Emotional Homework : A Systematic Literature Review of Patients ’ Intersession Experiences

Patients’ processing of therapy between sessions can be planned and deliberate (homework) or spontaneous and incidental (intersession experiences). Aiming to inform clinical practice and future research, this review synthesizes empirical findings relating to intersession experiences, including their types, prevalence, and associations with patient characteristics, therapeutic processes, and outcomes. Searches of electronic databases identified 781 journal articles, 14 met the inclusion criteria of investigating adult psychotherapy patients’ unplanned between-session experiences relating to their therapy and therapist, a further 4 were identified through hand searches and contact with authors. All 18 articles included in the review were written in either English or German language, 17 used quantitative and 1 qualitative methods. Their methodological quality was assessed using tools developed for the purpose of this review. Most patients report a range of intersession experiences, including recreating therapeutic dialogue, imagining interactions, images and dreams. Intensity and type were associated with patient personality, diagnosis, phase of therapy, alliance and outcome. Study limitations included small sample sizes, the exploratory nature of some designs, and the limited generalizability of results. Clinical implications include the potential of intersession experiences to provide information about the therapeutic relationship, and their association to treatment outcomes and possible post-therapy gains.

Psychotherapy patients spend only a small proportion of their waking time in face-toface contact with their therapists (typically less than 1%, Hartmann, Orlinsky, Geller & Zeeck, 2003); but important processes happen between sessions (Orlinsky, Heinonen & Hartmann, 2015).Some of these are planned, deliberate, and instrumental, and are usually referred to as 'homework' (Kazantsis & Ronan, 2006); others are spontaneous, sometimes involuntary and often emotionally charged, commonly referred to as 'working through' in the psychodynamic tradition (Owen, Quirk, Hilsenroth & Rodolfa, 2012).The term 'intersession experiences' (ISE) has been used to describe the range of mental representations patients may have about therapy between sessions.These include imagined interactions, fantasies, thoughts, feelings, dreams and images, specifically about the therapy or therapist and have been conceptualized as reflections of in-session processes that patients 'take home' (Orlinsky, Geller, Tarragona & Farber, 1993).
The generic model of psychotherapy (Orlinsky & Howard, 1987) is arguably the most comprehensive transtheoretical framework for integrating psychotherapy research findings.
Originally it accounted for organizational (contract), technical (operations), interpersonal (bond), intrapersonal (self-relatedness) and clinical (in-session impacts) facets of therapeutic process, butbeginning with its 1994 revision by Orlinsky, Grawe and Parksa temporal facet has been included, drawing attention to the function of sequential process patterns, encompassing stages or the whole course of therapies (Orlinsky, Rønnestad & Willutzki, 2004) .The model now acknowledges ISE as an aspect of the therapeutic process (Orlinsky, 2009(Orlinsky, , 2014) ) and considers that they may result in 'micro' outcomes, such as patients' better management of problematic situations they encounter outside of sessions.It is largely unknown how therapy is internally implemented by patients; however it is likely that therapy process and outcome are linked by means of internalization, with ISE integral to this process.Orlinsky et al. (1993) suggest that patient ISE reflect in-session interactions and serve as a vehicle for in-session processes to be transferred to patients' lives outside of therapy.Orlinsky and Geller (1994) argue that representations may reflect the therapeutic relationship, and constitute the "psychological connective tissue between successive therapy sessions" (p.23).In addition, they may influence the course of therapy, having significant impact on therapeutic process.For example, a patient's recreation of the therapeutic dialogue to solve a problem outside of the session may in turn strengthen the therapeutic relationship within sessions, implying that ISE interact with important aspects of the therapy process.Bohart and Wade (2013) provide a narrative summary of studies on the broader concept of clients' activities outside of therapy.While the definition of intersession experience is largely agreed in the literature, the current systematic review of empirical studies to date aims to synthesize the characteristics and correlates of factors of ISE, structured by four questions:

Current Focus and Questions
1. What are the types of ISE and how common are they?2. How are ISE related to patient/therapist characteristics?3. How are ISE associated with therapy stage, length and setting? 4. How are ISE associated with the therapeutic relationship and treatment outcome?
In addition, the quality of studies and instruments was of interest leading to a further question: 5. How dependable are existing measures and studies of ISE?
Past studies have been shaped by the measures they employed.All use retrospective self-report, but direct measures differ in the timeframe they are aimed at: The Therapist Representation Inventory (TRI, Geller, Cooley & Hartley, 1981) explores patients' representations of their therapist at a single point in time, focusing on their complexities, sensory modalities and function for the patient.The Intersession Experience Questionnaire (IEQ, Orlinsky and Lundy, 1986a,b) and its German translation, the Inter-Session-Fragebogen (ISF, Hartmann et al. 2003) are intended for repeated use.They explore between-session experiences over the course of therapy, asking patients about the type, frequency, content, context and emotional consequences of representations, and how much they talked about therapy with others.All three measures examine patient dreams about their therapist.A further instrument, the Disclosures About Therapy Inventory (DATI, Khurgin-Bott & Farber, 2011) is not a direct measure of intersession experience, but explores the perceived impact that patients' disclosures about therapy to significant others have on therapy process.

Method
Three strategies were used in a systematic and comprehensive search of the literature: (a) online searching of electronic databases, (b) checking citations and reference lists, and (b) contacting key authors.Studies were included if they met the following criteria: 1. Participants had accessed psychological therapy.
2. Patient representations of therapy or therapist between sessions were investigated.

Peer reviewed articles or dissertations.
Dissertations were located and considered within the search process; however, only one related article (Zeeck, Hartmann, Balke & Kuhn, 2003) met the criteria.
Studies were excluded if they: 1. Examined populations outside the adult age range (18 +).
2. Examined patients' representations only about themselves or significant others.

Investigated only the neuropsychological processes underpinning therapy.
No restriction on the date of publication was applied; however, database start dates limited the search.

Electronic Search Strategy
Four databases (EMBASE, MEDLINE, PsycINFO, and ASSIA) were searched with the same strategy up to and including 15.04.2014.Three broad concepts -patient, therapy/therapist, and intersession experience -were identified as relevant to the review.Key terms were searched individually, using Boolean operators to group searches within each concept (OR).They were then combined (AND) to produce a total search number for each database.Key terms used were: intersession experience*, intersession process* mental representation* between session experience*, patient* client*, therapist* therapy.Medical Subject Headings (MeSH) related to the key terms were also identified and exploded.
Headings varied however, examples included 'fantasy' as a MeSH for 'mental representation*' (EMBASE), and 'patients/or inpatients/or outpatients/' as a MeSH for 'patient*' (MEDLINE).ASSIA does not have this function and therefore only key terms were used.Key terms were intentionally broad, using known labels for the concept of ISE.They did not focus on specific themes, such as patient characteristics or measures of intersession experience, to ensure that only patient between-session experiences related to therapy were identified.Wzontek, Geller and Farber (1995) explored post-therapy, rather than betweensession, experiences; however, their study was included as its focus on patients' internal representations matched the review questions.A QUORUM diagram (Figure 1) details the search process.

Data abstraction
Authors, participants, methodology, measures, aims and key findings of papers selected for review are summarized in Table 1.Each of the existing tools for assessing the methodological quality of research has its limitations.Whilst they can be beneficial in reducing subjectivity and improving reliability, the validity of the conclusions they lead researchers to draw has been widely criticized (Juni, Witschi, Bloch & Egger, 1999).To assess the quality of quantitative studies, a coding frame was developed using an amalgamation of the most relevant elements of two established quality scales within the literature (Critical Appraisal Skills Programme, University of Oxford, 2004; Newcastle-Ottawa Scale, Wells et al., 2010).This helped guide the development of the quality assessment tool, but allowed for adaptations in line with the design of studies in the review.
The tool contained seven areas of potential bias (definition; participants; assessment; design; results; generalisability; and implications).Sub-questions considered measures, recruitment strategies, statistical reporting and procedure, and aims to reduce bias.The qualitative paper was evaluated based on criteria outlined by Tracy (2010) regarding qualitative best practice.This was adapted in line with the study being reviewed and is consistent with other qualitative assessment criteria (Yardley, 2000).No overall score is provided given that the reduction of a study's quality to a single dichotomous judgment is likely to obscure the important differences between aspects of study design (Cooper, 2010).
The quality of all studies was rated independently by both authors who agreed on 96% of judgments; consensus for the remaining 4% was achieved through discussion.Table 2 shows quality results for the quantitative studies, Table 3 for the qualitative study.To explore patient representations of therapists through their dreams.
 No significant difference in frequency, mood of dream or success of therapy between those in therapy and those whose therapy has terminated. Within dream content: 13.4% reported aggressive interactions between patient/therapist, 16.9% reported friendly interactions between patient/therapist, 7.5% reported sexual interactions between patient/therapist.Studenborgen completed after each session.

Differences in intersession experience
How is intersession experience related to therapy phase, outcome and personality?
 No differences between BP and NP in intensity of intersession experience overall. During phase two (weeks 3-6) BP had a higher intensity of intersession experience than NP** (t=2.77) and more negative ISE in all three phases of therapy.

Key Findings
Frequency and type.All studies noted that clients reported a range of ISE.Knox, Goldberg, Woodhouse and Hill (1999) noted that some participants discussed the idea of ISE as being 'between-session mini-sessions'.Other types of ISE include invoking a literal recreation of a therapy conversation to cope with an anxiety provoking situation, experiencing dreams about the therapist, and talking through what the therapist may say in future sessions to help manage distress.This is consistent with quantitative studies which report recreating the therapeutic dialogue as the most common intersession experience (Hartmann, Orlinsky, Weber, Sandholz & Zeeck, 2010;Hartmann et al., 2011;Rosenzweig et al., 1996;Wzontek et al., 1995), in addition to imagined interactions, images and dreams (Geller et al., 1981) and discussing therapy experiences with significant others (Khurgin-Bott & Farber, 2011).
Furthermore, factors such as recreating the therapeutic dialogue and applying therapy were noted cross-culturally and related to positive emotions.The study by Rohde, Geller and Farber (1992) focused specifically on dreams and found that many related to feeling separated or rejected, seduced or antagonized, protective or responsive and receiving praise from the therapist.The only data on pervasiveness of ISE are provided by Orlinsky et al. (1993), who found that more than 90% of their respondents reported having them.Specifically for dreams about the therapist, Geller et al. (1981) found an incidence of 33% in their sample.This finding has not been replicated in a recent study by Hill, et al. (2015) who found only 3% of clients reported dreams about their therapy or therapist1 .
Associations with patient characteristics.
The main patient characteristics explored in relation to ISE were gender and personality types.A range of demographic information was collected within all studies; however, this varied greatly and was generally not incorporated into the analyses.
Gender was not found to be associated with frequency of ISE; however, Farber and Geller (1994) noted that women report holding on to ISE for longer than men.A number of personality characteristics were associated with ISE.Bender et al. (2003) categorized personality types, reporting that patients with schizotypal traits had the most ISE throughout all stages of therapy; both positive and negative in tone.Patients categorized as high in borderline personality traits had the most difficulties in recreating images of their therapist.
This is similar to findings by Zeeck and Hartmann (2005), who reported patients with borderline traits to have greater negative ISE and difficulties in recreating positive therapeutic dialogue in the time between sessions.In addition, those with neurotic traits had significantly greater frequency and intensity of experience.This may relate to neurotic patients' high levels of anxiety meaning they think through the therapy much more between sessions, and borderline patients' fluctuations of positive regard for the therapist, resulting in greater negative ISE. Anorexic patients who recreated the therapeutic dialogue with negative emotion were less likely to report positive outcome (Hartmann et al., 2010), which may be attributable to personality factors.
One study (Brenner, 1992 as cited in Orlinsky et al. 1993) examined the link with patients' level of functioning and found that poorer functioning was associated with more frequent ISE.Although some studies reported therapist characteristics, these were commonly limited to gender and experience, and not used within analyses (Hartmann et al., 2010;Owen et al., 2012;Zeeck et al., 2005;Wzontek et al., 1995).
Associations with therapy stage, length and setting.
follow up.Wzontek et al. (1995) and Geller et al. (1993) noted that patients continue to experience a range of spontaneous thoughts, feelings and images about their therapist in the years following completion; however, changes were not tracked over time.When tracked over therapy (Hartmann et al., 2010;Hartmann et al., 2011;Owen et al., 2012;Rosenzweig et al., 1996;Zeeck, Hartmann & Orlinsky, 2006), frequency of ISE significantly increased.Zeeck et al. (2003) assessed variation in intersession experience across the course of individual treatment finding that while there was little variation in intensity of ISE over time; problem solving ISE increased within the early stages of therapy only and the frequency of ISE was significantly related to emotionally charged contexts across all stages of therapy.Similar findings were noted by Rosenzweig et al. (1996) who sampled psychotherapists in therapy.
They reported that positive emotions evoked about the therapy/therapist increased over the course of therapy, and recreating the therapeutic dialogue between sessions was associated with reductions in patient distress in later stages of therapy.In a longitudinal study across different therapy settings, Hartmann et al. (2003) found that 'applying therapy' during intersession intervals was associated with positive emotions.Zeeck et al. (2003) note that more positive than negative emotions are associated with ISE, with positive emotions found to increase towards the end stages of therapy.Hartmann et al. (2003) also distinguished differences between settings, with ISE being more pronounced in inpatient settings.

Therapeutic relationship and treatment outcome.
The relationship between intersession experience and outcome is a theme within seven studies, measured indirectly through the therapeutic alliance, by self-report of progress, psychometric assessment, or observable measures (such as weight gain in eating disorder populations).Continuation of the therapeutic dialogue correlates significantly with patient perceptions of benefit during therapy (Geller et al, 1982;Geller et al, 1993) and after termination (Wzontek et al., 1995).In addition, type and frequency of ISE was associated with significant weight gain in patients with anorexia (Zeeck et al., 2005).ISE that have negative emotions are associated with poor outcome for bulimia patients in the initial and mid stage of therapy (Hartmann et al., 2010).Owen et al. (2012) found that the therapeutic alliance, as measured by the Working Alliance Inventory (Hatcher & Gillaspy, 2006) was positively correlated with the quantity of ISE.Hartmann et al. (2011) also found that ISE associated with positive emotions showed a strong relationship with in-session alliance, whereas negative emotions showed a strong inverse relationship.The qualitative study by Knox et al. (1999) reported that clients liked having ISE, and felt they influenced the course of therapy and significantly strengthened the therapeutic relationship.

Methodological Characteristics.
Of the 18 studies within the review, 17 used quantitative methods.Only one qualitative study (Knox et al., 1999) met the criteria for the review.One other study (Arnd-Caddigan, 2012) was located; however, this was excluded due to focusing on therapist ISE only.

Quantitative studies (table 2).
The quality of studies varied; selection and recruitment of participants and sample representativeness ranged from good (recruiting patients from a range of settings) to moderate (only recruiting psychotherapists that were in therapy).Some samples were reported as being "highly ambivalent about being involved in treatment" (Zeeck et al., 2005, p.245) and therefore may have felt pressured to engage.Wzontek et al. (1995) reported postal mailing, potentially resulting in self-selection bias, although later stated that some recruitment had been through 'personal networking'.Quantitative study sample sizes ranged from 43-769 and limited demographic information was generally reported.Most studies were retrospective, although a number of studies did track changes over therapy (Rosenzweig et al., 1996;al., 2010).All studies used self-report methods.
All studies used one of four formal measures: the TRI, IEQ, ISF or DATI.The internal consistency of all is reported to be within the acceptable/good range (Geller et al., 1993;Hartmann, Orlinsky & Zeeck, 2011;Orlinsky et al., 1993;Rosenzweig, Farber & Geller, 1996, Khurgin-Bott & Farber, 2011).However, the TRI was validated using a sample of psychotherapists in therapy, potentially limiting its generalizability (Geller et al., 1981;Orlinsky et al., 1993).Hartmann et al. (2003) explored the factor structure of the ISF comparing 3,778 inter-session intervals in 249 therapy episodes across three therapy settings.
The ISF was stable across therapy and therapy settings with high internal consistency reported for most factors.Similarities across all direct measures of intersession experience (ISF, TRI, IEQ) were reported with some direct mapping of items across the questionnaires.The DATI (Khurgin-Bott & Farber, 2011), measures the impact the discussion about therapy experiences between sessions has on perception of therapy.Whilst good internal consistency is reported, this measure is yet to be assessed across therapy settings and length.
Several studies considered a range of variables within the analyses, yet only small sample sizes were recruited (Owen et al., 2012;Rosenzweig et al., 1996;Rohde et al., 1992) .
Only three studies made corrections for multiple comparisons, thereby increasing the chance of Type I errors.However, Hartmann et al. (2010), Zeeck et al. (2005) and Zeeck et al. (2006) explicitly state that their study design was exploratory.
Unless reporting correlations, all studies did not clearly state effect sizes.Whilst a range of studies highlighted strong associations between ISE and patient/therapy characteristics, correlation and regression do not identify causation.In addition, significant findings may be attributable to confounding variables that were not controlled for, such as therapist factors or the events outside of therapy.
Qualitative studies (table 3).Knox et al. (1999) was the only qualitative study included within this review.Overall the study's methodology either fully (yes) or partially (moderate) achieved the quality assessment criteria (Table 3).A strong rationale was provided and the sample representation was good, recruiting therapists to access a range of patients.The study provided a good methodological description, detailing data collection, overall research process, transcription and analysis.To minimize bias, researchers recorded their expectations of the results prior to data collection; however, only researcher interpretations were reported rather than direct quotations.The study provided meaningful coherence in achieving its aims; however, some ethical safeguards, such as debriefing participants, were not reported.

Discussion
Patients' therapy-related experiences between sessions have been the subject of empirical study for over 30 years.Our review surveyed their parameters and correlates as reported in the literature.ISE have been examined mainly in relation to psychodynamic therapies, but have also been established following cognitive-behavioral interventions (Owen et al., 2012).They are likely to be common to all forms of therapy, though this has yet to be documented.
The only available data on prevalence suggest that ISE are near ubiquitous.None of the qualitative studies relying on postal mailing of questionnaires reported return rates; however, it seems likely that at least some of their non-responders had no ISE to relate.
Patients not reporting ISE, even if they did constitute only 10% of the population, would be of particular interest in further studies.
There are differences in intersession experience based on individual characteristics.
Some of these, such as the highly negative ISE of borderline patients, could impact on their engagement within sessions (Zeeck et al., 2005).Exploring these occurrences, alongside the traditional focus on in-session process, may assist clinicians in addressing common obstacles to therapy.
The variety and frequency of ISE may relate to the varieties of therapy and of the therapists delivering the sessions.These factors were neglected within all studies.Therapists' ISE have been documented (Schröder, Wiseman & Orlinsky, 2009) and, given that ISE are conceptualized as relational, there may well be an interaction effect between therapist and patient experiences.Furthermore, ISE may reflect what happens inside sessions (Orlinsky et al., 1993), yet so far comparisons have not been made between intersession experience and insession content.
Measures of intersession experience have developed from time-and labour-intensive composite questionnaires to shorter and simplified scales that lend themselves to repeated measurement, sacrificing some complexity but gaining functionality.The existing measureshave been reported as generally reliable and valid, with most studies using either the TRI or the IEQ.However, the generalizability of the TRI may be somewhat limited due to the use of psychotherapist samples to validate the tool.Whilst the one qualitative study in this review (Knox et al., 1999) corroborated the quantitative findings, it also provided a richer understanding of how participants made sense of their ISE by capturing their original perspectives rather than relying on pre-existing categories.Future qualitative research may provide a greater understanding of the role of ISE in therapy process and outcome.
Generally, ISE appear to increase over the course of therapy in frequency, if not intensity, and spontaneous representations of therapy and therapist persist after termination.
Given their association to the therapeutic relationship, ISE may be conceptualized as a mechanism by which the therapeutic alliance is continued or reactivated in the patients' life outside of (and after) therapy, reflecting the strength and quality of the patient/therapist bond.
The emotional quality associated with ISE emerges as significant for both process and outcome.The amount of positive or negative ISE patients have over the course of therapy may be a reflection of how well the therapeutic relationship is developing (Hartmann et al., 2010;Hartmann et al., 2011;Khurgin-Bott & Farber, 2011;Owen et al., 2012;Rosenzweig et al., 1996).Negative experiences would be associated with strains in the alliance, indicating problematic process potentially leading to poor outcomes, while positive ISE would suggest a robust alliance, explaining why they are associated with therapeutic gains.If ISE can be considered as either helpful or detrimental to process and potential therapy outcome, depending on whether they are associated with positive or negative emotions, then their exploration in the session would be a source of valuable feedback.In addition, positive ISE may influence whether patients continue to apply and retain skills after therapy completion, thereby predicting longer term treatment effects.
In summary, we conclude from our review that ISE have the potential to be an important source of information about therapy process across different types and modalities of psychotherapy and may influence immediate and long term outcomes.Future research would usefully: 1. Investigate types and frequency of ISE in theoretical orientations other than psychodynamic and cognitive-behavioral.
2. Explore the relationship of ISE to preceding session content.
3. Explore the influence of ISE on subsequent in-session process.
4. Employ qualitative and mixed methods designs to explore personal meanings that patients and therapists attribute to ISE.
5. Study the interaction of patients' and therapists' ISE.

Limitations of the Review
The inclusion of only one qualitative paper may have impacted on the synthesis of findings due to its significantly different methodology and the lack of comparison with other qualitative studies.Nevertheless, quantitative findings were supported and the exclusion of this study would have ignored key findings.Whilst search terms were limited to broad key phrases related to intersession experience to exclude large numbers of unrelated studies, the terminology may have selectively privileged psychodynamic literature.

Table 1 :
Studies examining the properties of patient intersession experiences.
 Over 90% reported having ISE, mostly presession. Good internal consistency of the TRI (ranging from α=.70 to .86) and the IEQ (ranging from α = .57to .81). Function of representation: a source of emotional support, to master and manage conflict exposed during therapy.

Table 3 .
Qualitative methodological quality