promoting participation in physical activity using framed...

Copyright © The British Psychological SocietyReproduction in any form (including the internet) is prohibited without prior permission from the Society

Promoting participation in physical activity usingframed messages: An application of prospecttheory

Amy E. Latimer1*, Tara A. Rench1, Susan E. Rivers1,Nicole A. Katulak1, Stephanie A. Materese2, Lisa Cadmus1,2,Althea Hicks3, Julie Keany Hodorowski3 and Peter Salovey1

1Department of Psychology, Yale University, New Haven, USA2Department of Epidemiology and Public Health, Yale University, New Haven, USA3National Cancer Institute’s Cancer Information Service, Memorial Sloan-KetteringCancer Center, New York, USA

Objectives. Messages designed to motivate participation in physical activity usuallyemphasize the benefits of physical activity (gain-framed) as well as the costs of inactivity(loss-framed). The framing implications of prospect theory suggest that the effectivenessof these messages could be enhanced by providing gain-framed information only. Wecompared the effectiveness of gain-, loss-, and mixed-framed messages for promotingmoderate to vigorous physical activity.

Design. Randomized trial.

Method. Sedentary, healthy callers to the US National Cancer Institute’s CancerInformation Service (N ¼ 322) received gain-, loss-, or mixed-framed messages on threeoccasions (baseline, Week 1, and Week 5). Social cognitive variables and self-reportedphysical activity were assessed at baseline, Week 2, and Week 9. Separate regressionanalyses were conducted to examine message effects at each assessment point.

Results. At Week 2, gain- and mixed-framed messages resulted in stronger intentionsand greater self-efficacy than loss-framed messages. At Week 9, gain-framed messagesresulted in greater physical activity participation than loss- or mixed-framed messages.Social cognitive variables at Week 2 did not mediate the Week 9 framing effects onphysical activity participation.

Conclusions. Using gain-framed messages exclusively may be a means of increasingthe efficacy of physical activity materials.

Most people are not active enough to attain the health benefits of physical activity

(World Health Organization, 2004). In light of growing public health concern,

* Correspondence should be addressed to Dr Amy Latimer, School of Kinesiology and Health Studies, Queen’s University, 69Union Street, Kingston, ON K7L 3N6 Canada (e-mail: [email protected]).

TheBritishPsychologicalSociety

659

British Journal of Health Psychology (2008), 13, 659–681

q 2008 The British Psychological Society

www.bpsjournals.co.uk

DOI:10.1348/135910707X246186


educational campaigns have been launched in an effort to persuade sedentary adults to

become more active (reviewed by Cavill & Bauman, 2004). The lack of behaviour

change following these campaigns has rendered unanswered the question of how to

communicate effectively physical activity recommendations (Randolph & Viswanath,

2004). Past public health initiatives aimed at increasing other health behaviours indicate

that framing messages appropriately in terms of gains or losses is necessary for achievingbehaviour change (Randolph & Vishwanath, 2004).

Prospect theory suggests that individuals respond differently to factually equivalent

information depending on whether it is framed in terms of costs (loss-framed) or

benefits (gain-framed; Tversky & Kahneman, 1981). Individuals tend to seek out risks

when considering the potential negative consequences (or losses) associated with a

choice, and tend to avoid risk when considering the benefits (or gains) associated with a

decision. Applying the tenets of prospect theory to the health promotion domain,

Rothman and Salovey (1997) proposed that loss-framed messages persuade individualsto engage in detection behaviours, which are inherently risky given their potential to

indicate the presence of disease. This prediction has been supported in research

encouraging a variety of detection behaviours including Pap testing (Rivers, Salovey,

Pizarro, Pizarro, & Schneider, 2005), HIV testing (Apanovitch, McCarthy, & Salovey,

2003), and mammography (Schneider et al., 2001). Conversely, researchers have

hypothesized that gain-framed messages convince individuals to engage in prevention

behaviours which are not inherently risky. This prediction has garnered empirical

support in studies encouraging behaviours such as sunscreen use (Detweiler,Bedell, Salovey, Pronin, & Rothman, 1999), smoking cessation (Steward, Schneider,

Pizarro, & Salovey, 2003), and dental hygiene (Rothman, Martino, Bedell, Detweiler, &

Salovey, 1999).

Physical activity is associated with reducing the risk of chronic disease and

physical impairment in generally healthy individuals (World Health Organization,

2004). Thus, using gain-framed messages should optimize the effectiveness of

campaigns promoting this prevention behaviour. However, research testing this

hypothesis has generated mixed findings (Jones, Sinclair, & Courneya, 2003; Jones,Sinclair, Rhodes, & Courneya, 2004; McCall & Martin Ginis, 2004; Robberson &

Rogers, 1988). For example, Robberson and Rogers found that among sedentary

female college students some gain-framed messages elicited stronger intentions to

engage in physical activity than loss-framed messages. Whether or not this effect

extended to behaviour is unclear; the follow-up survey was administered immediately

after message delivery and did not assess actual physical activity. Accounting for this

limitation, Jones and colleagues used a 2-week prospective design to examine the

effects of framed messages on participation in physical activity among undergradu-ates. Gain-framed messages delivered by a credible source resulted in greater physical

activity participation rates than loss-framed messages. However, Jones and colleagues

failed to replicate this finding perhaps because prior to receiving the intervention

almost half of the participants met the current physical activity recommendations.

Physical activity messages are most effective among sedentary individuals (Kroeze,

Werkman, & Brug, 2006; Vandelanotte, De Bourdeaudhuij, Sallis, Spittaels, & Brug,

2005). Thus, the active participants might have been less receptive to the messages,

reducing the likelihood of detecting message framing effects and accounting for theinconsistent behavioural results.

The inconsistent findings may also be attributable to the sample (college students),

short follow-up durations (2 weeks), or single message exposures. McCall and Martin

660 Amy E. Latimer et al.


(2004) examined message framing using a longitudinal design and prolonged message

exposure in a population for whom participation in physical activity is particularly

relevant – patients recovering from coronary artery bypass graft surgery who were

recommended to engage in cardiac rehabilitation. Patients who received gain-framed

messages (both in a pamphlet and on a magnet for their home refrigerator) attended

more cardiac rehabilitation sessions over a 3-month period than participants whoreceived loss-framed messages; however, this difference was not statistically significant.

Participants in the gain-framed group did attend significantly more sessions than

participants in a no-message control group. The inclusion of a control group

strengthened the study design by demonstrating the advantages of gain-framed

messages over standard practice.

Overview and hypothesesThe findings from the extant research suggest possible positive results from the use

of gain-framed messages in promoting physical activity participation; however, more

research is needed to account for the limitations of previous studies. Specifically,

there is a need to test framed physical activity messages in a receptive audience using

(a) multiple message exposures; (b) an extended follow-up period; and (c) a standard

message control group. The present experiment was designed to address these goals.The primary objective was to compare the effectiveness of gain-, loss-, and mixed-

framed messages for promoting moderate to vigorous physical activity over a 9-week

period following multiple message exposures. In accordance with the framing

postulates of prospect theory, we hypothesized that the effectiveness of physical

activity messages would be enhanced by providing gain-framed information only.

A secondary objective of the current study was to identify the social cognitive

mechanisms underlying the effectiveness of framed messages. Little research has been

conducted in this area. Understanding the factors mediating the link between themessage and the behaviour is necessary to advance message framing theory and practice

(Frazier, Tix, & Barron, 2004). Investigating social cognition as a potential mediator is a

good starting-point. Substantial evidence indicates that self-efficacy, outcome

expectancies, and intentions are important determinants of health behaviour and

specifically physical activity (Sallis & Owen, 1998). Furthermore, preliminary evidence

suggests that social cognitive variables (e.g. self-efficacy, outcome expectancies,

intentions) derived from self-efficacy theory (Bandura, 1997) may be potential mediators

of framing effects. For example, among some participants, the effects of gain-framedsmoking cessation messages on intentions to quit were mediated by self-efficacy (i.e. the

belief that one has the ability to perform the requisite behaviour; Steward et al., 2003).

Within the message framing and physical activity domain, mediational

relationships with social cognitive variables have not been established. However,

gain-framed messages have resulted in more positive outcome expectancies (beliefs

about whether or not a given behaviour produces a given outcome; Jones et al.,

2004; McCall & Martin, 2004) and stronger intentions to engage in physical activity

(Jones et al., 2003; Robberson & Rogers, 1988) than loss-framed messages. In thecurrent study, self-efficacy, outcome expectancies, and intentions were tested as

mediators of the effects of framed messages on physical activity behaviour.

Specifically, we predicted that gain-framed messages would result in greater physical

activity outcome expectancies, self-efficacy, and intentions, which in turn would lead

to greater participation in physical activity.

Physical activity and message framing 661


Method

ParticipantsFive hundred and seventeen sedentary (i.e. no more than 2 days/week of moderateto vigorous physical activity for greater than 20min/day; Rodgers & Gauvin, 1998)

men and women were recruited from the National Cancer Institute’s (NCI) Cancer

Information Service (CIS) in the United States. The CIS is the NCI’s link to the

information needs of the American public (i.e. cancer survivors and their family and

friends), interpreting and explaining research findings in a clear manner, and

providing personalized responses to specific questions about cancer (Squiers &

Treiman, 2005). Participants accessed the CIS by calling a toll-free number. Self-

reported exclusion criteria were: (a) being a cancer survivor; (b) awaiting testresults about a possible cancer diagnosis; (c) being less than 18 or more than 69

years of age; (d) calling the CIS for physical activity information; (e) participating in

the study already; (f) having a physical impairment or doctor’s recommendation

contraindicating unsupervised physical activity participation; (g) being a non-English

speaker or international caller; (h) exhibiting significant distress during the CIS

service call; and (i) engaging in greater than 20min/day of moderate to vigorous

physical activity on three or more days/week. From a research perspective, this

stringent criterion for physical activity participation was used because theindividuals who participate in physical activity two or fewer times per week have

less favourable physical activity-related social cognition (e.g. self-efficacy; outcome

expectancies) than individuals who exercise three or more times per week (Rodgers

& Gauvin, 1998). Given that the secondary objective of the study was to examine

the effects of framed messages on social cognition, a group with less favourable

expectations is a good target for affecting change. From a practical perspective,

these criteria were selected because they could be assessed quickly and easily. An

age restriction was imposed because the instrument used to assess physical activityhas only been validated among adults ranging in age from 18 to 69 years

(International Physical Activity Questionnaire, 2006).

ProcedureAt baseline, following their CIS service call, consenting callers completed the screening

interview. Eligible callers were invited to participate in a 9-week study examining their

opinions about new physical activity print messages. The eligible callers who consented

to participate in the study completed a brief baseline interview, listened to a randomly

assigned gain-, loss-, or mixed-framed physical activity message dictated by a CISInformation Specialist, and responded to a manipulation check question (baseline post-

message). All screening and baseline procedures were conducted by CIS Information

Specialists using a custom-made, secure, on-line survey system. A random number

generator built into the survey system automatically assigned all screened callers

(regardless of their eligibility or whether they consented to participate in the 9-week

study) to a message frame condition. The logistical oversight of randomly assigning

callers prior to determining eligibility and consent resulted in an unbalanced

distribution of eligible and consenting participants across conditions (see Figure 1).To ensure adherence to study protocol, the CIS supervisory staff completed weekly

quality assurance checklists evaluating interview administration and message delivery.

Data from the checklists were compiled in a monthly report; no deviations from the

protocol affecting data quality were noted.



Figure

1.

CO

NSO

RT

flow

char

t.



Within one working day of their baseline interview, participants were mailed a

physical activity guide that was framed according to their randomly assigned condition.

Two weeks after baseline, participants were contacted for the Week 2 follow-up

telephone interview. Upon completion of the Week 2 phone interview or after 15

attempts were made to contact the participant over a 4-week period, participants were

sent an unframed letter thanking them for their study participation to date andinforming them of their eligibility to receive a free pedometer. Participants were

instructed to return the enclosed pre-stamped pedometer request postcard to receive

this free gift.

At Week 4, all participants were mailed a pedometer, a pedometer instruction

brochure, and a 4-week pedometer logbook regardless of whether they returned their

request postcard. Each of these items included framed information appropriate to the

participant’s condition assignment. The purpose of providing all participants with a

pedometer was twofold. First, it was necessary to provide all participants pedometers tomaintain intervention consistency. Second, providing a gift is a means of increasing

participant response rates for subsequent assessments (Salant & Dillman, 1994).

Nine weeks after the baseline interview, a research assistant contacted participants

to complete the Week 9 follow-up interview. If the participants were not reached for the

Week 9 follow-up interview after 15 call attempts, they were mailed a brief survey

including the physical activity measures (IPAQ and pedometer use) and a pre-stamped,

pre-addressed return envelope.1

Message developmentAlthough a wealth of physical activity information is available to the American public, a

standard set of print materials does not exist unlike in other countries (e.g. Canada’s

Physical Activity Guide is a standard print message). Thus, the study materials were

developed by collecting physical activity information from websites and print materials

sponsored by government agencies across North America (CDC, NCI, National Heart,Lung and Blood Institute, and Health Canada). The information collected included gain-

framed, loss-framed, and unframed messages, and was rooted in social cognitive theory.

It addressed the benefits of activity and the costs of inactivity (i.e. outcome expectancy

information), included encouraging slogans, and provided strategies to incorporate

physical activity into a daily routine (i.e. self-efficacy related information).

Using existing framed and unframed information, a set of mixed-framed materials

emphasizing outcome expectancies and self-efficacy was created. These mixed-framed

materials served as the messages for participants randomly assigned to the controlcondition. The set of materials included: (a) a baseline telephone message describing

general health and specific cancer-related benefits of physical activity and outlining

the CDC’s physical activity recommendation; (b) a print physical activity guide

detailing the health and cancer-related benefits of physical activity, highlighting the

health costs of physical inactivity, reiterating the CDC’s physical activity recommen-

dation, and listing tips for overcoming physical activity barriers (Health Canada, 2004);

1 The timeline for the Week 4 mailing and Week 9 follow-up interview is only approximate. The occurrence of these follow-upactivities hinged on the Week 2 date of completion (i.e. the Week 4 mailing was always scheduled 2 weeks after thecompletion of the Week 2 interview, and the Week 9 follow-up interview was always scheduled 7 weeks after the Week 2interview). Although all participants were scheduled to complete the Week 2 follow-up 2 weeks after baseline, some werecontacted as many as 15 times over a 4-week period to complete the interview.



(c) a pedometer brochure restating the costs of inactivity, the benefits of physical

activity, and the CDC physical activity recommendations, as well as providing

instructions for pedometer use; and (d) a slogan for the logbook and pedometer cover

encouraging participation in physical activity. The source of the information was noted

either within the body of the message (e.g. : : : the Centers for Disease Control

encourage participation in regular physical activity : : : ) or a reference section on theback of the printed material. Providing these details may have enhanced the credibility

of the messages.

The mixed-framed materials included 77 frameable physical activity statements (i.e.

could be framed); 14% were gain-framed, 3% were loss-framed, and 82% were unframed.

To create the gain-framed materials, all of the loss-framed and unframed statements in

the mixed-framed materials were reframed to emphasize the benefits of engaging in the

recommended amount of physical activity or mentioned the undesirable outcomes that

would be prevented from adequate activity. To create the loss-framed materials, all of thegain-framed and unframed statements from the mixed-framed message were reframed to

emphasize the risks of inactivity or mention desirable outcomes that would be missed

from lack of activity (Rothman & Salovey, 1997). This approach of reframing the mixed-

frame materials to create the gain- and loss-framed materials ensured that all the

materials were factually equivalent. Table 1 includes samples of messages of each type.

In a pilot test of the messages, 29 adults (79% women; Mage ¼ 27:96,SD ¼ 12:11 years) recruited from the community were assigned randomly to review the

gain-, loss-, or mixed-framedmessages. The contentwas rated as similar on all dimensions(e.g. informative, convincing, and understandable, MANOVA Fð4; 44Þ ¼ 0:71, p ¼ :75,Pillai’s Trace ¼ 1:77) except message tone (i.e. separate ratings of positive tone, negative

tone, and message focus, MANOVA Fð8; 38Þ ¼ 2:02, p ¼ :03, Pillai’s Trace ¼ 0:98). Asexpected, the loss-framed messages were evaluated as having a more negative tone and a

greater emphasis on the costs of inactivity when compared with the gain- and mixed-

framed messages. The positive tone for the gain- and mixed-framed messages was rated

similarly. The negative tone was rated as stronger for the mixed-framed message than the

gain-framed message confirming the mixed-frame of the control message.

Baseline measuresThe baseline interview, conducted by CIS Information Specialists, assessed behaviour,

social cognitive variables (outcome value, outcome expectancies, intentions, self-

efficacy), and perceptions of the baseline message.

Physical activity was assessed using the International Physical Activity Ques-tionnaire (IPAQ) telephone-administered short form (Craig et al., 2003). The

participants reported the number of days and amount of time spent walking and

engaging in moderate and vigorous activities over the previous 7 days for at least 10

minutes at a time. Walking and moderate and vigorous activities were defined using the

standard IPAQ descriptions. Walking included walking at work and home, and walking

for transport, recreation, leisure, or sport. Moderate activity included activities that

required moderate physical effort and made the participant breathe somewhat harder

than normal. Vigorous activity included activities that required hard physical effort andmade the participant breathe much harder than normal.

Separate estimates of energy expenditure were calculated for walking, and moderate

and vigorous physical activity by weighting the reported minutes/week by category-

specific metabolic-equivalent values (i.e. METs). The weighted MET (minutes/week)



Table

1.

Sam

ple

mes

sage

s

Gai

nLo

ssC

onto

l

Tele

phone

mes

sage

Acc

um

ula

ting

atle

ast

30

min

ute

sof

moder

ate-

inte

nsi

typhy

sica

lac

tivi

tyon

five

or

more

day

softh

ew

eek,

or

atle

ast

20

min

ute

sofvi

goro

us-

inte

nsi

typhy

sica

lac

tivi

tyon

thre

eor

more

day

softh

ew

eek

can

lead

tom

any

hea

lth

ben

efits

Faili

ng

toac

cum

ula

teat

leas

t30

min

ute

sofm

oder

ate-

inte

nsi

typhy

sica

lac

tivi

tyon

five

or

more

day

softh

ew

eek,

or

atle

ast

20

min

ute

sofvi

goro

us-

inte

nsi

typhy

sica

lac

tivi

tyon

thre

eor

more

day

softh

ew

eek

can

lead

tom

any

hea

lth

pro

ble

ms

Itis

reco

mm

ended

todo

moder

ate-

inte

nsi

typhy

sica

lac

tivi

tyfo

rat

leas

t30

min

ute

son

five

or

more

day

softh

ew

eek,

or

vigo

rous-

inte

nsi

typhy

sica

lac

tivi

tyfo

rat

leas

t20

min

ute

son

thre

eor

more

day

softh

ew

eek

Phy

sica

lac

tivi

tygu

ide

Ben

efits

ofre

gula

rac

tivi

tyin

clude

bet

ter

hea

lth

and

reduce

dri

skofco

lon

and

bre

ast

cance

r

Hea

lth

risk

sofin

activi

tyin

clude

colo

nan

dbre

ast

cance

ran

dpoor

hea

lth

Bet

ter

hea

lth

isa

ben

efit

ofre

gula

rac

tivi

ty.

Hea

lth

risk

sofin

activi

tyin

clude

colo

nan

dbre

ast

cance

rPe

dom

eter

bro

chure

and

logb

ook

Phy

sica

lac

tivi

tyca

nim

pro

veyo

ur

hea

lth

–ge

tac

tive

!Phy

sica

lin

activi

tyca

nca

use

hea

lth

pro

ble

ms

–ge

tac

tive

!G

etac

tive

!

Pedom

eter

sloga

nPhy

sica

lac

tivi

tyim

pro

ves

hea

lth

Phy

sica

lin

activi

tyw

ors

ens

hea

lth

Phy

sica

lac

tivi

tyin

fluen

ces

hea

lth



were summed across the moderate and vigorous activity categories to create a weighted

index of moderate and vigorous physical activity. An index of moderate and vigorous

physical activity rather than a total activity score corresponds with the message focus

and the Centers for Disease Control’s (CDC) physical activity recommendation:

Accumulate at least 30 minutes of moderate physical activity on$5 days/week or at

least 20 minutes of vigorous physical activity on $3 days/week (Centers for DiseaseControl, 2006). The IPAQ short form has acceptable measurement properties for

monitoring physical activity via telephone interview (Craig et al., 2003). Test–retest

reliability coefficients range from 0.65 to 0.88 with a pooled Spearman correlation

coefficient of 0.76. IPAQ scores correlate positively with accelerometer step counts

r ¼ :30. The magnitude of this correlation is consistent with median criterion validity

correlations reported for other self-reported measures.

Social cognitive variablesBecause CIS guidelines limited the length of the baseline interview to 7 minutes, eachsocial cognitive variable was measured using a single item rated on a 5-point scale. The

measures were adapted from research evaluating the reach of a physical activity mass

media campaign (Plotnikoff et al., 2006). Participants were provided the following

definition of regular physical activity: ‘Moderate and vigorous activities performed at

least three times/week for a total of 20 minutes or more/day. Activities can be

accumulated through the day in 10 minute periods’.

Outcome valuewas assessed with the item ‘Improving my health by participating in

regular physical activity over the next 2 weeks is : : : ’ (1 ¼ not at all important;5 ¼ extremely important; Sallis, Hovell, Hofstetter, & Barrington, 1992). Outcome

expectancy was assessed with the item ‘How likely is it that your health would improve

if you participated in regular physical activity over the next 2 weeks?’ (1 ¼ not likely;

5 ¼ extremely likely; Sallis et al., 1992). Self-efficacy was evaluated by the level of

agreement with the statement ‘If it were entirely up to me, I am confident that I would

be able to participate in regular physical activity over the next 2 weeks’ (1 ¼ disagree a

lot; 5 ¼ agree a lot; Armitage & Conner, 1999). Intentions were evaluated by the level

of agreement with the statement ‘I intend to participate in regular physical activity overthe next 2 weeks’ (1 ¼ disagree a lot; 5 ¼ agree a lot; Ajzen, 2002).

Post-message manipulation checkThe post-message interview included a single-item manipulation check rated on a

5-point scale: ‘Would you say the information I just read you : : : ’ (1 ¼ focused heavily

on the benefits of physical activity; 5 ¼ focused heavily on the risks of inactivity). This

item has been used in previous framing studies and is sensitive to differences in message

frame (Toll et al., 2007).

Week 2 follow-up measuresThe Week 2 follow-up interview, conducted by research staff, assessed the participants’

perceptions of the last week, physical activity behaviour, outcome expectancies,

intentions, self-efficacy, and message evaluations. Specifically, participants indicated

whether the last week had been typical (0 ¼ no, 1 ¼ yes). Physical activity was

assessed using the IPAQ described above.



Social cognitive variablesOutcome expectancies were assessed using the single item administered at baseline.

The intentions, self-efficacy, and manipulation check measures were expanded to

include the baseline interview questions and additional supplemental questions. It was

possible to use longer measures because unlike the baseline interview, the Week 2

interview was not subject to a time limit. Of benefit, using multi-item measuresincreases the reliability of the assessment tool (Streiner & Norman, 1999). The baseline

interview questions for each social cognitive variable (except self-efficacy) preceded

any additional questions related to the construct. This order was arranged to minimize

the potential impact of using an expanded multi-item scale on participants’ responses to

the initial baseline questions. All items were rated on a 5-point scale.

The longer intentions measure included a second item: ‘I will try to engage in

regular physical activity over the next two weeks’ (1 ¼ disagree a lot; 5 ¼ agree a

lot; Armitage, 2004). The two intentions items correlated highly, rð179Þ ¼ :71, p ,

:01 and have demonstrated a predictive validity in health behaviour research

(Armitage, 2004).

The longer measure of self-efficacy included three supplemental items: (a) ‘How

confident are you that you will be able to participate in regular physical activity over the

next 2 weeks?’ (1 ¼ not at all confident; 5 ¼ extremely confident); (b) ‘I believe I have

the ability to participate in regular physical activity over the next 2 weeks’

(1 ¼ disagree a lot; 5 ¼ agree a lot); and (c) ‘To what extent do you see yourself as

being capable of participating in regular physical activity over the next 2 weeks?’(1 ¼ not likely; 5 ¼ extremely likely; Armitage & Conner, 1999). The 4-item scale was

internally consistent (Cronbach’s a ¼ :82) and has exhibited predictive validity in the

physical activity domain (Rhodes & Courneya, 2003).

The five supplemental manipulation check items evaluated the extent to which the

message was (a) read (1 ¼ none, 5 ¼ all); (b) believable (1 ¼ not at all believable,

5 ¼ extremely believable); (c) interesting (1 ¼ not at all interesting, 5 ¼ extremely

interesting); (d) informative (1 ¼ not at all informative, 5 ¼ extremely informative);

and (e) perceived as negative or positive (message tone: 1 ¼ extremely negative,5 ¼ extremely positive). These items have been used widely in message framing

research to demonstrate the difference between gain- and loss-framed messages (Brug,

Steenhuis, Van Assema, & De Vries, 1996; Toll et al., 2007).

Mailed follow-upWhether or not participants returned a pre-stamped postcard requesting a

free pedometer was used as an indirect indicator of physical activity intentions.

Similar indirect intentions measures have been used effectively in framing research

(e.g. requesting bottles of sunscreen [Detweiler et al., 1999], encouraging smoking

cessation [McKee et al., 2004]).

Week 9 follow-up measuresThe final follow-up interview, conducted by research staff, repeated all of the questionsfrom the Week 2 follow-up interview and assessed pedometer use. Pedometer use was

assessed through the self-report, open-ended item: ‘In the past month, how many

days did you use the pedometer for the entire day?’ (Craig, Cragg, Tudor-Locke, &

Bauman, 2006).



Analyses

Managing outliersTo manage the non-normal distribution typical of physical activity reports, the IPAQ data

were treated according to steps prescribed in the IPAQ scoring manual (International

Physical Activity Questionnaire, 2006). Physical activity data were removed for

individuals whose total walking and moderate and vigorous physical activity time weregreater than 960min/day. Time variables exceeding 180 minutes were truncated to be

equal to 180 minutes. Subsequently, all IPAQ scores were submitted to a square-root

transformation to further remedy the non-normal distribution (Tabachnick & Fidell,

2001). In testing the assumptions of regression, three additional outlying physical

activity scores were identified (i.e. the residual value was 3 SD above or below the

predicted value) and eliminated from further analysis (Kleinbaum, Kupper, Muller, &

Nizam, 1998).

Analytic approachAnalyses of variance (ANOVA) for continuous data and chi-squared analyses for

categorical data were completed initially to ensure effective randomization, confirm

manipulation success (independent variable: experimental condition), and examine

participant characteristics associated with study drop-out (independent variable: study

adherence). Next, regression analyses with planned contrasts were conducted to testthe primary hypothesis that gain-framed messages would be more effective than loss-

and mixed-framed messages in affecting physical activity participation and beliefs.

Experimental condition was dummy-coded with the gain-framed message condition as

the reference group. In each regression model, relevant covariates (e.g. baseline values)

were entered first, followed by the dummy-coded contrast variables. This planned

contrast regression approach necessitated separate analyses for each follow-up period.

A benefit of conducting separate analyses for each follow-up is that it allows participants

who only completed one follow-up to be included in the analysis, thus increasingstatistical power.

Finally, in order to test the secondary hypothesis of mediation, an additional series of

regression analyses was conducted using a product coefficient approach (Mackinnon,

Fairchild, & Fritz, 2007; Mackinnon, Lockwood, Hoffman, West, & Sheets, 2002). The

product coefficient method examines whether the indirect effect of the independent

variable (i.e. message type) on the dependent variable (i.e. physical activity behaviour) is

significant. To conduct this test of significance, a z score was calculated by multiplying

the unstandardized beta coefficients (a,b) for the equation regressing the dependentvariable (social cognitive variables; coefficient b) on the independent variable (message

type; coefficient a) and then dividing the product of coefficients a and b by a standard

error term (sab ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffis2ab

2 þ s2ba

2 þ s2a þ s2

b

p; Mackinnon et al., 2007). The calculated z

score was then compared with a standard normal distribution.

Results

Study attritionAs described in Figure 1, 517 eligible callers were recruited and randomized to a message

condition (Ngain ¼ 158, N loss ¼ 165, Nmixed ¼ 194). Out of the 517 participants

randomized, 195 were excluded as a result of misreporting activity level in the



screening interview. Among the excluded participants, baseline IPAQ scores failed to

corroborate screening reports that indicated participation in fewer than three 20 minute

bouts of moderate to vigorous activity/week. The excluded participants

(M ¼ 3; 201:54^ 2; 906:37METmin=wk) engaged in significantly more moderate and

vigorous activity than the retained participants (M ¼ 395:84^ 512:88METmin=wk,

Fð1; 514Þ ¼ 579:01, p , :001, d ¼ 1:34). It was necessary to enforce this eligibilitycriterion given the evidence that messages are only effective for sedentary segments of

the population (Vandelanotte et al., 2005). The final sample consisted of 322

participants (Ngain ¼ 101, N loss ¼ 106, Nmixed ¼ 115). Out of these participants, 56.2%

completed the Week 2 assessment and 51.6% completed the Week 9 assessment. The

participants who dropped out (i.e. could not be reached for either telephone follow-up)

or withdrew were older and more likely from a non-white ethnic group than those who

completed at least one follow-up, x2ð1; N ¼ 322Þ ¼ 5:60, p ¼ :02.

DemographicsThe demographic characteristics of the participants retained for the subsequent

analyses are reported in Table 2. There were no significant differences in demographiccharacteristics between the three experimental groups, p values . :05. Most

participants were white (67.4%), female (76.1%) and had completed at least some

college (75.8%). The average age of the sample was 47.40 years (SD ¼ 12:03).

Manipulation checkAs expected, no differences in general evaluations of the messages emerged. The gain-,

loss-, and mixed-framed print materials were rated as being similarly believable,

interesting, and informative, p values . :05. Furthermore, participants indicated

Table 2. Demographic characteristics

Experimental condition

Full(N ¼ 322)

Gain(N ¼ 101)

Loss(N ¼ 106)

Control(N ¼ 115)

Age (years [SD ]) 47.40 (12.03) 49.29 (11.17) 45.31 (12.51) 47.67 (12.10)Gender (% male) 13.04 10.89 11.32 16.52Race (% white) 67.39 69.31 64.52 68.70Education (%)a

# HS grad 24.22 25.74 27.36 20.00Some college 31.99 30.69 32.08 33.04College grad 24.22 23.76 21.70 26.96Post grad 16.46 16.83 15.09 17.39

Income (%)b

, $40,000 44.10 48.51 48.11 36.52$40-59,000 15.84 14.85 11.32 20.87$60-79,000 12.11 10.89 14.15 11.30$80,000 þ 21.12 18.81 22.64 21.74

Note. aThere are 10 people missing from this analysis because they did not report this information.b There are 22 people missing from this analysis because they did not report this information.SD, standard deviation; HS, high school.



reading similar amounts of information across the three conditions, p values . :05.However, consistent with the objectives of the study, significant differences

in participants’ ratings of the tone and focus of the baseline telephone message

and physical activity guide emerged (Fbaselineð2; 313Þ ¼ 38:96, p , :001;MANOVA FWeek 2 ð4; 336Þ ¼ 4:22, p ¼ :002, Pillai’s Trace ¼ 0:10). As presented in

Table 3, participants in the loss-framed condition indicated that the messagesemphasized the risks of inactivity, whereas participants in the gain- and mixed-framed

conditions indicated that the messages emphasized the benefits of physical activity.

Additionally, the loss-framed messages were rated as having a more negative tone than

the gain- and mixed-framed messages. Although the multivariate effects for the model

comparing the tone and focus of the pedometer brochure and slogan did not reach

conventional levels of statistical significance ( p ¼ :15), the loss-framed message tended

to be rated as having a more negative tone and putting a greater emphasis on the costs of

inactivity than the gain- and mixed-framed messages.

Primary hypothesis testing

Moderate and vigorous physical activityThe hierarchical linear regression models testing the hypothesis that gain-framed

messages are more effective than loss- and mixed-framed messages for promoting

moderate and vigorous physical activity controlled for four variables: (a) moderate and

vigorous physical activity at baseline, which strongly affects the possibility for behaviour

change (i.e. people who engage in less activity at baseline have greater room to improve

than people who report more activity); (b) outcome value at baseline, which is an

important social cognitive determinant of behaviour change (Bandura, 1997) nottargeted specifically in the intervention; (c) walking at follow-up, which may be an

alternate physical activity competing with moderate and vigorous activities (the

outcome of interest); and (d) the extent to which the week preceding follow-up

represented typical activity levels (many participants were caring for cancer survivors

and experienced interruptions in their regular routines).

The overall models predicting the square root of moderate and vigorous

physical activity at the Week 2 and Week 9 follow-up were significant

(FWeek 2 ð6; 141Þ ¼ 5:47, p , :001; FWeek 9 ð6; 162Þ ¼ 11:89, p , :001). The contrastvariables explained incremental variance in the outcome at the Week 9 follow-up,

R2change ¼ :03, Fð2; 162Þ ¼ 3:06, p ¼ :05. Both the gain versus loss and the gain

versus mixed contrast variables provided a unique contribution (bgain vs: loss ¼ 2:15,p ¼ :05, CI ¼ :14 to 2 12:80; bgain vs:mixed ¼ 2:17, p ¼ :02, CI ¼ 21:08 to 2 13:93).As Table 4 shows, the predictions were supported: participants in the gain-framed

condition reported greater physical activity at the Week 9 follow-up when compared

with participants in the loss- and mixed-framed conditions. As described in Table 4, the

Week 9 regression model was not significant for the unadjusted means; however, thepattern of findings was consistent with the pattern for the adjusted means. These

contrast effects were not significant at the Week 2 follow-up.

Secondary hypothesis testingSeparate hierarchical linear regression analyses controlling for baseline scores were

conducted to examine the hypothesis that gain-framed messages are more effective

than loss- and mixed-framed messages for changing self-efficacy, intentions, and



Table

3.

Raw

and

adju

sted

mea

ns

for

man

ipula

tion

chec

kitem

s

Bas

elin

eW

eek

2W

eek

9

Gai

nLo

ssd

Gve

rsus

LM

ixed

dG

vers

usM

Gai

nLo

ssd

Gve

rsus

LM

ixed

dG

vers

usM

Gai

nLo

ssd

Gve

rsus

LM

ixed

dG

vers

usM

Mes

sage

tone

Inac

tivi

tyri

sks/

activi

tyben

efits

4.2

42.7

60.9

0*

4.4

32

0.1

44.6

74.2

60.5

6*

4.4

80.3

04.6

24.4

60.0

34.5

30.1

5

(SD

)(1

.43)

(1.8

3)

(1.2

2)

(0.5

7)

(0.8

7)

(0.7

0)

(0.6

3)

(0.7

1)

(0.6

2)

Neg

ativ

e/posi

tive

3.9

13.3

90.3

8*

4.1

02

0.1

43.6

13.2

00.6

63.8

52

0.1

8(S

D)

(1.4

2)

(1.3

7)

(1.2

6)

(1.4

2)

(1.4

6)

(1.3

2)

Mes

sage

appea

lA

mount

read

4.2

93.8

30.4

14.1

60.1

34.3

34.1

30.0

34.1

70.1

2(S

D)

(0.9

1)

(1.3

0)

(1.0

2)

(1.1

7)

(1.1

7)

(1.3

5)

Bel

ieva

ble

4.6

74.3

50.5

24.4

80.2

84.5

74.5

02

0.0

24.4

30.1

8(S

D)

(0.5

4)

(0.6

8)

(0.7

7)

(0.7

7)

(0.6

8)

(0.8

0)

Inte

rest

ing

4.1

43.9

80.1

93.9

80.1

84.2

34.0

22

0.0

44.0

70.1

8(S

D)

(0.9

0)

(0.7

6)

(0.8

7)

(0.8

9)

(1.0

4)

(0.9

3)

Info

rmat

ive

4.4

04.1

50.3

34.2

00.2

54.3

64.2

62

0.0

74.1

30.2

4(S

D)

(0.7

2)

(0.7

7)

(0.8

6)

(0.9

0)

(0.8

3)

(0.9

9)

Not

e.*p

,:0

5.SD

,st

andar

ddev

iation;d,

Cohen

’sd;

Gve

rsus

L,co

mpar

ison

ofth

ega

in-

vers

uslo

ss-f

ram

edm

essa

geef

fect

s;G

vers

usM

,co

mpar

ison

ofth

ega

in-

vers

usm

ixed

-fra

med

mes

sage

effe

cts.



Table

4.

Raw

and

adju

sted

mea

ns

for

phy

sica

lac

tivi

ty-r

elat

edoutc

om

eva

riab

les

Bas

elin

eW

eek

2W

eek

9

Gai

nLo

ssM

ixed

Gai

nLo

ss

dG

vers

usL

Mix

ed

dG

vers

usM

Gai

nLo

ss

dG

vers

usL

Mix

ed

dG

vers

usM

Mod.an

dV

ig.PA

(MET

S·m

in·w

eek2

1)

Raw

(SD

)395.8

4328.8

7356.6

71,1

56.0

0814.4

70.2

61,0

78.0

40.0

51,3

98.6

4947.0

20.3

0a

844.9

30.4

0b

(512.

88)

(476.

85)

(469.

79)

(1,5

78.6

8)

(960.5

5)

(1,5

77.0

9)

(1,6

02.5

5)

(1,3

66.6

0)

(1,0

89.8

3)

Adju

sted

(SD

)1,1

55.5

8819.1

40.5

91,0

61.0

80.1

41,3

99.7

8947.2

50.7

1*

842.5

00.8

1*

(593.

09)

(538.0

8)

(747.

31)

(648.4

1)

(624.8

5)

(722.

20)

Wal

king

(MET

S·m

in·w

eek2

1)

Raw

(SD

)662.6

1715.3

8812.1

7809.4

9480.7

50.4

1692.7

40.1

1982.7

6913.8

70.0

61,0

27.1

32

0.0

3

(1,0

83.3

0)

(1,0

60.3

5)

(1,1

94.0

5)

(983.

39)

(544.0

5)

(1,0

73.3

5)

(1,2

30.0

9)

(1,1

34.3

5)

(1,4

29.3

5)

Adju

sted

(SD

)800.7

7479.9

80.6

5692.0

60.1

8980.0

3915.4

60.1

31,0

20.6

52

0.0

9

(542.

86)

(430.4

5)

(635.

43)

(483.4

9)

(489.1

5)

(447.

60)

Pedom

eter

reques

t

(%re

turn

ed)

41.5

841.5

139.1

3

Pedom

eter

use

(#ofday

s[S

D])

10.2

610.8

12

0.0

510.8

82

0.0

5

(11.4

2)

(12.0

2)

(11.9

5)

Not

e.*p

,:0

5.SD

,st

andar

ddev

iation.

aB¼

2:1

7,p¼

:06.

bB¼

2:1

7,

p¼

:05;

how

ever

,th

eove

rall

model

pre

dic

ting

unad

just

edm

oder

ate

and

vigo

rous

phy

sica

lac

tivi

tyw

asnot

sign

ifica

nt

(Fð2;166Þ¼

2:4

5,

p¼

:09;

d¼

Cohen

’sd

),G

vers

usL,

com

par

ison

ofth

ega

in-

vers

uslo

ss-f

ram

edm

essa

geef

fect

s;G

vers

usM

,com

par

ison

ofth

ega

in-

vers

usm

ixed

-fra

med

mes

sage

effe

cts.



outcome expectancies. The overall model for each variable was significant at both

the Week 2 and Week 9 follow-up, p values , :05. The contrast variables contributed

incremental variance in the prediction of intentions (R2change ¼ :03,Fð2; 175Þ ¼ 3:48, p ¼ :03) and self-efficacy (R2change ¼ :03, Fð2; 174Þ ¼ 3:46,p ¼ :03) at the Week 2 follow-up. The beta weights indicated differences in

intentions and self-efficacy between the gain- and loss-framed conditions only(b values ¼ 2:20, p values # :01, 2:23 $ CI values $ 22:75). As hypothesized,

participants who received gain-framed messages had greater self-efficacy and

intentions at the Week 2 follow-up than those who received loss-framed messages.

As described in Table 5, the gain- versus loss-framed contrast effect for the unadjusted

means was significant. The pattern of findings for unadjusted intentions was

consistent with the pattern for the adjusted means; the unadjusted model was not

significant. The contrast effects were not significant for self-efficacy and intentions at

the Week 9 follow-up or for outcome expectancies at either time point.

Mediational analysesz Scores were calculated to examine whether the significance of the indirect effects of

message frame on Week 9 moderate to vigorous intensity physical activity was mediatedby Week 2 self-efficacy and intentions. Week 9 physical activity behaviour and Week 2

self-efficacy and intentions were included in these models because these were the

variables affected by gain-framed messages, a necessary condition for establishing

mediation according to Baron and Kenny (1986). A comparison of the z scores with a

standard normal distribution revealed that neither the indirect effect mediated by self-

efficacy (zgain vs: loss ¼ 2:50, zgain vs:mixed ¼ :51) nor the effect mediated by intentions

(zgain vs: loss ¼ 2:29, zgain vs:mixed ¼ 2:22) was significant, p values . :05. A failure to

demonstrate significance suggests that the framing effects observed at Week 9 mustbe mediated by variables other than intentions, self-efficacy, and outcome expectancies.

Ancillary analyses

Walking activityA hierarchical linear regression model was conducted to test whether gain-framed

messagesweremore effective than loss- andmixed-framedmessage in promotingwalking.

Covariates included: (a) walking activity at baseline; (b) outcome value; (c) moderate and

vigorous physical activity at follow-up; and (d)whether the week preceding follow-upwas

typical. The overall models at Week 2 and Week 9 were significant

(FWeek2 ð6; 141Þ ¼ 10:53, p , :001; FWeek 9 ð6; 163Þ ¼ 5:92, p , :001); however, the

contrast effects were not. As shown in Table 4, gain-framed messages did not differentially

promote walking activity compared to the loss- and mixed-framed messages.

Pedometer request and useThe binary logistic regression analysis with the likelihood of requesting a pedometer

regressed on the predictor variables was not significant. As indicated by the percentagesin Table 4, gain-framed messages did not increase the likelihood of requesting a

pedometer more so than the loss- or mixed-framed messages. The hierarchical linear

model regressing the number of self-reported days of pedometer use on the predictor

variables was not significant. As shown in Table 4, participants in the gain-framed



Table

5.

Raw

and

adju

sted

mea

ns

for

pote

ntial

soci

alco

gnitiv

em

edia

tor

vari

able

s

Bas

elin

eW

eek

2W

eek

9

Gai

nLo

ssM

ixed

Gai

nLo

ssd

Gve

rsus

LM

ixed

dG

vers

usM

Gai

nLo

ssd

Gve

rsus

LM

ixed

dG

vers

usM

Inte

ntions

Raw

(SD

)3.7

23.8

03.7

32.0

81.8

70.3

7a

2.0

20.1

21.9

22.0

22

0.0

11.9

52

0.0

5(1

.21)

(1.1

9)

(1.1

2)

(0.5

0)

(0.6

3)

(0.4

7)

(0.6

4)

(0.5

1)

(0.5

5)

Adju

sted

(SD

)2.0

81.8

70.8

2*

2.0

20.2

41.9

22.0

22

0.0

11.9

52

0.1

3(0

.26)

(0.2

6)

(0.2

5)

(0.2

4)

(0.2

7)

(0.2

5)

Self-

effica

cyR

aw(S

D)

4.3

94.1

34.2

54.1

83.7

20.4

7*

4.0

10.2

03.9

73.8

80.1

73.8

80.1

0(0

.95)

(1.0

1)

(1.0

7)

(0.9

2)

(1.0

2)

(0.7

5)

(1.0

0)

(0.8

6)

(0.8

8)

Adju

sted

(SD

)4.1

83.7

21.3

4*

4.0

10.4

33.9

73.8

70.4

03.8

70.4

5(0

.33)

(0.3

6)

(0.4

3)

(0.1

9)

(0.2

2)

(0.2

4)

Outc

om

eex

pec

tanci

esR

aw(S

D)

3.8

73.9

44.0

94.4

44.2

20.2

24.2

10.2

34.3

04.2

42

0.0

24.1

90.1

0(1

.29)

(1.1

5)

(1.0

6)

(0.9

3)

(0.9

9)

(1.0

4)

(1.2

2)

(1.1

0)

(1.1

2)

Adju

sted

(SD

)4.4

44.2

30.5

54.2

10.5

84.3

04.2

42

0.0

84.1

90.3

3(0

.45)

(0.3

2)

(0.3

3)

(0.3

6)

(0.3

1)

(0.3

1)

Not

e.*p

,:0

5.SD

,st

andar

ddev

iation.

aB¼

2:1

8,p

¼:0

4;h

ow

ever

,the

ove

rall

model

pred

icting

inte

ntio

nsun

adju

sted

for

base

line

cova

riat

esw

asno

tsi

gnifi

cant

(Fð2;17

6Þ¼

2:4

0,p

¼:0

9;d¼

Cohe

n’s

d),

Gve

rsus

L,co

mpa

riso

nofth

ega

in-

vers

uslo

ss-fra

med

mes

sage

effe

cts;

Gve

rsus

M,c

om

pari

son

ofth

ega

in-

vers

usm

ixed

-fra

med

mes

sage

effe

cts.

The

valu

esfo

rse

lf-ef

ficac

yan

din

tent

ions

shoul

dno

tbe

com

pare

ddir

ectly

ove

rtim

egi

ven

that

asi

ngle

item

mea

sure

was

used

atba

selin

ean

dan

expa

nded

mea

sure

was

used

atth

eoth

ertim

epo

ints

.



condition wore their pedometer as frequently as participants in the loss- and mixed-

framed conditions.

Discussion

The framing postulates of prospect theory suggest that behaviours not associated with

much risk (e.g. physical activity) are best promoted with gain-framed messages

(Rothman & Salovey, 1997). Due to limitations in research design, extant research in the

physical activity domain offered mixed support for this prediction (Jones et al., 2003,

2004; McCall & Martin, 2004; Robberson & Rogers, 1988). Improving upon the

limitations of past research by (a) targeting a receptive audience (i.e. sedentary adults);

(b) providing multiple messages; (c) conducting an extended follow-up; and (d)

including a standard message control group, the current investigation examined theutility of gain-framed messages for promoting physical activity. Indeed, with this

improved design, the results of this study are theoretically important because they

corroborate the framing postulates of prospect theory. Specifically, among sedentary

adults, repeated exposure to gain-framed physical activity messages resulted in a greater

participation in moderate and vigorous activities than loss- and mixed-framed messages

at the Week 9 follow-up. These findings provide further evidence that the framing

postulates are useful guidelines for constructing effective public health messages. In

particular, existing physical activity messages, which are largely mixed-framed, shouldbe revamped to emphasize gain-framed information.

Of interest, framing effects for physical activity behaviour only emerged at theWeek 9

follow-up after repeated exposure to the framedmaterials. Indeed, delivering an adequate

dose of information may be important for optimizing the impact of a message. This dose-

response relationship is exemplified in the Centers for Disease Control’s VERB campaign

– a multi-million dollar mass media initiative promoting physical activity for American

children 9 to 13 years of age. At the 2-year follow-up it was determined that as message

exposure increased, physical activity behaviour increased (Huhman et al., 2007). Thus,similar to the findings in the current study, repeated exposure to the campaign messages

lead to greater behaviour change. Evidence from laboratory-based research (Cacioppo &

Petty, 1989; Claypool, Mackie, Garcia-Marques, McIntosh, & Udall, 2004) suggests that

information is processed more thoroughly and carefully as messages are repeated. This

style of thinking about and scrutinizing all issue relevant details is ideal for eliciting

behavioural change (Brinol&Petty, 2006). In the current study, itmay be thatwith greater

exposure to the messages, participants began to think deeply about the information,

the effects of which were manifest at the Week 9 follow-up.Of note, Jones and colleagues (2003) did report an increase in physical activity

following a single exposure to a gain-framed message delivered from a credible source

(i.e. a medical professional). Emphasizing a credible message source is a means of

increasing the persuasiveness of framed messages and is a potential strategy for

increasing message scrutiny (Grewal, Gotlieb, & Marmorstein, 1994). Thus, in the Jones

et al. study, presenting the information from a credible source may have functioned in a

similar manner as repeated message exposure in the current study. Although the

messages in the current study were cited as being from reputable organizations, it ispossible that the credibility of these organizations was not as influential as a medical

professional. Medical professionals are a preferred and trusted source of health

information (Sillence, Briggs, Harris, & Fishwick, 2007). It is also important to note that

the results of Jones et al., 2003 were somewhat tenuous. This group of researchers was



not able to replicate their findings in an identical follow-up study. Had this series of

studies implemented a repeated message exposure design, the effects may have been

more robust (Brinol & Petty, 2006).

In addition to physical activity behaviour, intentions and self-efficacy were affected

by message frame. At the Week 2 follow-up, participants who received gain-framed

messages had stronger intentions and greater self-efficacy to engage in physical activitythan those in the loss-framed condition. Conceptually, determining that gain-framed

messages elicited more positive social cognition at the Week 2 follow-up and greater

behaviour at the Week 9 follow-up is consistent with the notion that change in cognition

precedes a change in behaviour. Nonetheless, mediational analyses testing this

possibility were not significant. Previous research has reported similar failed attempts to

establish social cognitive variables as mediators of message framing effects (Banks et al.,

1995; Detweiler et al., 1999; McCall & Martin, 2004). In the current study, a lack of

evidence for mediation might be the result of poor measurement correspondence.The referent time frame for the social cognitive variables was ‘the next two weeks’, yet

these variables were used to predict behaviour five weeks later.2 Although greater

measurement correspondence would increase the likelihood of establishing intentions

and self-efficacy as mediators of framing effects, given the past failures, possible

mediators beyond the ones used here (e.g. personal relevance, depth of information

processing) also should be considered (cf. Rothman, Bartels, Wlaschin, & Salovey,

2006). In addition to examining mediating variables, emerging research suggests

that individual differences in how people think about the framed materials should beconsidered as moderators (Updegraff, Sherman, Luyster, & Mann, 2007).

The gain-framed messages did not result in greater walking behaviour, likelihood

of requesting a pedometer, pedometer use, or outcome expectancy when compared

with the loss- or mixed-framed messages. These null findings may be related to message

content. The print materials advocated moderate to vigorous physical activity and

not walking or pedometer use per se. With research indicating that the effects of

print messages do not extend past the specific factors targeted in a message

(Chatzisarantis & Hagger, 2005), it is not surprising that framing effects were evident formoderate and vigorous activities but not walking-related activities. The specificity of

message effects should be considered in future research and physical activity promotion

campaigns.

Messages highlighting outcomes salient to the recipient group result in positive

outcome expectancies (Chatzisarantis & Hagger, 2005). Although the messages in the

current study highlighted numerous benefits associated with physical activity, it is

possible that they did not emphasize salient outcomes. Moreover, only health-related

outcome expectancies for a 2-week period were assessed.2 An expanded measurecapturing expectancies related to all of the salient outcomes highlighted in the messages

and using a longer time frame would have increased the likelihood of detecting framing

effects.

In addition to measurement limitations, this study has other limitations. First,

consistent with the demographics of the Cancer Information Service (CIS) clientele, the

2 According to Ajzen (2002), a specific time frame should be provided when assessing social cognition. However, this approachcreated a measurement conundrum when attempting concurrently to (a) demonstrate change in the mediators from baselineto the Week 2 follow-up (a precondition for testing for mediation and hence the reason for two-week referent) and (b) predictbehaviour assessed five weeks after the Week 2 follow-up.



majority of participants were relatively well-educated, non-Hispanic, white, and women

calling for health information. The sample homogeneity limits the generalizability of the

findings to the broader population. Also, it remains unclear whether gain-framed

messages are effective for promoting physical activity among people who are less

motivated to seek out health information than typical CIS callers.

Second, only a small portion of the people screened was randomized. Among thepeople randomized, some participants were excluded as a result of misreporting their

activity level during the screening interview, and a large portion of participants (44% at

Week 2 and 48% at Week 9) could not be reached for follow-up. The misreporting of

physical activity may reflect the susceptibility of the IPAQ to overreporting (Rzewnicki,

Vanden, & De Bourdeaudhuij, 2003). Perhaps, an alternate measure of physical activity

(e.g. Godin Leisure Time Exercise Questionnaire; Godin & Shephard, 1985) would be

more appropriate for this population. The observed low recruitment and high attrition

rates are common in CIS behaviour change research (Heimendinger et al., 2005). Manycallers are not eligible to participate because they are cancer survivors (i.e. cancer

survivors make up the largest portion of CIS callers). Those callers who are eligible often

decline participation or are unable to complete the study because of the stress they are

experiencing while managing the care of a sick or dying relative (e.g. a spouse with

cancer for whom they initially contacted the CIS). Taken together, these participant-

related factors limit the generalizability of the study findings and highlight the need to

replicate the study in a general community sample.

Despite these limitations, this study provides evidence that emphasizing the healthbenefits of engaging in physical activity is effective for promoting behaviour change.

As predicted, repeated exposure to gain-framed messages resulted in greater physical

activity than loss- or mixed-framed messages. With obesity and inactivity being highly

prevalent in society, these findings have implications for satisfying the desperate need

for effective strategies promoting physical activity to the public. Our findings suggest

that the traditional appeals including both loss- and gain-framed messages should be

reframed to emphasize gain-framed information only. Furthermore, the study findings

suggest that other factors such as message exposure and content should be consideredcarefully to optimize further message effectiveness.

Acknowledgements

Research reported in this article was funded through a grant from the National Cancer Institute

(R01-CA68427) awarded to Peter Salovey. Althea Hicks and Julie Keany Hodorowki are funded

through a National Cancer Institute Contract (HHSN261200511001C, ADB N0. N02-CO-51101).

Any Latimer was supported by a Postdoctoral Fellowship from the Social Sciences and Humanities

Research Counsel of Canada.

Amy Latimer is now an Assistant Professor in the School of Kinesiology and Health Studies at

Queen’s University, Kingston, ON Canada. Tara Rench is now a graduate student in the

Department of Psychology at Michigan State University, East Lansing, MI.

References

Ajzen, I. (2002). Construction of a standard questionnaire for the theory of planned behavior.

Retrieved May 13, 2004, from the University of Massachusetts Website: www.-unix.oit.

umass.edu/, aizen/tpb

Apanovitch, A. M., McCarthy, D., & Salovey, P. (2003). Using message framing to motivate HIV

testing among low-income, ethnic minority women. Health Psychology, 22, 60–67.



Armitage, C. J. (2004). Evidence that implementation intentions reduce dietary fat intake: A

randomized trial. Health Psychology, 23, 319–323.

Armitage, C. J., & Conner, M. (1999). Distinguishing perceptions of control from self-efficacy:

Predicting consumption of a low-fat diet using the theory of planned behavior. Journal of

Applied Social Psychology, 29, 72–90.

Bandura, A. (1997). Self-efficacy: The exercise of control. New York: Freeman.

Banks, S. M., Salovey, P., Greener, S., Rothman, A. J., Moyer, A., Beauvais, J., et al. (1995). The

effects of message framing on mammography utilization. Health Psychology, 14, 178–184.

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, 1173–1182.

Brinol, P., & Petty, R. T. (2006). Fundamental processes leading to attitude change: Implications for

cancer prevention. Journal of Communication, 56, s81–s104.

Brug, J., Steenhuis, I., Van Assema, P., & De Vries, H. (1996). The impact of a computer-tailored

nutrition intervention. Preventive Medicine, 25, 236–242.

Cacioppo, J. T., & Petty, R. E. (1989). Effects of message repetition on argument processing, recall,

and persuasion. Basic and Applied Social Psychology, 10, 3–12.

Cavill, N., & Bauman, A. (2004). Changing the way people think about health-enhancing physical

activity: Do mass media campaigns have a role? Journal of Sports Science, 22, 771–790.

Centers for Disease Control (2006). Physical activity for everyone: Recommendation. Retrieved

July 24, 2006, from the Centers for Disease Control Website: http://www.cdc.gov/nccdphp/

dnpa/physical/recommendations/index.htm

Chatzisarantis, N. L. D., & Hagger, M. S. (2005). Effects of a brief intervention based on the theory

of planned behavior on leisure-time physical activity participation. Journal of Sport and

Exercise Psychology, 27, 470–487.

Claypool, H. M., Mackie, D. M., Garcia-Marques, T., McIntosh, A., & Udall, A. (2004). The effects of

personal relevance and repetition on persuasive processing. Social Cognition, 22, 310–335.

Craig, C. L., Cragg, S. E., Tudor-Locke, C., & Bauman, A. (2006). Proximal impact of Canada on the

move: The relationship of campaign awareness to pedometer ownership and use. Canadian

Journal of Public Health-Revue Canadienne de Sante Publique, 97, S21–S27.

Craig, C. L., Marshall, A. L., Sjostrom, M., Bauman, A. E., Booth, M. L., Ainsworth, B. E., et al.

(2003). International physical activity questionnaire: 12-country reliability and validity.

Medicine and Science in Sports and Exercise, 35, 1381–1395.

Detweiler, J. B., Bedell, B. T., Salovey, P., Pronin, E., & Rothman, A. J. (1999). Message framing and

sunscreen use: Gain-framed messages motivate beach-goers. Health Psychology, 18, 189–196.

Frazier, P. A., Tix, A. P., & Barron, K. E. (2004). Testing moderator and mediator effects in

counseling psychology research. Journal of Counseling Psychology, 51, 115–134.

Godin, G., & Shephard, R. J. (1985). A simple method to assess exercise behavior in the

community. Canadian Journal of Applied Sports Science, 10, 141–146.

Grewal, D., Gotlieb, J., & Marmorstein, H. (1994). The moderating effects of message framing and

source credibility on the price-perceived risk relationship. Journal of Consumer Research, 21,

145–153.

Health Canada (2004). Canada’s Physical Activity Guide for Active Living. Retrieved September

24, 2004, from the Public Health Agency of Canada website: http://www.phac-aspc.gc.ca/

pau-uap/paguide/

Heimendinger, J., O’Neill, C., Marcus, A. C., Wolfe, P., Julesburg, K., Morra, M., et al. (2005).

Multiple tailored messages are effective in increasing fruit and vegetable consumption among

callers to the cancer information service. Journal of Health Communication, s65–s82.

Huhman, M. E., Potter, L. D., Duke, J. C., Judkins, D. R., Heitzler, C. D., & Wong, F. L. (2007).

Evaluation of a national physical activity intervention for children: VERB campaign,

2002–2004. American Journal of Preventive Medicine, 32, 38–43.



International Physical Activity Questionnaire (2006). At a glance: IPAQ scoring protocol.

Retrieved January 10, 2006, from the International Physical Activity Questionnaire website

http://www.ipaq.ki.se/IPAQ.asp?mnu_sel ¼ EEF&pg_sel ¼

Jones, L. W., Sinclair, R. C., & Courneya, K. S. (2003). The effects of source credibility and message

framing on exercise intentions, behaviors, and attitudes: An integration of the elaboration

likelihood model and prospect theory. Journal of Applied Social Psychology, 33, 179–196.

Jones, L. W., Sinclair, R. C., Rhodes, R. E., & Courneya, K. S. (2004). Promoting exercise behaviour:

An integration of persuasion theories and the theory of planned behaviour. British Journal of

Health Psychology, 9, 505–521.

Kleinbaum, D. G., Kupper, L. L., Muller, K. E., & Nizam, A. (1998). Applied regression analysis and

other multivariate methods (3rd ed.). Boston, MA: Duxbury Press.

Kroeze, W., Werkman, A., & Brug, J. (2006). A systematic review of randomized trials on the

effectiveness of computer-tailored education on physical activity and dietary behaviors.

Annals of Behavioral Medicine, 31, 205–223.

Mackinnon, D. P., Fairchild, A. J., & Fritz, M. S. (2007). Mediation analysis. Annual Review of

Psychology, 58, 614.

Mackinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A comparison

of methods to test mediation and other intervening variable effects. Psychological Methods, 7,

83–104.

McCall, L. A., & Martin Ginis, K. A. (2004). The effects of message framing on exercise adherence

and health beliefs among patients in a cardiac rehabilitation program. Journal of Applied

Biobehavioral Research, 9, 122–135.

McKee, S. A., O’Malley, S., Steward, W. T., Neveu, S., Land, M., & Salovey, P. (2004). How to word

effective messages about smoking and oral health: Emphasize the benefits of quitting. Journal

of Dental Education, 68, 569–573.

Plotnikoff, R. C., Spence, J. C., Tavares, L. S., Rovniak, L. S., Bauman, A., Lear, S. A., et al. (2006).

Characteristics of participants visiting the Canada on the move website. Canadian Journal of

Public Health-Revue Canadienne de Sante Publique, 97, S28–S35.

Randolph, W., & Viswanath, K. (2004). Lessons learned from public health mass media campaigns:

Marketing health in a crowded media world. Annual Review of Public Health, 25, 419–437.

Rhodes, R. E., & Courneya, K. S. (2003). Self-efficacy, controllability and intention in the theory of

planned behavior: Measurement redundancy or causal independence? Psychology and

Health, 18, 79–91.

Rivers, S. E., Salovey, P., Pizarro, D. A., Pizarro, J., & Schneider, T. R. (2005). Message framing and

pap test utilization among women attending a community health clinic. Journal of Health

Psychology, 10, 65–77.

Robberson, M. R., & Rogers, R. W. (1988). Beyond fear appeals: Negative and positive persuasive

appeals to health and self-esteem. Journal of Applied Social Psychology, 18, 277–287.

Rodgers, W. M., & Gauvin, L. (1998). Heterogeneity of incentives for physical activity and self-

efficacy in highly active and moderately active women exercisers. Journal of Applied Social

Psychology, 28, 1016–1029.

Rothman, A. J., Bartels, R. D., Wlaschin, J., & Salovey, P. (2006). The strategic use of gain- and loss-

framed messages to promote healthy behavior: How theory can inform practice. Journal of

Communication, 56, S202–S220.

Rothman, A. J., Martino, S. C., Bedell, B. T., Detweiler, J. B., & Salovey, P. (1999). The systematic

influence of gain- and loss-framed messages on interest in and use of different types of health

behavior. Personality and Social Psychology Bulletin, 25, 1355–1369.

Rothman, A. J., & Salovey, P. (1997). Shaping perceptions to motivate healthy behavior: The role of

message framing. Psychological Bulletin, 121, 3–19.

Rzewnicki, R., Vanden, A. Y., & De Bourdeaudhuij, I. (2003). Addressing overreporting on the

international physical activity questionnaire (IPAQ) telephone survey with a population

sample. Public Health Nutrition, 6, 299–305.

Salant, P., & Dillman, D. A. (1994). How to conduct your own survey. New York: Wiley.



Sallis, J. F., Hovell, M. F., Hofstetter, C. R., & Barrington, E. (1992). Explanation of vigorous physical-

activity during 2 years using social-learning variables. Social Science and Medicine, 34, 25–32.

Sallis, J. F., & Owen, N. (1998). Physical activity and behavioral medicine. Thousand Oaks, CA:

Sage Publications.

Schneider, T. R., Salovey, P., Apanovitch, A. M., Pizarro, J., McCarthy, D., Zullo, J., et al. (2001).

The effects of message framing and ethnic targeting on mammography use among low-income

women. Health Psychology, 20, 256–266.

Sillence, E., Briggs, P., Harris, P. R., & Fishwick, L. (2007). How do patients evaluate and make use

of online health information? Social Science and Medicine, 64, 1853–1862.

Squiers, L., & Treiman, K. (2005). The cancer information service research agenda. Retrieved

September 12, 2006, from the National Cancer Institute’s Cancer Information Service website:

http://cis.nci.nih.gov/research/agenda_book.pdf

Steward, W. T., Schneider, T. R., Pizarro, J., & Salovey, P. (2003). Need for cognition moderates

responses to framed smoking-cessation messages. Journal of Applied Social Psychology, 33,

2439–2464.

Streiner, D. L., & Norman, G. R. (1999). Health measurement scales: A practical guide to their

development and use (2nd ed.). New York: Oxford University Press.

Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th ed.). Boston, MA: Allyn

and Bacon.

Toll, B. A., O’Malley, S., Katulak, N. A., Wu, R., Dubin, J. A., & Latimer, A. E. et al. (2007).

Comparing gain- and loss-framed messages for smoking cessation with sustained-release

bupropion: A randomized controlled trial. Psychology of Addictive Behaviors, 21, 531–544.

Tversky, A., & Kahneman, D. (1981). The framing of decisions and the psychology of choice.

Science, 211, 453–458.

Updegraff, J., Sherman, D. K., Luyster, F. S., & Mann, T. (2007). The effects of message quality and

congruency on perceptions of tailored health communications. Journal of Experimental

Social Psychology, 43, 249–257.

Vandelanotte, C., De Bourdeaudhuij, I., Sallis, J. F., Spittaels, H., & Brug, J. (2005). Efficacy of

sequential or simultaneous interactive computer-tailored interventions for increasing physical

activity and decreasing fat intake. Annals of Behavioral Medicine, 29, 138–146.

World Health Organization (2004). Global Strategy on Diet, Physical Activity, and Health.

Retrieved February 15, 2007, from theWorld Health Organization’s website: http://www.who.

int/dietphysicalactivity/strategy/eb11344/en/index.html

Received 24 January 2007; revised version received 1 September 2007


promoting participation in physical activity using framed...

Documents