Jonathan D, Haidt -...

76
What a "Facebook Smile" Reveals About Future Happiness: The Expression of Positive Affect in Facebook Photographs Predicts Long-Term Well-Being James Patrick Seder New York, NY Bachelor of Arts, The City College of New York, 2002 Master of Arts, University of Virginia, 2007 A Dissertation presented to the Graduate Faculty of the University of Virginia in Candidacy for the Degree of Doctor of Philosophy Department of Psychology University of Virginia August, 2010 Shigehiro Oishi /Jonathan D, Haidt James ETBurroughs

Transcript of Jonathan D, Haidt -...

What a "Facebook Smile" Reveals About Future Happiness: The Expression of Positive Affect in Facebook Photographs

Predicts Long-Term Well-Being

James Patrick Seder New York, NY

Bachelor of Arts, The City College of New York, 2002 Master of Arts, University of Virginia, 2007

A Dissertation presented to the Graduate Faculty of the University of Virginia in Candidacy for the Degree of

Doctor of Philosophy

Department of Psychology

University of Virginia August, 2010

Shigehiro Oishi

/Jonathan D, Haidt

James ETBurroughs

UMI Number: 346223

All rghts reserve

INFORMATION TO ALL USER The qualty of this reproduction is dependent on the quality of the copy su

In the unlikely event that the author did not send a complete man and there are missing pages, these will be noted, Also, if material had to be

a note will indicate the deleti

UMI 346223

Copyright 2011 by ProQuest L

All rghts reserved. This edition of the work is protected a unauthorized copyng under Title 17, United States C

ProQuest LLC 789 East Eisenhower Parkws

P.O. Box 134 Ann Arbor, Ml 48106-1

2

Abstract

Does smile intensity in people's Facebook photos predict long-term well-being? In order

to answer this question, we coded the extent to which college students were smiling in

photographs they posted to their Facebook accounts during their first semester at college.

We used these behavioral data to predict long-term self-reported life satisfaction and

relationship satisfaction. In Study 1 and Study 2, we showed that smile intensity coded

from a single Facebook profile photograph retrieved during participants' first semester at

college was a predictor of well-being 3.5 years later—as the participants were about to

graduate from college. These results held even when we controlled for the influence of

personality (e.g., Big Five), and for life satisfaction reported during the first semester. In

Study 3, we explore the utility of using multiple photographs in order to predict long-

term life satisfaction.

3

What A "Facebook Smile" Reveals About Future Happiness:

The Expression of Positive Affect in Facebook Photographs

Predicts Long-Term Well-Being

In a recent large-scale study involving nearly two thousand adults, Davidson,

Mostofsky, and Whang (2010) showed that display of positive affect on the part of

patients during (relatively short) patent-nurse interviews predicted decreased incidence of

coronary heart disease ten years later. The possibility that "momentary" displays of

positive affect can predict future well-being has long been an intriguing—and largely

under-explored—area of psychological research.

In the most well-known example of this type of research, Harker and Keltner

(2001) coded the extent to which a cohort of women were smiling in their college

graduation yearbook photos. The women had graduated from the all-female Mills

College (in California) and were participants in a 30-year longitudinal study sponsored by

the college. As part of the study, all of the women completed follow-up surveys once per

decade; each of these surveys included questions about the participants' well-being.

Thus, in addition to the codings for the expressions of positive affect (i.e., the yearbook

photos), Harker and Keltner had access to self-report data from multiple time points

beyond graduation from college, with the final survey being completed when participants

were 52 years old.

Results showed that the intensity with which the women were smiling in their

yearbook photos predicted self-reported life satisfaction at each of the four future time

points. Smile intensity was also correlated with participants' likelihood of marrying in

their 20s and with their marital satisfaction at age 52 (the latter question was only

4

included in the final survey). Thus, the "momentary" expression of positive affect

captured in a single (formal) photograph that was taken and made public when

participants were just 21 years old conveyed surprisingly rich information about the

participants' short-term and long-term well-being and other important life outcomes.

Harker and Keltner (2001) did not test for mediation. They did, however,

speculate that at least two types of processes might be likely to mediate the relationship

between smile intensity and their participants' long-term well-being. First, the affective

display captured in the photographs might have been a marker of personality type.

Extraversion, for example, is typically associated with subjective well-being (e.g.,

Diener, Oishi, & Lucas, 2003) and relationship satisfaction (e.g., Scollon & Diener,

2006), and may be characterized by display of positive affect (i.e., smiling). Thus,

extraversion could be one plausible candidate. Second, Harker and Keltner noted that

intensity of affective display in the photos may have also been an indication of the degree

to which participants were likely to display positive emotion in their "real life"

interactions—and the degree to which interaction partners might be likely to perceive this

display and respond in kind. Display of positive affect in relationships has, for example,

been shown to be associated with relationship quality (e.g., Gottman, Levenson, &

Woodin, 2001) and relationship satisfaction. Thus, there is ample past research to

suggest that the relationship between smile intensity and long-term well-being could have

been at least partially mediated by relationship satisfaction.

This work by Flarker and Keltner (2001) generated much interest within the field,

Indeed, the single-study paper is cited in numerous introductory psychology textbooks (e.g.,

Baumeister & Bushman, 2007; Kalat, 2007; Myers, 2007). Owing perhaps to the time

5

required in order to conduct longitudinal research (or the lack of existing archival data

suitable for such research questions), few attempts to replicate and extend the work were

immediately forthcoming in the literature. In fact, the only "replication and extension" paper

published within six years of Harker & Keltner's study failed to replicate the findings

between smile intensity and well-being—this time using high school graduation yearbook

photos (Freese, Meland, & Irwin, 2007).

In the last few years, however, interest in this paradigm (i.e., the predictive utility of

expressions of positive affect "captured by" photographs) has been renewed. Two recently-

published studies have used intensity of smiling (as captured in photographs) as a predictor

variable for long-term outcomes. Hertenstein, Hansel, Butts, and Hile (2009) showed that

smile intensity in photos taken before the age of 22 (which included yearbook photos in

Study 1 and non-yearbook photos contributed by participants in Study 2) was correlated with

decreased likelihood of divorce later in life. In another paper, Abel and Kruger (2010) coded

smile intensity in archival photographs of early-career professional baseball players and

showed that smile intensity predicted longevity; players who displayed an intense smile

(smile intensity was rated on a scale ranging from 1-3) lived an average of 7 years longer

than those who showed minimal display of positive affect. These findings held even when

the researchers controlled for several other potentially relevant behavioral variables (e.g.,

career longevity; whether or not the players graduated from college). While neither study

assessed self-reported happiness, the direction of both outcomes would appear to be in line

with other research showing a positive relationship between happiness and relationships

(Diener, & Seligman, 2002) and happiness and longevity (Veenhoven, 2008).

6

One unresolved question within this small but intriguing literature involves the null

findings reported by Freese and colleagues (2007) in their attempt to replicate the

relationship between smile intensity and well-being found by Harker and Keltner (2001).

Because Freese and colleagues used coding methods that were somewhat different from

those used by Harker and Keltner1, there has been speculation as to whether the null results

may have been (at least somewhat) attributable to the coding method used. The recent study

by Abel and Kruger (2010), however, successfully employed the coding method that was

developed by Freese and colleagues. Although Abel and Kruger examined longevity and not

subjective well-being, their results do suggest that the 3-point scale method of evaluating

smile intensity can be sound . Regardless, the recently-published work by Abel and Kruger

highlights the need to replicate Flarker and Keltner (2001) in new samples—as well as the

need to understand and address the "differences" that currently exist between the different

coding schemes that have been validated and used within the literature (e.g., Freese, Meland,

8c Irwin, 2007; Harker & Keltner, 2001; Tsai, Louie, Chen, & Uchida, 2007).

Revisiting Harker and Keltner (2001) in the age of Facebook

Prior to the internet age, the vast majority of U.S. college students were much like

the participants in Harker and Keltner's 2001 study—a single graduation yearbook photo

was the only public photographic record of their time in college. With the advent of the

internet, however, college students have been able to publicly share a near-limitless

number of photographs (and other personal information) with little to no cost involved or

technical know-how required. Further, because such photos are typically informal and

shared with "friends," it seems plausible that the photos in question may also be likely to

be more naturalistic, and far more "information rich" than the posed, yearbook photos

7

used in past research (e.g., Harker & Keltner, 2001). Indeed, the vast majority of photos

posted to social networking websites are "social" (i.e., multiple people typically appear in

the photographs). Because the contexts of the photos are usually informal, it also seems

likely that most are taken by someone known by the person(s) posing. This latter

dynamic could be contrasted with the environment typical of photography studios in

which a person is required to "pose" in front of a stranger (i.e., the professional

photographer).

While many online venues for public content sharing exist, the social networking

website Facebook has become the current destination of choice for college students.

Started in 2004 and available at most U.S. colleges by 2005, Facebook now has in excess

of 400 million regular users—which includes the vast majority of college students

(Facebook, 2010). In line with the focus of the current research, the Facebook website

also reports that "more than 3 billion photos are uploaded to the site each month"

(Facebook, 2010).

Given the quantity and diversity of personal information being archived and

shared publicly on the Internet by college students (and non-students), a steadily-growing

but small body of research has begun to examine what such information is likely to reveal

about the people who post it. While several programs of research exist that make use of

information provided by participants on the Internet (e.g., via coding the content for some

variable of interest), much focus has been devoted to understanding the ways in which

users' personalities are likely to be revealed via the content shared. For instance, Vazire

and Gosling (2004) found that assessments of the web users' personality made by

observers of their web pages were moderately correlated with users' own personality

8

ratings; results also showed that extraversion and agreeableness were the traits most

easily detected. Buffardi and Campbell (2008) showed that observers were able to

predict the self-reported narcissism scores of Facebook users by simply looking at the

content displayed in users' profiles. Recent research employing both U.S. and European

samples also showed that people's Facebook profiles are typically an accurate

representation of their actual (rather than simply their "ideal") selves (Back, Stopfer,

Vazire, Gaddis, Schmukle, Egloff, & Gosling, 2010). Taken as a whole, then, this

research—although certainly a small and very "new" literature—suggests that people's

web pages (and the photos contained therein) are likely to be a rich and accurate source

of information about their "offline" lives. Thus, use of photos posted on people's social

networking web pages within the context of our program of research seems likely to be

methodologically sound.

While researchers have sought clues about personality by examining the content

of people's Facebook pages, little work exists related to the use of posted content in order

to infer or predict subjective well-being. Indeed, to our knowledge little research has

addressed the degree to which observers—untrained or trained—are capable of

perceiving the well-being of web site owners via behavioral indicators. However, a few

"clues" do exist about ways in which Facebook content may be indicative of users'

"offline" well-being. Mikami, Szwedo, Allen, Evans, and Hare (2010), for example,

coded the messages posted to the "Wall" of college students' Facebook profiles for

indications of social support. Results showed that the quantity and quality of the displays

of social support posted online were correlated with the quality of participants' "offline"

social relationships. Seder and Oishi (2009) coded the ethnic/racial diversity of college

9

students' on-campus Facebook friends and showed that among European Americans, a

homogeneous Facebook social network was associated with higher well-being at the end

of the first semester at college. Finally, recent research by Kramer (2010) showed that

use of positive words by Facebook users in their "status updates" (i.e., short messages

posted in order to communicate with friends on the system) was correlated with self-rated

happiness.

While intriguing, much research is obviously needed in order to understand the

degree to which the expression of positive affect at one point in time is likely to predict

long-term subjective well-being and relationship satisfaction—and why such a predictive

relationship might exist. Given the considerable body of research that has documented

causal links between happiness and numerous vital life outcomes (see Lyubomirsky,

King, & Diener, 2005)—and given the specific health outcomes predicted by Davidson

and colleagues (2010) from short-duration displays of positive affect in patient-nurse

interactions—this behavioral variable may have the potential to make important

contributions to our understanding of long-term life satisfaction and relationship

satisfaction.

Overview of Studies

The purpose of the current research is to conduct a systematic investigation of the

relationship between expressions of positive affect (i.e., smiling behavior) as coded from

participants' Facebook profile photos, and participants' short- and long-term self-reported

life satisfaction and relationship satisfaction. A primary goal will be to determine whether

and when the display of positive affect in the photographs is likely to convey information

about well-being that has unique predictive utility. This goal is especially important because

10

while the past research cited above (e.g., Harker & Keltner, 2001) has shown that the

expression of positive affect in photographs can predict longitudinal outcomes, results have

been reported as simple correlations.

We examined these questions in three longitudinal, mixed-sex studies. In each, we

tracked cohorts of college students beginning in their first semester at college. Thus, the

participants in each study were facing the challenge of adapting to the same general physical,

social, and educational environment at roughly the same time—regardless of where they

lived prior to coming to university. An additional benefit to our use of cohorts is that each of

our cohorts (presumably) "adapted" to the use of Facebook during the same general time

period. Thus, our design may have also helped to mitigate the influence of any changes that

took place with the website during the course of the study (e.g., new features; norm changes

for number of photos posted or how frequently people update their profile photos). The latter

point is important for us to consider; given that we are coding profile photos from the

website, we must be alert for any possibility that "the medium" (i.e., Facebook itself) could

influence the content of participants' photos.

In Study 1 and Study 2, we track students (two separate cohorts, one year apart)

through four years of college. Participants in both studies contributed data twice in semester

one, and once toward the end of their final semester at college. In a direct replication of

Harker and Keltner (2001), we coded Smile Intensity (i.e., the extent to which people

displayed positive affect) using a single profile photo (and via the coding scale developed

and validated by Harker and Keltner). It should be noted that the sample sizes in Study 1 and

Study 2 were modest. In this paper, we could have collapsed the data across the two studies

in order to increase power. Because this research is "new" in a variety of respects, however,

11

we felt it was important to look carefully for any hints of differences between cohorts that

might have the potential to influence results (for better or for worse). In order to address the

issue of power, however, we have included an "Overview Analyses" section after Study 2 in

which we conduct mediation analyses using the combined data from Study 1 and Study 2.

Study 3 represents an initial attempt to expand upon the design of the first two

studies. First, we coded multiple profile photos rather than a single photo in order to

calculate Smile Intensity. This enables us to examine the wilhin-person mean and standard

deviation of Smile Intensity codings. Given that self-consistency has been shown to be

related to well-being within Western culture (e.g., Oishi, Diener, Scollon, 8c Biswas-Diener,

2005), it seems plausible that people (especially our quintessentially American college

students) who exhibit more variability in their display of affective behavior across different

situations could experience lowered well-being. Conversely, however, lack of variability

could have the potential to be a marker of lack of social competence (i.e., the proverbial "bull

in the china shop"), which could be associated with lower well-being. In a second change of

design, the participants in Study 3 completed their "Time 3" follow-up study at the end of

their second year at college (rather than at the end of their fourth year). This will enable us to

address the question of whether (and to what extent) Smile Intensity will predict life

satisfaction and relationship satisfaction at the mid-point of students' college careers rather

than only at "the end"—which could be associated with an increase in self-reported in well-

being (e.g., Fredrickson, 2000).

Study 1

Method

Participants

12

Participants were 152 first-year, first-semester undergraduates at the University of

Virginia (57 male) in fall 2005. Only first-year students who had an active Facebook account

were eligible to participate in this study. In addition, this study involved self-report data

collected at three time points over the course of four years, and it required participants to

have at least one code-able profile photograph on Facebook. Participants with substantial

missing data at Time 2 or Time 3 and/or those with no code-able Facebook photo {n - 110)

were dropped from the final dataset. This left a total of 48 participants (20 male).3 The age

of these participants ranged from 17 to 19, with a mean age of 17.96 (SD = .41). In terms of

ethnicity, 3 (6.3%) were African-American, 11 (22.9%) Asian, 27 (56.3%) European

American, 1 (2.1%) Hispanic, 1 (2.1%) Middle Eastern, and 5 (10.4%) were categorized as

"Multi-Racial/Ethnic." All participants received partial fulfillment of their course

requirement in exchange for participation at Time 1 and Time 2; those who completed Time

3 were entered into a lottery to win one of five cash prizes.

Procedure

The current study was run in three sessions over the course of 3.5 years. At Time 1

(the beginning of participants' first semester at college), a majority of the participants (90%;

n - 43) completed several self-report measures of subjective well-being and personality. At

Time 2 (toward the end of the first semester at college), all participants (n - 48) completed

numerous self-report measures to assess overall life satisfaction. At Time 3 (toward the end

of participants' final semester in college), all participants again completed self-report

measures to assess subjective well-being. In addition, we measured current relationship

satisfaction (e.g., satisfaction with friendships and social events). As exploratory variables,

we also measured anticipated future behaviors related to relationship satisfaction (e.g.,

13

likelihood of attending reunions and of donating money to the university as an alumni), and a

behavioral indicator of relationship satisfaction (i.e., number of On-Campus Facebook

Friends). Finally, at Time 2 all participants gave consent to have their Facebook profile and

accompanying photos retrieved and analyzed. The web pages containing participants' main

profile photos were retrieved and saved toward the end of their first semester at college

(Time 2).

Materials

Time 1 (Beginning of first semester at college). A majority of participants (n ~ 43)

completed the Satisfaction with Life Scale (SWLS) (Diener, Emmons, Larsen, & Griffin,

1985; l=Strongly disagree to 7=Strongly agree; a = .86), and the Felt Understanding (1-Not

at all to 7=A lot in the past month: understood, appreciated, validated; a =.91) and Felt

Misunderstanding Scales (l=Not at all to 7=A lot in the past month: alienated,

misunderstood, ignored; a = .86 ), (Oishi, Krochik, & Akimoto, 2010). In addition, we

assessed Big Five personality traits (Brody & Ehrlinchman, 1997; extraversion a = .87;

neuroticism a - .91; agreeableness a = .80; conscientiousness a = .77; openness a = .56. ; a

possible range of 1 to 5). Finally, participants answered two questions that asked about the

amount of content posted to their Facebook pages (l=Very little to 4=A lot), and how often

they update the content on their Facebook page (1 =Rarely to 4=Frequently).

Time 2 (End of first semester at college). All participants (n - 48) completed the

SWLS (a - .87), and the Felt Understanding (a = .92) and Felt Misunderstanding Scales (a

- .87). In addition, they completed the Positive Negative Affect Schedule (PANAS; Watson,

Clark, & Tellegen, 1988; PA« = .86 ; NA a = .74). At this time, we also retrieved

14

participants' Facebook pages, and calculated the number of On-Campus Facebook Friends

and Other-School Facebook Friends listed in their account.

Time 3 (End of final semester at college; 3.5 years later). All participants (n = 48)

again completed the SWLS (a - .89), and the Felt Understanding (a - .92) and Felt

Misunderstanding Scales (a = .86). They also completed questions related to relationship

satisfaction (e.g., friendships, social events; l=Exlremely dissatisfied to 7=Extremely

Satisfied) and anticipated behaviors related to relationship satisfaction (e.g., likelihood of

attending future reunions, and of donating money to the university as alumni; l=Extremely

unlikely to 7=Extremely likely). Further, they completed a question which asked them to

rate their satisfaction with how well they "Fit in" during their time at college (l~Extremely

dissatisfied to 7==Extremely Satisfied).

Procedure for Coding the Expression of Positive Affect in Photographs

Participants' Facebook pages and photos were retrieved and saved toward the end of

their first semester at college (Time 2). In this study, we coded the main Facebook profile

photo (i.e., the photo chosen by the participant to be their most "public" presentation of self;

it is the first photo seen by viewers and the largest photo displayed in the account) that was in

each participant's account on the day we accessed and saved their profiles. Only photos that

provided a clear view of the participants' face (including eyes and mouth) were coded. In

all cases, the identity of the person in the photo was verified to be the participant.

Compared with the types of photographs used in past research on display of facial

affect (e.g., yearbook photos), photos uploaded to Facebook are typically less formal in pose,

and they vary on numerous aesthetic qualities (e.g., lighting, composition, camera angle).

That said, Facebook profile photos are typically of high resolution, and they are capable of

15

displaying in a web browser (e.g., Firefox) at 6" x 6" (or larger) with little loss of image

quality. Thus we were able to view most photos quite well. When an even more detailed

view was required, photos were imported into Adobe Photoshop. When necessary, we were

also able to use Photoshop to adjust lighting (e.g., darken or lighten) or to sharpen slightly

blurred images.

We adopted the coding procedure used by Harker and Keltner (2001), which draws

on the Facial Action Coding System (FACS; 1976,1978) developed by Eckman and Friesen.

We coded the intensity of action in the two muscle action units associated with smiling. The

first, AU6 (orbicularis oculi), causes raised cheeks and bagging/squinting around the eyes.

The second, AU12 (zygomatic major), causes smiling via raising the corners of the mouth in

an upward motion. Each action unit was coded individually by using a scale which ranged in

intensity from a 1 (Minimal intensity) to 5 (Maximum Intensity/Extreme). As per the

procedures used by Harker and Keltner, we then added the two scores together in order to

create a continuous score for Smile Intensity. Thus, the most intense smile would receive a

Smile Intensity score of 10, whereas a person with minimal activation of the muscles would

receive a score of 2.

All photos were coded by the author and at least one other trained research assistant.

A random subset (20%) were also coded by at least four trained coders. Only the author had

access to self-report information about the participants (although those data were kept

separate during the coding process). The other coders did, however, have access to

participants' Facebook pages. In order to minimize the chance for biases (and in order to

facilitate comparison across photographs and across studies), a master coding document was

created which contained multiple examples (i.e., photos) for each possible score; thus (over

16

time) we created a system in which we were able to compare any given photograph with

those previously scored. Inter-rater reliability was high (a ~ .89).

Results

Smile Intensity

The mean for coded Smile Intensity in participants' Facebook profile photos was 8.18

(SD = 2.10), with a possible range of 2 to 10. A majority of participants (87.4%) were

engaged in some sort of smile in the photo coded. But only 15 of 48 (31.3%) were rated as

being in the top category of Smile Intensity.6

Smile Intensity and Life Satisfaction

Our first goal was to determine whether the Smile Intensity codes would be correlated

with the various self-report measures of life satisfaction. Well-being was assessed at three

time points: the beginning of the first semester at school (Time 1), toward the end of the first

semester at school (Time 2), and toward the end of participants' final semester at college

(Time 3; 3.5 years later). Photos used to code Smile Intensity were retrieved at Time 2.

Because Time 1 occurred at the beginning of the first semester at school—two months prior

to retrieval of the photos used to code Smile Intensity—we were less firm in our predictions

about the strength of relationship between Smile Intensity and self-reported well-being at

Time 1. This matter was further complicated by the fact that only 43 of 48 participants

completed the measures at Time 1. Nevertheless, we did expect the relationships to be

positive. We also predicted that the well-being measures collected at Time 2 and Time 3

(and completed by all participants) would be correlated with Smile Intensity.

As predicted, Smile Intensity was correlated with all measures of subjective well-

being across each of the three time points. Specifically, Smile Intensity was associated with

17

higher scores on the SWLS at Time 1 (r - .27, p - .09), Time 2 (r - .36, p =- .01), and at

Time 3 (r = ,46>p < .01). These relationships were supported by similar findings for Felt

Understanding. Specifically, Smile Intensity was correlated with Felt Understanding at Time

1 (r - .29,/? - .06), Time 2 (r = .28,/? = .06), and at Time 3 (r = .40,/? < .01). Smile

Intensity was also marginally and negatively correlated with Felt Misunderstanding at Time 1

(r ™ -.27,/? =.08). There was, however, no statistically significant relationship between the

Smile Intensity and Felt Misunderstanding at Time 2 (r - -.12,/? = .40) or at Time 3 (r = -

.21,/? =.15). Finally, Smile Intensity was correlated with Positive Affect (r = .28,/? = .057),

but not with Negative Affect (r = -.19,/? = .19); both of these variables were measured at

Time 2 only. Overall, then, results indicated that Smile Intensity coded from a single

Facebook profile photo at Time 2 (i.e., toward the end of the first semester at college) was a

moderate predictor of short-term life satisfaction, and a robust predictor of long-term life

satisfaction as assessed via numerous validated scales.

Next, we examined whether changes in self-reported well-being (as measured by the

SWLS) from the first year to the fourth year would also be predicted by Smile Intensity

(measured in the first semester of the first year). Specifically, we predicted the Time 3 (final

semester of college) SWLS scores from Time 2 (end of the first semester in college) SWLS

scores and Smile Intensity. Results showed that 39% of the variance in SWLS scores at

Time 3 (R2 ~ .39, p < .01) was explained by a model consisting of SWLS scores at Time 2 (p

- .462, t = 3.71,/? < .01) and Smile Intensity at Time 2 (p = .287, t = 2.30,/? = .026). In

other words, a change in life satisfaction (as measured by the SWLS) from the first semester

of college to the final semester of college was predicted by the degree to which participants

were smiling in their Facebook profile photo from their first semester of college.

18

Next we conducted similar analyses to examine changes in Felt Understanding. The

model showed that 25% of the variance in Felt Understanding at Time 3 (R2 = .25, p < .01)

was explained by a model consisting of Felt Understanding at Time 2 (f3 = .299, t - 2.21,/? =

.032) and Smile Intensity at Time 2 (p - .320, t = 2.37, p = .022). Thus, a change in Felt

Understanding from the first year to the last year of college was also predicted by Smile

Intensity assessed in the first semester at college.

As might be expected given that Smile Intensity was not significantly correlated with

Felt Misunderstanding at Time 2 or Time 3, only Felt Misunderstanding at Time 2 (p = .323,

t - 233, p - .024) was a significant predictor of Felt Misunderstanding at Time 3 (R3- .15,

/? ~ .025). Taken together, these results show that Smile Intensity was a unique, behavioral

indicator of life satisfaction over time—even when controlling for the influence of two types

of life satisfaction assessed nearly 3.5 years earlier.

Smile Intensity and Relationship Satisfaction

In the current study, we examined three indicators of relationship satisfaction. Our

primary measure involved self-reported responses to questions about satisfaction with

friendships and social events). In addition, we measured self-reported likelihood of

participation in future events that were presumed to be related to current relationship

satisfaction (i.e., likelihood of attending future reunions and of donating money to the school

as an alumni). The third type of relationship satisfaction was indirect—we counted the

number of On-Campus Facebook Friends listed in participants' Facebook accounts at Time 2

(i.e., toward the end of the first semester at school). We hypothesized that this number would

be a behavioral indication of students' attempts to "network" and "find community" with

19

other students during the first four months at college, and that it would (therefore) be

correlated with the self-report relationship satisfaction variables.

Smile Intensity and self reported Relationship Satisfaction. For reference, the

correlations between Smile Intensity and questions measuring self-reported satisfaction with

social relationships are reported in Table 3. All of the self-report variables were assessed at

Time 3. Smile Intensity (measured in the first semester of college) was positively correlated

with Satisfaction with Friendships (r = .37,/? < .01) and Satisfaction with Social Events (r

=.38,/? < .01) in the final semester of college. In addition, Smile Intensity was correlated

with self-rated likelihood of attending future UVA reunions (r - .29, p = .05) and likelihood

of donating to the university as alumni (r ~ .39,/? < .01). Thus, participants' expressions of

positive affect (i.e., smiles) as coded from a photo taken and made public on Facebook in

their first semester at college were a robust predictor of self-rated satisfaction with social

relationships 3.5 years later, and they also predicted self rated likelihood of engagement with

the university community after graduation.

We next tested whether Smile Intensity would predict long-term relationship

satisfaction even when controlling for self-reported life satisfaction at Time 2. To do this, we

first calculated the mean of satisfaction with Friendships and Social Events (M- 5.88, SD =

.10), and used it as the outcome variable. We considered this our primary measure of

relationship satisfaction. A regression analysis showed that 28% of the variance in the mean

for Satisfaction with Friendships and Social Events at Time 3 (R2 - .28,/? < ,01) was

explained by a model consisting of Smile Intensity (p = .319, t = 2.34, p - .024) and SWLS

at Time 2 (p ~ .316, t = 2.31, p < .025). Thus, participants' Satisfaction with Friendships and

Social events in their final semester of college was predicted by Smile Intensity measured in

20

the first semester of college—while controlling for life satisfaction in the first semester of

college. In a second analysis, we incorporated the Felt Understanding scale as the measure

of well-being at Time 2. This time, 21% of the variance in Satisfaction with Friendships and

Social Events at Time 3 (R3 = .21, p < .01) was explained by a model containing Smile

Intensity and Felt Understanding at Time 2; however, only Smile Intensity (p = .389, / =

2.82, p < .01) was a significant predictor. Thus, we showed that Smile Intensity (measured

at Time 2) was a strong and unique predictor of Satisfaction with Friendships and Social

Events in the final semester at school (i.e., 3.5 years later), even when accounting for the

influence of two types of subjective well-being measured during the first semester at school.

Next we took the mean of the two questions which asked participants to estimate their

likelihood of engaging in future behaviors presumed to be related to current relationship

satisfaction (i.e., attending reunions and donating money to the university; M= 4.60, SD -

2.07). We considered this a "secondary" measure of relationship satisfaction. We tested

whether Smile Intensity would predict the Reunions/Donations variable while controlling for

first-semester (Time 2) life satisfaction. Smile Intensity was not, however, a unique predictor

of this variable when paired with either SWLS at Time 2 or Felt Understanding at Time 2.7

Smile Intensity and Number of Facebook Friends. We reasoned that one indicator of

social engagement at college (and also potentially relationship satisfaction) would be the

number of On-Campus Facebook Friends listed in participants' Facebook accounts by the

end of their first semester at school.8 The mean number of On-Campus Facebook Friends

was 120.53 (SD = 55.82), whereas the mean for number of Other-School Friends was 136.49

(SD-71.64).9

21

As displayed in Table 3, the number of On-Campus Facebook Friends (Facebook

Friends OC; variable 5) was correlated with several of the self-reported relationship

satisfaction variables. Thus, it was (as predicted) an indicator of relationship satisfaction.

Smile Intensity was positively correlated with the number of On-Campus Facebook Friends

(r = .32,/? - .03). There was, however, no statistically significant relationship between

Smile Intensity and number of Other-School Facebook Friends (r = .12, p = .45).

Because the number of On-Campus Facebook Friends was highly correlated with the

number of Other-School Friends (r = .57,/? < .01), we wished to determine whether Smile

Intensity would remain a significant predictor of number of On-Campus Friends in a model

that also included number of Other-School Friends. The regression analysis showed that

38% of the variance in number of On-Campus Facebook Friends at Time 2 (R3 - .38,/? <

.01) was explained by a model consisting of Smile Intensity (p = .245, t = 2.05,p < .05) and

Number of Other-School Facebook Friends (p = .538, t ~ 4.50,/? < .01). Thus, Smile

Intensity was a unique predictor of the number of first semester On-Campus Facebook

Friends even when controlling for number of Other-School Friends.

Smile Intensity and Personality

One possibility, acknowledged by Harker and Keltner (2001), is that the expressions

of positive affect captured in photographs may be largely reducible to the "influence" of

personality traits. From this perspective, Smile Intensity would be a proxy for (in the

example given earlier) Extraversion. Thus it is important to address such an explanation.

Doing so within the realm of well-being research is, however, further complicated by the fact

that many measures of life satisfaction are correlated (at least to some extent) with Big 5

personality traits.

22

For reference, correlations between Smile Intensity and the various Big 5 variables

(measured at Time 1) appear in Table 4. Results show that Smile Intensity was positively

correlated with Conscientiousness (r = .36,/? - .016), and negatively correlated with

Neuroticism (r - -.31,/? = .04). Recall, however, that our Big 5 measure was completed by

only 43 of 48 participants. In spite of this loss of power, we wished to conduct a series of

exploratory analyses in order to determine whether Smile Intensity would remain a

significant predictor of future (i.e., Time 3) life satisfaction and relationship satisfaction

when controlling for the influence of traits measured by the Big 5. In order to mitigate the

influence of correlations among the Big 5 items (see Anusic, Schimmack, Pinkus, &

Lockwood, 2009), we created a Big 5 HALO term by taking a mean of the Big 5 scores (with

a reverse score for neuroticism). In doing so, our main interest was to determine whether the

long-term predictive relationship between Smile Intensity and the key well-being variables at

Time 3 would hold, controlling for the general positive responding tendency captured by the

Big 5 HALO index. As expected, Smile Intensity was marginally positively correlated with

the Big 5 HALO index (r = .28,/? « .074).

Predicting Life Satisfaction at Time 3. A regression analysis showed that 25% of the

variance in SWLS scores at Time 3 (R2 - .24,/? < .01) was explained by a model consisting

of Smile Intensity (p = .309, t = 2.15,/? = .037) and the Big 5 Halo score (p - .304, t = 2.12,

p = .040). That is, even when controlling for the general positive responding tendency of the

Big 5, Smile Intensity measured in the first semester of college predicted life satisfaction (as

measured by the SWLS) in the final semester of college (i.e., 3.5 years later). We used this

same approach in order to predict Felt Understanding at Time 3. The regression analysis

showed that 20% of the variance in Felt Understanding at Time 3 (R3 - .20,/? - .012) was

23

explained using a model consisting of Smile Intensity and the Big 5 HALO; however, only

Smile Intensity was a significant predictor (p = .303, t = 2.06, p - .046), whereas the

influence of the Big 5 HALO was only marginally significant (p - .253, r - 1.71,/? - .094).

Thus, Smile Intensity measured in the first semester at college was a unique predictor of two

types of life satisfaction measured 3.5 years later—even when controlling for the influence of

personality.

Predicting Relationship Satisfaction at Time 3. In order to explore whether Smile

Intensity would remain a significant predictor of the relationship satisfaction variables if

personality was included as a co-predictor, we conducted an additional series of regression

analyses. We first predicted the mean of our primary measure of relationship satisfaction

(Satisfaction with Friendships and Social Events). Results showed that 21% of variance in

relationship satisfaction (R2~ .21, p < .01) was explained by a model that included Smile

Intensity and the Big 5 HALO; however, only Smile Intensity was a significant predictor (p -

.437,1 = 3.00,/? < .01). Thus Smile Intensity assessed in the first semester at college was a

strong predictor of self-reported relationship satisfaction 3.5 years later—even after

controlling for the effects of personality as measured by the Big 5.

Next we predicted the mean of our secondary relationship satisfaction measure

(participants' self-reported likelihood of attending future reunions and of donating money to

the university in the future) (M- 4.60, SD = 2.07) from Smile Intensity and personality.

While the overall model explained 41% of the variance on this variable (R2 =- .41,/? < .01),

only the Big 5 HALO (p = .556, t - 4.40, p < .01) was a statistically significant predictor.

Finally, we predicted Number of On-Campus Facebook Friends at Time 2 from Smile

Intensity, Number of Other-School Facebook Friends, and the Big 5 HALO score. Results

24

showed that 39% of the variance (R2 = .39,/? < .01) in the number of On-Campus Facebook

Friends at Time 2 could be explained by a model consisting of Smile Intensity (p = .238,1 ~

2.16, p - .037), number of Other-School Facebook Friends (p == .548, t ~ 3.88,/? < .01), and

the Big 5 HALO (P = -.050, t = -.346, p = . 731). Thus, Smile Intensity predicted Number of

On-Campus Facebook Friends, even when controlling for Number of Other-School Facebook

Friends and for the positive responding tendencies captured by the Big 5 HALO score.

Mediation Analyses

Harker and Keltner (2001) speculated that Smile Intensity might have a beneficial

effect on relationships, and that this could be positively associated with long-term life

satisfaction. In order to address this hypothesis, we tested whether our primary measure of

relationship satisfaction (i.e., the mean for Satisfaction with Friendships/Social Events)

mediated the relationship between Smile Intensity and participants' SWLS scores at Time 3.

We used the SPSS macro for simple mediation (Preacher & Hayes, 2004), and followed the

steps outlined by Baron and Kenny (1986). Results showed that no mediation was present.

This was confirmed by a Sobel test (Sobel - 1.39,/? = .18).

In a second model, we tested whether relationship satisfaction (i.e., Satisfaction

Friendships/Social Events) mediated the relationship between Smile Intensity and Felt

Understanding at Time 3. We again used the steps outlined by Baron and Kenny. Again, the

results showed that no mediation was present in the model (Sobel ~ 1.64, p ~ 117). In this

particular sample, then, relationship satisfaction was not a statistically significant mediator of

the relationship between Smile Intensity and long-term life satisfaction.

Discussion

25

In the current longitudinal study, we showed that the expression of positive affect

captured by and coded from a single Facebook profile photo from students' first semester at

college predicted self-reported life satisfaction 3.5 years later, as the participants were about

to graduate from college. This finding held even when we controlled for the influence of

personality (as measured by the Big 5), and even when we controlled for the influence of two

measures of self-reported life satisfaction (i.e., SWLS and Felt Understanding) measured at

Time 2. Thus, Smile Intensity was shown to be a strong and unique predictor of future life

satisfaction.

In addition, we showed that Smile Intensity was uniquely related to two types of

relationship satisfaction. At the end of participants' first semester at school, for example,

Smile Intensity predicted the number of On-Campus Facebook Friends (a behavioral measure

of relationship satisfaction). This finding held even after controlling for the number of

Other-School Facebook Friends, and for Big 5 personality. Further, Smile Intensity at Time

2 also predicted our primary measure of relationship satisfaction (i.e., self-reported

Satisfaction with Friendships/Social Events) 3.5 years later (when participants were about to

graduate from college). This finding held even when we controlled for the influence of

personality and for both types of life satisfaction measured at Time 2.

Taken together, then, this longitudinal study provides new and compelling evidence

that the expression of positive affect in informal photographs—specifically those posted to

Facebook as profile photos during participants' first semester at college—can predict current

and future life satisfaction and relationship satisfaction. Thus, we did indeed replicate the

overall findings reported by Harker and Keltner (2001). Further, we showed that these

effects held even when controlling for two types of life satisfaction at Time 2 and for the

26

influence of personality. Results showed, however, that relationship satisfaction was not a

statistically significant mediator of the relationship between Smile Intensity and long-term

life satisfaction.

Limitations

While intriguing, the results reported in Study 1 had several key limitations. First and

foremost was the modest size of our sample. A second limitation was the loss of power

incurred because 10% of the participants did not complete the measure to assess Big 5 traits

(i.e., at Time 1). This latter issue was particularly troublesome in that it is important for us to

be able to have a thorough accounting of the influence of personality—especially at this early

stage of our research program.

A third limitation of the study was, in essence, the "downside" of one of the likely

"upsides" of this sort of longitudinal design: namely, that these results are "the product" of a

single cohort of college students and based on behaviors and events that took place during a

particular four-year period in their lives. While the use of cohorts in research can be

extremely useful for mitigating the effects of various types of "real world" influences, this

design can also have the potential to be somewhat of a "challenge" when attempting to

generalize findings to individuals who were not part of the cohort. This issue may be

especially likely during periods of time in which mass events or substantial cultural shifts

have taken place that could make one cohort "different" from another.10

With the latter point in mind, it could be argued that the cohort of participants in the

current study were indeed participants in a major cultural shift—they were members of the

first national cohort of college students to use Facebook (as a vehicle for social networking

and for the sharing of personal information) from their first to their final semester at college.

27

Why might this matter? Perhaps it does not. However, within one year of the start of the

current study (2005), Facebook usage increased exponentially (amongst college students, and

later the general public). Thus it is at least possible that the rapid expansion of and upgrades

to Facebook that took place during the period in which our study was conducted (i.e., 2005-

2009) could make for as-yet-unknown limitations in our current results. For example,

changing norms about the "content" of people's Facebook profile photos or how often they

"should" be updated (or even how easy it is for people to do the updates) could conceivably

have the ability to alter the predictive utility of such photographs with regard to our research.

In one such scenario, an increase in the frequency of updates might signify less endorsement

of any given photo as a representation of "self." As we will discuss later in the paper, the

sheer volume of data shared by today's college students (including photos of themselves)

may be both a boon and a great challenge when conducting studies that attempt to code

photographs—whether online or offline. Thus, more research is needed in order to

understand whether the findings reported in the current study are likely to generalize to other,

subsequent cohorts—and to people in general.

Study 2

To our knowledge, Study 1 is the first study to utilize people's Facebook profile

photographs as a way to measure the expression of positive affect while incorporating that

behavioral "information" into a longitudinal study to predict well-being. Several of the

results reported are highly promising and may—if replicated—have the potential increase

and broaden the relevance of this variable (i.e., smile intensity) as a distinct marker of long-

term well-being. Given such incentives (and also the stakes of being too quick to go public

with "preliminary" results), we felt that it was important to attempt to replicate the overall

28

design and findings reported in Study 1 in a second cohort of college students. While doing

so, we extended our initial work in several ways. In the current study, we again track a

cohort of college students from their first semester at college (2006) until their final semester

(just before graduation, 3.5 years later). Thus, we are again able to incorporate and to

"control for" any of the Facebook-related changes that took place during that period of time.

In addition, we were able to incorporate Big 5 data from all of our participants.

Method

Participants

Participants were 109 first-year, first-semester undergraduates at the University of

Virginia in fall 2006. Only first-year students who had an active Facebook account were

eligible to participate. In addition, this study involved self-report data collected at three time

points over the course of four years, and it required participants to have a code-able

photograph on Facebook. Participants with substantial missing data at Time 2 or Time 3

and/or those with no code-able Facebook photo (n - 73) were dropped from the final dataset.

This left a total of 36 participants (13 male). The age of these participants ranged from 17 to

19, with a mean age of 17.89 (SD = .32). In terms of ethnicity, 2 (5.6%) were African-

American, 2 (5.6%) Asian, 27 (75%) European American, and 5 (13.9%) were categorized as

"Multi-Racial/Ethnic." All participants received partial fulfillment of their course

requirement in exchange for participation at Time 1 and Time 2; those who completed Time

3 were entered into a lottery to win one of five cash prizes.

Procedure and Materials

The overall procedures for the current study were largely identical to those used for

Study 1. However, the Satisfaction With Life Scale (SWLS) was the only measure of life

29

satisfaction to be given at each of the three time points in the study. Further, the Big 5

measure was given at Time 3 rather than at Time 1.

Time 1 (Beginning of first semester at college). A majority of participants (n - 26)

completed the Satisfaction with Life Scale (SWLS) (Diener, Emmons, Larsen, & Griffin,

1985; 1-Strongly disagree to 7=Slrongly agree; a = .84). In addition, participants answered

one question that asked how often they update the content on their Facebook page (l=Rarely

to 4-Frequently).

Time 2 (End of first semester at college). All participants (n = 36) completed the

SWLS (a = .92). In addition, they completed the Positive Negative Affect Schedule

(PANAS; Watson, Clark, & Tellegen, 1988; PA a - .81 ; NA a = .83). At this time,

participants' Facebook profiles (including main profile photographs and number of Facebook

friends) were retrieved and saved for coding.

Time 3 (Toward end of final semester at college; 3.5 years later). All participants (n

~ 36) completed the SWLS (a ~ .91), and the Felt Understanding (a = .85) and Felt

Misunderstanding Scales (a - .77) (Oishi, Krochik, & Akimoto, 2010). In addition, we

assessed Big Five personality traits (Gosling, Rentfrow, 8c Swann, 2003; extraversion a =

.80; neuroticism a — .64; agreeableness a - .48; conscientiousness a = .72; openness a -

.71; a possible range of 1 to 6). Finally, participants also completed numerous questions

related to relationship satisfaction (e.g., friendships; social events; and the degree to which

they feel they "fit in"; l=Extremely dissatisfied to 7=Extremely Satisfied), as well as

anticipated behaviors related to relationship satisfaction (e.g., likelihood of attending future

reunions, and of donating money to the university as alumni). Participants were also asked to

estimate the frequency with which they login to their Facebook account and the frequency

30

with which they update content on their Facebook page (l=Multiple times per day to 7=Less

than once per month)

Procedure for Coding the Expression of Positive Affect in Photographs

The procedures used for coding the photographs were the same as those used in Study

1. Inter-rater reliability was high (a - .84). As per Harker and Keltner (2001), we again

used a single Facebook profile photo to code for Smile Intensity.

Results

Smile Intensity

The mean for coded Smile Intensity in participants' Facebook profile photos was 6.78

(SD = 2.37), with a possible range of 2 to 10. A majority of participants (77.8%) were

engaged in some sort of smile in the photo coded. But only 13 of 36 (36%) were rated as

being in the top category of Smile Intensity.

Smile Intensity and Well-Being

Smile Intensity was correlated with all measures of subjective well-being across each

of the three time points. Specifically, Smile Intensity was associated with higher scores on

the SWLS at Time 1 (r « .46,/? - .02), Time 2 (r = 35, p = .037), and at Time 3 (r - .51,/?

< .01). These relationships were supported by similar findings for Positive Affect (r =* .50,/?

< .01) at Time 2, and for Felt Understanding (r =.45,/? < .01) at Time 3. In addition, Smile

Intensity was negatively correlated with NA at Time 2 (r - -.30,/? - .08) and Felt

Misunderstanding at Time 3 (r - -.47,/? <.01). Overall, then, these initial results replicated

the correlational findings reported in Study 1 in a second cohort, and indicated that

expressions of positive affect coded from a single Facebook profile photo in students' first

31

semester at college were a robust predictor of short-term and long-term life satisfaction (as

assessed by several validated measures of life satisfaction).

Next, we examined whether changes in participants' self-reported life satisfaction

from Time 2 to Time 3 (i.e., from the first to the fourth years) would also be predicted by

Smile Intensity. Specifically, we predicted the Time 3 (final semester of college) SWLS

scores from the Time 2 (end of the first semester in college) SWLS scores and Smile

Intensity. Results showed that 60% of the variance in life satisfaction at Time 3 (R2 = .60,/?

< .01) was explained by a model consisting of life satisfaction at Time 2 (p = .617, t = 5.22,/?

< .01) and Smile Intensity at Time 2 (P = .295, t = 2.50,/? = .018). On par with the results

reported in Study 1, then, the change in life satisfaction from the first semester of college to

the final semester of college was predicted by Smile Intensity assessed in the first semester of

college.

Smile Intensity and Relationship Satisfaction

As in Study 1, we assessed two types of relationship satisfaction variables: self-

report and indirect (i.e., number of Facebook friends).

Smile Intensity and self-reported Relationship Satisfaction. Smile Intensity in the

first semester of college (Time 2) was positively correlated with both Satisfaction with Social

Events (r- .37, p — .029) and Satisfaction with Friendships (r- .42, p = .011). In a marked

divergence from the results of Study 1, however, Smile Intensity was neither correlated with

self-rated likelihood of attending future UVA reunions (r = .23,/? = .19) nor with likelihood

of donating to the university as an alumni (r - .07, /? = .70). Indeed, the alumni donation

question was uncorrected with all measures of well-being. While there may be many

reasons for such a change from the first cohort—including the smaller sample size in the

32

current study—it also seems plausible that the downturn in the American economy and the

tight job market that continued in the year 2010 (as the current participants were graduating

from college) could have made these individuals less able to imagine having "money to

spare" in the future (for donations or for return trips to attend reunions). Regardless of such

issues, Smile Intensity as coded from participants' first semester Facebook profile photos did

remain a robust predictor of self-reported satisfaction with Friendships and Social Events 3.5

years later. As in Study 1, this was our primary measure of relationship satisfaction.

We next tested whether Smile Intensity would predict long-term relationship

satisfaction (Time 3) even when controlling for self-reported life satisfaction at Time 2. To

do this, we again calculated the mean of Satisfaction with Friendships and Social Events (M

= 5.83, SD = 1.27) and used it as the outcome variable. A regression analysis showed that

20% of the variance in the mean for Satisfaction with Friendships and Social Events at Time

3 (R2 = .20, p = .025) was explained by a model consisting of Smile Intensity (p - .385, t =

2.32,/? - .027) and SWLS at Time 2 (p = .131, t = .792, p = .43). Thus, Satisfaction with

Friendships and Social events in the final semester of college was (as in Study 1) predicted

by expressions of positive affect coded from profile photos uploaded to Facebook during the

first semester of college; this finding held even when controlling for life satisfaction

measured during that first semester.

Smile Intensity and Number of Facebook Friends. In Study 1, the number of On-

Campus Facebook friends was predicted by Smile Intensity. One major change that occurred

in the year between our first and second cohorts, however, was the rapid expansion and

evolution of Facebook as a venue for social networking amongst college students. In the

current study, the mean number of On-Campus Facebook Friends (at Time 2) was 187.39

33

(SD ~ 124.26), and the mean for number of Other-School Friends was 205.69 (SD = 115.56).

Thus, participants in Study 2 began their college life with an average of 65% more Facebook

"friends" (of both types) than the first year students who started college the previous fall. As

with Study 1, we expected and found no correlation between Smile Intensity and number of

Other-School Facebook Friends (r = .16,/? = .36). Unlike Study 1, however, this time we

found only a marginal positive relationship between Smile Intensity and the number of On-

Campus Facebook Friends (r = .28,/? ™ .10). Smile Intensity, then, was no longer a

compelling predictor of number of On-Campus Facebook Friends (our "indirect" measure of

relationship satisfaction).

Regardless of this difference, we did replicate the most important findings from Study

1. Namely, Smile Intensity coded from photographs during students' first semester at college

was a strong and unique predictor of our primary measure of relationship satisfaction (i.e.,

self-rated Satisfaction with Friendships and Social Events)—even when controlling for the

influence of life satisfaction at Time 2.

Smile Intensity and Personality

One improvement in the current study was that all participants contributed Big 5

ratings. We next conducted a series of analyses in order to determine whether Smile

Intensity would remain a significant predictor of the various indicators of well-being when

controlling for the influence of traits measured by the Big 5. To do this, we again created a

Big 5 HALO term by taking a mean of the Big 5 scores (with a reverse score for neuroticism;

see Anusic, Schimmack, Pinkus, & Lockwood, 2009). As in Study 1, Smile Intensity was

marginally positively correlated with the Big 5 HALO index (r = .30, p - .077).

34

Predicting Life Satisfaction at Time 3. Our first task was to predict life satisfaction

(SWLS) at Time 3. A regression analysis showed that 31% of the variance in SWLS scores

at Time 3 (R2~ .31,/? < .01) was explained by a model consisting of Smile Intensity (p =

.439, / = 2.90,/? < .01) and the Big 5 Halo score (P = .240, t - 1.58,/? < .038). Next, we

predicted Felt Understanding at Time 3. The analysis showed that 32% of the variance in

Felt Understanding at Time 3 (R2 = .32,/? < .01) was predicted by Smile Intensity (p = .348, /

- 2.30,/? = .028) and the Big 5 Halo score (P - .349, i = 2.30,/? - .027). Thus, using two

different measures of life satisfaction we again showed that expressions of positive affect

measured during the first semester at college predicted life satisfaction 3.5 years later—even

when controlling for the general positive responding tendency of the Big 5.

To show in the most stringent manner that Smile Intensity represents a unique

predictor of life satisfaction, we next predicted the change in SWLS scores at Time 3 from

the SWLS scores at Time 2, along with Smile Intensity and the Big 5 Halo. The regression

analysis showed that 63% of the variance in SWLS at Time 3 (R2 = .63,/? < .01) was

explained by the model; however, only Smile Intensity (p - .248, / = 2.08,/? - .046) and

SWLS at Time 2 (p = .599, t = 5.16,/? < .01) were significant predictors. Thus, because we

had full Big 5 data for all participants in the current study, we were able to show that Smile

Intensity continued to predict the change in self-reported life satisfaction from Time 2 to

Time 3 even while controlling for the influence of personality.

Predicting Relationship Satisfaction at Time 3. We also wished to explore whether

Smile Intensity would remain a significant predictor of satisfaction with Friendships and

Social Events (i.e., our primary measure of relationship satisfaction) if the Big 5 HALO was

included in the model. Results showed that 21% of the variance in this variable (R2 - .21,/?

35

< .01) was explained by a model that included Smile Intensity and the Big 5 HALO;

however, only Smile Intensity was a significant predictor (p = .368, t = 2.29, /? = .028).

Thus, the expression of positive affect coded from Facebook profile photos from students'

first semester at college predicted relationship satisfaction (i.e., Satisfaction with Friendships

and Social Events) 3.5 years later—even when controlling for the influence of personality.

Mediation Analyses

As in Study 1, in order to test whether relationship satisfaction mediated the

relationship between Smile Intensity and SWLS scores at Time 3 we used the SPSS macro

for simple mediation (Preacher & Hayes, 2004). We also followed the steps outlined by

Baron and Kenny (1986). First, Smile Intensity was a significant predictor of SWLS at Time

3 (p = .510,/? < .01). Second, Smile Intensity predicted relationship satisfaction, the

proposed mediator (p = .431,/? < .01). Third, relationship satisfaction predicted SWLS

scores, controlling for Smile Intensity (p = .357, p - .027). Fourth, because the effect of

Smile Intensity on SWLS scores remained significant when controlling for relationship

satisfaction (p = 351, p - .027), this suggested that partial mediation occurred. A Sobel test

confirmed that the model was marginally significant (Sobel = 1.71,/? = .086).

Next, using the same methods as above, we tested whether relationship satisfaction

mediated the relationship between Smile Intensity and Felt Understanding at Time 3. First,

Smile Intensity was a significant predictor of Felt Understanding at Time 3 (p = .452,/? <

.01). Second, Smile Intensity was a significant predictor of the mediator (i.e., relationship

satisfaction; p = .431,/? < .01). Third, the mediator significantly predicted Felt

Understanding, controlling for Smile Intensity (p = .522,/? < .01). Finally, because the effect

of Smile Intensity on Felt Understanding did not remain significant when controlling for the

36

mediator (p = .227, p < .130), this suggested that full mediation occurred in the model. A

Sobel test confirmed that the model was significant (Sobel = 2.33,/? = .023).

Unlike Study 1, then, the results of the current study did provide evidence for

mediation. Specifically, relationship satisfaction fully mediated the relationship between

Smile Intensity and Felt Understanding (Time 3). Additionally, the model showing that

relationship satisfaction partially mediated the relationship between Smile Intensity and

SWLS scores at Time 3 was marginally significant.

Discussion

The primary objective of the current study was to attempt to replicate the major

findings from Study 1—that expressions of positive affect coded from our participants' main

Facebook profile photos in their first semester at college could be used to predict long-term

life satisfaction and relationship satisfaction. To do this, we recruited a second cohort of

college students in their first semester at college (these students started college a year later

than those run in Study 1), and tracked them for 3.5 years (from their first until their final

semester at school). In addition, because we considered it highly important to address the

issue of personality (i.e., how personality might relate to Smile Intensity), we took extra steps

to insure that all participants in Study 2 completed the Big 5 measure.

All of the major findings from Study 1 were replicated in the current study.

Expressions of positive affect from a single Facebook profile photo retrieved during students'

first semester at college predicted self-reported life satisfaction 3.5 years later (i.e., from

Time 2 to Time 3), as the participants were about to graduate from college. This finding

again held even when we controlled for the influence of personality (as measured by the Big

5), and even when we controlled for the influence of self-reported life satisfaction (i.e., the

37

SWLS) measured at Time 2. Thus, Smile Intensity was again shown to be a strong and

unique predictor of future life satisfaction.

We also showed that Smile Intensity (measured in the first semester at college)

predicted self-reported relationship satisfaction (i.e., Satisfaction with Friendships/Social

Events) 3.5 years later (when participants were about to graduate from college); this finding

held even when we controlled for life satisfaction and also for personality.

In a marked contrast from the results of Study 1, mediation analyses showed that

relationship satisfaction was a mediator between Smile Intensity and subjective well-being.

Specifically, Felt Understanding (measured at Time 3) fully mediated the relationship.

Additionally, the model to test whether relationship satisfaction mediated the relationship

between Smile Intensity and Satisfaction With Life Scale (SWLS) scores at Time 3 was

marginally significant and indicated that partial mediation occurred. Given that the current

study was under-powered (n — 36), it seems plausible that the marginal findings could be due

(at least in part) to sample size. Further, because these results were different from those in

Study 1, we wished to learn whether evidence for mediation would be present if we collapsed

the data for Study 1 and Study 2. We will address this issue in the following section.

Combined Mediation Analyses: Study 1 and Study 2

Using the combined data from Study 1 and Study 2(n~ 84), we tested whether our

main measure of relationship satisfaction (the mean for Satisfaction with Friendships/Social

Events) mediated the relationship between Smile Intensity and SWLS scores at Time 3. To

do this, we again used the Preacher and Hayes (2004) SPSS Macro for Simple Mediation; we

also followed the Baron and Kenny (1986) steps. First, Smile Intensity was a significant

predictor of SWLS scores at Time 3 (p = .446,/? < .01). Second, Smile Intensity predicted

38

the mediator (i.e., relationship satisfaction; p = .420,/? < .01). Third, relationship satisfaction

predicted SWLS scores, controlling for Smile Intensity (p = .286,/? < .01). Fourth, because

the effect of Smile Intensity on SWLS scores remained significant when controlling for

relationship satisfaction (p - .326,/? = .01), this suggests that partial mediation occurred. A

Sobel test confirmed that the model was statistically significant (Sobel = 2.38,/? = .019).

Next we tested whether relationship satisfaction mediated the relationship between

Smile Intensity and Felt Understanding at Time 3 in the combined dataset. Smile Intensity

was a significant predictor of Felt Understanding at Time 3 (p = .406,/? < .01). Smile

Intensity was also a significant predictor of the mediator (i.e., relationship satisfaction; p =

.420,/? < .01), and the mediator significantly predicted Felt Understanding, controlling for

Smile Intensity (p = .386,/? < .01). Finally, because the effect of Smile Intensity on Felt

Understanding remained significant when controlling for the mediator (i.e., relationship

satisfaction; p ~ .244, /? =.021), this again suggested that partial mediation occurred in the

model. A Sobel test confirmed that the model was indeed significant (Sobel = 2.79, p < .01).

Discussion

One concern about Study 1 and Study 2 was that the size of each sample was

modest (n = 48 and n ~ 36 respectively). Additionally, the mediation results were

conflicting. Specifically, the results for Study 2 showed evidence for statistically

significant mediation, whereas the results for Study 1 did not. In attempt to bring about a

"tie breaker," we collapsed the data for the two studies (thereby increasing power). We

then used the combined dataset in order test for mediation. Results showed that the

relationship between Smile Intensity and long-term life satisfaction (as measured by both

the Satisfaction with Life Scale and the Felt Understanding Scale) was partially mediated

39

by relationship satisfaction.11 This is an intriguing finding which we will address in

greater detail in the General Discussion.

Study 3

One of the biggest remaining "limitations" of the current program of research would

appear to be the use of a single photograph in order to code the expression of positive affect

and to predict life satisfaction and relationship satisfaction. In this study, we take an initial

step toward addressing the issue.

In the current study, we track a third cohort of college students from the beginning of

their first semester at college. In this study, however, we coded Smile Intensity using

multiple Facebook profile photographs from the first semester (rather than only a single

photo). Use of multiple photographs will enable us to examine the within-person mean and

the within-person standard deviation for Smile Intensity, and to consider the degree to which

both relate to short- and long-term life satisfaction and relationship satisfaction.

Additionally, this cohort participated in their "Time 3" follow-up study at the end of

their second year (rather than at the end of their fourth year). Thus, in this study we will also

be able to address questions about whether Smile Intensity coded in the first semester at

college will predict well-being at the mid-point of students' college careers.

Method

Participants

Participants were 82 first-year, first-semester undergraduates at the University of

Virginia in fall 2008. Only first-year students who had an active Facebook account were

eligible to participate in the study. Participants were required to have multiple code-able

photographs on Facebook. Those who had substantial missing data at Time 2 or Time 3

40

and/or those who had too few Facebook profile photos (n =14) were dropped from the final

dataset. This left a total of 68 participants (13 male). All participants received partial

fulfillment of their course requirement in exchange for participation at Time 1 and Time 2;

those who completed Time 3 were entered into a lottery to win one of five cash prizes.

Procedure and Materials

With the exception of the inclusion of multiple photographs in order to code Smile

Intensity and the shortened (2-year) time frame of the study, the procedures for the current

study were largely identical to those used in Study 2.

Time 1 (Beginning of first semester at college). A majority of participants (n - 66)

completed the Satisfaction with Life Scale (SWLS; Diener, Emmons, Larsen, 8c Griffin,

1985; l=Strongly disagree to 7=Strongly agree; a = .86). In addition, participants answered

one question that asked how often they update the content on their Facebook page (1 =Rarely

to 4=Frequently).

Time 2 (End of first semester at college). All participants (n ~ 68) completed the

SWLS (a = .90). In addition, they completed the Positive Negative Affect Schedule

(PANAS; Watson, Clark, & Tellegen, 1988; PA a = .87 ; NA a = .82).

Time 3 (Toward end of their second year at college; 1.5 years later). All participants

(n = 68) completed the SWLS (a = . 77), and the Felt Understanding (a = .90) and Felt

Misunderstanding Scales (a ^ .76) (Oishi, Krochik, & Akimoto, 2010). We also assessed

Big Five personality traits (Gosling, Rentfrow, & Swann, 2003; extraversion a - .82;

neuroticism a ~ .71; agreeableness a — .53; conscientiousness a = .78; openness a ~ .74; a

possible range of 1 to 6), and the Subjective Happiness Scale (Lyubomirsky & Lepper, 1999;

a - .78). Finally, participants also completed numerous questions related to relationship and

41

personal satisfaction (e.g., friendships; social events; the degree to which they feel they "fit

in"; and how satisfied they felt with their "smile"; l~Extremely dissatisfied to 7~Bxtremely

Satisfied), as well as anticipated behaviors related to relationship satisfaction (e.g., likelihood

of attending future reunions, and of donating money to the university as alumni). Finally,

participants were asked to estimate the frequency with which they login to their Facebook

account and the frequency with which they update content on their Facebook page

(l=Multiple times per day to 7=Less than once per month)

Procedure for Coding the Expression of Positive Affect in Multiple Photographs

The general procedures used for coding photographs were the same as those used in

the prior two studies. However, in this study we coded multiple profile photos for each

participant. Because participants had varying numbers of code-able photographs (ranging

from 3 to dozens), we settled on the use of three photos for this initial study. The decision to

use this specific number of photos was largely practical—it enabled us to retain the

maximum number of participants (i.e., many would have been eliminated because they had

only three photos). In the majority of cases, we coded the profile photos in the order in

which they appeared in participants' history starting from the day the data were saved (i.e.,

we coded the three most recent photos). Inter-rater reliability was adequate (a ~ .74).

Results

Smile Intensity

We took the mean and the standard deviation for the three coded photos. The mean

for Smile Intensity was 7.79 (SD - 1.35). The mean of the standard deviation for Smile

Intensity was 1.32 (SD - .77). A majority of participants (97%) were engaged in some sort

42

of smile in the photo coded, but only 22 of 68 (32%) were rated as being in the top category

of Smile Intensity.

Smile Intensity and Well-Being

Well-being was assessed at three time points: the beginning of the first semester at

college (Time 1), toward the end of the first semester at college (Time 2), and toward the end

of participants' second year at college (Time 3; 1.5 years later). As in the prior two studies,

we predicted that the well-being measures collected at all time points would be correlated

with Smile Intensity.

Smile Intensity was, as predicted, correlated with the results for the SWLS at each of

the three time points. Specifically, Smile Intensity was associated with higher scores on the

SWLS at Time 1 (r = .31,/? - .01), Time 2 (r = 33, p < .01), and at Time 3 (r = .27,/? =

.029). Contrary to predictions, however, Smile Intensity was not correlated with Positive

Affect (r ~ .19,/? = .13), Felt Understanding (r - -.08/? =.50), or with the Subjective

Happiness Scale (r — .04, p = .76). In addition, Smile Intensity was not significantly related

to Negative Affect at Time 2 (r ~ .00,/? = .94), or to Felt Misunderstanding at Time 3 (r = -

.12,/? -.34). The direction of the relationship with Felt Misunderstanding was, however, in

the predicted direction (i.e., it was negative).12 Although Smile Intensity did predict the

SWLS scores at each time point, the other data patterns were highly atypical if viewed within

the context of the findings for the past two studies. As the results shown in Table 5 attest,

however, the relationships among the self-report well-being variables appeared to be rather

typical (e.g., the SWLS was positively correlated with the other well-being measures such as

the Subjective Happiness Scale and PA).

43

We next examined whether changes in life satisfaction from the first semester to the

end of the second year would be predicted by Smile Intensity (as measured during the first

semester). To do this, we predicted the Time 3 (fourth semester in college) SWLS scores

from the Time 2 (first semester in college) SWLS scores and Smile Intensity. Results

showed that 34% of the variance in SWLS at Time 3 (R2 - .34, p < .01) was explained by the

model; however, only SWLS at Time 2 (p = .549, t - 5.14,/? < .01) was a significant

predictor. Smile Intensity, again showed a highly atypical pattern (p =.083, t - .780, p -

.438).

Smile Intensity and Relationship Satisfaction

We examined the associations between Smile Intensity and the relationship

satisfaction variables. Again, the patterns were atypical given what we observed in the past

two studies. Smile Intensity was uncorrelated with the mean for Satisfaction with

Friendships and Social Events (r - -.02,/? = .87), as well as with the mean for likelihood of

attending reunions and for donating to the university in the future (r ~ .14,/? = .26). Both of

these variables, however, showed typical associations (i.e., they were correlated) with the

other well-being variables.

Additional Exploratory Analyses

Given that we had no outliers in the dataset and that the relationships among the self-

report variables were all in keeping with trends observed in the past studies, it seemed that

the "issue" must have been related to the means for our Smile Intensity codings. To

investigate, we checked multiple times for data entry errors and also for coding irregularities.

In addition, we even re-coded a sample of the photographs (25%, and then a full 50%) for

comparison purposes—all to no avail. Examination of the individual means for the three

44

profile photos from which the Smile Intensity means were calculated produced almost

identical (i.e., highly atypical) relationships with the self-report variables of interest. Thus,

nothing appeared to be objectively "wrong" with either our coding of the photographs or with

the self-report data. We were, at that point, forced to concede that there simply appeared to

be little "understandable" relationship between mean Smile Intensity in the current study and

the self-report variables of interest (i.e., life satisfaction and relationship satisfaction).

We next examined whether our hypothesis would be supported in subgroups of this

sample. Specifically, we performed a median split in order to create a "Low" Smile Intensity

group (« = 31) and a "High" Smile Intensity group (n - 37).

First, we conducted independent samples /-tests in order to look for mean differences

between the Low Smile Intensity and High Smile Intensity groups. On the self-report

variables of interest, only two differences between the groups were evident. Participants

categorized in the "Low" Smile Intensity group had marginally lower SWLS scores at Time

3 (M= 5.45, SD = .63) than did those in the "High" Smile Intensity group (Af« 5.79, SD =

.78; t(66) = -1.98, p - .052, d = -.49). Additionally, the "Low" group members also had

marginally lower scores on the mean for likelihood of attending reunions and for donating to

the university in the future (W= 4.19, SD - 1.54) than did those in the "High" Smile

Intensity group (Af = 4.85, SD - 1.37; t(66) - -1.87,/? = .066, d= -.04).

Next, we ran correlations for the two groups. Surprisingly, the overall results for

participants in the Low Smile Intensity group (Table 6) appeared to be at least somewhat

typical of the relationships found in the past two studies. Smile Intensity was, for example,

positively correlated with the three SWLS scores, and with Positive Affect (r ~ .38,/? =

.036). Further, the other relationships (with the exception of Felt Understanding) were all in

45

the "expected" directions. However, when we attempted to predict SWLS scores at Time 3

from Smile Intensity and SWLS at Time 2, only SWLS at Time 2 was a significant predictor

(p = .561, t = 3.12,/? < .01; R2 = .32,/? < .01). Thus, these data were still not fitting the

patterns established in either Study 1 or Study 2.

Looking at the correlations for the "High" Smile Intensity group showed results that

were quite contrary to those found in the "Low" group. First, for participants in the High

Smile Intensity group (Table 7), Smile Intensity was negatively related to each of the

"positive" well-being scales. This included a marginally significant negative correlation with

Happiness (r ~ -.30,/? = .074). Second, Smile Intensity was also significantly and positively

correlated with Felt Misunderstanding (r - .42,/? < .01), and with Negative Affect (r = .34,/?

- .042). Third, although not displayed in Table 7, participants in the High Smile Intensity

group also showed a marginally significant negative correlation between Smile Intensity and

their response to how well they "fit in" at college (r - -.30, p ~ .072). Fourth, Smile Intensity

was negatively and significantly correlated with the mean for anticipated Reunions/Alumni

donations (r ™ -.36,/? - .031). This finding was particularly surprising given that (as reported

above) the Low Smile Intensity group had a lower mean score on this variable.14 Fifth,

participants in the High Smile Intensity group showed a highly significant positive

correlation between Smile Intensity and satisfaction with their own smile (r ~ .44, p -

.006).15 Participants in the "Low" Smile Intensity group, however, showed a positive but

nonsignificant relationship with that particular variable (r = .25,/? = .17). This unexpected

and intriguing finding indicates that the degree to which the "High" group was smiling in

their photos was strongly related to how satisfied they happened to be with their smile.

Because Smile Intensity was also positively and significantly correlated with negative affect

46

(i.e., Felt Misunderstanding and NA) for those in the "High" Smile Intensity group, this

appears to suggest that the "meaning" of their smiling behavior may have been quite different

than we would predict (in general, and based on the results of previous studies).

Looking for additional clues as to what—if anything—"observable" might be driving

these results, we retrieved (from our archives) the most current Facebook profile photographs

for all of the participants. For many, we had photos that had been archived just before or just

after they participated in the follow-up study (April, 2010). We sorted the participants by

group (High vs. Low), and we simply "looked" at the photographs. Although purely

speculation at this point, it appeared that a substantial number of the participants who seemed

to be "driving" these effects (i.e., those in the High group) were "big" social smilers in their

photographs—and it was the type of smile that should (technically) be coded as high (or

relatively high) for mouth and eye activation.

Discussion

It is difficult to know what, if any, conclusions can be drawn from these data.

Are the results erroneous? Might they have merit—and even be pointing to unique

relationships between Smile Intensity and individual differences that we were unable to

identify? Might our high smile satisfaction participants be members of that segment of

the population who happen to be good at feigning Duchenne smiles (e.g., Krumhuber &

Manstead, 2009)—in general or/and in order to camouflage (or perhaps to help

themselves to cope with) negative emotions?

At this time, we cannot claim to know answers to any of the above questions. It

does seem worth noting, however, that if we had been content to analyze these data using

the simple correlational methods that have been employed in much of the prior published

47

research in this area of inquiry, we would (most likely) have been moved to consider the

initial results of this study to be at least a moderate success. Why? Because our Smile

Intensity score (measured at Time 2) was positively and significantly correlated with

Satisfaction With Life Scale scores at Time 3 (r ~ . 27, p = . 029). Thus, by some

established standards Smile Intensity did indeed "predict" future happiness for these

participants.

It was because we have made use of multiple measures of life satisfaction (ones

that captured the positive and the negative), sophisticated data analysis techniques, and

insisted upon replication (in general, and as a point of comparison for results across

studies) that we were able to avoid a considerable error—of the gravest type within

psychological research.

General Discussion

In many respects, the current program of research began as an attempt to conduct

a (conceptual) replication and extension of the work of Harker and Keltner (2001). Their

longitudinal study—which tracked women for three decades post-college and assessed

subjective well-being once per decade—continues to be a compelling example of the

potential for momentary displays of positive affect (i.e., those which were conveyed by

the college yearbook photos coded by Harker and Keltner) to predict long-term life

satisfaction and relationship satisfaction. In spite of the passing of nearly a decade since

its publication, however, this highly-respected study had not been replicated (even in

shorter-term work). Given the strong links shown in the research literature between

happiness and long-term health, success, and relationship satisfaction (for a

comprehensive review, see Lyubomirsky, King, & Diener, 2005), it seemed ever-more

important to revisit this work—and to attempt to replicate the basic findings (i.e., that the

expression of positive affect in one context will predict future well-being and

relationship satisfaction).

One major weakness of the work by Harker and Keltner (2001) was that the data

analytic techniques used in the study—simple correlations between smile codings and

self-reported future well-being—left most of the "process" questions unaddressed. As

the result, little explanation was given for why smile intensity was associated with future

well-being. Harker and Keltner (2001) did acknowledge that the associations they found

between smile intensity at age 21 and long-term self-reported well-being could have

been fully or partially attributable to a variety of processes including personality and

relationship factors. However, further research was required in order to begin to address

such issues.

Our interest in attempting to replicate and extend this work reached a peak in late

2004—just after the website Facebook was launched. We recognized that the

photographs (and other behavioral data) being made public by college students on the

website might make for a rich environment in which to conduct such research. It helped

that we (e.g., Seder & Oishi, 2009) were already in the process of making use of

Facebook (including photos posted to Facebook) for research related to well-being.

Thus we were in the right place at the right time, and began the current research in the

fall of 2005.

Although it has taken five years to bring this project to fruition, the findings from

the first two cohorts were indeed quite promising. We did conceptually replicate the

basic findings of Harker and Keltner (2001) in two separate cohorts. In addition, were

able to extend their preliminary work in a host of ways.

In Study 1 and Study 2, for example, we showed that display of positive affect

coded from participants' main Facebook profile photos during their first semester at

college was a robust predictor of (two types of) self-reported life satisfaction 3.5 years

later (as the students were about to graduate from college). This finding held even when

self-reported well-being (from the first semester) was included in the model. Thus, we

were able to show that the "behavior" instantiated in participants' display of positive

affect during the first semester was not simply correlated with long-term well-being—it

was a unique predictor of long-term well-being. We also showed that the predictive

utility of Smile Intensity coded from the photographs held even when controlling for the

effects of personality (as captured by our Big 5 measure). This finding was informative

in that smile intensity could be an indicator of personality processes (e.g., extraversion),

and thus a likely mechanism for the link with subjective well-being. However, our

results—which were replicated in two separate cohorts—showed that this was not the

case. While intrigued by these findings, we do acknowledge that there are numerous

ways to measure personality—and we utilized only one method of doing so in the current

studies. Thus, future research should incorporate other (and perhaps more in-depth)

measures of personality.

An additional contribution made by the current research was that we were able to

address the issue of mediation with enough power (by collapsing the data from Study 1

and Study 2). These results showed that relationship satisfaction partially mediated the

relationship between Smile Intensity and participants' long-term life-satisfaction (as

50

measured by both the SWLS and the Felt Understanding scale). This finding represents

the first indication of a causal link between Smile Intensity and long-term subjective

well-being. However compelling this might be, the finding is also ambiguous in that the

models showed evidence of partial mediation. Thus, there is great need follow up on

these initial findings by addressing specific social/relationship processes in future

research.

In order to begin to address specific processes, however, it seems essential for

future research to first ascertain the extent to which people's affective displays in their

photographs are likely to be typical of their "real world" display behaviors. A robust

relationship between the two has been predicted (e.g., Keltner, 2004), but it has yet to be

proven. Future research could utilize informant reports (from friends and family). It

also seems likely that in-lab interaction studies could be a promising way to test these

relationships. For example, two strangers could engage in a series of interaction tasks

and then look at various photos from each others' Facebook pages in order to rate the

similarity between "in-person" affective displays and "online" displays. These

interactions could be videotaped so that multiple observers (of the interactions and of the

photographs) could also rate the correspondence (or lackthereof) between the

participants' in-person displays and their in-photo displays.

If research shows that people's affective displays in their photographs do predict

their real-world display behaviors, several relationship-oriented mechanisms are then

likely to be important to consider. First, displays of positive affect may act as a "signal"

to potential (or actual) interaction partners that people are good-natured and

approachable. Signaling of this sort could attract like (i.e., positive) others, and thus

51

could be associated with the potential for increased relationship satisfaction in the short-

term and in the long-term (e.g., Fridlund, 1994). This hypothesis could be tested using

the interaction tasks just detailed.

"Signaling" of this sort is typically considered only within the context of "real

world" interactions. However, it seems plausible to us that technology has expanded the

potential "reach" and significance of this particular variable. Rather than simply being a

proxy for "real world" behaviors, for example, affective "signaling" via one's Facebook

photographs could also have the potential to influence real-world interactions and

relationships. How might this happen? Todays's college students (one of the many

groups that use Facebook) spend an inordinate amount of time consulting the Facebook

pages of those around them. It would hardly be surprising to find that people may look

to Facebook photos as a way to help them to interpret ambiguous affective "signals"

conveyed by individuals during real-time interactions. Given that many people have

near-constant access to the internet and to Facebook, this influence could be pervasive

(especially within certain demographic groups, and at certain times—such as when

people are adapting to a new environment or social network). Further, Facebook pages

are increasingly dynamic (i.e., users update photos and content frequently). Thus, using

web-posted photos in order to look for clues about in-person behavior could be adaptive

(e.g., "She usually looks very happy, but in the past week she has seemed dour—in

person and in her 'Facebook photos.' Something must be wrong."). In short, there may

be numerous ways—"traditional" and "new"—in which signaling via the expression of

positive affect could be associated with relationship satisfaction and long-term well-

being. This is likely to be a rich topic for future research.

A second possible mechanism—presumably related to the first—is that displays

of positive affect could have the potential to elicit similar responses in others. This could

take the form of smiling, or experienced positive affect, or both. Much research has

shown that shared positive affect can have a positive influence on relationships (e.g.

Gottman, et al , 2001). In order to address this as a specific mechanism within the

current program of research, future research could utilize in-lab interactions similar to

those described above but with more of an emphasis on experienced affect. Experience

sampling methodology could also prove useful for tracking this in the context of real-

world interactions. In addition, a longitudinal study which tracks affective processes and

relationships among first year college students who are randomly assigned to dorm

rooms or suites could be a highly informative way to investigate such processes.

On a related note, it also seems plausible that the "fit" between a person's

affective display "style" and the display norms of those around them could have

implications for relationship satisfaction and for overall well-being. For example, how

might a person who scores high on our "smile intensity" scale be likely to respond to

living for an extended period of time amongst people who all score on the low end of the

scale? Could lack of "fit" in such a scenario be likely to be associated with lowered

well-being and a greater sense of felt misunderstanding on the part of the "big smile"

individual? Could it lead to the potential for conformity to the group norm as a way to

"fit in"—and/or could it lead to the potential for the "smiley guy" to be perceived to be

an outsider from day one? Further, might such a "display culture" clash be likely to

operate (largely) outside of conscious awareness? Each of these scenarios seems

plausible to us, and the notion that these particular affective display norms may have the

53

potential to influence well-being at the group level seems worthy of future research.

Indeed, we may even be able to address this in a preliminary way by comparing the

affective display codings of our participants with the affective displays of the people

they listed as their closest friends and acquaintances. One hypothesis would be that lack

of fit would be negatively associated with relationship satisfaction and overall well-

being.

In Study 3, we began the challenge of addressing the question of whether use of

multiple photos (containing multiple, code-able affective expressions) will have the

potential to increase the predictive utility of coding for Smile Intensity. Multiple photos

will enable us to consider both the mean level of positive affective displays as well as the

degree with which people vary across display contexts. It is conceivable that variation

could impact well-being along with or separate from mean-level smile intensity. The

results of Study 3 were inconclusive. We will, however, continue to address this

important issue in future research.

In addition to the need to explore the various mechanisms likely to connect smile

intensity with long-term well-being, future research must address and rectify certain

methodological challenges. One clear need is for increased examination of the

relationships between the existing schemes used to code affective displays in

photographs (e.g., Freese, Meland, & Irwin, 2007; Harker & Keltner, 2001; Tsai, Louie,

Chen, & Uchida, 2007) in order to understand how well they measure the object of

interest, and whether it is possible to find ways to "translate" scores between the scales.

In our Study 2, for example, we originally intended to code multiple Facebook album

photographs and to compare these with Smile Intensity coded from only the profile

54

photos. (Album photos are more plentiful on Facebook, and the content of album photos

is typically more varied than those used as "main profile photos"). Coding was

problematic, however, because the pictorial resolution on the album photos was of

substantially lower quality than the resolution on the profile photos. Thus, after a

considerable investment of time and effort, we were forced to concede that we could not

use the Harker and Keltner scale (which codes both mouth and eyes on a precise scale

from 1-5) in order to code those photographs. However, we could (presumably) have

used the scale developed by Freese, et al. (2007), because it requires less precision (i.e., a

1-3 rating) in order to represent relative smile intensity. We were hesitant to invest the

time and resources to do so, however, because we had little idea of how such scores

would be likely to relate to our current coding method. Additionally, we have in the past

made use of the coding scheme developed by Tsai and colleagues (2007) that employs a

1-4 scale for mouth and for eyes, and it produced very interesting results. However,

because we were attempting to replicate and extend the work of Harker and Keltner

(2001), we opted to use only their scale. Given that photos have different resolutions

and researchers have different research agendas (i.e., a relative measure of smile

intensity may be adequate within some research paradigms), coordinating these scales—

and developing new methods of extracting useful information—can only improve the

potential for more and better research. Further, this will also be likely to increase the

ability to seek and find commonalities (and differences) across research findings.

Conclusions

In many respects, the current research represents a sizeable step forward in the

quest to understand how and under what conditions the expression of positive affect

55

captured in photographs at one point in time can predict future well-being and

relationship satisfaction. Although we have indeed "replicated" and extended the work

of Harker and Keltner (2001) in two studies, this overall program of research is only in

the early stages of development. Many questions remain. In addition, many promising

uses may soon become apparent for this paradigm. Thus far, for example, we have only

focused on methods to predict life satisfaction and relationship satisfaction—anchoring

on the happiness component rather than on lack of happiness. It seems plausible,

however, that this research could be adapted for use in programs to predict (and maybe

even to prevent) likelihood of future depression, anxiety, or other psychological

disorders in adults and in children. Could such outcomes be predicted from photographs

or from examination of short video clips of teacher-student or parent-child interactions?

Indeed, consideration of the utility of such a program of research within, for example,

school systems (e.g., using affective displays in class photos as an indicator of risk) may

be one of many promising areas for future research.

56

References

Abel, E. L., & Kruger, M. L. (2010). Smile intensity in photographs predicts longevity.

Psychological Science, doi: 10.1177/0956797610363775

Ambady, N., Hallahan, M., & Conner, B. (1999). Accuracy of judgments of sexual orientation

from thin slices of behavior. Journal of Personality & Social Psychology, 77(3), 538-

547.

Ambady, N., Krabbenhoft, M. A., & Hogan, D. (2006). The 30-sec sale: Using thin-slice

judgments to evaluate sales effectiveness. Journal of Consumer Psychology, 7(5(1), 4-13.

Ambady, N., & Rosenthal, R. (1992). Thin slices of expressive behavior as predictors of

interpersonal consequences: A meta-analysis. Psychological Bulletin, 111(2), 256-274.

Ambady, N., & Rosenthal, R. (1993). Half a minute: Predicting teacher evaluations from thin

slices of nonverbal behavior and physical attractiveness. Journal of Personality and

Social Psychology, 64(3), 431-441.

American College Testing Program. (2007). Unpublished tabulations, derived from statistics

collected by the Census Bureau, 1960 through 1969. U.S. Department of Commerce,

Census Bureau. Current Population Survey (CPS), October, 1970 through 2006.

Retreived from http://nces.ed.gov/programs/digest/d07/tables/dt07_191 .asp

Anusic, L, Schimmack, U., Pinlcus, R. T., & Lockwood, P. (2009). The nature and structure of

correlations among Big Five ratings: The halo-alpha-beta model. Journal of Personality

and Social Psychology, 97(6), 1142-1156.

Babad, E., Avni-Babad, D., & Rosenthal, R. (2004). Prediction of students' evaluations from

brief instances of professors' nonverbal behavior in defined instructional situations.

Social Psychology of Education, 7(1), 3-33.

Back, M. D., Schmukle, S. C , & Egloff, B. (2010). Why are narcissists so charming at first

sight? Decoding the narcissism-popularity link at zero acquaintance. Journal of

Personality and Social Psychology, 98(1), 132-145.

57

Back, M. D., Stopfer, J. M., Vazire, S., Gaddis, S., Schmulde, S. C , Egloff, B., & Gosling, S. D.

(2010). Facebook profiles reflect actual personality not self-idealization. Psychological

Science, 21, 372-374.

Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social

psychological research: Conceptual, strategic, and statistical considerations. Journal of

Personality and Social Psychology, 51(6), 1173-1182.

Baumeister, R. F., & Bushman, B. J. (2007). Social psychology and human nature (1st ed.).

Belmont, CA US: Thomson Wadsworth.

Boehm, J. K., & Lyubomirsky, S. (2008). Does happiness promote career success? Journal of

Career Assessment, 16(1), 101-116.

Borkenau, P., Brecke, S., Mattig, C, & Paelecke, M. (2009). Extraversion is accurately

perceived after a 50-ms exposure to a face. Journal of Research in Personality, 43(4),

703-706.

Borkenau, P., Mauer, N., Riemami, R., Spinath, F. M., 8c Angleitner, A. (2004). Thin slices of

behavior as cues of personality and intelligence. Journal of Personality and Social

Psychology, 86(4), 599-614.

Brissette, L, Scheier, M. F,, & Carver, C. S. (2002). The role of optimism in social network

development, coping, and psychological adjustment during a life transition. Journal of

Personality and Social Psychology, 82( 1), 102-111.

Brody, N., & Ehrlichman, H. (1997). Personality psychology: The science of individuality.

Upper Saddle River, NJ: Prentice Hall.

Buffardi, L. E., 8c Campbell, W. K. (2008). Narcissism and social networking web sites.

Personality and Social Psychology Bulletin, 34,1303-1314.

Carver, C. S., Pozo, C, Harris, S. D., Noriega, V., Scheier, M., Robinson, D., 8c et al. (1993).

How coping mediates the effect of optimism on distress: A study of women with early

stage breast cancer. Journal of Personality and Social Psychology, 65, 375-390.

58

Cohn, J. F., Ambadar, Z., & Ekman, P. (2007). Observer-based measurement of facial

expression with the Facial Action Coding System. In Handbook of emotion elicilation

and assessment., Series in affective science (pp. 203-221). New York, NY US: Oxford

University Press.

Cote, S., & Morgan, L. M. (2002). A longitudinal analysis of the association between emotion

regulation, job satisfaction, and intentions to quit. Journal of Organizational Behavior,

23($), 947-962.

Curhan, J. R., & Pentland, A. (2007). Thin slices of negotiation: Predicting outcomes from

conversational dynamics within the first 5 minutes. Journal of Applied Psychology,

92(3), 802-811.

Davidson, K. W., Mostofsky, E., & Whang, W. (2010). Don't worry, be happy: Positive affect

and reduced 10-year incident coronary heart disease: The Canadian Nova Scotia Health

Survey. European Heart Journal, doi: 10.1093/eurheartj/ehp603

Diener, £., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The satisfaction with life scale.

Journal of Personality Assessment, 49(1), 71-75.

Diener, E., Nickerson, C , Lucas, R. E., & Sandvik, E. (2002). Dispositional affect and job

outcomes. Social Indicators Research, 59(3), 229-259.

Diener, E., Oishi, S., & Lucas, R. E. (2003). Personality, culture, and subjective well-being:

Emotional and cognitive evaluations of life. Annual Review of Psychology, 54, 403-425.

Diener, E., 8c Seligman, M. E. P. (2002). Very happy people. Psychological Science, 75,81-

84.

Dunnett, C. W. (1980). Pairwise multiple comparisons in the unequal variance case. Journal of

the American Statistical Association, 75, 796-800.

Egloff, B., Schmukle, S. C, Burns, L. R., Kohlmann, C, & Hock, M. (2003). Facets of

dynamic positive affect: Differentiating joy, interest, and activation in the Positive and

Negative Affect Schedule (PANAS). Journal of Personality and Social Psychology,

85(3), 528-540.

Eckman, P., 8c Friesen, W. V. (1976). Measuring facial movement. Journal of Environmental

Psychology and Nonverbal Behavior, 1, 56-75.

Eckman, P., & Friesen, W. V. (1979). Facial Action Coding System: A technique for the

measurement of facial movement. Palo Alto, CA: Consulting Psychologists Press.

Ekman, P., & O'Sullivan, M. (1991). Who can catch a liar? American Psychologist, 46(9), 913-

920.

Ellison, N. B., Steinfield, C, & Lampe, C. (2007). The benefits of Facebook'friends:'Social

capital and college students' use of online social network sites. Journal of Computer-

Mediated Communication, 12(4), 1143-1168.

Eschleman, K. J., & Bowling, N. A. (2010). Facing the limitations to self-reported well-being:

Integrating the facial expression and well-being literatures. In New developments in

theoretical and conceptual approaches to job stress, Research in occupational stress and

well-being (Vols. 1-1, Vol. 8, pp. 199-235). United Kingdom: Emerald Group

Publishing Limited.

Facebook.com Press Room Statistics. (2010). Retrieved from

http://www.facebook.com/press/info.php7statistics

Fowler, K. A., Lilienfeld, S. O., & Patrick, C. J. (2009). Detecting psychopathy from thin

slices of behavior. Psychological Assessment, 21(1), 68-78.

Fredrickson, B. L. (1998). What good are positive emotions? Review of General Psychology, 2,

300-319.

Fredrickson, B. L. (2000). Extracting meaning from past affective experiences: The importance

of peaks, ends, and specific emotions. Cognition and Emotion, 14, 577-606.

Fredrickson, B. L. (2001). The role of positive emotions in positive psychology: The broaden-

and-build theory of positive emotions. American Psychologist, 56, 218-226.

Fredrickson, B. L., & Levenson, R. W. (1998). Positive emotions speed recovery from the

cardiovascular sequelae of negative emotions. Cognition and Emotion, 12,191-220.

Fredrickson, B. L., Mancuso, R. A., Branagan, C , & Tugade, M. M. (2000). The undoing

effect of positive emotions. Motivation and Emotion, 24, 237-258.

Freese, J., Meland, S., & Irwin, W. (2007). Expressions of positive emotion in photographs,

personality, and later-life marital and health outcomes. Journal of Research in

Personality, 41, 488-497.

Fridlund, A. (1994). Human facial expression: An evolutionary view. New York: Academic

Press.

Frisch, M. B., Clark, M. P., Rouse, S. V., Rudd, M. D., Paweleck, J. K., Greenstone, A., &

Kopplin, D. A. (2005). Predictive and treatment validity of life satisfaction and the

Quality of Life Inventory. Assessment, 12(1), 66-78.

Gable, S. L., & Haidt, J. (2005). What (and why) is Positive Psychology? Review of General

Psychology, Positive Psychology, 9(2), 103-110.

Gasper, K., & Clore, G. L. (2002). Attending to the big picture: Mood and global versus local

processing of visual information. Psychological Science, 13(1), 34-40.

Gosling, S. D., Rentfrow, P. J., 8c Swann, W. B. (2003). A very brief measure of the Big Five

personality domains. Journal of Research in Personality, 37, 504-528.

Gosling, S. D., Ko, S. J., Mannarelli, T., & Morris, M. E. (2002). A room with a cue:

Personality judgments based on offices and bedrooms. Journal of Personality and Social

Psychology, 82, 379-398.

Gottman, J., Levenson, R., Woodin, E. (2001). Facial expressions during marital conflict.

Journal of Family Communication, 7(1), 37-57.

Grasmuck, S., Martin, J., 8c Zhao, S. (2009). Ethno-racial identity displays on Facebook.

Journal of Computer-Mediated Communication, 75(1), 158-188.

Haidt, J. (2006). The happiness hypothesis: Finding modern truth in ancient wisdom. New

York, NY US: Basic Books.

61

Harker, L., 8c Keltner, D. (2001). Expressions of positive emotions in women's college

yearbook pictures and their relationship to life outcomes across childhood. Journal of

Personality and Social Psychology, 80, 112-124.

Hertenstein, M. J., Hansel, C. A., Butts, A. M.,&Hile, S. N. (2009). Smile intensity in

photographs predicts divorce later in life. Motivation and Emotion, 33(2), 99-105.

Isen, A. M., & Daubman, K. A. (1984). The influence of affect on categorization. Journal of

Personality and Social Psychology, 47(6), 1206-1217.

Isen, A. M., Daubman, K. A., & Nowicki, G. P. (1987). Positive affect facilitates creative

problem solving. Journal of Personality and Social Psychology, 52(6), 1122-1131.

Isen, A. M., 8c Levin, P. F. (1972). Effect of feeling good on helping: Cookies and kindness.

Journal of Personality and Social Psychology, 21(3), 384-388.

Kalat, J. W. (2007). Introduction to Psychology (8th ed.). Belmont, CA: Wadsworth

Publishing.

Keltner, D. (2003). Expression and the course of life: Studies of emotion, personality, and

psychopathology from a social-functional perspective. In Emotions inside out: 130 years

after Darwin's: The expression of the emotions in man and animals., Annals of the New

York of Sciences; Vol. 1000 (pp. 222-243). New York, NY US: New York University

Press.

Keltner, D., (2004). Ekman, emotional expression, and the art of empirical epiphany. Journal of

Research in Personality, 38, 37-44.

Keltner, D., Anderson, C , & Gonzaga, G. C. (2002). Culture, emotion, and the good life in the

study of affect and judgment. Psychological Inquiry, 13(1), 65-67.

Keltner, D., Ekman, P., Gonzaga, G. C, &Beer, J. (2003). Facial expression of emotion. In

Handbook of affective sciences., Series in affective science (pp. 415-432). New York,

NY US: Oxford University Press.

62

Kim-Prieto, C, Diener, E., Tamir, M., Scollon, C , & Diener, M. (2005). Integrating the diverse

definitions of happiness: A time-sequential framework of subjective well-being. Journal

of Happiness Studies, 6(3), 261-300.

Kluemper, D. H., & Rosen, P. A. (2009). Future employment selection methods: Evaluating

social networking web sites. Journal of Managerial Psychology, 24(6), 567-580.

Kramer, A. D. I. (2010, March 23). How happy are we? Message posted to

http://blog.facebook.com/blog.php?post=150162112130

Kraus, M. W., & Keltner, D. (2009). Signs of socioeconomic status: A thin-slicing approach.

Psychological Science, 20(1), 99-106.

Krumhuber, E. G., & Manstead, A. S. (2009). Can Duchenne smiles be feigned? New

evidence on felt and false smiles. Emotion, 9(6), 807-820.

Kunzmann, U., Stange, A., 8c Jordan, J. (2005). Positive affectivity and lifestyle in adulthood:

Do you do what you feel? Personality and Social Psychology Bulletin, 31(4), 574-588.

LaFrance, M., Hecht, M. A., & Paluck, E. L. (2003). The contingent smile: A meta-analysis of

sex differences in smiling. Psychological Bulletin, 129(2), 305-334.

Lai, J. C. L., Evans, P. D.,Ng, S. FI.,Chong,A. M. L., Siu, O. T., Chan, C. L. W., Ho, S.

M. Y., et al. (2005). Optimism, positive affectivity, and salivary Cortisol. British

Journal of Health Psychology, 10(4), 467-484.

Lyubomirsky, S., King, L., 8c Diener, E. (2005). The benefits of frequent positive affect: Does

happiness lead to success? Psychological Bulletin, 131(6), 803-855.

Lyubomirsky, S., & Lepper, H. S. (1999). A measure of subjective happiness: Preliminary

reliability and construct validation. Social Indicators Research, 46, 137-155.

Lyubomirsky, S., Tkach, C , & DiMatteo, M. R. (2006). What are the differences between

happiness and self-esteem? Social Indicators Research, 78, 363-404.

Marcus, B., Machilek, F., 8c Schutz, A. (2006). Personality in cyberspace: Personal web sites as

media for personality expressions and impressions. Journal of Personality and Social

Psychology, 90, 1014-1031.

63

Matsumoto, D., Keltner, D., Shiota, M. N., O'Sullivan, M., & Frank, M. (2008). Facial

expressions of emotion. \n Handbook ofemotions (3rd ed). (pp. 211-234). New York,

NY US: Guilford Press.

McCray, R. R., & Costa, P. T. (1986). Personality, coping, and coping effectiveness in an

adult sample. Journal of Personality, 54, 385-405.

Messinger, D., & Fogel, A. (2007). The interactive development of social smiling. In Advances

in child development and behavior (Vol 35)., Advances in child development and

behavior; 0065-2407 (pp. 327-366). San Diego, CA US: Elsevier Academic Press.

Mikami,A. Y., Szwedo, D. E., Allen, J. P., Evans, M. A., & Hare, A. L. (2010). Adolescent

peer relationships and behavior problems predict young adults' communication on social

networking websites. Developmental Psychology, 46(1), 46-56.

Muise, A., Christofides, E., 8c Desmarais, S. (2009). More information than you ever wanted:

Does Facebook bring out the green-eyed monster of jealousy? Cyber Psychology &

Behavior, 12(4), 441-444.

Myers, D. G. (2008). Exploring psychology (7th ed). New York, NY US: Worth Publishers.

Naumann, L. P., Vazire, S., Rentfrow, P. J., 8c Gosling, S. D. (2009). Personality judgments

based on physical appearance. Personality and Social Psychology Bulletin, 55(12), 1661-

1671.

Oishi, S., 8c Koo, M. (2008). Two new questions about happiness: 'Is happiness good?' and 'Is

happier better?'. In The science of subjective well-being, (pp. 290-306). New York, NY

US: Guilford Press.

Oishi, S., Diener, E., Scollon, C. N., & Biswas-Diener, R. (2004). Cross-situational consistency

of affective experiences across cultures. Journal of Personality and Social Psychology,

86(3), 460-472.

Oishi, S., Krochik, M., 8c Akimoto, S. (2010). Felt understanding as a bridge between close

relationships and subjective well-being: Antecedents and consequences across

individuals and cultures. Social and Personality Psychology Compass, 4(6), 403-416.

Orr, E. S., Sisic, M., Ross, C , Simmering, M. G., Arseneault, J. M., & Orr, R. R. (2009). The

influence of shyness on the use of Facebook in an undergraduate sample.

Cyber Psychology & Behavior, 12(3), 337-340.

Oveis, C, Gruber, J., Keltner, D., Stamper, J. L.,&Boyce, W. T. (2009). Smile intensity and

warm touch as thin slices of child and family affective style. Emotion, 9(4), 544-548.

Papa, A., & Bonanno, G. A. (2008). Smiling in the face of adversity: The interpersonal and

intrapersonal functions of smiling. Emotion, 8(1), 1-12.

Pelted, L. H., & Xin, K. R. (1999). Down and out: An investigation of the relationship

between mood and employee withdrawal behavior. Journal of Management, 25, 875-

895.

Peluchette, J., 8c Karl, K. (2010). Examining students' intended image on Facebook: 'What

were they thinking?!'. Journal of Education for Business, 85(1), 30-37.

Pempek, T. A., Yermolayeva, Y., 8c Calvert, S. L. (2009). College students' social networking

experiences on Facebook. Journal of Applied Developmental Psychology, 30(3), 227-

238.

Penley, J. A., 8c Tomaka, J. (2002). Associations among the Big Five, emotional responses,

and coping with acute stress. Personality and Individual Differences, 32, 1215-1228.

Preacher, K. J., & Hayes, A. (2004). SPSS and SAS procedures for estimating indirect effects

in simple mediation models. Behavior Research Methods, Instruments, & Computers,

36(4), 717-731.

Proffitt, D. R. (2006). Embodied perception and the economy of action. Perspectives on

Psychological Science, 1(2), 110-122.

Ross, C, Orr, E. S., Sisic, M., Arseneault, J. M., Simmering, M. G., & Orr, R. R. (2009).

Personality and motivations associated with Facebook use. Computers in Human

Behavior, 25(2), 578-586.

65

Schnall, S., Harber, K. D., Stefanucci, J. K., & Proffitt, D. R. (2008). Social support and the

perception of geographical slant. Journal of Experimental Social Psychology, 44(5),

1246-1255.

Scollon, C. N., 8c Diener, E. (2006). Love, work, and changes in extraversion and neuroticism

over time. Journal of Personality and Social Psychology, 91, 1152-1165.

Seder, J. P., & Oishi, S. (2009). Ethnic/racial homogeneity in college students' Facebook

friendship networks and subjective well-being. Journal of Research in Personality,

43(3), 438-443.

Sheldon, P. (2008). The relationship between unwillingness-to-communicate and students'

Facebook use. Journal of Media Psychology: Theories, Methods, and Applications,

20(2), 67-75.

Staw, B. M., Sutton, R. I., 8c Felled, L. H. (1994). Employee positive emotion and favorable

outcomes in the workplace. Organization Science, 5( 1), 51 -71.

Steinfield, C, Ellison, N. B., & Lampe, C. (2008). Social capital, self-esteem, and use of online

social network sites: A longitudinal analysis. Journal of Applied Developmental

Psychology, 29(6), 434-445.

Strathearn, L., Li, J., Fonagy, P., & Montague, P. R. (2008). What's in a smile? Maternal brain

responses to infant facial cues. Pediatrics, 722,40-51. doi: 10.1542/peds.2007-1566

Tong,S. T.,Van,D. H., Langwell, L., & Walther, J. B. (2008). Too much of a good thing?

The relationship between number of friends and interpersonal impressions on Facebook.

Journal of Computer-Mediated Communication, 13(3), 531-549.

Todorov, A., Mandisodza, A. N., Goren, A., 8c Hall, C. C. (2005). Inferences of competence

from faces predict election outcomes. Science, 308,1623-1626.

Tsai, J. L., Louie, J. Y., Chen, E. E., & Uchida, Y. (2007). Learning what feelings to desire:

Socialization of ideal affect through children's storybooks. Personality and Social

Psychology Bulletin, 33, 17-30.

Tufekci, Z. (2008). Grooming, gossip, Facebook and Myspace: What can we learn about these

sites from those who won't assimilate? Information, Communication & Society, 11(4),

544-564.

Valenzuela, S., Park, N., & Kee, K. F. (2009). Is there social capital in asocial network site?:

Facebook use and college students' life satisfaction, trust, and participation. Journal of

Computer-Mediated Communication, 14(4), 875-901.

Vazire, S., 8c Gosling, S. D. (2004). e-Perceptions: Personality impressions based on personal

websites. Journal of Personality and Social Psychology, 87(1), 123-132.

Veenhoven, R. (2008). Healthy happiness: Effects of happiness on physical health and the

consequences for preventive health care. Journal of Happiness Studies, 9(3), 449-469.

Walther,J. B., Van, D. H., Kim, S., Westerman, D., &Tong, S. T. (2008). The role of

friends' appearance and behavior on evaluations of individuals on Facebook: Are we

known by the company we keep? Human Communication Research, 34(1), 28-49.

Wang,S. S., Moon, S., Kwon, K. H., Evans, C. A., & Stefanone, M. A. (2010). Face off:

Implications of visual cues on initiating friendship on Facebook. Computers in Human

Behavior, 26(2), 226-234.

Watson, D., Clark, L. A., Mclntyre, C. W., & Hamaker, S. (1992). Affect, personality, and

social activity. Journal of Personality and Social Psychology, 63, 1011-1025.

Watson, D., Clark, L. A., 8c Tellegen, A. (1988). Development anad validation of brief

measures of positive and negative affect: The PAN AS scale. Journal of Personality and

Social Psychology, 54, 1063-1070

Weisbuch, M., Ivcevic, Z., & Ambady, N. (2009). On being liked on the web and in the 'real

world': Consistency in first impressions across personal webpages and spontaneous

behavior. Journal of Experimental Social Psychology, 45, 573-576.

Wilson, T. D. (2002). Strangers to ourselves: Discovering the adaptive unconscious.

Cambridge, MA: Harvard University Press.

67

Wright, T. A., 8c Cropanzano, R. (1998). Emotional exhaustion as a predictor of job

performance and voluntary turnover. Journal of Applied Psychology, 83, 486-493.

Wright, T. A., & Staw, B. M. (1999). Affect and favorable work outcomes: Two longitudinal

tests of the happy-productive worker thesis. Journal of Organizational Behavior, 20, 1-

23.

Young, S., Dutta, D., & Dommety, G. (2009). Extrapolating psychological insights from

Facebook profiles: A study of religion and relationship status. CyberPsychology &

Behavior, 12(3), 347-350.

Zaki, J., Bolger, N., & Ochsner, K. (2009). Unpacking the informational bases of empathic

accuracy. Emotion, 9(4), 478-487.

Zywica, J., 8c Danowski, J. (2008). The faces of Facebookers: Investigating social enhancement

and social compensation hypotheses; predicting Facebook and offline popularity from

sociability and self-esteem, and mapping the meanings of popularity with semantic

networks. Journal of Computer-Mediated Communication, 14(\), 1-34.

68

Table 1

Descriptive statistics for key self report variables

Variable Time 1 Time 2 Time 3

M (SD) n M (SD) n M (SD) n

SWLS

Felt Understanding

Felt Misunderstanding

Positive Affect

Negative Affect

Need to Belong

Extraversion \ Neuroticism

Agreeableness

Conscientiousness

Openness

Facebook Updates (Freq)

Facebook Content (Amt)

5.00(1.12) 41

5.03(1.28) 42

3.15(1.35) 42

3.60(0.62)42

3.73(0.85)43

3.04(0.99)43

4.12(0.60)43

3.76(0.68)43

3.79 (0.48) 43

2.10(0.63)48

2.96(0.68)48

5.21 (1.12)48

4.92(1.27)48

2.78(1.31)48

5.19(1.08)48

3.32(0.98)48

5.24(1.24) 48

5.04(1.29)48

2.56(1.33)48

Table 2

Correlations Between Smile Intensity in Facebook Profile Photos and Self-Reported

Well-Being at Three Times During College

Variable Time 1 Time 2 Time 3

(n = 43) (n - 48) (n = 48)

SWLS .27+

Felt Understanding .29+

Felt Misunderstanding -.27+

Positive Affect (PA)

Negative Affect (NA)

.36*

.29+

-.12

.28+

-.19

.46**

.40**

-.21

+p<A0. *p<.05. **/><.01.

70

Table 3

Correlations Between Smile Intensity in Facebook Profile Photos, Subjective Weil-Being,

and Self-Reported Satisfaction With Social Relationships, Study 1

1 2 3 4 5 6 7 8 9 10 11

1. Smile Intensity

2. SWLS

(T3)

3. SWLS (T2)

10. Donate as Alumnus

11. Attend Reunions

.45"

.38" .57"

„ * * _ , * * 4SWLS(T3) .28 .77 .74

LXot, * ••»• - •»» •• F„:rs •« •» •» •» •"" -7. Friendship .38" .36* .44" .20 .40** .14 -Sat.(T3)

8 Sat. w. Social Life .38" .25 .32* .24 .16 .07 .51** --(T3)

9. "Fit In" .27 .48" .43" .33* .27 .16 .54" .25

.39" .48" .64" .55" .19 .29* .42" .39" .40*

.30* .34* .50" .46".37* .45* .33* .25 .37** .67**

+p<.10. */><.05. **p<M.

71

Table 4

Correlations between Smile Intensity and Big 5 Personality Variables.

5 6 7 Variables

1. Smile Intensity

2. Extrav

3. Agree

4. Consc

5. Neurot

1

.21

-.09

.36*

-.31*

2

-

.50"

.26

-.27

3

-

.01

-.0'

6. Open -.11 .42" .57** .08 .57 * -

L ^ f -2^.78" .60" .50** .40" .14 HALO

+p<A0. V<.05 . **p<-01.

Table 5

Correlations Between Smile Intensity (mean) and main Life Satisfaction Variables,

Study 3 (n^ 68)

1 2 3 4 5 6 7 8 9

1. Smile Intensity —

2. SWLS(T3) .27* -

3. SWLS(T2) .33** .58" -

* _-,** 4 SWLS(Tl) .31 .49 .78

5. Underst. -.08 .49" .25* .22+

6. Misund. -.12 - .51" -.40** -.36** -.42**

7. Happiness .04 .59*" .43" .42" .49** -.56**

8 PA .19 .32 .70 .67 .40 -.17 .40

9. NA .01 -.18 -.51 -.47* -.18 .28* -.34 -.56**

+/?<.10. */?<.05. **/?<.01.

73

Table 6

Correlations Between Smile Intensity in Facebook Profile Photos and Life Satisfaction

Variables for the "Low" Group, Study 3 (n~ 31)

1 2 3 4 5 6 7 8 9

1. Smile Intensity

2. SWLS (T3) .29

3. SWLS(T2) .50** .57** --

4. SWLS(Tl) .52" .60" .78" --

5. Underst. .01 .65** .31+ . 27

6. Misund. -.18 -.44** -.45* -.52** - .31"

7. Happiness .20 .65** .56** . 66"" .50** -.39

8 PA .38" .48** .79** .65" .51" -.36+ .59"

9. NA -.13 -.32+ -.65** -.51 -.25 .21 -.56 -.69 *#

+p<.10. *p<.05. **/?<.01.

74

Table 7

Correlations Between Smile Intensity in Facebook Profile Photos and Life Satisfaction

Variables for the "High" Group, Study 3 (n = 37)

1 2 3 4 5 6 7 8 9

1. Smile Intensity

2. SWLS(T3) -.03

3. SWLS(T2) -.05 .61"

4. SWLS(Tl) -.17 .40* .77"

5. Underst. -.10 .43" .22 .22

6. Misund. .42** -.53** -.30+ -.13 -.58s

7. Happiness -.30+ .56** .34** .21 .51** -.68**

8 PA -.07 .21 .62** .74** .32* -.02 .28+

9. NA .34* -.08 -.35* -.43" -.13 .33* -.19 -.45**

+p<.10. V < . 0 5 . * V < . 0 1 .

75

Harker and Keltner (2001) used a scale which ranged from 2-10 in total smile intensity. Freese and colleagues (2007) used a 3-point scale (e.g., l=no smile, 2=Standard smile, 3=Intense smile). 2 Depending, perhaps, on sample size and on the types of research questions involved. 3 All participants were recruited from the Participant Pool in the Department of Psychology at University of Virginia in Fail 2005. During this semester, a total of 1,161 members of the Pool were first year students who had an active Facebook account; thus, they were eligible to become participants in our study. Of those, 152 became participants. Because the design of the study required us to drop a substantial number of participants (n =110), we wished to determine whether the 48 participants used in the final dataset differed in any substantial ways on the self-report variables of interest in the study. In addition, we wished to know the extent to which our participants—those retained and those dropped—differed from the other eligible members of their cohort (n = 1009). A majority of Participant Pool members complete a mass testing session at the beginning of the semester. Included in the session were the variables we used for Time 1 in our study: Satisfaction With Life Scale, Felt Understanding and Felt Misunderstanding, Big 5, and the Need to Belong scale. In addition, there was one question that asked participants to rate the amount of content they had in their Facebook account (1-4), and how often they made updates to the content (1-4). A one-way between groups ANOVA was conducted to examine the impact of group on each of the variables of interest. Because our data contained unequal sample sizes, we used Dunnett's T3 multiple comparisons test (Dunnett, 1980). No significant differences were found between the three groups on any of the variables. Thus, our 48 participants appear to be typical of the other members of their Facebook-using cohort. 4 In five cases, we felt that the existing profile photo was not suitable for coding (e.g., the participant was wearing sunglasses and therefore we could not see their eyes). For each of these, we retrieved an alternate photo from the participant's Facebook account. Care was taken to choose an alternate photo that was similar in facial expression to the original. 5 This is a highly confidential document because participants did not give us permission use their data for presentations.

This is similar to results reported in other studies. 7 Results showed that 34% of the variance in the mean for Likelihood of Attending Reunions/Donating to the University (R2 = .34, p < .01) was explained by a model consisting of Smile Intensity (/? = .201, p = .152) and SWLS at Time 2 (fi = A92,p < .01). In addition, 33% of the variance for this variable was explained by Smile Intensity (/? - .240, p = .066) and Felt Understanding at Time 2(fi- .457, p < .01). 8 We expected that the number of Other-School Facebook Friends would be correlated with overall relationship satisfaction. However, we reasoned that a majority of such individuals would probably be friends from high school. Thus, we considered it likely to be less related to in-college relationship satisfaction. 9 A paired-samples t-test indicated that this was a marginally significant mean difference (/(46) = -1.79, p = .08, d = -.24). Thus, our first year students did have marginally fewer On-Campus Facebook Friends than Other-School Facebook Friends. 10 For example, the experience of cohorts of first-year college students at Tulane before vs. after Hurricane Katrina. 11 Additionally, all of the major findings reported in Study 1 and Study 2 held when the analyses were re­run using the combined dataset. 12 This had not been the case for Felt Understanding. 13 Note that the correlation between Smile Intensity and SWLS at Time 3 is .29 (p = . 119); thus the relationship is in the expected direction even if it is not a statistically significant relationship. . 14 Note that for the Low Smile Intensity group, Smile Intensity is "positively" related to the mean for future Reunions/Donations (r ~ . 115, p = 539). 15 That p-value is not a typo.