back to projects

Healing Words:  Textual Analysis of 

Picasso, Nu De DosAutobiographical Accounts of Trauma  


Sarah A. Novak


The University of Texas at Austin


Experimental Psychology Final Paper  





            Previous studies have found particular language patterns to be indicative of improved physical and mental health for research participants whose writing disclosed intense personal trauma (Pennebaker & Seagal, 1999).  Health improvements have not been seen in writers whose topics were trivial.  This study sought to determine whether professional writers’ works would follow these patterns as well.  The key indicators include an increase and relatively high level of positive emotion, a steady, moderate amount of negative emotion, and an increase in insight words over time.   A text-analysis computer program called LIWC was used to analyze and compare autobiographical works about trauma to works that are not about trauma by the same authors.  It was hypothesized that professional writers who discuss their own trauma in autobiographical form will show these indicators of health more than they would in works on other subjects.  A 2x6x3 (type of work x section x indicator) within-subjects design was employed to compare the results of the text-analysis.  The results show that between the two different books by each author, there were not significant overall differences in either the percentages of positive emotion words or negative emotion words, but the level of insight words had a near-significant increase in the autobiographical works.  These results did not support the hypothesis, but under less stringent standards, the finding for insight would reflect the prediction.  It was concluded that published autobiographical works about trauma are not as similar to data collected from research subjects on key linguistic factors that have been shown to predict improved health.  This implies that for professional writers, disclosing trauma may not be as therapeutic physically and mentally as has been observed under experimental conditions.  

Healing Words: Textual Analysis of Autobiographical Accounts of Trauma  

            The concept of disclosing traumatic experiences in written form has been approached from both literary and psychological perspectives.  The autobiographical and semi-autobiographical works by authors who have experienced personal trauma are at the intersection of these different fields.  The purpose of the current investigation is to quantitatively integrate findings from both psychological and literary analysis to show to what degree autobiographies containing elements of trauma are consistent with the psychological evidence that when disclosure writing contains particular linguistic patterns, it is much more likely to be a therapeutic process that can lead to improvements in physical and mental health.

            Numerous studies have demonstrated the value of writing as a form of therapy.  Pennebaker and Francis (1996) demonstrated that writing about deeply emotional topics for as little as ten minutes for three to five days can have an impact on physical health for months after the writing has occurred.  Their studies have been theoretically based on and supportive of the inhibition-confrontation model which holds that inhibiting emotions and thoughts is mentally and physically stressful and can result in negative impact on physical health (Pennebaker, Colder & Sharp, 1990).  Additionally, experiences that are not understood and assimilated cognitively can be especially problematic for individuals, particularly for young victims of trauma.  This is more common when traumatic experiences are not discussed and therefore not organized into language.  Reasons for not discussing experiences include being unable or unwilling, and fear of blame or humiliation frequently plays a part (Pennebaker & Seagal, 1999).  In studies, many of the subjects who wrote or talked about major upheavals such as adjusting to college or losing a job showed greater cognitive changes as they progressed with the writing assignments.  Subjects have reported that writing forced them to think about events differently (Pennebaker & Francis, 1996).  Pennebaker and Seagal (1999) have suggested that the underlying mechanism for the observed benefits is the process of re-evaluating and organizing a disjointed, problematic experience into a coherent, insightful narrative.

Writing about traumatic experiences has been demonstrated to improve health in specific situations (i.e. medical students under pressure), but similar effects have been seen in studies of non-specific experiences in which participants are simply instructed to write about any traumatic experience that has deeply affected them (Pennebaker & Seagal, 1999).  Because of this finding, the current investigation will be able to include the stories of both men and women who have coped with a variety of traumatic experiences ranging from the Vietnam War to incest.  The wide variety of autobiographical accounts of trauma reflects the prevalence of this form of writing and its popularity in literary discourse.

            In literary analysis, some critics focus mainly on fitting autobiographies into political and social movements (e.g. Tal, 1996).  However, in so-doing, these authors site reasons to explain why survivors of sexual abuse write about their trauma including the feminist movement, political statements, and a desire to place public blame on the aggressor.  Meanwhile, they neglect motivations such as psychological healing.  Henke (1998) has put forth a model for understanding autobiographies of trauma that incorporates historical relevance as well as theories of writing as an instinctive step toward healing body and mind.  She sites prominent psychological theories to support her assertions and blends them into her literary analysis.  By using actual published autobiographies and thus, entering the typical realm of literary criticism, the present study seeks to further legitimate Henke’s arguments and substantiate her claims that autobiographical writing is therapeutic by providing quantitative evidence from textual analysis.

            The goal of this study is to bridge the gap between literary and psychological understanding of autobiographical works.  It will be determined whether published autobiographical accounts of trauma are representative of narratives written by trauma survivors in the course of psychological research.  This link between these fields will be made by way of Linguistic Inquiry and Word Count (LIWC) analysis.  Pennebaker and Francis (1999) created the computer program to perform textual analyses on 72 different variables tabulating linguistic dimensions, psychological constructs, relativity, and personal concerns.  Studies have established LIWC’s external validity and demonstrate that the program can be used to quantify positive and negative emotions, cognitive strategies, thematic content, and elements of language composition. 

As seen in a number of experiments, text analyses have shown that there are three main linguistic patterns used by those who benefit from writing about deeply emotional experiences (Pennebaker, 1997).  The first of these is a generally high and occasionally increasing amount of positive emotion words (e.g. happy, laugh, love, and good). The second indicator is a moderate amount of negative emotion words (e.g. hurt, ugly, and angry).  Those who use extremely high or low amounts of negative words were more likely to experience worse health after writing.   The greatest effects were demonstrated by an increase the number of cognitive words (e.g. understand, know, and realize) throughout the writing sessions, showing that thought and reflection are occurring at higher levels by the end of the sessions (Pennebaker, 1997).  Thus, if the analyses of published autobiographies on trauma reflect the patterns established by others who have experienced health benefits through writing, the hypothesis will be confirmed and the autobiographies will further support the relationship between writing about trauma and improved health.



            A total of 3 pairs of books were sampled for this experiment.  One book by each author was an autobiography that included traumatic experiences, while the other was on a different, unrelated topic.  The autobiographies served as experimental subjects while the unrelated books provided controls.  The list of authors and books in each category is provided in the Appendix.

            The LIWC was used for textual analysis.  Of the 74 preset dimensions, only 3 (positive emotion, negative emotion, and insight words) were highlighted in the results.  The program compares text to its dictionary, word by word, and words that are identified are then categorized and counted.  The categories into which any given word is placed were determined by agreement between independent judges (Pennebaker & Francis, 1999). 


The six books were selected by the following criteria: 1) the author experienced some form of serious personal trauma, 2) the author wrote a book unrelated to trauma before writing about trauma, and 3) the author published an autobiography describing the personal trauma.  One autobiography and one previously written, unrelated work were selected from each author.  Each work was divided into six equal sections from beginning to end.  A sample of approximately eight pages from the latter portion of each section was scanned into text file format and analyzed on 74 dimensions using the LIWC computer program. 


            Writing about trauma did not significantly affect the number of positive or negative emotion words, although insight words did show a slightly increasing trend, particularly within the last sections.  Analysis of variance determined that there were no significant effects for positive or negative emotion words.  Figure 1 shows that there was not a significant increase between the percentage of positive emotion words in the unrelated book and the book on trauma, F(5,30) = .75, p = .59), and Figure 2 shows no significant difference for negative emotion words, F(5,30) = .979, p = .447). 


                                                Insert Figure 1 about here

                                                Insert Figure 2 about here


Although the difference in insight words was not statistically significant, F(5,30) = 2.2, p = .08), it is much closer than the other results.  As shown in Figure 3, section six is a particularly clear example of the expected difference in insight.


                                                Insert Figure 3 about here



            The hypothesis that the linguistic patterns of autobiographical works about trauma would resemble those of research participants who benefited physically and mentally from writing, was not supported by the results.  This suggests that perhaps Henke (1998) is incorrect in asserting that writing about trauma, professionally or not, is intuitively therapeutic.  There may be differences between professional writers and previous study participants that had not been considered before.  Because this study did not use research participants and instead relied on published works, a link has not been demonstrated between the literary world and the phenomenon elucidated in psychological research

            A number of considerations must be made when generalizing these results.  First and foremost, a sample of three authors can hardly represent the great diversity found in the entire population.  In addition, study that used a greater number of works from each author could helpful in further determination of any trends. Perhaps the act of writing several books changes the way authors think about events in their own life.  This might make them more capable of readily organizing their thoughts into a narrative when compared to non-authors.  Another possibility is that the strong propensity to create narratives out of both life experience and fiction is what both motivates individuals to become authors and distinguishes them from other research participants.

Each pair of works had the same order of publishing – unrelated work before autobiography.  Perhaps the process of writing many books over time diminishes the effects seen in studies of non-writers.  Looking at each author’s complete body of work might help explain whether additional writing has an effect on trends observed in other studies.  Also, the experimental studies have looked only at the writing process over a few days.  Perhaps looking at a longer, finished work is not the same as analyzing the first stages of the writing process.  This study was also done with the assumption that authors write books from beginning to end, and that publishing dates reflect the order in which separate books are written.  However, this may not always be the case, and the true order of writing may help explain why previous research findings were not confirmed in this study. 

            Previous studies have suggested reasons for the ability or inability to benefit from writing about trauma.  Pennebaker (1997) states that most of the people who do not benefit are unable to re-evaluate their painful experiences, and instead ruminate over them without reaching any conclusion.  This does not mean that they do not construct logical, coherent narratives, it means that they never revise their story and do not gain insight from the experience (Pennebaker & Seagal, 1999).  In the present study, the fact that the authors did not reflect the trends found in previous investigations may indicate that these authors may be the type of people who are ruminating.  Thus, these individuals would not be expected to show the physical and mental benefits from writing that others have demonstrated.  However, the one near-trend that was found was for increasing insight toward the end of the books about trauma.  This seems to show that writing about their own experiences gave these writers more insight and clarity than when writing fictional stories.  Further experiments on more authors may reveal that this small difference in insight does play a part in improving health among professional writers.


Henke, S. (1998).  Shattered Subjects: Trauma and Testimony in Women’s Life-Writing.  New York: St. Martin’s Press.  

Pennebaker, J.W. (1997).  Opening Up: The Healing Power of Expressing Emotions.  New York: The Guilford Press.


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. & Francis, M.E. (1999).  Linguistic Inquiry and Word Count.  Mahwah, NJ: Lawrence Erlbaum Associates, Inc.


Pennebaker, J.W. & Seagal, J.D. (1999).  Forming a Story: the health benefits of narrative.  Journal of Clinical Psychology, 55, 1243-1254.


Tal, K. (1996).  Worlds of Hurt.  Cambridge, MA: Cambridge University Press.


Author Unrelated Work Autobiography
Allison, Dorothy Gospel Song (1988) Bastard Out of Carolina (1992)
Duras, Marguerite  Madame Dodin (1954) The Lover (1984)  
Harrison, Kathryn   Poison (1995)   The Kiss (1997)  

Figure Captions

Figure 1.  The percentage of positive emotion words used across the six sections.  The first bar for each author represents the autobiographical work about trauma (Auto), while the second represents a work by the same author that is not related (Not). 

Figure 2.  The percentage of negative emotion words used across the six sections.  The first bar for each author represents the autobiographical work about trauma (Auto), while the second represents a work by the same author that is not related (Not). 

Figure 3.  The percentage of insight words used across the six sections.  The first bar for each author represents the autobiographical work about trauma (Auto), while the second represents a work by the same author that is not related (Not).