Construct Validity of a Measure of Affective Communication in Psychotherapy

Abstract

The present study evaluates the psychometric properties of a therapist measure for evaluating the affective communication created between the patient and therapist during the initial stages of treatment. The Affective Communication Questionnaire (ACQ) was administered to a sample of 81 therapists, each rating a single patient, and principal component analysis indicated the measure has coherent dimensions with strong internal consistency. The construct validity of the ACQ was then established in a sample of 16 therapists rating 73 patients diagnosed with Borderline Personality Disorder (BPD). The measure was found to have a strong relationship to the related constructs of transference, countertransference, and affect experience in predicted directions. The measure also was found to have a modest relationship to independent assessments of patient functioning; most notably more negative affect was significantly related to more odd/eccentric (cluster A) and less anxious/fearful (cluster C) personality disorder symptoms, and greater narrative coherence. Differences in affective communication as a function of treatment type were also evaluated. The clinical and research implications of the findings are discussed.

One of the predominant theories of change in psychoanalytically-oriented psychotherapy, particularly with regard to more disturbed patients and patients organized at the borderline level, focuses on the movement from the discharge of impulses to the symbolization of affective experience (Bion, 1961; Fonagy, Gergely, Jurist, & Target, 2002; Freedman & Berzofsky, 1995; Kernberg, 1975; Mayes & Cohen, 1992; Pine, 1985). While there are differences among the proponents of such theories, one of the many unifying components of such models focus on the distinction between unconscious versus conscious affective experience, both in terms of how it is experienced internally by the patient and how it becomes expressed in the psychotherapy. The task of metabolizing, organizing, and giving meaning to the patient’s unconscious affect being communicated to the therapist nonverbally is viewed as essential to psychotherapeutic intervention. However, to date researchers have not systematically measured this affective communication between the patient and therapist in psychotherapy. The purpose of the present research is to evaluate a method for reliably assessing affective communication in psychotherapy, and to examine its relationship to patient characteristics that bear significantly on psychotherapeutic outcome.

Affective Communication in Psychotherapy

While describing a predominant view of the process of change in psychotherapy risks smoothing over essential differences in theory and technique, it can be broadly stated that object relations-oriented theorists would describe treatment for patients organized at the borderline level in terms of the movement from the discharge of impulses to the symbolization of affective experience (Bion, 1961; Kernberg, 1975). Due to impairments in the internalization of representations of self in emotion-laden interactions with others, the internal worlds of patients with borderline pathology are thought to be characterized by unintegrated affect-laden mental representations that may be unconsciously communicated to the therapist through the patient’s affective experience. When the patient is filled with painful affects and self states that are too intolerable to experience consciously, rather than locate these affects within the self, the patient projects these affects onto the therapist (Blatt & Ford, 1994; Fonagy et al., 2002; Kernberg, 1984).

Due to the contagious nature of affect, the therapist is likely to identify with, experience, and possibly enact the feelings that the patient is evoking (Westen & Gabbard, 2002). The therapist may experience him/herself as filled with intense feelings, thoughts, images, and fantasies that over time the therapist can come to identify and eventually verbalize to the patient. The therapist internally processes and eventually translates the patient’s unconscious affect into more symbolic representations of affect. In doing so, the therapist metabolizes the affective experience, gives organization and meaning to it, and gives it back to the patient in a more tolerable form (Bion, 1961; Kernberg, 1975; Winnicott, 1971).

Through the process of organizing and representing the patient’s affective experience, there is a shift from discharge of affects through action to symbolic expression through language (Pine, 1985). The development of higher-order order representations of constitutional affect states is thought to be not only essential for the delay and control of strong affects, but is also related to the developmental maturity of mental representations of self and significant others. (Diamond & Blatt, 1994; Levy, 2000).

Kernberg (1975, 1984) has emphasized that there are three channels by which the patient communicates his/her internal experience to the therapist: the content of the patient’s language, the patient’s nonverbal communication, and the therapist’s countertranference. While feelings that the patient can identify and articulate are primarily communicated through the first channel, aporetic feelings are primarily communicated through the latter channels. Kernberg argues that due to impairments in the internalization of representations of self in emotion-laden interactions with others, the internal experience of patients with borderline pathology are characterized by unconscious affect-laden mental representations that are split off and unintegrated with one another. As a result, the patient may be unaware that the content of their language is not in integrated with, and often contradicted by, what is communicated through affect and action.

Jurist (2005) also makes a distinction between affects within and outside of awareness when he distinguishes between strong feelings, in which the patient knows what he or she feels and can identify and articulate the experience, and aporetic feelings, in which the patient’s affect is experienced as vague, diffuse, often contradictory, and confusing. This does not suggest that aporetic feelings are not strongly felt; in fact they are often quite powerful, but their unformulated quality belies the patient’s capacity to know how he or she feels. Thus such feelings are not communicated through the patient’s conscious expression through language, but rather are inferred through the therapist’s affective experience of the patient in sessions.

Westen (1990; 1998; Westen & Gabbard, 2002a, b) has integrated psychoanalytic models with findings from cognitive neuroscience to elaborate the mental structures that underlie distinctions between conscious and unconscious affective experience. These literatures each contribute to a growing consensus that cognitive, affective, and motivational components of mental representations are distinct but interacting, and often competing, systems. Further, any of these components of a representation can be highly active and influential on behavior and yet not available to consciousness.

To demonstrate the influence of unconscious affects on behavior, Westen (1998) cites a rich body if research on prejudice. He cites Fazio, Jackson, Dunton, and Williams’ (1995) groundbreaking study in which they demonstrated that implicit racial attitudes measured by reaction times to adjectives when presented with Black and White faces predicted the degree to which a Black confederate later rated the subject as friendly versus cold during a brief interaction. The subjects’ demeanor towards the Black confederate was unrelated to conscious racial attitudes, but powerfully related to implicit affective associations. This research suggests that while an individual’s affective experience may not be consciously represented, such affect may nonetheless exert influence on behavior in ways the individual is not aware of, but indications of this affect are most evident in a relational context, and perhaps most identifiable by the recipient of this unconscious affect.

Such findings are consistent with clinical observation that the most telling information on the patient’s affective experience often comes not from the conscious report of their affect, but rather inferred through the therapist’s own affective experience of the patient. Further, the degree to which there is less integration of the patient’s affect-laden introjects, the more the therapist’s phenomenological experience of the patient must be relied upon to interpret the affect being communicated. Because this affective experience is co-created between patient and therapist, affective communication can be best conceptualized as an “interactional structure” in the therapy (Jones, 2000).

Bucci (1997), also integrating psychoanalytic and cognitive models, nicely describes the phenomenological experience of the therapist when listening to a patient whose language, imagery, and emotional experience are highly integrated, indicative of high referential activity. By way of example, she draws a parallel to a reader’s experience of being drawn into a story – as the reader engages the narrative, pictures and fantasized sensory images come to mind. In a similar fashion, if a patient were to use language that is imagistic during a session, the therapist would feel enlivened as images and associations were evoked in the therapist. This experience would reflect the multiple sensory and information neural networks integrated to create vividness to the patient’s narrative.

Freedman and Berzofsky (1995) make a similar observation when they differentiate between talking and “language deployed in the service of a communicated wish…a wish to impart inner objects – the introjects – into the consciousness of the listening analyst and therapist as object” (p. 366), the latter being indicative of what they call a communicated transference. This manifests in the patient’s capacity to present a rich and coherent narrative in which feelings are put into words, and ambivalence is tolerated. The therapist in turn feels enlivened by the patient and experiences the richness of his/her language as filled with imagery and affect. They contrast this with less integrated narratives that are disorganized or difficult to follow, barren, and deplete of affect and imagery, leading the therapist to feel disengaged from the patient, experience boredom and sleepiness.

Overall this literature suggests that when patients arouse an affective experience in psychotherapy, positive or negative, they may be unconsciously communicating their unintegrated affect-laden introjects to the therapist. The degree to which therapists feels that their attention and curiosity has been engaged, that they experience the patient’s language as fresh and vivid, or alternately feel bored and disengaged from the material, this interaction represents an essential affective communication that allows for eventual the articulation and integration of what previously could not be consciously represented but only felt intersubjectively.

Research on Affective Experience in Psychotherapy

In the clinical psychodynamic tradition, while there has been much rich exploration of affective experiences between patient and therapist in psychotherapy from the standpoint of theory, case studies, and qualitative studies, very little empirical research has focused on constructs related to this affective communication we have described. In the psychotherapy research tradition, while there has been much research focused on the patient’s experiencing of emotion in session (Greenberg & Safran, 1987), less attention has been paid to affective experience as an “interactional structure” created between patient and therapist (Jones, 2000).

Westen and colleagues have examined transference and countertransference patterns in psychotherapy that relate in significant ways to affective communication (Betan, Heim, Conklin, & Westen, 2005; Bradley, Heim, & Westen, 2005). Their research, in which clinicians are asked to fill out questionnaires on their patients, indicates that affective experience in sessions differs as a function of DSM-IV personality symptoms. The transference patterns associated with cluster A symptoms were characterized by a cold and unengaged stance towards the therapist, and the therapists’ countertransferential experience was one of feeling criticized and mistreated. The transference patterns associated with cluster B symptoms were characterized as an angry, entitled, and sexualized stance towards the therapist, and therapists’ countertransferential experience was one of feeling overwhelmed, disorganized, helpless, and disengaged. The transference patterns associated with cluster C symptoms were characterized as anxious and preoccupied, and therapists’ countertransferential experience was one of feeling parental and protective towards the patient. This research suggests that the nature of the patient’s personality pathology will differentially contribute to the way in which the patient’s unconscious affective experience is communicated in treatment.

Westen and colleagues have also examined affect regulation and experience patterns in patients through the use of clinician ratings (Westen, Muderrisoglu, Fowler, Shedler, & Koren, 1997, Conklin, Bradley, & Westen, 2006). They have identified affect patterns that differentiate BPD from Dysthymic Disorder (DD) patients; with higher dysregulation, externalization, and disorganization in BPD patients, and higher affective availability and reality-focused coping in DD patients (Conklin et al., 2006). Affect patterns have also been related to behavioral criteria, such as increased past suicide attempts (with lower positive affect, socialized negative affect, reality-focused responses, avoidant defenses, and with higher intense negative affect and externalizing defenses) and past hospitalizations (with lower socialized negative affect, reality-focused responses, and with higher intense negative affect; Westen et al., 1997).

Jones and colleagues have evaluated psychotherapy process by having independent evaluators rate videotaped psychotherapy sessions using a Q-sort procedure (Jones, 2000). Their research has shown differences in affective experience as a function of treatment type; psychodynamic therapy was found to elicit more affective experience in sessions as compared to Cognitive Behavioral Therapy (CBT; Jones & Pulos, 1993). However, Interpersonal Therapy (IPT) was not found to elicit more affective experience in sessions than CBT (Coombs, Coleman, and Jones (2002). They have also shown differences in outcome as a function of affective experience; while negative affect was negatively correlated with favorable outcome in CBT, it was positively correlated with therapist ratings of outcome in psychodynamic therapy (Jones & Pulos, 1993). In a separate study of CBT and IPT, higher observer-rated painful negative affect in the treatment was negatively related to greater collaborative emotional exploration and to favorable outcome in these treatments (Coombs et al., 2002). Taken together, this research suggests that the degree to which the presence of painful negative affect impedes favorable outcome may differ as function of treatment type, with the presence of negative affect potentially enriching psychodynamic therapy while impeding CBT and IPT.

While these programs of research have each made invaluable contributions to our understanding of the affective experience between patient and therapist in sessions, each has examined this phenomenon only to the degree to which such experiences were represented in specific factors among the breadth of many psychotherapy experiences being evaluated in these treatments. As a result the specific affective communication between patient and therapist has not been evaluated in depth. The ability to measure such phenomenon with greater specificity than has been previously available would be of great benefit to psychotherapy research.

The Present Studies

The present studies examine initial data on a self-report questionnaire for therapists evaluating an affective communication created between the patient and therapist. The primary aim of the first study was to establish the psychometric properties of the affective communication questionnaire (ACQ). The ACQ was administered to a large sample of therapists, each rating a single patient, in order to examine the measure’s reliability through the internal consistency of its component structure in a sample in which each rating represents an independent observation. We hypothesized that the ACQ would have a cohesive factor structure representing distinct and interpretable domains of affective experience in sessions, both in terms of intensity and valence. The aims of the second study were to 1) examine the construct validity of the measure in the context of a nomological network by demonstrating its relationship to other measures of similar constructs, such as transference, countertransference, and affect regulation and experience, and 2) establishing the measure’s external validity through its relationship to independent measures of patient functioning, such as personality symptomatology, attachment variables, and retention in psychotherapy. We hypothesized that the ACQ factors would positively relate to similar or near neighbor constructs, with specific hypothesized relationships bolded in Table 3. Lastly, we were interested in evaluating whether ACQ factors differ between types of treatment, with prominent affective characteristics expected to mirror the putative foci of that treatment’s early stages. Specifically, we hypothesized that Transference Focused Psychotherapy (TFP) therapists would identify more negative affect in the first four months of treatment, while Dialectical Behavior Therapy (DBT) therapists would identify more engagement in treatment. Taken together, this research aims to validate a measure of the affective experience between patient and therapist that reflects both clinical richness and psychometric rigor.

Table 3.

Correlation between ACQ and Transference, Countertransference, and Affect Components, Total DSM-IV Personality Disorder Cluster Symptoms, and AAI Ratings

	Disengaged	Full Range	Neg. Affect	Enlivened
Transference:
Hostile	.28*	−.10	.60 **	.22
Narcissistic	.27*	.00	.43 **	.30*
Preoccupied	.23	−.11	.29*	.11
Positive Alliance	−.28*	.32 **	−.35**	.52 **
Avoidant/Dismissing	.44 **	−.28*	.49 **	.04
Sexualized	.19	−.07	.38**	.25*
Countertransference:
Hostile/Mistreated	.35**	−.17	.79 **	.12
Helpless/Inadequate	.36**	−.23	.67 **	.15
Positive/Satisfying	−.35**	.44 **	−.62**	.44 **
Parental/Protective	−.07	.20	−.32**	.36 **
Overwhelmed/Disorganized	−.02	.15	.29 *	.55**
Special/Overinvolved	−.08	.22	−.13	.62**
Sexualized	.16	.06	.17	.29*
Disengaged	.81 **	−.49**	.51**	−.41**
Affect Regulation and Experience:
Affect Regulation:
Emotional Avoidance:	.40 **	−.31**	.52**	−.02
Externalization	.22	−.07	.57 **	.13
Healthy Coping:	−.14	.21	−.39**	.32 **
Passive Avoidance:	.19	−.19	.13	−.14
Affective Experience:
Negative Affect:	.14	−.19	.39 **	−.09
Affect Availability:	−.59**	.70 **	−.62**	.30 **
Emotional Dysregulation:	.17	.07	.44 **	.26*
Positive Affect:	−.38**	.54 **	−.49**	.43 **
DSM-IV Personality DO
Cluster Symptoms:
Cluster A1	.12	−.13	.33 *	.03
Cluster B2	−.04	−.03	−.07	−.08
Cluster C3	.13	−.12	−.25*	−.29*
AAI Ratings:
Coherence4	−.09	.23	−.28*	.13
Reflective Function5	−.08	.00	−.15	−.09

Note: Hypothesized correlations in bold

¹Partial correlations controlling for Cluster B and C scores on a subset of 65 subjects for whom both IPDE and ACQ data were available

²Partial correlations controlling for Cluster A and C scores on a subset of 65 subjects for whom both IPDE and ACQ data were available

³Partial correlations controlling for Cluster A and B scores on a subset of 65 subjects for whom both IPDE and ACQ data were available

⁴Pearson correlations on a subset of 59 subjects for whom both AAI and Coherence data were available

⁵Pearson correlations on a subset of 70 subjects for whom both AAI and RF data were available

^**Significant at the 0.01 level (2-tailed)

^*Significant at the 0.05 level (2-tailed)

Study 1: Internal Consistency of the ACQ

Method

Participants

The 81 participants in the present study were comprised of two samples of therapists, each rating a single patient. The first sample comprised of 65 therapists in training who filled out measures, including the ACQ, on a “difficult patient” of their choice that they were currently working with. Their participation was part of a larger unrelated study evaluating characteristics of psychotherapist that relate to positive working alliance. Therapists were recruited and participated online through clinical listserves and were compensated with a $25 gift card. This sample consisted of 55 female and 10 male therapists of eclectic theoretical orientations, with 53.9% psychodynamic, 50.8% cognitive-behavioral, 23.0% humanistic, 21.5% family systems, and 12.3% other (primarily Interpersonal Therapy and Dialectical Behavior Therapy) as the orientation they “often” or “always” utilize (though the categories were not mutually exclusive). Most therapists were currently in training in clinical psychology doctoral programs (47.7% Ph.D. and 29.2% Psy.D.), though a minority also attended school or counseling psychology programs at the doctoral level (15.4%) and the Master’s level (4.6%) as well as psychiatry residency programs (3.1%). Most therapists were late in their training, in the 5^th (35.4%) or 6^th year and beyond (21.5%), though some therapists were earlier in their training (3.1% in the 1^st, 15.4% in the 2^nd, 13.8% in the 3^rd, and 10.8% in the 4^th year).

The second sample consisted of 16 therapists treating patients in a randomized controlled trial for Borderline Personality Disorder (BPD) at a large university-affiliated medical center. The 16 therapists were of diverse theoretical orientations, with 6 therapists practicing Transference Focused Psychotherapy (TFP), 6 therapists practicing a psychodynamically-oriented supportive psychotherapy (SPT), and 4 therapists practicing Dialectic Behavioral Therapy (DBT). All were experienced therapists, with between two and fifteen years of experience treating BPD patients, and specific training in their respective manualized treatments (see Clarkin, Levy, Lenzenweger, & Kernberg, 2007 as well as Levy, Meehan, Kelly, Reynoso, Weber, Clarkin, & Kernberg, 2006 for additional therapist characteristics). In the present study, therapists’ ACQ ratings on a single patient each were selected; because ratings were used to establish the measure’s reliability it did not matter which of the many patients rated was chosen, so long as each rating represented an independent observation. For consistency, we chose to include the first patient on the therapists’ caseload that was randomized and treated for a minimum of 6 months (to ensure they had enough experience with the patient to make accurate ratings).

Measures

Affective Communication Questionnaire. (ACQ; Meehan, 2004) The ACQ is a 28-item self-report measure that asks therapists to rate their patients in terms of the degree to which they felt enlivened and engaged by the patient, the nature of the affect experienced in sessions with the patients, and the degree to which the patient imbued his/her language with affect. Each statement is rated on a 5-point scale, ranging from 1 (not true) to 5 (very true) in terms of how much the given statements characterized the therapist’s work with the patient.

The 28 items in the measure were derived from clinical and empirical literature that focused on implicit communication of affect in the therapeutic context, most notably the three channels of communication described by Kernberg (1975, 1984), the communicated transference described by Freedman and Berzofsky (1995), referential activity described by Bucci (1997), and interactional structures described by Jones (2000). Items were designed to measure the affective experience co-created between patient and therapist, and thus they evaluate multiple levels of this affective experience in terms of intensity and valence, including how both patient and therapist feel (“I find that this patient brings a full range of emotions to sessions ”, “I find that this patient tends to stir up mostly negative feelings in me”), think (“I notice this patient often referring back to emotionally meaningful moments in previous sessions”, “I tend to think about this patient quite a bit in-between our sessions”), and act (“I often experience the patient’s language as repetitive and monotonous”, “I find it hard to give this patient my full attention”). Items were written to be unambiguous and easily identifiable experiences for therapists of all theoretical orientations, and thus could be utilized a wide range of clinical contexts in which the affective quality of the relationship could be evaluated. Several senior clinicians were asked to then critique the items in terms of clarity and clinical relevance, which led to further editing and refinement of the items until agreement was reached on their clarity and comprehensiveness for capturing the target construct.

Results

Component Structure of the ACQ

To identify the component structure of the ACQ, a principal component analysis with varimax rotation was conducted on the questionnaire items. Because our sample size was under 150, we adopted a conservative standard of including only those items with loadings of 0.60 as recommend by Stevens (2002). Five main factors emerged using Kaiser’s criteria of eigenvalues greater than 1, accounting for 68.90% of the variance in scores. However, scree testing indicated a four factor solution, and the fifth factor only included one item with a loading over 0.60. Due to the strong interpretability and internal consistency of the first four factors, indicating that they represent cohesive constructs, a four-factor solution was retained (see Table 1).

Table 1.

Component Structure of the ACQ

	1	2	3	4
Component 1: Disengaged
13. My mind often wanders to things other  than what this patient is talking.	.83
10. I often feel bored during sessions with  this patient.	.80
14. I don’t feel fully engaged with this  patient during sessions.	.73
3. I find it hard to give this patient my full  attention.	.72
25. I often experience the patient’s language  as repetitive and monotonous.	.68
4. I often experience the patient’s language  as stagnant.	.68
20. I often have difficulty remembering  what happened in the last session with this  patient.	.67
12. I often experience the patient’s language  as barren.	.65
Component 2: Full Range of Emotions
1. I often experience the patient’s  descriptions of events as full of emotion.		.77
5R. I often experience this patient as  showing very little emotion, even when  describing what sound like very emotional  events.		−.77
16R. I often experience this patient as  emotionally constricted.		−.75
21. I experience this patient as very  emotionally present in sessions.		.74
8. I find that this patient brings a full range  of emotions to sessions.		.72
Component 3: Negative Affect
19R. I find that this patient tends to stir up  warm, positive feelings in me as well as  frustrated, angry feelings.			−.74
9. I find that this patient tends to stir up  mostly negative feelings in me.			.68
28. I find this patient difficult to engage  during sessions.			.62
27. I find that this patient brings mostly  negative emotions to sessions.			.60
Component 4: Enlivened
24. The things this patient says tend to stick  with me well after our sessions.				.82
22. I tend to think about this patient quite a  bit in-between our sessions.				.76
23. I find this patient stimulating to work  with.				.64
Eigenvalue	10.87	2.85	2.57	1.91
Cumulative % of Variance	38.80	48.98	58.14	64.97

As indicated in Table 1, the first component included 8 items representing therapists experiencing their patients and themselves as each feeling disengaged from the treatment (α = .91). The second component included 5 items representing the therapist experiencing a full range of emotion in the treatment (α = .88). The third component included 4 items representing the therapist experiencing a predominance of negative affect in the treatment (α = .77). The fourth component included 3 items representing therapists experiencing their patients and themselves as feeling enlivened in the treatment (α = .80).

As for the relationship between factors, as shown in Table 2, the correlations were in expected directions, with full range and enlivened positively correlated to one another, and each negatively correlated with negative affect and disengaged. The components were found to have moderate to large correlations with one another, with coefficients ranging from .24 to .57. The fact that many of these correlations are strong indicates that these factors are interrelated, and therefore independence cannot be assumed among these components. While the dependence between components is high, it is not inconsistent with theoretical expectations.

Table 2.

Correlation between ACQ Components

	Disengaged	Full Range	Neg. Affect	Enlivened
Disengaged	-	−.57**	.49**	−.46**
Full Range		-	−.44**	.24*
Neg. Affect			-	−.39**
Enlivened				-

^**Significant at the 0.01 level (2-tailed)

^*Significant at the 0.05 level (2-tailed)

Study 2: Construct Validity of the ACQ

Method

Participants

The participants in the present study were therapists treating patients in a randomized controlled trial for Borderline Personality Disorder (BPD) at a large university affiliated medical center, as discussed in Study 1. The patients were randomized to one of the three treatment conditions for a no cost 1-year outpatient treatment: Transference Focused Psychotherapy (TFP), Supportive Psychotherapy (SPT), and Dialectic Behavioral Therapy (DBT). To meet criteria for the RCT, patients must have met at least five of the nine DSM-IV criteria for BPD, must not have met criteria for schizophrenia or any psychotic disorders, bipolar disorder, organic pathology, or mental retardation, and have been between the ages of 18 and 50. Ninety BPD patients were randomized to one of the three treatment cells for a one-year treatment (see Clarkin et al., 2007 as well as Levy et al., 2006 for a full description of the study design).

In the present study, therapists from the three treatments retrospectively filled out self-report questionnaires on their patients treated in the RCT. Sixteen therapists from the three treatments completed questionnaires retrospectively on 73 patients: 4 DBT therapists on 28 patients, 6 SPT therapists on 19 patients, and 6 TFP therapists on 26 patients. One patient was excluded because she met criteria for a psychotic disorder, 2 patients were not rated because their therapist was deceased prior to the beginning of the present study, and 14 patients dropped out of treatment too soon for therapists to rate the quality of the therapeutic relationship. Therapists were asked to rate what was typical of first 4 months of the treatment. These ratings were then compared to baseline assessments of personality disorder symptoms (IPDE) and attachment variables (coherence and RF on the AAI).

Measures

Affective Communication Questionnaire. (ACQ; Meehan, 2004) As in study 1, the ACQ asks therapists to rate their patients along the aforementioned dimensions. The factor structure from study 1 was retained, with α values reported for these factors in the present sample: disengaged (α = .91), full range of emotion (α = .82), negative affect (α = .73), and enlivened in the treatment (α = .85).

Therapist-Rated Comparison Measures

Psychotherapy Relationship Questionnaire. (PRQ; Westen, 2000; Bradley et. al., 2005) The Psychotherapy Relationship Questionnaire is a 90-item self-report measure that asks therapists to rate the thoughts, feelings, and behaviors expressed by their patients in order to characterize the pattern of transference in the therapeutic relationship.

Previous research has identified six transference factors (with α values reported for these factors in the present sample): hostile (α = .94; e.g., “Feels critical of the therapist”), narcissistic (α = .80; e.g., “Needs excessive admiration from the therapist”), compliant/anxious preoccupied (α = .64; e.g., “Is overly compliant”), positive working alliance (α = .75; e.g., “Works hard in therapy”), avoidant/dismissing attachment (α = .69; e.g., “Tries hard not to feel needy or dependent in therapy”), and sexualized (α = .69; e.g., “Is sexually attracted to the therapist”).

Countertransference Questionnaire. (CTQ; Zittel & Westen, 2003; Betan et. al., 2005) The Countertransference Questionnaire is a 79-item self-report measure that asks therapists to rate their patients in terms of the emotional and countertransferential reactions evoked in the therapist during the treatment.

Previous research has identified eight countertransference factors (with α values reported for these factors in the present sample): hostile/mistreated (α = .95; e.g., “I feel criticized by him/her”), helpless/inadequate (α = .88; e.g., “I feel incompetent/inadequate working with him/her”), positive satisfying (α = .89; e.g., “S/he is one of my favorite patients”), parental/protective (α = .85; e.g., “I feel like I want to protect him/her”), overwhelmed/disorganized (α = .87; e.g., “I feel overwhelmed by his/her needs”), special/overinvolved (α = .77; e.g., “I disclose my feelings with him/her more than other patients”), sexualized (α = .67; e.g., “I feel sexually attracted to him/her”), and disengaged (α = .73; e.g., “My mind often wanders to things other than what she is talking about”).

Affect Regulation and Experience Q-sort, Questionnaire Version (AREQ-QV; Westen et al., 1997, Conklin et al., 2006) The AREQ-QV is a 98-item self-report measure that asks therapists to rate their patients along multiple dimensions of affective experience and affect regulation. Therapists were asked to rate what was typical of their patient’s affect during the first 4 months of the treatment.

Previous research has identified four affect regulation factors (with α values reported for these factors in the present sample): emotional avoidance (α = .82; e.g., “Tends to have difficulty acknowledging feeling sad”), externalization (α = .84; e.g., “Tends to lash out at others when distressed or angry”), healthy coping (α = .88; e.g., “Able to anticipate problems and develop realistic plans for dealing with them”), and passive avoidance (α = .70; e.g., “Tends to remain at home or restrict travel or activities when distressed”). Four affective experience factors have previously been identified as well: negative affect (α = .76; e.g., “Tends to feel guilty”), affect availability (α = .69; e.g., “Displays emotion appropriate in quality and intensity to the situation at hand”), emotional dysregulation (α = .82; e.g., “Feelings tend to change rapidly from moment to moment”), and positive affect (α = .41; e.g., “Tends to feel excited or energized”).

Patient Comparison Measures

International Personality Disorder Examination. (IPDE; Loranger, 1995) The IPDE is a semi-structured diagnostic interview for diagnosing personality disorders. It consists of 99 items arranged in six categories (e.g., Self or Work), along with a detailed scoring manual (Loranger et al., 1994). The IPDE generates probable (subthreshold number of DSM-IV criteria met) and definite diagnoses for each of the DSM-IV diagnoses. It also generates dimensional scores for each diagnosis by adding the ratings on all the criteria composing a diagnosis. Reliability of assessment interviews indicated good to excellent levels of inter-rater reliability; the Kappa for BPD diagnosis was .64 and the ICC for dimensional criteria ratings was .86. All Kappa and ICC coefficients were in the good to excellent range (Fleiss, 1971).

Adult Attachment Interview (AAI; George, Kaplan, & Main, 1985). The Adult Attachment Interview is a semi-structured clinical interview designed to elicit thoughts, feelings and memories about early attachment experiences, and to assess the individual’s state of mind or internal working model with regard to early attachment relationships. The AAI is transcribed verbatim and coded using subscales ratings. This study focused on the narrative coherence subscale of the AAI, which assesses the subject’s capacity to provide a succinct yet and thorough and organized description of early attachment experiences that are supported by specific and relevant examples. The coherence subscale has been found to be the best predictor of attachment security, r = .96, p < .001 (Waters, Treboux, Fyffe, & Crowell, 2001). Raters were blind to all identifying characteristics of the subjects, and reliability was established with different raters coding a subset of each other’s transcripts (n= 22, ICC=.88).

Reflective Function (RF; Fonagy et al., 1998). The AAI was also scored with the Reflective Function (RF) Scale, an 11-point scale that evaluates the quality of mentalization in the context of attachment relationships. Passages are rated on a scale of −1 (negative RF, concrete, totally barren of mentalization or grossly distorting of the mental states of others) to 9, (exceptional RF, complex, elaborate or original reasoning about mental states), and these scores are then aggregated to provide an overall score for the transcript. Raters were blind to all identifying characteristics of the subjects, and reliability was established with different raters coding a subset of each other’s transcripts (n= 26, ICC=.86).

Results

Sample Characteristics

The 73 participants rated by their therapists in this study included 67 women and 6 men. Participants had a mean age of 31.0 years (SD = 7.9) and were 66% Caucasian, 10% African-American, 10% Hispanic, 7% Asian, 4% mixed ethnicity/race, and 4% other. In terms of marital status, 8% were married, 13% were living with a partner, 44% were divorced, and 24% were in relationship. Participants were fairly well educated, with 49% having a 4-year college degree and 88% having completed some college, but underemployed, with 62% in some employment and 33% in full-time employment. No significant differences were found between this subset of 73 patients and the 17 patients not evaluated in terms of demographic variables [age (t(88) = −.22, ns), gender (χ²(1) = .03, ns), ethnicity (χ² (5) = 3.10, ns)] or in terms of level of psychopathology [Time 1 Total IPDE symptoms (t(88) =.13, ns)].

Construct Validity of the ACQ

Cronbach and Meehl (1955) have argued that the construct validity of a measure must be demonstrated in the context of a nomological network by demonstrating the relationship of that measure to other measures of similar constructs. Therefore the ACQ components were correlated with measures of transference (PRQ; Westen, 2000), countertransference (CTQ; Zittel & Westen, 2003), and affect regulation and experience (AREQ; Conklin et al., 2006).

Transference

Table 3 shows the correlation between the ACQ and PRQ components, with hypothesized relationships to similar or near neighbor constructs indicated in bold. Overall the data strongly supports the hypothesized relationships (e.g., positive working alliance correlating with the relationship being characterized as enlivened and with a fuller range of affect; avoidant/dismissing attachment correlating with the relationship being characterized as disengaged and with more negative affect). In terms of relationships not hypothesized, patients rated high on complaint/anxious preoccupation, while expected to withdraw in sessions creating a predominantly disengaged climate (only a trend level association), instead had relationships predominantly characterized by negative affect. Conversely, while patients rated high on hostility had relationships strongly characterized by negative affect as expected, hostility was also associated with a disengaged emotional climate. Thus there seems to be more covert expression of affect in overt hostility than hypothesized, while there is more overt expression of affect in covert anxious preoccupation than hypothesized.

Further, while patients rated as narcissistic did have relationships characterized by negative affect as expected, narcissistic transference was also associated with a more disengaged emotional climate and yet surprisingly therapist reported feeling more enlivened. Also contrary to expectations was the finding that therapists experiencing their patients as sexualized correlated with therapists experiencing more negative affect and yet also a more enlivened emotional climate.

Countertransference

Table 3 also shows the correlation between the ACQ and CTQ components, with hypothesized relationships to similar or near neighbor constructs indicated in bold. Overall the data strongly supports the hypothesized relationships (e.g., positive countertransference correlating with the relationship being characterized as more enlivened and with a fuller range of affect; disengaged countertransference correlating with the relationship being characterized as more disengaged and with greater negative affect). To highlight some surprising non-hypothesized relationships, therapists’ feeling countertransferentially disorganized by, overinvolved with, and sexually attracted to the patient each correlated with feeling more enlivened in the treatment. Also not hypothesized but consistent with clinical theory, while therapists’ feeling countertransferentially mistreated and made to feel inadequate did have relationships characterized by negative affect as expected, each was also associated with a disengaged emotional climate.

Affect Regulation and Experience

Table 3 also shows the correlation between the ACQ and AREQ components, with hypothesized relationships to similar or near neighbor constructs indicated in bold. Overall the data strongly supports the hypothesized relationships (e.g., patient positive affect and greater affect availability correlating with the relationship being characterized as more enlivened and having a more full range of emotional expression; patient negative affect correlating with the relationship being characterized as having more negative affect). To highlight some non-hypothesized relationships, while patients rated high on emotional avoidance did have relationships characterized by a disengaged emotional climate, it was also associated with more negative affect. Also not hypothesized but consistent with clinical theory, while patients rated as emotionally dysregulated did have relationships characterized by negative affect as expected, it was also associated with therapists feeling more enlivened.

Comparison of ACQ to Patient Variables

Despite the strong relationship demonstrated between the ACQ and measures of transference, countertransference, and affect regulation, a limitation of this methodology is that ratings on these questionnaires may not be independent of one another in that all are based on therapist report. Therefore, in order to demonstrate the external validity of the ACQ, components were also compared to independent measures of patient functioning.

ACQ and Personality Disorder Symptoms

First, as shown in Table 3, ACQ components were compared to the total DSM-IV personality disorder cluster symptoms as rated by an independent evaluator on the IPDE, with hypothesized relationships to similar or near neighbor constructs indicated in bold identified. Because of the high comorbidity of Axis II pathology in patients diagnosed with BPD, symptoms were analyzed at the level of clusters (with the “odd or eccentric” symptoms of schizoid, schizotypal, and paranoid PDs in cluster A; the “dramatic, emotional, or erratic” symptoms of borderline, antisocial, histrionic, and narcissistic PDs in cluster B; and the “anxious or fearful” symptoms of avoidant, dependent, and obsessive-compulsive PDs in Cluster C; APA, 2000). The total number of symptoms in each cluster was summed, and partial correlations were used to control for comorbidity across clusters. This same methodology was used by Westen and colleagues (Bradley et al., 2005; Betan et al., 2005) to demonstrate the construct validity of the PRQ and CTQ measures.

Higher Cluster A symptoms were significantly associated with therapists experiencing the relationship with the patient as being predominated by negative affect, but contrary to expectations this cluster was not associated with experiencing a more disengaged climate and less full range of emotions. Further, there was a significant relationship between Cluster C symptoms and the therapist experiencing the relationship as less enlivened and with less negative affect, but contrary to expectations this cluster was not associated with a more disengaged emotional climate. Notably, ACQ components were not found to relate to Cluster B symptoms, which is contrary to previous research (Betan, Heim, Conklin, & Westen, 2005; Bradley, Heim, & Westen, 2005).

ACQ and Attachment Variables

Second, as also shown in Table 3, ACQ components were compared to ratings of coherence and RF on the AAI by an independent coder, with hypothesized relationships identified based on the item content of the components. Higher coherence was significantly associated with therapists experiencing the relationship with the patient as having less negative affect, though contrary to expectations coherence was not associated with a more enlivened emotional climate, and there was only a trend towards a fuller range of affect. Contrary to expectations higher RF was not associated with feeling more enlivened or a fuller range of affect.

ACQ and Treatment Completion

As shown in Table 4, ACQ components were compared to completer status of twelve months of treatment in the RCT, with all components hypothesized to be related to treatment completion. For patients who completed treatment, their therapists characterized the treatments as having more negative affect, and there was a trend towards having a fuller range of emotions as compared to the treatments of patients who dropped-out, though contrary to expectations completion was not associated with a less disengaged and more enlivened emotional climate.

Table 4.

Relationship between ACQ Components and Completer Status

	Completer (N=45)		Drop-out (N=28)
	M	(SD)	M	(SD)	t (71)
Disengaged	2.19	.94	2.24	.69	−.25
Full Range	3.04	.95	2.67	.71	1.74
Negative Affect	2.78	.88	3.25	.74	−2.31*
Enlivened	2.69	.89	2.42	1.02	1.17

^**Significant at the 0.01 level (2-tailed)

^*Significant at the 0.05 level (2-tailed)

ACQ and Treatment Group Differences

Finally, as shown in Table 5, Analysis of Variance with Post Hoc testing (Bonferroni) for differences between treatment groups on ACQ factors indicate that TFP therapists were found to report experiencing more negative affect in the treatment as compared to DBT and SPT therapists. In addition, SPT therapists were found to report feeling more enlivened in the treatments than TFP therapists, and each reported feeling more enlivened in the treatments than DBT therapists. No other between group differences was found on the ACQ factors.

Table 5.

Analysis of Variance of Treatment Group Differences on ACQ

	1- TFP (n=26)		2- DBT (n=27)		3- SPT (n=20)
	Mean	SD	Mean	SD	Mean	SD	F(2,70)	Post hoc
Disengaged	2.47	0.69	2.06	0.82	2.13	1.03	1.69
Full Range	2.75	0.69	2.94	0.87	3.08	1.08	0.83
Neg. Affect	3.37	0.72	2.77	0.89	2.75	0.85	4.52*	1>2,3
Enlivened	2.67	0.65	1.95	0.89	3.27	0.87	15.58**	3>1>2

^**Significant at the 0.01 level.

^*Significant at the 0.05 level.

Discussion

The present study examined initial data on a therapist report measure for evaluating the affective communication created between the patient and therapist during treatment. We evaluated the structure of the ACQ through principal component analysis, showing it to have coherent dimensions and a strong relationship to similar constructs. Four dimensions of affective communication were identified, with high internal consistency and interpretability, indicating that they represent cohesive constructs.

Further we found that the four affective communication dimensions related to transference, countertransference, and affect dimensions in predicted directions. There were two patterns across these relationships that are of note. First, components characterizing overt expression of negative affect in the treatment (e.g., hostility, being made to feel inadequate) were found to be related to more covert forms of affective communication (disengaged relatedness) than expected. Conversely, components characterizing more covert expression of affect in the treatment (e.g., avoidance, anxious preoccupation) were found to be related to more overt negativity in the affective communication than expected. This finding is consistent with Kernberg’s (1975, 1984) assertion that negative affect and aggression are often communicated in both covert and overt forms, and that patients with personality disorders may oscillate between these modes of expression.

Second, feeling enlivened by treatment was not found to occur exclusively in therapeutic relationships that were predominated by positive transference and countertransference. Rather, intense and at times overwhelming affect, even if not necessarily positive in valence, may draw both the patient and therapist into an interchange that feels very “live in the here and now”. Contrary to expectations, patients presenting as emotionally dysregulated, and therapists’ feeling disorganized by, overinvolved with, and sexual attraction towards patients each related to feeling more enlivened in the treatment. Each of these transference, countertransference, and affective dimensions involve experiences in which therapists feels flooded by emotion. A number of writers have noted that the transference and countertransference with BPD patients is frequently intense and stormy in a manner that can be experienced by the therapist as both engrossing and overwhelming, often leaving the therapist feeling intensely but ambivalently involved with the patient. However, findings did not simply indicate that more affect of any kind in the treatment correlates with feeling more enlivened in the treatment. Therapists whose countertransferential experience of the patient left them feeling helpless, attacked, mistreated, and deskilled by the patient was associated with a disengaged emotional climate in treatment.

One example of this is the finding that patients with narcissistic transference patterns had relationships characterized by more negative affect, more disengaged emotional climate, and yet therapists also noted feeling more enlivened. This apparent contradiction can be understood in the context of the object relational experience of these patients, whose unconscious affect-laden mental representations that are split off and unintegrated with one another may vacillate during the course of treatment (or even within a session hour) (Kernberg, 1975). As a result the affective experience of the clinician may vacillate as well, with moments of intense engagement followed by barren negativity.

This vacillation in the affective experience with the patient may be particularly true of sexual feelings in the therapeutic relationship with BPD patients, which contrary to the subtle longing seen in the therapies of higher functioning patients, is often expressed in overt, primitive and highly chaotic ways (Gabbard, 1996; Kernberg, 1984). It may be the case that such sexuality is both alluring and unnerving to therapists, leaving them simultaneously feeling enlivened by and trying to defensively disengage from such feelings (see the case of “Nicole” in Diamond, Stovall-McClough, Clarkin, & Levy, 2003 for a clinical illustration of this process). Future research attempting to replicate these findings on a non-personality disordered sample would be needed to further explore this question. Future research should also explore the impact of patient’s and therapist’s gender on processes related to sexual attraction.

It was also found that dimensions of affective communication related to independently assessed patient characteristics, specifically baseline assessments of personality disorder symptoms, attachment variables (Coherence on the AAI), and treatment completion. Negative affect was positively related to Cluster A symptoms but negatively related to Cluster C symptoms. Clinical experience with Cluster A symptomatology suggests that such treatments are characterized by object representations related to hateful persecution (Kernberg, 1975), and the present finding of heightened negative affect as well as Betan et. al’s (2005) finding of therapists’ countertransferentially feeling criticized and mistreated by these patients are consistent with this clinical picture. In contrast, clinical experience with Cluster C symptomatology suggests that aggression is often defensively denied and yet often expressed in passive communications (Kernberg, 1975), and the present finding of low negative affect as well as therapists’ characterizing the treatment as less enlivened is consistent with this clinical picture. Notably, ACQ components were not found to relate to Cluster B symptoms, although it should be noted that all subjects in this study met criteria for BPD, so ceiling effects may obscuring the relationship between these variables. Future research should evaluate Betan et. al’s (2005) finding in a more diverse personality disordered sample.

Negative affect was also negatively related to Coherence on the AAI, and there was a trend towards full range of affect being positively correlated with Coherence. This finding is consistent with previously research demonstrating that the quality and effectiveness of affect regulation in adults was significantly related to the developmental maturity of individuals’ representations of self and significant others (Levy, 2000). In treatments that are predominated by negative affect, such strong emotion may interfere with the patient’s capacity to construct a coherent narrative.

It was also found that therapists reporting less negative affect and a fuller range of emotions related to completion of treatment. This finding is consistent with Coombs and colleagues’ (2002) finding that collaborative emotional exploration was positively related to outcome. This is also consistent with literature suggesting that when patients and therapists each feel involved in a process in which a rich array of feelings are experienced and put into words, each look to continue this work to its completion (Bucci, 1997; Freedman & Berzofsky, 1995).

Lastly, consistent with expectations, TFP therapist more negative affect in the treatment as compared to SPT and DBT. This finding is consistent with the treatment approach’s emphasis on aggression, and deemphasis on the supporting and validating interventions characteristic of SPT and DBT respectively. One unexpected finding was that DBT therapists reported feeling less enlivened in treatment than TFP therapists, who in turn felt less enlivened than SPT therapists (although this finding should be viewed cautiously in light of the small number of subjects in each treatment cell). Given that the focus of DBT is on supporting and bolstering “effective” functioning on the part of the patient, it would reason that DBT therapists would feel less enlivened by the “ineffective” behaviors and emotional displays of patients with BPD during the initial stage of treatment. In contrast, dynamic clinicians, with their focus on identifying and exploring maladaptive object relational patterns within the treatment may feel enlivened by these same “ineffective” behaviors and emotional displays as “grist for the mill”. Future research with larger samples of therapists of varying orientations should further explore this question.

Limitations and Future Directions

With regard to the limitations of this study, as previously noted, despite the strong relationship demonstrated between the ACQ and measures of like constructs (transference, countertransference, and affect regulation), these ratings may not be independent of one another in that all are based on therapist report. It should be noted that studies relying exclusively on self-report measures suffer from the same problem, and yet are ubiquitous in our field (Westen & Weinberger, 2005). A strength of the present study was that the ACQ was related to independent measures of patient functioning (personality disorder symptoms, coherence, RF, treatment retention), although the relationships were of a lesser magnitude. However, such independent measures do not directly assess in-session characteristics and therefore would be expected to relate less strongly to ACQ factors. Future research should further the construct validity of the ACQ through a more stringent comparison to independent measures of in-session affective exchanges, such as observer ratings of transference and countertransference in videotaped sessions.

Further, because the therapists were asked to rate characteristics of their patients, it may be difficult to tease apart patient variables from therapist variables. While the construct of an affective communication was compared to independent measures of patient characteristics, there is no independent assessment of therapist characteristics that might impact findings. As the work of Dozier and colleagues (Dozier et al., 1994; Bernier & Dozier, 2002) suggests, the attachment organization of the patient may interact with the attachment organization of the therapist. Future research should further evaluate the interaction between the patient’s and therapist’s relational style or personality organization on the patient’s functioning or change in treatment. Therapist variables may be particularly relevant in the second sample because each of the therapists in this study had been chosen based on his/her experience in working with patients with BPD. A therapist who finds working with patients with BPD gratifying must presumably be comfortable working with intense affect, both coming from the patient and within oneself. Therefore the therapist’s own level of comfort with intense affect may be an important therapist characteristic to further evaluate.

Therapists in the second sample were asked to make retrospective assessments of their patients, which relies on the memory of the therapists to recall a treatment. However, therapists were provided with a brief demographic description of the patient and also asked to review their notes from the initial stages of treatment. Further, therapists were asked to also rate how difficult it was to remember this patient, and the modal score was “1, very easy” with a strong positive skew in the distribution. Further, the relationship between therapist ratings and independent measures suggests the validity of these ratings.

Another limitation of this study concerns the use of principal component analysis on a small sample to assess the construct of affective communication. There are a number of standards suggested for adequate sample size in principal component analysis, most related to either overall sample size or number of variables in the measure to be factored. By each standard this study meets what would be considered the most minimal requirements (with a sample size of 81, a participant-to-component ratio of 20:1, and a participant-to-variable ratio of 3:1). However, others suggest that adequacy of sample size should be determined by the results of communalities of the components (Field, 2005). As can be seem from the component loadings on rotated component matrix for the ACQ (see Table 1) the communalities tended to be high, and we adopted a conservative standard of including only items with loadings of 0.60 or more in a component, which increases the stability of the component structure (Stevens, 2002). Further, in the present study the strong internal consistency and clear interpretability of the component structure contraindicates the possibility of having arrived at an arbitrary component solution. Nonetheless, future research should test this component structure in a larger sample using confirmatory factor analysis.

It should also be noted that the construct validity of the ACQ was assessed in a personality disordered sample. Future research should evaluate whether the identified convergent relationships replicate in a more neurotic-level patient population, or whether the construct of an affective communication is as relevant to treatment course and outcome in a non-personality disordered patient population.