Content uploaded by Sheldon Cohen
Author content
All content in this area was uploaded by Sheldon Cohen on Jan 30, 2014
Content may be subject to copyright.
Content uploaded by Sheldon Cohen
Author content
All content in this area was uploaded by Sheldon Cohen on Jan 30, 2014
Content may be subject to copyright.
Content uploaded by Sarah D Pressman
Author content
All content in this area was uploaded by Sarah D Pressman
Content may be subject to copyright.
Health Psychology
Positive Emotion Word Use and Longevity in Famous
Deceased Psychologists
Sarah D. Pressman and Sheldon Cohen
Online First Publication, September 19, 2011. doi: 10.1037/a0025339
CITATION
Pressman, S. D., & Cohen, S. (2011, September 19). Positive Emotion Word Use and
Longevity in Famous Deceased Psychologists. Health Psychology. Advance online
publication. doi: 10.1037/a0025339
Positive Emotion Word Use and Longevity in Famous
Deceased Psychologists
Sarah D. Pressman
University of Kansas
Sheldon Cohen
Carnegie Mellon University
Objective: This study examined whether specific types of positive and negative emotional words used in
the autobiographies of well-known deceased psychologists were associated with longevity. Methods: For
each of the 88 psychologists, the percent of emotional words used in writing was calculated and
categorized by valence (positive or negative) and arousal (activated [e.g., lively, anxious] or not activated
[e.g., calm, drowsy]) based on existing emotion scales and models of emotion categorization. Results:
After controlling for sex, year of publication, health (based on disclosed illness in autobiography), native
language, and year of birth, the use of more activated positive emotional words (e.g., lively, vigorous,
attentive, humorous) was associated with increased longevity. Negative terms (e.g., angry, afraid,
drowsy, sluggish) and unactivated positive terms (e.g., peaceful, calm) were not related to longevity. The
association of activated positive emotions with longevity was also independent of words indicative of
social integration, optimism, and the other affect/activation categories. Conclusions: Results indicate that
in writing, not every type of emotion correlates with longevity and that there may be value to considering
different categories beyond emotional valence in health relevant outcomes.
Keywords: emotions, longevity, arousal, text analysis, positive affect
Emotion has long been held to be an important determinant of
health and disease (e.g., Hippocrates 466 –375 B.C.), however, the
focus has almost exclusively been on the role of negative feelings
in detrimental health outcomes. Only recently has attention moved
toward the influence of positive emotions on health and well-being
(e.g., Chida & Steptoe, 2008; Fredrickson, 1998; Pressman &
Cohen, 2005; Seligman & Csikszentmihalyi, 2000). To date, the
strongest and most provocative evidence has been that connecting
“trait” emotional styles (reflecting a person’s typical emotional
experience) to longevity with findings revealing benefits of 2 to 23
extra years of life in those individuals with higher levels of
positive emotional style (PES) (e.g., Levy, Slade, Kunkel, & Kasl,
2002; Maier & Smith, 1999; Ostir, Markides, Black, & Goodwin,
2000; Parker, Thorslund, & Nordstrom, 1992). Although these
associations typically withstood statistical controls for an array of
potential confounds (e.g., baseline health, age, sex, race, socioeco-
nomic status), this is still considered only a suggestive literature
with many conceptual and methodological limitations (Cohen &
Pressman, 2006; Pressman & Cohen, 2005).
One seriously neglected issue in the field has to do with the
types of positive emotions that may influence health. Existing
work merely differentiates between positive and negative emo-
tions; however, there are many other characteristics to consider.
One critical typology of affect for health researchers is that of
arousal, which is thought to be a primary pathway linking emo-
tional experiences to health (Cohen, Kessler, & Gordon, 1995;
Krantz, Glass, Contrada, & Miller, 1981; Pressman & Cohen,
2005). Circumplex theories are one example of models of emotion
that place importance on arousal by using two orthogonal dimen-
sions, activation and valence, as the central components of affect
(e.g., Russell, 1980). Affect adjectives can be categorized by
where they fall on a two-dimensional plane resulting in four
quadrants (high activated positive affect [e.g., excitement], unac-
tivated positive affect [e.g., calm], high activated negative affect
[e.g., anxiety], and unactivated negative affect [e.g., sadness]).
Similarly, Thayer (1967, 1970) factor analyzed mood adjectives
into an Activation-Deactivation check list with the goal of repre-
senting physiological arousal (e.g., skin conductance, heart rate)
via self-report. Emotions were categorized as “General Activation”
(e.g., full-of-pep, energetic), “High Activation” (e.g., fearful, jit-
tery), “General Deactivation” (leisurely, calm), “Deactivation-
Sleep” (e.g., drowsy, tired). While not specifically delineated by
the author, these subscales map on to the circumplex dimensions.
This issue remains relevant today given that the Positive Affect
and Negative Affect Schedule (PANAS; Watson & Clark, 1988),
one of the most commonly used emotions scales in the health
psychology field, only assesses activated adjectives creating a
concern if unactivated emotions are an important contributor to
health.
Sarah D. Pressman, Department of Psychology, University of Kansas;
Sheldon Cohen, Department of Psychology, Carnegie Mellon University.
Research reported here was funded by a grant to the Pittsburgh NIH
Mind-Body Center (HL65111 and HL65112). We thank Brian Potetz for
his work on the WordCount program and Michal Kollnesher for her
assistance in data collection. Finally, the authors would also like to express
their appreciation to Peggy Clark, Vicki Helgeson, Anna Marsland, and
Joel Greenhouse for their comments and assistance on the dissertation
work that was the basis for this paper and Emily Hooker and Denise
Janicki-Deverts for their comments on the final manuscript.
Correspondence concerning this article should be addressed to Sarah D.
Pressman, Department of Psychology, University of Kansas, 1415 Jayhawk
Blvd, Lawrence, KS 66045. E-mail: pressman@ku.edu
THIS ARTICLE HAS BEEN CORRECTED. SEE LAST PAGE
Health Psychology © 2011 American Psychological Association
2011, Vol. ●●, No. ●, 000– 000 0278-6133/11/$12.00 DOI: 10.1037/a0025339
1
Another significant concern in the emotion-longevity literature
arises from the (moderate) negative correlation generally found
between positive and negative emotions. Because greater levels of
negative emotional style (NES) have been frequently associated
with poorer health outcomes (see reviews by Carney, Freedland,
Miller, & Jaffe, 2002; Herbert & Cohen, 1993; Kiecolt-Glaser,
McGuire, Robles, & Glaser, 2002; Krantz & McCeney, 2002), it is
possible that the reported association of PES and increased lon-
gevity may merely be a product of persons with higher PES having
lower NES. Existing studies of emotional styles and health gen-
erally include one or the other measure making it impossible to test
whether associations with health are attributable to variations in
NES, PES, or both.
There is also concern that measures of PES in this literature may
be markers of social and/or positive cognitive dispositions such as
sociability and optimism, which are also thought to be important
predictors of health outcomes (Cohen, Alper, Doyle, Treanor, &
Turner, 2006; Pressman & Cohen, 2005; Scheier, Carver &
Bridges, 1994). In general, these factors have moderate associa-
tions with PES, but few existing studies control for the possibility
that they and not PES are responsible for reported associations of
PES and health.
Finally, most studies on the role of emotional styles in health
base scores on self-reports of emotions. Because it is socially
desirable to report more positive emotions (e.g., Diener & Suh,
1999) and self-reports of affect can be influenced by response
styles (Lucas, Diener, & Larsen, 2003) it is advantageous to use
other modes of affect measurement when possible (Diener, Suh,
Lucas, & Smith, 1999). Assessing the components of writing has
been proposed as an independent and meaningful way to study
various trait differences (McAdams, 2001; Pennebaker & King,
1999), and when collected outside of a research context, may result
in less self-presentation bias (Fazio & Olson, 2003). This proce-
dure was used by Danner, Snowdon, and Friesen (2001) who hand
coded short autobiographies written by a group of nuns when they
were in their 20s. They found that more positive words in the
writing samples were associated with lower mortality 65 years
later but that negative words were not associated with longevity.
Although extremely provocative, the generalizability of this study
is uncertain given that the sample was made up of women belong-
ing to a religious order and interpretation is complicated by the
lack of health data at baseline. Moreover, their dictionaries distin-
guished word based on valence but not activation. More recent
studies of emotion word usage in text have used the Linguistic
Inquiry and Word Count program (LIWC; e.g., Pennebaker, Fran-
cis, & Booth, 2001) because of the program’s ability to quickly
and automatically calculate the different categories of words used
in digitized texts although the standard emotions word lists utilized
(i.e., dictionaries) do not differentiate words on activation level
and include adjectives and behaviors in addition to emotions. This
makes the existing LIWC emotion dictionaries unsuitable for the
specific types of affect questions remaining in the PES-health
literature.
The current study assesses emotional styles using scales drawn
from established emotion measures that tap both the valence and
activation dimensions of affect and automated word counting
technology. The primary measure is consistent with a circumplex
model of emotion and included words falling into quadrant com-
binations of positive/negative valence and activation/unactivation
(Mackay, Cox, Burrows, & Lazzerini, 1978). For convergent va-
lidity and comparative purposes, the expanded version of the
PANAS (Watson & Clark, 1994) was utilized to examine other
variants of emotion not included in the Mackay scale and the
LIWC was used to contrast general affect approaches from this
study’s more specific emotion technique. Finally, the LIWC affect
subscales (sadness, anxiety, and anger) were included along with
an author-created humor scale since this is a high activation
emotion type that has been previously tied to health (e.g., Martin,
2001) but was untapped by other scales. This study tests the
potential independent association of different types of emotion
with longevity, and whether any found effects can be accounted for
by demographics, native language, or emotion–related constructs.
While objective health data at baseline was not available, health
was controlled by examining the use of words describing self-
ascribed illness.
Method
Participants
The sample included 88 influential researchers who each con-
tributed an autobiographical piece to a seven volume series entitled
“A History of Psychology in Autobiography” (e.g., Murchison,
1961). Individuals were chosen because of their great influence on
different areas of contemporary psychology with some falling
within the mainstream of the field and others picked from outside
the field because they contributed to moving psychological re-
search and theory in a new direction. The subjects’ names are
listed in the appendix. To obtain a sample of emotion words that
broadly reflect life experience (and is not confounded with how
many words an author used to discuss his or her work), analyses
were restricted to the personal segments of the autobiographies.
Out of 96 possible subjects, seven were excluded because they
omitted personal information and focused solely on research. One
additional subject was excluded for being alive at the time of
analysis. In the final sample, individuals were born between 1843
and 1926, lived an average of 79 years, were 66.7 at the age of
publication of their autobiographies, and lived on average 12.3
years postpublication. The final sample was primarily male (three
females) and entirely White.
Biographic and Demographic Information
Biographic (birth date and death date) and demographic (sex,
race) information was collected from the autobiographies them-
selves, online encyclopedias, obituaries, and psychology depart-
ments. Whenever possible, more than one source was used to
provide this information. Year of publication of each volume was
used to estimate age of the author when the autobiography was
written. Longevity was defined as the number of days between the
date of birth and date of death.
Word Counts
Several different word count measures were used. In all cases
“word counts” represent the number of times words from a given
subcategory (dictionary) were mentioned divided by the total num-
ber of words in the text resulting in a percentage score.
2PRESSMAN AND COHEN
Measures of activation and emotional style. Word count
dictionaries were created using items directly from a model of
affect designed to assess self-reported activation (Mackay, Cox,
Burrows, & Lazzerini, 1978). This 34-item scale was based on
factor analyses of previous scales designed to assess the arousal
continuum within mood adjective checklists (Nowlis, 1965;
Thayer, 1967). While the words were originally differentiated only
on the arousal continuum, for the purpose of this paper, words
were assigned into categories based on positive or negative va-
lence and low or high activation. Activated and unactivated pos-
itive words were each assessed with eight items (active, energetic,
vigorous, alert, lively, activated, stimulated, and aroused for acti-
vated PES; peaceful, relaxed, cheerful, contented, pleasant, com-
fortable, calm, and restful for unactivated PES). Activated negative
words were measured via 11 items (tense, worried, apprehensive,
bothered, uneasy, dejected, uptight, jittery, nervous, distressed,
fearful) while unactivated negative words were based on seven
items (drowsy, tired, idle, sluggish, sleepy, somnolent, passive). In
all dictionaries, when the emotion adjectives had more than one
tense of use, word stems were used to account for all versions of
the adjective (e.g., “relax” stem would include relax, relaxed,
relaxation etc.).
Coding dictionaries were also created by using the exact items
from the PANAS-X (Watson & Clark, 1994). Items in the PANAS
are generally considered to represent “activated” emotions and are
assessed with 10 items each (negative affect [NA]: e.g., afraid,
guilty, upset; positive affect [PA]: e.g., active, enthusiastic, in-
spired). Also included were the seven basic emotions subscales
from the PANAS-X including self-assurance (6 items; e.g., proud,
confident), attentiveness (4 items; e.g., alert, determined), joviality
(8 items; e.g., cheerful, happy) for the positive subscales, and fear
(6 items; e.g., afraid, nervous), hostility (6 items; e.g., angry,
irritable), guilt (6 items; e.g., guilty, ashamed), and sadness (6
items; e.g., lonely, downhearted) for the negative subscales. Sub-
scales have minimal item overlap with the general scales. As is
evident from the words used, the specific positive subscales are all
activated. The negative subscales are a mix of activated (fear,
hostility, guilt) and unactivated (sadness) items.
It was important to use a measure that has previously been used
in linguistic analysis and one that contained a larger number of
words, therefore the LIWC 2001 measures of emotion were also
included (Pennebaker et al., 2001). Positive (261 words) and
negative (435 words) emotion scores are not differentiated on
activation and include some nonemotion words (e.g., adjectives,
behaviors). Importantly, the majority of the words from the shorter
affect scales described above were included within these broader
LIWC scales. The LIWC negative emotion subscales were also
included that are: anxiety (62 words; e.g., afraid, sickened, vul-
nerable), anger (121 words; e.g., wicked, temper, ugly, fight, hate),
and sadness (72 words; e.g., low, useless, weep, lonely). There are
no LIWC positive emotion subscales, however, a humor/laughter
dictionary was created by the researchers based on 10 terms drawn
from the LIWC positive emotions (chuckle, laugh, funny, humor,
giggle, hilarious, fun, hilarity, jolly, silly). Please refer to Figure 1
for a list of all emotions measures and their theorized activation
levels.
Cognitive and social measures. Two other concepts that are
often closely associated with positive emotions were assessed:
optimism and social integration. Optimism was measured using the
optimism dictionary from the LIWC which includes a number of
cognitive, emotional, and descriptive adjectives (69 items: e.g.,
bold, certainty, hope, optimism, pride, superior, win). Social inte-
gration/sociability was measured with a social role index (103
terms) which included words such as husband, wife, family, friend,
Figure 1. The theorized arousal levels of the different emotions scales utilized in this study. Word lists (directly
derived from existing affect measures) are placed on a scale of high to low arousal and positive to negative
valence. This model parallels existing circumplex models of affect (e.g., Russell, 1980) although the location of
the specific affect measures is not based on data from these exact scales. Locations of affect types within the
clusters are not meant to be informative but only the general location within each quadrant (high, middle, low).
3
POSITIVE EMOTION WORDS & LONGEVITY
partner, grandparent derived from an existing writing measure of
social networks (Pressman & Cohen, 2007).
Participant Characteristics
Illness. Each autobiography was manually read and tabulated
for mentions of illnesses that were suffered by the author (e.g.,
sick, cancer, hypertension, ill, emphysema, disease, diabetes, heart
attack, angina, stroke, ulcer, epilepsy). The assumption was that
individuals who had suffered from a serious disease would write
about it. Manual counting was used on only this occasion because
computerized counting would capture the illness occurrences of
others.
Language. It is feasible that the native language of the author
could influence emotional expression in writing. A variable that
indicated whether the life story was translated was created (N⫽20
translated).
Procedure and Apparatus
Since the current questions focused on emotional word use, text
analyses were done on personal segments of the autobiographies.
Specifically, the sections devoted to childhood, education, family,
culture, and growing up as opposed to sections focused on later
research. This was necessary because several subjects study emo-
tions (confounding the predictor variables of interest) and also
because research was discussed in a less emotional fashion.
Research assistants identified these personal sections from each
autobiography then scanned them into an image file (PDF/TIF)
using a Gestetner 7502 Xerox machine with 1,200 ⫻1,200 dots
per inch (dpi) resolution. Autobiographies were then analyzed with
ABBYY Sprint 5.0 object character recognition technology. Each
autobiography was proofread for scan errors and to remove super-
fluous information (e.g., page numbers, captions, footnotes, head-
ers). A word count software program was developed to calculate
the extent to which different categories of words were used. The
“Word Count” program counts the total number of words as well
as the frequency of words and words stems in a specified diction-
ary.
Autobiography personal content ranged widely from 280 words
to 9,814 words (M⫽1,862.6; SD ⫽1,405.4 words), however,
total words were not associated with longevity (p⫽.97). Al-
though not statistically significant, the seven excluded participants
without personal content in their writing samples (discussed ear-
lier) lived on average, 2.8 fewer years than their personal-content
included counterparts (F⫽.578, p⫽.46).
Primary Statistical Approach
The primary analytical technique for this study was to examine
whether different types of emotion categories (Independent Vari-
ables [IVs]) were associated with longevity after controlling for
the appropriate covariates in Hierarchical Multiple Regression
models. This approach tested three types of affect variables as the
IVs: (1) The MacKay scale with separate testing of the four
valence and activation quadrants; (2) PANAS measures, and (3)
LIWC emotion scales as the IVs. The goal of these analyses was
to determine which types of emotion were tied to longevity. A
secondary analytical question was whether the negative and pos-
itive valence measures had similar explanatory power and whether
they were independent of one another. This was tested by includ-
ing positive and negative variables in the same regression and
examining the decrease in explained power of the initial IVs. Next
those variables that were shown to be significant were forced into
a regression model with variables thought to be closely related to
PES. Covariates for assessment of longevity including sex, year of
birth, health (illness words), language, and age at time of publi-
cation were included in ALL regression models irrespective of
associations with the IVs or longevity.
Results
Based on Pearson’s correlations, age at publication was the only
control variable correlated with longevity (r⫽.40, p⬍.01),
however, the use of more self-ascribed illness words was margin-
ally associated with a shorter life span (r⫽⫺.16, p⫽.1). Control
variables were generally not correlated with the four main affect
word count variables, with the exception of associations between
age at publication and unactivated PES (r⫽⫺.22, p⬍.05) and
activated NES (r⫽⫺.25, p⬍.05).
Word usage in the Mackay emotion scales was skewed for three
out of the four scales with skewness values of 2.4 (unactivated
PES), 3.1 (unactivated NES), and 3.8 (for activated NES). This
was because of greater than 58% of individuals not using these
word types in their autobiographies, or specifically 76 and 72
nonusers for unactivated and activated NES respectively (range of
0 –5 and 0 –2 uses in text), and 58 nonusers for unactivated PES
(0 –3 uses in text). Because of this irregular distribution in word
use, data could not be normalized and therefore variables were
dichotomized into those who did and did not use the words.
Activated PES had a skewness statistic of .60 allowing the data to
be analyzed as a continuous variable (range of 0 –9 uses in text).
However, to enable comparisons to the other dichotomous word
categories it was also dichotomized into words used or not used
(N⫽23 not used). Phi coefficients were examined to determine
the association between the dichotomized emotion categories re-
vealing a correlation between unactivated NES and unactivated
PES (⫽.32, p⬍.01) and a weak association between activated
NES and unactivated NES (⫽.19, p⫽.07).
Results of individual linear regression analyses of each of the
four scales appear in Table 1. As apparent from the table, only
activated PES was significantly associated with increased longev-
ity (p⬎.2 for other three scales) accounting for approximately 6%
of the variance in longevity. The same result was found whether
examining activated PES as a continuous or dichotomous variable.
There was a 5-year advantage for those who used activated posi-
tive words versus those that did not (see Figure 2). To determine
whether the association of activated PES and longevity was inde-
pendent of the other scales, a single regression equation was
utilized including all four Mackay emotion measures. Activated
PES remained significantly associated with longevity when as-
sessed continuously (⫽.22, p⬍.05) or when assessed as a used
or not used variable (⫽.20, p⬍.05).
Next the PANAS dictionaries were tested. General PA from the
PANAS (i.e., the standard 10 item scale) was normally distributed
(skewness rating of .84, 8% nonusage, range of 0 – 43 uses) but NA
had 62% nonusage (skewness rating of 4.37, range of 0 – 6 uses).
The PANAS-X subscales were not normally distributed with a
4PRESSMAN AND COHEN
range 30 – 65% nonusage. To assess the association of these scales
with longevity, continuous measures were utilized when appropri-
ate (normally distributed data with skewness ratings ⬍1.5), the
data were dichotomized for those variables with greater than 50%
nonuse, and tertiles were used for those variables in the middle
(i.e., with approximately one-third nonusage). Variables that were
tertiled were dummy coded for analysis. As is apparent from Table
1, neither the PANAS PA nor NA scale was associated with
longevity. However, two of the PANAS-X positive subscales were
associated with increased longevity (one marginal) accounting for
approximately 4% of the variance in life span each. This repre-
sented a 3-year advantage for those who used attentive words and
a 2.7-year advantage for those who used self-assurance words (vs.
nonusers). Joviality was not associated with longevity nor were the
NA subscales.
Finally, when examining LIWC affect measures, both positive
and negative scales were normally distributed (skewness under
1.5; 2–161 and 0 –92 uses, respectively) but neither was associated
with longevity, nor were the LIWC negative subscales (all ps⬎
.2). The humor/laughter words subscale usage was skewed with
72% of the population not using them (0 –3 uses) and was therefore
analyzed as a used/not used category. As is apparent from Table 1,
use of these words was associated with increased longevity, ac-
counted for 8% of its variance, and amounted to a 6-year advan-
tage for humor word users.
Other Plausible Explanatory Factors
Optimism was not associated with longevity (p⬎.2) and
including it as a covariate did not alter any of the associations
between significant positive measures and longevity (all betas
remained the same and with p⬍.1). As noted in the methods, the
optimism dictionary included some PES/PA words. When these
were excluded from the word list (5 overlapping constructs), it
remained uncorrelated with longevity (p⬎.5). Using more social
words was associated with greater longevity (⫽.24, p⬍.05) per
previous findings (Pressman & Cohen, 2007); however, inclusion
of this factor in the separate regressions had little to no effect on
the association between any of our activated positive variables
(Activated PES, humor/laughter, and PANAS-X attentiveness and
self-assuredness) with all betas and significance levels remaining
approximately the same.
Given the small number of females (N⫽3), the primary PES
findings were reanalyzed without them in the sample. No major
changes occurred with their removal with the largest change being
a standardized beta weight change from .22 (self-assured) to .20
(new p⬍.1) and vice versa for attentiveness (change from .20 to
.22, new p⬍.05).
Finally, there may be some question as to whether there is a
significant effect in years lived after writing the autobiography (or
whether findings are an artifact of living a longer life and therefore
being able to contribute a life story). Because analyses control for
date of birth, as well as age at publication, this had been statisti-
cally controlled to some extent, however, to be conservative results
were additionally tested with the outcome years lived postpubli-
cation (controlling for standard covariates). Analyses revealed no
major differences in any emotion associations. For example, the
activated PES finding revealed a significant association with lon-
gevity (⫽.23, p⬍.05) nearly identical to the past association
(⫽.22, p⬍.05) and humor showed no substantial change (s
were .31 with new assessment vs. .36 for old, both ps⬍.01).
Discussion
This study found that including different types of activated
positive emotion words in one’s autobiography was associated
with approximately three to six additional years of life for those
who included at least one of these word types. These associations
withstood accounting for covariates such as date of birth, sex, age
at publication of autobiography, health, NES, and language. To
help the reader understand what it looks like to be high in written
activated PES, two examples from the text are included: (1) “By
Table 1
Independent Hierarchical Multiple Regression Analyses
Predicting Longevity From Different Emotion Types and
Dictionaries (Adjusted for Covariates)
Predictor 
Block 1 (covariates entered together)
Date of birth
1
0.16
Age at publication
1
.40
ⴱⴱ
Sex
2
(1 ⫽male, 2 ⫽female) ⫺0.01
Health
1
⫺0.07
Translated
2
(1 ⫽yes, 0 ⫽no) ⫺0.17
Block 2 (each entered in separate equation)
Positive emotional style
(High arousal)
Activated PES
2
(MacKay) .22
ⴱ
Humor/laughter
2
(created from LIWC) .36
ⴱⴱ
PANAS PA
1
0.02
PANAS-X attentiveness
3
0.20
†
PANAS-X self-assured
2
0.24
ⴱ
PANAS-X joviality
3
⫺0.05
(Neutral/mixed)
LIWC positive emotion
1
0.12
(Low arousal)
Unactivated PES
2
(MacKay) 0.10
Negative emotional style
(High arousal)
Activated NES
2
(MacKay) ⫺0.11
PANAS NA
2
0.05
PANAS-X hostility
2
0.05
PANAS-X fear
2
0.10
LIWC anger
1
0.01
LIWC anxiety
2
0.07
(Neutral/mixed)
PANAS-X guilt
2
0.07
LIWC negative emotion
1
0.06
(Low arousal)
LIWC sadness
2
0.03
PANAS-X sadness
2
0.17
Unactivated NES
2
(MacKay) ⫺0.08
Note. Results are sorted based on valence and arousal level. All emo-
tional style types were entered independently in separate equations after
covariates. N⫽88; PES ⫽positive emotional style; NES ⫽negative
emotional style; PANAS ⫽Positive and Negative Affect Schedule; NA ⫽
negative affect; PA ⫽positive affect; LIWC ⫽Linguistic Inquiry and
Word Count.
1
Entered as a continuous variable.
2
entered as a dichotomous vari-
able.
3
entered as dummy variables in tertiles (reported as high group
versus else).
†
p⬍.10.
ⴱ
p⬍.05.
ⴱⴱ
p⬍.01.
5
POSITIVE EMOTION WORDS & LONGEVITY
the spring, after lying awake all night in the excitement of genuine
vocational choice ...I decided to take an MA in psychology.” and
(2) “I entered with zest all those extracurricular activities which
make up college life in a small town. Life was vital and enjoyable.”
Because previous studies of affective styles and longevity have
not classified affective styles based on activation, there are no
comparable data. However, it is possible that previously published
associations between PES and longevity were driven by the acti-
vated components of the scales. This is especially likely given that
many health studies rely on scales heavily weighted with aroused
adjectives such as the PANAS (Watson & Clark, 1988) and the
Profile of Mood States (McNair, Lorr & Droppleman, 1971).
These data are however consistent with those from a prospective
viral-challenge study on the role of emotional styles in suscepti-
bility to the common cold. There, among the four quadrants of
emotion, activated positive emotions were the primary predictors
of resistance to illness among healthy people exposed to a cold
virus (Cohen et al., 2006). Together with the evidence from the
cold study, these data suggest that the valence/activation distinc-
tion may play an important role in understanding the association
between affect and health.
It was surprising that negative emotions were not associated
with longevity in this study given the vast literature showing that
higher scores on negative traits such as depression, anxiety and
hostility are associated with poorer health (e.g., see reviews by
Carney et al., 2002; Hemingway & Marmot, 1999; Kiecolt-Glaser
et al., 2002). As is apparent from Figure 2, there is a trend whereby
those who use more activated NES words have shorter lives. That
this association does not approach significance is likely because
of the rarity of negative emotion usage in these samples. Given that
these are published autobiographies, it may be that some individ-
uals curbed negative emotion expression for self-presentation pur-
poses. This may have also been the case for the nuns in the Danner
and colleagues study (2001) given that there were comparatively
low levels of NES expression (in comparison to PES) and also no
found association of NES with longevity. However, it also may
have to do with the increasing evidence indicating a positivity bias
in the memories of older adults similar to our sample (e.g., Mather
& Carstensen, 2005).
Activated but not unactivated PES was tied to longevity. Con-
sistent with this finding, emotions researchers have purported that
low arousal emotions are actually the absence of emotion (e.g.,
Watson & Clark, 1988) lending credibility to the current finding
that activated PES is the critical ingredient to better health out-
comes. That being said, the multitude of studies on relaxation
interventions and the like would argue that feeling calm is also
critical to health and well-being (e.g., Gruzelier, 2002), however,
given that these studies frequently start with stressed individuals,
the benefits of calm in healthy unstressed populations remain
unclear. While it is tempting to assume that low usage of words is
responsible for the lack of unactivated PES findings in this study,
some high activation scales had greater nonusage (e.g., humor) but
were still tied to longevity.
Why are activated positive words associated with increased
longevity? One possibility is that there are physiological changes
that occur in conjunction with arousing positive emotional expe-
riences (e.g., improved vagal tone, increased endogenous opioids)
that contribute to better health (Pressman & Cohen, 2005; Ryff &
Singer, 2000). These changes may improve health through their
influences on immune and cardiovascular function and may also
aid in buffering physiological stress responses. It is also possible
that activated PES terms may be tapping the author’s perception of
their health status (e.g., vigorous, active, energetic; Pressman &
Cohen, 2005). If this is the case, the association of activated PES
Figure 2. The unadjusted associations between activated and unactivated measures of emotion (PES and NES)
with longevity. For the figure, emotion categories are dichotomized into those who did and did not use the words
in each category. Error bars represent standard errors.
6PRESSMAN AND COHEN
and longevity may merely represent perceived health predicting
longevity. This possibility was controlled for, albeit imperfectly,
by adding a covariate for the percent of words referring to the
author’s own illnesses. Activated PES might also reflect physical
fitness, however, a post hoc analysis of usage of 30 exercise words
(not shown) did not reveal associations with longevity. Along the
same lines, it is possible that those who aged well felt more
positive. However, because neither objective age (date of birth,
age of publication), nor perceived health, nor fitness could explain
the association, this seems unlikely. While it is possible that subtle
affective markers pick something up about the authors’ health not
represented by these terms, there is existing evidence from both
mortality (Parker, Thorslund, & Nordstrom, 1992) and morbidity
(Cohen et al., 2006) studies that PES-health associations are not
explicable in terms of traditional perceived health measures. Fi-
nally, given the strong ties between social relationships, health,
and positive emotions (Lyubomirsky, King, & Diener, 2005; Holt-
Lundstad, Smith, & Layton, 2010) it was also possible that social
relationships were responsible for the found emotion effects. The
PES-longevity associations however, were independent from the
influence of social integration (as measured in writing) raising
the possibility of distinct pathways connecting emotion versus
social relationships to better health. While there are hypotheses
about separate mediators connecting these variables to health
outcomes (e.g., oxytocin for social measures), to our knowledge no
work has tested in a single study whether there are unique paths.
There was an interestingly strong association between age at
time of publication and longevity, where those who wrote their
autobiography later in life (e.g., in their 80s) lived longer. Part of
this is likely because of the artifact of these individuals already
outliving their life expectancy (and therefore living longer than the
average for the sample), however, it is also interesting to consider
that these individuals who are older may be different in some way.
Analyses revealed them to have lower levels of activated NES
indicating that they are less distressed which may produce some
overall well-being benefit. There was however no association with
any of the activated positive variables tested (all ps⬎.4) indicat-
ing that it was not simply living longer that was responsible for the
found PES-longevity associations. This issue was further ad-
dressed by examining years lived postpublication as a dependent
variable (as opposed to overall longevity) and all of existing
activated PES findings persisted despite this new modeling tech-
nique. While there was no PES association with age at publication,
because some of these subjects were still productive and working
late in their lives, it is possible that this life purpose is also
extending their longevity (e.g., Boyle, Barnes, Buchman, & Ben-
nett, 2009). While “life purpose” is not easily testable in writing,
previous work has shown that finding meaning in the context of an
expressive writing paradigm (tested by insight words and causal
language) is tied to better health (e.g., Pennebaker, Mayne, &
Francis, 1997). When these variables were tested in the current
data (analyses not shown) they were not tied to increased life span
(ps⬍.3).
This study has several strengths such as its indirect assessment
of emotional style, assessment of different emotion types, and its
controls for multiple alternative explanations including year of
publication, year of birth, language, baseline health (via disclosed
illness in writing), negative emotional style, optimism, and social
integration. There is still the possibility that other variables such as
unreported illness in writing, openness to experience, other per-
sonality characteristics, or other psychological and biological vari-
ables not assessed could account for both greater PES and longev-
ity. This study is also limited by the lack of objective health data
at the time of writing (e.g., those individuals who aged “well” may
have been more positive) and the use of a self-reported health
assessment in a published writing sample. This limits this study’s
ability to say anything about the causality or directionality of the
findings. Additionally, while no psychologist in this study was
listed as committing suicide in obituaries or other encyclopedia
entries, it is impossible to say with 100% certainty that all causes
of death were natural. The generalizability of these findings may
also be limited given that famous psychologists differ from the
general population. Nevertheless, it remains impressive that this
sample of highly educated, and primarily male psychologists
showed similar associations with writing PES and longevity as did
the entirely female Danner et al. (2001) nun sample.
It may be that specific types of emotion are relevant to health in
different scenarios, and that by diluting these constructs with other
emotions (e.g., by looking at overall PES instead of subscales),
researchers are missing out on potentially interesting findings. This
study raises the possibility that it is not only important to distin-
guish between the effects of positive and negative emotions, but
also the specific types of emotions within these domains.
References
Boyle, P. A., Barnes, L. L., Buchman, A. S., & Bennett, D. A. (2009).
Purpose in life is associated with mortality among community-dwelling
older persons. Psychosomatic Medicine, 71, 574 –579. doi:10.1097/
PSY.0b013e3181a5a7c0
Carney, R. M., Freedland, K. E., Miller, G. E., & Jaffe, A. S. (2002).
Depression as a risk factor for cardiac mortality and morbidity: A review
of potential mechanisms. Journal of Psychosomatic Research, 53, 897–
902. doi:10.1016/S0022-3999(02)00311-2
Chida, Y., & Steptoe, A. (2008). Positive psychological well-being and mortality:
A quantitative review of prospective observational studies. Psychosomatic Med-
icine, 70, 741–756. doi:10.1097/PSY.0b013e31818105ba
Cohen, S., Alper, C. M., Doyle, W. J., Treanor, J. J., & Turner, R. B.
(2006). Positive emotional style predicts resistance to illness after ex-
perimental exposure to rhinovirus or influenza a virus. Psychosomatic
Medicine, 68, 809 – 815. doi:10.1097/01.psy.0000245867.92364.3c
Cohen, S., Doyle, W. J., & Skoner, D. P. (1999). Psychological stress,
cytokine production, and severity of upper respiratory illness. Psycho-
somatic Medicine, 61, 175–180.
Cohen, S., Kessler, R. C., & Gordon, L. U. (1995). Strategies for measuring
stress in studies of psychiatric and physical disorders. In S. Cohen, R. C.
Kessler, & Gordon, L. U. (Eds.), Measuring stress: A guide for health
and social scientists (pp. 3–26). London: Oxford University Press.
Cohen, S., & Pressman, S. (2006). Positive affect and health. Current
Directions in Psychological Science, 15, 122–125. doi:10.1111/j.0963-
7214.2006.00420.x
Danner, D. D., Snowdon, D. A., & Friesen, W. V. (2001). Positive
emotions in early life and longevity: Findings from the nun study.
Journal of Personality & Social Psychology, 80, 804 – 813. doi:10.1037/
0022-3514.80.5.804
Diener, E., & Suh, E. M. (1999). National differences in subjective well-
being. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being:
The foundations of hedonic psychology (pp. 434 –450). New York, NY:
Russell Sage Foundation.
Diener, E., Suh, E. M., Lucas, R. E., & Smith, H. L. (1999). Subjective
well-being: Three decades of progress. Psychological Bulletin, 125,
276 –302. doi:10.1037/0033-2909.125.2.276
7
POSITIVE EMOTION WORDS & LONGEVITY
Fazio, R. H., & Olson, M. A. (2003). Implicit measures in social cognition
research: Their meaning and uses. Annual Review of Psychology, 54,
297–327. doi:10.1146/annurev.psych.54.101601.145225
Fredrickson, B. L. (1998). What good are positive emotions? Review of
General Psychology, 2, 300 –319. doi:10.1037/1089-2680.2.3.300
Gruzelier, J. H. (2002). A review of the impact of hypnosis, relaxation,
guided imagery and individual differences on aspects of immunity and
health. Stress, 5, 147–163. doi:10.1080/10253890290027877
Hemimgway, H., & Marmot, H. (1999). Psychosocial factors in the aeti-
ology and prognosis of coronary heart disease: Systematic review of
prospective cohort studies. British Medical Journal, 38, 1460 –1467.
Herbert, T. B., & Cohen, S. (1993). Depression and immunity: A meta-
analytic review. Psychological Bulletin, 113, 472– 486. doi:10.1037/
0033-2909.113.3.472
Holt-Lunstad, J., Smith, T. B., & Layton, B. (2010). Social relationships
and mortality risk: A meta-analytic review. PLoS Medicine, 7, 1–20.
doi:10.1371/journal.pmed.1000316.g001
Kiecolt-Glaser, J. K., McGuire, L., Robles, T. F., & Glaser, R. (2002).
Emotions, morbidity and mortality: New perspectives on psychoneuro-
immunology. Annual Review of Psychology, 53, 83–107. doi:10.1146/
annurev.psych.53.100901.135217
Krantz, D. S., Glass, D. C., Contrada, R., & Miller, N. E. (1981). Behavior
and Health: Mechanisms and Research Issues. Social Science Research
Council: Items, 36, 1– 6.
Krantz, D. S., & McCeney, M. K. (2002). Effects of psychological and
social factors on organic disease: A critical assessment of research on
coronary heart disease. Annual Review of Psychology, 53, 341–369.
doi:10.1146/annurev.psych.53.100901.135208
Levy, B. R., Slade, M. D., Kunkel, S. R., & Kasl, S. V. (2002). Longevity
increased by positive self-perceptions of aging. Journal of Personality
and Social Psychology, 83, 261–270. doi:10.1037/0022-3514.83.2.261
Lucas, R. E., Diener, C., & Larsen, R. J. (2003). Measuring positive
emotions. In S. J. Lopez & B. K. Snyder (Eds.), Positive psychological
assessment: A handbook of models and measures (pp. 201–218). Wash-
ington, DC: American Psychological Association. doi:10.1037/10612-
013
Lyubomirsky, S., King, L., & Diener, E. (2005). The benefits of frequent
positive affect: Does happiness lead to success? Psychological Bulletin,
131, 803– 855.
Mackay, C., Cox, T., Burrows, G., & Lazzerini, T. (1978). An inventory
for the measurement of self-reported stress and arousal. British Journal
of Social & Clinical Psychology, 17, 283–284. doi:10.1111/j.2044-
8260.1978.tb00280.x
Maier, H., & Smith, J. (1999). Psychological predictors of mortality in old
age. Journal of Gerontology: Psychological Sciences, 54B, 44 –54.
doi:10.1093/geronb/54B.1.P44
Martin, R. A. (2001). Humor, laughter and physical health: Methodological
issues and research findings. Psychological Bulletin, 127, 504 –519.
doi:10.1037/0033-2909.127.4.504
Mather, M., & Carstensen, L. L. (2005). Aging and motivated cognition:
The positivity effect in attention and memory. Trends in Cognitive
Sciences, 9, 496 –502. doi:10.1016/j.tics.2005.08.005
McAdams, D. P. (2001). The psychology of life stories. Review of General
Psychology, 5, 100 –122. doi:10.1037/1089-2680.5.2.100
McNair, D. M., Lorr, M., & Droppleman, L. F. (1971). Profile of mood
states manual. San Diego, CA: Educational Testing Service.
Murchison, C. (Ed.). (1961). A history of psychology in autobiography
(Vol. 1). New York, NY: Russell & Russell.
Nowlis, V. (1965). Research with the mood adjective checklist. In S. S.
Tomkins & C. E. Izard (Eds), Affect, cognition and personality (pp.
352–389).New York, NY: Springer.
Ostir, G. V., Markides, K. S., Black, S. A., & Goodwin, J. S. (2000).
Emotional well-being predicts subsequent functional independence and
survival. Journal of the American Geriatrics Society, 48, 473– 478.
Parker, M. G., Thorslund, M., & Nordstrom, M. L. (1992). Predictors of
mortality for the oldest old. A 4-year follow-up of community-based
elderly in Sweden. Archives of Gerontology and Geriatrics, 14, 227–
237. doi:10.1016/0167-4943(92)90023-W
Pennebaker, J. W., Booth, R. E., & Francis, M. E. (2007). Linguistic
inquiry and word count: LIWC2007– operator’s manual. Austin, TX:
LIWC.net.
Pennebaker, J. W., Francis, M. E., & Booth, R. J. (2001). Linguistic Inquiry
and Word Count (LIWC): LIWC2001. Mahwah, NJ: Erlbaum Publishers.
Pennebaker, J. W., & King, L. A. (1999). Linguistic styles: Language use
as an individual difference. Journal of Personality and Social Psychol-
ogy, 77, 1296 –1312. doi:10.1037/0022-3514.77.6.1296
Pennebaker, J. W., Mayne, T. J., & Francis, M. E. (1997). Linguistic
predictors of adaptive bereavement. Journal of Personality and Social
Psychology, 72, 863– 871. doi:10.1037/0022-3514.72.4.863
Pressman, S. D., & Cohen, S. (2005). Does positive affect influence health?
Psychological Bulletin, 131, 925–971. doi:10.1037/0033-2909.131
.6.925
Pressman, S. D., & Cohen, S. (2007). The use of social words in autobi-
ographies and longevity. Psychosomatic Medicine, 69, 262–269. doi:
10.1097/PSY.0b013e31803cb919
Russell, J. A. (1980). A circumplex model of affect. Journal of Personality
& Social Psychology, 39(Suppl 6), 1161–1178.
Ryff, C. D., & Singer, B. (2000). Interpersonal flourishing: A positive
health agenda for the new millennium. Personality and Social Psychol-
ogy Review, 4, 30 – 44. doi:10.1207/S15327957PSPR0401_4
Scheier, M. F., Carver, C. S., & Bridges, M. W. (1994). Distinguishing
optimism from neuroticism (and trait anxiety, self-mastery, and self-
esteem): A reevaluation of the Life Orientation Test. Journal of Person-
ality and Social Psychology, 67, 1063–1078. doi:10.1037/0022-
3514.67.6.1063
Seligman, M. E. P., & Csikszentmihalyi, M. (2000). Positive psychology:
An introduction. American Psychologist, 55, 5–14. doi:10.1037/0003-
066X.55.1.5
Thayer, R. E. (1967). Measurement of activation through self-report.
Psychological Reports, 20, 663– 678.
Thayer, R. E. (1970). Activation states as assessed by verbal report and
four psychophysiological variables. Psychophysiology, 7, 86 –94. doi:
10.1111/j.1469-8986.1970.tb02278.x
Watson, D., & Clark, L. A. (1988). Development and validation of brief
measures of positive and engative affect: The PANAS scale. Journal of
Personality and Social Psychology, 54, 1063–1070. doi:10.1037/0022-
3514.54.6.1063
Watson, D., & Clark, L. A. (1994). The PANAS-X: Manual for the positive
and negative affect schedule-expanded form. Iowa City, IA: The Uni-
versity of Iowa.
(Appendix follows)
8PRESSMAN AND COHEN
Appendix
The subjects included in this study are listed below in rank order based on age at
death, with the longest life span being 98.07 years (Viteles) and the shortest
being 55.04 years (Klemm)
M. Viteles (98.1) G. Murphy (84.0) J. Angell (80.0) R. Dodge (71.0)
E. Hilgard (97.3) M. Bentley (83.8) H. Carr (80.0) G. Allport (70.0)
H. Murray (93.8) K. Groos (83.1) R. Yerkes (79.7) J. Konorski (69.8)
W. Miles (93.2) B. Bourdon (83.0) H. Helson (79.0) D. Katz (69.0)
O. Klineberg (93.1) C. Seashore (83.0) L. Terman (79.0) L. Thurstone (68.4)
R. Woodworth (93.1) H. Pieron (83.0) J. Drever (77.0) M. Washburn
1
(68.3)
R. Cattell (92.9) A. Michotte (82.5) M. Mead
1
(77.0) D. Krech (68.0)
A. Anastasi
1
(92.4) R. Elliott (82.4) F. Beach (76.1) C. Hull (68.0)
S. Pressey (91.1) K. Marbe (82.2) W. Wirth (76.0) E. Claparede (67.6)
J. Guilford (90.8) E. Boring (82.1) A. Luria (75.0) D. Broadbent (67.0)
C. Stumpf (88.7) C. Spearman (82.0) C. Osgood (75.0) S. Stevens (67.0)
C. Burt (88.7) F. Bartlett (81.8) J. Gibson (75.0) H. Warren (67.0)
P. Janet (87.8) F. Kiesow (81.6) O. Mowrer (75.0) Q. McDougall (67.0)
H. Hoffding (87.1) H. Eysenck (81.6) L. Carmichael (74.9) W. Stern (66.9)
J. Dashiell (87.1) T. Newcomb (81.1) G. Thomson (74.0) M. Calkins
1
(66.9)
T. Ziehen (86.8) F. Allport (81.0) E. Thorndike (73.9) W. Hunter (65.4)
W. Pillsbury (86.7) D. Hebb (81.0) C. Judd (73.4) C. Graham (65.0)
B. Skinner (86.5) R. Sears (81.0) C. Myers (73.0) G. Ferrari (64.3)
Q. McNemar (86.1) A. Gesell (81.0) E. Tolman (73.0) S. Franz (58.6)
K. Goldstein (85.7) J. Jastrow (81.0) S. De Sanctis (73.0) O. Klemm (55.0)
C. Rogers (85.1) J. Watson (80.8) J. Baldwin (72.8)
H. Simon (84.7) J. Frobes (80.6) K. Dunlap (72.7)
K. Dallenbach (84.2) F. Geldard (80.6) H. Zwaardemaker (72.4)
J. Piaget (84.2) A. Gemelli (80.4) W. Bingham (72.0)
C. Morgan (84.0) E. Scripture (80.1) G. Heymans (71.8)
1
Female.
9
POSITIVE EMOTION WORDS & LONGEVITY
Correction to Pressman and Cohen (2011)
In the article “Positive emotion word use and longevity in famous deceased psychologists,” by Sarah
D. Pressman and Sheldon Cohen (Health Psychology, Advance online publication. September 19,
2011. doi: 10.1037/a0025339), the bars in Figure 2 were mislabeled. The black bars should have
been labeled the “Used” words category and the grey bars should have been labeled as the “Did Not
Use” category. All versions of this article have been corrected.
DOI: 10.1037/a0026071
A preview of this full-text is provided by American Psychological Association.
Content available from Health Psychology
This content is subject to copyright. Terms and conditions apply.
