Behavior Is Learned; New Behavior Can Be Learned Through Early Cue Detection and Practice

  Behavior therapies are based on an assumption that individuals have learned to respond and act in the ways they habitually respond and act through identifiable principles of learning. As noted in Chapter 2, behavior therapies evolved from experimental research that detailed these learning principles. Modern behavior therapies are similarly informed by newer developments in experimental research that have identified complexities in principles of learning (e.g., Bouton, Mineka, & Barlow, 2001; Craske et al., 2008). An in-depth discussion of these principles and complexities is beyond the scope of this book (see Bouton, Woods, Moody, Sunshay, & García-Gutiérrez, 2006; Craske & Mystkowski, 2006; O’Donohue & Fisher, 2009), but we provide a summary so that therapists can use these principles to guide implementation of behavior therapies.

Learning Through Association

  Both humans and animals learn to associate stimuli that frequently appear together. Classical conditioning refers to the process through which a previously neutral stimulus becomes associated with a stimulus that evokes certain responses (either aversive or appetitive). Through being repeatedly paired with an unconditioned stimulus (US) that naturally evokes a given response, the conditioned stimulus (CS) becomes a cue for the US and elicits similar or related responses. This process is clearly evolutionarily adaptive in that organisms learn that the presence of certain stimuli indicates that a threat is likely to appear or that something desirable is likely to appear, and respond accordingly. Once a stimulus has been conditioned, it can lead to new learning by being paired with another previously neutral stimulus, which will in turn come to be associated with the CS and elicit similar or related responses.¹ Through this process of higher order conditioning, more stimuli come to be associated with undesirable or desirable events. Also, through stimulus generalization, stimuli that are similar to the CS also become learned cues, so that eventually a broad range of stimuli are associated and evoke similar responses. For instance, a learned fear of a bright red shirt might lead an individual to respond with anxiety or fear to anything red in the environment.

A client, Monique, can be used to illustrate these principles. Monique presented for therapy reporting that she was anxious and uneasy in social situations. A functional analysis, including monitoring of her symptoms and exploration of specific incidents of anxiety during the previous week, revealed that she responded with physiological arousal and anxious thoughts when she interacted with people who looked or sounded critical. She described her father as extremely critical when she was growing up and stated that he would often turn his attention to other people or walk away after he had criticized her for something. In this example, this withdrawal of attention and affection from a parent was a US that would have naturally elicited fear in a child. Its pairing with criticism from her father led Monique to respond to her father’s criticism with anxiety because she anticipated the removal of his attention and affection. Gradually, these associations generalized, and she came to have similar responses to any instances of perceived criticism, leading her to feel anxious in a broad range of social situations.

People are particularly prone to learning threatening cues because it is evolutionarily adaptive to identify markers for potential harm and danger so that individuals can avoid this harm or danger. In addition, some individuals are probably biologically predisposed to learn threat more easily and robustly and are therefore more prone to anxiety (e.g., Lonsdorf et al., 2009). Prior experiences with threat, or modeling of fear behavior by significant role models, may also make it more likely that an individual will easily learn to fear cues, and those responses will generalize (Mineka & Zinbarg, 2006). Biology, prior experiences, and modeling likely play a role in other kinds of learning, such as the reinforcing properties of alcohol and drugs (e.g., Enoch, 2007).

Initial models of associative learning identified the conditions under which learned associations (to CSs) could be extinguished such that an organism no longer responded to the CS as though it were associated with the US. Further study has indicated that the term extinction is a misnomer because associations are not, in fact, unlearned. Instead, new, competing, nonthreatening associations are learned. So, in the case of fear conditioning, repeated exposure to the CS in the absence of the US will lead to a new, nonthreatening association to the CS, such that fear is no longer the predominant response. Extinction can therefore be thought of as inhibitory learning (Craske et al., 2008) in that an association that inhibits the previous association is learned. Rescorla and Wagner (1972) noted that learning is an adjustment that occurs when there is a discrepancy between the outcome that is expected and the outcome that occurs. So extinction trials promote new learning in that the expected association does not occur, so that the CS comes to be associated with “not US” instead of the US.

Bouton et al. (2006) recently reviewed the literature that suggests that conditioned associations, as well as conditions that are likely to make extinction or inhibitory learning more robust, are not unlearned. Animal research has demonstrated that even after extensive extinction of fearful associations, the continued presence of these associations is demonstrated by (a) a renewal effect, in which a learned association to a CS returns when the CS is presented in a different context from the extinction trials; (b) spontaneous recovery, in which a learned association to a CS returns after the passage of time; (c) reinstatement, in which a learned association to a CS returns after the US is presented alone and the CS is presented later; and (d) rapid reacquisition, in which an association to a previously extinguished CS is learned much more rapidly in new conditioning trials. All of these phenomena suggest that a learned fearful association is maintained despite successful extinction. Bouton et al. interpreted these findings as evidence that extinction learning is context specific, which makes sense from an evolutionary standpoint—people learn cues for fear easily, and generalize them, yet learning of inhibitory responses to feared stimuli is more context specific. This serves an important survival function in that individuals will not prematurely learn that a given stimulus is safe simply because it was safe in a specific context. However, it makes it more likely that learned fears will recur, making it important for therapists to address relapse prevention in therapy, so that clients are prepared for these recurrences and are able to continue to approach feared stimuli to promote more robust extinction learned across multiple contexts. Researchers have also suggested that the presence of retrieval cues during extinction trials will help extinction (or inhibitory learning) to generalize to novel contexts (Craske et al., 2008).

Although associative learning is often described in terms of learned associations to external stimuli, there is also extensive evidence that organisms learn associations to internal stimuli as well (for an extensive review of this literature in the context of panic disorder, see Bouton et al., 2001). As a result, people’s own internal sensations can become threat cues, leading them to respond with anxiety, which strengthens the cue, potentially leading to a spiral of anxiety or panic. From a behavioral perspective, thoughts can also become associated with a US. As such, thoughts or memories of a traumatic event can elicit posttraumatic responses, even in the absence of the event itself. Thoughts can also have appetitive associations, so that a thought of a drink can lead to a powerful conditioned response of craving for an individual addicted to alcohol. Because these internal cues are beyond people’s instrumental control (they cannot avoid thoughts of drinking or anxious sensations completely), these associations are particularly likely to lead to clinical problems. As such, learning new associations to these cues is often an important target of treatment (as is learning not to respond to them behaviorally, as discussed later).