2. Implicit emotion and emotional trauma

Breuer and Freud (Reference Breuer, Freud and Strachey1895/1955) believed that the critical pathogenic element in hysteria was strangulated affect. Consistent with Janet's concept in the late nineteenth century (Van der Kolk & Van der Hart Reference Van der KolK and Van der Hart1989), trauma was conceptualized as an experience that was psychologically overwhelming because of the intensity of the affect that was activated, not because it was an event that was inherently life-threatening (as is specified in current Diagnostic and Statistical Manual [DSM-V] criteria for Post-Traumatic Stress Disorder [PTSD]) (American Psychiatric Association 2013). They believed that there was a lack of affective expression at the time of a trauma that kept the memory of the traumatic event alive for years. Once this emotion was experienced, expressed and put into words in the therapeutic context it would be curative. This conceptualization was consistent with the Freudian concept of unconscious mental representation, which was that mental contents including emotions were fully formed in the unconscious, were revealed in conscious awareness only when defenses were removed or overcome (Schimek Reference Schimek1975), and that the goal of therapy was to “make the unconscious conscious” (Breuer & Freud Reference Breuer, Freud and Strachey1895/1955; Freud Reference Freud and Strachey1923/1961).

A century of research has altered our understanding of unconscious mental representation. We now understand that memories and feelings do not reside in the unconscious fully formed waiting to be unveiled when the forces of repression are overcome (Lane & Weihs Reference Lane and Weihs2010; Levine Reference Levine2012). In contrast to a model of the unconscious as a cauldron of forbidden impulses and wishes, the “adaptive unconscious” (Gazzaniga Reference Gazzaniga1998) is conceptualized as an extensive set of processing resources that execute complex computations, evaluations, and responses without requiring intention or effort. Much of this processing may be unavailable to conscious awareness, or at least, awareness is unnecessary for such processing to occur. More commonly, cognitive psychology refers to implicit processes to differentiate them from explicit processes that are engaged during intentionally driven and goal-directed tasks. The distinction between implicit and explicit processing has been applied in some form to virtually all areas of cognition, including perception, problem solving, memory and, as we will discuss, emotion, leading Gazzaniga (Reference Gazzaniga1998) to suggest that 99% of cognition is implicit. Importantly, some psychoanalysts believe that this new way of understanding the unconscious as fundamentally adaptive calls for a revision of classic psychoanalytic models of the unconscious mind (Modell Reference Modell and Petrucelli2010).

In the memory domain, implicit memory refers to the impact of prior experience on subsequent behavior in the absence of explicit recall or awareness of that prior experience (for review, see Schacter et al. Reference Schacter, Chiu and Ochsner1993). In the laboratory a brief exposure to a specific word, for example, increases the likelihood that a person will respond with that particular word during various language-based tasks such as completing a three-letter word stem (Graf & Schacter Reference Graf and Schacter1985; Schacter & Graf Reference Schacter and Graf1989) or producing exemplars belonging to a semantic category (Ryan et al. Reference Ryan, Cox, Hayes and Nadel2008a). A different form of implicit learning is the acquisition of complex sets of rules that govern predictions (Reber Reference Reber1989), allow categorization of novel objects and concepts (Seger & Miller Reference Seger and Miller2010), and guide social interactions (Frith & Frith Reference Frith and Frith2012). Importantly, this learning occurs regardless of whether the individual is explicitly aware of the rules that have been acquired or that learning has even taken place. In the social domain, this learning consists of the semantic rules, expectations, and scripts for behavior that provide the basis for the self-concept (Markus & Wurf Reference Markus and Wurf1987).

The distinction between implicit and explicit processes, a cornerstone of modern cognitive neuroscience, has also been applied to emotion (Kihlstrom et al. Reference Kihlstrom, Mulvaney, Tobias, Tobis, Eich, Kihlstrom, Bower, Forgas and Niedenthal2000; Lane Reference Lane, Lane and Nadel2000). Emotions are automatic, evolutionarily older responses to certain familiar situations (Darwin Reference Darwin1872). Emotion can be understood as an organism's or person's mechanism for evaluating the degree to which needs, values, or goals are being met or not met in interaction with the environment and responding to the situation with an orchestrated set of changes in the visceral, somatomotor, cognitive, and experiential domains that enable the person to adapt to those changing circumstances (Levenson Reference Levenson, Elkman and Davidson1994). Implicit processes apply to emotion in two important senses. First, the evaluation of the person's transaction with the environment often happens automatically, without conscious awareness, and is thus implicit. Importantly for this discussion, this implicit evaluation is based on an automatic construal of the meaning (implications for needs, values or goals) of the current situation to that person (Clore & Ortony Reference Clore, Ortony, Lane and Nadel2000). Second, the emotional response itself can be divided into bodily responses (visceral, somatomotor) and mental reactions (thoughts, experiences). The latter include an awareness that an emotional response is occurring and an appreciation of what that response is. A foundational concept of this paper is that emotional responses can be implicit in the sense that the bodily response component of emotion can occur without concomitant feeling states or awareness of such feeling states.

There is now considerable evidence supporting an implicit view of emotion (Kihlstrom et al. Reference Kihlstrom, Mulvaney, Tobias, Tobis, Eich, Kihlstrom, Bower, Forgas and Niedenthal2000; Lambie & Marcel Reference Lambie and Marcel2002; Lane Reference Lane2008). Indeed, 25 years of research has demonstrated the occurrence of spontaneous affective reactions associated with changes in peripheral physiology and/or behavior that are not associated with conscious emotional experiences (Ledoux Reference LeDoux1996; Quirin et al. Reference Quirin and Lane2012; Winkielman & Berridge Reference Winkielman and Berridge2004; Zajonc Reference Zajonc and Forgas2000). For example, one can activate emotions with subliminal stimuli and demonstrate that the emotional content of the stimuli influences subsequent behavior, such as consummatory behavior, without the person being aware of such influences on behavior (Winkielman & Berridge Reference Winkielman and Berridge2004).

Furthermore, many decades of research preceding the modern era of neuroimaging demonstrated the evocation of visceral and somatomotor expressions of emotion in brainstem stimulation studies of laboratory animals (Ledoux Reference LeDoux1996). Although these phenomena cannot be linked to reportable experiences in animals without language, they nevertheless are the physical manifestation of emotion. We believe that implicit emotion, consisting of these visceromotor and somatomotor expressions of emotion, constitute the foundation upon which differentiated emotional experience is built. Moreover, subcortical structures including the thalamus, hypothalamus, amygdala, and periaqueductal grey likely contribute to the generation of these undifferentiated emotional responses that are not associated with specific emotional experiences (Lane Reference Lane2008).

Lambie and Marcel (Reference Lambie and Marcel2002) distinguish among three different conditions: an emotional state with no phenomenal experience; the first-order phenomenal experience of emotion, which is expressible; and a second-order experience of emotion associated with awareness, which is reportable. Implicit emotion, or bodily felt sensations, can be transformed into discrete conscious experiences of specific emotions or feelings by putting the felt sensations into words (Barrett et al. Reference Barrett, Mesquita, Ochsner and Gross2007; Lane Reference Lane2008). Through this process individuals can feel specific, differentiated emotions and “know” what it is that they are feeling. Thus, the term “explicit” is used to refer to states of awareness that are symbolized and known. In relation to emotion the term “implicit” refers to automatic bodily responses that are unconscious in the sense that they are not associated with awareness, are not consciously symbolized and are not known (but could include the unattended conscious or phenomenal experience of the bodily state). Note that the unconscious can further be differentiated into that which has never been mentally represented and needs to be formulated for the first time versus that which has previously been represented or known but is not consciously accessible at the moment (Levine Reference Levine2012). “Emotion processing” refers to any change in either the implicit or explicit components of the emotional response. “Cognitive processing of emotion” includes attending to the experience, symbolizing it (e.g., in words or images) and reflecting upon what the experience means (e.g., determining what one needs), or some combination thereof.

Based on these conceptual distinctions, one can revisit the concept of trauma as described by Breuer and Freud (Reference Breuer, Freud and Strachey1895/1955). Trauma may consist of experiences that are emotionally overwhelming in the sense that the ability or resources needed to cognitively process the emotions (attend to, experience and know them) are exceeded. Trauma may consist of a single event but more commonly consists of a repeated pattern of abuse or mistreatment that is emotionally painful to the victim. In the context of growing up as a child in a family in which abuse repeatedly occurs, one makes cognitive and emotional adaptations to keep the subjective distress to a minimum. This helps to keep attention and other conscious resources available for other tasks (see Friston Reference Friston2010). The victim learns to accept certain kinds of mistreatments in order to continue in relationships, which appear to be (and often are) necessary for survival. The needed adjustments include tuning out awareness of one's own emotional responses or taking for granted certain things about the self (such as “you're no good and deserve to be punished”). Later in life, related situations are interpreted implicitly based on the implicit learning that occurred from these experiences (Edelman Reference Edelman1989). One might conjecture that the more intense the abuse the more implicit evaluations in distantly related contexts are influenced by the trauma.

All too commonly, perhaps as a result of direct physical threats, shame or lack of available confidants, these experiences are never discussed with anyone. When a parent is the instigator of abuse it is often a “double whammy,” first because of the violation or harm and second because the parent is not available to assist the victim in dealing with it (Newman Reference Newman2013). The lack of an available caregiver to provide comfort and support may be a critical ingredient in what makes the experience(s) overwhelming or traumatic. What this means emotionally is that the implicit emotional responses were never brought to the conscious level of discrete feeling through mental representation, as in language. As a result, the traumatized individual knew the circumstances of the trauma but did not know how it affected him emotionally. This lack of awareness contributes to the tendency to experience traumatic threats in an overly generalized manner that reflects the inability to distinguish circumstances that are safe from those that are not. It is often only in therapy when the experiences are put into words that the emotional responses are formulated for the first time (Lane & Garfield Reference Lane and Garfield2005; Stern Reference Stern1983).

Although Breuer and Freud believed that expressing the emotion was critical, this alternative perspective highlights the importance of becoming aware of the emotional impact of the experience(s) through symbolization and contextualization (narrative formation) (Liberzon & Sripada Reference Liberzon and Sripada2008) and using this awareness in the promotion of more adaptive responses (that is, converting implicit emotional responses to explicit emotional responses). When the trauma is first recalled, the description of experience is likely to include strong emotions, such as fear, that were experienced at the time and contributed to strong encoding of the event. As the therapy process unfolds, the events are recalled in the context of a supportive therapist who also helps the client to attend to contextual information that may not have been available to the client at the time of the trauma (in part because of temporary hippocampal dysfunction [Nadel & Jacobs Reference Nadel and Jacobs1998]; see sect. 4). This new information in therapy contributes to a construction of the events in a new way that leads to emotions that had not been experienced before, for example, experiencing anger at abuse that could not be expressed or experienced at the time because the threat was so severe. The anger is a signal that one needs to be protected. In that sense, the emotional response is adaptive to the circumstances: It probably was not permissible at the time of the trauma to experience or express it. Experiencing and describing anger in therapy helps create a coherent narrative account of what occurred. Doing so is not the same as Freudian catharsis (release of pent up energy) but rather the creation of a more complete picture of what happened, how one responded, what one experienced, and how it could have been different (Greenberg Reference Greenberg2010).

Having another person such as a therapist participate in and facilitate this mentalization process in adulthood may be essential (Allen Reference Allen2013). The capacity for self-observation is limited, and more so if empathic and responsive parenting was limited during development (Paivio & Laurent Reference Paivio and Laurent2001). Just as having a teacher/coach/observer is helpful in the development and refinement of any athletic, intellectual, or musical skill, in the case of psychotherapy the therapist is potentially able to view a given situation from a different, if not a broader, perspective, making it possible to construe the situation, and the client's emotional response to it, differently. This relates to the “coaching” aspect of helping someone to get in touch with feelings of which they were previously unaware (Greenberg Reference Greenberg2002).

The guiding thesis of this article is that the therapy experience provides new information and that the old memory (or memories) is reconsolidated with this new information. Different therapy modalities focus on different kinds of information that are inherent in the therapy experience (see sects. 3 and 9). The discussion above focused on new information consisting of expansion of the client's understanding of what they experienced emotionally. The new information consists, in part, of both the conscious experience of emotions not previously experienced originally or during prior retellings of the event, and an understanding of what these experiences are and what they mean. Another source of new information, which is a common denominator across modalities, is the therapeutic alliance with the therapist (Horvath & Luborsky Reference Horvath and Luborsky1993). Experiencing the safety, support, caring, and compassion of the therapist in the context of recalling adverse experiences permits incorporation of this interpersonal experience, a type of information, into the traumatic memories, which often involve being alone and unprotected. The experience of comfort and support may be sensed and responded to implicitly without being brought to explicit awareness through attention, reflection, and verbal description.

Therapy modalities differ in the emphasis placed on self-exploration and the importance of the interpersonal connection with the therapist. A person's ability to be aware of and process her own emotions, and to engage with a therapist, may be a function of the degree to which caregivers succeeded in providing this function during childhood in a way that matched the needs of the child in question (Steklis & Lane Reference Steklis, Lane, Watanabe and Kuczaj2013). In the case of emotion as a subjective experience, there is no information in the external environment that corresponds to the child's internal experience except that which is provided by an attuned other. For example, one cannot typically see one's own facial expressions and in infancy such expressions may not be recognized as one's own even if looking in the mirror. This may be contrasted with the example of self-initiated movement (Keysers & Gazzola Reference Keysers and Gazzola2006). The basic coordination of intention with actual motor movement can occur without help from other people because one can see what happens when one intentionally moves one's arm. This visual input goes beyond the feedback provided by proprioceptive sensation. Thus, in early development the ability to link subjective experience of emotion with an understanding of its behavioral manifestations in the real world requires input from others (Gergely & Watson Reference Gergely and Watson1996).

For example, a very young child may recoil and appear frightened when a puppy approaches. A parent may say, “Don't worry. There's nothing to be afraid of (parent pets the puppy). See how friendly he is.” A somewhat older child may manifest avoidance behavior in anticipation of an upcoming event at school. Recognition by a parent that the avoidance behavior may be an expression of fear, labeling it as such, discussing with the child what he perceives as threatening and discussing ways to deal with it all contribute to the child's ability to experience fear and use it as a cue for adaptive responding in similar situations in the future. If input such as this from significant others is missing during development the capacity to know what one is feeling will be impaired in childhood and persist into later life, creating a predominance of implicit emotional responses relative to explicit emotional experiences and a greater propensity for being overwhelmed (traumatized) or unable to cognitively process one's own emotions later in life.

The expansion of awareness in therapy is not unlike that induced by a physical therapist who helps extend the range of motion of a joint by facilitating movements that are associated with tolerable but not excessive levels of pain and discomfort. It is difficult to extend oneself in these ways on one's own on account of self-protective mechanisms (the same ones that led to avoidance of emotional pain through regulatory actions). Parenthetically, the origin of restricted movement in a joint typically arises from inflammatory mechanisms designed to respond to and repair the original injury, just as psychological adjustments are made to limit access to “the part that hurts.” In psychotherapy, according to our formulation, expanding awareness involves experiencing, labeling, reflecting upon and using emotions that were originally associated with the trauma, but which by definition originally exceeded the person's capacities for assimilation and coping. Thus, new information brought in or facilitated by the therapist, available for reconsolidation, includes new ways of construing and responding to the client as a person, a new perspective on the originally traumatic events, and the facilitation of new emotional experiences.

From this perspective, what constitutes traumatic stress varies from person to person. This also helps to explain why trauma early in life predisposes to trauma later in life. As we'll see in section 3, the role of implicit memory in the construal of current situations based on past experience provides another perspective on how we might currently interpret what Breuer and Freud meant when they said that the lack of affective expression at the time of the trauma kept the memory of the traumatic experience alive.

3. Role of implicit emotion in different therapeutic modalities

Based on the considerations above, implicit emotion plays a critical role in a variety of psychotherapy modalities. In this section we will briefly discuss how implicit processes are relevant to behavioral (sect. 3.1), cognitive–behavioral (sect. 3.2), experiential (sect. 3.3), and psychodynamic psychotherapies (sect. 3.4).

4. Evidence that emotional arousal is critical to psychotherapeutic success

As noted in the introduction, there is good evidence that emotional arousal appears to be important for the success of many different forms of psychotherapy. Although this appears to be the case for BT, CBT, EFT, and psychodynamic psychotherapy, there are important caveats to consider.

Numerous behavior therapies based on exposure have been shown to be effective for treating trauma and anxiety-related disorders. A meta-analytic review of the literature found that exposure therapy is the most effective treatment for PTSD, and that its effectiveness is based on emotional processing (Foa et al. Reference Foa, Rothbaum and Furr2003). Patients with anxiety disorders who are best able to experience anxiety during the therapy session are most likely to benefit from therapy, including those with phobias (Borkovec & Sides Reference Borkovec and Sides1979), agoraphobia (Watson & Marks Reference Watson and Marks1971), obsessive-compulsive disorder (Kozak et al. Reference Kozak, Foa and Steketee1988), and PTSD (Foa et al. Reference Foa, Riggs, Massie and Yarczower1995). In a series of studies on behavioral exposure (Foa et al. Reference Foa, Riggs, Massie and Yarczower1995; Jaycox et al. Reference Jaycox, Foa and Morral1998), positive outcome for PTSD from rape was predicted by the arousal of fear and its expression while narrating memories of the trauma during the first exposure session and by reduction of distress over the course of treatment. Findings like this show that emotional arousal while engaging in imaginal exposure is an aspect of the mechanism of change. In studies of recovery patterns in sexual and nonsexual assault victims, long-term recovery in general was found to be impeded if the indispensable emotional engagement with traumatic material in therapy was delayed (Gilboa-Schechtman & Foa Reference Gilboa-Schechtman and Foa2001). As Greenberg & Pascual-Leone (Reference Greenberg and Pascual-Leone2006) note, research on behavioral exposure (e.g., Jaycox et al. Reference Jaycox, Foa and Morral1998) has shown that only some individuals actually engaged in the exposure task and therefore only some were able to benefit from the treatment.

Jones and Pulos (Reference Jones and Pulos1993) found that the strategies of evocation of affect, and the bringing of troublesome feelings into awareness, were correlated positively with outcome in both cognitive-behavioral and dynamic therapies. Another study (Coombs et al. Reference Coombs, Coleman and Jones2002) by this group examining the therapists' stance in CBT and interpersonal therapy of depression showed the importance of focusing on emotion regardless of orientation. Reviews of process–outcome studies in psychotherapy show a strong relationship between in–session emotional experiencing, as measured by the Experiencing Scale (Klein et al. Reference Klein, Mathieu-Coughlan, Kiesler, Greenberg and Pinsof1986), and therapeutic gain in dynamic, cognitive, and experiential therapies (Castonguay et al. Reference Castonguay, Goldfried, Wiser, Raue and Hayes1996; Goldman et al. Reference Goldman, Greenberg and Pos2005; Orlinsky & Howard Reference Orlinsky, Howard, Garfield and Bergin1986; Silberschatz et al. Reference Silberschatz, Fretter and Curtis1986).

A survey (Pilero Reference Pilero2004) investigated clients' experience of the process of affect-focused psychotherapies. The clients had participated in one of three emotion-focused therapies: Accelerated Experiential Dynamic therapy (Fosha Reference Fosha2000), Intensive Short Term Dynamic therapy (Abbass Reference Abbass2002), and Emotion-Focused Therapy (Greenberg Reference Greenberg2002). Clients' experiences were assessed retrospectively. Client reports of having experienced deep affect in therapy were clearly related to both satisfaction with therapy and feeling that change had occurred. There was a significant relationship between clients' recognition of their therapist's affect-eliciting techniques and feelings of satisfaction and change. Pilero (Reference Pilero2004) concluded that emotional experiencing may be the final common pathway to therapeutic change.

In studies of EFT for depression higher mid-therapy emotional arousal was found to significantly predict outcome, whereas a client's ability to use internal experience to make meaning and solve problems added to the outcome variance over and above middle phase emotional arousal (Missirlian et al. Reference Missirlian, Toukmanian, Warwar and Greenberg2005). In addition, in a study of emotion-focused therapy of depression a curvilinear relation between emotional arousal and outcome was found showing that too much or too little arousal when emotion was being processed was not as predictive of outcome as was arousal 25% of the time (Carryer & Greenberg Reference Carryer and Greenberg2010). Thus, it appears that a combination of emotional arousal and reflecting on the emotion is a better predictor of outcome than either alone. In addition, productivity of aroused emotional expression as measured by the ability to mentalize and work with the aroused emotion was found to be an excellent predictor of outcome (Auszra et al. Reference Auszra, Greenberg and Herrmann2013; Greenberg Reference Greenberg2010).

In studies of EFT for trauma good client process early in trauma therapy has been found to be particularly important because it sets the course for therapy and allows maximum time to explore and process emotion related to traumatic memories (Paivio et al. Reference Paivio, Hall, Holowaty, Jellis and Tran2001). One practical implication of this research is the importance, early in therapy, of facilitating clients' emotional engagement with painful memories. Being able to symbolize and explain traumatic emotional memories in words helps promote their assimilation into one's ongoing self-narrative (van der Kolk Reference Van der Kolk and Alpert1995). This form of putting emotion into words allows previously unsymbolized experience in emotional memory to be assimilated into peoples' conscious, conceptual understandings of self and world, where it can be organized into a coherent story. Timing is also important, as there is strong evidence that debriefing immediately after a trauma has occurred is harmful in that such debriefing increases rather than decreases the likelihood that PTSD will develop (McNally et al. Reference McNally, Bryant and Ehlers2003). The activation of emotion in therapy for trauma appears useful only after PTSD has set in.

Regarding psychodynamic psychotherapy, emotional arousal is part of clinical lore. Vividly re-experiencing emotions in the transference is thought to contribute to therapeutic change (Luborsky Reference Luborsky1984; Spezzano Reference Spezzano1993), but objective evidence that this is an essential ingredient to psychodynamic therapeutic success may not be available. Monsen et al. (Reference Monsen, Odland, Faugli, Daae and Eilertsen1995) conducted a five-year follow-up study on personality-disordered patients who had been treated using a psychodynamic psychotherapy that had a particular focus on patient's consciousness of affect. Both during treatment and five years post treatment, researchers found significant and substantial changes in the degree to which patients were aware of affect, characterological defenses, and symptoms. Moreover, at the end of treatment, nearly three-quarters of the patients who met DSM-III criteria for both Axis I and Axis II diagnoses no longer met these criteria. This finding suggests that intensive psychotherapy focusing on warded-off affect is helpful to a group of patients, in whom most studies report only moderate to poor outcome.

In sum, the evidence from the psychotherapy research just reviewed indicates that the explicit, conscious reported experience of emotion is an important ingredient in therapeutic success across all of the modalities listed above, including those such as behavior therapy and CBT that do not in theory explicitly rely on such experiences. Yet, a coherent theory of the cognitive processing of emotion in therapy must account for the fact that emotion activated in therapy may be adaptive or maladaptive. As illustrated by the work on emotional expression during debriefing immediately after trauma versus after PTSD has set in, emotions at times need to be regulated and modified and at other times accessed and used as guides.

This balance can be understood if one hypothesizes that the relationship between the degree of arousal and the ability to create mental representations of one's own emotional state is quadratic (inverted U) rather than linear. If emotional arousal is too intense, the mentalizing function mediated by a network including the medial prefrontal cortex (Amodio & Frith Reference Amodio and Frith2006) goes off-line, limiting the capacity for reflection in emergency situations. If arousal is very low, then cognitive processing of emotion is not likely to occur. Arousal needs to be moderate in the psychotherapy session: more arousing than non-emotive therapies, but less arousing than the trauma itself. This inverted-U effect for emotion and medial prefrontal cortex function is parallel to that for dopamine agonism or antagonism and activation in the dorsolateral prefrontal cortex during spatial working memory (Vijayraghavan et al. Reference Vijayraghavan, Wang, Birnbaum, Williams and Arnsten2007). In order to reflect upon a given situation the context needs to be recalled and brought into working memory, which has been shown to involve this inverted-U function for dopamine. This inverted-U relationship is also expressed in the Yerkes–Dodson (Diamond et al. Reference Diamond, Campbell, Park, Halonen and Zoladz2007) law of arousal and performance (both motoric and intellectual), which states that performance level for complex tasks is best when arousal level is neither too high nor too low. Neuroimaging evidence of the important role of superior medial prefrontal cortex in mentalizing on the one hand (Amodio & Frith Reference Amodio and Frith2006), and the positive correlation between activity in this region and vagal tone (which indicates that when arousal is high and vagal tone is low the activity in this region is reduced) on the other (Thayer et al. Reference Thayer, Åhs, Fredriskon, Sollers and Wager2012), are consistent with the hypothesis that the process of creating mental representations of emotional experience is compromised when arousal level is too high.

The applicability of the inverted-U relationship between arousal level and mentalizing can be carried a step further in relation to the encoding of the original trauma. At extremely high levels of arousal during the traumatic event, one's ability to know what one was feeling at the time would be very limited. This is consistent with and extends the hypothesis (Nadel & Jacobs Reference Nadel and Jacobs1998) that high levels of arousal during trauma interfere with amygdala–hippocampal interaction such that encoding of context is impaired. As a result, when recalling what one experienced at the time the emotional content would be limited in detail and complexity. At the opposite end of the continuum, the nature of the trauma may have been emotional neglect associated with depressed or listless affect associated with low arousal. Both extremes could contribute to the lack of encoding of what one experienced at the time of the trauma and later lead to impoverished detail when recalling the emotions associated with the trauma.

There are several principles that follow from this. First, if there are deficits in emotional encoding at the time of the trauma a complete account of what one experienced originally needs to be formulated for the first time in therapy. Second, if during therapy a client is at the low end of the arousal curve during recall, the arousal level needs to be increased to achieve sufficient emotion activation, whereas if the arousal level is too high, the arousal level needs to be decreased. Third, in conditions of high arousal, such as an emergency or traumatic situation, attention is narrowed and emotional experience, if it occurs, is simplified and streamlined, whereas the kind of corrective experience that leads to change is a more complex blend of emotions, such as feeling accepted and cared for while simultaneously fearing criticism and rejection, which requires more moderate levels of arousal. Effective psychotherapy occurs in conditions of safety promoted by a therapeutic alliance in which the client can rely on the therapist to facilitate experiences that are new but not overwhelming.

5. Interactions of emotion, stress, and memory

It has long been understood that memory is influenced by the presence of both emotional arousal and physiological stress, which are inherent components of distressing events. An extensive cognitive behavioral literature exists on the influence of emotion on attention and memory (for review, see Hoscheidt et al. Reference Hoscheidt, Dongaonkar, Payne, Nadel and Reisberg2013; LaBar & Cabeza Reference Labar and Cabeza2006; McGaugh Reference McGaugh2003; Roozendaal et al. Reference Roozendaal, McEwen and Chattarji2009). For example, a person experiencing an emotional state will selectively attend to and process information that is consistent with her present emotional state, an attentional effect referred to as “emotional congruence.” Additionally, when a person experiences an event in a particular emotional state, the event is remembered best when the person is in a similar emotional state, referred to as emotion-dependent memory or more broadly, state-dependent memory (Eich et al. Reference Eich, Macaulay and Ryan1994). The intensity of emotion experienced during the original event, regardless of positive or negative valence, increases the likelihood that the memory will be recalled vividly and the original emotion re-experienced, including the visceral or bodily manifestations of that emotion (Talarico et al. Reference Talarico, LaBar and Rubin2004). These behavioral effects are likely mediated by interactions among many brain systems, including two that play an important role in mediating emotion and memory, the amygdala and the hippocampus. Considerable research with both animals and humans has shown that emotional arousal results in increased physiological interaction between the amygdala and hippocampus, which leads to enhanced encoding and long term consolidation of emotionally arousing information (Cahill Reference Cahill2000; Murty et al. Reference Murty, Ritchey, Adcock and Labar2010; Phelps Reference Phelps2004; Vyas et al. Reference Vyas, Mitra, Shankaranarayana Rao and Chattarji2002).

The additional influence of stress on emotional memory is complex, sometimes resulting in enhanced memory for prior events, and sometimes resulting in impaired recollection (Kim & Diamond Reference Kim and Diamond2002; Lupien et al. Reference Lupien, Friocco, Wan, Maheu, Lord, Schramek and Tu2005). The effects of stress are due in part to the activation of the hypothalamic–pituitary–adrenal (HPA) axis, which results in a cascade of stress hormones culminating in the release of glucocorticoids (cortisol) from the adrenal cortex. Many of the brain regions important for emotional memory (hippocampus, prefrontal cortex, amygdala) have dense concentrations of glucocorticoid receptors and the function of these brain regions is influenced by elevated stress hormones (de Quervain et al. Reference de Quervain, Henke, Aerni, Treyer, McGaugh, Berthold, Nitsch, Buck, Roozendaal and Hock2003). Although prolonged exposure to stress interferes with memory function, acute increases in glucocorticoids enhance the encoding and consolidation of new emotional memories (for review, see Lupien et al. Reference Lupien, Friocco, Wan, Maheu, Lord, Schramek and Tu2005; Maren Reference Maren1999; McGaugh & Roozendaal Reference McGaugh and Roozendaal2002). Interestingly, at the same time as enhancing memory for emotional experience, stress hormones may actually impair memory for the neutral elements of the same event (de Quervain et al. Reference de Quervain, Roozendaal, Nitsch, McGaugh and Hock2000, Reference de Quervain, Henke, Aerni, Treyer, McGaugh, Berthold, Nitsch, Buck, Roozendaal and Hock2003). In a study by Payne et al. (Reference Payne, Jackson, Ryan, Hoscheidt, Jacobs and Nadel2006), participants were subjected to a stressful social situation that resulted in increased cortisol levels, and then shown a narrated slide show that included both emotionally arousing and neutral information. Participants were impaired in recalling the neutral elements of the event immediately after the event, whereas memory for the emotionally salient and arousing information in the event was preserved relative to a no-stress control group. Subsequently, Payne et al. (Reference Payne, Jackson, Hoscheidt, Ryan, Jacobs and Nadel2007) showed that after one week, memory was further enhanced for emotionally arousing material whereas memory for closely matched neutral material was impaired. These findings are consistent with the notion of tunnel memory, where high levels of arousal facilitate memory for central details (presumably those most relevant to the emotional content of the event) at the expense of peripheral details (Burke et al. Reference Burke, Heuer and Reisberg1992; Christianson Reference Christianson1992; Christianson & Loftus Reference Christianson and Loftus1991).

Relevant to the present discussion are those cases where severe stress is experienced during an emotionally arousing event, such as rape, combat, witnessing an accident, or another personally traumatic event. In these circumstances, stress appears to enhance the encoding and subsequent memory for the emotionally-salient aspects of the experience. The emotionality of that subsequent recollection is probably further enhanced by the fact that neutral elements of the same event are less likely to be recalled later on. However, the encoding of the emotional responses at the time of a traumatic event may be compromised if the arousal level at the time is sufficiently high. For example, a large study of memory for the events of 9/11/2001 revealed that recall of the emotions experienced at the time event was worse than recall of the factual details (Hirst et al. Reference Hirst, Phelps, Buckner, Budson, Cuc, Gabrieli, Johnson, Lustig, Lyle, Mather, Meksin, Mitchell, Ochsner, Schacter, Simons and Vaidya2009).

It is important to emphasize that participants in Payne et al. (Reference Payne, Jackson, Ryan, Hoscheidt, Jacobs and Nadel2006; Reference Payne, Jackson, Hoscheidt, Ryan, Jacobs and Nadel2007) were exposed to social stress immediately prior to experiencing the to-be-remembered event, and therefore the results inform how stress affects the initial acquisition and early consolidation of emotional memories. In contrast, stress experienced during recollection of prior events consistently produces memory impairment, regardless of emotionality of the material (Kuhlmann et al. Reference Kuhlmann, Kirschbaum and Wolf2005a; Reference Kuhlmann, Piel and Wolf2005b). This is consistent with many real-world examples where stress during memory retrieval can have negative consequences, such as taking an examination or speaking in front of an audience. This interference with memory retrieval may actually have a beneficial effect during exposure therapy. Cortisol has a facilitative effect on fear extinction (Soravia et al. Reference Soravia, Heinrichs, Aerni, Maroni, Schelling, Ehlert, Roozendaal and de Quervain2006). Roozendaal and colleagues (Roozendaal et al. Reference Roozendaal, Okuda, Van der Zee and McGaugh2006; see also Vocks et al. Reference Vocks, Legenbauer, Wachter, Wucherer and Kosfelder2007) have suggested that the fear extinction during exposure therapy may be further enhanced by the role of cortisol in attenuating retrieval of past traumatic events.

These mechanisms are relevant to the encoding and storage of highly emotional and stressful experiences that are later recalled and discussed in psychotherapy. For past events not associated with either extremely high or low arousal at the time of occurrence, emotional memories will be easily accessible later on, and recollection will likely emphasize the emotional elements of the memory, to the detriment of neutral information. Recollection under these circumstances is also likely to reinstate the emotional experience, including the visceral components of that emotion that were experienced originally. For past traumatic events at the extremes of arousal, however, accessing of details including what one experienced at the time may be much more limited because of the influence of arousal on initial encoding. Our discussion highlights the integral relationship between past memories and ongoing emotional responses, and also helps to explain how recollection of prior memories can be distorted over time as emotional components of experience take precedence over other, possibly more moderating, information. In fact, Rubin et al. (Reference Rubin, Berntsen and Johansen2008) have proposed that PTSD symptoms derive not from the emotional experience of the original event per se, but from the explicit memory for that event that is constructed and reconstructed through subsequent recollections. This leads us to a broader discussion of the dynamic nature of memories.

6. The dynamic nature of memory

Following the experience of an event, the memory for that event undergoes a process of stabilization, often referred to as consolidation, that renders the memory more resistant to interference from similar experiences, and more likely to be successfully recollected later on (Dudai Reference Dudai2004; McGaugh Reference McGaugh2000). Consolidation, however, does not result in a memory representation that is immutable. Memories are not a perfect record of the past, but undergo revision and reshaping as they age and, importantly, are recollected. The notion of memory retrieval as a dynamic and constructive process rather than a mere replay of the original event has substantial empirical support, beginning with Bartlett's (Reference Bartlett1932) famous “War of the Ghosts” study. Using what he called the method of repeated reproduction, Bartlett showed that repeated recollections of the story typically led to a shortened, more stereotyped version of it, with details either discarded, transformed, or added. Bartlett's observational study was replicated empirically by Bergman and Roediger (Reference Bergman and Roediger1999), who also found that participants distorted information and imported novel propositions into the story, most prominently after a delay of six months.

Studies such as these focusing on memory for newly acquired short stories, or lists of words, pictures, or scenes, may have limited relevance to the remembrance of the rich and personally relevant emotions and details associated with autobiographical memories. In contrast to Bartlett's (Reference Bartlett1932) observations, autobiographical memories that are highly emotional and hold importance for the individual often become increasingly consistent in the manner they are recalled, even “scripted,” across repeated recollections (Nadel et al. Reference Nadel, Campbell and Ryan2007; Neisser & Harsch Reference Neisser, Harsch, Winograd and Neisser1992). Neisser and Harsch (Reference Neisser, Harsch, Winograd and Neisser1992) suggest that repeated retellings of these memories gives structure to the narrative that improves consistency over time. Interestingly, the retelling of these stories may also result in an increasing number of details being recalled across repeated retrieval sessions, even after a year (Campbell et al. Reference Campbell, Nadel, Duke and Ryan2011). Whether those additional details are accurate, however, is impossible to tell. Studies of autobiographical memory retrieval can rarely assess accuracy, because there is rarely a veridical account of the original event available for comparison. One notable exception is Ulrich Neisser's (Reference Neisser1981) analysis of the testimony of John Dean regarding his involvement in the Watergate scandal during the Nixon administration. Neisser compared Dean's exhaustive accounts of intensely emotional and important meetings that transpired in the White House oval office with the original tape recordings of the very same meetings, made in secret by Nixon. Neisser found that Dean's accounts were generally devoid of correct details, despite his high confidence in the accuracy of his recollections. Nevertheless, Neisser noted that the core information contained in Dean's memories – who knew what, who did what – was accurate, even if each of the event memories themselves had been revised and reconstructed to a surprising degree, a phenomenon that Neisser dubbed “repisodic memory.”

Also relevant to this discussion is the recollection of flashbulb memories – vivid, long-lasting memories for emotionally arousing, often shocking events that carry strong social importance. These memories contain both an “event” portion and an “autobiographical” component – you remember what happened in New York on 9/11/2001 (the event) but also where you were, who you were with, who you told, and the emotional reactions of you and others around you (the autobiographical part). Participants are usually asked on two subsequent occasions to recall key pieces of personal information, such as where they were when they heard the news, and who told them the news. Even a year or more after the first recollection, 75% to 80% of people provide consistent answers to these questions (Berntsen & Thomsen Reference Berntsen and Thomsen2005; Cohen et al. Reference Cohen, Conway and Maylor1994; Davidson et al. Reference Davidson, Cook and Glisky2006). However, consistency over time is not equivalent to accuracy. Pezdek (Reference Pezdek2003) found that nearly three-quarters of participants incorrectly reported that on 9/11/2001 they saw a videotape of the first plane striking the first tower. Similarly, Ost et al. (Reference Ost, Vrij, Costall and Bull2002) reported that 45% of their United Kingdom sample reported that they had seen a videotape of the car crash that killed Diana, Princess of Wales. In both cases, no videotaped record exists of the incidents. Coluccia et al. (Reference Coluccia, Bianco and Brandimonte2006) suggests that, particularly after repeated recollections, additional information is incorporated into the memory that is either self-generated or experienced through other sources after the event (see also Neisser & Harsch Reference Neisser, Harsch, Winograd and Neisser1992). The revised version of the memory is then recalled consistently over time. Interestingly, these changes in the details of the memory have no bearing on the confidence of the person remembering – what people “remember” at any given time is vivid and emotionally engaging, regardless of the veridicality of their recollection (Neisser & Harsch Reference Neisser, Harsch, Winograd and Neisser1992; Talarico & Rubin Reference Talarico and Rubin2003). A similar pattern is observed for traumatic event memories. A review (Van Giezen et al. Reference Van Giezen, Arensman, Spinhoven and Wolters2005) of 17 studies of memories for both combat and noncombat traumatic experiences identified inconsistencies when participants were asked to recall the memories on two different occasions (for discussion, see Rubin et al. Reference Rubin, Berntsen and Johansen2008).

The reconstructive nature of memory could be construed as a design flaw. Indeed, in the flashbulb memory literature, these changes in memory are referred to not merely as revisions, but as errors and inaccuracies. Why would a memory system exist that does not provide a stable and faithful representation of past events? Cognitive research has clearly demonstrated that people make errors during recollection, even when they are extremely confident in their attributions, and that these errors increase with time and repeated recollections. However, the dynamic nature of memory can also be construed as beneficial: It provides an important mechanism for understanding how existing knowledge can be updated in light of new information. Klein et al. (Reference Klein, Cosmides, Tooby and Chance2002) describe the importance of memory as an adaptive function, one that can (and should) be updated over time depending upon new experiences and changes in the environment. Adaptive behavior, according to Klein et al. (Reference Klein, Cosmides, Tooby and Chance2002) depends on an interaction between decision rules derived from multiple experiences that guide the behavior of the individual (semantic memory) and the recollection of specific events (episodic memories) that provide boundary conditions or expectations to those rules. Relevant to the current discussion, the notion of a dynamic and adaptive memory system is critical to understanding how memories that are painful or disturbing might be transformed through the process of psychotherapy and the corrective experience. The sections that follow discuss how memory updating and the interactive nature of episodic and semantic memories may provide insights into the mechanisms underlying therapeutic change.

7. Memory reconsolidation

As mentioned earlier, consolidation refers to the idea that event memories undergo a stabilization process that renders the memory less susceptible to interference from similar experiences, and more likely to be successfully recollected after the passage of time. In recent years, two rather different versions of what happens in the brain during consolidation have emerged. One version, often referred to as the “standard model of memory consolidation” (Squire & Alvarez Reference Squire and Alvarez1995) emphasizes that the brain structures mediating retrieval shift over time, from medial temporal lobe structures including the hippocampus, to neocortical structures including the prefrontal cortex. Importantly, as this consolidation process and concomitant transition takes place, the content of memories presumably remains unchanged.

Nadel and Moscovitch (Reference Nadel and Moscovitch1997; Moscovitch & Nadel Reference Moscovitch and Nadel1999) developed an alternative theory of memory consolidation, known as the multiple trace theory (MTT). Rather than focusing on the mere passage of time, the theory addresses the question of how repeated recollections of prior events lead to strengthening of the memory representation for the original event. Similar to the standard model of consolidation, MTT posits that the establishment of long-term memories involves a lengthy interaction between the hippocampal region of the medial temporal lobes and neocortical regions. Unlike standard theory, MTT posits that the hippocampus remains an integral part of the memory trace and is thus always involved in retrieval of long-term episodic memories regardless of the age of the memory. Evidence supporting this view comes from neuroimaging studies showing that retrieval of detailed episodic memories activates the hippocampus no matter how old these memories are, even after 40 or more years (e.g., Ryan et al. Reference Ryan, Nadel, Keil, Putnam, Schnyer, Trouard and Moscovitch2001; see Moscovitch et al. Reference Moscovitch, Nadel, Winocur, Gilboa and Rosenbaum2006 for review).

The standard view of memory consolidation suggests that immediately after learning there is a period of time during which the memory is fragile and labile, but that after sufficient time has passed, the memory is more or less permanent. During this consolidation period, it is possible to disrupt the formation of the memory, but once the time window has passed, the memory may be modified or inhibited, but not eliminated. In contrast, MTT suggests that every time a memory is retrieved, the underlying memory trace once again enters into a fragile and labile state, and thus requires another consolidation period, referred to as “reconsolidation” (Nadel et al. Reference Nadel, Samsonovich, Ryan and Moscovitch2000). The reconsolidation period provides an additional opportunity to amend or, under appropriate circumstances, even disrupt access to the memory.

MTT proposes that each time an episodic memory is recollected or retrieved, a new encoding is elicited, leading to an expanded representation or memory trace that makes the details of the event more accessible and more likely to be successfully retrieved in the future. This process is primarily initiated by active retrieval or recollection, although off-line reactivation that occurs during sleep and indirect reminder-induced reactivation can also trigger it (Hardt et al. Reference Hardt, Einarsson and Nader2010; Hupbach et al. Reference Hupbach, Gomez, Hardt and Nadel2007; Nadel et al. Reference Nadel, Campbell and Ryan2007; Wilson & McNaughton Reference Wilson and McNaughton1994). Critically, each time an event is recollected and re-encoded, an updated trace is created that incorporates information from the old trace but now includes elements of the new retrieval episode itself – the recollective experience – resulting in traces that are both strengthened and altered. This altered trace may incorporate additional components of the context of retrieval, new relevant information pertaining to the original memory, or even new information that is inadvertently (perhaps incorrectly) generated during the act of retrieval, as in the case of the flashbulb memories described earlier. In this regard, MTT holds that memories are not a perfect record of the original event but undergo revision and reshaping as memories age and, importantly, are recollected. The reconsolidation process, by this view, results in memories that are not just stabilized and strengthened, but are also qualitatively altered by the recollective experience.

This dynamic interplay between retrieval of the memory and reconsolidation has been demonstrated experimentally both in animals and humans. Animal studies have shown that well-established, supposedly consolidated, memories can be disrupted after reactivation (Nader et al. Reference Nader, Schafe and Le Doux2000), even when that reactivation is nothing more than a reminder of the spatial context of the original event. The utility of this to control fearful responses emerged from a study by Nader et al. (Reference Nader, Schafe and Le Doux2000). In this study, rats were conditioned to fear a tone, and then up to 14 days later were presented with an unreinforced presentation of the tone. This “reminder” was followed immediately by an injection into the amygdala of anisomycin, a protein synthesis inhibitor that blocks the consolidation of fear memories. Even though the time window of consolidation had passed, the rats that received anisomycin following the reminder failed to show evidence of conditioned fear on subsequent test trials. Rats injected with saline following a reminder showed normal conditioned fear. These results suggest that fear memories undergo reconsolidation every time they are retrieved, and that this reconsolidation process can be disrupted, apparently eliminating the previously well-learned fear responses.

In discussing memory reconsolidation it is important to distinguish it from the behavioral phenomenon of extinction. In animal studies of both reconsolidation and extinction, an element of the learning situation (the context, or a conditional stimulus – CS) is presented without its previous consequence – the unconditioned stimulus (US). In most of the experiments with rats the US is a shock administered through the grid floor. Because of this similarity, there has been some question about how to separate the two – and this has considerable importance in the present context, because reconsolidation is assumed to actually change components of the reactivated memory, whereas extinction is assumed to merely create a new memory that overrides the previously trained response (Milad & Quirk Reference Milad and Quirk2002). Thus, an “extinguished” response is not really gone, because it can spontaneously recover over time, or be reinstated if the organism is exposed to a relevant cue in a new context. Recent work has shown that the cellular/molecular cascades in these two cases are different, and that whether reconsolidation or extinction is initiated depends upon the temporal dynamics of the test procedure, and how recently the memory in question was formed and/or reactivated (de la Fuente et al. Reference de la Fuente, Freudenthal and Romano2011; Inda et al. Reference Inda, Muravieva and Alberini2011; Maren Reference Maren2011). At this time we can be certain that reconsolidation and extinction represent distinct reactions to reactivating a memory, but the conditions eliciting one or the other remain to be fully determined.

In humans, Hupbach et al. (Reference Hupbach, Gomez, Hardt and Nadel2007, Reference Hupbach, Hardt, Gomez and Nadel2008) have shown that when memories are reactivated through reminders, they are open to modification through the presentation of similar material that then becomes incorporated into the original event memory. Using a simple interference paradigm, Hupbach et al. (Reference Hupbach, Gomez, Hardt and Nadel2007) had participants learn a set of objects during the first session. Forty-eight hours later, one group of participants was reminded of the first session and immediately afterward learned a second set of objects. A second group received no reminder and simply learned a second set of objects. Another 48 hours later, they were asked to recall the first set of objects only, that is, the objects they learned during the first session. Participants in the “reminder” condition showed a high number of intrusions from the subsequently learned object set, whereas those who had not been reminded showed almost no intrusions. The results demonstrated that updating of pre-existing memories can occur in humans, and that this updating is dependent upon reactivation of the original memory. Hupbach et al. (Reference Hupbach, Hardt, Gomez and Nadel2008) subsequently showed that reminders of the spatial context of the original event were the most effective in triggering the incorporation of new information into the existing memory.

The processes of reactivation and re-encoding, and consolidation and reconsolidation, have important clinical implications for understanding the psychotherapeutic process. MTT provides a way of understanding how distressing emotional memories can be both strengthened over time and also altered through the corrective experience. Consider, for example, an emotionally distressing event such as a betrayal or abandonment. As we have seen, the emotional reaction is an integral component of the memory, connected via the spatial and temporal contexts to the event and bound to the self, forming an autobiographical memory. The more highly arousing the emotional reaction, the more likely the evoking situation will be remembered later on (McGaugh Reference McGaugh2003). When a memory is recalled, the emotional response is re-engaged and the amygdala reactivates the sympathetic response. According to MTT, the recollected event and its newly experienced emotional response will be re-encoded into a new and expanded memory trace. Thus, memory for the original traumatic incident is strengthened, making it (and the now intensified emotional response) even more likely to be accessed in the future.

MTT also provides a mechanism for understanding how this same emotional memory might be revised. During therapy, patients are commonly asked to recall and re-experience a painful past event, often eliciting a strong emotional reaction. If the psychotherapy process leads to a re-evaluation of the original experience, a new, more adaptive and perhaps more positive, emotional response may ensue. The corrective experience occurs within a new context, the context of therapy itself, which can then be incorporated into the old memory through reconsolidation. It is important to reiterate that MTT suggests this process is not simply attributable to a new memory trace being created, but that the original event memory itself is transformed in fundamental ways. It is conceivable that once this transformation has taken place the original memory, including the associated emotional response, will no longer be retrievable in its previous form. By this view, psychotherapy is a process that not only provides new experiences, but also changes our understanding of past experience in fundamental ways through the manipulation of memory.

If experiencing a qualitatively different emotion during recollection can have a modifying effect on subsequent emotional responsivity to memories, it is plausible that drug-induced blockade of the new memory formation can lead to the same result. Taking the notion of reconsolidation one step further in humans, researchers have now begun to investigate the possibility of modifying previously acquired traumatic memories by using drugs to block the emotional response during recollection. For example, administration of propranolol, a beta-adrenergic antagonist, may block reconsolidation of fear memory in rats by indirectly influencing protein synthesis in the amygdala (Debiec & Ledoux Reference Debiec and Ledoux2004). The effect of propranolol in altering the reconsolidation of emotional memories has been demonstrated in humans in an fMRI study (Schwabe et al. Reference Schwabe, Nader, Wolf, Beaudry and Pruessner2012). Propranolol has been administered to individuals immediately after experiencing a traumatic event (Pitman et al. Reference Pitman, Sanders, Zusman, Healy, Cheema, Lasko, Cahill and Orr2002), and also to PTSD patients immediately after they recall traumatic memories (Brunet et al. Reference Brunet, Orr, Tremblay, Robertson, Nader and Pitman2008), blocking the emotional response to the memory and, in both studies, leading to decreased emotional responsivity during subsequent recollections. Although promising, this treatment has initiated heated debate regarding the legal and ethical issues associated with “memory dampening,” as it has been called (Kolber Reference Kolber2006; Tenenbaum & Reese Reference Tenenbaum and Reese2007).

8. Semantic memory is integrated with autobiographical memory

Autobiographical and semantic memory seem, at least phenomenologically, quite different from one another. Autobiographical or episodic recollection involves thinking about a past event – it is personal, emotional, imbued with detail, and temporally and spatially unique; and it often has great relevance to our sense of self and the meaning of our lives. Semantic memory, on the other hand, has to do with the knowledge and rules governing behavior that have been acquired through a lifetime of experiences – it is factual and typically devoid of emotion or reference to the self or specific times and places. Although semantic knowledge conveys meanings, it is rarely the kind of personal meaning embodied in autobiographical and episodic memories. Instead, it provides us with expectations and enables us to predict the outcomes of new situations using the generic knowledge gained from similar situations in the past. The distinction, as outlined by Tulving (Reference Tulving1983), focused originally on the different types of information processed by the two systems, unique spatial-temporal contexts for episodic memory, and facts and concepts for semantic memory. More recently, Tulving (Reference Tulving2002; Reference Tulving, Terrace and Metcalfe2005) has emphasized that what distinguishes episodic memory is not so much the type of information being processed, but instead the phenomenal experience of remembering, or autonoesis. According to Tulving (Reference Tulving2002; p. 5) “It [episodic memory] makes possible mental time travel through subjective time, from the present to the past, thus allowing one to re-experience, through autonoetic awareness, one's own previous experiences. Its operations require, but go beyond, the semantic memory system.” This updated formulation suggests that episodic and semantic memory are representational systems that together capture both the regularities and irregularities of the world, allowing people to create concepts and categories (semantic memories) and also capture the time and place when one particular combination of entities was experienced, yielding an episode that may or may not be consistent with one's prior expectations.

It has long been assumed that episodic and semantic memories are relatively independent of one another, both functionally and anatomically (Aggleton & Brown Reference Aggleton and Brown1999; Schacter & Tulving Reference Schacter, Tulving, Schacter and Tulving1994; Schacter et al. Reference Schacter, Wagner, Buckner, Tulving and Craik2000; Tulving & Markowitsch Reference Tulving and Markowitsch1998). Recent research, however, has called this independence into question (see Ryan et al. Reference Ryan, Hoscheidt, Nadel, Dere, Easton, Huston and Nadel2008b for review). In a series of functional MRI studies, Ryan and colleagues demonstrated that both semantic and episodic retrieval results in a similar pattern of hippocampal activation, particularly when the tasks were matched for spatial content (Ryan et al. Reference Ryan, Cox, Hayes and Nadel2008a, Reference Ryan, Lin, Ketcham and Nadel2010; Hoscheidt et al. Reference Hoscheidt, Dongaonkar, Payne, Nadel and Reisberg2013). Consistent with Tulving (Reference Tulving2002), semantic memory and episodic memory are seen as interactive and complementary systems. Both semantic structures and singular episodic memories are important for identifying familiar circumstances, interpreting novel events and predicting outcomes, and choosing appropriate behaviors in response to situations and personal interactions. Barsalou (Reference Barsalou, Neisser and Winograd1988) has long championed the idea that semantic knowledge is embedded within a network of autobiographical memories. Episodes are represented as single events that are connected to other related episodes. Semantic memory is essentially derived from similar event memories that can be convolved to emphasize common information that is experienced across contexts, giving rise to what we call semantic memory. This idea is the basis of latent semantic analysis models (Landauer & Dumais Reference Landauer and Dumais1997). By this view, semantic information may be indistinguishable from episodic memory at the level of the brain when it is first acquired, and only later becomes differentiated as similar experiences accumulate and structural regularities and rules are derived. The information can then be retrieved separately from a specific context if necessary.

According to Barsalou (Reference Barsalou, Neisser and Winograd1988), “There are no invariant knowledge structures in memory. Instead, people continually construct unique representations from loosely organized generic and episodic knowledge to meet the constraints of particular contexts” (p. 236). Instead of focusing on abstracted concepts, Barsalou emphasizes the critical role of personally relevant instances for generating semantic knowledge. A concrete example comes from a study by Ryan et al. (Reference Ryan, Cox, Hayes and Nadel2008a). They asked participants to generate exemplars to the cue “kitchen utensils,” and then asked people to describe what they were thinking about as they generated the items. Their responses were at the same time similar and yet uniquely personal. Every participant reported something like, “I pictured myself standing in my kitchen, looking around the room, opening drawers and then looking in the cabinets.” Note that it is “my kitchen” that guided the responses of the individual, rather than a disembodied “typical kitchen,” leading a few individuals to give items like “espresso maker” a prominent place in the list. The observation is interesting because it suggests that semantic memory is not simply a stable record of past learning but something that is generative, flexible, contextually bound, and subject to revision through personal experience. Semantic memory is generated anew each time it is required, in much the same way as Bartlett (Reference Bartlett1932) and others (Bergman & Roediger Reference Bergman and Roediger1999; Nadel et al. Reference Nadel, Campbell and Ryan2007) have noted: Episodic memories are reconstructed and revised over time through multiple retrievals. This stands in contrast to the classic distinction between episodic and semantic memories and the assumption that semantic memory, at least, is a faithful record of prior learning.

What are the implications of viewing episodic and semantic memory as interactive systems for understanding psychotherapy? It suggests that distressing or traumatic event memories that elicit emotional responses are incorporated into semantic structures that are used to predict the outcomes of subsequent experiences and to choose appropriate (or inappropriate) emotional and behavioral responses in novel situations. It is easy to see how highly emotional and accessible memories from the past become the dominant basis for maladaptive responses in novel circumstances that share some characteristics with the original distressing event.

Importantly, this formulation also suggests that there are multiple routes to behavioral change and the “working through” process. The new experiences in therapy that update prior event memories through reconsolidation also contribute to a change in semantic structures. Applying the new knowledge and experiencing the results in a variety of contexts can be conceptualized as creating multiple episodic experiences that will broaden the range of applicability of new knowledge encoded in semantic memory. Linkage to emotional responses (as proposed in our integrated memory model) is expected to translate into greater adaptive flexibility and success relative to the difficulties that led the client to seek treatment.

9. Implications of this integrative synthesis

To reiterate, emotional responses, autobiographical memories, and semantic structures derived from them are inextricably linked. Together they form an integrated memory structure that can be accessed by many cues – emotional responses including action tendencies and behaviors expressive of emotion, perceptual details associated with the event(s), and the derived principles, rules, and schemas used to interpret novel situations. All of those elements have the ability to activate the memory structure, and importantly, once activated, any one of the components has the potential to update other components of the structure via reconsolidation. Emotional responding is not separate from the event memories that occurred when that response was first experienced. Nor are semantic structures accessed without reinstating personally relevant information, and, particularly under circumstances where the memory was strongly reconsolidated, the specific memories that add unique information to that structure.

Given these considerations, it becomes possible to understand each of the major modalities discussed above as focusing on a particular way of entering or engaging the integrated memory structure (see Fig. 1). Behavior therapy initially engages emotional responding with a greater emphasis on implicit rather than explicit processes. EFT also initially preferentially engages emotional responding but with a greater emphasis on explicit than implicit processes. CBT engages semantic memory initially, and Psychodynamic Psychotherapy, with its emphasis on the here and now in the transference situation and its relation to past experiences, focuses on autobiographical memory as a point of entry. As we discuss in sections 9.1–9.4, however, a more comprehensive understanding of how each modality works requires consideration of the entire memory structure.

Figure 1. Points of entry into the integrated memory structure for four types of psychotherapy.

The integrated memory model provides an opportunity to develop a common language that spans disciplines and a common mechanism underlying change in all psychotherapeutic modalities. We suggest that the 450 forms of therapy (MacLennan Reference MacLennan1996), to the extent that they are effective, address different aspects of a common phenomenon, and the success of practitioners of a given modality depends upon their ability to access an integrated memory structure that may include aspects of experience not typically emphasized in the formal explication of that modality. The model provides a way to highlight the unique aspects of each modality that may explain why each type of therapy is best suited to address particular types of maladaptive behaviors and distressing emotions. The model may also provide new information by suggesting ways to optimize change within each therapy modality. Although we cannot address each and every type of psychotherapy here, we will briefly discuss the implications of the integrated memory model for the four therapy modalities highlighted in this paper that focus primarily on one or another component of the model: autobiographical memory, emotional responding, the semantic structure, or some combination of these. The implications that we present in sections 9.1–9.5 are theory-driven and remain to be empirically tested.