How Do I Love Thee? Let Me
Count the Words
The Social Effects of Expressive Writing
Richard B. Slatcher and James W. Pennebaker
The University of Texas at Austin
ABSTRACT—Writing about emotional experiences is asso-
ciated with a host of positive outcomes. This study ex-
tended the expressive-writing paradigm to the realm of
romantic relationships to examine the social effects of
writing. For 3 consecutive days, one person from each of
86 dating couples either wrote about his or her deepest
thoughts and feelings about the relationship or wrote
about his or her daily activities. In the days before and
after writing, instant messages were collected from the
couples. Participants who wrote about their relationship
were significantly more likely to still be dating their ro-
mantic partners 3 months later. Linguistic analyses of the
instant messages revealed that participants and their
partners used significantly more positive and negative
emotion words in the days following the expressive-writing
manipulation if the participants had written about their
relationship than if they had written about their daily
activities. Increases in positive emotion words partially
mediated the relation between expressive writing and re-
Over the past two decades, multiple studies have demonstrated
the positive beneﬁts of expressive writing in domains as diverse
as health, achievement, and well-being. Most of these studies
have used a relatively straightforward procedure in which par-
ticipants write about their deepest thoughts and feelings about a
particular topic for 20 min a day, 3 or 4 days in a row. The
ﬁndings indicate that expressive writing can result in fewer
doctor visits, fewer depressive symptoms, enhanced immune
system functioning, better grades, and a host of other positive
outcomes (for reviews, see Lepore & Smyth, 2002, and Smyth,
Researchers now are examining the social effects of expres-
sive writing, working from the assumption that confronting
conﬂicting or complex emotions or thoughts can facilitate social
interactions. The preliminary ﬁndings suggest that expressive
writing may be particularly beneﬁcial for people in romantic
relationships. For example, when people write expressively
about recent relationship breakups, they are somewhat more
likely than control participants to reunite with their partners
(Lepore & Greenberg, 2002). Similarly, when married couples
recovering from inﬁdelity write emotionally expressive letters to
each other, they experience reductions in depression, anger, and
marital distress (Gordon, Baucom, & Snyder, 2004). The social
effects of writing need not be limited to people recovering from a
relationship breakup or to those whose relationship is in dis-
tress. People in healthy and committed romantic relationships
also might beneﬁt from expressive writing.
There are a number of ways in which one could measure the
effects of expressive writing on the functioning of relationships.
One way is to examine the impact of writing on relationship
stability. Results from the few studies examining the social ef-
fects of expressive writing indicate that it may enhance rela-
tionship stability among couples in distress (Gordon et al., 2004;
Lepore & Greenberg, 2002). Expressive writing could lead to
improved stability for individuals in normal, healthy relation-
ships as well.
Although previous studies have addressed the potential
beneﬁts of expressive writing for couples, none have examined
the underlying mechanisms that might mediate writing’s effects
on relationship outcomes. One potential mediator is natural
language use (Pennebaker & Graybeal, 2001; Sillars, Shellen,
McIntosh, & Pomegranate, 1997). By measuring the words that
people use with others in their social worlds, researchers can
gain insight into the social changes that occur after expressive
writing. The use of emotion words may be particularly rele-
vant. Positive emotion words—such as happy and love—and
negative emotion words—such as angry and nervous—can
Address correspondence to Richard B. Slatcher, Department of
Psychology A8000, The University of Texas at Austin, Austin, TX
78712, e-mail: email@example.com.
660 Volume 17—Number 8Copyright r2006 Association for Psychological Science
reveal deeply felt emotions (Pennebaker, Mehl, & Niederhoffer,
2003; Twenge, Catanese, & Baumeister, 2003). Expressive
writing provides an opportunity for reﬂection, and, accordingly,
one would expect it to lead to increased emotional expressive-
ness with other people—reﬂected in increases in emotion words
during social interactions. Such increases in emotion words
might mediate higher-level changes in relationship functioning,
such as improvements in relationship stability.
There are various ways to measure the words that people use.
One technology—instant messaging (IM)—holds great promise.
For many people (53 million adults, including 30% of all In-
ternet users over the age of 40), IM is quickly replacing e-mail as
a preferred mode of on-line dyadic communication (Shiu &
Lenhart, 2004). Unlike e-mail, IM allows users to chat with each
other so that a conversation can unfold in much the same way
that spoken conversation does. The analysis of IM conversations
can allow researchers to examine the ebb and ﬂow of people’s
natural language use.
In the present study, we sought to investigate the social effects
of expressive writing. Individuals in committed romantic rela-
tionships were randomly assigned either to write about their
deepest thoughts and feelings about their relationship or to write
about a superﬁcial topic for 20 min a day, 3 days in a row. Three
predictions were tested. First, we predicted that those who wrote
about their relationship would be more likely to be dating their
partners 3 months after the experiment. Second, we predicted
that participants who wrote about their relationship would
change in the way in which they communicated with their
partners. Specifically, we expected those in this group to in-
crease their use of positive and negative emotion words in their
daily IM conversations. Finally, we predicted that changes in
use of emotion words would mediate the relation between ex-
pressive writing and relationship stability.
Eighty-six undergraduate psychology students (55 women, 31
men; mean age 518.7, SD 51.0) and their partners (31 women,
55 men; mean age 519.3, SD 51.6) were recruited on the basis
that they were in a committed heterosexual romantic relation-
ship and that they engaged in IM conversation with each other
every day. The couples had been dating an average of 1.31 years
One member of each couple participated in the writing phase of
the study. An experimenter met with participants individually or
in small groups of 2 to 5 to give them an overview of the study.
They were told that its purpose was to quantify the types of words
that people use in writing and everyday interactions.
Participants were instructed to forward to a secure e-mail
address all daily IM conversations between themselves and their
partners during the 10 days of the study. Considerable effort was
taken during the introductory session to reassure participants
that their messages would be completely conﬁdential.
Questionnaires assessing demographic information, rela-
tionship status, and relationship satisfaction were completed by
participants and their partners on-line from home on the ﬁrst day
of the study. Satisfaction was measured using the Relationship
Assessment Scale (RAS; Hendrick, 1988). The RAS consists of
seven items, such as ‘‘In general, how satisﬁed are you with your
relationship?’’ Ratings are made on a 7-point Likert scale.
Participants were randomly assigned to the experimental (n5
44) or control (n542) writing condition and were instructed to
set aside 20 min per day on the fourth, ﬁfth, and sixth days of the
study for completion of the on-line writing exercises. Partici-
pants in the experimental condition were instructed to write
about their deepest thoughts and feelings about their current
romantic relationship, ‘‘to really let go and explore your very
deepest emotions and thoughts about your relationship.’’ Those
in the control condition were instructed to write in detail about
their daily activities—a standard control condition for expres-
sive-writing studies. The purpose of the writing assignments was
left intentionally vague to reduce the possibility of demand
characteristics; participants were told that the writing assign-
ments would be used to assess basic word use.
Three months later, participants completed a brief on-line
questionnaire that assessed relationship status and satisfaction.
The IM conversations were analyzed using the Linguistic In-
quiry and Word Count program (LIWC; Pennebaker, Francis, &
Booth, 2001). All conversations from couples in both conditions
were spell-checked prior to being submitted to LIWC. They
were converted to text ﬁles, categorized by speaker (participant
or partner), and subcategorized according to whether they dated
from before or after the writing manipulation, resulting in a total
of four separate IM text ﬁles per couple.
Relationship Stability and Language Use
Expressive writing was significantly related to the long-term
stability of relationships, odds ratio 53.09, p
d50.54. Thirty-four participants (77%) in the experimental
condition were still dating their partners at the 3-month follow-
up mark, compared with 22 participants (52%) in the control
condition. There were no significant differences in baseline
relationship satisfaction between the conditions (experimental
M56.21, SD 50.63; control M56.01, SD 50.66), nor was
condition related to satisfaction at the 3-month follow-up (ex-
perimental M55.87, SD 51.50; control M55.94, SD 50.94).
Volume 17—Number 8 661
Richard B. Slatcher and James W. Pennebaker
Note that the follow-up satisfaction measures are difﬁcult to
interpret because they were available only for couples still in
Hierarchical linear modeling (HLM; Raudenbush & Bryk,
2002) was employed to analyze changes in use of emotion words
within couples’ IM conversations as a function of experimental
condition. Initial analyses focused on the effects of expressive
writing on positive emotion words in couples’ conversations.
Level 1 (individual-level) predictors were baseline use of pos-
itive emotion words and speaker (participant or partner). Only
baseline use of positive emotion words was a significant Level 1
>.99, d51.19. Level 2 (couple-level)
predictors were gender of the participant, experimental condi-
tion, and a Gender Condition interaction term. Only experi-
mental condition was a significant Level 2 predictor, g
5.93, d50.46. Couples in the expressive-writing condi-
tion were more likely than those in the control condition to in-
crease their use of positive emotion words with their romantic
partners in the days following the manipulation. Additionally,
participants and partners in the experimental condition in-
creased their use of positive emotion words at similar rates as
they communicated with each other via IM, even though only the
participants were involved in the expressive-writing manipu-
Next, the effect of expressive writing on negative emotion
words was analyzed. Level 1 predictors were baseline use of
negative emotion words, as well as the speaker variable. Only
baseline use of negative emotion words was a significant Level 1
5.98, d50.46. Level 2 predictors were
gender of the participant, experimental condition, and a Gender
Condition interaction term. Significant Level 2 predictors
were experimental condition, g
and the Gender Condition interaction, g
d50.61. Compared with control couples, couples in the ex-
pressive-writing condition were more likely to increase their
use of negative emotion words with their romantic partners in the
days following the writing manipulation. This effect was mod-
erated by the gender of the person in each couple who took part
in the writing manipulation. Specifically, couples in which the
male wrote about the relationship increased their use of negative
emotion words significantly compared with control couples,
whereas couples in which the female wrote about the relation-
ship did not change in their use of negative emotion words
compared with control couples.
Mediation Effects of Changes in Use of Emotion Words
A strength of this design is that it allowed us to explore the degree
to which measures of language may reﬂect the social psycho-
logical processes underlying the effects of expressive writing.
Two sets of mediation analyses were conducted—one with pos-
itive emotion words as the potential mediator, the other with
negative emotion words as the potential mediator. In both cases,
the outcome measure was relationship stability at the 3-month
follow-up. Means and standard deviations for couples’ use of
emotion words in their IM conversations are shown in Table 1.
Positive Emotion Words
Experimental condition and use of positive emotion words be-
fore and after the writing manipulation were entered in separate
steps into a logistic regression in which relationship stability
was the dependent variable. Higher levels of postwriting posi-
tive emotion words were associated with higher levels of couple
stability, odds ratio 53.04, p
5.99, d50.95. After con-
trolling for changes in positive emotion words, the effects of
experimental condition on relationship stability were no longer
significant, suggesting possible mediation. The results of a Sobel
ztest (Sobel, 1982) supported this analysis (z52.00, p
Thus, the salutary effects of writing about one’s relationship
were at least partially mediated by increases in use of positive
emotion words in daily IM conversations.
Negative Emotion Words
No relationship was found between increases in negative emo-
tion words and couple stability (p
<.50). This also was the
case in a separate analysis for just those couples in which the
male was the participant. The effects of expressive writing on
stability therefore were not mediated by changes in use of
negative emotion words.
The relatively simple act of writing about their romantic rela-
tionship changed the way in which participants communicated
Couples’ Use of Emotion Words in InstantMessages as a Function
of Time and Condition
Positive emotion words
Experimental condition 4.51 1.29 4.94 1.54
Control condition 4.39 1.42 4.31 1.36
Negative emotion words
Experimental condition 1.84 0.61 2.36 0.96
Control condition 1.91 0.82 1.97 0.93
Note. Experimental condition n544; control condition n542.
Exploratory analyses also were conducted on changes in use of ﬁrst-
person plural pronouns (e.g., we,us,our); no significant effects were found.
Previous studies (e.g., Sillars et al., 1997; Wegner, 1982) have found positive
correlations between relationship functioning and use of we when participants
discussed their relationships with an interviewer. Note that this is a very different
context from closed interactions between the two members of a couple, during
which the use of we may not reﬂect intimacy between them (e.g., it may reﬂect
condescension, emotional distancing, or the referencing of someone close out-
side the relationship).
662 Volume 17—Number 8
Social Effects of Expressive Writing
with their partners in IM conversations; it also changed the way
in which the partners communicated with the participants and
improved relationship stability. Couples in the expressive-
writing condition were more likely to increase their use of pos-
itive emotion words than were couples in the control condition;
couples in the expressive-writing condition in which the male
was the participant increased in their use of negative emotion
words as well. Increases in positive emotion words partially
mediated the association between writing and relationship
Taken together, these ﬁndings shed light on processes un-
derlying interactions in close relationships. In particular, the
ﬁndings relating to increases in use of emotion words illuminate
previous research (e.g., Butler et al., 2003; Gottman & Leven-
son, 2000; Laurenceau, Barrett, & Pietromonaco, 1998) sug-
gesting that increased expression of positive emotions can result
in better outcomes for relationships. This study is one of the ﬁrst
to go beyond self-reports to demonstrate that couples’ increased
emotional expressiveness may have the power to improve
objective relationship outcomes.
An advantage of the current design is that it allows one to
unobtrusively measure interpersonal processes underlying
dyadic interactions. By analyzing the words that couples use in
daily IM interactions, researchers can assess the extent to which
the couples’ language patterns are predictive of relationship
quality and stability. Given the growth in electronic communi-
cation in recent years, IM technology may serve as a promising
tool in examining real-time, on-line interactions in naturalistic
That people may enhance their romantic relationships by
simply writing down their thoughts and feelings about those
relationships has clear implications for clinicians. The use of
expressive writing as a tool for relationship enhancement could
be applied to a broad range of relationships, including those in
families, circles of friends, and even work groups. Expressive
writing may serve to strengthen the relational connections of a
broad array of social channels, particularly for persons who have
not had extensive experience expressing emotions to others.
There are some potential limitations of this study. First, it is
conceivable that the effects of writing on changes in word use
may have been partially affected by demand characteristics. We
believe that the emphasis on keeping the purpose of the study
vague to participants and the fact that writing also inﬂuenced
relationship stability make this an unlikely possibility. Second,
it is unknown whether it was the act of writing itself that led to
positive relationship outcomes, or whether simply directing
people to mentally attend to and explore relationship issues
would be equally beneﬁcial; future studies should address this
Unlike previous expressive-writing studies, this is the ﬁrst to
demonstrate some of the social processes that may underlie the
effects of expressive writing. Further, this study points to the
advantages gained in using current on-line technologies such as
IM for psychological research. Such technologies now allow
researchers to examine natural interactions in a relatively
simple, inexpensive, and straightforward manner.
Acknowledgments—Portions of this research were funded by a
grant from the National Institutes of Health (MH52391). We
would like to thank Greg Hixon, Amy Kaderka, and Girish
Tembe for their assistance on this project and Amie Green,
Timothy Loving, Matthew Newman, William Swann, and Simine
Vazire for their helpful comments on an earlier draft of this
Butler, E.A., Egloff, B., Wilhelm, F.H., Smith, N.C., Erickson, E.A., &
Gross, J.J. (2003). The social consequences of expressive sup-
pression. Emotion,3, 48–67.
Gordon, K.C., Baucom, D.H., & Snyder, D.K. (2004). An integrative
intervention for promoting recovery from extramarital affairs.
Journal of Marital and Family Therapy,30, 213–231.
Gottman, J.M., & Levenson, R.W. (2000). The timing of divorce: Pre-
dicting when a couple will divorce over a 14-year period. Journal
of Marriage and the Family,62, 737–745.
Hendrick, S.S. (1988). A generic measure of relationship satisfaction.
Journal of Marriage and the Family,50, 93–98.
Laurenceau, J.-P., Barrett, L.F., & Pietromonaco, P.R. (1998). Intimacy
as an interpersonal process: The importance of self-disclosure,
partner disclosure, and perceived partner responsiveness in in-
terpersonal exchanges. Journal of Personality and Social Psy-
Lepore, S.J., & Greenberg, M.A. (2002). Mending broken hearts: Ef-
fects of expressive writing on mood, cognitive processing, social
adjustment and health following a relationship breakup. Psy-
chology and Health,17, 547–560.
Lepore, S.J., & Smyth, J. (2002). The writing cure. Washington, DC:
American Psychological Association.
Pennebaker, J.W., Francis, M.E., & Booth, R.J. (2001). Linguistic In-
quiry and Word Count (LIWC). Mahwah, NJ: Erlbaum.
Pennebaker, J.W., & Graybeal, A. (2001). Patterns of natural language
use: Disclosure, personality, and social integration. Current Di-
rections in Psychological Science,10, 90–93.
Pennebaker, J.W., Mehl, M.R., & Niederhoffer, K. (2003). Psycholog-
ical aspects of natural language use: Our words, our selves. Annual
Review of Psychology,54, 547–577.
Raudenbush, S.W., & Bryk, A.S. (2002). Hierarchical linear models:
Applications and data analysis methods (2nd ed.). Newbury Park,
Shiu, E., & Lenhart, A. (2004). How Americans use instant messaging.
Retrieved October 14, 2005, from http://www.pewinternet.org/
Sillars, A.L., Shellen, W., McIntosh, A., & Pomegranate, M.A. (1997).
Relational characteristics of language: Elaboration and differen-
tiation in marital conversations. Western Journal of Communica-
Smyth, J.M. (1998). Written emotional expression: Effect sizes, outcome
types, and moderating variables. Journal of Consulting and
Clinical Psychology,66, 174–184.
Volume 17—Number 8 663
Richard B. Slatcher and James W. Pennebaker
Sobel, M.E. (1982). Asymptotic intervals for indirect effects in struc-
tural equations models. In S. Leinhart (Ed.), Sociological meth-
odology 1982 (pp. 290–312). San Francisco: Jossey-Bass.
Twenge, J.M., Catanese, K.R., & Baumeister, R.F. (2003). Social ex-
clusion and the deconstructed state: Time perception, meaning-
lessness, lethargy, lack of emotion, and self-awareness. Journal of
Personality and Social Psychology,85, 409–423.
Wegner, D.M. (1982, August). When does the intimate group come to
mind? Paper presented at the annual meeting of the American
Psychological Association, Washington, DC.
(RECEIVED 6/16/05; REVISION ACCEPTED 12/21/05;
FINAL MATERIALS RECEIVED 1/17/06)
664 Volume 17—Number 8
Social Effects of Expressive Writing