Journal of Consulting and Clinical Psychology Copyright 2007 by the American Psychological Association

2007, Vol. 75, No. 6, 875– 887 0022-006X/07/$12.00 DOI: 10.1037/0022-006X.75.6.875

Emotional Processing in Experiential Therapy:

Why “the Only Way Out Is Through”

Antonio Pascual-Leone Leslie S. Greenberg

University of Windsor York University

The purpose of this study was to examine observable moment-by-moment steps in emotional processing
as they occurred within productive sessions of experiential therapy. Global distress was identified as an
unprocessed emotion with high arousal and low meaningfulness. The investigation consisted of 2 studies
as part of a task analysis that examined clients processing distress in live video-recorded therapy sessions.
Clients in both studies were adults in experiential therapy for depression and ongoing interpersonal
problems. Study 1 was the discovery-oriented phase of task analysis, which intensively examined 6
examples of global distress. The qualitative findings produced a model showing: global distress, fear,
shame, and aggressive anger as undifferentiated and insufficiently processed emotions; the articulation
of needs and negative self-evaluations as a pivotal step in change; and assertive anger, self-soothing, hurt,
and grief as states of advanced processing. Study 2 tested the model using a sample of 34 clients in global
distress. A multivariate analysis of variance showed that the model of emotional processing predicted
positive in-session effects, and bootstrapping analyses were used to demonstrate that distinct emotions
emerged moment by moment in predicted sequential patterns.

Keywords: emotional processing, emotion-focused therapy, distress, task analysis, change mechanisms

Within the field of psychotherapy research a growing number of major approaches to therapy, the claim still needs elaboration from
investigators have suggested that the most essential questions to be both phenomenological and developmental perspectives. Further-
addressed refer to the process of psychotherapy. In other words, more, it will be important to elaborate the role of emotional
“how” does change happen (Orlinsky, Grawe, & Parks, 1994)? processing in different treatment approaches, because facets of
Experimental research by Hunt (1998) has concluded that when it emotional processing likely apply differently across treatment
comes to recovering from dysphoria, “the only way out is models (Greenberg & Pascual-Leone, 2006). To formulate some
through.” The purpose of the current article, which borrows from specific model of how experiencing emotion actually changes
her title, is to show why that conclusion may be true in the context people— or how “emotion changes emotion” to use Greenberg’s
of psychotherapy. This article looks at how distress is changed in (2002) phrasing—researchers must examine the mechanisms un-
psychotherapy. derlying this process directly.
Emotional processing is described as a central process in In their study on psychodynamic processes, Jones, Pare, and
exposure-based therapies (i.e., Foa & Kozak, 1998) as well as in Pulos (1992) highlighted the need for more “in motion’ descriptors
integrative approaches to short-term psychodynamic psychothera- of the evolution of change processes that permit an understanding
pies (i.e., Fosha, 2000; McCullough et al., 2003). The role of of process as a sequence of patterns that extend over time” (p. 14).
processing underlying emotion has recently emerged as important Again, in an article that applied dynamic systems to cognitive–
to cognitive behavioral formulations of problems like generalized behavioral therapy (CBT) processes, Hayes and Strauss (1998)
anxiety (see Borkovec, Alcaine, & Behar, 2004) and in cognitive argued that treatment research should study growth curve process
therapy in general (Samoilov & Goldfried, 2000). Moreover, El- in an effort to generate trajectories of change (see also Hayes et al.,
liott, Watson, Goldman, and Greenberg (2004) provided a broad 2007). The current study follows this call to describe and examine
framework for how emotional change occurs through the intrap- patterns of client change, this time with respect to emotional
ersonal tasks described in emotion-focused therapy (EFT). How- processing in experiential therapy.
ever, although emotion has been championed in recent years as How clients might process highly aroused states of negative
playing a crucial role in the process of client change in several feeling is a question already considered by Greenberg and Paivio
(1997). Based on both clinical theory and practice, they proposed
a model for the restructuring of what they referred to as “bad
feelings.” Greenberg and Paivio’s model of client processes re-
Antonio Pascual-Leone, Department of Psychology, University of flects the treatment steps for evoking, exploring, and restructuring
Windsor, Windsor, Ontario, Canada; Leslie S. Greenberg, Department of
unhealthy emotion. As such, it stands as the null hypothesis for the
Psychology, York University, Toronto, Ontario, Canada.
This article is partly based on the doctoral thesis of Antonio Pascual-
current study. According to their model, restructuring bad feelings
Leone. begins with attending to the aroused feelings followed by explor-
Correspondence concerning this article should be addressed to Antonio ing the cognitive–affective sequences that generate those bad feel-
Pascual-Leone, Department of Psychology, University of Windsor, 401 ings. Eventually this leads to the experience of the output of some
Sunset, Windsor, Ontario N9B 3P4 Canada. E-mail: [email protected] core maladaptive emotion scheme. Often, this is a form of fear,

875
876 PASCUAL-LEONE AND GREENBERG

shame, guilt, or sadness that represents an unhealthy and painful way Clients were selected from clinical trials of the resolution of
of perceiving and experiencing oneself; one that is regrettably familiar interpersonal emotional problems (Greenberg & Malcolm, 2002;
to the client, like an age-old emotional wound. Greenberg and Paivio Paivio & Greenberg, 1995). The treatment protocol was EFT (also
described the source for changing maladaptive emotion as being the known as process experiential therapy) to help individuals resolve
introduction of something new, a new adaptive emotion that leads to long-standing interpersonal grievance. Clients had given informed
a new perspective. In short, the essence of emotional reorganization is consent to have their sessions studied for psychotherapy research.
the creation of some healthy emotional alternative (i.e., assertiveness The sessions selected ranged between Sessions 2 and 9 (M ⫽ 5.2,
or appropriate grieving) in the face of a dysfunctional feeling and SD ⫽ 2.9). Each client’s session was selected on the basis of
belief. The key transformational process in accelerated experiential expert consensus by two assessors that it was a good representation
dynamic psychotherapy, according to Fosha (2000), bears a lot of of the treatment protocol, demonstrated substantial emotional dis-
similarity in that it involves moving from distress through state tress, and would therefore be fruitful for a discovery-oriented
transformation into an embodied (visceral, somatic) experience of examination. In the second round of discovery-oriented analyses,
core affect without defense, anxiety, or shame. three of the original six cases were then examined across the
The proposed models of emotional transformation would benefit course of their treatments (12–14 sessions).
from a more detailed moment-by-moment description of how
emotions actually interact and produce change. These ideas also Procedure
need to be tested empirically. For a model of emotional processing
to be applied and used by researchers and/or clinicians it must be The discovery-oriented phase of task analysis involves a sys-
concretized by observable phenomena and accompanied by some tematic program for constructing a specific model of client change
form of measurement. Finally, as indicated by Hayes et al. (2007), process. This phase of task/process analysis was done in three
to show its broader clinical relevance, researchers must elaborate steps, by developing (a) a rational model; (b) an empirical model;
a moment-by-moment model of emotional processing in how it and, finally, (c) a synthesized model (for a detailed description of
relates to the development of therapeutic changes sustained across this method, see Greenberg, 2007).
treatment. In other words, the role of emotional processing must be
considered across treatment as well as in the moment and in the Step 1: Rational Model
session. This project addressed these issues in two separate studies
by developing a model, testing it, and studying its process of Drawing on the literature, investigators specified a model of
change over the whole treatment. how they thought the process of change occurred. Both investiga-
Aside from work in EFT (Greenberg, 2002; Greenberg & Paivio, tors hold doctorates and are clinical psychologists, practicing and
1997; Greenberg & Safran, 1987), psychotherapies have not clearly teaching experiential therapy. The model was developed within a
articulated how people actually deal with their feelings of distress. On framework of emotion research.
one hand, ideas of exposure and habituation do not adequately ac- Specifying the task. The rational analysis began by delineating
count for the emergence of categorically new experiences. On the the emotional processing task, circumscribed by a problem state
other hand, explanations that simply appeal to a growth tendency still and an eventual end marker for when the task is achieved. Global
beg for some type of refinement and elaboration. Furthering research- distress marker refers to the problem state. This category of feeling
ers’ understanding of precisely how individuals shift their experience (a) is an aversive state of suffering, (b) is high in expressive
from emotional distress (i.e., the problem state) to some more mean- arousal, (c) is nonagentic and without a clear sense of direction,
ingful and adaptive resolution state will open the way to advancement and (d) does not entail a clear and specific internal referent of
in both clinical theory and methods of intervention for any approach experience. This emotion state is best characterized as an emo-
to therapy that targets emotion. tionally expressed distress reaction, with little or no substantive
The task-analytic approach temporarily puts the therapist in the meaning elaboration. Commonly expressed examples of this type
conceptual background so as to focus on optimal client process of emotion include undifferentiated hurt, hopelessness, helpless-
(Greenberg, 1999, 2007). So, this study is not about the role of the ness, resignation, self-pity, unelaborated loneliness, vague self-
therapist in facilitating client emotional change but rather about blame, confusion, and undifferentiated complaint. Note that an
how clients actually change. Productive client processes are not expression of sadness may or may not be global distress, depend-
always self-evident, especially when it comes to negotiating pain- ing on the manner in which it is expressed.
ful emotion. Thus, the issue investigated here was whether emo- This sort of emotional state can be described as being of high
tional sequences exist during the in-session resolution of global expressive arousal and low meaningfulness in regard to some
distress. In this study, both discovery-oriented and validation personally sensitive theme and/or life event. Ratings of this state
phases of a task analysis were used to examine this change process reach a minimum of 4 out of 7 on the Emotional Arousal Scale
(Greenberg, 2007). (Warwar & Greenberg, 1999); or the client must verbally report
that his or her arousal is high, indicating that the emotion is
STUDY 1: DISCOVERY-ORIENTED PHASE activated. Moreover, the person experiences himself or herself as
a passive recipient of this state. This state is nonagentic, it has no
Method adaptive action tendency associated with it, and it usually lacks a
sense of direction. The object of distress is never well elaborated;
Participants
it is either unknown, minimally elaborated, or not addressed be-
In the first round of discovery-oriented analyses, the sample yond a subjective sense of victimhood. Global distress has been
consisted of six single sessions, each from a different client. referred to as secondary emotion in EFT and is often referred to as
 EMOTIONAL PROCESSING IN SESSION 877

defensive emotion or unformulated experience by

psychodynamic– experiential theorists (i.e., Fosha, 2000; Stern,
1997). Thus, both of these approaches discuss global distress as an
emotional reaction to some other, unarticulated, perhaps unformu-
lated, affective-meaning processes. The reliability of identifying
the marker state, global distress, was high, with reliability coeffi-
cients ranging from .76 to .91 (see Study 2).
Advanced processing indicators serve as end markers for the
task. Because emotional processing is an open-ended problem with
no specific end state, the advanced processing marker or end state
cannot be defined in the same structured manner as the start
marker, global distress. The “solution” for the transformation of
global distress through a task that is referred to as emotional
processing lies in meeting certain performance criteria that char-
acterize skillful affective-meaning making.
Advanced emotional processing is indicated by any of the
following: (a) new feelings being freshly experienced in the mo-
ment, expressed clearly in an integrative and affirmative fashion;
(b) productive meaning making associated with the ongoing ex-
perience; (c) positive self-evaluation (either implicit or explicit);
(d) agentic stance with an adaptive action tendency; (e) the iden-
Figure 1. Rational/empirical model: A state-transition diagram for the
tification of clear and specific object(s) of emotional concern. As
emotional processing of distress.
with global distress, the reliability of identifying states that met
criteria as advanced processing markers was high, ranging from
.76 to .91 (see Study 2).
Specifying the task environment. The environment in which grounded findings, each of the components in Figure 1 is a
participants attempt this emotional processing task (a) is a validat- necessary stepping-stone in a nonlinear ladder toward emotional
ing, empathic interpersonal environment, supportive of emotional processing. Nevertheless, getting “stuck” in any of these states is
processes; (b) has minimal content directiveness; (c) has minimal what makes for disordered regulation and inadequate processing.
interruption of emotional continuity; and (d) focuses attention on Detailed descriptions and case transcripts elaborating the phenom-
the client’s bodily felt experience. enology of this differentiation process can be found in A. Pascual-
Leone and Greenberg (2006).

Steps 2 and 3: Empirical and Synthesized Models Affective-Meaning States in Processing Distress
Empirical analysis required the creation of models for trans-
Global Distress
forming distress by discerning the steps of advanced processing as
observed in actual client performances. This was paralleled with At the top of the model is global distress; this client marker state
developing criteria for the measurement of those steps. Thus, video is one of feeling pain, hopelessness, helplessness, complaints,
recordings of the six psychotherapy cases were systematically self-pity, confusion, despair, and so on. As described previously,
observed to discover the essential steps in the advanced processing this is a state of high expressive arousal (i.e., tears, emotional
of distress. Then, based on the full cycle of rational and empirical voice, etc.) and low specificity in meaning (i.e., the object of
analyses, a final synthesized model was developed. distress is often unknown, the client has no sense of direction).
Characteristically the marker of global distress emerges suddenly,
Results the person becomes dysregulated, and the specific concern at hand
is often very vague and global. Sometimes, when therapists ini-
Task analysis of the emotional processing of global distress tially explore this, clients explicitly state that they do not know
produced a synthesized rational/empirical model that represented why they are feeling so inundated with distress.
the final results of the discovery-oriented phase of research. The For example, in a prototypical statement of global distress,
diagram in Figure 1 shows key findings regarding clients who start Client 407 described, “I feel hurt, miserable, and angry and I’m
with an expression of nonspecific emotional distress and progress tired of it. It’s so overwhelming,” followed by, “I don’t know what
to the productive emotional processing of that distress. A measure that pain is” (9 min into Session 12). Client 076 said in an
was also developed for identifying model components called the emotional voice, “I feel terrible today . . . I burst into tears 10 times
Classification of Affective-Meaning States (CAMS; see A. a day,” but when the therapist inquired as to what was going on,
Pascual-Leone & Greenberg, 2005). the client replied, “I’m not sure . . . I feel all alone and stuff like
At the starting point of this model, the client in global distress that” (2 min into Session 7). When clients in a state of global
is already emotionally aroused and engaged. From then on the path distress begin to elaborate and differentiate their thoughts and
toward emotional processing is nonlinear and reveals that there is feelings, they subsequently move in one of two directions. The
more than one way in which clients therapeutically transform path on the right side of the model (toward fear or shame) is
states of global distress. However, according to these empirically described first (see Figure 1).
878 PASCUAL-LEONE AND GREENBERG

Fear/Shame at the same time, I feel like I desperately need that love/validation
[need].” This statement captures several aspects of a client’s dy-
This component state is both maladaptive and specific (com- namic construction of meaning in the processing of emotional
pared to the global nature of the distress state). Although they are experience. Notice that both the need and negative evaluation are
distinct emotions, fear and shame are functionally equivalent in partly evoked by the same releasing cues (i.e., the preceding state
their role here and moreover are often complementary aspects of of fear and shame), and yet these two parts are contradictory.
the same specific and maladaptive emotional experience. In other These conditions of both simultaneous emergence and inherent
words, as experiential states they embody similar pieces of infor- opposition make the negative evaluation and the need a dialectical
mation (i.e., affectively laden meaning) about the self vis-à-vis the pair. The client will subjectively experience this internal state of
environment. Clients in fear and shame experience themselves as contradiction as an impossible situation, finding it both very dis-
feeling inadequate, empty, lonely, unable, and so on. This state tressing and disorienting.
represents a familiar source of deep and enduring pain and often Eventually, the contradiction is overcome by the creation of new
carries some implicit meaning of “I am defective/bad” or “I am meaning through a dialectical construction that extracts invari-
weak/insecure.” The meaning in this state refers to the self, and the ances across the two positions. Within the context of a supportive
action tendency associated with it is universally one of “closing relationship in which the need is attended to and activated, a new
down” or withdrawing. Part of what makes this “deep and endur- more positive evaluation of the self emerges as a synthesized
ing” is that the affective meaning is formulated by the client as a outcome (i.e., the conclusion that “I am entitled to be loved/
generic autobiographical memory—a collation of repeated occur- valued”). This evaluation regards the self in a new and positive
rences within a specified type of interpersonal situation/context way. By contrast, it also makes an appraisal of the circumstances
(Angus, Lewin, Bouffard, & Rotondi-Trarisan, 2004). As such, as being unfortunate or unsatisfactory. When the client experiences
meaning entailed in this state is much more specific and differen- the self and the need positively, this awareness also entails an
tiated than that in the examples of global distress cited earlier (for implicit acknowledgement of loss or injury (i.e., “I still don’t have
which there is very little autobiographical context). what I need, and I miss what I deserved”). Bridging two categories
For example, in describing his chronic sense of personal inad- of meaning in this way to create a third, higher order meaning
equacy and shame, Client 505 covered his face and said, “I have to (through a dialectical construction) involves two important shifts.
monitor everything I say, even while I’m saying it because I’m . . . First, there is a shift in perspective from the other to the self as the
I know, or feel that everything I say is a bit off . . . People will do judge of self-worth. Second, the new meaning entails a shift from
a double take when I speak and disregard me as a nutcase” (33 min the possible present (i.e., “I need/want/hope to be lovable/
into Session 5). In a statement of fear, Client 516 shuddered and validated”) to the actual present (i.e., “I now experience myself as
said of her mother, with whom she had a pernicious and emotion- being lovable/validated”).
ally abusive relationship, “I just have this sinking feeling, that—
She’ll get me” (52 min into Session 8). In an example of feeling
shame, guilt, and fear, Client 500 talked about being neglected as
Rejecting Anger
a child and stated “but then I still feel it’s my fault. I must deserve Rejecting anger is a generic type of emotion (not yet specific in
this . . . to be punished” (at the end of Session 9). Then she broke its meaning). It is characterized by feelings of protest, repulsion,
into sobbing tears. hate, disgust, and so forth. This is usually expressed with relatively
high arousal (i.e., angry tears, shaking head, fists, etc.) Moreover,
Dialectical Construction 1: “I Am Lovable/Worthy” the action tendency is one of either distancing or sometimes even
destroying. For example, in a prototypic statement of rejecting
The next stage in processing that occurs is a dynamic dialectical anger, Client 505 said of his father, “I’m disgusted, I want to get
construction, represented by the first set of three arrows that are rid of him” (40 min into Session 7); and on a different occasion,
marked by a positive evaluation (see Figure 1). This is produced by “He’s repulsive, I’d like to punch him out” (21 min into Session 3).
first a progressive unfolding or differentiation of the presenting What initially makes this emotion adaptive is that it is an agentic
fear/shame state into both an existential need and a core negative rejection of some noxious experience, as opposed to shrinking
evaluation about the self, and then a dynamic synthesis of these. away or closing down, which as we have seen is characteristic of
The three arrows are used to indicate that the subsequent step is the maladaptive fear and shame.
product of an interactive process (rather than a simple state-to-state Although more differentiated than global distress (in that this
transformation). Thus, when a client begins to move beyond the emotion has a clearer action tendency, meaning, etc.), the state of
experience of shame and fear, a critical moment of constructive rejecting anger is still not very specific or well differentiated in its
abstraction takes place. Constructive abstraction has been de- meaning. A client who expresses rejecting anger essentially pro-
scribed in the cognitive developmental literature as reading or tests the noxiousness of a given situation but does not go on to
collating invariances across different types of situations or con- elaborate the subjective experience of that situation. This, of
tents to create higher order truths (Greenberg & Pascual-Leone, course, is a step in the right direction toward adaptive self-
1995; A. Pascual-Leone & Greenberg, 2006; J. Pascual-Leone, organization and self-protection, but as an emotion it remains
1990). generic and unanchored in personal and idiosyncratic information.
The essence of statements that a client might make at this point In the left-hand path of Figure 1, further differentiating rejecting
based on feeling ashamed and/or scared might be as follows: “On anger can transform the affective-meaning state of a client into a
the one hand, I have the sense that I am a really unlovable/ new type of emotion: assertive anger. In the right-hand path,
unworthy person [negative evaluation]—and on the other hand and clients who seem not to have a strong enough positive self-
 EMOTIONAL PROCESSING IN SESSION 879

evaluation or are unable to access existing emotional resources are led to an unfulfilled need. For this reason assertive anger and
faced with the difficult task of literally constructing a positive self-soothing act as one side of the coin, whereas grief and hurt
self-evaluation from aspects of their present emotional experience, make up the other. Clients often oscillate between these two sets of
which is a more difficult, laborious, and painful process. emotions, and so it is not surprising that assertive anger (or
self-soothing) often transforms into sadness.
Assertive Anger
Grief/Hurt
Assertive anger has been grouped with self-soothing as a single
component in the model in Figure 1. This is because although they The emotions of grief and hurt constitute a single component in
are very distinct states that can be easily differentiated by an the model of Figure 1. Again, this is because they are functionally
outside observer, these two affective-meaning states are function- equivalent in the model, but nevertheless what clients often ex-
ally equivalent for processing emotional distress (just as fear and press is a mixture of the two. Thus, this emotional state is char-
shame are also functionally equivalent). Assertive anger is essen- acterized by either sadness over a loss or recognition of one’s hurt
tially anger that has enough differentiation to embody a positive (i.e., woundedness)— or both. However, the emotional tone must
self-evaluation and the clear assertion of that evaluation or of some be without blaming, self-pity, or resignation (which are character-
personal need. This means that assertive anger is not just about istic of global distress).
pushing away something noxious. More than that, it is about Emotional arousal is present in this state, but it remains facili-
setting boundaries and engaging in a fight for one’s rights and/or tative of the healing process. That is to say, it is enough to keep
existential needs. Some prototypical statements that represent as- pushing the client along but not so much that it is disregulating or
sertive anger could be “I won’t accept this, I have value! I have disorienting. As with assertive anger or self-soothing, the explo-
been mistreated” or “I am different from you. I exist, I deserve.” ration of this state entails great specificity and idiosyncratic detail
Expressions of this state are both very specific and clearly adap- regarding its meaning about the self. In a prototypical example,
tive. Client 509 expressed grief over a serious falling out he had had
As a case in point, after the expression of rejecting anger cited with his siblings and with one sister, in particular: “Tremendous
earlier, Client 505 went on to differentiate the meaning of his anger sadness . . . It means we won’t ever get together again, to have a
in an imaginary dialogue with his father: “It’s not right. I should swim, to have a BBQ, to . . . talk. . .. I’m very sad about losing
have had a father who loved me . . . I didn’t deserve the way you [her]. Oh! . . . She more than anybody” (Session 3, starting at 45
treated me, it wasn’t a way to treat a child” (Session 7). In another min). Client 505 reflected on something missing from his child-
instance of assertive anger Client 516 stated, “I can love, I am hood relationships: “In many ways I never had a family. If it
loveable. I have been mistreated and abused by you. . .. It’s my weren’t for extended family it wouldn’t have been much . . . I’m
right and I’m walking away, I want you to leave [me] alone!” missing that particular type of love . . . the love that only a parent
(Session 8). could give” (45 min into Session 7).
One should note that although assertive anger/self-soothing and
hurt/grief can elicit one another, both of these model components
Self-Soothing
exist and emerge as a result of some positive evaluation. In other
On some other occasions, after experiencing fear and shame and words, they are two parts elaborated from the same ongoing
creating a positive evaluation through the dialectical synthesis of a experience. Assertive anger and self-soothing are states that rep-
negative evaluation and an existential need, some clients do not resent the experience of, or lead to, need fulfillment, whereas hurt
assert their need but rather directly attend to that need. (Thus, some and grief are states that represent the very opposite—the lack of
clients may not express anger over the course of processing their need fulfillment. As before, the conditions of both simultaneous
emotional distress.) Self-soothing or self-nurturing is the affective- emergence and inherent opposition make the components of as-
meaning state that is characterized by fulfilling certain expressed sertive anger/self-soothing versus hurt/grief a dialectical pair (rep-
needs oneself. This serves the same function as assertion, but resented by the second positive evaluation in Figure 1).
instead of directing the emotion outward (i.e., in anger) it ex-
presses the emotion as tenderness or caring turned inward toward Dialectical Construction 2: “I Will Survive and Can
the self. Cope”
This state can appear in several forms, including explicit self-
soothing, attributed self-nurturing (as when the client roleplays a It follows, then, that the final stage toward resolution is a second
caring significant other), or acknowledgment of and reflection on dialectical construction, this time between the experiences of as-
existing resources (such as available social support, past personal sertive anger/self-soothing and the experiences of hurt/grief. On
successes, etc.). Some prototypical examples include when Client one hand, the client has the experience of “I feel wounded, I have
521 spoke to herself reassuringly from the role of her parents, lost the love” (hurt/grief), whereas on the other hand the client is
saying, “I love you. It’s going to be alright” (26 min into Session at the same time aware that “I am strong and deserve to be
10); or when Client 076 spoke to herself and considered her loved loved/validated or I am currently being loved/validated” (assertive
ones: “You deserve to be treated well. You have your husband who anger/self-soothing). The dialectical synthesis of these two states
loves you, your sister . . .” (40 min into Session 7). creates the new experience of acceptance and agency: “I will
As we have seen, a positive evaluation embodied by assertive survive and can cope. I accept the past wounds/losses. I can let go
anger or self-soothing highlights the client’s healthy need. How- and move on with the rest of my life.” This is a second positive
ever, the counterpoint of this is the unfortunate circumstances that evaluation of the self as confident and future oriented. If in the first
880 PASCUAL-LEONE AND GREENBERG

positive evaluation (discussed earlier) one of the shifts was from Sequential Processing Hypothesis (Hypothesis 2)
the possible to the actual present, then at this point there is a shift
from the present to the future. At this point the client is in a This research hypothesis addresses the following question: Are
position for acceptance and agency of the distressing issues, which components of the model sequentially ordered? In other words, are
meets all of the criteria of performances characterized as advanced the model’s components really structured in a stepwise and se-
emotional processing (see the bottom of Figure 1). quentially ordered fashion, or is the ordering of the model’s
components more reflective of chance? This hypothesis states that
the first emergence of either hurt/grief or assertive anger or self-
Successful Emotional Processing soothing (i.e., a Level 4 state) is always preceded by either nega-
tive evaluation or need (i.e., a Level 3 state), which in turn is
Acceptance and agency is not an end state but a momentary always preceded by either fear/shame or rejecting anger (i.e., a
crystallization of “skillful” performance in emotional processing, Level 2 state), which in turn is always preceded by global distress
aspects of which are also present in some other states. To resolve (Level 1).
this type of open-ended problem (i.e., emotional processing), the
client must maintain engagement among affective-meaning states
Method
near the bottom of the model. Every successfully performing client
who begins in global distress is likely to go through each of the
Participants
different states described previously, to reach other states that meet
performance criteria for advanced emotional processing. In this Sample
sense, all of the states are “components” of a model for emotional
processing. Of course, a given client will not usually accomplish The sample in this study of single sessions comprised 34 clients.
emotional processing in a single session, or in one fell swoop, but This sample came from several larger participant pools originally
rather in a “two steps forward, one step backward” fashion. This recruited for four clinical trials completed at a large urban univer-
saw-tooth pattern of improvement has been observed both in single sity psychotherapy research clinic between 1991 and 2002. All
sessions and across sessions. clients had given informed consent to have their sessions studied
for psychotherapy research.

STUDY 2: VALIDATION PHASE Sample Demographics

Hypotheses About the Structure of the Model The research sample of 34 clients included 28 women and 6 men
between the ages of 27 and 59 (M ⫽ 42, SD ⫽ 10.4). In all, 9
The validation phase is the second part of a task/process anal-
clients were single, 20 were married or common law, and 5 were
ysis. In this phase the model already developed in the discovery
separated or divorced. The education level of the selected sample
phase (Study 1) was reexamined using empirical criteria to test a
included 12 graduates of college or university, 3 clients with some
larger sample of cases. This phase of research established predic-
postgraduate education, and 9 clients who had postgraduate or
tive and external validity of the synthesized model using single
professional school educations. Information on ethnicity was not
sessions that were not part of the original model building. The
systematically collected in each of the original participant pools,
microprocess research in this study did not attempt to predict final
but the diversity of client ethnicities in the current sample is
treatment outcome. Rather, the purpose of this study was to predict
believed to be representative of a large multicultural urban area.
good in-session process. Most would agree that good in-session
According to pretreatment assessments, the mean Global As-
events are incrementally important to producing a good treatment
sessment of Functioning score for the sample was 69.2 (SD ⫽
outcome. The contribution of this research, then, is a prior step: to
10.7, range ⫽ 53–90). This mean score indicates mild to moderate
determine what emotional processes predict good events in ther-
symptoms of impairments in psychological, social, and occupa-
apy. The assumption, of course, is that good in-session events are
tional functioning. In the sample, 20 clients (58.9%) suffered
the building blocks of good treatment and, as such, are highly
primarily from mood disorders; another 6 clients (17.6%) suffered
relevant to clinical work.
primarily from anxiety disorders. Five clients (14.7%) out of the
total sample suffered from a secondary, comorbid Axis I disorder.
Advanced Processing Hypothesis (Hypothesis 1) There were 8 clients (23.5%) who did not receive any Axis I
diagnosis at all, based on the Diagnostic and Statistical Manual of
The first test of predictive validity addresses the following Mental Disorders (3rd ed., rev.; American Psychiatric Association,
question: Can good within-session effects (i.e., events that end in 1987). Based on the Structured Clinical Interview for DSM–IV,
reduced distress and increased meaning making) be discriminated assessment for personality disorders indicated that 13 clients
from poor within-session effects (such as events in which clients (38.2%) in this sample had Axis II disorders (American Psychiatric
do not make new meaning) on the basis of model components? Association, 1994).
This hypothesis tests whether cases with positive within-session
effects enter advanced model states more often than cases with Therapists and Final Treatment Outcome
poor within-session outcomes. In other words, it is predicted that
clients who show good within-session effects are more likely to The relevance of providing final treatment outcome at this point,
enter the states of hurt/grief, assertive anger, and self-soothing than even if it is not pertinent to the design of this study, is to provide
those clients with poor within-session outcome. context. When describing a sample of selected sessions, the suc-
 EMOTIONAL PROCESSING IN SESSION 881

cessfulness of the treatment from which cases were taken is as and although they are theoretically and empirically regarded as
relevant as the type of treatment or sample demographics. The productive, such microprocesses do not necessarily translate im-
psychotherapy in each of the four clinical trials differed slightly mediately into postsession mood or functioning.
according to the nature of the original study. The therapeutic
modalities included 32 clients in EFT (Greenberg, Rice, & Elliott,
Procedure
1993) and another 2 clients in client-centered therapy (Rogers,
1957), both treatments that embody highly humanistic and expe- Event Selection
riential therapies. In the sample of 34 cases, treatment was con-
ducted by 19 therapists (17 women and 2 men). Of these, 17 The first step in the procedure of data collection was selecting
therapists were advanced doctoral students of clinical psychology events relevant to the topic of inquiry. The unit of analysis in this
and 2 were registered clinical psychologists. All therapists had a study was a within-session emotional event of a naturalistic length
minimum of 3 years of clinical experience. Twenty clients in the that varied from client to client. This unit was not necessarily the
sample had enjoyed good final treatment outcomes, and 14 had had whole session and was never more than one session. The model
poor final treatment outcomes (details on final outcome measures describes the affective-meaning state of global distress as the
can be found in the articles cited earlier as sources for the sample). initial marker of the event. Accordingly, the criteria defining
global distress served as identifying markers for the events to be
studied. (See the coding criteria for global distress in the CAMS
Process Measure measure in A. Pascual-Leone & Greenberg, 2005.) The selection
The Classification of Affective-Meaning States (CAMS; A. of global distress events was done in a series of convergent
Pascual-Leone & Greenberg, 2005) is an instrument for the sys- appraisals through the combined efforts of therapists, two inde-
tematic observation of emotion and affective-meaning states and pendent raters, and the principal investigator. A total of 102
was designed in the discovery-oriented phase of the task analysis. session videos from approximately 60 cases were initially re-
It measures discrete and specific emotional states that are unique to viewed for the presence of the marker. Markers were found in 34
the model and that are central to the hypotheses of this study. cases, and only one session was taken per client.
Reliability coefficients for different facets of this measure are high
and are reported in detail for the first time here (see the Results Emotion Process Ratings
section of Study 2).
Ratings began with a confirmation of the preselected global
distress marker, which served as the rater’s first code. Using the
Measure for Within-Session Event Effects
CAMS, the rater then coded emotion event continuously for the
Client Experiencing Scale. The Experiencing Scale (Klein, presence of any one of the given possible emotion codes. Raters
Mathieu-Coughlan, & Kiesler, 1986) is a 7-point scale that mea- were instructed to make codes as they saw them unfold sequen-
sures the degree to which clients orient to and symbolize their tially and indicated the time at which they initiated a change in
internal experience and use this felt experience as information to code. This procedure is well documented in ethology and is tech-
resolve their problems. It makes no reference to discrete emotions nically referred to as continuous cross-classification (Bakeman &
or affective states. The 7-point scale was reduced to a binary scale Gottman, 1986; Martin & Bateson, 1986).
of low (Levels 1– 4) versus high (Levels 5–7). Higher experiencing
ratings at or near the end of a given emotion event were used as Identifying the Effects of In-Session Events
indicators of good event outcome with respect to the immediately
preceding client processes. The Experiencing Scale is widely con- Raters using the Experiencing Scale were operationally inde-
sidered to be the gold standard of good experiential process and pendent from raters using the CAMS. High experiencing in the last
remains one of the most extensively studied and validated mea- 15% of each event was used as an indicator of good within-session
sures of productive in-session process in psychotherapy research. outcome. An additional, confirmatory criterion for discriminating
High ratings on this measure have been shown to be predictive of between good versus poor event outcome was the clinical judg-
good treatment outcome across most major schools of psychother- ment of an outside expert. The expert appraisal of event outcomes
apy, including client-centered, cognitive behavioral, psychody- confirmed the identification of each good event outcome case. In
namic, and emotion-focused psychotherapy (see Greenberg & only 1 case out of 34 clinical judgments was there disagreement
Pascual-Leone, 2006). Interrater reliability coefficients for the between the judge and the Experiencing Scale ratings in discrim-
measure have been reported to range from .76 to .91 (Klein et al., inating between in-session event outcomes. In that case, the dis-
1986). tress event had high experiencing in the last 15%, but the expert
Expert clinical judges. The second measure for determining clinical judgment evaluated the event as having a poor within-
good versus poor effects of within-session events was the binary session event outcome. However, the high experiencing that oc-
evaluation of expert clinical judges, which provided clinical va- curred was very brief in that event, and a second expert clinical
lidity from a separate set of criteria. Note that client self-report judge independently confirmed the appraisal that it was indeed a
measures of mood or functioning completed outside the session poor-outcome event case.
were not appropriate for this study because good versus poor Because it so happened that if and when high experiencing
effects were with respect to moment-by-moment events that were occurred it only occurred in the last 15% of the event, sorting
nested within a given session. Moreover, positive in-session ef- criteria did not produce any intermediate event outcome cases. In
fects are similar in kind to an insight or productive cognitive shift, this way, the procedures resulted in the full sample (N ⫽ 34) being
882 PASCUAL-LEONE AND GREENBERG

subdivided post hoc into two equal groups. Thus, 17 cases were distribution. The likelihood of the observed data having occurred
identified as good and 17 as poor within-session effects. by chance was subsequently measured using an empirical p value.
In this way, it was possible to calculate the probability of the
observed findings having occurred by chance (out of 100,000
Analyses
simulations) without making assumptions of normality.
Advanced processing hypothesis (Hypothesis 1). A total of 17
cases with good within-session effects were compared with 17 Results
poor within-session effects on the number of times each group
initiated different affective-meaning states in the model. This was Reliability
done for all model components except the global distress marker,
with which all cases began on account of the event selection. A Reliability of CAMS Process Measure
7 ⫻ 2 multivariate analysis of variance used seven predictor Of the total sample of 34 cases used in this study, 27 cases
variables to test for differences on a single outcome variable. Thus, (79.4%) were randomly selected and independently rerated in their
one rating for each predictor variable per selected event was used entirety by a second rater for the purpose of establishing reliability.
to predict event outcomes. Using this analysis allowed for both an With respect to the number of actual codes made during data
omnibus measure of the model as well as F tests for each model collection, 352 out of 395 individual codes (89.1%) were rerated.
component. Ratings produced by continuous cross-classification require differ-
The seven predictor variables were dichotomous; each indicated ent forms of reliability depending on how data are being used.
the presence or absence of either (a) fear/shame, (b) rejecting First, the continuous nature of ratings required that independent
anger, (c) negative evaluation, (d) need, (e) hurt/grief, (f) self- rater agreement on what constitutes a discrete change in state be
soothing, or (g) assertive anger. A client entering a given affective- demonstrated; this is referred to as the unitization of observations
meaning state at least once in the duration of the sampled emotion and reached an agreement of 85.9%. Second, the classification of
event determined the “presence” of a state. If a client never entered 11 different codes in the measure required that the sequential
a given state, then that affective-meaning state was considered ordering of those classes also be reliable in the ratings (which they
“absent” from the emotion event. The event outcome variable was were; Cohen’s ␬ ⫽ .91). Third, the continuous nature of ratings
also dichotomous: good versus poor within-session outcome. As required that an agreement be demonstrated between independent
indicated, good and poor outcomes of selected in-session events raters on the duration of any given unit of coding, and this was also
were determined by high and low ratings on the Experiencing shown to be high (r ⫽ .76). According to Fleiss (1981), levels of
Scale at the end of the distress event (last 15%) and by expert agreement above .75 can be considered excellent agreement above
clinical judgments. chance. Thus, the measure demonstrated high overall reliability.
Sequential processing hypothesis (Hypothesis 2). Data were
recoded into a 4-point scale: (1) global distress, (2) fear/shame or
Reliability of Experiencing Scale Ratings (Measure of
rejecting anger, (3) negative evaluation or need, (4) assertive
anger or self-soothing or hurt/grief. Following this procedure, if a
Within-Session Effects)
given affective-meaning state in the data (a) emerged for the first One half of the data set (17 out of 34 cases) was randomly
time and (b) was preceded by affective-meaning state(s) from each chosen to be coded by two independent raters to establish the
and every one of the lower level(s) as specified by the 4-point reliability of high versus low experiencing ratings (i.e., on a binary
scale, then it was considered to be in sequential concordance with scale). The interrater reliability was high (Cohen’s ␬ ⫽ .88), which
the model. If, in contrast, a state (a) emerged for the first time but was consistent with the measures of reliability from psychotherapy
(b) was not preceded by hypothesized state(s) from each of the cases reported by Klein et al. (1986).
previous level(s) in the prescribed order, then this was considered
to be a disagreement and not in concordance with the model. If Description of Groups: Good Versus Poor In-Session
state(s) from a particular affective-meaning level did not occur at
Effects
all, it was considered neither a concordance nor discordance with
the model. Clients with good within-session effects had longer emotional
Given that according to the research hypothesis the base rates events (M ⫽ 32.0 min, SD ⫽ 12.8) as compared to clients with
for each level of component are predicted to be different (meaning poor event effects (M ⫽ 21.5 min, SD ⫽ 12.7), and that difference
not all combinations have equal probability), one cannot assume was significant, t(32) ⫽ 2.40, p ⬍ .025. There was no evidence
there is a normal distribution. In essence, there is no simple that shorter and less effective events were truncated by running
traditional inference method for this data set. For this reason the into the end of the therapy hour. There was no difference between
significance of observed findings compared to chance was deter- groups (n ⫽ 17 vs. 17) on demographic variables of age, marital
mined using a bootstrapping technique (for an overview of this, see status, or education (using t test and chi-square analyses; all ps ⬎
Hesterberg, Moore, Monaghan, Clipson, & Epstien, 2006). .45). This indicated that the two subgroups were representative of
To do this, we randomized the sequential ordering of the four a homogeneous client population with respect to basic demograph-
levels of emotion states within each of the 34 cases, thereby ics.
producing a new “simulated” data set. Next, the number of times The final treatment outcome and therapy from which events
the hypothesis was fulfilled by chance was counted. This resam- were selected served as a descriptive variable for the research
pling procedure was repeated with replacement 100,000 times, sample. Although raters were blind to the final treatment outcomes
thereby producing a probability curve known as a bootstrapping of the 34 cases, 13 out of 17 with good within-session effects, as
 EMOTIONAL PROCESSING IN SESSION 883

contrasted with 4 cases out of 17 with poor within-session effects, anger, as opposed to 0 out of 17 clients with poor event outcomes
went on to eventually end in good overall treatment outcomes. ( p ⬍ .001). Overall, the aforementioned findings supported the
This was a significant positive relationship between in-session advanced processing hypothesis (Hypothesis 1).
effects (i.e., task resolution measured by high experiencing) and
good overall treatment outcome, ␹2(1) ⫽ 4.37, p ⬎ .05. Thus,
outcome of single within-session events had some predictive Sequential Processing Hypothesis (Hypothesis 2) Findings
power over treatment outcomes (explaining nearly 13% of the final
Affective-meaning states coded by the CAMS were grouped
outcome variance, ␾2 ⫽ 0.129). Note that the predictive power of
into four levels in the model and then examined for their emer-
these selected sessions, in the context of 12 to 14 session treat-
gence in the correct or incorrect sequence according to the stated
ments, was not the focus of this study. However, it does provide
hypothesis. This was done for the distress event experienced by
descriptive information about the sample of sessions and their
each of the 34 clients. If the first instance of a newly emerging
relative importance to the clinical process.
state was preceded by a prerequisite state from each of the earlier
levels as indicated by the hypothesis, this was counted as a
Advanced Processing Hypothesis (Hypothesis 1) Findings concordance. Table 2 indicates the number of clients who experi-
Presence of specific affective-meaning states coded by the enced a given state in the hypothesized sequence over the number
CAMS was compared across both groups using a 7 ⫻ 2 multivar- of clients who experienced that state (irrespective of sequence).
iate analysis of variance (see Table 1). Clients with good as Although the preceding analyses showed that advanced states
opposed to poor within-session effects used the model components occurred more often in the experiences of clients with good event
to significantly different degrees (Wilks’s ⌳ ⫽ .325), F(7, 26) ⫽ outcomes, if and when they occurred in either event outcome, they
7.723, p ⬍ .001. Seven univariate F statistics (one for each model generally occurred with the correct prerequisites (and this at a
component) determined the nature of group differences (see comparable rate for both good and poor in-session effects).
Table 1). Note that all clients experienced Level 1, global distress, before
The emotional performance of all clients was statistically equiv- any other state entirely as an artifact of the process by which
alent with respect to fear/shame ( p ⬎ .70). Contrary to the hy- events were selected (see column marked Level 1). Because that
pothesis, rejecting anger occurred more often in good as compared particular model component was the “marker,” it was excluded
to poor event outcomes ( p ⬍ .05). Clients who had good within- from any calculations of sequential concordance. Although the
session event outcomes expressed an existential need significantly second level in the model (fear/shame and/or rejecting anger) did
more often than clients who had poor in-session events (good ⫽ not necessarily occur for every client, when it did occur it was
14/17 vs. poor ⫽ 6/17; p ⬍ .005). There was no significant always preceded by the previous level in the hypothesized
difference in negative evaluations between groups (good ⫽ 10/17 sequence (i.e., global distress). If it occurred, the concordance
vs. poor ⫽ 9/17; p ⬎ .70). As predicted by the hypothesis, clients of a Level 2 occurrence in the correct sequence was 100%
with good within-session event outcomes more frequently initiated because there were no degrees of freedom (34/34 cases started
the three advanced model components of hurt/grief, self-soothing, with Level 1).
and assertive anger (see Table 1). Clients with good event out- Table 2 only shows the number of clients who did experience a
comes were almost 5 times more likely to engage the experience of given state. However, Level 2 was not experienced at all by 4 of
hurt/grief ( p ⬍ .001) and 6 times more likely to self-soothe ( p ⬍ 34 cases (all of which eventually resulted in poor within-session
.001) than those clients who had poor within-session event out- effects). Level 3 was not experienced by 8 of 34 cases (2 resulted
comes. In addition, 9 out of 17 clients with good within-session in good within-session effects, 6 in poor effects). Level 4 was not
event outcomes entered the affective-meaning state of assertive experienced by 14 of 34 cases (all of which were poor event

Table 1
Presence of Model States as Summed Across Clients and Their Between-Event Differences

7 ⫻ 2 MANOVA:
Univariate
In-session effects Fs (df ⫽ 1)

State Good (n ⫽ 17) Poor (n ⫽ 17) F p Partial ␩2

Fear/shame 13 12 0.14 .708 0.004

Rejecting anger 12 5 6.53 .016* 0.170
Negative evaluation 10 9 0.11 .738 0.004
Need 14 6 9.48 .004** 0.229
Hurt/grief 14 3 23.05 .000*** 0.419
Self-soothing 12 2 17.78 .000*** 0.357
Assertive anger 9 0 18.00 .000*** 0.360

Note. Global distress was the event marker and as such was present at the beginning of all 34 events.
MANOVA ⫽ multivariate analysis of variance.
*
p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.
884 PASCUAL-LEONE AND GREENBERG

Table 2
Rates of Sequential Concordance

Level 1 Level 2 Level 3 Level 4

Client in-session Global Rejecting Negative Assertive

outcome distress Fear/shame anger evaluation Need Hurt/grief Self-soothing anger

Good (n ⫽ 17) 17/17 13/13 12/12 5/10 12/14 8/14 10/12 7/9
Poor (n ⫽ 17) 17/17 12/12 4/4 7/9 6/7 1/2 1/2 0/0
All (N ⫽ 34) 34/34 25/25 16/16 12/19 18/21 9/16 11/14 7/9

Note. The denominator indicates the number of clients who experienced a given state; the numerator indicates the number of clients who experienced that
state in the predicted order. Some clients experienced more than one state at the same level (i.e., fear/shame and rejecting anger, both of which are at Level
2). This table does not contain information on the number of clients who did not experience a given level at all (i.e., neither fear/shame nor rejecting anger
for Level 2). That information is in the text.

outcomes). Thus, model components in Levels 2, 3, and 4 did not ing: “What type of anger?” or “What type of sadness?” Several
necessarily occur. authors have called for a more elaborate differentiation of anger
The sequential processing hypothesis was tested using nested types and anger-related problems (see Linden et al., 2003). This
arguments. In this very stringent test of concordance, a Level 4 study identified observable, reliable criteria for rejecting anger and
affective-meaning state was considered to have occurred in the assertive anger and described their respective roles in the healing
correct sequence if and only if it was preceded by a Level 3 state, process of therapy. Differentiating these two types of anger is one
and that Level 3 state was in turn preceded by a Level 2 state. important contribution of this work.
Thus, this approach to testing the hypothesis required that three The model also discriminates between three types of sadness:
qualitatively different kinds of emotion represented in the model one global and undifferentiated, one specific and maladaptive (i.e.,
emerge in a specified series, like links in a chain. Based on blended with fear/shame), and one specific and adaptive (i.e.,
100,000 randomized bootstrapping simulations, the findings in all authentic grief/hurt). Work by Greenberg (2002; Greenberg &
34 cases (irrespective of within-session outcome) were compared Paivio, 1997) has made the powerful distinction between primary
against chance. The hypothesized pattern was found to occur in 9 versus secondary sadness in the therapeutic process. Secondary
out of 18 cases (50%). Based on simulations of the sample, the sadness, which is an emotional reaction to some more primary and
randomized probability of this particular pattern in emotional underlying feeling, would be included here as global distress. The
experience taking place 50% of the time was shown to be less than model suggests that global distress, on account of its ill-defined
1 in every 250 samples, thereby indicating the statistical signifi- quality, is a prevalent form of maladaptive sadness. Moreover,
cance of this finding (empirical p ⬍ .004). This finding supported when global distress as a form of sadness is differentiated in terms
the sequential processing hypothesis (Hypothesis 2). of feelings and meanings, it consistently leads to emotional states
that are at least tinged with either fear (i.e., as in fear–sadness, such
DISCUSSION as feeling lonely) or with shame (i.e., as in shame–sadness, such as
feeling worthless or empty). In short, the core facet of a maladap-
Prior to this investigation, the most thorough emotion-based tive emotional experience seems to be fear and/or shame based
model of how clients’ experiences of bad feelings actually change rather than being the simple experience of sadness and loss.
in therapy was that of Greenberg and Paivio (1997). Consequently,
many of the steps in that model served as null hypotheses for the
discovery phase regarding components for processing emotional Emotional Experience as a Mechanism of Change
distress. The new model both confirms this model and develops an
original model of how bad feelings are restructured through emo- The advanced processing hypothesis (Hypothesis 1) examined
tional processing. Overall, progression through the model de- the model as a whole in terms of its relevance for the outcome of
scribes a differentiation of emotion states from global to more in-session events. A good in-session effect was defined as high
specific feelings/meanings (see vertical axis in Figure 1). This experiencing at the end of a selected emotion event and was
supports work by Stern (1997), who argued within contemporary verified by expert clinical judges as being “therapeutically produc-
psychoanalytic theory that working through distress begins by tive.” When examining the probability of entering any given
helping clients interpret raw, unformulated, “global” experiences. emotional state, it was especially clear that although clients who
enjoyed good in-session effects were having longer emotional
experiences, they were also having qualitatively different emo-
Not All Emotions Are Created Equal
tional experiences than clients with poor event outcomes. In gen-
Discerning the key affective-meaning qualities of emotions has eral, good in-session effects were preceded by a more diverse array
demonstrated that not all forms of “sadness,” or not all forms of of qualitatively distinct emotional processes.
“anger,” are equally useful emotional experiences. A fine grain of As predicted by the model, most early model components were
refinement in the identification of emotion qualities, tone, and shown to be present in both good and poor within-session events.
context is important when working with emotions, on both clinical For the most part this means that clients experience both global
and research frontiers. The question to be answered is the follow- distress and fear/shame regardless of the outcome of the event.
 EMOTIONAL PROCESSING IN SESSION 885

Rejecting anger, in contrast, was unexpectedly significantly more (either negative evaluation or need), which in turn was preceded by
related to good event outcomes, albeit with the smallest effect size. an early state (either fear/shame or rejecting anger), which in turn
However the case may be, sequential findings were significant in was preceded by global distress. Thus, when the emotional expe-
the final analyses, lending support to the description of rejecting rience of clients included all four levels in the model, that expe-
anger as an early model component. rience unfolded in a step-by-step sequence that was predicted by
Intermediate model components that represented the level of the model and occurred significantly more often than chance. This
personal evaluation and reevaluation proved to be a point of sequential processing hypothesis was supported by the sample as a
critical distinction between good and poor in-session effects. State- whole, irrespective of good or poor in-session outcome.
ments of negative evaluation about the self were shown to be Overall, these findings suggest that when certain prerequisite
present in virtually the same number of cases across event out- affective-meaning states occur it increases the likelihood that emo-
come. Thus, painful and heartfelt statements about feeling worth- tional processing of global distress will reach the final stage. One
less, frail, or unlovable could not be predicted by in-session explanation why these canonical moment-by-moment patterns
outcome—just as the experience of fear/shame could not distin- may exist is that emotional experiences develop in predictable
guish between event outcomes. However, a heartfelt statement patterns due to the psychological proximity between affective-
expressing an existential need to feel valuable, lovable, safe, or meaning states involved. These findings support the hypothesis
alive was predictive of good in-session event outcomes (i.e., high that model components are organized in a sequential structure that
experiencing levels). is empirically grounded.
This finding is consistent with Greenberg and Paivio’s (1997) In recent works in EFT, Greenberg (2002) introduced the con-
model and supports the view held in EFT that the expression of a ceptualization of two- and three-step emotional change processes.
need—a wish for attachment, personal agency, or survival—is the Clients undergoing a three-step change process, according to
“gateway” to deeper emotional experiencing (Greenberg, 2002; Greenberg (2002), find themselves confronted with more deeply
Greenberg & Paivio, 1997; Greenberg et al., 1993). Of course, rooted emotional obstacles than simply secondary/defensive emo-
even though the expression of a need may herald more adaptive tion. The three-step process is perhaps the single most important
emotion, it is not synonymous with adaptive emotion and does not contribution of emotion-focused interventions and theory to un-
ensure deeper levels of experiencing. This is indicated by the fact derstanding emotional processing. In this series of emotional
that approximately one third (6/17) of the clients who had poor shifts, successful clients move in their moment-by-moment expe-
in-session events expressed a core need. The key differentiating riences from (a) a secondary emotion to (b) a primary maladaptive
factor between event outcome groups is likely to be whether the emotion and eventually to (c) a primary adaptive emotion that is
expression of a need is embodied and followed by the emotional activated through the articulation of an existential need.
impetus of advanced model components (as discussed below). As discussed in the results of Study 1, global distress represents
Even so, this finding offers an empirically supported clarification a secondary emotion; fear/shame represents primary maladaptive
for CBT: Identification of a core negative belief (i.e., a negative emotions; and assertive anger, self-soothing, and hurt/grief each
evaluation) is indicative of therapeutic progress (Beck, 1995), but represent primary adaptive emotions. Thus, the three-step process
more so in experiential therapy if it is followed by the explicit of change conceptualized by Greenberg (2002), that is, how emo-
articulation of an existential need. tion changes emotion, can be described concretely using affective-
The overwhelming majority of clients with good within-session meaning states from the current model. Note that “emotion” is
events were characterized by the presence of advanced model understood here as the complex synthesis of both affective feelings
components. Thus, engaging in the affective-meaning experiences and the meaning embodied by those experiences (hence “affective-
of hurt/grief, self-soothing, and/or assertive anger was predictive meaning” states). More important, for the first time, the three-step
of good in-session outcome. These findings support the advanced model of emotional transformation in EFT is empirically sup-
processing hypothesis (Hypothesis 1) and validate both the models ported by these findings as a general change process. The ordering
of Greenberg and Paivio (1997) and the current model that was of states and dialectical syntheses (i.e., transformations) in this
derived through task/process analysis. The model is empirically empirically supported model also bear some resemblance to the
related to good in-session effects, and clients who have good clinical and theoretical observations made in accelerated experi-
versus poor effects experience components of the model in sys- ential dynamic psychotherapy, in which core affects are accessed
tematically distinct ways. and transformed (Fosha, 2000).
The sequential processing hypothesis (Hypothesis 2) findings
Sequences of Productive Emotion are critical in demonstrating that this model of emotional process-
ing is actually a process model rather than a model of “attainment”
This part of the study begins to address the call made by a or “missing ingredient.” Clients with poor in-session effects are
number of psychotherapy researchers (i.e., Hayes et al., 2007; missing the experience of the advanced model components, and,
Jones et al., 1992) for the description of growth curves and the for clients with good event effects, those components are attained
analysis of sequential patterns in client change. Findings supported by following a particular sequential process. This study speaks to
the sequential processing hypothesis (Hypothesis 2), validating productive client processes without addressing its causal determi-
that the first time a given model component occurred it followed at nants. Whether sequential patterns are primarily explained by
least one component from each of the preceding model levels spontaneous self-organizing processes (within a described thera-
(irrespective of lag or other interjected states). In concrete terms, peutic context) or by participation in some specific task is a
the first emergence of advanced states (either hurt/grief or asser- separate question that is beyond the scope of this inquiry. There-
tive anger or self-soothing) was preceded by an intermediate state fore, one could also interpret the current findings as characteristic
886 PASCUAL-LEONE AND GREENBERG

of intervention approaches generally employed by experiential distress. First, therapists familiar with this model may recognize
therapists, which might imply that more directive approaches, such their client’s unfolding states of feeling and meaning by using the
as CBT, could generate different sequences. criteria of model components described here. Second, familiarity
with the sequences of productive emotional processing will help
Productive Emotion or Unproductive Emotion? therapists identify those facets of complex client experience that
are viable alternative states of self-organization and that, therefore,
As the model shows, some emotions, such as hurt/grief or would be most helpful to explore. Thus, this model could be used
assertive anger, will be painful and unpleasant but will promote within different approaches to inform which type of interventions
healing. This makes it impossible to separate positive and negative will be most fruitful according to the client’s presenting state and
emotions as if they were healthy and unhealthy, respectively. proximal zone of emotional development at any given time.
Assuming that unpleasant or negative emotion is unhealthy over-
looks the primary function of emotions as being adaptive (Frijda,
1986; Greenberg & Safran, 1987; LeDoux, 1996). In addition, one Limitations
must also be careful not to consider more superficial states of
The client population in this study was restricted to those who
affective meaning, such as rejecting anger, as categorically dys-
suffered primarily from depression and/or long-standing interper-
functional in therapy. Rejecting anger is often a useful approxi-
sonal grievances and who had been screened for suitability for
mation toward assertive anger, and it represents a step in the right
brief treatment. Moreover, all clients were treated with experiential
direction toward self-assertion and eventual resolution. As the
therapies, and the investigation was done within an a priori frame-
process model suggests, none of the affective-meaning states are
work of emotion research. Therefore, these findings are not gen-
inherently pathogenic. However, getting stuck or persistently “re-
eralizable to the population at large or to other therapeutic con-
peating” any one of these components will cause emotional dis-
texts. It will be important for future research to show that the same
order, whether that component is global distress, fear/shame, re-
sequences can be observed in globally distressed clients receiving
jecting anger, hurt/grief, or even a sense of need. In this way, the
psychodynamic or cognitive behavioral therapies. Even so, the
client’s moment-by-moment experiential history is what most
speculation that these findings represent aspects of emotional
clearly tells which model components are main sources of diffi-
development is provocative and suggests that the model might
culty and which others will usher in new experiential meaning for
reflect characteristics of how humans process emotional distress.
the client.
Some emotional experiences propel the client on a healthy
self-organizing trajectory that reaches its completion as a mean- References
ingful, emotionally differentiated, and integrative experience.
However, before such an experience is carried forward it may American Psychiatric Association. (1987). Diagnostic and statistical man-
begin as a sense of intense, poorly regulated, and ill-defined global ual of mental disorders (3rd ed., rev.). Washington, DC: Author.
malaise. Thus, potentially injurious affective-meaning states such American Psychiatric Association. (1994). Diagnostic and statistical man-
ual of mental disorders (4th ed.). Washington, DC: Author.
as global distress, fear/shame, or negative evaluations could be
Angus, L., Lewin, J., Bouffard, B., & Rotondi-Trarisan, D. (2004). What’s
painful and stagnant or could be rendered into important affective the story? Working with narrative in experiential psychotherapy. In L.
meaning to be developed forward into an emotional experience Angus & J. McLeod (Eds.), The handbook of narrative and psychother-
that eventually results in healing. In the sequential order of change, apy: Practice, theory and research (pp. 87–102). Thousand Oaks, CA:
working with painful emotion is accepted as contributing to emo- Sage.
tional processing and resolution. Bakeman, R., & Gottman, J. M. (1986). Observing interactions: An intro-
duction to sequential analysis. New York: Cambridge University Press.
Beck, J. S. (1995). Cognitive therapy: Basics and beyond. New York:
A Guide for Therapists Guilford Press.
Borkovec, T. D., Alcaine, O., & Behar, E. (2004). Avoidance theory of
In conclusion, these findings show that certain emotional pat-
worry and generalized anxiety disorder. In R. G. Heimberg, C. L. Turk,
terns predicted well processed emotion during in-session events.
& D. S. Mennin (Eds.), Generalized anxiety disorder: Advances in
Moreover, the nature of these events was positively related to research and practice (pp. 77–108). New York: Guilford Press.
treatment outcome. The implication is that good in-session events Elliott, R., Watson, J., Goldman, R. N., & Greenberg, L. S. (2004).
(measured by high experiencing) are eventually important and Learning emotion focused therapy: A process experiential approach to
relevant to good overall outcomes in therapy, something that is change. Washington, DC: American Psychological Association.
well supported by the literature on the relationship between client Fleiss, J. L. (1981). Statistical rates for measures and proportions (2nd
experiencing and outcome (Greenberg & Pascual-Leone, 2006). ed.). New York: Wiley.
Thus, the major contribution of this study is to show that a Foa, E. B., & Kozak, M. (1998). Clinical applications of bioinformational
processing sequence of emotions posited by the theory of a spe- theory: Understanding anxiety and its treatment. Behavior Therapy, 29,
cific approach (EFT) can empirically predict a peak in productive 675– 690.
Fosha, D. (2000). The transforming power of affect: A model of accelerated
emotional processing according to the unrelated measure depth of
change. New York: Basic Books.
experiencing, which itself has repeatedly been demonstrated to Frijda, N. H. (1986). The emotions. Cambridge, England: Cambridge
predict good treatment outcomes across types of treatment. University Press.
The findings of this project are formulated in terms of clients’ Greenberg, L. S. (1999). Ideal psychotherapy research: A study of signif-
concrete emotional experiences (i.e., states and transitions), and, as icant change processes. Journal of Clinical Psychology, 55, 1467–1480.
such, they offer a model for guiding therapeutic work with global Greenberg, L. S. (2002). Emotion-focused therapy: Coaching clients to
 EMOTIONAL PROCESSING IN SESSION 887

work through their feelings. Washington, DC: American Psychological LeDoux, J. (1996). The emotional brain: The mysterious underpinnings of
Association. emotional life. New York: Simon & Schuster.
Greenberg, L. S. (2007). A guide to conducting a task analysis of psycho- Linden, W., Hogan, B. E., Rutledge, T., Chawla, A., Lenz, J. W., & Leung,
therapeutic change. Psychotherapy Research, 17, 15–30. D. (2003). There is more to anger coping than “in” or “out.” Emotion, 3,
Greenberg, L. S., & Malcolm, W. (2002). Resolving unfinished business: 12–29.
Relating process to outcome. Journal of Consulting and Clinical Psy- Martin, P., & Bateson, P. (1986). Measuring behaviour: An introductory
chology, 70, 406 – 416. guide. New York: Cambridge University Press.
Greenberg, L. S., & Paivio, S. C. (1997). Working with emotions in McCullough, L., Kuhn, N., Andrews, S., Kaplan, A., Wolf, J., & Hurley,
psychotherapy. New York: Guilford Press. C. L. (2003). Treating affect phobia: A manual for short term dynamic
Greenberg, L. S., & Pascual-Leone, A. (2006). Emotion in psychotherapy: psychotherapy. New York: Guilford Press.
A practice friendly research review. Journal of Clinical Psychology, 62, Orlinsky, D. E., Grawe, K., & Parks, B. K. (1994). Process and outcome in
611– 630. psychotherapy—noch einmal. In A. E. Bergin & S. L. Garfield (Eds.),
Greenberg, L. S., & Pascual-Leone, J. (1995). A dialectical constructivist Handbook of psychotherapy and behavioural change (4th ed., pp. 270 –
approach to experiential change. In R. A. Neimeyer & M. J. Mahoney 376). Toronto, Ontario, Canada: Wiley.
(Eds.), Constructivism in psychotherapy (pp. 169 –191). Washington, Paivio, S. C., & Greenberg, L. (1995). Resolving “unfinished business”:
DC: American Psychological Association. Efficacy of experiential therapy using empty-chair dialogue. Journal of
Greenberg, L. S., Rice, L. N., & Elliott, R. (1993). Facilitating emotional Consulting and Clinical Psychology, 63, 419 – 425.
change: The moment-by-moment process. New York: Guilford Press. Pascual-Leone, A., & Greenberg, L. (2005). Classification of Affective-
Greenberg, L. S., & Safran, J. D. (1987). Emotion in psychotherapy. New Meaning States. In A. Pascual-Leone, Emotional processing in the
York: Guilford Press. therapeutic hour: Why “the only way out is through” (pp. 289 –367).
Hayes, A. M., Laurenceau, J. P., Feldman, G., Strauss, J. L., & Cardaciotta, Unpublished doctoral thesis, York University, Toronto, Ontario, Can-
L. (2007). Change is not always linear: The study of non-linear and ada.
discontinuous patterns of change in psychotherapy. Clinical Psychology Pascual-Leone, A., & Greenberg, L. S. (2006). Insight and awareness in
Review, 27, 715–723. experiential therapy. In L. G. Castonguay & C. E. Hill (Eds.), Insight in
Hayes, A. M. & Strauss, J. L. (1998). Dynamic systems theory as a psychotherapy (pp. 31–56). Washington, DC: American Psychological
paradigm for the study of change in psychotherapy: An application to Association.
cognitive therapy for depression. Journal of Consulting and Clinical Pascual-Leone, J. (1990). An essay on wisdom: Toward organismic pro-
Psychology, 66, 939 –947. cesses that make it possible. In R. J. Sternberg (Ed.), Wisdom: Its nature,
Hesterberg, T., Moore, D. S., Monaghan, S., Clipson, A., & Epstien, R. origins, and development (pp. 244 –278). New York: Cambridge Uni-
(2006). Bootstrap methods and permutation tests. In D. S. Moore & G. P. versity Press.
McCabe (Eds.), Introduction to the practice of statistics (5th ed., pp. Rogers, C. R. (1957). The necessary and sufficient conditions of therapeu-
14.1–14.70). New York: Freeman. tic personality change Journal of Consulting Psychology, 21, 95–103.
Hunt, M. G. (1998). The only way out is through: Emotional processing Samoilov, A., & Goldfried, M. (2000). Role of emotion in cognitive
and recovery after a depressing life event. Behaviour Research and behavior therapy. Clinical Psychology: Science & Practice, 7, 373–385.
Therapy, 36, 361–384. Stern, D. B. (1997). Unformulated experience: From dissociation to imag-
Jones, E. E., Pare, L. A., & Pulos, S. M. (1992). How is therapy conducted ination in psychoanalysis. Hillsdale, NJ: Analytic Press.
in the private consultation room: A multidimensional description of brief Warwar, S. H., & Greenberg, L. S. (1999). Emotional Arousal Scale III.
psychodynamic treatments. Psychotherapy Research, 2, 16 –30. Unpublished manuscript, York University, Toronto, Ontario, Canada.
Klein, M. H., Mathieu-Coughlan, P., & Kiesler, D. J. (1986). The Expe-
riencing Scales. In L. S. Greenberg & W. M. Pinsof (Eds.), The psy- Received April 23, 2007
chotherapeutic process: A research handbook (pp. 21–71). New York: Revision received August 15, 2007
Guilford Press. Accepted September 10, 2007 䡲