J Primary Prevent (2012) 33:19–31

DOI 10.1007/s10935-012-0263-2

ORIGINAL PAPER

A Theory of Planned Behavior Research Model

for Predicting the Sleep Intentions and Behaviors
of Undergraduate College Students
Adam P. Knowlden • Manoj Sharma •

Amy L. Bernard

Published online: 1 February 2012

 Springer Science+Business Media, LLC 2012

Abstract The purpose of this study was to opera- participants reported receiving insufficient sleep
tionalize the constructs of the Theory of Planned (M = 407.3 min, SD = 100.75). Multiple regression
Behavior (TPB) to predict the sleep intentions and modeled perceived behavioral control, subjective
behaviors of undergraduate college students attending norm, and attitude toward adequate sleep behavior
a Midwestern University. Data collection spanned on behavioral intention. Collectively, the significant
three phases. The first phase included a semi-struc- predictors produced an R2adjusted value of .362. Further
tured qualitative interview (n = 11), readability by specification of the model identified behavioral inten-
Flesch-Kincaid, face and content validity by a panel of tion as a significant predictor of sleep behavior
six experts. The second phase included stability (R2adjusted = .185). As a population, undergraduate
reliability by test–retest (n = 37). The final phase college students are not achieving adequate sleep.
included construct validation applying confirmatory The TPB was found to be a useful framework for
factor analysis, internal consistency by Cronbach’s predicting the sleep intentions and behaviors of
alpha, and predictive validity (n = 197) employing undergraduate students. Practical implications and
multiple regression analysis. The majority of the recommendations for future research are discussed.

Keywords Sleep health
Undergraduate college

A. P. Knowlden (&)
students
Theory of planned behavior
Health Promotion & Education Program,
University of Cincinnati, 526 Teachers College,
P.O. Box 210068, Cincinnati, OH 45221-0068, USA
e-mail: [email protected]
Introduction
M. Sharma
Health Promotion & Education Program & Public Health
Sciences, University of Cincinnati, 526 Teachers Sleep Health
College, P.O. Box 210068, Cincinnati, OH 45221-0068,
USA Sleeping 7–8 h on a daily basis is an essential element
e-mail: [email protected]
of optimum health (Belloc & Breslow, 1972). Despite
A. L. Bernard this, a mounting body of evidence suggests that sleep
Health Promotion & Education Program & Public Health restriction is on the rise in the society, both in the
Sciences, University of Cincinnati, 526 D Teachers general public and among college populations (Bixler,
College, P.O. Box 210068, Cincinnati, OH 45221-0068,
USA 2009; Lund, Reider, Whiting, & Prichard, 2010).
e-mail: [email protected] Epidemiological research has uncovered the health

123
20 J Primary Prevent (2012) 33:19–31

ramifications of this trend, associating both insuffi- Factor Surveillance System. Analysis of the data
cient (\7 h) and excessive ([8 h) sleep durations with found 35.3% of respondents (n = 75,571) reported
increased rates of morbidity and mortality (Hall et al., obtaining less than 7 h of sleep on average during a
2008; Hublin, Partinen, Koskenvuo, & Kaprio, 2007). 24 h period, 37.9% reported unintentionally falling
Injurious health outcomes associated with inadequate asleep during the day at least 1 day in the preceding
sleep duration include cardiovascular disease (Ayas 30 days, and 4.7% reported falling asleep while
et al., 2003), diabetes mellitus (Gangwisch et al., driving in the preceding 30 days (CDC, 2011).
2007), depression (Strine & Chapman, 2005), auto- The sleep behaviors of college populations mirror
mobile and occupational accidents (Lockley et al., the deleterious trends reported in the general public.
2004), as well as learning and memory problems Data collected by Hicks, Fernandez, and Pellegrini
(Stickgold, Hobson, Fosse, & Fosse, 2001). (2001b) over a period of three decades revealed a 14%
The causal factors explicating the association decrease in college students’ (n = 9,543) median
between inadequate sleep duration and an increased hours of sleep: 7.75 h in 1969; 7.13 h in 1979; 6.75 h
risk for all-cause mortality remain unclear (Hublin in 1989; and 6.65 h in 2001. In a recent investigation,
et al., 2007; Youngstedt & Kripke, 2004). As sleep is a Lund et al. (2010) reported that only 29.4% of college
modulator of hormone release, cardiovascular func- students (n = 1,125) received 8 h or more of total
tion, and glucose regulation, Bixler (2009) postulated sleep per night. Strikingly, a quarter of the participants
that short sleep duration contributes to an increased indicated they received less than 6.5 h of sleep per
risk of cardiometabolic disorders. Conversely, Young- night. A significant amount of the study population
stedt and Kripke (2004) have proposed that mortality reported erratic sleeping patterns including staying up
rates associated with excessive sleep duration are a a full 24 h at least once in the past month (20%) and
consequence of sleep fragmentation and daytime staying up until 3:00 a.m. at least once per week
lethargy. Current evidence indicates a bidirectional (35%).
relationship between sleep and health such that In addition to reduced sleep duration, college
inadequate sleep contributes to the progression of a students are also experiencing reduced sleep quality
number of medical and psychiatric disorders, and that (Jensen, 2003). In their analysis of the prevalence of
these same disorders, in turn, contribute to poor sleep sleep difficulties in college students, Buboltz Jr,
quality (Zee & Turek, 2006). Brown, and Soper (2001) noted that 73% of students
(n = 191) experienced occasional sleeping problems
Societal Sleep Patterns and that only 11% met the criteria for good sleep
quality; in their investigation, difficulties with sleep
Sleep restriction is commonplace in society. Evidence latency, morning fatigue, and waking too early were
indicates behavioral dynamics and sleep disorders are cited as primarily sleep distruptors. Similarly, data
at the core of curtailed sleep durations (Monk, Buysse, collected by Hicks, Fernandez, and Pellegrini (2001a)
Rose, Hall, & Kupfer, 2000; Tamakoshi & Ohno, indicated that students (n = 9,543) in 2000 were 2.96
2004). Behavioral components are interlinked to times more likely to report sleep dissatisfaction than
evolving societal norms and encompass lifestyle students in 1978. The data set also revealed a linear
factors such as shift work, prolonged working hours, increase in sleep dissatisfaction: 24% in 1978, 53% in
jet lag, and maintaining irregular sleep schedules 1988, and 71% in 2000.
(Caldwell, Caldwell, & Schmidt, 2008). Conse- Primary symptoms of sleep deprivation among
quently, sleep researchers Zee and Turek (2006) have college populations include impairment of neurocog-
cited voluntary sleep restriction as a primary factor nitive and psychomotor performance, emotional
responsible for insufficient sleep in modern society. imbalance, and overall lower life satisfaction (Curcio,
A study examining population-based data Ferrara, & De Gennaro, 2006). Interestingly, students
(n = 110,441) collected between 2004 and 2007 are often unaware of the impact sleep restriction has on
revealed 28.3% of US adults slept six or fewer hours their ability to perform cognitive tasks (Brown,
per night (Krueger & Friedman, 2009). In 2009, the Buboltz Jr, & Soper, 2006). A study conducted by
Centers for Disease Control and Prevention (CDC) Pilcher and Walters (1997) found that sleep-deprived
incorporated a sleep module into the Behavioral Risk college students performed considerably worse on a

123
J Primary Prevent (2012) 33:19–31 21

complex task than non-deprived participants; yet, interventions. Given the efficacy of the TPB, the
deprived participants reported significantly higher purpose of this study was to operationalize the
levels of estimated performance and concentration constructs of the TPB to predict the sleep intentions
than did non-deprived participants. and behaviors of undergraduate college students
Brown et al. (2006) cautioned educators to take into attending a Midwestern University. An additional
consideration the general lack of sleep health knowledge study objective was to assess the self-reported sleep
exhibited among college students as opposed to attrib- durations of the sample of college students.
uting students’ poor sleeping habits to deviant lifestyle
choices. Concurrently, satisfactory sleep appears to be a
important health concern among college students. A Methods
survey conducted by the American College Health
Association (ACHA) revealed that 76.4% (n = 29,703) Participants and Instrument Psychometrics
of students had not received information on sleep
difficulties from their colleges, yet 50.9% (n = 29,360) Participant inclusionary criteria for this investigation
were interested in receiving information on this impor- included undergraduate students between the ages of
tant health issue (ACHA, 2010). 18 and 24 years who were unmarried and did not
reside with a parent or legal guardian. These criteria
Theoretical Framework were selected to analyze the sleep patterns of
traditional undergraduate college students. Three
The Theory of Planned Behavior (TPB; Ajzen, 1991) phases of data collection were employed to develop
is highly predictive of human functioning, serving as the TPB-based instrument. Phase I involved conduct-
the model for more than 1,200 empirical studies of ing a semi-structured qualitative interview to generate
behavioral prediction and change (Ajzen, 2011). From theoretical construct indicators. Phase II entailed
the perspective of primary prevention, the TPB has collecting data for the purpose of conducting instru-
been applied to numerous areas of research including mentation test–retest procedures. Phase III data col-
the prediction of safer sex practices, nutrition behav- lection assessed the ability of the instrument to predict
iors, and physical activity (Asare & Sharma, 2009– the sleep intentions and behaviors of the sample. Prior
2010; Ickes & Sharma, 2011). The TPB posits that the to initializing data collection, Institutional Review
intention to perform a behavior is the immediate Board permission to conduct the study was sought and
antecedent of the given behavior. Intentions are obtained. All data were analyzed using Predictive
postulated to capture the motivational factors that Analytics Software Statistics Grad Pack 18.0 data
influence behavior (Ajzen, 1991). Behavioral inten- analysis software.
tion, in turn, is the function of three latent variables:
perceived behavioral control, subjective norms, and Data Collection
attitude toward the behavior (Francis et al., 2004).
Perceived behavioral control determines the extent to Phase I
which an individual believes they are capable of
performing a specific behavior. Subjective norm Phase I procedures entailed conducting a semi-struc-
encompasses an individual’s perception of the social tured qualitative interview for the purpose of instru-
pressure to perform a given behavior. Attitude toward ment item generation (Munhall, 2010). A systematic
a behavior refers to an individual’s overall feeling of random sample of every tenth student who walked into
like or dislike toward a given behavior. the main entrance of the university cafeteria was
The TPB assumes human behavior is rational and solicited to participate in the study. Those who agreed
guided by logical thought processes. The causal chain completed open-ended questionnaires querying their
of the TPB implies that altering behaviorally specific opinions about sleeping 7–8 h each night. The ques-
beliefs can assist in ameliorating unhealthy behaviors. tions were framed according to the universal defini-
This underpinning, along with its parsimonious nature, tions of the TPB constructs of perceived behavioral
has established the TPB as an effective theoretical control, subjective norm, and attitude toward the
framework for developing primary prevention behavior. Sample questions included ‘‘what do you

123
22 J Primary Prevent (2012) 33:19–31

believe are the advantages of sleeping 7–8 h a night’’ recommendations for crafting TPB questionnaires
and ‘‘what factors or circumstances would enable you were consulted.
to sleep 7–8 h a night.’’
The research team independently analyzed the Sleep Behavior Sleep behavior was defined in terms
interview responses. The responses were labeled and of a single, observable action with a specific target,
listed in order of frequency into three categories: action, context, and time (TACT; Francis et al. 2004).
perceived behavioral control, subjective norm, and For the purposes of this study, adequate sleep behavior
attitude toward the behavior. Applying Crabtree and was defined as undergraduate students (target)
Miller’s (1999) five stages of qualitative analysis, data achieving 420–480 min (7–8 h; time) of sleep
saturation was achieved with a sample size of 11. Once (action) every night (context). Sleep behavior was
data saturation was confirmed, instrument indicators assessed through self-report employing 24-h recall of
were developed based on the results of the qualitative the previous night’s sleep (CDC, 2011). Participants
analysis. Items were then combined to produce the first were asked, ‘‘In the past 24 h, how many hours and
draft of the instrument. Next, the draft instrument minutes did you sleep at night time?’’ During data
was subjected to the Flesch Reading Ease test analysis, responses for the sleep behavior construct
(score = 77.6) and Flesch-Kincaid Grade Level Test were transformed from total hours and minutes to total
(score = 5.7 grade level) to assess readability. minutes of sleep accrued at night time in the past 24 h.
Finally, a panel of six experts confirmed face validity, In this context, sleep behavior had a potential range of
content validity, and readability of the instrument over 0–1,440 total minutes of sleep. Scores within the range
a two round process. The panel was comprised of two of 420–480 total minutes of sleep (equivalent to 7–8 h)
population experts, two TPB experts, and two sleep were considered to have met this study’s operational
experts. Based on the recommendations of the panel, definition of adequate sleep behavior; conversely,
one item was eliminated from the subjective norm sub- scores outside of this range were considered to have
scale. Additional recommendations included reword- met this study’s operational definition of inadequate
ing of item stem statements and item endpoints. sleep behavior. The operational definition of adequate
The final version of the instrument was comprised sleep was derived from the current epidemiological
of 29 items: 18 items measuring TPB constructs and evidence (Bixler, 2009).
11 demographic items. TPB constructs were measured
directly in accordance with the primary objective of
Perceived Behavioral Control Four 7-point
this research to develop a predictive model of sleep
semantic differential scale items were employed to
behavior. In application, there are two methods for
measure perceived behavioral control to obtain
measuring constructs of the TPB. The first approach is
adequate sleep. The items tapped the respondents’
to measure the constructs of the theory directly. For
self-efficacy and controllability to engage in adequate
example, to directly measure attitude, participants
sleep behavior. Sample items included ‘‘if I wanted to,
would be requested to indicate whether their attitude
I am confident that I could sleep for 7–8 h every night:
toward a behavior was favorable or unfavorable. The
completely disagree—completely agree’’ and ‘‘for me,
second method is to measure the constructs indirectly.
to sleep 7–8 h every night is: not practical—
To indirectly measure attitude, respondents would be
practical.’’ A range of scores from 4 to 28 were
asked about their specific beliefs. The specific beliefs
possible for the construct.
would then be combined using theoretical principles
and inferences drawn regarding the favorability or
unfavorability of the participants’ attitude toward a Subjective Norm Four 7-point semantic differential
behavior (Francis et al., 2004). For prediction it is scale items were used to assess this construct. Sample
generally accepted that direct measurement is suffi- items included ‘‘most people who are important to me
cient (Ajzen & Fishbein, 1980). Subsequently, con- think that I should sleep 7–8 h every night: completely
structs were measured directly in accordance to the disagree—completely agree’’ and ‘‘my friends want
guidelines developed by Francis et al. (2004). This me to sleep 7–8 h every night: completely disagree—
manual provides in-depth instruction in developing completely agree.’’ A range of scores from 4 to 28
TPB based questionnaires. In addition, Ajzen’s (2002) were possible for construct.

123
J Primary Prevent (2012) 33:19–31 23

Attitude Toward the Behavior Bipolar adjectives Prior to launching the instrument, it was determined
were applied to evaluate attitude toward obtaining data collection for the study would cease once the
adequate sleep. The stem statement that preceded the requirement of 200 fully completed questionnaires
list of adjectives was, ‘‘For me to sleep for 7–8 h was satisfied. Respondents for Phase III were recruited
every night would be….’’ Both instrumental and through an e-mail invitation sent from the university
affective items were included to fully gauge how the registrar. Eligibility to participate in the study was
respondents evaluated the behavior. Sample endpoints determined through a series of inclusionary logic
included ‘‘not enjoyable—enjoyable,’’ ‘‘detrimental to questions built into the electronic instrument. All
my social life—beneficial to my social life,’’ and participants were required to indicate their consent and
‘‘bad—good.’’ Six 7-point semantic differential scale understanding of the study terms and conditions prior
items assessed attitude toward the behavior. to attempting the instrument. A message at the end of
the consent form recommended the participants print
Behavioral Intention Three 7-point semantic the information sheet and retain the form for their
differential scale items were used to gauge the records. Contact information for the research team was
participants’ behavioral intention to obtain adequate included in the form. Participation was voluntary, and
sleep. The phrases ‘‘I intend,’’ ‘‘I will try,’’ and ‘‘I no incentives were offered.
plan’’ were employed to gauge the intention of the The instrument was administered on a weekday in
participants to perform the target behavior within the mid-February of 2011. Based on the university’s
next 24 h. The possible score range for behavioral quarterly calendar, this time frame fell 1 week after
intention was 3–21. The mean of the item scores was mid-term examinations. Respondents were pooled
calculated to provide an overall construct score. from a sampling frame of 4,436 undergraduate
Higher scores reflected a greater intention to perform students between the ages of 18 and 24. A total of
the target behavior. 211 students participated in the study. The full sample
was acquired on the first day the instrument was
launched for a calculated response rate of 4.7%.
Phase II
Fourteen responses were discarded during the data
screening process: 11 for missing data, one for failure
Test–retest participants were recruited from health
to provide consent, and two responses for failing to
promotion and education classrooms. Data collection
satisfy inclusion criteria, leaving a final sample size of
included having the same group of volunteering partic-
197 for data analysis purposes.
ipants (n = 37) complete a pencil-and-paper version of
Instrument refinement during Phase III included
the instrument twice, 2 weeks apart. Acceptable stabil-
Cronbach’s alpha for internal consistency and confir-
ity coefficient values were set a priori at 0.70. The TPB
matory factor analysis (CFA) for construct validity.
constructs of sleep behavior, r(37) = 0.806, p \ .01;
Acceptable Cronbach’s alpha coefficient values were
behavioral intention, r(37) = 0.833, p \ .001; per-
set a priori at .70. The constructs of behavioral
ceived behavioral control, r(37) = 0.905, p \ .01;
intention (a = .925), perceived behavioral control
subjective norm r(37) = 0.803, p \ .01; and attitude
(a = .865), subjective norm (a = .803), and attitude
toward the behavior r(37) = 0.885, p \ .01, exceeded
toward the behavior (a = .801) exceeded the accept-
the acceptable a priori condition.
able a priori criterion. Responses for the item repre-
senting the sleep behavior construct were converted
Phase III from total hours and minutes to total minutes of sleep
accrued at night time in the past 24 h. As the TPB
For Phase III, a power analysis was conducted to sleep behavior construct was represented by a single
determine a sufficient sample size for evaluating the item, testing for internal consistency was not
predictive validity of the instrument. In determining applicable.
sample size, an alpha of .05, a power of .80, and a Construct validation was determined by CFA
population correlation coefficient of .20 were consid- applying the maximum likelihood method. Signifi-
ered. In applying this methodology a sample size of cance of factor loadings for CFA was considered in
200 was found to fit the criteria (Polit & Beck, 2004). light of Stevens (1996) recommendation of ±0.36 for

123
24 J Primary Prevent (2012) 33:19–31

a sample size of 200. Individual CFAs were applied to Table 1 Distribution of means standard deviations for tested
each of the constructs to confirm the a priori TPB theoretical constructs for the sample of undergraduate students
the sample of undergraduate students (n = 197)
framework. Scale items which loaded on more than
one factor were inspected and considered for removal. Range
Applying this benchmark resulted in one factor Construct Possible Observed M SD
solutions for each of the TPB constructs.
Sleep behavior 0–1,440 180–720 407.34 100.75
Behavioral intention 3–21 3–21 13.49 5.89
Perceived behavioral 4–28 4–28 17.08 6.56
control
Results
Subjective norm 4–28 4–28 20.35 4.71
Attitude toward the 6–42 6–42 35.92 5.64
Data Screening
behavior

Prior to analysis, all data were examined for accuracy

of data entry and missing values. Construct normality Theoretical Construct Descriptive Statistics
was evaluated using the Kolmogorov–Smirnov (K–S)
test. The initial K–S test failed to reject the null Table 1 summarizes the distribution of means and
hypotheses that the distributions for the behavioral standard deviations of the constructs evaluated in this
intention and attitude toward the behavior constructs investigation. Examining adequate sleep behavior, the
were normal. Visual inspection of the construct mean minutes of total sleep at night time of the sample
histograms revealed negative skews. Square root was 407.34 with a standard deviation of 100.75. The
transformation with reflection was employed to cor- mean sleep behavior score of the sample fell below the
rect for non-normality of the identified constructs. minimal acceptable score of 420 min set forth in this
study, indicating a large portion of the sample received
inadequate sleep. Closer examination of the data
Demographics revealed only 48 (24.37%) of the participants achieved
adequate sleep behavior as defined by this study.
For Phase III, the instrument was administered to a Among the pool of participants, 108 (54.80%)
sample (n = 197) of undergraduate students (66.0% received insufficient sleep and 41 (20.81%) obtained
female; 91.9% Caucasian; Mage = 20.27 years, SD = excessive sleep. Associations between the dependent
1.537) attending a Midwestern university. The major- and independent variables were tested using Pearson
ity of the students were of full-time status (n = 187; product-moment correlations. Results of correlational
94.9%). Respondents were comprised of first year analyses are provided in Table 2.
(n = 54; 27.4%), second year (n = 54; 27.4%), third
year (n = 31; 15.7%), fourth year (n = 43; 21.8%), Predictive Validity
and fifth year and above (n = 15; 7.6%) students.
Most participants worked either part-time (n = 91; Predictive validity of the instrument was tested using
46.2%) or full-time (n = 18; 9.1%). The mean number multiple linear regression. When fully operational-
of hours worked among respondents who indicated ized, the TPB incorporates behavioral intention as
they were employed (n = 109) was 11.31 (SD = both a predictor and outcome variable. In its role as an
13.46). Further analysis revealed the mean hours of outcome variable, behavioral intention is regressed on
respondents who were employed part-time (n = 92) perceived behavioral control, subjective norm, and
was 16.41 (SD = 7.55) and the mean hours of attitude toward the behavior. As a predictor, behav-
participants who were employed full-time (n = 17) ioral intention is regressed on behavior. To the extent
was 42.29 (SD = 4.83). A significant amount of the to which perceived behavioral control is an accurate
sample were unemployed (n = 88; 44.7%). The reflection of actual behavioral control, perceived
overwhelming proportion of participants (n = 194; behavioral control can be combined with intention to
98.2%) did not have any children. Only three (1.5%) predict behavior (Ajzen, 1991). Generally, the less
individuals reported having children. volitional control an individual has over a behavior,

123
J Primary Prevent (2012) 33:19–31 25

Table 2 Theoretical construct Pearson product-moment correlation coefficients for the sample of undergraduate students (n = 197)
Variable 1. 2. 3. 4. 5.

1. Sleep behavior construct –

2. Behavioral intention constructa 0.434** –
3. Perceived behavioral control construct 0.570** 0.517** –
4. Subjective norm construct 0.206** 0.312** 0.151* –
5. Attitude toward the behavior constructa 0.129 0.357** 0.143* 0.294** –
a
Variable re-reflected
* p \ .05, ** p \ .01

the more significant perceived behavioral control behavioral intention and perceived behavioral control
becomes in the prediction of behavior (Maddux & on the outcome variable sleep behavior. The a priori
DuCharme, 1997). criteria of probability of F to enter a predictor in the
In keeping with these assumptions, two models model was less than or equal to 0.05 and for removing
were developed to specify a complete TPB model. a predictor was greater than or equal to 0.10. Both
Model 1 specified a framework for predicting behav- behavioral intention (p = .005) and perceived behav-
ioral intention. Model 2 provided a complete picture of ioral control (p \ .001) were identified as significant
the TPB by specifying paths between intention, predictors of sleep behavior, F(1, 196) = 52.659,
perceived behavioral control, and behavior. Regres- p \ .001, explaining 34.5% of the variance of the
sion diagnostics on both models confirmed satisfaction participants’ sleep behavior. Table 3 summarizes the
of the underlying regression assumptions of homo- parameter estimates for models 1 and 2. Figure 1
scedasticity of the variance, linearity between the illustrates the final specified TPB model for predicting
predictors and outcome variable, absence of outliers, the sleep intentions and behaviors of the sample.
and absence of multicollinearity.

Model 1 Additional Analyses

Multiple linear regression applying the backward elim- One-way analysis of variance (ANOVA) tests were
ination strategy modeled the theoretical predictors of performed to evaluate differences in mean scores and
perceived behavioral control, subjective norm, and effect sizes of behavioral intention, perceived behav-
attitude toward the behavior on behavioral intention. ioral control, subjective norm, and attitude toward the
The a priori criteria of probability of F to enter a behavior based on inadequate and adequate sleep
predictor in the model was less than or equal to 0.05 and behavior group membership. Groups were dichoto-
for removing a predictor was greater than or equal to mized in accordance with the operational definitions
0.10. Each of the three predictors regressed on behav- of adequate and inadequate sleep behavior employed
ioral intention were deemed significant: perceived in this study. Cohen’s d was calculated to gauge effect
behavioral control (p \ .001), subjective norm (p = size of the differences between groups. In interpreting
.003), and attitude toward the behavior (p \ .001). Cohen’s d, Cohen’s (1988) criteria of .20 for small, .50
Collectively, the significant predictors produced an for moderate, and .80 for large effect sizes were
R2adjusted value of .362, F(3, 196) = 38.133, p \ .001, considered. Wolf’s (1986) criterion of effect sizes
suggesting the model accounted for 36.2% of the greater than or equal to .50 for determining practical/
variance in the behavioral intention to obtain adequate clinical significance was also referenced.
sleep in the sample of participants. Scores of behavioral intention differed significantly
between the inadequate and adequate sleep behavior
Model 2 groups, F(1, 196) = 8.866, p = .003, Cohen’s d =
.50. Those who obtained adequate sleep reported
Linear regression applying the backward elimination higher behavioral intention (M = 15.65, SD = 5.63)
method regressed the theoretical predictors of than those who did not (M = 12.79, SD = 5.82). The

123
26 J Primary Prevent (2012) 33:19–31

Table 3 Parameter estimates for model 1 and model 2 for the sample of undergraduate students (n = 197)
Variable B SE B b t p

Model 1
Constant -1.070 0.440 -2.434 .016
Perceived behavioral control 0.076 0.010 0.457 7.882 .000
Subjective norm 0.041 0.014 0.179 3.000 .003
Attitude toward the behavior 0.044 0.011 0.231 3.865 .000
Model 2
Constant 236.661 17.875 13.240 .000
Behavioral intention 17.691 6.280 0.190 2.817 .005
Perceived behavioral control 7.248 1.037 0.472 6.988 .000
Model 1 tested perceived behavioral control, subjective norm, and attitude toward the behavior regressed on behavioral intention,
R2adjusted = .362, F(3, 196) = 38.133, p \ .001; Model 2 tested behavioral intention and perceived behavioral control regressed on
sleep behavior, R2adjusted = 0.345, F(1, 196) = 52.659, p \ .001. Combined, Models 1 and 2 provide a complete TPB model

Fig. 1 Specified theory of

planned behavior model for
predicting the sleep
intentions and behaviors for
undergraduate college
students. Regression
weights are standardized
beta coefficients; R2 values
are adjusted

effect size for behavioral intention between the two toward the behavior found no significant differences
groups was moderate (Cohen, 1988) and practically based on sleep behavior, F(1, 196) = 0.166, p =
significant (Wolf, 1986). Scores of perceived behav- .684. Those who obtained adequate sleep did not
ioral control also differed significantly between the report significantly higher attitude toward the behavior
inadequate and adequate sleep behavior groups, F(1, (M = 36.21, SD = 4.58) than those who reported
196) = 25.163, p \ .001, Cohen’s d = .81. Those inadequate sleep (M = 35.83, SD = 5.96).
who obtained adequate sleep reported higher per-
ceived behavioral control (M = 20.98, SD = 6.39)
than those who did not (M = 15.83, SD = 6.39). The Discussion
effect size for perceived behavioral control between
the two groups was large (Cohen, 1988) and practi- The purpose of this study was to operationalize the
cally significant (Wolf, 1986). TPB to predict the sleep intentions and behaviors of a
ANOVA tests calculated for subjective norm found sample of undergraduate students attending a Mid-
no significant differences based on sleep behavior, western university. A three-phase instrumentation
F(1, 196) = 3.423, p = .066. Those who obtained process led to the development of a reliable and valid
adequate sleep did not report significantly higher instrument for accomplishing this objective. The
subjective norm (M = 21.44, SD = 4.39) than those current study extends previous application of the
who reported inadequate sleep (M = 20.00, SD = TPB by advancing a theory-based model for predict-
4.77). Similarly, ANOVA tests calculated for attitude ing sleep intentions and behaviors. To the best of the

123
J Primary Prevent (2012) 33:19–31 27

authors’ knowledge this is the first report examining responsibilities. As sleep is largely considered an
college student sleep health from a TPB perspective. expendable requirement by college students, it is often
According to the TPB, perceived behavioral con- disregarded in favor of schoolwork or social activities
trol, subjective norm, and attitude predict behavioral (Brown et al., 2006). Time management education is
intention. Model 1 affirmed that all three independent needed to reverse these perceptions (Vranesh, Madrid,
variables were significant in the prediction of inten- Bautista, Ching, & Hicks, 1999). Lund et al. (2010)
tion. Among the predictors, perceived behavioral cited stress as a salient predictor of poor sleep quality
control was the strongest (b = 0.457), followed by among undergraduate students. Stress management
attitude toward the behavior (b = 0.231) and finally relaxation techniques including deep breathing and
subjective norm (b = 0.179). Collectively, the pre- journaling can assist in alleviating pre-sleep mental
dictors accounted for 36.2% of the variance in the stress (Romas & Sharma, 2010). In conjunction with
behavioral intention to obtain adequate sleep in the stress management, financial management education
sample of participants. can minimize the number of work hours students
The TPB assumes that behavioral intention is the require to meet their financial obligations, allowing
immediate antecedent of behavior. Concurrently, the more time for sleep. Finances are a significant stressor
less volitional control an individual has over a behav- for students and thus should be included in an
ior, the more significant perceived behavioral control intervention to help mitigate overall stress (Romas &
will be in the prediction of the given behavior (Maddux Sharma, 2010).
& DuCharme, 1997). In the specified model, perceived Interventionists should include goal setting as the
behavioral control and intention accounted for 34.5% cornerstone of sleep health programs. Behavior
of the variance in sleep behavior. As demonstrated in change is difficult and without clear, objective goals
Table 3, the standardized beta weight for perceived it is unlikely students will succeed in modifying their
behavioral control (b = 0.472) was more than twice as unhealthy sleep behaviors. Goal setting can increase
large as that of behavioral intention (b = 0.190). perceived behavioral control and improve the likeli-
Additionally, the ANOVA tests conducted to deter- hood that students will successfully implement stress,
mine group differences between adequate and inade- time, and financial management strategies. Addition-
quate sleep behavior group membership demonstrated ally, goal setting can also be used as a measurement
a strong relationship between behavior and perceived device to determine the success of an intervention.
behavioral control. However, subjective norm and There are a variety of goal setting paradigms available
attitude were not directly related to sleep behavior. to program planners. The authors of this paper
These findings suggest that the participants had lower recommend the specific, measurable, action-oriented,
volitional control over sleep behavior. realistic, and time-bound (SMART) goal orientation
template. For more information on goal setting, refer
Interventions to Support Changes in Sleeping to Locke and Latham’s (2002) seminal paper on goal
Behavior setting theory. Modifying perceived behavioral con-
trol environmental antecedents will primarily involve
Health promotion and education interventions are offering more options for class scheduling. Buboltz Jr
strongly needed to improve the sleep health of et al. (2001) have recommended that campus admin-
undergraduate college students. The results of this istrators offer later class times to students. This may be
study suggest practitioners should focus intervention difficult for universities to implement, especially
efforts on increasing perceived behavioral control. As universities with high enrollment. For such universi-
a predictor of behavior, perceived behavioral control ties, flexible classroom alternatives, such as on-line
considers volitional and non-volitional factors that classes, should be considered.
enable the enactment of a behavior. To maximize Assertive communication techniques can be incor-
modification of this construct, both behavioral and porated into intervention sessions to improve sub-
environmental determinants should be addressed. jective norms. To obtain sufficient sleep, it is
Many students underestimate the amount of time imperative that students acquire the ability to com-
university-level work requires and do not properly municate their need to engage in 7–8 h of sleep each
allot sufficient time to fulfill their academic night to significant others in their lives. Role play

123
28 J Primary Prevent (2012) 33:19–31

scenarios employing refusal skills are one possible Recommendations for Future Research
medium health professionals can employ when deliv-
ering an intervention. Additional techniques such Although it is one of the three primary pillars of
as Rational Emotive Behavior Therapy (Ellis and optimal health, sleep has received minimal attention in
MacLaren, 2005) can be employed to refute irrational the primary prevention literature. Yet, from a behav-
beliefs regarding the social pressure to implement self- ioral and environmental perspective, there is much
imposed sleep restriction. room for modification of sleep behaviors through
For students living in shared living environments, health interventions. The Healthy People 2020 report
such as residence halls or campus housing, dormitory has begun to shed light on this critical health topic by
quiet hours, ear plugs, and white noise machines can establishing a national goal to increase the percentage
assist students in overcoming noise barriers. of adults who receive sufficient sleep from the 2008
In addressing attitude, relevant information regard- baseline level of 69.6–70.9% by 2020 (United States
ing the importance of sleep for learning and stress Department of Human Services, 2010). Research
coping should be highlighted. A variety of channels investigating the underlying behavioral and social
can be employed to improve attitude. Classic promo- determinants of voluntary sleep restriction is strongly
tional tools such as educational literature and visual encouraged.
media can be incorporated into an intervention. It is Longitudinal research examining sleep behavior
also recommended that health practitioners incorpo- change across the life span would considerably
rate electronic media platforms. Digital technology increase the capacity of primary prevention interven-
affords the opportunity for interactive activities to tions. In the current context, there is a need to
assist in improving attitude. Examples include inter- understand the changes students go through from high
active websites and social media. From an environ- school to college that lead to the degradation of
mental perspective, university campuses must healthy sleep patterns. Currently, much of this tran-
emphasize the importance of sleep health to improve sition is based on speculation which makes interven-
students’ attitude towards sleep. The American Col- tion development difficult. There is also a need to
lege Health Association Fall 2010 report revealed that determine if unhealthy sleeping patterns garnered
more than 75% of college students have not received during college years continue into students’ profes-
information related to sleep difficulties from their sional lives.
universities. Sleep research indicates the final 2 h of an A need exists for more studies that explore the
8 h sleep epoch as the most critical for memory attitudes, opinions, and beliefs underlying sleep
consolidation and learning (Buboltz Jr et al., 2001). health. Past evidence has found that most college
Subsequently, students who habitually receive fewer students view sleep as an expendable requirement and
than 7–8 h of sleep per night may be depriving that they underestimate the impact sleep deprivation
themselves of their maximum learning potential. has on their neurocognitive performance and psycho-
Universities should make students cognizant of this motor vigilance. Research exploring this phenomenon
fact and emphasize to students their need to prioritize could augment the ability of interventions to result in
sleep. behavior change.
Several channels are available to universities to To reduce measurement error, objective methods
highlight the importance of sleep. One outlet is to for monitoring sleep behaviors should be incorporated
incorporate sleep health education into student orien- in future research. In the present study, self-report was
tations. In this context, a brief intervention based on utilized; however, the accuracy of this method is
the TPB could be employed. To perpetuate an ongoing suspect (Magee, Iverson, Huang, & Caputi, 2008).
positive attitude toward obtaining adequate sleep, Furthermore, the types of self-report items used by
college academic advisors should query students about sleep researchers are inconsistent, confounding the
their sleeping patterns during consultation sessions. ability to conduct meta-analyses. While polysomnog-
Advisors can offer sleep health education literature to raphy is considered the gold standard for collecting
students and, if needed, can refer students to university sleep behavior data (Markov & Goldman, 2006), this
mental health counselors for further assessment and technique is highly specialized and largely impractical
treatment (Jensen, 2003). for primary prevention interventions. Alternatively,

123
J Primary Prevent (2012) 33:19–31 29

actigraphy is a non-invasive technique for monitoring There are a multitude of knowledge gaps in the
and recording sleep-wake patterns (Sadeh & Acebo, relationship between sleep and health. As a field of
2002). Employment of objective measurement tools science, the study of sleep health is in its infancy. As
such as actigraphs would strengthen the findings of sleep science advances, the assumptions of sleep
future sleep health research initiatives. health presented in this report may become outdated.
The advancement of sleep health from a primary For example, some researchers argue that the tradi-
prevention perspective could benefit tremendously tional 7–8 h of sleep considered necessary for optimal
from qualitative research. Understanding the lived health is an average duration and that sleep length
experiences of college students could bolster inter- need is normally distributed in the population (Ferrara
ventionists’ ability to increase perceived behavioral & De Gennaro, 2001). Horne (1988) has contended
control and identify additional factors that impact that humans require only 5–6 h of sleep per night and
sleep behavior. Finally, due to the complexity and that additional sleep is optimal; alternatively, Bonnet
multifactorial nature of sleep health, advanced mod- and Arand (1995) have maintained that humans
eling techniques, such as structural equation model- require 9–10 h each night. As it stands, many of the
ing, are recommended to provide more comprehensive basic elements of sleep health have not been resolved
models of sleep behavior. among prominent sleep researchers.
There are limitations inherent to the TPB. Primar-
Limitations ily, the TPB does not explain affective processing. An
underlying assumption of the TPB is that humans base
There are limitations to this study which should be behavior-related decisions on rational thought pro-
considered when evaluating the findings of this report. cesses; however, unhealthy behaviors are often rooted
Participants were pooled from non-random samples of in irrationality. For example, it is well known that
students at a Midwestern university. Therefore, out- sleep deprivation impairs mental capacity; yet,
comes cannot be generalized beyond the study partic- research has demonstrated that college students
ipants. It is possible that a greater proportion of frequently underestimate the cognitive ramifications
respondents who were less satisfied with their sleep of sleep restriction.
volunteered to participate in the study. Such a bias
could threaten the validity of the descriptive statistics
of the sleep behavior construct. A cross-sectional Conclusions
study design was utilized in this investigation thereby
inhibiting the ability to establish cause-and-effect The present study represented the first attempt to
relationships between variables. The findings of this investigate the sleep behaviors of college students
study were based on the self-reporting accuracy, applying the TPB. The results of this study specified a
integrity, and honesty of the participants. Subse- theory-based model for predicting the sleep intentions
quently, misinterpretation of instrument items may and behaviors of undergraduate college students.
have skewed participant responses. Analysis of the model found that perceived behavioral
The present study considered sleep within the control was a strong predictor of both behavioral
context of voluntary sleep restriction. In this context, intention and sleep behavior. Primary prevention
sleep behavior was limited to self-reported sleep interventions targeting the sleep behaviors of college
duration. The instrument did not address sleep quality students should focus on improving the perceived
or medically diagnosed sleep disorders. In addition, behavioral control to obtain 7 to 8 h of sleep each night.
sleep behavior was operationalized in accordance with
the TPB TACT (time, action, context, and target)
principle; however, sleep is a highly complex behav-
ior, which cannot be fully reified through a single
variable. Sleep behavior data was collected through References
self-report. Sleep researchers have questioned the
Ajzen, I. (1991). The theory of planned behavior. Organiza-
accuracy of this data collection method for measuring tional Behavior and Human Decision Processes, 50(2),
sleep behavior. 179–211. doi:10.1016/0749-5978(91)90020-T.

123
30 J Primary Prevent (2012) 33:19–31

Ajzen, I. (2002). Constructing a TPB questionnaire: Conceptual Ferrara, M., & De Gennaro, L. (2001). How much sleep do we
and methodological considerations. Retrieved from http:// need? Sleep Medicine Reviews, 5(2), 155–179. doi:10.
people.umass.edu/aizen/pdf/tpb.measurement.pdf. 1053/smrv.2000.0138.
Ajzen, I. (2011, March). The theory of planned behavior: A bib- Francis, J. J., Eccles, M. P., Johnson, M., Walker, A., Grimshaw,
liography. Retrieved from http://www.people.umass.edu/ J., Foy, R., et al. (2004). Constructing questionnaires based
aizen/tpbrefs.html. on the theory of planned behavior: A manual for health
Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and services researchers. Newcastle upon Tyne, UK: Centre
predicting social behavior. Englewood Cliffs, NJ: Pre- for Health Services Research, University of Newcastle.
ntice-Hall. Retrieved from http://people.umass.edu/aizen/pdf/
American College Health Association. (2010). National College Francis%20etal.TPB%20research%20manual.pdf.
Health Assessment: Fall 2010 reference group data report. Gangwisch, J. E., Heymsfield, S. B., Boden-Albala, B., Buijs, R.
Retrieved from http://www.acha-ncha.org/docs/ACHA- M., Kreier, F., Pickering, T. G., et al. (2007). Sleep duration
NCHA-II_ReferenceGroup_DataReport_Fall2010.pdf. as a risk factor for diabetes incidence in a large US sample.
Asare, M., & Sharma, M. (2009–2010). Using the theory of Sleep, 30(12), 1667–1673.
planned behavior to predict safer sexual behavior by Hall, M. H., Muldoon, M. F., Jennings, J. R., Buysse, D. J.,
Ghanaian immigrants in a large Midwestern U.S. city. Flory, J. D., & Manuck, S. B. (2008). Self-reported sleep
International Quarterly of Community Health Education, duration is associated with the metabolic syndrome in
30(4), 321–335. midlife adults. Sleep, 31(5), 635–643.
Ayas, N. T., White, D. P., Manson, J. E., Stampfer, M. J., Hicks, R. A., Fernandez, C., & Pellegrini, R. J. (2001a). Striking
Speizer, F. E., Malhotra, A., et al. (2003). A Prospective changes in the sleep satisfaction of university students over
study of sleep duration and coronary heart disease in the last two decades. Perceptual and Motor Skills, 93(3),
women. Archives of Internal Medicine, 163(2), 205–209. 660. doi:10.2466/pms.93.7.660-660.
doi:10.1001/archinte.163.2.205. Hicks, R. A., Fernandez, C., & Pellegrini, R. J. (2001b). The
Belloc, N. B., & Breslow, L. (1972). Relationship of physical changing sleep habits of university students: An update.
health status and health practices. Preventive Medicine, Perceptual and Motor Skills, 93(3), 648.
1(3), 409–421. doi:10.1016/0091-7435(72)90014-X. Horne, J. A. (1988). Why we sleep: The functions of sleep in
Bixler, E. (2009). Sleep and society: An epidemiological per- humans and other mammals (p. 1988). Oxford, UK: Oxford
spective. Sleep Medicine, 10(Supplement 1), S3–S6. University Press.
doi:10.1016/j.sleep.2009.07.005. Hublin, C., Partinen, M., Koskenvuo, M., & Kaprio, J. (2007).
Bonnet, M. H., & Arand, D. (1995). We are chronically sleep Sleep and mortality: A population-based 22-year follow-up
deprived. Sleep, 18, 908–911. study. Sleep, 30(10), 1245–1253. Retrieved from http://
Brown, F. C., Buboltz, W. C., Jr, & Soper, B. (2006). Devel- www.ncbi.nlm.nih.gov.
opment and evaluation of the sleep treatment and education Ickes, M. J., & Sharma, M. (2011). Does behavioral intention
program for students (STEPS). Journal of American Col- predict nutrition behaviors related to adolescent obesity?
lege Health, 54(4), 231–237. doi:10.3200/JACH.54.4.231- ICAN: Infant. Child, & Adolescent Nutrition, 3(1), 38–48.
237. doi:10.1177/1941406410395017.
Buboltz, W. C., Jr, Brown, F., & Soper, B. (2001). Sleep habits Jensen, D. R. (2003). Understanding sleep disorders in a college
and patterns of college students: A preliminary study. student population. Journal of College Counseling, 6(1),
Journal of American College Health, 50(3), 131–135. 25.
Caldwell, J. A., Caldwell, J. L., & Schmidt, R. M. (2008). Krueger, P. M., & Friedman, E. M. (2009). Sleep duration in the
Alertness management strategies for operational contexts. United States: A cross-sectional population-based study.
Sleep Medicine Reviews, 12(4), 257–273. doi:10.1016/ American Journal of Epidemiology, 169(9), 1052–1063.
j.smrv.2008.01.002. doi:10.1093/aje/kwp023.
Centers for Disease Control and Prevention. (2011). Unhealthy Locke, E. A., & Latham, G. P. (2002). Building a practically
sleep-related behaviors—12 States, 2009. Morbidity & useful theory of goal setting and task motivation: A 35-year
Mortality Weekly Report, 66(SS-08), 233–266. Retrieved odyssey. American Psychologist, 57(9), 705–717. doi:10.
from http://www.cdc.gov/mmwr/preview/mmwrhtml/ 1037/0003-066X.57.9.705.
mm6008a2.htm. Lockley, S. W., Cronin, J. W., Evans, E. E., Cade, B. E., Lee, C.
Cohen, J. (1988). Statistical power analysis for behavioral sci- J., Landrigan, C. P., et al. (2004). Effect of reducing
ences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum Asso- interns’ weekly work hours on sleep and attentional fail-
ciates Inc. ures. New England Journal of Medicine, 351(18),
Crabtree, B. F., & Miller, W. L. (Eds.). (1999). Doing qualita- 1829–1837. doi:10.1056/NEJMoa041404.
tive research in primary care: Multiple strategies Lund, H. G., Reider, B. D., Whiting, A. B., & Prichard, J. R.
(pp. 163–177). Newbury Park, CA: Sage Publications. (2010). Sleep patterns and predictors of disturbed sleep in a
Curcio, G., Ferrara, M., & De Gennaro, L. (2006). Sleep loss, large population of college students. Journal of Adolescent
learning capacity and academic performance. Sleep Med- Health, 46(2), 124–132. doi:10.1016/j.jadohealth.2009.
icine Reviews, 10(5), 323–337. doi:10.1016/j.smrv.2005. 06.016.
11.001. Maddux, J. E., & DuCharme, K. A. (1997). Behavioral inten-
Ellis, A., & MacLaren, C. (2005). Rational emotive behavior tions in theories of health behavior. In D. S. Gochman
therapy: A therapist’s guide (2nd ed.). San Luis Obispo, (Ed.), Handbook of health behavior research (Vol. 1,
CA: Impact Publishers. pp. 133–151). New York, NY: Plenum Press.

123
J Primary Prevent (2012) 33:19–31 31

Magee, C. A., Iverson, D. C., Huang, X. F., & Caputi, P. (2008). Stickgold, R., Hobson, J. A., Fosse, R., & Fosse, M. (2001).
A link between chronic sleep restriction and obesity: Sleep, learning, and dreams: Off-line memory reprocess-
Methodological considerations. Public Health, 122(12), ing. Science, 294(5544), 1052–1057. doi:10.1126/
1373–1381. doi:10.1016/j.puhe.2008.05.010. science.1063530.
Markov, D., & Goldman, M. (2006). Normal sleep and circadian Strine, T. W., & Chapman, D. P. (2005). Associations of fre-
rhythms: Neurobiologic mechanisms underlying sleep and quent sleep insufficiency with health-related quality of life
wakefulness. Psychiatric Clinics of North America, 29(4), and health behaviors. Sleep Medicine, 6(1), 23–27.
841–853. doi:10.1016/j.sleep.2004.06.003.
Monk, T., Buysse, D., Rose, L., Hall, J. A., & Kupfer, D. (2000). Tamakoshi, A., & Ohno, Y. (2004). Self-reported sleep duration
The sleep of healthy people: A diary study. Chronobiology as a predictor of all-cause mortality: Results from the
International: The Journal of Biological and Medical JACC study. Japan. Sleep, 27(1), 51–54.
Rhythm Research, 17(1), 49. U.S. Department of Health and Human Services (2010). Healthy
Munhall, P. A. (2010). Nursing research: A qualitative per- people 2020: The road ahead. Retrieved from http://www.
spective (5th ed.). Sudbury, MA: Jones and Bartlett. healthypeople.gov/2020/topicsobjectives2020/overview.
Pilcher, J. J., & Walters, A. S. (1997). How sleep deprivation aspx?topicid=38.
affects psychological variables related to college students’ Vranesh, J. G., Madrid, G., Bautista, J., Ching, P., & Hicks,
cognitive performance. Journal of American College R. (1999). Time perspective and sleep problem. Percep-
Health, 46(3), 121–126. tual and Motor Skills, 88(1), 23–24. doi:10.2466/pms.
Polit, D. F., & Beck, C. T. (2004). Nursing research: Principles 88.1.23-24.
and methods (7th ed.). Hagerstown, MD: Lippincott Wil- Wolf, F. M. (1986). Meta-analysis: Quantitative methods for
liams & Wilkins. research synthesis. Beverly Hills, CA: Sage.
Romas, J. A., & Sharma, M. (2010). Practical stress manage- Youngstedt, S. D., & Kripke, D. F. (2004). Long sleep and
ment. A comprehensive workbook for managing change mortality: Rationale for sleep restriction. Sleep Medicine
and promoting health (5th ed.). San Francisco, CA: Ben- Reviews, 8(3), 159–174. doi:10.1016/j.smrv.2003.10.002.
jamin Cummings. Zee, P. C., & Turek, F. W. (2006). Sleep and health: Everywhere
Sadeh, A., & Acebo, C. (2002). The role of actigraphy in sleep and in both directions. Archives of Internal Medicine,
medicine. Sleep Medicine Reviews, 6(2), 113–124. 166(16), 1686–1688. doi:10.1001/archinte.166.16.1686.
doi:10.1053/smrv.2001.0182.
Stevens, J. (1996). Applied multivariate statistics for the social
sciences (3rd ed.). Mahwah, NJ: Lawrence Erlbaum
Associates, Inc.

123