Does Written Emotional Disclosure about Stress Improve College Students’ Academic Performance? Results from Three Randomized, Controlled Studies

Abstract

Several early studies and subsequent reviews suggested that written emotional disclosure (WED)—writing repeatedly about personal stressful experiences—leads to improved academic performance of college students. Yet a critical review of available studies casts some doubt on this, so we conducted three randomized, controlled experiments of the effects of WED versus control writing on grade point average (GPA) of college students. In all three studies, WED writing was implemented effectively—it contained more negative emotion language and generated more negative mood than did control writing. In Study 1, WED did not influence GPA during either the writing semester or subsequent semester among 96 students with headaches. In Study 2, WED had no effect on GPA compared with either control writing or no writing conditions among 124 students with unresolved stress. In Study 3, WED did not influence GPA or retention among 68 academically at-risk ethnic minority students, although secondary analyses suggested some benefits of WED among students who wrote more than once, particularly men. These three studies challenge the belief that WED improves academic performance of college students, and research should examine subgroups of students who might benefit from WED.

Does Written Emotional Disclosure about Stress Improve College Students’ Academic Performance? Results from Three Randomized, Controlled Studies

Academic performance and retention of college students are influenced by many factors, but psychological stress in particular appears to cause poorer performance and attrition (Haines, Norris, & Kashy, 1996; Meilman, Manley, Gaylor, & Turco, 1992; Misra, McKean, West, & Russo, 2000). Stressful life experiences and resulting emotional struggles may impair academic performance and increase attrition by interfering with adaptive behaviors, such as studying or class attendance, or by hindering vital cognitive processes, such as concentration.

The consequences of stressful events, however, hinge on how the experience and associated negative emotions are processed and resolved (Borkovec, Roemer, & Kinyon, 1995; Horowitz, 1986). In general, inhibiting or avoiding thoughts and feelings related to stressors serves to prolong or exacerbate the stress response and has maladaptive emotional, cognitive, and behavioral effects. In contrast, resolution of stressors is facilitated when people voluntarily recall a stressful event, experience and express the emotions related to it, and then explore how it has changed their thoughts and beliefs (Lepore, Greenberg, Bruno, & Smyth, 2002).

The technique of written emotional disclosure (WED) about stressful life events, also called expressive writing, may help people process and resolve stressful experiences and perform better academically. In 1986, Pennebaker and Beall introduced an experimental paradigm to test this idea. In this paradigm, participants are randomized to write for three or four days, for 15-30 minutes each day, about either stressful experiences (WED group) or emotionally neutral topics (control group), and the functioning of the two groups is compared at follow-up several months later. An early meta-analysis of the first 13 WED studies concluded that WED leads to more improvements than control writing, including better grade point average (GPA), with a moderate effect size of nearly half of a standard deviation improvement in outcomes (Smyth, 1998).

These early studies took the field by storm. The notion that privately writing for a few days about personal, stressful experiences—in the absence of a professional therapist, emotional support, feedback, or skills building—can improve one's health, functioning, and even academic performance is very encouraging—but also provocative. Commentators began to advocate for this technique, particularly because it is inexpensive, portable, and easily prescribed, and generated seemingly substantial benefits (Pennebaker, 1997). For example, “writing about trauma, particularly about feelings and emotions related to trauma, appears to have a long-term positive impact on a variety of physiological and psychological health outcomes” (Sheese, Brown, & Graziano, 2004, p. 457). Pennebaker and Keough (1999) opined that “the writing paradigm itself is sufficiently powerful to statically swamp most individual difference measures” (p. 109). An article targeting mental health professionals claimed that “long term benefits of expressive writing include... higher students’ grade point average” (Baikie & Wilhelm, 2005, p. 339). Indeed, a study by Klein and Boals (2001) showed that WED improved students’ working memory, probably by decreasing stress-induced intrusive thoughts, and this finding was used to support a possible mechanism by which WED might lead to better academic performance.

To our knowledge, only six published studies have examined the effects of WED on college students’ academic performance, and in all but one of these articles (Klein & Boals, 2001), they authors concluded that WED led to better GPAs. A more careful examination of these studies, however, shows that their results are more ambiguous and do not necessarily merit the highly positive claims noted above. We briefly review these studies.

Pennebaker, Colder, and Sharp (1990) randomized 124 new college students to write about their “very deepest thoughts and feelings about coming to college” (WED) or to a control group that wrote about the day's activities. Writing was done in groups at various times during the semester. Analyses compared the two groups over the writing and subsequent semester, and the authors reported a non-significant group by semester interaction (p = .11) in which the WED group maintained their GPA from one semester to the next (2.78 to 2.79), whereas controls dropped slightly (2.79 to 2.64), resulting in a net difference between groups of 0.16 of a point.

Pennebaker and Francis (1996) had 61 new college students write about the same WED topic as above, or about neutral control topics. Writing was done in large groups during the last week of October. Analyses again showed a marginally significant group by semester interaction (p = .07). The WED group's GPA increased from the writing semester to the subsequent semester (from about 2.90 to 3.10), whereas the controls dropped slightly (from 2.90 to 2.85), resulting in a net difference between groups of 0.25 of a point during the subsequent semester.

Cameron and Nicholls (1998) randomized 121 first year students to standard WED (using the same instructions as above), to an enhanced WED condition that added the consideration of coping options to disclosure (labeled “self-regulation”), or to a control group. Students wrote in small groups during the 4^th, 5^th, and 6^th week of classes, and GPAs were obtained at the semester's end. Analyses showed that the standard WED group had a higher GPA (2.99) than either the enhanced WED group (2.54) or control group (2.68). These results are difficult to interpret, however, in that the condition that was hypothesized to be most effective—the enhanced WED group—did indeed result in the best self-reported psychological adjustment and mood, but at the same time had the lowest GPA. To account for these unexpected and seemingly contrasting findings, the authors speculated adding coping options to the WED may have led to increased socialization (social coping), which may have aided psychological development but interfered with academic performance.

Lumley and Provenzano (2003) had 74 students with elevated physical symptoms write individually for four days about either their most stressful life experience (WED) or how they manage their time (control). Analyses showed a significant group by semester interaction; the WED students maintained their GPA from the writing semester to the following semester (2.69 to 2.72) whereas the controls showed a substantial drop in GPA from the writing to the subsequent semester (2.86 to 2.34), resulting in a net group difference of 0.55 of a point.

Finally, Klein and Boals (2001) conducted two studies. In Study 1, 52 first semester college students wrote about the stress of coming to college (WED) or about daily activities during weeks 5, 6, and 7 of the fall semester. Analyses across semesters showed no difference between groups in how GPA changed—for the sample overall, it dropped from 3.18 to 2.97. In Study 2, 88 students wrote about a negative major life event, a positive event, or time management three times during weeks 5 through 8 of the semester. It appears that there was no group effect on GPA, although this is not clearly stated in the article.

Overall, the published literature on the effects of WED on GPA of college students is mixed, and we think that the conclusion that WED leads to better grades is premature. One study (with two samples) found no effect, two studies found only marginally significant effects of WED, two studies reported effects that were due largely to worsening among controls rather than improvements in WED students—raising concerns about the reliability of the effects—and one study had a WED group unexpectedly perform much better on GPA than another, enhanced WED group. Clearly this literature needs additional research to determine whether WED can indeed improve the academic performance of college students.

In this paper, we present three randomized experiments of the effects of WED on academic performance in college students. In all three studies, we analyzed the content of the writings to verify that, as expected, WED generated more negative emotional language than did control writing. We also examined post-writing mood, to confirm that WED activated negative feelings, which is typically found in WED studies (Smyth, 1998) and is consistent with the hypothesized mechanisms of change (Sloan, Marx, & Epstein, 2005). The three studies that we report varied in the type of sample, specific disclosure instructions, location of the writing, and control groups used. These variations allowed us to test different conditions under which WED might, or might not, influence grades. As with all research to date in this literature, our first two studies were designed primarily to test whether WED influences mental and physical health, and the results of these health outcome analyses have been published (D'Souza, Lumley, Kraft, & Dooley, 2008; Radcliffe, Lumley, Kendall, Stevenson, & Beltran, 2007). For these two studies, we now present the analyses of WED effects on academic performance. The third study was designed specifically to test WED's effects on academic performance. For each study, we provide a brief introduction and then describe its methods and results. Following Study 3, we present a general discussion.

Study 1: Students with Headaches

Our first study tested the effects of four sessions of laboratory-based WED versus control writing among students with headaches. We examined the effects of WED in the full sample of students as well as sub-samples that had either migraine headaches or tension headaches, because the primary study suggested that the effects of stress management techniques may differ for these two headache subtypes (D'Souza et al., 2008).

Study 1 Methods

Participants

Participants were 96 undergraduate psychology students who had either migraine headaches or chronic tension headaches without migraines. The sample included 81 women (84%) and 15 men (16%), had a median age of 20 years, and was 53.1% European American, 20.8% African American, 13.5% Arabic, 4.2% Hispanic, 3.1% East Asian, 3.1% Multiethnic, and 2.1% Native American. Of the 96 students, 62 (55 women, 7 men) had migraine headaches, and 34 (26 women, 8 men) had tension headaches. All students were recruited from a much larger pool of students who were screened for headaches in classes or by a departmental website. If they met headache inclusion criteria, they were contacted by telephone or e-mail and invited to the laboratory where they provided written consent to the institutional review board-approved protocol, which included a statement that their academic transcripts would be released to the research team. Students who did not meet standard headache diagnostic criteria or who were currently in psychotherapy or counseling were excluded.

Procedures

Recruitment into the study occurred over several consecutive semesters, from February 2001 to March 2002, and participation in the writing exercise ran from week 4 to week 11 of the semester. After a baseline assessment of health variables, students were stratified by headache type and randomized to experimental groups. Participants were seated in a private room and given a sealed packet that contained specific writing instructions. The first of four writing sessions occurred then, and participants returned to the laboratory three more times during the next two weeks for the remaining three writing sessions. Each session lasted for 20 minutes, and students rated their mood before and immediately after writing each session. Students returned their writings to the researchers at the end of each session, and were given course credit or payment for participating.

Writing Conditions

A similar rationale was presented to both writing conditions—writing was described as a potential way to manage stress with possible health and functioning benefits.

Written emotional disclosure (n = 48)

Participants were given standard WED instructions, slightly enhanced with the addition of encouragement to write a story and relate the stressor to one's life and health:

“I would like you to write about a trauma or upheaval or stressful experience that you may be experiencing right now, or that you experienced at some other time in your life. The event that you write about should be the one you consider to be the most stressful that you have experienced and is the most significant to you. When you write about the event, I would like you to discuss the facts surrounding it and write about your deepest feelings related to the issue. Try to make your memories as vivid as possible, including thoughts, emotions and bodily sensations that you experienced. You should try to write about the same event for all four writing days, but this is not a requirement. Again, ideally the topic will be one that you have not talked about in detail with others. When you write, try to tell a story, including the facts of what happened, your deepest feelings, and how the experience as affected your life. You might write about how the event has affected your relationships, or you might write about how the event has affected your health, including your headaches.”

Time management control (n = 48)

We chose the most commonly used neutral writing control condition, which is to write about how one manages one's time. This condition is a credible stress management technique that also controls for the time and effort made by participants in the WED condition. On Day 1, participants were instructed to write about their behavioral activities for the past week. On Day 2, they wrote about their activities for the past 24 hours. On Day 3, they wrote about their planned activities for the next 24 hours, and on Day 4, they wrote about their planned activities for the next week. Participants were instructed to write about only their actions and to refrain from writing about their feelings or opinions.

Measures

Grade Point Average (GPA)

Academic transcripts were obtained for each student in the study. We extracted GPA values (range of 0.0 = F to 4.0 = A, with GPA weighted for the number of credits each class was worth) for both the writing semester and the subsequent semester. Class grades of pass or fail, successful or unsuccessful, incomplete, withdrawal, or “not reported” were not averaged into the GPA, consistent with university policy. However, when students failed or withdrew from all of their classes for a given semester, they received a GPA of 0.0 for that semester.

Linguistic analysis of writings

We transcribed all of the writings and submitted them to analysis using specialized software, Linguistic Inquiry and Word Count-Second Edition (SLIWC; Pennebaker, Francis, & Booth, 2001). This program counts the proportion of total words that occur in different language categories. We focused specifically on negative emotion words as a manipulation check to verify that the WED writings contained more negative emotion than did control writings.

Writing-related mood

Before and after writing, participants rated how they felt “right now” on mood adjectives using an abbreviated version of the Positive and Negative Affect Schedule-Expanded Version (Watson & Clark, 1994). Items were rated from 1 (not at all) to 7 (a great deal), and we focused specifically on adjectives in four negative mood categories (anger, guilt, sadness, fear). These four moods were highly correlated (e.g., alpha = .84 for the mean mood ratings), so we averaged the four ratings into one negative mood score, and calculated a mean “before writing” rating and a mean “after writing rating” across the writing days.

Study 1 Results

All but three of the 96 participants completed all four assigned writing days; two students in the WED group and one in the control group (all in the migraine sub-sample) wrote for only the first day but did not return for subsequent sessions. We retained these students in analyses because we had their academic performance outcome data, and we sought to use the conservative “intent-to-treat” approach of analyzing all randomized participants. Although all 96 students had outcome GPA for the semester of writing, 11 students (6 WED group and 5 control group) did not enroll in classes at the university during the subsequent semester; thus, analyses that included the subsequent semester were conducted on only 85 students.

Manipulation check analyses

The percentage of negative emotion words (averaged over the writing days) in the WED writings (M = 2.70, SD = 0.86) was 4.5 times higher than in the control writings (M = 0.57, SD = 0.49), t(94) = 14.93, p < 0.001. In addition, change scores were calculated for negative mood levels from post to pre-writing, which showed the WED writings (M = 0.44 SD = 0.81) resulted in significant increases in negative mood compared to the control writing (M = -0.19, SD = 0.36), t(93) = 4.77, p < .001. Both of these findings confirm expected differences between the WED and control writing groups and suggest that WED was conducted as expected.

The left columns of Table 1 show the GPA for the full sample of 96 students, including GPAs for the writing and subsequent semesters. Independent samples t-tests indicated that the GPA of the WED group was not different than that of the controls during either the writing semester, t(94) = 0.88, p = .38, or subsequent semester, t(83) = -0.36, p = .72. A repeated measures ANOVA, comparing both groups over the two semesters, was not significant for the group effect, F(1, 78) = 0.03, p = 0.87, or the group × semester effect, F(1, 78) = 0.48, p = 0.49. The middle columns of Table 1 show that for the 62 students with migraines, the WED group had was not significantly different in GPA than the controls during the writing semester, t(60) = 1.58, p = .12, or the subsequent semester, t(51) = 0.44, p = .66. The repeated measures ANOVA showed that there was no group effect, F(1, 46) = 1.50, p = 0.23, or group × semester interaction, F(1, 46) = 0.17, p = 0.68. The right columns of Table 1 show that for the 34 students with tension headaches, the two groups did not differ in GPA during either writing semester, t(32) = - 0.60, p = .56, or subsequent semester, t(30) = -1.24, p = .22. Again, there was no group effect, F(1, 30) = 1.15, p = 0.29, or group × semester interaction, F(1, 30) = 0.60, p = 0.44.

Table 1.

Grade Point Average (M, SD) for Writing Semester and Subsequent Semesters for the Written Emotional Disclosure (WED) and Control Groups, for the Full Sample as well as Migraine and Tension Headache Sub-samples

	Full Sample		Migraine Headache		Tension Headache
	WED (n = 48)	Control (n = 48)	WED (n = 31)	Control (n = 31)	WED (n = 17)	Control (n = 17)
Writing semester	2.91 (0.84)	2.76 (0.88)	3.04 (0.82)	2.71 (0.86)	2.66 (0.84)	2.84 (0.93)
Subsequent semestera	2.87 (0.92)	2.94 (0.98)	3.00 (0.95)	2.88 (0.99)	2.63 (0.83)	3.04 (0.99)

^aFor the subsequent semester, full sample: WED (n = 42), Control (n = 43); Migraine: WED (n = 27), Control (n = 26); Tension: WED (n = 15), Control (n = 17).

Study 2: Students with Unresolved Stress

Study 1 had several limitations. We studied students with headaches, rather than who reported unresolved stress, and it is possible that WED is simply not effective for such a sample; indeed, our primary study found little benefit of WED on outcome measures related to headaches and health (D'Souza et al., 2008). Also, recent research indicates that disclosure writing about the same stressful experience across days leads to larger effects than does changing topics (Sloan et. al, 2005), but in Study 1, our instructions did not require students to remain on the same topic over days, and writers changed topics. Another limitation is that control writing may not have been inert. Although this control topic is commonly used, writing about time management may actually have reduced stress, especially for students who struggle with managing their time, or because some writing about one's daily activities inadvertently contains elements of emotional disclosure (e.g., writing about yesterday's fight with one's partner or visiting an ill parent). Such a control condition could reduce the comparative benefits of WED. Alternatively, the control condition could be somehow aversive, and it is noteworthy that several studies of WED on academic performance in college students found control group worsening rather than WED group improvement in grades (Lumley & Provenzano, 2003; Pennebaker et al., 1990), suggesting the possibility that the control condition actually led to worsening of grades.

We redressed these concerns in Study 2. We recruited students who theoretically should benefit from WED—those who report having some unresolved stressor in their lives—and instructed the students to write on the same stressful topic over days. Both of these changes should lead to positive WED effects, and in our primary study, the WED group did indeed show substantial health improvement and stress reduction compared with controls (Radcliffe, Lumley, Kendall, Stevenson, & Beltran, 2007). Study 2 also included a second control condition—a no-writing control group—to test whether any observed disclosure effect on GPA was due to some unexpected influence of the neutral writing control condition. No prior studies of WED and academic performance have used an additional control condition.

Study 2 Methods

Participants

Participants were 124 undergraduate psychology students who reported an unresolved stressful experience. At the start of the semester, students completed a web-based screening survey that included the items, “Have you ever experienced a traumatic, stressful, or bothersome event?” and, if so, “How much does the event still bother you?” (answered not at all, somewhat, moderately, or very much). We recruited only students who reported experiencing a stressful event that continued to bother them moderately or very much. The final sample had 103 women (83.1%) and 21 men (16.9%), had a median age of 19 years, and was 57.3% European American, 28.2% African American, 6.5% East Asian, 2.4% Hispanic, and 5.6% other.

Procedure

Recruitment into the study occurred over several consecutive semesters from October 2002 to November 2003, and participation in the writing exercise ran from week 4 to week 11 of the semester. Potential participants were contacted by email, invited to participate, and brought to the lab where they provided written consent. After completing baseline questionnaires (not part of the current report), participants were stratified by gender and randomized to groups using a random numbers table. Those randomized to the no-writing control group were dismissed. Those randomized into one of the writing groups were placed in a private room and given instructions for their assigned topic. Participants were left alone and wrote by hand on their topic for the first of four, 20-minute sessions. At the end of the first session, participants were given writing journals and instructed to write in a private place of their choosing (such as at home) for 20 minutes for the next two sessions. Writing for session 4 occurred in the laboratory, one week after the first writing session, and participants brought their writings from sessions 2 and 3 back with them, and returned all writings to the researchers. Participants received course credit or payment for participating.

Experimental Groups

Written emotional disclosure (n = 42)

This WED group was given a rationale and instructions that were similar to those used in Study 1. An additional instruction for this study was that the student should “work on and resolve the stressful experience, and this means that you should write about the same stressful experience over all 4 days.” Also, the student was instructed, “As you write, try to make sense of your stressful experience—to understand its meaning. This might include trying to answer questions about why the experience occurred, how it has affected you—such as your beliefs, your relationships, or your actions—and ways that you might cope with the experience now.”

Neutral writing control (n = 42)

The rationale and instructions for this group were identical to those used in Study 1.

No-writing control (n = 40)

This group did not participate in any writing, but they provided GPA data.

Measures

As in Study 1, we obtained transcripts and calculated GPA for the semester in which writing was conducted and the subsequent semester. GPA was determined in the same manner as in Study 1. Also, as in Study 1, writing content was analyzed with the SLIWC and mood ratings were obtained before and after writing on an abbreviated version of the PANAS-X.

Study 2 Results

All but seven of the 82 students in the two writing conditions completed all four writing days; five students in the WED group and two in the control group wrote for only the first day. As in Study 1, we analyzed all randomized students. For the subsequent semester, 9 students (1 in WED, and 4 in both the neutral writing and the no-writing control groups) did not return to the university, so analyses that included the subsequent semester were conducted on 115 students.

Manipulation check analyses

These analyses were conducted only on the WED and control writing, because these were the only groups that provided mood data and writing samples. The percentage of negative emotion words (averaged over the writing days) in the WED writings (M = 1.13, SD = 0.41) was 4.5 times higher than in the control writings (M = 0.33, SD = 0.28), t(79) = 10.14, p = 0.04. Again, change scores (post-writing minus pre-writing) were calculated for negative mood. These scores indicate that WED writings (M = 0.65, SD = 0.92) resulted in significantly more negative affect, compared with control writings (M = -0.17, SD = 0.30), t(77) = 5.29, p < 0.001 . These findings confirm expected differences between WED and control writing, and suggest the WED was conducted as expected.

GPA analyses

Table 2 presents the GPA data for all three experimental groups, for both the writing semester and subsequent semester. An ANOVA comparing the three groups indicated that the groups did not differ in GPA during either the writing semester, F(2, 121) = 0.49, p = .61; or during the subsequent semester, F(2, 112) = 0.05, p = .95, and a repeated measures ANOVA revealed no group effect, F(1, 112) = 0.36, p = 0.70, or group × semester interaction, F(1, 112) = 0.43, p = 0.52. When t-tests were used to compare the WED group to each control group separately for each semester, the WED group did not differ from either control condition, for either semester (all p > .38).

Table 2.

Grade Point Average (M, SD) for the Writing Semester and Subsequent Semester for the Written Emotional Disclosure (WED), Control Writing, and No-writing Control Groups for Students with Unresolved Stressors

	WED (n = 42)	Writing Control (n = 42)	No-writing Control (n = 40)
Writing semester	2.84 (0.93)	3.01 (0.82)	2.99 (0.85)
Subsequent semestera	2.91 (1.07)	2.87 (1.08)	2.94 (0.98)

^aFor the subsequent semester: WED (n = 41), Writing control (n = 38), and No-writing control (n= 36).

Study 3: At-Risk Minority Students

Studies 1 and 2 were adequate tests of the potential of WED to improve academic performance, but they had various limitations. First, generalizability was limited by studying only students taking psychology classes, and particularly only students meeting special screening criteria—headaches or unresolved stress. Second, many students in Studies 1 and 2 were making good progress in college; only about one-third were freshman, who might be expected to need more help adjusting and managing stress. Third, those students—and especially students in the six previously published studies of WED and academic performance—were predominantly White and middle or upper middle class, which further limits generalizability. Fourth, both studies were interested primarily in whether WED could reduce stress symptoms and improve health, and academic performance was a secondary outcome. Fifth, we conducted WED throughout the semester rather than at the start of the semester, which likely attenuated any benefits during that semester, and required us to track students to the next semester, by which point, any academic benefits might have waned. Finally, meta-analyses of WED studies suggest that men tend to respond better than women to emotional disclosure (Frattaroli, 2006; Smyth, 1998), perhaps because men are socialized to inhibit the sharing of stressful emotional experiences. No studies, however, have examined whether the genders differ in how WED influences academic performance, and Studies 1 and 2 had too few men for gender analyses.

To redress these limitations, Study 3 tested the effectiveness of classroom-based WED at the start of the first college semester in a different population—ethnic minority students enrolled in a special program for academically at-risk students. Although colleges spend considerable resources trying to improve the performance and retention of such students, low grades and high drop-out are common (Thayer, 2000). We examined whether WED conducted in the classroom at the start of a semester would improve GPA and reduce course failures and drop-outs during the semester. Also, because a number of students did not write beyond the first session, we analyzed not only all students who were randomized but also those who adherent and wrote for at least two days. Finally, we also examined the performance of women and men separately to test the hypothesis that men would respond better than women to WED.

Study 3 Methods

Participants

Participants were new college students enrolled in one of four sections of an orientation course that teaches basic college learning and adjustment skills and which meets once per week. Students admitted to this program fall short of usual admissions standards, have a high school GPA between 2.0 and 2.5, and have not taken college admissions tests. A total of 70 students were invited to participate; two declined, and the remaining 68 consented and were randomized. There were 38 women (55.9%) and 30 men (44.1%), and all students but two (one Hispanic and one Middle Eastern student) were African American. The median age was 19 years.

Procedures

Instructors of the four classes during the winter, 2004 semester agreed to make the writing exercise part of their course. In week 2 of the semester, a researcher came to class, described the study to the students, and invited participation; the study was described as seeking to test the academic effects of two writing exercises. After written consent was obtained, we distributed to the students prepared and randomized (via a random numbers table) writing packets which contained instructions for either WED or the control topic. Four writing sessions were scheduled, once per week, for 20 minutes at the end of the class period, and held in the classroom. Sessions started in week 2 and lasted through week 5 of the semester. At the end of each session, packets were retained by the researcher and distributed to students during the subsequent session; the researcher kept the writings after the last session. Thus, students could write up to four times, but only if they were in class for each of the four weeks; however, as noted below, attendance in this high-risk group varied widely.

Writing Topics

The rationale and instructions for the two conditions were similar to those used in Studies 1 and 2, with slight modifications based on guidelines for increasing the impact of WED (Broderick, Stone, Smyth, & Kaell, 2004; Gidron et al., 2002).

Written emotional disclosure (n = 32)

The WED group wrote about unresolved stressful experiences and their deepest feelings, and for sessions 2 through 4, the instructions were more explicit than in Studies 1 and 2. For session 2, the instructions included encouragement to write about the stressful experience in a narrative, with a beginning, middle, and end. Session 3 instructions asked the student to add “how the experience has affected you, such as your beliefs or attitudes, your actions, and your relationships.” Finally, session 4 instructions asked the student to include how “you might cope with the event or problem—that is, how you might change the situation, change your thoughts or actions, or change your relationships so that it is no longer stressful or problematic.”

Neutral writing control (n = 36)

This condition was a variation of time management that dealt with future plans, and which we used in our previous study that demonstrated academic benefits of WED (Lumley & Provenzano, 2003). Students were asked to write about their planned actions or behaviors for the next day (session 1), next week (session 2), next year (session 3), and next decade (session 4). Students were instructed to “try to be objective and write only about your planned actions, not about feelings or opinions.”

Measures

With the students’ consent, we obtained their academic transcripts at the end of the semester and derived the following measures of academic performance:

Grade point average (GPA)

This was calculated as in Studies 1 and 2.

Classes and semester failed

The number of classes that each student failed (grade of E) or did not complete (grades of incomplete, withdrawal, or no grade reported) was divided by the number of classes that each student took; this yielded the percentage of classes failed for each student. In addition, we also classified students as to whether they had failed the semester. Students who had failure grades, withdrawals, incompletes, or “no grade reported” for all of their classes were considered to have failed the semester. Those students with a passing grade (i.e., D or better) in at least one of their classes did not fail the semester.

Also, as in Study 1, writing content was analyzed with the SLIWC and mood ratings were obtained before and after writing on an abbreviated version of the PANAS-X.

Study 3 Results

Manipulation check analysis

The percentage of negative emotion words (averaged over the writing days) in the WED writings (M = 1.00, SD = 0.47) was nearly 4 times higher than in the control writings (M = 0.25, S D= 0.26), t(55) = 7.69, p = 0.02. Changes scores for negative mood, from post to pre-writings, show that WED writing (M = 0.25, SD = 0.71) resulted in significantly more negative emotion, compared with control writing (M = -0.05, SD = 0.27), t(41) = 1.91, p = .003. This finding confirms expected differences between WED and control writing groups, and suggests the WED condition was conducted as expected.

Adherence to the Writing Exercise

Although we planned to have students write for four consecutive weeks, we found much variability in the number of writings completed, due largely to students’ absences. Of the 32 students randomized to WED, two (6.2%) did not write at all, six (18.8%) wrote for the first session only, eight (25%) wrote for two sessions, eight (25%) wrote for three sessions, and eight (25%) wrote for all four sessions. Of the 36 students randomized to the control group, four (11.1%) wrote just once, seven (19.4%) wrote for two sessions, seven (19.4%) wrote for three sessions, and 18 (50%) wrote all four sessions.

Analyses of Academic Performance in the Full Sample

We first analyzed all randomized students, regardless of the amount of writing completed. The first set of data in Table 3 shows the academic performance of the two groups. Although the WED group's GPA was 0.22 better than the controls’, the variation was large, and this difference was not significant, t(66) = 0.61, p = .55. The WED group also did not differ significantly from the control group in the percent of classes failed, t(66) = 0.69, p = .49, or the percent of students who failed the semester, χ²(1, N = 68) = 0.26, p = .61.

Table 3.

Academic Performance for Written Emotional Disclosure (WED) and Control Groups for all Randomized Students as well as Adherent Students (Those Who Wrote for at Least Two Sessions), and for Adherent Women and Men Separately

	All Randomized Students		All Adherent Students		Adherent Women		Adherent Men
Academic Measure	WED (n = 32)	Control (n = 36)	WED (n = 24)	Control (n = 32)	WED (n = 14)	Control (n = 19)	WED (n = 10)	Control (n = 13)
GPA (M, SD)	1.99 (1.53)	1.77 (1.44)	2.43 (1.31)	1.82 (1.42)	2.51 (1.44)	1.94 (1.38)	2.32 (1.19)	1.65 (1.53)
% Classes failed (M, SD)	42.29 (40.91)	49.26 (42.24)	30.35 (35.43)	46.67 (41.62)	35.36 (37.85)	37.37 (40.29)	23.33 (32.35)	60.26 (41.27)
Failed semester (n %)	8 25%	11 30.6%	3 12.5%	9 28.1%	4 14.3%	6 21.1%	1 10.0%	5 38.5%

Analyses of Academic Performance in the “Adherent Sample.”

As noted above, 12 participants in the study did not write beyond the first session, when the researcher initially came to the class and recruited the students. We decided that writing for at least 2 days was a minimally acceptable “dose” of writing to be considered actively participating or adherent, resulting in 24 disclosure writers and 32 control writers in the adherent sample.

The second set of data in Table 3 shows the academic performance data of adherent students. Here, there is a clearer difference between the groups, more strongly favoring the WED group. The WED group's GPA was 0.61 points higher than the controls', t(54) = 1.63, p = .11, WED's class failure rate was reduced almost 16%, t(54) = 1.54, p = .13, and WED had two-fold lower semester failure rate, χ²(1, N = 56) = 1.99, p = .16. Finally, Table 3 also shows the results separately for women and men who were adherent (wrote for at least two sessions). For women, the WED group's GPA was 0.57 higher than the control group's, t(31) = 1.15, p = .26; and for men, the WED group's GPA was 0.67 higher, t(21) = 1.14, p = .27. For class and semester failure, WED helped men but not women. For men, WED led to a significant reduction of 37% in classes failed, t(21) = 2.33, p = .03, and a 3.8 times lower likelihood of failing the semester, χ²(1, N = 23) = 2.38, p = .12. There were no effects for the women on these variables, t(31) = 0.15, p = .89; and χ²(1, N = 33) = 0.25, p = .62, respectively.

General Discussion

These three randomized, controlled studies suggest that the effect of four days’ of written emotional disclosure, or expressive writing, about stress on the objective academic performance of college students is negligible. In none of the studies was there a significant effect of WED on GPA during either the semester of writing (all three studies) or the subsequent semester (Studies 1 and 2). Admittedly, studies without effects are open to a host of explanations that might explain the lack of findings. Yet, we believe that these studies address many of the potential reasons typically offered for null findings. Most importantly, the lack of a WED effect on GPA was replicated across three independent studies. Furthermore, in all of the studies, manipulation checks indicated that the WED writing contained much higher levels of negative emotional language, and created a more negative mood, than did control writing, which suggests that the WED writing was engaged in adequately. Sample sizes in the three studies were as large or larger than previously published studies and large enough to identify relatively small effects, if present; yet, there were not even trends for WED to lead to better grades than the control condition, at least in the first two studies. The three studies also complement each other by testing different samples, using different control conditions, varying the location of the writing (lab, home, classroom), and applying different yet theoretically powerful WED instructions. Thus, we think that these studies raise substantial doubts about the ability of WED to improve academic performance in general. As we discuss below, however, we suspect that WED may be more effective for some subgroups of students, or under certain conditions.

Study 1, on students with headaches, found that four sessions of WED in the controlled conditions of the laboratory resulted in no differences in GPA than four sessions of time management writing. There were no effects specific to a subtype of headache either. Study 2 targeted students with unresolved stress—students who theoretically should benefit from WED, and indeed, did so quite substantially in our primary study (Radcliffe et al., 2007). Yet, WED clearly had no effect on GPA, during the writing or subsequent semester, and this is true whether an active writing control, or a no-writing control was used. Our prior study of students with physical symptoms found that the writing control group deteriorated in GPA over two semesters, whereas the GPA of the WED group was unchanged, suggesting an apparent benefit for the WED group (Lumley & Provenzano, 2003). Studies 1 and 2, however, raise questions about the reliability of the earlier findings.

Study 3 targeted ethnic minority students at risk for academic problems and focused exclusively on improving their academic performance by having writing occur in classrooms at the start of the semester, in contrast to the other two studies, which dealt primarily with improving health or stress and secondarily addressed academic performance. Study 3 also found that neither grades nor failure rates improved among those in the WED group. Secondary analyses among those who wrote for at least two sessions suggested some academic benefits of WED, specifically for men—a finding that is consistent with meta-analyses of the larger WED literature (Frattaroli, 2006; Smyth, 1998). It has been hypothesized that men might benefit more than women from WED because men have been socialized to avoid expressing the vulnerable emotions of fear, sadness, guilt, and shame, whereas women may feel more comfortable sharing such emotions with others. Thus, WED may be particularly useful for men because it provides a partially anonymous, confidential, and socially acceptable method of expressing these feelings. Yet, it should be noted that these gender analyses are limited by their post-hoc nature and the differential attrition of the two groups, rendering them less comparable.

It has been more than 20 years since the first controlled study of WED was published. Early narrative reviews of the WED literature were very optimistic, concluding that there were robust effects of WED, including improvements to college students’ GPA (Pennebaker, 1997; Pennebaker & Keough, 1999). Although an early meta-analysis concluded that WED led to more benefits than control writing, including higher GPAs (Smyth, 1998), more recent meta-analyses, which include a much larger number of studies, are less supportive of the benefits of WED, concluding either that WED has not been proven to be effective (Meads & Nouwen, 2005), or that the effects are present but quite small (Frattaroli, 2006; Frisina, Borod, & Lepore, 2004), or inconsistent across samples (Harris, 2006).

The recent meta-analysis of Frattaroli (2006) is particularly noteworthy. She reported that 21 studies had investigated outcomes relevant to academic performance, and that there was an overall positive, yet very small effect of WED. Most of these 21 studies, however, are unpublished doctoral dissertations, and the academic outcomes that were studied varied widely, including objective measures such as GPA, but also self-reported behaviors such as class absences, and subjective outcomes such as college adjustment and test anxiety. Although Frattaroli combined all of these outcomes in her meta-analysis, it is likely that different academic outcomes show different effects, and perhaps WED improves subjective outcomes more strongly than objective outcomes, such as GPA.

We believe that the published literature has no unequivocally positive findings of the benefits of WED on college students’ grades. Two studies in one article showed no effect (Klein & Boals, 2001), one study had only a marginal effect (Pennebaker & Francis, 1996), two studies showed significant (Lumley & Provenzano, 2003) or marginal effects (Pennebaker et al., 1990) due primarily to unexpected worsening of the control group, and the remaining study had a surprising effect in which standard WED led to the highest GPA, but enhanced WED led to the lowest GPA. Taken together with the three studies that we report here, it appears that there is little support for the proposal that WED improves college student's grades.

Perhaps it should not be a surprise that outcomes such as grades are not improved by WED. First, it appears that WED has rather modest benefits in general. Second, academic performance is influenced by many factors, and unresolved stress—the target of WED—probably has a relatively small influence compared with socioeconomic, environmental, personality, and intellectual factors. Furthermore, grades are a relatively distal outcome of stress reduction and are less likely to change compared to more proximal mediators, such as improvements in working memory and better health. It is possible that WED can improve such proximal effects of unresolved stress, but that outcomes further down the causal chain, including ultimately grades, are less likely to be affected by WED, and likely need more comprehensive and direct interventions.

Two other concerns regarding the use of WED should be noted. First, participants randomized to the WED group may be slightly more likely to drop out of this writing assignment than those randomized to control writing. This trend was found in the current three studies. Although the actual frequency of attrition in the two groups was rather small and did not reach statistical significance in any given study, the ratio of dropping out from WED versus control after one day of writing was approximately 2:1 in the Study 1, 5:2 in study 2, and over 2:1 in Study 3. We suspect that this is due primarily to emotional discomfort associated with WED, such as shame or guilt triggered by certain experiences, discomfort experiencing inhibited emotions, and concern over privacy and social repercussions. These experiences are associated with a second concern about WED, which is that participants tend to experience a short-term increase in negative affect or worsening of mood when engaging in WED, and this was verified in our manipulation check analyses. This stands in sharp contrast to the relief and calmness usually stemming from stress management interventions such as relaxation training, mindfulness meditation, or cognitive reappraisal. These two limitations further reduce the likelihood that WED will be suggested routinely for students or others to reduce their stress.

There are several additional limitations of the current set of three studies. First, all three studies had students write about general life stressors, whereas an early review of the WED literature hypothesized that academic benefits might be limited to writing that focused exclusively on the stress of college adjustment (Pennebaker & Keough, 1999). Although our review of available studies does not support this hypothesis, we recommend that it be tested more directly by randomizing students to write specifically about college adjustment stress versus general life stress. Second, our university has minimal selection criteria for undergraduate admissions, which results in a very wide range of academic performance among the students. This was reflected in the relatively large standard deviations of their GPAs, which increased the difficulty of reaching statistical significance. Other studies that purported to find support for WED on GPA (Pennebaker et al., 1990; Pennebaker & Francis, 1996; Cameron & Nichols, 1998) were conducted at more selective colleges, which may have reduced performance variability, making it easier to find significant effects of WED.

More importantly, we believe that WED can improve stress, health, and possibly academic performance, but only for certain people or under certain conditions. This is suggested by the suggestive findings of Study 3—that men might uniquely benefit from WED. Indeed, the current focus of WED research in general appears to be shifting away from trying to demonstrate overall benefits of WED to an attempt to identify moderators; that is, the types of people for whom, and the conditions under which, WED is effective (Norman, Lumley, Dooley, & Diamond, 2004; Smyth & Pennebaker, 2008; Stroebe, Schut, & Stroebe, 2006). For example, we have proposed a model of individual differences in emotional experience and expression that may moderate the effects of WED (Lumley, Tojek, & Macklem, 2002). This model proposes that WED is best-suited for people who have are able to acknowledge having an unresolved stressor and who have negative emotions that they have inhibited or suppressed because of ambivalence about expression or the lack of social support for disclosing. We hypothesize that effects of WED can be enhanced by targeting those people for whom it is best-suited, and that alternative stress reduction interventions might be indicated for other types of people. We encourage research on these hypotheses.