviors selected for expression were influenced by the characteristics (or affordances) and demands of a given situation (Bolles & Fanselow, 1980). In subsequent elaborations, physical or psychological distance from threat (termed predatory/threat imminence) was regarded as the most important guiding factor for the selection of the appropriate form of defensive behavior, with the affordances of a given situation or termination of an aversive stimulus playing a minimal role in response selection (Fanselow & Lester, 1988)(see section 2.2.1). This reconceptualization brought the use of the term avoidance in the psychological literature closer to the second dictionary meaning of the term to avoid.
Clinical or diagnostic definitions of avoidance, such as the one provided in 1New responses could be learned through instrumental contingencies, sometimes with
diffi-culty, if the pre-wired behaviors do not fulfill their function (Bolles, 1970). Opposite responses (e.g., moving away when moving towards is the natural response) seem, however, almost impos-sible to acquire (Hershberger, 1986).
the Diagnostic and Statistical Manual of Mental Disorders - 5thedition (DSM-5;
American Psychiatric Association, 2013), similarly emphasize the distance from threat signals: “the act of keeping away from stress-related circumstances: a ten-dency to circumvent cues, activities and situations that remind the individual of a stressful event experienced.” It is important to note that in contrast to ear-lier conceptualizations in psychology, this definition of avoidance has no response contingency component. The diagnostic criterion of avoidance for most anxiety disorders further clarifies that if a threatening or stressful stimulus is not actively avoided, it is endured with significant distress (American Psychiatric Associa-tion, 2013). Endurance of threat is usually accompanied by an array of actions that serve the function of safety seeking (Thwaites & Freeston, 2005), such as remaining in proximity to a trusted companion, superstitious objects, or certain medications. These behaviors are often coined safety behaviors, because they increase the subjective perception of safety.
Social psychology researchers utilize a much broader concept of avoidance mo-tivation, which can be elicited by any negative stimulus rather than threat stimuli in particular (e.g., Chen & Bargh, 1999). Nonetheless, distance from the nega-tive stimulus is again viewed as underlying automatic avoidance tendencies, as described in more detail below (Krieglmeyer, De Houwer, & Deutsch, 2013). Sim-ilarly, emotion theorists have suggested that behaviors are primarily modes of distance regulation (Frijda, Kuipers, & ter Schure, 1989) and that any behavior that increases the distance from a negative stimulus can be seen as avoidance.
What these definitions and conceptualizations all share is the idea that avoid-ance increases the distavoid-ance of the agent from negative stimuli more broadly or threatening stimuli more specifically. Given our goal of developing a model that explains avoidance response selection and execution in the presence of threat, we propose the following definition of avoidance: “any covert or overt action that functions to physically (spatially or temporally) or psychologically distance the agent from perceived or actual threat.” We focus on avoidance as having a spe-cific function of distancing from threat, which may include actions that increase psychological distance without affecting physical distance. The concept of psy-chological distance in animal models of response selection pertains to two main variables: the type of predator (e.g., big or small) and the direction of its move-ments (e.g., towards or away; Fanselow, 1997). We later elaborate on the factors appraised during threat imminence estimations in humans (see section 2.6). An example of a human action that serves to increase psychological distance without changing the physical properties of threat is the ascertainment of the presence of a friend when one is under the threat of having to walk the street at night, for example. In the same scenario, an avoidance response that changes the physical distance from threat is to stay off the streets at night altogether. Thus, our def-inition encompasses safety behaviors, which function to psychologically distance oneself from threat and increase feelings of safety, as well as withdrawal and flight behaviors, which function to increase physical distance and might result in the cancelation of threat.
In contrast to the early models of avoidance learning proposed by Mowrer (1939) and Miller (1941), our definition does not make reference to reinforcement of the current avoidance response. Feedback about whether the current avoidance action has fulfilled its function does not occur before the response is executed,
and thus cannot affect its selection. We do not differentiate between avoidance and escape at the point of response selection. A cognitive representation of the anticipated consequences of a particular behavioral response (Lovibond, 2006), however, might play a role in the decision making process for avoidance response selection (as described in more detail in section 2.6). The perceived consequence in some circumstances may be based on the success of prior responses undertaken under similar circumstances (Bouton, Winterbauer, & Todd, 2012). Overall, the instrumentality of the behavior is not of primary concern within our definition, because not all avoidance behaviors are instrumentally motivated (for further clarification and examples of actions for which instrumental learning might (not) play a role, see later sections).
Finally, we maintain that the specific form of the avoidance behavior can be shaped by the specific threat. Thus, using our definition, avoidance can occur in response to a CS (as a warning signal of future threat) or a US (as a present threat). The concept of threat in our model is operationalized as any object, person or event (internal or external) that might endanger one’s physical health (e.g., a weapon that can inflict wounds) or psychological well-being (e.g., an event that can lead to financial losses and disappointment). We acknowledge that other behaviors, such as fighting or bonding with other individuals (S. E. Taylor, 2006), can also play a role in dealing with threat (Fanselow & Lester, 1988), but possibly under very specific circumstances. Here, we focus on the phenomenon of threat avoidance, because of its clinical significance; we address in subsequent sections how the occurrence of other behaviors can be explained through the TeARS model as well.
2.2 Response selection to threat
In this section, we address the issue of response selection in animals and hu-mans, and use findings from research on response selection as a foundation for the TeARS model. The learning of avoidance responses has been the focus of multi-ple associative and instrumental process theories (e.g., Lovibond, 2006; Mowrer, 1939). Avoidance learning research focuses on specific factors that influence the expression of a particular response towards a particular stimulus. An in-depth discussion of avoidance learning is beyond the scope of this review (see Bolles, 1972 and Krypotos, Effting, et al., 2015). Instead, we address the processes that take place from the moment of stimulus encounter to avoidance expression and the determination of the form of the avoidance response.
2.2.1 Evidence from animal research
Defensive responding and avoidance has been examined in-depth in a number of species (e.g., horses, Keiper & Berger, 1982; birds, Schaller & Emlen, 1962; dogs, Solomon & Wynne, 1953). Research in rats, in particular, has resulted in well-specified threat response-selection theories (Fanselow & Lester, 1988). Thus, we turn first to results from animal studies for clues about important factors in threat avoidance selection.
An immediate action is mandated in the presence of threat in order to preserve chances for survival (Bolles & Fanselow, 1980; Fanselow, 1997; Ozono, Watabe, &
Yoshikawa, 2012). Initially, laboratory animal research studied avoidance in rela-tion to threat intensity and reinforcement schedules (Miller, 1941; Mowrer, 1939). However, theoretical propositions about the role of reinforcement contingencies of defensive behavior (avoidance, escape and fear reduction) did not account fully for the variance observed in natural animal behavior. This led Bolles (1970) to propose his influential species-specific defense response (SSDR) theory. Ac-cording to the SSDR framework, an animal’s behavioral repertoire following a threat encounter is limited due to the activation of a defensive motivational net-work. The available defense reactions include fleeing, freezing and fighting even before any behavioral learning has taken place. Other, incompatible behaviors are suppressed due to the dominance of the defensive motivational network. For example, pain-motivated recuperative behaviors (such as resting and healing) are inhibited through analgesia, because these behaviors are incompatible with defen-sive responses (Bolles & Fanselow, 1980). The SSDR model further posits that defensive behavior in reaction to a threat encounter is primarily guided by extrin-sic feedback from the environment regarding whether the situation has changed as a result of response execution, rather than by the actual termination of threat (Bolles, 1970). The SSDR theory thus represented a major shift from earlier mod-els of avoidance as an action that prevents the occurrence of a negative event to avoidance as the execution of any of a range of evolutionary pre-wired behaviors. In an expansion of the SSDR framework, Bolles and Fanselow (1980) included the behavior of animals following their return to safety. When preservation has been achieved but defense has not been optimal and the animal has been injured, the animal enters a recuperative stage where injury is of primary concern. Once recuperation is complete, the animal returns to its preferred activity pattern until another (potential) threat is detected.
The combination of the perception-defense-recuperation loop (PDR; Bolles & Fanselow, 1980), the SSDR framework, and laboratory animal research with varied threat intensities and reinforcement schedules, led to the conclusion that predatory imminence is the primary determinant for threat response selection (Fanselow, 1997; Fanselow & Lester, 1988). Predatory imminence is determined by the temporal and geographic position of the predator relative to the prey and by the prey’s perception of the predator’s psychological distance, based on the type of predator and the direction of its movements. Behavior is thus oriented towards mitigating threat imminence by maintaining distance from the predator and allowing the animal to return to its preferred pattern of activity in the absence of predatory potential (Fanselow & Lester, 1988).
In the predatory imminence model, activation of the defensive motivational network at different levels of the threat imminence continuum constrains the be-havioral repertoire of the animal and determines the form of the behavior being executed (see Figure 2.1). As long as no threat has been encountered, the animal organizes its behavior so that the risk of encountering a predator is controlled. For example, animals will reduce the number of meals, while increasing meal size, in order to maintain a balanced energy intake, when food procurement is associated with an increased probability of predatory encounter (Fanselow, Lester, & Helm-stetter, 1988). Behavior changes dramatically once threat has been detected. For the rat, the predominant behavior at this stage is full-body freezing. Later, upon close contact with the threat, circa-strike behaviors characterized by an activity
burst and loud vocalizations occur. Once the animal has returned to safety, if injury has been sustained the animal engages in recuperative behaviors aimed at promoting healing. This theory retains ideas from the SSDR framework in which behaviors that are inappropriate for the degree of threat imminence are inhibited in order to give way to appropriate behaviors (Fanselow, 1997). In addition, the theory offers a specific and testable framework regarding the factors that guide threat-related behaviors in rats and has received extensive empirical support over the years (for summaries, see Fanselow, 1994; Fanselow & Lester, 1988).
The most extensively investigated of those animal defensive behaviors is freez-ing, which consists of muscle tension and inhibition of all voluntary movement2
(Fanselow, 1984). This response occurs when a threat signal (CS) is encountered (i.e., threat detection), and has been studied from an associative learning per-spective. That is, Pavlovian conditioning procedures, where the animal is trained to associate a CS with an aversive US, are widely used in both animal and hu-man research because they elucidate processes that underlie the learning of threat signals.
Freezing occurs in safe places, meaning that the animal transitions from the position where the threat is detected to a quickly and easily accessible safer posi-tion where it freezes. Locomoposi-tion from the place of threat-detecposi-tion to the place of freezing can be regarded as a form of flight behavior (Fanselow, 1997; Fanselow & Lester, 1988). In fact, a form of flight has been observed within most de-fense reactions along the predatory imminence continuum and flight may play an important role in facilitating other defensive responses (Fanselow, 1997). Flight behavior physically increases the distance between the predator and the prey, thus clearly representing an avoidance response. Freezing also can be seen as an avoid-ance response since it diminishes the possibility of prey detection, thus providing protection, and therefore reduces the probability of aversive impact. As seen at the stage of pre-encounter, animals shape their behavior so that they can maxi-mize threat distance and consequently, minimaxi-mize interference with the preferred activity pattern.
An animal is able to rapidly switch between two avoidance response types; for example, from freezing (post-encounter) to an activity burst (circa-strike) and back (Fanselow, 1994; Fanselow, DeCola, & Young, 1993; Mobbs et al., 2007). When a warning signal is detected, threat imminence rises from the level of no predatory potential and the rat freezes, but the moment the negative consequence of that warning signal occurs (e.g., shock) the rat immediately engages in circa-strike behaviors (Fanselow, 1982) to return to freezing thereafter. Such rapid switching would only have been preserved through evolution if it offered a com-petitive advantage. The presence of fear-potentiated startle during freezing also suggests that the animal is prepared to rapidly engage in circa-strike behaviors at any moment, because startle has the ability to interrupt ongoing activity (Gra-ham, 1979) and allow the initiation of a more appropriate behavior. This offers support to the notion that a flight tendency is present even if the circa-strike be-havior of active flight is not being executed. That is, the threat imminence model suggests that the animal can be in a state of readiness for flight even when overt flight is not observed. The notion of a readiness to act is a critical factor in our TeARS model.
Figure 2.1: Threat imminence continuum and response in the rat after Fanselow and Lester (1988) No predatory potential Predatory potential Predator detected Predator makes contact Predator makes the kill
Preferred activity pattern of nonaversively motivated behaviors Pre-encounter defensive behavior Post-encounter defensive behavior Circa-strike defensive bahavior Point of no return Recuperative behavior BEHAVIORAL CONSEQUENCE Increasing predatory imminence LEVELS
In summary, animal research suggests that threat activates a defensive mo-tivational network. Avoidance in the presence of threat is seen as evolutionarily predisposed, can be automatic and does not need shaping by instrumental learn-ing contlearn-ingencies. Thus, the organism is prepared for the execution of avoidance at the moment of threat encounter. Hereafter, we refer to such automatically primed readiness for avoidance as an avoidance tendency. The specific form of the response to threat depends on the imminence of threat, which can be actual (such as when a US is present) as well as anticipated (such as when a CS is present). Animals are prepared to engage in the extreme form of defense or avoidance (fight/flight circa-strike behaviors), but under lower levels of threat imminence, these are inhibited and other forms of threat avoidance (e.g., pre-encounter de-fense) are executed. We now examine the empirical support for similar phenomena in humans.
2.2.2 Evidence from human research
Threat stimuli are presumed to activate a defensive motivational network, which improves survival chances through ensuring protection, in contrast to positive stimuli that are presumed to activate an appetite motivational network, which
ensures survival through supporting food procurement and procreation (Dickin-son & Dearing, 1979; Lang, 1995). These two opposing networks elicit opposing underlying action tendencies: approach and avoidance (Frijda, 1986) aimed at pro-tection or procurement, respectively. Responding guided by these networks can occur in three distinct domains: language-based subjective report such as ver-balized emotional experiences, psychophysiological responding such as the startle response, and behavior indicative of approach or avoidance (Lang, 1995; Lang & Davis, 2006).
Activation of the defensive motivational network in humans has been widely studied through the observation of physiological responses during passive viewing of pictures of different valence (Lang et al., 1997). It has been shown that defensive responses are primed when individuals are faced with stimuli of negative valence (Lang, Bradley, & Cuthbert, 1990). Multiple studies have demonstrated that star-tle is increased when humans passively view negative relative to positive images or neutral images (Lang, 1995; Lang et al., 1997). Potentiated startle in humans is a defensive response, which can interrupt ongoing behavior (Graham, 1979) that might be inappropriate to the level of threat present. Even though it has multiple components (Blumenthal et al., 2005), potentiated startle is commonly measured in the laboratory as the intensity of the reflexive eye blink to loud auditory stimuli (Blumenthal et al., 2005). This reflexive eye blink further protects the eye from injury (Graham, 1979). Similar startle potentiation is observed in rodents at the post-encounter stage of threat imminence (Fanselow, 1989). Lang and Bradley (2013) regard the physiological responses elicited under valenced conditions as serving a preparatory function for action to be undertaken, thus enabling rapid enactment of the action or, more specifically, rapid switching to approach and avoidance modes of responding. It has been shown that loud auditory stimuli, which provoke startle responses, are associated with a biphasic cardiac defense. The first acceleration/deceleration phase of heart rate serves to interrupt ongoing activity and initiate threat analysis, while the second phase prepares the organ-ism for action (Vila et al., 2007). Enhancement of the startle reflex and the heart rate pattern associated with it while passively viewing negative images provides indirect support for an automatic action tendency (or, priming of a particular action), akin to elevated startle during freezing at the post-encounter phase in rodents.
Despite the importance of the behavioral component of appetitive and defen-sive motivational networks, its priming has been overlooked until recently, mainly due to the lack of appropriate methodology for its study and the complexity of the phenomenon under question. In some animals, the only option for executing the action tendencies associated with each motivational system is instinctive and easily observable: reflexive approach (reducing the distance between the body and a pleasant object) and reflexive withdrawal (increasing the distance between the body and an unpleasant object) (Lang & Davis, 2006; Schneirla, 1959), which sim-plifies understanding of the behavioral impact of network activation. In humans, the networks are believed to guide behavior in a much more complex way (Lang & Bradley, 2013). For example, in some contexts when threat is not imminent, both the appetitive and defensive systems can be activated simultaneously and the observed behavioral patterns may be complex, creative and sometimes unpre-dictable (Bouton, Mineka, & Barlow, 2001; Lang, 1995; Lang et al., 1997) (e.g.,
faking an illness to prevent attending school on the day of an exam). However, human behavior becomes more aligned with the defensive motivational state at high levels of threat imminence (Bouton et al., 2001; Lissek et al., 2006) when active fleeing is the only behavioral response that will definitely assure survival.
The issue of which response is selected when the defensive motivational net-work is activated in humans has been addressed but to a limited degree. Drawing from the work of Bolles and Fanselow, Lang et al. (Lang & Bradley, 2013; Lang et al., 1997) theorize that threat imminence is indeed the crucial factor deter-mining specific human behavioral responses, with defensive activation triggering alert behaviors during pre-encounter and focusing of attention and freezing at post-encounter. At this latter stage, it is proposed that physiological changes occur to prepare for the following stage where mobilization increases and action potentially occurs (circa-strike stage; Fanselow, 1989). Extending from their ani-mal work, Rau and Fanselow (2007) proposed that the functioning of individuals with post-traumatic stress exemplifies inappropriate activation of the adaptive defensive motivational network, as they react with post-encounter behavior when pre-encounter behavior is more appropriate and with circa-strike behavior at times when post-encounter behavior is the needed mode of defense. They attribute pat-terns of inappropriate responding to an inability to make correct threat-imminence judgments due to the trauma experience. Craske (1999; 2003) applied the model of threat imminence to human anxiety, and posited that predatory encounter parallels the individual’s detection of threat, and is associated with worry, mus-cle tension and avoidant behaviors, whereas the circa-strike behaviors parallel fear/panic and associated autonomic activation and escape behaviors. However, none of those models directly addresses the question of how a response is selected and which factors guide the execution of a particular behavior.
The strongest evidence for the role of threat imminence in humans comes from the work of Mobbs and colleagues (2009; 2007). Following evidence for the neu-rological pathways of defense responses in rodents (Fanselow, 1994; Fanselow et al., 1993), they examined activation of human brain regions while participants were playing a symbolic computer game where they could control the movement of a prey which was chased by a predator that administered electric shock to the participant upon symbolic contact. When the participant reduced the distance between the predator and the prey, brain activity increased in the periaquadec-tal gray, a brain region that controls reflexive responding; when the distance did not reduce, brain activity mostly occurred in the ventromedial prefrontal cortex (vmPFC, Mobbs et al., 2009, 2007). The neurological activation pattern for hu-mans was thus similar to that seen in rodents in reaction to imminent threat, with brain activation switching from evolutionarily newer to older brain structures as distance to the threat decreased (Fanselow, 1994). Recent work focused on the aspect of physical distance and its influence on the modulation of physiological defensive responses. It was observed that startle is potentiated by proximal pres-ence of objects and participants chose to maintain a larger distance of the self from threat predictors as compared to predictors of safety in both familiar and novel contexts (˚Ahs, Dunsmoor, Zielinski, & LaBar, 2015). Last, but not least, this line of research showed that proximity enhances the valence of stimuli and the respective motivational network activation. Startle seems to be potentiated when avoidance is required and inhibited when endurance is required for
predic-tors of shock at close distances (˚Ahs et al., 2015). Other evidence shows indirectly that changes in threat distance might capture attention, with participants being much faster to categorize the approach of an angry face (aggressor/threat) than of a fearful face (victim, R. B. Adams, Ambady, Macrae, & Kleck, 2006; but see van Peer, Rotteveel, Spinhoven, Tollenaar, & Roelofs, 2009). However, no stud-ies to date have directly linked the neural or attentional mechanisms associated with distance from threat (i.e., threat imminence) with response selection and execution of behavioral responses.
2.2.3 Summary
Theories and findings on human defense responding reiterate many of the propo-sitions of the threat imminence model established in rodents. The robustness of the findings attests to the universality of defensive responding, the importance of threat imminence and the evolutionary advantage of rapid switching from less to more vigorous responding. However, empirical evidence has been limited to phys-iological, neurological and subjective responses, and has lacked measurement of behavioral responses (Beckers et al., 2013). Research from social and experimental psychology can offer clues about what shapes the behavioral response component of responding to threat. We will turn to that research in the next section.
2.3 Action tendencies
As already discussed, valenced stimuli can activate the appetitive or the defensive motivational network (e.g., Lang et al., 1997) and provoke the related action tendency of approach or avoidance. Gaining more insight into this phenomenon is crucial for a comprehensive understanding of threat-related responding and its regulation (Elliot, Eder, & Harmon-Jones, 2013). The concept of behavioral predispositions can be found in social psychology research on attitudes (seen as emotionally ridden ideas; Chen & Bargh, 1999) and emotions. Regardless of theoretical orientation, there is convergence on the idea that specific tendencies to act will correspond with the perceived motivational orientation of an encountered stimulus (i.e., defensive or appetitive), but that the primed action need not be executed (Elliot et al., 2013; Frijda, 2010). The expected consequence of exposure to a negative stimulus and corresponding defensive motivational orientation is the priming of an avoidance tendency (Eder & Hommel, 2013; Lang & Bradley, 2013). In other words, negatively valenced stimuli will limit the behavioral repertoire by activating the defensive motivational network and priming an avoidance tendency (much like their effect on rodents; Bolles, 1970), but will not precisely determine the manner in which that avoidance tendency would be executed in terms of its intensity and form. Response selection would be guided by other factors, discussed in later sections.
Two seminal studies set the stage for a detailed examination of activation of approach-avoidance tendencies in response to valenced material. The first in-structed participants to sort cards with positive and negative words by pushing them away (a motoric movement indicative of a defensive motivational state) or pulling them towards (a motoric movement indicative of an appetitive motiva-tional state) the body (Solarz, 1960). Participants who were required to respond
to positive words with a pull movement and to negative words with a push move-ment were significantly faster than participants who responded in the opposite manner. The results were interpreted as evidence that motivationally compati-ble motoric movements (pull appetitive, push defensive) are executed faster, due to matching between stimulus valence and the movement. These findings were the first to empirically show that the mere presentation of a negatively valenced stimulus primes an avoidance tendency, with priming reflected in speed of re-sponding. Chen and Bargh (1999; but see Rotteveel et al., 2015) replicated and extended these findings using a different task. In two experiments, participants pushed or pulled a lever in response to positive or negative words with or without evaluating the meaning of the words. They concluded that valence of the words resulted in facilitation of either pulling (approach) or pushing (avoidance) regard-less of whether conscious evaluation of stimulus valence was required. Those data suggest that an avoidance tendency might emerge even in the absence of overt evaluative processing of threat.
The compatibility of approach and avoidance tendencies with positive and negative valenced stimuli, respectively, has since been corroborated in a number of studies using a variety of tasks and stimuli. In more recent versions of Chen and Bargh’s task, participants push or pull a joystick in response to computer-generated stimuli (joystick approach-avoidance task, JAAT), and in some versions of this task the image increases or decreases in size to strengthen the percep-tion of approach or avoidance (joystick approach-avoidance task with feedback, JAAT-F). This task has been used primarily with positive and negative words (e.g., Krieglmeyer & Deutsch, 2010), pictorial stimuli of positively and nega-tively valenced objects (Krieglmeyer & Deutsch, 2010; Rinck & Becker, 2007) and faces (e.g., Roelofs, Elzinga, & Rotteveel, 2005; Roelofs, Minelli, Mars, van Peer, & Toni, 2008; Vrijsen, Van Oostrom, Speckens, Becker, & Rinck, 2013) and the results have been consistent over studies. A recent meta-analysis of the effects of task, stimuli, valence and instructions on valence-action tendency com-patibility concluded that there is evidence for the link between valence and ap-proach/avoidance tendencies (possibly indirect with appraisals playing a crucial role; Phaf, Mohr, Rotteveel, & Wicherts, 2014).
Similar results have been obtained using a variant of the affective Simon task. In this task, participants emit verbal responses to valenced cues. Regardless of instructions to ignore the valence of the stimulus, responding to congruent trials (e.g., saying the word positive to a positive cue) is faster than incongruent trials (e.g., saying the word negative to a negative cue), even when responding on the basis of a stimulus dimension different from valence (e.g., saying positive to nouns and negative to adjectives). This so-called Simon effect has been extended to non-verbal responses of approach and avoidance (De Houwer, Crombez, Baeyens, & Hermans, 2001). With this methodology of moving a manikin either towards or away from a valenced word (manikin approach-avoidance task, MAAT; De Houwer et al., 2001), researchers also found that valence is differentially linked to action tendencies. Similar effects were obtained when pictorial stimuli are used within MAAT (e.g., Krieglmeyer & Deutsch, 2010). In addition, previously neutral stimuli that become conditioned stimuli through pairing with an aversive outcome (i.e., shock) prime an avoidance tendency in a modified MAAT when compared to neutral stimuli never paired with an aversive consequence (Krypotos
et al., 2014). This finding suggests that approach-avoidance tendencies extend beyond innately aversive or innately appetitive stimuli to stimuli that acquire such properties through association; in other words, approach-avoidance tendencies can be learned. The crucial difference between the MAAT and the JAAT is that participants use a button press in order to approach or move away from the threat stimulus in the former paradigm. Thus, differential action tendencies towards both learned and innately valenced objects can be observed across behaviors that albeit topographically different (i.e., button press and joystick), function to withdraw from threat.
Finally, compatibility between a behavioral tendency of approach or avoidance and valence has also been examined using a movement platform that records small changes in distribution of body weight or full-body movements. Stins et al. (2011) found that participants were faster at initiating a step forward (approach) when faced with a portrait picture of a smiling face (positive valence) than of an angry face (negative valence); however, no significant difference was found for backward stepping (avoidance). Eerland, Guadalupe, Franken and Zwaan (2012) also found a preferential forward body leaning towards positive pictures and more backward leaning in response to negative pictures. One study on full-body movements, however, failed to find this compatibility effect (Stins, Lobel, Roelofs, & Beek, 2014). Last but not least, a recent study showed a backward displacement of body posture while viewing negative material relative to neutral material (Lelard et al., 2014). Thus, the compatibility of pleasant-approach and unpleasant-avoid has been extended to full-body responses, which may possess greater ecological validity than joystick or button press responses, although the results are not entirely consistent.
2.3.1 Mechanisms underlying approach-avoidance
compatibility effects
Three major hypotheses exist for why an approach tendency is primed by posi-tively valenced stimuli and an avoidance tendency by negaposi-tively valenced stim-uli. First, Cacioppo, Priester and Berntson (1993) and Chen and Bargh (1999) expanded the initial interpretation of motoric movement as an index of valence-primed action suggested by Solarz (1960) and argued that arm-flexion/extension (muscle activation as well as arm movement) is linked to valence. However, ob-servation of compatibility effects when using button presses rather than arm movements within tasks such as the MAAT argue against this hypothesis (see Krieglmeyer et al., 2013 for a more detailed account and Phaf et al., 2014 for a rebuttal of the flexion/extension account). In a recent meta-analysis of studies focusing on muscle activation, it was concluded that two other explanations of the compatibility effects are more convincing: namely the evaluative coding and distance regulation accounts (Laham, Kashima, Dix, & Wheeler, 2015).
The evaluative coding hypothesis suggests that valenced stimuli prime con-gruently valenced responses (Eder & Rothermund, 2008). In the Theory of Event Coding (TEC; Hommel, M¨usseler, Aschersleben, & Prinz, 2001; Lavender & Hom-mel, 2007), the label of the response required in the task as positive or negative (based on the situation, the task or current goals; Eder & Rothermund, 2008) determines what response will be primed. This theory was initially focused on
ex-plaining goal-directed action and steered clear from the question of consciousness (Hommel et al., 2001). TEC was later expanded to explain affective compatibility by emphasizing the role of goals and intentions in action control, thus arguing against the automaticity of these valence compatibility effects (Lavender & Hom-mel, 2007). This theory prompted the introduction of feedback (i.e., zooming in and out of the stimulus) in JAAT, because feedback facilitates the labeling of motoric movements as approach or avoidance, and prevents participants from re-framing their pushing and pulling response as away from the body or towards the body rather than the intended toward the object and away from the object (Rinck & Becker, 2007). However, this account does not readily explain the observation of congruency effects with valence-irrelevant instructions, when participants are required to respond to another feature of the presented stimulus (Krieglmeyer & Deutsch, 2010; Krypotos et al., 2014) and when no valence labels are given to responses (Krieglmeyer, Deutsch, De Houwer, & De Raedt, 2010)(both tested within the MAAT).
Krieglmeyer, De Houwer and Deutsch (2011) investigated a third hypothe-sis, i.e., that distance change determines which response will be preferred. They proposed that every response that increases the distance from a negative object would be executed faster than any response that reduces this distance. The re-verse would occur for positive objects. Krieglmeyer et al. (2011) modified the MAAT such that initial responding was incompatible with the ultimate distance between the manikin and the object at the end of the trial (i.e., the manikin has to approach the object initially, in order to ultimately be further away from it). They found that ultimate distance was the factor that determined response speed. These results mirrored findings by Seibt et al. (2008), who found distance-change compatibility effects even when arm-flexion and extension were mismatched. Ref-erence frame also seems to have an effect on the compatibility findings (Saraiva, Sch¨u¨ur, & Bestmann, 2013); more specifically, the primed movements depend on whether participants receive self-referent (towards-away from the body) or object-referent (towards-away from the object) instructions. This suggests that the executed behavior can be quite flexible as long as it fulfills its distance regu-lating function. In addition, information-processing biases have been found on the basis of perceived or actual approach/avoidance. That is, individuals categorized positive words faster while performing or perceiving approach and negative words faster while performing or perceiving avoidance (Neumann & Strack, 2000). Thus, perception of distance change also primes the identification of the presence of a negative stimulus, even when the distance change is not controlled by the individ-ual. Finally, it has been recently shown that arm-flexion (indicative of a move-ment reducing the distance between the self and the object) increases appetitive physiological responding to positive stimuli (startle attenuation) and defensive re-sponding to negative stimuli (startle potentiation) as compared to arm-extension (Deuter, Best, Kuehl, Neumann, & Sch¨achinger, 2014). In our view, the empirical evidence supports a distance change account above other accounts (as also argued by Krieglmeyer et al., 2013). This theory, further, closely matches motivational theories discussed earlier, in which activation of the defensive motivational net-work enables the organism to execute an avoidance action faster in order to assure survival by increasing the distance between the self and the threat.
2.3.2 Automaticity of avoidance action tendencies
According to the theories outlined above, avoidance tendencies to threat should represent an automatic mode of behavior (De Houwer & Moors, 2012), which can operate even when conscious processing is not available (Chen & Bargh, 1999). However, operationalization of the term automatic has been the source of much controversy (for a detailed review, see Moors & De Houwer, 2006). Here, we use the framework of Moors and De Houwer(2006)3and define automatic processes as
processes that are “unintentional, uncontrolled, unconscious, efficient, and fast”. In contrast, controlled processes are intentional, controlled, conscious, inefficient and slow (De Houwer & Moors, 2012; Moors & De Houwer, 2006). Each of these features lies on a continuum and each contributes uniquely to the automaticity of a response. In the next section, we review the evidence for the automatic-ity of the triggering of avoidance tendencies as observed in approach-avoidance compatibility effects.
Approach-avoidance compatibility effects are “fast”, since the effects can be observed at reaction times below 700 ms (Krieglmeyer et al., 2013; Krypotos et al., 2014)4. Priming of approach or avoidance may be “unconscious”, since the
task does not depend on explicit evaluation instructions (Chen & Bargh, 1999). Some researchers disagree with this conclusion (Krieglmeyer et al., 2013), because, for example, tendencies can be modified based on reference frame instructions, suggesting that they are under conscious control (Saraiva et al., 2013). The dis-agreement may stem from different usage of the term “unconscious” (Moors & De Houwer, 2006). Phaf et al. (2014) conclude that there may be a mix of conscious (such as appraisals or re-appraisals) and non-conscious evaluative processes in the valence-tendency link. More research is needed to clarify to what extent valenced stimuli trigger approach-avoidance tendencies outside consciousness (see Moors & De Houwer, 2006).
The degree of “efficiency” of avoidance tendency preparation, which refers to the amount of cognitive resources occupied by the phenomenon, has not been assessed as of yet (Moors & De Houwer, 2006). Furthermore, the degree to which compatibility effects are “unintentional” is unclear. Some studies show that the effects remain stable regardless of the instructions given to participants, which suggests that they cannot be easily overridden by intentions, and as such are unintentional (Krieglmeyer et al., 2013; Krypotos et al., 2014). However, others have found that instructions do have an effect on responding (Lavender & Hom-mel, 2007). Eder and Hommel (2013) propose that every stimulus primes one or more associated goals and the desired consequences in relation to these goals determine the primed response. Eder (2011) also showed that incompatible imple-mentation instructions, or planning what precisely to do in a specific situation, can change slightly the pattern of these tendencies. On the other hand, the difficulty in reversing the tendencies through implementation instructions might partially depend on the relative strength of the goals of the individual, with appetitive and defensive motivational network goals being stronger than task goals. Further, 3The framework of Moors and De Houwer (2006) is the most comprehensive and detailed
analysis of automaticity to date. Therefore, we have chosen to rely heavily on its propositions in the examination of approach-avoidance compatibility effects.
4Moors and De Houwer (2006) recognize that no objective criterion can be set for this feature
since avoidance is biologically predetermined, it might obstruct goal-oriented be-havior. A study by Fishbach and Shah (2006) indeed suggests that the relative strength of goals might be the crucial factor. They showed that individuals with stronger and more long-term goals (e.g., avoiding sweets when dieting) were faster at executing responses compatible with their long-term goals than responses com-patible with the stimulus. However, Fishbach and Shah (2006) researched objects that were not immediately related to survival (e.g., those involved in studying and dieting5). Hence, their findings might not pertain to primary stimuli that
link directly to activation of the defensive or appetitive system. When a threat to self is concerned, the goal associated with the defensive system activation should be stronger than many other self-relevant goals, because survival is so essential to the organism. Thus intentionality might be a function of goal strength and the intentional component of the goal.
The “controllability” of approach-avoidance tendencies depends on the goals of participants, whose responses are measured with the AAT. The process mea-sured by AAT has been proposed to be uncontrolled , because individuals have the exact same goals on both compatible and incompatible trials, those goals do not relate to stimulus valence, but rather to speed and accuracy of the executed response, and the compatibility effects are observed regardless (Krieglmeyer et al., 2013). Bossuyt, Moors, and De Houwer (2014) investigated precisely the impor-tance of goals for the preferential execution of avoidance or approach tendencies in response to a stimulus associated with emotions of fear or anger. They tested the hypothesis that fear is associated with the superordinate goal of protection and that avoidance, defined as increasing physical distance from threat, is a sub-ordinate goal of protection, which would predominate only when it serves that su-perordinate goal. Movement of the manikin (in a MAAT) away from a symbolic opponent, when defined in the instructions to participants as “aggressive stub-bornness,” was not the primed response to the angry opponent stimuli. Rather, moving towards the threat (the opponent) was primed, because it was defined as “submissive begging.” However, the definition of avoidance used by this group of researchers focused on the topography of the response and not its function. In social situations, “begging” might serve the function of increasing psychological as well as temporal distance from threat, because the opponent needs to consider the plea of the subject, thus delaying any aggressive action. Thus, begging can be seen as a safety behavior, which is included in our definition of avoidance. From animal research, the topography of the response to more distal threat can change from active fleeing to other responses serving the avoidance function (Fanselow & Lester, 1988). Thus, the controllability of an avoidance response may depend on the threat imminence of the situation.
A symbolic threat (i.e., another manikin), as used by Bossuyt et al. (2014), might not replicate what happens when actual feared stimuli are encountered. In a laboratory fear-conditioning study, Krypotos et al. (2014) showed that a warn-ing stimulus (CS) previously associated with an aversive consequence produces an avoidance tendency even when no aversive consequence can follow it and when the instructions required a stimulus-irrelevant feature to be evaluated for responding. 5Even though food is seen as a primary stimulus for the activation of the appetitive system,
food stimuli might not result in a strong activation of this system if the subject is already satiated.
The presentation of a CS in the MAAT resulted in strong defensive system ac-tivation, probably more so than the presentation of a symbolic opponent in the experiment of Bossuyt et al. (2014). The data from Krypotos et al. (2014) suggest that the threat-avoidance link is strong and impervious to subsequent task infor-mation (such as the removal of shock electrodes or extinction) and thus probably uncontrolled.
In the terminology of Moors and De Houwer (2006), we believe that the extant data suggests that avoidance tendencies in response to threat are automatic in the sense that they are relatively fast and uncontrollable. However, evidence regarding other features of automaticity (such as dependence on consciousness or sensitivity to intentions) is mixed. We are not aware of any evidence regarding the feature of efficiency. As argued by Moors and De Houwer (2006), a process can possess features of automaticity and controlled features at the same time; if a process exhibits at least some features of automaticity, it should be considered at least partially automatic. Our interpretation of the existing data suggests that avoidance tendencies are mostly automatic, a conclusion also supported by Krieglmeyer et al. (2013).
2.3.3 Summary
In summary, the link between defensive motivation (as provoked by negative stim-ulus valence) and avoidance tendencies seems to be strong and robust across stud-ies using various experimental methodologstud-ies, from symbolic manikin computer tasks to full-body movement measurements. A distance-regulation motivational account of these compatibility effects seems most plausible, especially when dis-tance is defined according to our proposal (including psychological disdis-tance). Last, but not least, the process seems to have at least some features of automaticity. Overall, experimental evidence on approach-avoidance tendencies supports the idea that activation of the defensive motivational network automatically primes avoidance tendencies.
2.4 Avoidance actions
The priming of avoidance tendencies in the face of threat is adaptive and evolu-tionarily predisposed. However, priming of avoidance tendencies is not identical with the actual execution of an avoidance response. In this section, we discuss observational (section 2.4.1) and experimental (section 2.4.2) studies on avoidance in order to partial out the factors that might play a role in the transformation of primed and automatic avoidance tendencies into an actual avoidance action.
2.4.1 Avoidance “in the real world”: withdrawal and flight
Ethologists who monitor animal behavior in the wild believe that human behav-ior closely resembles that of other mammals and specifically primates (Weisfeld, 1997). Cook and Mineka (1987) also suggested that studying primate responding to threat would give us better insight into threat processing in humans. Cook and Mineka (1989) noted frequent displays of avoidance of threat in rhesus monkeys within the confines of the cage. On some occasions, animals suddenly moved to
the back of the cage and pressed their bodies to its bars, thus achieving maximal distance from the threat presented to them. Other more or less expressive behav-iors were also recorded. In a famous study of Seligman and Maier (1967), dogs that were repeatedly unable to escape (i.e., avoid) an electrical shock exhibited a large range of dysfunctional behaviors, which have an astonishing resemblance to the human presentation of clinical depression6. Thus, avoidance is a primary
mode of animal responding and the continuous inability to execute avoidance in response to threat can have severe consequences for general functioning, as seen in the Seligman and Maier’s experiment (1967).
What humans do in the presence of physical threat has not been systemati-cally studied in natural settings7, however, behavior under social threat such as a
stranger encounter has been examined in some detail (p. 170-187, Eibl-Eibesfeldt, 1989). When exposed to a stranger, children show an approach-avoidance con-flict. In these stranger-encounter situations, both the appetitive (establishing social contact) and the defensive (preserving integrity) motivational networks are activated. Children show initial avoidance of the stranger (e.g., a child burying his face in his mother), indicative of the primacy of the defensive system over the appetitive system. If threat imminence does not increase over time (distance has remained stable) or perceived threat imminence has reduced due to the ob-servation of pleasant exchanges between a trusted adult and the stranger, the child might establish contact anew. If the stranger approaches (threat imminence increases), however, the child reacts with outright withdrawal, panic and crying, much like the circa-strike behaviors observed in rats (Fanselow & Lester, 1988). Video recordings from one seminal study on human fear learning also show Little Albert (J. B. Watson & Rayner, 1920) attempting to crawl away crying when he was presented with a stimulus predictive of threat. The evidence from youths suggests that withdrawal and flight from threat is a pre-wired behavior (Frijda et al., 1989) that has dominance over other behaviors (Frijda et al., 1989). It seems that only as threat imminence decreases, withdrawal and flight can be inhibited and substituted for alternative behaviors.
Further evidence for the innate or pre-wired nature of certain defensive behav-iors comes from the universality of behavior in response to threatening situations. Some emotion theories posit that behavior elicited by specific emotions is more stable across cultures than the self-report of emotional experiences (Mosquera, Manstead, & Fischer, 2000; Soto, Levenson, & Ebling, 2005). Other emotion theories suggest that cross-cultural differences are more readily observed in emo-tionally provoked overt behavior than in underlying action tendencies, because the former are more prone to influences of social acceptability, norms and circum-stances (Mesquita & Frijda, 1992).
Research on cross-cultural emotional action is not as rich in evidence as the study of emotional facial expressions conducted by Izard (e.g., 1994) and Ekman (e.g., 1971). Also, motoric action might be less universally linked to emotions than facial expressions (Neumann, Lozo, & Kunde, 2014), because it can be more varied and its pattern can be more complex. Nevertheless, a few notable studies 6Attempted, but non-successful avoidance is also related to distress in humans (Andrews,
Troop, Joseph, Hiskey, & Coyne, 2002).
7It is difficult to create ethically acceptable levels of threat that will provoke extreme
avoid-ance behaviors in humans. In addition, it is challenging to objectively record such behaviors outside of the laboratory, since a researcher can hardly remain uninvolved when threat is posed.
of overt behaviors are worth mentioning8. Wallbott and Scherer (1986) asked
participants from 27 countries to report emotional events and their responses to them; moving away was a response associated with all negative emotions. This might suggest that flight is related to a more general activation of the defensive motivational network. Consedine, Strongman and Magai (2003) asked partici-pants to identify emotions based on action statements embedded in reports of behaviors associated with various emotions collected from another set of par-ticipants. Significant consensus was observed cross-culturally, especially for the basic emotions of fear, anger, sadness and happiness. The same was true for the statements describing behavioral urges (i.e., action tendencies) associated with emotions, with no differences across cultures in the rate of emotion identification. In other words, people from different cultures converge on identifying the emotion that is associated with a specific action9. Rim´e, Boulanger, Laubin, Richir, and
Stroobants (1985) also showed that participants from different cultures do not differ in their inferences about emotions based on the movements of human-like figures. Finally, participants from Japan and the United States successfully in-ferred the emotions portrayed by actors’ kinetic movement (Sogon & Masutani, 1989). This suggests that the primary association between defensive motivational network activation (or the negative emotions most closely associated with it, such as fear) and avoidance can be found in many cultures.
On the other hand, a certain level of cultural variability may exist as well (Mesquita & Frijda, 1992). One example of such variation is the predominant avoidance response of falling asleep among Balinese (Bateson and Mead, 1942 as cited in Mesquita & Frijda, 1992). In another cross-cultural study on emotional responding, as much approach as withdrawal was reported for episodes of fear, whereas other negative emotions predisposed individuals to move away much more reliably (e.g., sadness and disgust) and there were small to medium differences between cultures on expression of these emotions (Scherer & Wallbott, 1994). Cul-tural variation may result from the interaction across the motivational networks activated, level of threat imminence, and the behavioral repertoire of particular people and the circumstances in which their emotions are experienced.
2.4.2 Experimental studies of avoidance
As mentioned earlier, learning theorists use the term avoidance to describe the response that prevents the onset of an aversive stimulus and the term escape to describe the response that terminates the aversive stimulus. Using our definition of avoidance, both responses can be seen as forms of avoidance since they are prompted by the defensive motivational network and their function is to increase the distance from threat. The forefathers of behaviorism, Bechterev (1913) and J. B. Watson (1916) conducted experiments on the learning of avoidance, even though they referred to the observed responses as “conditioned reflexes” (Ole-son & Cheer, 2013) and did not recognize the presence of an avoidance contin-gency (Bolles, 1972). In these experiments, participants could avoid mild electrical 8We are not aware of any cross-cultural studies of action tendencies that utilize reaction time
tasks or other appropriate measures of action tendencies.
9Even though these studies do not measure avoidance behavior per se and possibly tap into
social conventions, the intercultural consensus suggests somewhat more rigidity in responding for these emotions.
stimulation to the hand by moving their fingers in response to a warning signal (Bechterev, 1913), thus increasing the distance between the threatening surface and the threatened part of the body - a clear physical withdrawal. In recent human studies, however, a button press on a computer keyboard functions as an avoid-ance response to cavoid-ancel a predicted shock (e.g., Delgado, 2009; Dymond, Schlund, Roche, De Houwer, & Freegard, 2012; Lovibond & Colagiuri, 2013) or prevent another form of aversive consequence such as money loss (Schlund & Cataldo, 2010)10. Most individuals readily learn to perform the response required by the
task (Lovibond, Saunders, Weidemann, & Mitchell, 2008) with some individual difference variables affecting the rate of responding (e.g., gender, Aupperle, Sulli-van, Melrose, Paulus, & Stein, 2011; McLean & Hope, 2010; Sheynin, Beck, Pang, et al., 2014; Sheynin, Beck, Servatius, & Myers, 2014). The ability of humans to learn and perform almost any arbitrary response in a laboratory in order to match the requirements of the activated defensive motivational network suggests that the topography of the avoidance response under conditions of lower threat imminence might be very flexible even if its function remains constant (i.e., to increase distance from threat). Responses that do not increase spatial or phys-ical distance, but rather increase psychologphys-ical distance through decreasing the feelings of threat are named here safety behaviors. These, together with physical withdrawal or flight constitute the large range of avoidance actions that we try to explain within our model.
Extant research on avoidance learning has not directly examined the factors that drive the performance of the avoidance response following its selection. How-ever, in both the cognitive theory of avoidance (Seligman & Johnson, 1973) and the expectancy account of avoidance learning (Lovibond, 2006), performance is guided by comparison of the expected consequences of executing the response to those of not executing the response. These models presuppose that higher-order cognitive processing and evaluation of the response guides active avoidance, which is instrumentally motivated. Direct empirical evidence for this conclusion is lacking. However, indirect support is found in studies using virtual reality.
Virtual reality offers the opportunity to immerse an individual in a highly re-alistic, but controlled environment, which can be navigated and explored freely (e.g., Glotzbach, Ewald, Andreatta, Pauli, & M¨uhlberger, 2012; Lissek et al., 2006; Rinck et al., 2010). Grillon et al. (2006) exposed participants to virtual re-ality environments that contained either predictable shocks, unpredictable shocks, or no shocks. Later, participants were given the opportunity to freely navigate between two of these three environments to retrieve money while not being threat-ened by shock administration. Participants were more likely to avoid (keep dis-tance between the avatar and the room) the environment where unpredictable shocks were encountered, which mirrors findings with rodents (Fanselow, 1980). In another study, participants showed more subsequent avoidance of a dangerous context (with shocks) than of a safe (without shocks) or neutral (new) context (Glotzbach et al., 2012). In yet another study, where the relationship between avoidance and subsequent extinction was examined, participants spent less time examining a virtual reality context (i.e, more avoidance) that predicted an aver-sive event (Cornwell, Overstreet, Krimsky, & Grillon, 2013). The results of these 10Mineka and Oehlberg (2008) note that these paradigms most likely measure the same
studies suggest that prior experience informs judgments regarding the likelihood of encountering threat within a specific context, which in turn influences the ex-ecution of avoidance responses.
In the procedure used by Grillon et al. (2006), behavior was reinforced by the retrieval of money, which might have activated the appetitive system. Hence, the results might be interpreted as an inhibition of goal-oriented appetitive behavior under unpredictable threat, rather than avoidance. Recent studies (Pittig, Brand, Pawlikowski, & Alpers, 2014; Pittig, Schulz, Craske, & Alpers, 2014; van Meurs et al., 2014) suggest that individuals engage in avoidance when in the presence of a threatening signal even when avoidance is associated with a loss of reward. In other words, the approach-avoidance conflict is resolved in an avoidant way, which implies primacy of the defensive motivational network over the appetitive motiva-tional network. However, it is possible that individual differences might modulate this primacy effects and some individuals, such as those willing to “accept risks as the price for the reward” (sensation seekers; Zuckerman, 1994), might show more approach in the aforementioned studies.
Another important aspect of the virtual reality studies is their focus on context anxiety (i.e., the virtual reality environment), which might differ from the fear and avoidance elicited by a discrete cue (e.g., a pointed gun within a virtual reality environment) that predicts an aversive stimulus (Glotzbach et al., 2012). As seen in the experiment of Grillon et al. (2006), participants were less avoidant of a context in which a cue predicted the occurrence of an aversive event than of a context where the same aversive event occurred without being signaled by a specific cue; thus, it is very likely that both the context and the cue determined the precise behavior. Stimulus and context might both shape avoidance behavior by affecting evaluation of the expected consequences of defensive responding within a specific context as compared to the expected consequences of not responding, in accordance with the ideas put forth by Lovibond (2006).
Overall, laboratory studies show that participants can perform a range of safety behaviors to serve the function of avoidance when physical withdrawal/flight is either not needed, or not available. Safety behaviors, like physical withdrawal or flight, function to increase distance from threat, with threat imminence being an important factor in establishing the precise behavior executed. Furthermore, avoidance seems capable of inhibiting appetitively-motivated behavior and to be dependent on the context as well as the cue it is linked to. Finally, avoidance actions are hypothesized to depend on higher-order processing of the likelihood of threat, as proposed by Seligman and Johnson (1973) and Lovibond (2006).
2.4.3 Summary
Physical withdrawal/flight is readily identified as an action associated with fear, the primary emotion linked to the activation of the defensive motivational net-work, and other negative emotions. The findings suggest that overt avoidance in the face of threat is universal. However, safety behaviors that serve the func-tion of avoidance may be more varied and dependent on higher order processing such as task instructions in laboratory paradigms, context affordances, social and cultural demands, as well as a “cost-benefit” analysis of engagement in an avoid-ance behavior, as proposed by Lovibond (2006). Avoidavoid-ance in all its forms has the
