Written Emotional Expression and Well-Being

The Australasian Journal of Disaster
and Trauma Studies
ISSN:  1174-4707
Volume : 2002-1

Written Emotional Expression and Well-Being:
Result From a Home-Based Study

David Sheffield Psychology Department Staffordshire University Stoke-on-Trent ST4 2DE, England, UK. Phone: 01782 294 679 Fax: 01782 745506 Email: d.sheffield@staffs.ac.uk	Elaine Duncan Department of Psychology, Glasgow Caledonian University, Cowcaddens Road, Glasgow, G4 0BA, Scotland, UK. Phone: 331 3476 FAX: 331 3636 Email: edu@gcal.ac.uk
Karen Thomson Department of Psychology, Glasgow Caledonian University, Cowcaddens Road, Glasgow, G4 0BA, Scotland, UK. Phone: 331 3476 FAX: 331 3636	Sarbjit S. Johal Department of Psychology, University of Bath, Claverton Down, Bath, BA2 3 AY, England, UK.

Abstract

This study examined whether written emotional expression of personal and emotional events in a home-based setting might result in improved psychological and physical well-being. Following completion of a number of health and psychological well-being questionnaires, forty-seven men and ninety-nine women were assigned to one of three conditions. Participants in the emotional expression writing group were asked to write about previously undisclosed personal and emotional events for 3 days; participants in the unemotive writing group were asked to write about pre-assigned superficial topics for 3 days; and participants in the control group did not write. Participants were asked to complete the initial questionnaires three, seven and thirty weeks later. The emotional expression group experienced an increase in physical symptoms (p =.04) and self-reported number of days taken off college due to illness (p =.03) at three weeks, and less anxiety and insomnia (p =.04) at 30 weeks. Written emotional expression was associated with short-term detrimental physical health effects and less substantial long-term psychological health benefits, cautioning against the adoption of written emotional expression alone in the home environment.

Written Emotional Expression and Well-Being:
Result From a Home-Based Study

Preamble

Studies evaluating disclosure of past traumatic events to a significant other or professional are well documented (e.g., Shalev, Bonne, & Eth, 1996; Smith & Glass, 1977). More recently, a number of studies have focused on the efficacy of written emotional expression (e.g., Pennebaker & Susman, 1988; Smyth, 1998). These studies usually involve a progressive writing task across three to five days. Variations exist in the method of inducing the writings but the studies all have a common goal - to test the power of disinhibition. It has been found that college students randomly assigned to write about personally upsetting topics in an emotional manner for three to five consecutive days subsequently evidenced fewer physician visits than a control group who wrote about superficial topics (Greenberg & Stone, 1992; Pennebaker & Beall 1986; Pennebaker, Colder & Sharp, 1990). For example, Pennebaker and Beall (1986) found that writing about the emotions surrounding a traumatic experience resulted in health benefits whereas simple unemotional disclosure did not. Studies using variations of this writing technique also found enhanced immune functioning (e.g. Pennebaker, Kiecolt-Glaser & Glaser 1988). Written emotional expression has also produced clinically relevant changes in health status in sufferers of moderate to severe asthma and those with rheumatoid arthritis (Smyth, Stone, Hurewitz, & Kaell, 1999). In addition, targeting participants who have experienced a recent relationship break-up and giving them the opportunity to write about their deepest thoughts and feelings improved upper respiratory levels and lowered levels of fatigue and tension in comparison to controls (Greenberg & Lepore, 1999). Prior to these latest studies positive outcomes were confirmed in a meta-analytic review of the written emotional expression literature. Smyth (1998) found that health was enhanced across a number of different types of outcome variables. These included physical health (self-reported, which may or may not have been substantiated by objective clinic visit records), self-reported psychological well-being, physiological functioning (immune system functioning) and academic or employment functioning.

Smyth (1998) also noted that studies differed in the type and gender of samples used, the nature of instructions priming expressive writing, the setting within which the expressive writing intervention took place, duration of the writing manipulation or intervention and the publication status of the study itself. Across all these moderating variables Smyth (1998) found a number of surprising features. First, student samples tended to produce the higher effect sizes in psychological well-being than non-student samples; however, there was no difference in the severity of the traumatic experiences that were written about. Second, it appeared that men benefited more from writing about traumatic events than women. Third, the publication status of the study was not biased in favor of positive disclosure effects. On the contrary, some of the unpublished manuscripts had the greatest effect sizes.

In spite of these methodological differences, Smyth (1998) concluded that written emotional expression might provide a viable alternative to interpersonal disclosure of traumatic events common to many psychological therapies. However, most studies have examined the effects of written emotional expression in laboratory settings and, thus, necessarily involve some interpersonal exchange (or "therapist"). In 1996 however, Davidson and Pennebaker did utilize a psychoeducational intervention of relaxation and introspective writing outside the laboratory using healthy participants. They found that the treatment group reported fewer visits to doctors; self-reported social closeness was the only factor contributing to health status. Since this intervention included both written emotional expression and relaxation, it is unclear why health benefits were seen. The purpose of the current study was to examine whether written emotional expression alone, in the home environment, would result in health benefits. Clearly, if unequivocal health benefits are seen in this study, one might advocate written emotional expression in an unstructured home environment.

The present study sought to replicate the findings of previous written disclosure studies (e.g., Pennebaker & Beall, 1986) using a home-based intervention design. The study had three main objectives. First, to assess whether a home-based, 'self-help' expressive writing package might have comparable, beneficial health effects to laboratory- based designs. Second, to examine whether changes in psychological well-being (use of coping strategies, appraisal of daily uplifts and hassles, social support levels) or mood differed by writing group over time. We expected participants in the written emotional expression (trauma writing) group to experience fewer negative emotions, make more use of problem-solving coping strategies (e.g., active coping, positive reinterpretation, growth and acceptance) and support resources, report fewer hassles, and make less use of mental disengagement and denial as coping strategies. Third, to examine whether changes in health variables were related to the emotionality of the written disclosures or to changes in mood during the disclosure sessions. In addition, we had wanted to examine whether psychological well-being or mood changes mediated any health benefits experienced by the trauma writing group, but the small number of participants who completed each of the follow up phases prevented us performing the appropriate analyses.

Method

Design
In contrast to other written disclosure studies, the present study did not take place within a controlled laboratory setting, but was designed as a home-based intervention. The study was carried out in three phases: initial recruitment, where all potential participants completed a battery of questionnaires; an intervention phase consisting of three 10 minute writing sessions; and a follow-up phase consisting of three follow-up periods.

Participants
One hundred and forty-six psychology students completed initial questionnaire packs. Participants were unpaid volunteers and received no course credit for participation. Participants were assured of total anonymity through the adoption of personal identification numbers and the employment of two independent judges from another institution to analyze the written disclosures.

Measures
Six measures were employed to assess changes in health; measures were recorded for the previous month to maximize variance, with the exception of the number of symptoms, which were assessed over the previous 14 days to aid recall. Three measures of psychological well-being were derived from the first three 7-item sub-scales of Goldberg's (1978) General Health Questionnaire (GHQ). These subscales tap somatic symptoms, anxiety / insomnia, and social dysfunction; the severe depression sub-scale was excluded for ethical reasons. Physical symptoms (e.g., chest pain, earache, flu) occurring in the last 14 days were assessed using a 17-item checklist, developed for use in the Whitehall II study of civil servants (Marmot, Davey Smith, Stansfeld, Patel, North, Head, White, Brunner, and Fenney, 1991). Positive responses to these items were summed to create a symptom score. Finally, participants were asked to report the number of days they were absent from college due to illness in the previous month, and the number of times they had visited their local physician in the previous month.

In addition, we examined changes in psychological variables that might act as mediators for changes in health status; these variables were assumed (explicitly or implicitly) to assess psychological states over the previous week. Social support was assessed with Blumenthal, Burg, Barefoot, Williams, Haney, & Zimet's (1987) 12-item Perceived Social Support Scale (Cronbach alpha =.99). The Students Assessment of Daily Experience Questionnaire (Evans, Pitts, & Smith, 1988), a 27 item check-list, was administered to determine how frequently participants experience a number of desirable and undesirable events; the number of items were summed to create two scores. The Modified COPE (Carver, Sheier, & Weintraub, 1989) was completed to examine changes in the strategies used to cope with stress. This 24-item questionnaire has subscales determining the current use of twelve different coping styles. To decrease our chances of type I errors associated with multiple tests we chose to focus attention on the six problem-solving coping strategies namely: active coping (Cronbach alpha =.92), seeking social support (Cronbach alpha =.92), positive reinterpretation and growth (Cronbach alpha =.92), acceptance (Cronbach alpha =.92), mental disengagement (Cronbach alpha =.99), and denial (Cronbach alpha =.92). Scoring is accomplished by summing the items of the 4-point Likert scales that comprise each sub-scale. Finally, positive and negative mood was assessed by a 30-item Mood Questionnaire adapted from Zevon and Tellegen's (1982) questionnaire. Scoring is accomplished by summing points on a 5-point Likert scale for the 15 items pertaining to each mood dimension (Cronbach's alpha for each scale was .99). Participants were requested to record their mood over the past week when completing the initial and final questionnaire packs and for each day of the written disclosure intervention stages.

Procedure
In the initial session, participants were approached during seminar classes and informed that we wished to recruit participants for a study whose purpose was to "examine the effects of writing about past events." Non-participants were asked to withdraw from the room for 20 minutes to allow participants to complete the questionnaire packs in privacy. Following completion of the initial questionnaires, a researcher, blinded to randomization, distributed envelopes containing one of three sets of instructions depending on the experimental condition.

Participants assigned to the control group were given the following instructions: "Four days after you have completed the questionnaires in the seminar class, we would like you to complete the brief questionnaires included in the envelope. The next day we would like you to return these questionnaires". Four days after the initial seminar session these participants completed the set of mood, hassles and uplifts, and coping questionnaires.

Participants in the unemotive writing group were asked to write about pre-assigned superficial topics for 10 minutes for 3 consecutive days, completing the mood questionnaire each day. Participants in this condition were given the following instructions: "During today's writing session, we would like you to spend about 10 minutes describing in detail (Day 1) what you have done since you woke up this morning" (Day 2), the most recent social event you went to, (Day 3) your plans for the remainder of the day. It is important that you describe things exactly as they occurred. Do not mention your own emotions, feelings or opinions. Your description should be as objective as possible. After you have finished writing please complete the questionnaire. The next day we would like you to return your writing and the questionnaires". Following the third writing session, four days after the initial seminar session, they also completed the set of mood, hassles and uplifts, and coping questionnaires.

Participants in the emotive writing group were asked to write about previously undisclosed traumatic events for 3 consecutive days and complete the mood questionnaire. These participants were made aware, in writing, that Student Counseling Services were on stand-by should they need more support or advice. On the first day, participants in this condition were instructed as follows: "During today's writing session, we would like you to spend about 10 minutes writing about a traumatic or upsetting experience that has happened to you. The important thing is that you write about your deepest thoughts and feelings. Ideally, whatever you write about should deal with an event or experience that you have not talked with others about in detail. After you have finished writing please complete the questionnaire." On the second and third days, participants in this condition were instructed as follows: "During today's writing session, we would like you to spend about 10 minutes describing in detail a traumatic or upsetting experience that has happened to you. You may write about the same event as yesterday or a different one. The important thing is that you write about your deepest thoughts and feelings. Ideally, whatever you write about should deal with an event or experience that you have not talked with others about in detail. After you have finished writing please complete the questionnaire". Following the third writing session, four days after the initial seminar session, they also completed the set of mood, hassles and uplifts, and coping questionnaires.

Follow-up data was collected at three-, seven- and thirty-week intervals. Initial recruitment questionnaire packs were again distributed during seminars at week 3 and 7 intervals. At 30-week follow-up, questionnaires were mailed to all potential participants.

Statistical analysis
Continuous variables were described with means ± standard errors, and categorical variables were described with frequencies and percentages. To increase power to detect effects, participants were included in analyses if they completed the initial questionnaires, the intervention and at least one set of follow-up questionnaires (including immediately post-writing task). Attention initially focused on various selection biases: whether there were any differences in initial measures between those included in any analyses (i.e., those participants completing at least one follow-up) and those not included; whether participants who completed each of the three follow up stages differed from those who did not; and whether there were differences in initial psychological and well-being measures between the groups. ANOVAs, in the case of continuous variables, and Fisher's exact tests, in the case of categorical variables, were used for this purpose. Differences in health status and potential psychological mediators between the groups at each of the follow-up sessions were then examined using univariate ANCOVAs (initial values were entered as covariates). Finally, the relationships between the emotionality of the written disclosure and changes in positive mood in response to writing, and to health status were examined using bivariate correlations. In all tests, the criterion for statistical significance was two-tailed p<.05. Since we wanted to maximize our chances of finding effects, we chose not to use any correction factors (e.g. Bonferroni), although many tests were performed (6 health status measures and 11 psychological mediators at each of the follow-up stages) and the likelihood of Type I errors was increased (see, Greenwald, Gonzalez, Harris, & Guthrie, 1996).

Results

Demographics
One hundred and forty-six psychology students, 47 men (32%) and 99 women (68%), completed the initial questionnaire package. Participants were aged 17 to 55 years (mean.S.E. = 22.94..66 years). Seventy-three students (50%), 18 men and 55 women, agreed to participate in the intervention; the mean.S.E. age of this sample was 23.53.1.00 years. Forty-six participants completed the 3-week follow-up, 37 completed 7-week follow up and 30 completed the 30-week follow up. However, only 10 participants completed the intervention and all three phases of follow-up. There no differences in age or sex distribution in the groups at any stage of the study (p>.10).

Table 1. Number of Participants Completing each Phase of the Study.

Selection Biases
The possibility that selection biases were operating was examined with three sets of analyses. First, we examined whether there were any differences in initial measures between those included in any analyses (i.e., those participants completing at least one follow-up) and those not included. Participants were more likely to turn to religion as a coping strategy and less likely to use alcohol or other drug-use as a coping strategy (F = 12.19, p<.01) than non-participants, suggesting that participants may have been more conscientious than non-participants; no other differences emerged (all p> .10).

Second, we compared participants who completed each of the three follow up stages with those who did not, to examine whether participants who dropped out of each group had different initial demographic, psychological or well-being scores. ANOVA revealed that only one interaction term (condition (control, write unemotive, write emotive) x completed follow up (yes, no)) was statistically significant. At three-week follow up, those who dropped out of the unemotive writing group were more likely to have visited a doctor at the initial stage (F = 5.75, p<.01). Given that 17 interaction terms were computed for each follow up stage, one statistically significant term would be no more than we might expect by chance. Thus, we concluded there was little difference between those participants in each group who completed follow-up at any stage and those who did not.

Finally, differences in initial psychological and well-being measures between the groups were examined. A number of differences between the groups approached or reached the criterion for statistical significance (see table 2). Participants in the unemotive writing group appeared to have poorer physical health at the onset of the study: they reported more visits to the doctor, and a tendency to report more days off college due to illness and more physical symptoms. In addition, participants in the unemotive writing group reported greater positive affect. Accordingly, initial measures were used as covariates when we examined differences between conditions at follow up.

Table 2. Psychological and Health Measures at Initial Stage by Writing Condition.

 

Measure	Control	Unemotive writing	Emotive writing	p-value
N	30	18	25
Number of visits to Doctor	.43±.13	1.06±.40	.20±.08	.03
Days off due to illness	1.18±.25	2.00±.52	.76±.26	.03
Physical symptoms	5.07±.35	6.72±.67	5.20±.49	.07
Somatic symptoms	5.84±.57	7.94±1.10	8.12±1.05	.05
Positive Affect	11.89±.43	12.83±.47	9.60±.54	.002

Disclosure characteristics
Two independent judges (blinded to the experimental condition) rated each essay in terms of how personal and emotional they were (7-points scales, 7=very personal/emotional). Interjudge correlations across essays averaged .83 for personal scale and .86 for emotional scale. Participants in the emotive writing group wrote more emotional and personal essays on all three days compared to participants in the unemotive writing group (all p<.001). In addition, the judges coded the essays into categories of event. Participants in the emotive writing group disclosed highly personal and upsetting experiences, including: coming to college (18%); death of relative/friend (17%); parental troubles including divorce, family quarrels and violence (14%); boyfriend/girlfriend conflict (12%); injury/illness through drug/alcohol abuse or other causes (9%); financial concerns (3%); sexual abuse (2%) and others (25%).

Participants in the writing groups completed five 5 point scales measuring: how much they thought about what they wrote, how much they talked about what they wrote, how much impact it had, how long its effects might last, and how valuable/meaningful the writing sessions were. Participants in the emotive writing group reported they had thought more about what they wrote (F = 8.45, p<.01), and had a tendency to find the sessions more meaningful (F = 3.55, p=.07) than participants in the unemotive writing group. Finally, participants completed a mood questionnaire every day. Participants in the emotive group reported fewer positive moods on the first writing day (F = 8.85, p<.01), even after adjusting for mood at the initial stage (F = 5.36, p<.05). Table 3 displays the change in positive mood during the writing period. No differences in negative mood were observed (p>.10).

Table 3. Positive mood (Mean±S.E.) by condition during writing days controlling for initial positive mood score.

Follow-up
At the end of the third day of the intervention period all participants (n=57) completed questionnaires assessing positive and negative moods, hassles and uplifts, and coping strategies. Of the 17 ANCOVAs calculated, only one achieved the criterion for statistical significance: participants in the unemotive writing (3.98±.26) were more likely to use humor as a strategy to cope with stressors (F = 5.51, p=.006) than participants in the emotive (2.98±.35) or control groups (2.93±.25).

After three weeks, 46 participants completed the full questionnaire pack. ANCOVA revealed differences in the number of days taken off college due to illness in the past month (F = 3.82, p=.03): the emotive writing group reported being absent from college for a full day more than the other two groups (see table 4). Further, as table 4 shows, the emotive writing group reported more physical symptoms than the other two groups (F = 3.43, p=.04). No other well-being measure, and none of the psychological measures, varied by condition (p>.10).

Table 4. Health Measures at Three-week Follow-up by Writing Condition.

At seven-week follow up (n=37), no measure of well-being varied by condition (p>.10). However, participants in the emotive writing group reported using positive reinterpretation and growth as a way of coping less frequently (3.29±.42) than participants in the unemotive writing (4.80±.37) or control groups (4.06±.40; F = 4.41, p=.02).

At thirty-week follow up (n=30), participants in the emotive group reported less anxiety and insomnia than participants in the unemotive writing or control groups (F = 3.68, p=.04) (see table 5).

Table 5. Health and Psychological Measures at Thirty-week Follow-up by Writing Condition.

Thus, our data suggest that individuals in the emotive group experienced a short-term drop in physical well-being as indexed by physical symptoms and self-reported number of days taken off college due to illness at three weeks. In contrast, the same participants appeared to have improved psychological well-being in the long-term; they reported less anxiety at thirty weeks.

Health status in the emotive group
Finally, we examined whether the nature of the written disclosure (mean emotional rating of disclosures) or changes in positive mood in response to writing (positive mood on the first day of writing minus initial mood score) were related to health status. Specifically, we examined whether responses to the writing session predicted the number of days taken off college due to illness in the past month, the number of physical symptoms at three-week follow-up, and anxiety/insomnia at 30 weeks. This was done particularly in view of the fact that the emotive writing group reported different levels of these measures at these times compared to the other two groups. Six correlation co- efficients were computed to examine these relationships; one reached the criteria for statistical significance. Ratings of emotional content were positively related to the number of physical symptoms at three-week follow-up (r = .49, p <.05): participants who used many emotional phrases in their disclosures had more illness symptoms at three-week follow-up.

Discussion

Participants in the emotive group wrote about highly personal and upsetting events. The events that participants disclosed were similar in nature to those in past studies (e.g., Pennebaker & Beall, 1986), and included concerns about coming to college, death of a relative or friend, parental troubles, boyfriend or girlfriend conflict, and injury or illness. The independent judges rated the essays written by participants in the emotive writing group as more emotional and more personal compared to those of the unemotive writing group. Further, participants in the emotive writing group reported they had thought more about what they wrote, and had a tendency to find the writing sessions more meaningful than participants in the unemotive writing group. Participants in the emotive group also reported reduced positive mood on the first writing day relative to participants in the unemotive writing group (cf. Francis & Pennebaker, 1993). Although there were no apparent shifts in negative mood, the finding that positive mood was significantly affected points to some level of distress being induced by the writing procedure, a common finding within written disclosure studies (Smyth, 1998). However, in contrast to most previous studies, no differences in negative mood were observed between the writing groups. This may suggest that the disclosures were less traumatic or distressing than those of previous studies, although most of our other data does not support this conclusion.

At three-week follow-up the emotive writing group reported more physical symptoms and more days absent from college due to illness, suggesting they had poorer physical health than the other two groups. Since we asked participants to report symptoms and absenteeism during the previous month, it is possible that these differences occurred before or at the time of writing. However, the different patterning of effects we observed from the initial questionnaires suggest that the differences we observed at three-week follow-up were not due to the poor health of the emotive writing group. Some studies (e.g., Greenberg & Stone, 1992) have reported an initial increase in physical symptoms in their trauma groups, whereas others have found no differences (e.g., Pennebaker & Beall, 1986). However, it is difficult to compare these data with past studies since few studies have had their short term follow up so soon after the intervention (e.g., Francis & Pennebaker, 1992). One exception is Gidron, Peri, Connolly & Shalev's (1996) study of post traumatic stress disorder (PTSD) patients. They found that patients reported significant increases in visits to health services and intrusive symptoms five weeks after written emotional expression compared to matched controls. These data suggest the short-term distress created by writing about traumatic events can result in decreased physical well-being within the first month of writing. Changes in immune function might mediate these effects (Esterling, Antoni, Kumar & Schneiderman, 1990; Pennebaker, et al., 1988). Numerous diary studies have found that minor stressors may precede increases in physical symptoms (Sheffield, McVey, & Carroll, 1996), especially symptoms of upper respiratory infection (e.g., Evans, et al., 1988). Many of the symptoms we indexed in this study were respiratory symptoms and the differences we observed in this might be attributable to those minor symptoms (as opposed to mental health symptoms). Further, participants in the emotive writing group who used many emotional phrases in their disclosures had more illness symptoms and days off due to illness at three-week follow-up. Participants in the emotive writing group reported thinking about what they had written about more often than participants in the unemotive writing group at the end of the writing period. Therefore it is possible that participants continued to think about their traumatic disclosures for some time after they wrote about them. Persistent rumination in the hope of finding meaning in the situation can in itself become maladaptive and disruptive to daily plans and functioning (Pennebaker & Suedfeld, 1997). Therefore, it is possible that the processes involved in organizing these memories, for example, assimilation, required physiological effort and compromised participants' well-being in the short-term (Greenberg, 1995).

At seven-week follow up, no measure of well-being varied by condition and at thirty-week follow up, only one measure of well-being varied by condition: participants in the emotive writing group reported less anxiety and insomnia than participants in the other groups. Participants in the emotive group also showed a tendency to report more positive affect at 30 weeks than participants in the other two groups. Thus, we found less evidence of an overall health benefit than previous studies and the improvements that emerged were restricted to psychological well-being. Failure to find overall uniform health benefits is not unusual. Greenberg & Stone (1992) found no significant differences over time between experimentally assigned disclosed-trauma groups and undisclosed trauma groups and control participants on measures of overall visits to medical professionals for illness and self-reported physical symptoms. Post-hoc analyses revealed that beneficial effects of written disclosure on health occurred for the participants who rated their trauma as severe compared with participants who rated their trauma as less severe. Sub-group analyses of our data by gender, emotionality of essay and mood changes during the writing period revealed no differences, however, cell sizes were small. Smyth's (1998) meta-analysis provides a plausible explanation for our findings. In tests of moderator variables, he found that the greatest improvements in overall well-being occurred in men rather than women, and when writing was spaced over longer time periods. Accordingly, we may have reduced our chances of finding differences between emotive writing group and the other groups since our study had few men and writing took place every day. Smyth (1998) also found that students showed larger improvements in psychological well-being than non-students, suggesting that our study was designed to more easily detect differences in psychological well-being than in physical well-being. Further, even with a liberal method for including participants in analyses, only a third of participants completed the thirty-week follow up, reducing our power to detect effects. Thus, the patterning of data may be a result of design features, which have previously been found to influence the efficacy of disclosure.

Alternatively, differences between our study and previous studies may have resulted in the different patterning of our data. Unlike past studies, the present writing intervention took place at home rather than in the laboratory, and so, we had little control over the choice of topic, recency of trauma or length of writing period. Rather the participants could choose what, when, and for how long they wished to write. For example, although participants were instructed to write for 10 minutes, the lengths of the essays clearly indicate that some participants spent more than 10 minutes writing. However, duration of writing has been found to be unrelated to outcomes (Smyth, 1998). In addition, both the location of the writing and the timing of writing may allow the participants to spend more time thinking about their personal and emotional disclosures or to re-read their writings throughout the three days. It is possible that the degree of uncontrolled exposure might be a factor contributing to the short-term distress. It is clear that being able to structure traumatic memories is associated with more successful adaptation and less distress (Greenberg, 1995; Foa, Molnar, & Cashman, 1995). The unstructured nature of our participants' exposure to their traumatic memories may therefore create additional distress in individuals who are less able to exert cognitive control over their memories. Since PTSD patients may have particular difficulty in structuring traumatic memories, it may be no surprise our data find much similarity to Gidron, et al.'s (1996) findings. In addition, participants in our emotive writing condition appeared to use fewer problem solving coping strategies and more repressive coping strategies at longer follow up periods, suggesting poor re-integration of the traumatic memories and further attempts at unsuccessful repression. It has also been proposed that short-term distress may be a necessary factor in the path to long-term recovery (Pennebaker, 1993). Greenberg, Wortman, & Stone (1996) found that a group primed to write about an imagined traumatic event experienced short-term distress, in addition to long-term health benefits, and that negative mood ratings predicted positive health benefits for the imagined trauma group. Thus, the short-term distress accompanying disclosure might be part of the process in which emotional expression results in long-term health benefits. Our data, along with the results of Smyth's (1998) meta-analysis, do not support this hypothesis: although our participants experienced a large amount of short-term distress and a decrease in physical well-being, the long-term benefits were small.

The results of this study echo Smyth's (1998) conclusions. That the risk of inducing short-term distress should caution against a rush to utilise written emotional expression as a self-help therapeutic tool. Moreover Panhuysen and Hezewijk (1999) proposed that confrontation with one's traumatic experiences in writing may have inherent risks. They theorized that adverse effects could be a function of cognitive biases associated with high anxiety and depression inherent in participants. However, it is unclear whether our findings are due to methodological differences, differences in the timing of follow up data collection or the attributes of participants themselves. Alternatively, the lack of clear health benefits might be attributable to certain weaknesses in our study. First, participant attrition decreased our power to detect effects and increased the possibility that selection biases were operating. However, use of liberal statistical significance criteria increased our chances of finding effects; since effects, when emerged, appeared on more than one measure it is unlikely that the health effects we describe are "false positives". While little evidence of selection biases were evident, it is possible that biases were operating through unmeasured variables. Finally, all of our health indices were based on self-report measures and accordingly, may be a function of various reporting biases (Stansfeld, Davey Smith & Marmot, 1993). However, previous studies have found similar effects of writing on both self-report and more objective measures (e.g., Greenberg & Stone, 1992).

In spite of these weaknesses, it is clear from our study that future studies should take into consideration the potential short-term detrimental effects of written-disclosure, particularly when administering writing outside the controlled dosage designs of laboratory studies. Since we found little evidence of changes in potential psychological mediators, it is apparent that we need a better understanding of the mechanisms that underlie long-term health outcomes of disclosure. Studies have begun to examine cognitive functioning during disclosure by analyzing the language used in participants' writings. Murray and his colleagues (Donelly & Murray, 1989; Murray, Lamnin & Carver, 1991) found that students who wrote or talked about trauma to a therapist evidenced greater cognitive changes across the four days of study than students who wrote or talked about superficial topics. Pennebaker & Francis (1996) examined the texts of students who wrote for three days about their thoughts and feelings about coming to college. Increases in words that suggested causal or insightful thinking (e.g., "because", "realize") over

the three days were related to health benefits in the following months. Thus, cognitive changes may account for the health benefits of written emotional expression of past traumas, either independently of disinhibition, or as part of it. Thus future studies should focus more closely on the processes that occur during (e.g., Pennebaker & Francis, 1996) and after writing (e.g., Greenberg, et al., 1996). Until we have a better understanding of the mechanisms underlying the health effects of expressive writing, it will be difficult to maximize the potential health benefits to clinical populations (Gidron, et al., 1996).

Acknowledgments

We would like to thank Johanna Doyle for her help in coding the essays. We gratefully acknowledge the participation of the staff and students in the department of psychology at Glasgow Caledonian University.

References

Blumenthal, J.A., Burg, M.M., Barefoot, J., Williams, R.B., Haney, T. & Zimet, G. (1987). Social support, Type A behavior, and coronary artery disease. Psychosomatic Medicine, 49, 331-339.

Carver, C.S., Sheier, M.F., & Weintraub, J.K. (1989). Assessing coping strategies: A theoretically based approach. Journal of Personality and Social Psychology, 56, 267-283.

Davidson, K.P., & Pennebaker, J.W. (1996). Psychological and physical health impacts of a short term psychoeducational intervention. Psychosomatic Medicine, 58, 92.

Donelly, D.A., & Murray, E.J. (1989). Cognitive and emotional changes in written essays and therapy interviews. Journal of Social and Clinical Psychology, 10, 334-350.

Esterling, B.A., Antoni, M.H., Kumar, M., & Schneiderman, N. (1990). Emotional repression, stress disclosure responses, and Epstein-Barr capsid antigen titers. Psychosomatic Medicine, 52, 397-410.

Evans, P.D., Pitts, M.K., & Smith, K. (1988). Minor infection, minor life events and the four day desirability dip. Journal of Psychosomatic Research, 32, 533-539.

Francis, M.E., & Pennebaker, J.W. (1992). Putting stress into words: The impact of writing on physiological, absentee, and self-reported emotional well-being measures. American Journal of Health Promotion, 6, 280-287.

Foa, E.B., Molnar, C., & Cashman, L. (1995). Change in rape narratives during exposure therapy for post-traumatic stress disorder. Journal of Traumatic Stress, 8, (4), 675-690

Gidron, Y., Peri T., Connolly, J.F., & Shalev, A.Y. (1996). Written emotional expression in posttraumatic stress disorder: is it beneficial for the patient? Journal of Nervous and Mental Disease, 184, 505-507.

Goldberg, D.P. (1978). Manual of the General Health Questionnaire. Windsor: NFER-Nelson.

Greenberg, M.A. (1995). Cognitive processing of traumas: The role of intrusive thoughts and reappraisals. Journal of Applied Social Psychology, 25, 1262-1296.

Greenberg, M.A. & Stone, A.A. (1992). Emotional disclosure about traumas and its relation to health: Effects of previous disclosure and trauma severity. Journal of Personality and Social Psychology, 63, 75-84.

Greenberg, M.A., Wortman, C.B., & Stone, A.A. (1996). Emotional expression and physical health: Reviving traumatic memories or fostering self-regulation. Journal of Personality and Social Psychology, 71, 588-602.

Greenberg, M. A., & Lepore S. J. (1999). Cognitive and emotional processes underlying the health benefits of disclosure. In A.J.M. Vingerhoets and I. Nyklicek (Eds.), The (Non) Expression of Emotions in Health and Disease. Tilburg, The Netherlands; Tilburg University Press.

Greenwald, A. G., Gonzalez, R., Harris, R. J., & Guthrie, D. (1996). Effect sizes and p values: What should be reported and what should be replicated? Psychophysiology, 33, 175-183.

Horowitz, M.J. (1976). Stress Response Syndromes. New York: Jacob Aronson.

Lutgendorf, S.K., Antoni, M.H., Kumar, M., & Schneiderman, N. (1994). Changes in cognitive coping strategies predict EBV-antibody titre change following a stressor disclosure induction. Journal of Psychosomatic Research, 38, 63-78.

Marmot, M. G., Davey Smith, G., Stansfeld, S., Patel, C., North, F., Head, J., White, I., Brunner, E., & Feeney, A. (1991). Health inequalities among British civil servants: The Whitehall II study. Lancet, 337, 1387-1392.

Murray, E.J., Lamnin, A.D., & Carver, C.S. (1991). Emotional expression in written essays and psychotherapy. Journal of Social and Clinical Psychology, 8, 414-429.

Panhuysen, G. & Van Hezewijk, R. (1999). The risks of writing on emotions. Abstract and Poster, Second International Conference on The (non) Expression of Emotions in health and Disease, Department of Psychology, Tilburg University, The Netherlands, 9-11 June 1999.

Pennebaker, J.W. (1993). Putting stress into words: Health, linguistic, and therapeutic implications. Behavioral Research Therapy, 31, 539-548.

Pennebaker, J.W., & Beall, S.K. (1986). Confronting a traumatic event: Toward an understanding of inhibition and disease. Journal of Abnormal Psychology, 95, 274-281.

Pennebaker, J.W., Colder, M., & Sharp, L.K. (1990). Accelerating the coping process. Journal of Personality and Social Psychology, 58, 528-537.

Pennebaker, J.W., & Francis, M.E. (1996). Cognitive, emotional, and language processes in disclosure. Cognition and Emotion, 10, 601-626.

Pennebaker, J.W., Hughes, C.F., & O' Heeron, R.C. (1987). The psychophysiology of confession: Linking inhibitory and psychosomatic processes. Journal of Personality and Social Psychology, 58, 528-537.

Pennebaker, J.W., Kiecolt-Glaser, J.K., & Glaser, R. (1988). Disclosure of traumas and immune function: Health implications for psychotherapy. Journal of Consulting and Clinical Psychology, 56, 239-245.

Pennebaker, J.W., & Susman, J.R. (1988). Disclosure of traumas and psychosomatic processes. Social Science and Medicine, 26, 327-332.

Pennebaker, J.W., & Suedfeld, P. (1997). Health outcomes and cognitive aspects of recalled negative life events. Psychosomatic Medicine, 59, (2), 172-177.

Shalev, A.Y., Bonne, O., & Eth, S. (1996). Treatment of posttraumatic stress disorder: A review. Psychosomatic Medicine, 58, 165-182.

Sheffield, D., McVey, C., & Carroll, D. (1996). Daily Events and Somatic Symptoms: Evidence of a lagged relationship. British Journal of Medical Psychology, 69, 267-269.

Smith, M. & Glass, G. (1977). Meta-analysis of psychotherapy outcome studies. American Psychologist, 32, 752- 760.

Smyth, J.M. (1998). Written emotional expression: A meta-analytical review. Journal of Consulting and Clinical Psychology.

Smyth, J. M., Stone, A.A, Hurewitz, A., & Kaell, A. (1999). Effects of writing about stressful experiences on symptom reduction in patients with asthma or rheumatoid arthritis: a randomized trial. Journal of American Medical Association , 281(14), 1304-9.

Stansfeld, S.A., Davey Smith, G. & Marmot, M.G. (1993). Association between physical and psychological morbidity in the Whitehall II study. Journal of Psychosomatic Research, 37, 227-238.

Zevon, M.A. & Tellegen, A. (1982). The structure of mood change: An idiographic / nomothetic analysis. Journal of Personality and Social Psychology, 43, 111-122.

Copyright