Research report

Nutrition self-efficacy is unidirectionally related to outcomeexpectations in children

Andrew L Larsen a John J McArdle a Trina Robertson b Genevieve F Dunton a

a University of Southern California 3620 S McClintock Ave SGM 501 Los Angeles CA 90089-1061 USAb Dairy Council of California 2151 Michelson Drive Suite 235 Irvine CA 92612-1339 USA

A R T I C L E I N F O

Article historyReceived 9 March 2014Received in revised form 15 September2014Accepted 7 October 2014Available online 18 October 2014

KeywordsSocial-cognitionAttitudesDietSchool-based researchHealth

A B S T R A C T

Objective To clarify the underlying relationship between nutrition self-efficacy and outcome expecta-tions because the direction of the relationship (unidirectional vs bidirectional) is debated in the literatureMethods Secondary data analysis of a 10-week 10-lesson school-based nutrition education interven-tion among 3rd grade students (N = 952) Nutrition self-efficacy (7 items) and nutrition outcomeexpectations (9 items) were measured through student self-report at intervention pre- (time 1) and post-(time 2) assessments A series of two time point multi-group cross-lagged bivariate change score modelswere used to determine the direction of the relationship Results A cross lag from nutrition self-efficacy at time 1 predicting changes in nutrition outcome expectations at time 2 significantly improvedthe fit of the model (Model 3) whereas a cross lag from nutrition outcome expectations at time 1 tochanges in nutrition self-efficacy at time 2 only slightly improved the fit of the model (Model 2) Fur-thermore adding both cross lags (Model 4) did not improve model fit compared to the model with onlythe self-efficacy cross lag (Model 3) Lastly the nutrition outcome expectations cross lag did not signifi-cantly predict changes in nutrition self-efficacy in any of the models Conclusions Data suggest that thereis a unidirectional relationship between nutrition self-efficacy and outcome expectations in which self-efficacy predicts outcome expectations Therefore theory-based nutrition interventions may considerfocusing more resources on changing self-efficacy because it may also lead to changes in outcome ex-pectations as well

copy 2014 Elsevier Ltd All rights reserved

Introduction

Self-efficacy and outcome expectations (and conceptually similarconstructs with alternate labels) are key components of several prom-inent models of self-regulation including Social Cognitive Theory (SCTBandura 1997 2004) the Health Action Process Approach (HAPASchwarzer 1992) the Theory of Planned Behavior (TPB Ajzen 1991)and Protection Motivation Theory (PMT Maddux amp Rogers 1983 Rogers1975) Self-efficacy is described as a personrsquos perceived competencyin a given domain (Bandura 1997) Outcome expectations are antici-pated outcomes that people expect their actions to produce such as

the belief that eating healthy food will make your body feel better(Bandura 1997) Outcome expectations and self-efficacy are explic-itly included in the SCT and HAPA models and similar constructs areincluded in the TPB and PMT such as response efficacy perceived be-havioral control and attitudes (Bandura 1997 Conner amp Norman 2005)Social cognitive theories of self-regulation hypothesize that people withhigh self-efficacy and high positive outcome expectations are more likelyto successfully self-regulate their behavior in a given domain (Bandura1997)

However the relationship between self-efficacy and outcome ex-pectations (as well as similar constructs from other models) isdebated (Fishbein et al 2000) SCT posits a unidirectional relation-ship from self-efficacy to outcome expectations HAPA posits abidirectional relationship and other models (eg PMT TPB) fail tospecify specific relationships (Bandura 1997 Conner amp Norman2005 Schwarzer 1992) The direction of the relationship is inte-gral for understanding processes of behavior change which the socialcognitive self-regulation literature has been criticized for failing todo in research (Leventhal amp Mora 2005) Additionally understand-ing processes of change helps to determine which constructs toprioritize in interventions particularly minimalist interventionsand interventions with limited resources If a unidirectionalrelationship exists in which self-efficacy predicts changes in

Abbreviations SCT Social Cognitive Theory HAPA Health Action Process Ap-proach TPB Theory of Planned Behavior PMT Protection Motivation Theory SEMstructural equation modeling df degrees of freedom dχ2 change in chi-square ddfchange in degrees of freedom CFI comparative fit index TLI TuckerndashLewis indexRMSEA root mean square error of approximation

Acknowledgements This work was supported by the Dairy Council of Califor-nia and the American Cancer Society (118283-MRSGT-10-012-01-CPPB) The studywas funded by the Dairy Council of California This manuscript is based on data ina previously published report (Dunton et al 2012)

Corresponding authorE-mail address allarsenuscedu (AL Larsen)

httpdxdoiorg101016jappet2014100130195-6663copy 2014 Elsevier Ltd All rights reserved

Appetite 84 (2015) 166ndash170

Contents lists available at ScienceDirect

Appetite

journal homepage wwwelseviercom locate appet

outcome expectations ndash as posited in SCT ndash then interventions mayprimarily target self-efficacy because changes in self-efficacy wouldpresumably lead to changes in outcome expectations However ifthere is a bidirectional relationship then interventions with limitedresources should target both constructs relatively equally

A weakness of past research has been an over-reliance onbetween-group analyses such as cross-sectional designs (Wein-stein 2007) Instead longitudinal within-subjects research is requiredfor testing processes of change ndash particularly for understanding tem-poral sequencing of construct effects (Rhodes amp Nigg 2011 CervoneShadel Smith amp Fiori 2006) ndash and is facilitated by modern statis-tical techniques The aim of the present research is to clarify theunderlying relationship between nutrition self-efficacy and outcomeexpectations by testing a series of longitudinal bivariate crossed-lagged latent change score models

Material and methods

The present work is a secondary analysis from a previously re-ported intervention (Dunton et al 2012) The data were collectedto evaluate a 10-week (10-lesson) school-based nutrition educa-tion program developed by the Dairy Council of California to teach3rd grade students the importance of healthy eating and physicalactivity behaviors and attitudes (see wwwHealthyEatingorgSMC) The regular classroom teachers all of whom received trainingprior to starting the intervention taught the lessons Lessons taught(1) the five food groups (2) the main nutrients and their roles inthe body (3) the importance of balanced meals (4) how to readfood labels (5) how to measure portion sizes (6) healthy bever-age choices and (7) how to be active for 60 minutes a day Self-efficacy was targeted through activities such as children practicingmaking balanced meals substituting healthy snacks for unhealthyones and by exposing students to a wide variety of foods to givethem confidence that they could find healthy foods to eat Outcomeexpectations were targeted through explaining benefits of healthyeating including the roles of the main nutrient groups impor-tance of balanced meals the importance of being physically activeand importance of reducing added sugar in their diet

Data were collected during the 2010ndash2011 school year The studyincluded pre- post- and follow-up assessments of nutrition knowl-edge outcome expectancies self-efficacy and dietary intake forIntervention and Control groups Only pre- (time 1) and post- (time2) assessments were included in the current analyses Post-assessments were completed immediately following the 10-weekprogram

A sample of 22 public elementary schools across California (con-sisting of 1147 students) who had either ordered the programmaterials from the Dairy Council of California in previous years orhad been recommended by district level contacts were recruitedto participate in the study Schools were selected so that the sampleresembled the state-level demographic profile of 3rd grade stu-dents attending public schools in California Within each school twothird grade classrooms were randomly selected for participation inthe evaluation Classrooms were considered eligible if they were nota combination grade classroom and did not teach other nutritioninformation as part of the regular classroom curriculum If a teacherdid not agree to be in the evaluation an additional classroom fromthat school was chosen as a replacement Within each school oneclassroom was randomly assigned to be in the Intervention groupand one in the Control group Six teachers did not agree to randomassignment and were placed in the Intervention group The Controlgroup completed the assessments only (and as part of eligibility re-quirements did not teach any other nutrition information) Studyprocedures were approved by the ethics committee of the Institu-tional Review Board at Independent Review Consulting Inc Writteninformed assent was obtained from students and a passive paren-

tal consent procedure was used (ie if the parent did not declineconsent the student was approached for consent)

Nutrition self-efficacy and outcome expectations were mea-sured using scales developed to gauge specific components of theintervention because few measures of nutrition self-efficacy andoutcome expectations exist for children The scales were based onpreviously validated scales typically used on different populationsand contexts (Hagler Norman Radick Calfas amp Sallis 2005 NationalCancer Institute 2005) Highly trained research staff used a stan-dardized scripted procedure to administer the questionnaires Self-efficacy was measured with seven items on a four item responsescale (Cronbachrsquos α = 071) with anchors I can and I canrsquot (eg ldquoIcan eat breakfast every day even if I am in a hurryrdquo ldquoI can chooseto drink water or milk instead of soda at restaurantsrdquo ldquoEvery dayI can eat foods that are low in sugar for meals and snacksrdquo) Outcomeexpectations were measured with nine items on a four item re-sponse scale (Cronbachrsquos α = 073) with anchors Yes and No (egldquoI think that I will feel healthier if I eat fewer sweetsrdquo ldquoI think thatskipping a meal will make me feel tiredrdquo ldquoI think that I will buildstrong muscles if I eat more meatrdquo) Both scales were pilot testedin 3rd graders (n = 57) and were determined to have adequate 7d testndashretest reliability with an r = 75 for nutrition outcome ex-pectations and r = 57 for nutrition self-efficacy (see Dunton et al2012) Additionally factor analyses of the self-efficacy and outcomeexpectations scales revealed that all items loaded gt40 onto a singlefactor for each scale Although a formative test of validity was notconducted predictive validity was evaluated by testing correla-tions between the self-efficacy and outcome expectation scales withother measured constructs within the present sample Both scaleswere positively correlated with nutrition knowledge (eg knowl-edge of food groups main nutrients types of foods rrsquos = 056 minus 374prsquos = 104 minus lt 001) positively correlated with a number of healthyfoods children reported consuming (eg fruits vegetablesrrsquos = 070 minus 185 prsquos = 033 minus lt 001) and negatively correlated witha number of unhealthy foods children reported consuming (eg sodajunk foods rrsquos ranged from minus153 to minus264 prsquos = lt 001) as mea-sured through the School Physical Activity and Nutrition (SPAN) 24-hour recall instrument (data not shown Hoelscher Day Kelder ampWard 2003 Thiagarajah et al 2008)

Statistical analysis

The data analyses were conducted in R version 2152 usingthe Lavaan package The goal of the analysis was to test the direc-tionality of the relationship between nutrition self-efficacy andoutcome expectations by comparing model fit of a series of fourmulti-group cross-lagged bivariate change score models (Fig 1) InStructural Equation Modeling (SEM) longitudinal latent changescores can be created by setting the regression path between Time1 and Time 2 equal to 1 implying that some portion of the Time 2score is exactly equal to Time 1 and the residual variable (eg DSEand DOE in Fig 1) is directly interpretable as a difference score(McArdle amp Nesselroade 1994) An arrow from Time 1 to the dif-ference score represents an association between the previous timepoint and any changes over time (eg being high at Time 1 couldbe associated with a decrease at Time 2)

Means and variances at the first time point were held invariantacross group (Intervention vs Control) to control for group differ-ences at the start of the study Based on significant differences inself-efficacy and outcomes expectancies observed between the In-tervention and Control groups (see Dunton et al 2012) allparameters were allowed to vary across study groups (ie meanchanges variance of changes covariance between changes and re-gression coefficients) The first model had no cross-lags the secondmodel contained one cross-lag from outcome expectations at time1 to changes in self-efficacy at time 2 (referred to as the outcome

167AL Larsen et alAppetite 84 (2015) 166ndash170

expectations cross lag) the third model contained one cross-lag fromself-efficacy at time 1 to changes in outcome expectations at time2 (referred to as the self-efficacy cross lag) and the last model con-tained both cross-lags

It is possible to determine the direction of the relationshipbetween self-efficacy and outcome expectations based on changesin model fit If each cross-lag individually improves the fit of themodel then we can conclude a bidirectional relationship If only onecross-lag improves the fit of the model then we can conclude a uni-directional relationship (Ferrer amp McArdle 2003)

Model fit was determined using the minimum function chi-square (χ2) degrees of freedom (df) change in chi-square (dχ2)

change in degrees of freedom (ddf) the Comparative Fit Index (CFI)the TuckerndashLewis Index (TLI) and the Root Mean Square Error ofApproximation (RMSEA) Conventional standards for good fit arevalues above 095 for the CFI and TLI values below 006 for theRMSEA and a failure to reject the minimum function χ2 test(Hu amp Bentler 1999) Missing data were handled using the FullInformation Maximum Likelihood function in the Lavaan packageof R an estimation procedure that uses all available data to mini-mize bias resulting from missing data

Results

A total of 1147 students completed pre-surveys 1019 studentscompleted post-surveys and 952 (females = 496 Interventiongroup = 520) students were able to be used in analysis Differ-ences in the number of student surveys at each time point are dueto absent children and students transferring to different schools Thesample had a mean age of 796 years (SD = 045) and had a higherproportion of Asian students (19 vs 9) a lower proportion of His-panic students (35 vs 51) and similar proportions of WhiteCaucasian (23 vs 27) and African American students (9 vs 8)as compared with 3rd graders attending public elementary schoolsacross the state of California

Independent samples t-tests revealed no difference in self-efficacy or outcome expectations between the Intervention andControl groups at baseline (p gt 05) The Intervention group hadmeans of 2314 (highest possible score = 28 SD = 384) for self-efficacy and 2882 (highest possible score = 36 SD = 388) for outcomeexpectations at pre-intervention while the Control group had meansof 2288 (highest possible score = 28 SD = 393) for self-efficacy and2885 (highest possible score = 36 SD = 380) for outcome expec-tations The Intervention group observed significant increases frompre- to post-intervention in both nutrition self-efficacy (mean changeof 0728 [SD = 378] vs minus0398 [SD = 399]) and outcome expecta-tions (mean change of 252 [SD = 397] vs 0302 [SD = 421])compared to the Control group (p lt 05 see Dunton et al 2012)

The parameter estimates and fit indices are presented inTable 1

The first model tested (Model 1) is the baseline model forcomparison to other models because it contained no cross-lags

Fig 1 Path diagram for the bivariate change model for self-efficacy and outcomeexpectations across two time points (Model 4) Observed variables are drawn assquares and unobserved (latent) variables are drawn as circles All model param-eters are shown as two-headed or one-headed arrows Two-headed arrows with bothheads pointing to the same variable represent variances A fixed variable set at 1and drawn as a triangle is included to represent mean scores Arrows deriving fromthis fixed variable represent group means SE 0 = Self-efficacy at baseline SE 1 = Self-efficacy post-intervention DSE = change in self-efficacy OE 0 = outcome expectationat baseline OE 1 = outcome expectations post-intervention DOE = change in outcomeexpectations

Table 1Numerical results for test of invariance in bivariate change score model of baseline self-efficacy predicting changes in outcome expectations

Model 1No cross lags

Model 2OE cross lag only

Model 3SE cross lag only

Model 4Both cross lags

Control Int Control Int Control Int Control Int

Parameters M S M S M S M S M S M S M S M S

Fixed parametersMean SE0 2304 596 = = 2304 596 = = 2303 597 = = 2303 597 = =Mean DSE 1219 291 1348 353 999 238 1180 309 1090 268 1218 329 916 225 1090 294SE0rarrDSE minus055 minus051 minus055 minus056 minus057 minus053 minus057 minus058 minus049 minus047 minus049 minus051 minus051 minus049 minus051 minus054OE0rarrDSE X X X X 010 009 008 008 X X X X 008 007 006 006

Mean OE0 2880 745 = = 2880 745 = = 2881 746 = = 2880 746 = =Mean DOE 1632 376 1849 461 1565 364 1782 450 1355 311 1475 368 1309 303 1433 362OE0rarrDOE minus056 minus049 minus056 minus054 minus053 minus048 minus053 minus052 minus062 minus055 minus062 minus060 minus061 minus054 minus061 minus059SE0rarrDOE X X X X X X X X 021 018 025 024 020 018 024 023DSEharrDOE 352 026 350 033 347 026 352 033 321 024 342 033 313 024 339 033Goodness of fit indicesχ2dfp value

579912p lt 001

510710p lt 001

95510p = 48

5328p = 72

dχ2ddf 00 6922 48442 52674CFITLI 093093 093092 100100 100101RMSEA 009 009 000 000

Note Paired equal signs indicate group invariance for that parameter X = parameter not estimated M = maximum likelihood estimate S = standardized estimates Int = in-tervention group SE0 = self-efficacy pre measure DSE = change in self-efficacy OE0 = outcome expectations pre measure DOE = change in outcome expectationsap gt 05

168 AL Larsen et alAppetite 84 (2015) 166ndash170

This model showed borderline good fit to the data with a χ2

(df = 12) = 5799 CFI = 093 TLI = 093 RMSEA = 009 Howeverthe minimum function χ2 test of goodness of fit rejected the nullhypothesis p lt 001 failing to support the notion of good fit to thedata

Model 2 included the outcome expectations cross lag onlyThis model showed borderline good fit to the data with a χ2

(df = 10) = 5107 CFI = 093 TLI = 092 RMSEA = 009 Howeverthe minimum function χ2 test of goodness of fit for Model 2 re-jected the null hypothesis p lt 001 failing to support the notion ofgood fit to the data Model 2 improved the fit of the data com-pared to the baseline model by dχ2 = 692 on ddf = 2 which is a smallbut significant improvement in fit (p = 03) Lastly the outcome ex-pectations cross lag was not significant for the Control group (β = 09p = 053) or the Intervention group (β = 08 p = 051)

Model 3 included the self-efficacy cross lag only This modelshowed very good fit to the data with a χ2 (df = 10) = 955 CFI = 100TLI = 100 RMSEA = 000 The minimum function χ2 test of good-ness of fit for Model 3 failed to reject the null hypothesis furthersupporting the notion of a good fit to the data (p = 48) Model 3 alsosignificantly improved the fit of the data compared to the base-line model by dχ2 = 4844 on ddf = 2 p lt 001 Model 2 and Model3 are not directly comparable because they share the same numberof df Finally the self-efficacy cross lag was significant for the Controlgroup (β = 21 p lt 001) and the Intervention group (β = 18 p lt 001)

Model 4 included both cross lags This model also showed verygood fit to the data with a χ2 (df = 8) = 532 CFI = 100 TLI = 101RMSEA = 000 The minimum function χ2 test of goodness of fit forModel 4 failed to reject the null hypothesis of a good fit to the datafurther supporting the notion of a good fit to the data (p = 72) Model4 significantly improved fit compared to the baseline model bydχ2 = 5267 on ddf = 2 p lt 001 However Model 4 did not improvethe fit of the data compared to Model 3 (the self-efficacy cross lagmodel) with a dχ2 = 423 on ddf = 2 p = 12 Finally the self-efficacy cross lag was significant for the Control group (β = 18p lt 001) and the Intervention group (β = 23 p lt 001) while theoutcome expectations cross lag was not significant for the Controlgroup (β = 07 p = 13) or the Intervention group (β = 06 p = 13)

Discussion

The present research provides strong support for a unidirec-tional relationship between nutrition self-efficacy and nutritionoutcome expectations in children in which changes in self-efficacyprecede changes in outcome expectations in both intervention andnon-intervention contexts Through a series of four multi-groupcross-lagged bivariate change score models the present researchfound that (a) the outcome expectations cross lag (Model 2) slightlyimproved model fit compared to a no cross lag model (Model 1)(b) the self-efficacy cross lag (Model 3) greatly improved model fitcompared to a no cross lag model (Model 1) and overall showedan extremely good fit to the data and (c) adding both cross lags(Model 4) failed to improve model fit compared to the self-efficacycross lag alone (Model 3) Additionally none of the outcome ex-pectations cross-lags in any of the models significantly predictedchanges in self-efficacy for either the Control or Interventiongroups Therefore self-efficacy predicted changes in outcome ex-pectations but outcome expectations did not predict changes inself-efficacy

To date the majority of research in this area has focused on per-ceived self-efficacy alone rather than the relationship between self-efficacy and outcome expectations (Rhodes amp Nigg 2011) Howeverempirical research consistently shows a strong association betweenthe two constructs (Plotnikoff Lippke Courneya Birkett amp Sigal2008 Ralf Schwarzer et al 2007 White Woacutejcicki amp McAuley 2012)There are also known moderators of the self-efficacy and outcome

expectations relationship For instance perceived ability to controlan outcome (ie high locus of control) has been shown to strength-en the relationship (Bandura 1997 Urbig amp Monsen 2012) Extremescores for outcome expectations can also moderate the relation-ship between self-efficacy and behavior (Rhodes amp Nigg 2011)in that someone can have high self-efficacy but may not performthe behavior if they have extremely low or negative outcomeexpectations

Similar analytic strategies using different methodologies havetested the relationship between self-efficacy and behavior (GilsonChow amp Feltz 2012 Seo amp Ilies 2009) but the authors are not awareof a study evaluating the temporal sequence of the relationshipbetween self-efficacy and outcome expectations particularly in anapplied setting like the present study The tests of self-efficacy andbehavior serve the purpose of validating the legitimacy of self-efficacy as a causal construct whereas the present research servesto update theory and inform the design of interventions

Only SCT specifically hypothesizes a unidirectional relation-ship in which self-efficacy leads to changes in outcome expectations(Bandura 1997 2004) SCT views self-efficacy and outcome expec-tations as a feed-forward loop (Bandura 1997 Maes amp Karoly 2005)Outcome expectations represent potential future results of per-forming a behavior while self-efficacy represents the likelihood theperson is able to perform the behavior (Bandura 1997) SCT con-tends that a personrsquos expected results (ie outcome expectations)depend on their perceived ability to perform the required action(ie self-efficacy Bandura 1997 2004) For example if a persondoesnrsquot feel capable of performing a behavior well enough he wouldnot expect that attempting to perform the behavior would bring val-uable gains On the other hand SCT posits that having high outcomeexpectations for a behavior while motivating in its own right wouldnot affect the personrsquos perceived capability of performing that be-havior (Bandura 1997 2004)

There are several limitations to this study First the study par-ticipants are elementary school students and may not generalizewell to other age groups or populations This study also focused solelyon self-efficacy and outcome expectations related to nutrition andmay not generalize to other contexts or behaviors The self-efficacy and outcome expectations measures were not previouslyvalidated An extensive search of the literature was conducted priorto creating the scales that were used and at the time of design(2008ndash2009) no standard scales were available for this age groupand study context In lieu of formative validity testing various formsof reliability (testndashretest Chronbachrsquos alpha factor analysis)and predictive validity were tested The study was also not truly arandomized controlled design as six teachers self-selected into theIntervention group prior to randomization However the pattern ofresults were identical for both study groups Strengths of the studywere the large sample size (N = 952) and the pattern of findingsindicating a highly consistent unidirectional relationship

Conclusion

Interventions utilizing theory have been shown to have stron-ger effects than those that do not (Dombrowski et al 2012 TaylorConner amp Lawton 2012) Theory enhances behavior change inter-ventions by explaining relationships between constructs howconstructs are supposed to change over time and which con-structs should be targeted for intervention (Michie amp Prestwich2010) Results from the current study suggest that there is a uni-directional relationship between self-efficacy and outcomeexpectations in which self-efficacy predicts changes in outcomeexpectations It is important to note that these findings donot invalidate any of the common social cognitive models ofself-regulation None of the mentioned models hypothesize thatoutcome expectations are the driving force behind a personrsquos self-

169AL Larsen et alAppetite 84 (2015) 166ndash170

efficacy Instead the current findings help to update and extendsocial-cognitive theories of self-regulation given the ambiguity ofthe interrelationships of self-efficacy and outcome expectations inthe literature These findings indicate that minimalist interven-tions and interventions with limited resources may primarily targetthe improvement of self-efficacy given that improvements inoutcome expectations were shown to follow improvements in self-efficacy This does not mean however that outcome expectationsare unimportant or that interventions should ignore them alto-gether because extremely low outcome expectations can bedemotivating even for people with high self-efficacy (Rhodes amp Nigg2011)

References

Ajzen I (1991) The theory of planned behavior Organizational Behavior and HumanDecision Processes 50(2) 179ndash211

Bandura A (1997) Self-efficacy The exercise of control New York Worth PublishersBandura A (2004) Health promotion by social cognitive means Health Education

amp Behavior 31(2) 143ndash164Cervone D Shadel W G Smith R E amp Fiori M (2006) Self-Regulation Remin-

ders and Suggestions from Personality Science Applied Psychology 55(3) 333ndash385

Conner M amp Norman P (2005) Predicitng health behavior A social cognitionapproach In M Conner amp P Norman (Eds) Predicting health behavior (2nd edpp 1ndash27) New York Open University Press

Dombrowski S U Sniehotta F F Avenell A Johnston M MacLennan G ampArauacutejo-Soares V (2012) Identifying active ingredients in complex behaviouralinterventions for obese adults with obesity-related co-morbidities or additionalrisk factors for co-morbidities A systematic review Health Psychology Review6(1) 7ndash32

Dunton G F Liao Y Grana R Lagloire R Riggs N Chou C P et al (2012)State-wide dissemination of a school-based nutrition education programme ARE-AIM (Reach Efficacy Adoption Implementation Maintenance) analysis PublicHealth Nutrition 17 422ndash439

Ferrer E amp McArdle J (2003) Alternative structural models for multivariatelongitudinal data analysis Structural Equation Modeling 10 493ndash524

Fishbein M Triandis H C Kanfer F H Becker M Middlestadt S E amp Eichler A(2000) Factors influencing behavior and behavior change In A S Baum T ARevenson amp J E Singer (Eds) Handbook of health psychology (pp 1ndash17) MahwahNJ Lawrence Erlbaum

Gilson T A Chow G M amp Feltz D L (2012) Self-efficacy and athletic squatperformance Positive or negative influences at the within-and between-levelsof analysis Journal of Applied Social Psychology 42(6) 1467ndash1485

Hagler A S Norman G J Radick L R Calfas K J amp Sallis J F (2005) Comparabilityand reliability of paper-and computer-based measures of psychosocial constructsfor adolescent fruit and vegetable and dietary fat intake Journal of the AmericanDietetic Association 105(11) 1758ndash1764

Hoelscher D M Day R S Kelder S H amp Ward J L (2003) Reproducibility andvalidity of the secondary level School-Based Nutrition Monitoring studentquestionnaire Journal of the American Dietetic Association 103(2) 186ndash194

Hu L amp Bentler P M (1999) Cutoff criteria for fit indexes in covariance structureanalysis Conventional criteria versus new alternatives Structural EquationModeling A Multidisciplinary Journal 6(1) 1ndash55

Leventhal H amp Mora P A (2005) Is there a science of the processes underlyinghealth and illness behaviors A comment on Maes and Karoly Applied Psychology54(2) 255ndash266

Maddux J E amp Rogers R W (1983) Protection motivation and self-efficacy A revisedtheory of fear appeals and attitude change Journal of Experimental SocialPsychology 19(5) 469ndash479

Maes S amp Karoly P (2005) Self-Regulation Assessment and Intervention in PhysicalHealth and Illness A Review Applied Psychology 54(2) 267ndash299

McArdle J J amp Nesselroade J R (1994) Using multivariate data to structuredevelopmental change In S H Cohen amp H W Reese (Eds) Life-span developmentalpsychology Methodological contributions (pp 223ndash267) Hillsdale NJ LawrenceErlbaum

Michie S amp Prestwich A (2010) Are interventions theory-based Development ofa theory coding scheme Health Psychology 29(1) 1ndash8

National Cancer Institute (2005) Food attitudes and behaviors (FAB) survey HealthBehaviors Research Branch Retrieved from lthttpcancercontrolcancergovbrpfabindexhtmlgt

Plotnikoff R C Lippke S Courneya K S Birkett N amp Sigal R J (2008) Physicalactivity and social cognitive theory A test in a population sample of adults withtype 1 or type 2 diabetes Applied Psychology 57(4) 628ndash643

Rhodes R E amp Nigg C R (2011) Advancing physical activity theory A review andfuture directions Exercise and Sport Sciences Reviews 39(3) 113ndash119

Rogers R W (1975) A protection motivation theory of fear appeals and attitudechange The Journal of Psychology 91(1) 93ndash114

Schwarzer R (1992) Self-efficacy in the adoption and maintenance of health behaviorsTheoretical approaches and a new model Washington DC Hemisphere PublishingCorp

Schwarzer R Schuumlz B Ziegelmann J P Lippke S Luszczynska A amp Scholz U(2007) Adoption and maintenance of four health behaviors Theory-guidedlongitudinal studies on dental flossing seat belt use dietary behavior and physicalactivity Annals of Behavioral Medicine 33(2) 156ndash166

Seo M amp Ilies R (2009) The role of self-efficacy goal and affect in dynamicmotivational self-regulation Organizational Behavior and Human Decision Processes109(2) 120ndash133

Taylor N Conner M amp Lawton R (2012) The impact of theory on the effectivenessof worksite physical activity interventions A meta-analysis and meta-regressionHealth Psychology Review 6(1) 33ndash73

Thiagarajah K Fly A D Hoelscher D M Bai Y Lo K Leone A et al (2008)Validating the food behavior questions from the elementary school SPANquestionnaire Journal of Nutrition Education and Behavior 40(5) 305ndash310

Urbig D amp Monsen E (2012) The structure of optimism ldquoControllability affectsthe extent to which efficacy beliefs shape outcome expectanciesrdquo Journal ofEconomic Psychology 33(4) 854ndash867

Weinstein N D (2007) Misleading tests of health behavior theories Annals ofBehavioral Medicine 33(1) 1ndash10

White S M Woacutejcicki T R amp McAuley E (2012) Social cognitive influences onphysical activity behavior in middle-aged and older adults The Journals ofGerontology Series B Psychological Sciences and Social Sciences 67 P18ndashP26

170 AL Larsen et alAppetite 84 (2015) 166ndash170

This model showed borderline good fit to the data with a χ2

(df = 12) = 5799 CFI = 093 TLI = 093 RMSEA = 009 Howeverthe minimum function χ2 test of goodness of fit rejected the nullhypothesis p lt 001 failing to support the notion of good fit to thedata

Model 2 included the outcome expectations cross lag onlyThis model showed borderline good fit to the data with a χ2

(df = 10) = 5107 CFI = 093 TLI = 092 RMSEA = 009 Howeverthe minimum function χ2 test of goodness of fit for Model 2 re-jected the null hypothesis p lt 001 failing to support the notion ofgood fit to the data Model 2 improved the fit of the data com-pared to the baseline model by dχ2 = 692 on ddf = 2 which is a smallbut significant improvement in fit (p = 03) Lastly the outcome ex-pectations cross lag was not significant for the Control group (β = 09p = 053) or the Intervention group (β = 08 p = 051)

Model 3 included the self-efficacy cross lag only This modelshowed very good fit to the data with a χ2 (df = 10) = 955 CFI = 100TLI = 100 RMSEA = 000 The minimum function χ2 test of good-ness of fit for Model 3 failed to reject the null hypothesis furthersupporting the notion of a good fit to the data (p = 48) Model 3 alsosignificantly improved the fit of the data compared to the base-line model by dχ2 = 4844 on ddf = 2 p lt 001 Model 2 and Model3 are not directly comparable because they share the same numberof df Finally the self-efficacy cross lag was significant for the Controlgroup (β = 21 p lt 001) and the Intervention group (β = 18 p lt 001)

Model 4 included both cross lags This model also showed verygood fit to the data with a χ2 (df = 8) = 532 CFI = 100 TLI = 101RMSEA = 000 The minimum function χ2 test of goodness of fit forModel 4 failed to reject the null hypothesis of a good fit to the datafurther supporting the notion of a good fit to the data (p = 72) Model4 significantly improved fit compared to the baseline model bydχ2 = 5267 on ddf = 2 p lt 001 However Model 4 did not improvethe fit of the data compared to Model 3 (the self-efficacy cross lagmodel) with a dχ2 = 423 on ddf = 2 p = 12 Finally the self-efficacy cross lag was significant for the Control group (β = 18p lt 001) and the Intervention group (β = 23 p lt 001) while theoutcome expectations cross lag was not significant for the Controlgroup (β = 07 p = 13) or the Intervention group (β = 06 p = 13)

Discussion

The present research provides strong support for a unidirec-tional relationship between nutrition self-efficacy and nutritionoutcome expectations in children in which changes in self-efficacyprecede changes in outcome expectations in both intervention andnon-intervention contexts Through a series of four multi-groupcross-lagged bivariate change score models the present researchfound that (a) the outcome expectations cross lag (Model 2) slightlyimproved model fit compared to a no cross lag model (Model 1)(b) the self-efficacy cross lag (Model 3) greatly improved model fitcompared to a no cross lag model (Model 1) and overall showedan extremely good fit to the data and (c) adding both cross lags(Model 4) failed to improve model fit compared to the self-efficacycross lag alone (Model 3) Additionally none of the outcome ex-pectations cross-lags in any of the models significantly predictedchanges in self-efficacy for either the Control or Interventiongroups Therefore self-efficacy predicted changes in outcome ex-pectations but outcome expectations did not predict changes inself-efficacy

To date the majority of research in this area has focused on per-ceived self-efficacy alone rather than the relationship between self-efficacy and outcome expectations (Rhodes amp Nigg 2011) Howeverempirical research consistently shows a strong association betweenthe two constructs (Plotnikoff Lippke Courneya Birkett amp Sigal2008 Ralf Schwarzer et al 2007 White Woacutejcicki amp McAuley 2012)There are also known moderators of the self-efficacy and outcome

expectations relationship For instance perceived ability to controlan outcome (ie high locus of control) has been shown to strength-en the relationship (Bandura 1997 Urbig amp Monsen 2012) Extremescores for outcome expectations can also moderate the relation-ship between self-efficacy and behavior (Rhodes amp Nigg 2011)in that someone can have high self-efficacy but may not performthe behavior if they have extremely low or negative outcomeexpectations

Similar analytic strategies using different methodologies havetested the relationship between self-efficacy and behavior (GilsonChow amp Feltz 2012 Seo amp Ilies 2009) but the authors are not awareof a study evaluating the temporal sequence of the relationshipbetween self-efficacy and outcome expectations particularly in anapplied setting like the present study The tests of self-efficacy andbehavior serve the purpose of validating the legitimacy of self-efficacy as a causal construct whereas the present research servesto update theory and inform the design of interventions

Only SCT specifically hypothesizes a unidirectional relation-ship in which self-efficacy leads to changes in outcome expectations(Bandura 1997 2004) SCT views self-efficacy and outcome expec-tations as a feed-forward loop (Bandura 1997 Maes amp Karoly 2005)Outcome expectations represent potential future results of per-forming a behavior while self-efficacy represents the likelihood theperson is able to perform the behavior (Bandura 1997) SCT con-tends that a personrsquos expected results (ie outcome expectations)depend on their perceived ability to perform the required action(ie self-efficacy Bandura 1997 2004) For example if a persondoesnrsquot feel capable of performing a behavior well enough he wouldnot expect that attempting to perform the behavior would bring val-uable gains On the other hand SCT posits that having high outcomeexpectations for a behavior while motivating in its own right wouldnot affect the personrsquos perceived capability of performing that be-havior (Bandura 1997 2004)

There are several limitations to this study First the study par-ticipants are elementary school students and may not generalizewell to other age groups or populations This study also focused solelyon self-efficacy and outcome expectations related to nutrition andmay not generalize to other contexts or behaviors The self-efficacy and outcome expectations measures were not previouslyvalidated An extensive search of the literature was conducted priorto creating the scales that were used and at the time of design(2008ndash2009) no standard scales were available for this age groupand study context In lieu of formative validity testing various formsof reliability (testndashretest Chronbachrsquos alpha factor analysis)and predictive validity were tested The study was also not truly arandomized controlled design as six teachers self-selected into theIntervention group prior to randomization However the pattern ofresults were identical for both study groups Strengths of the studywere the large sample size (N = 952) and the pattern of findingsindicating a highly consistent unidirectional relationship

Conclusion

Interventions utilizing theory have been shown to have stron-ger effects than those that do not (Dombrowski et al 2012 TaylorConner amp Lawton 2012) Theory enhances behavior change inter-ventions by explaining relationships between constructs howconstructs are supposed to change over time and which con-structs should be targeted for intervention (Michie amp Prestwich2010) Results from the current study suggest that there is a uni-directional relationship between self-efficacy and outcomeexpectations in which self-efficacy predicts changes in outcomeexpectations It is important to note that these findings donot invalidate any of the common social cognitive models ofself-regulation None of the mentioned models hypothesize thatoutcome expectations are the driving force behind a personrsquos self-

169AL Larsen et alAppetite 84 (2015) 166ndash170

efficacy Instead the current findings help to update and extendsocial-cognitive theories of self-regulation given the ambiguity ofthe interrelationships of self-efficacy and outcome expectations inthe literature These findings indicate that minimalist interven-tions and interventions with limited resources may primarily targetthe improvement of self-efficacy given that improvements inoutcome expectations were shown to follow improvements in self-efficacy This does not mean however that outcome expectationsare unimportant or that interventions should ignore them alto-gether because extremely low outcome expectations can bedemotivating even for people with high self-efficacy (Rhodes amp Nigg2011)

References

Ajzen I (1991) The theory of planned behavior Organizational Behavior and HumanDecision Processes 50(2) 179ndash211

Bandura A (1997) Self-efficacy The exercise of control New York Worth PublishersBandura A (2004) Health promotion by social cognitive means Health Education

amp Behavior 31(2) 143ndash164Cervone D Shadel W G Smith R E amp Fiori M (2006) Self-Regulation Remin-

ders and Suggestions from Personality Science Applied Psychology 55(3) 333ndash385

Conner M amp Norman P (2005) Predicitng health behavior A social cognitionapproach In M Conner amp P Norman (Eds) Predicting health behavior (2nd edpp 1ndash27) New York Open University Press

Dombrowski S U Sniehotta F F Avenell A Johnston M MacLennan G ampArauacutejo-Soares V (2012) Identifying active ingredients in complex behaviouralinterventions for obese adults with obesity-related co-morbidities or additionalrisk factors for co-morbidities A systematic review Health Psychology Review6(1) 7ndash32

Dunton G F Liao Y Grana R Lagloire R Riggs N Chou C P et al (2012)State-wide dissemination of a school-based nutrition education programme ARE-AIM (Reach Efficacy Adoption Implementation Maintenance) analysis PublicHealth Nutrition 17 422ndash439

Ferrer E amp McArdle J (2003) Alternative structural models for multivariatelongitudinal data analysis Structural Equation Modeling 10 493ndash524

Fishbein M Triandis H C Kanfer F H Becker M Middlestadt S E amp Eichler A(2000) Factors influencing behavior and behavior change In A S Baum T ARevenson amp J E Singer (Eds) Handbook of health psychology (pp 1ndash17) MahwahNJ Lawrence Erlbaum

Gilson T A Chow G M amp Feltz D L (2012) Self-efficacy and athletic squatperformance Positive or negative influences at the within-and between-levelsof analysis Journal of Applied Social Psychology 42(6) 1467ndash1485

Hagler A S Norman G J Radick L R Calfas K J amp Sallis J F (2005) Comparabilityand reliability of paper-and computer-based measures of psychosocial constructsfor adolescent fruit and vegetable and dietary fat intake Journal of the AmericanDietetic Association 105(11) 1758ndash1764

Hoelscher D M Day R S Kelder S H amp Ward J L (2003) Reproducibility andvalidity of the secondary level School-Based Nutrition Monitoring studentquestionnaire Journal of the American Dietetic Association 103(2) 186ndash194

Hu L amp Bentler P M (1999) Cutoff criteria for fit indexes in covariance structureanalysis Conventional criteria versus new alternatives Structural EquationModeling A Multidisciplinary Journal 6(1) 1ndash55

Leventhal H amp Mora P A (2005) Is there a science of the processes underlyinghealth and illness behaviors A comment on Maes and Karoly Applied Psychology54(2) 255ndash266

Maddux J E amp Rogers R W (1983) Protection motivation and self-efficacy A revisedtheory of fear appeals and attitude change Journal of Experimental SocialPsychology 19(5) 469ndash479

Maes S amp Karoly P (2005) Self-Regulation Assessment and Intervention in PhysicalHealth and Illness A Review Applied Psychology 54(2) 267ndash299

McArdle J J amp Nesselroade J R (1994) Using multivariate data to structuredevelopmental change In S H Cohen amp H W Reese (Eds) Life-span developmentalpsychology Methodological contributions (pp 223ndash267) Hillsdale NJ LawrenceErlbaum

Michie S amp Prestwich A (2010) Are interventions theory-based Development ofa theory coding scheme Health Psychology 29(1) 1ndash8

National Cancer Institute (2005) Food attitudes and behaviors (FAB) survey HealthBehaviors Research Branch Retrieved from lthttpcancercontrolcancergovbrpfabindexhtmlgt

Plotnikoff R C Lippke S Courneya K S Birkett N amp Sigal R J (2008) Physicalactivity and social cognitive theory A test in a population sample of adults withtype 1 or type 2 diabetes Applied Psychology 57(4) 628ndash643

Rhodes R E amp Nigg C R (2011) Advancing physical activity theory A review andfuture directions Exercise and Sport Sciences Reviews 39(3) 113ndash119

Rogers R W (1975) A protection motivation theory of fear appeals and attitudechange The Journal of Psychology 91(1) 93ndash114

Schwarzer R (1992) Self-efficacy in the adoption and maintenance of health behaviorsTheoretical approaches and a new model Washington DC Hemisphere PublishingCorp

Schwarzer R Schuumlz B Ziegelmann J P Lippke S Luszczynska A amp Scholz U(2007) Adoption and maintenance of four health behaviors Theory-guidedlongitudinal studies on dental flossing seat belt use dietary behavior and physicalactivity Annals of Behavioral Medicine 33(2) 156ndash166

Seo M amp Ilies R (2009) The role of self-efficacy goal and affect in dynamicmotivational self-regulation Organizational Behavior and Human Decision Processes109(2) 120ndash133

Taylor N Conner M amp Lawton R (2012) The impact of theory on the effectivenessof worksite physical activity interventions A meta-analysis and meta-regressionHealth Psychology Review 6(1) 33ndash73

Thiagarajah K Fly A D Hoelscher D M Bai Y Lo K Leone A et al (2008)Validating the food behavior questions from the elementary school SPANquestionnaire Journal of Nutrition Education and Behavior 40(5) 305ndash310

Urbig D amp Monsen E (2012) The structure of optimism ldquoControllability affectsthe extent to which efficacy beliefs shape outcome expectanciesrdquo Journal ofEconomic Psychology 33(4) 854ndash867

Weinstein N D (2007) Misleading tests of health behavior theories Annals ofBehavioral Medicine 33(1) 1ndash10

White S M Woacutejcicki T R amp McAuley E (2012) Social cognitive influences onphysical activity behavior in middle-aged and older adults The Journals ofGerontology Series B Psychological Sciences and Social Sciences 67 P18ndashP26

170 AL Larsen et alAppetite 84 (2015) 166ndash170

efficacy Instead the current findings help to update and extendsocial-cognitive theories of self-regulation given the ambiguity ofthe interrelationships of self-efficacy and outcome expectations inthe literature These findings indicate that minimalist interven-tions and interventions with limited resources may primarily targetthe improvement of self-efficacy given that improvements inoutcome expectations were shown to follow improvements in self-efficacy This does not mean however that outcome expectationsare unimportant or that interventions should ignore them alto-gether because extremely low outcome expectations can bedemotivating even for people with high self-efficacy (Rhodes amp Nigg2011)

References

Ajzen I (1991) The theory of planned behavior Organizational Behavior and HumanDecision Processes 50(2) 179ndash211

Bandura A (1997) Self-efficacy The exercise of control New York Worth PublishersBandura A (2004) Health promotion by social cognitive means Health Education

amp Behavior 31(2) 143ndash164Cervone D Shadel W G Smith R E amp Fiori M (2006) Self-Regulation Remin-

ders and Suggestions from Personality Science Applied Psychology 55(3) 333ndash385

Conner M amp Norman P (2005) Predicitng health behavior A social cognitionapproach In M Conner amp P Norman (Eds) Predicting health behavior (2nd edpp 1ndash27) New York Open University Press

Dombrowski S U Sniehotta F F Avenell A Johnston M MacLennan G ampArauacutejo-Soares V (2012) Identifying active ingredients in complex behaviouralinterventions for obese adults with obesity-related co-morbidities or additionalrisk factors for co-morbidities A systematic review Health Psychology Review6(1) 7ndash32

Dunton G F Liao Y Grana R Lagloire R Riggs N Chou C P et al (2012)State-wide dissemination of a school-based nutrition education programme ARE-AIM (Reach Efficacy Adoption Implementation Maintenance) analysis PublicHealth Nutrition 17 422ndash439

Ferrer E amp McArdle J (2003) Alternative structural models for multivariatelongitudinal data analysis Structural Equation Modeling 10 493ndash524

Fishbein M Triandis H C Kanfer F H Becker M Middlestadt S E amp Eichler A(2000) Factors influencing behavior and behavior change In A S Baum T ARevenson amp J E Singer (Eds) Handbook of health psychology (pp 1ndash17) MahwahNJ Lawrence Erlbaum

Gilson T A Chow G M amp Feltz D L (2012) Self-efficacy and athletic squatperformance Positive or negative influences at the within-and between-levelsof analysis Journal of Applied Social Psychology 42(6) 1467ndash1485

Hagler A S Norman G J Radick L R Calfas K J amp Sallis J F (2005) Comparabilityand reliability of paper-and computer-based measures of psychosocial constructsfor adolescent fruit and vegetable and dietary fat intake Journal of the AmericanDietetic Association 105(11) 1758ndash1764

Hoelscher D M Day R S Kelder S H amp Ward J L (2003) Reproducibility andvalidity of the secondary level School-Based Nutrition Monitoring studentquestionnaire Journal of the American Dietetic Association 103(2) 186ndash194

Hu L amp Bentler P M (1999) Cutoff criteria for fit indexes in covariance structureanalysis Conventional criteria versus new alternatives Structural EquationModeling A Multidisciplinary Journal 6(1) 1ndash55

Leventhal H amp Mora P A (2005) Is there a science of the processes underlyinghealth and illness behaviors A comment on Maes and Karoly Applied Psychology54(2) 255ndash266

Maddux J E amp Rogers R W (1983) Protection motivation and self-efficacy A revisedtheory of fear appeals and attitude change Journal of Experimental SocialPsychology 19(5) 469ndash479

Maes S amp Karoly P (2005) Self-Regulation Assessment and Intervention in PhysicalHealth and Illness A Review Applied Psychology 54(2) 267ndash299

McArdle J J amp Nesselroade J R (1994) Using multivariate data to structuredevelopmental change In S H Cohen amp H W Reese (Eds) Life-span developmentalpsychology Methodological contributions (pp 223ndash267) Hillsdale NJ LawrenceErlbaum

Michie S amp Prestwich A (2010) Are interventions theory-based Development ofa theory coding scheme Health Psychology 29(1) 1ndash8

National Cancer Institute (2005) Food attitudes and behaviors (FAB) survey HealthBehaviors Research Branch Retrieved from lthttpcancercontrolcancergovbrpfabindexhtmlgt

Plotnikoff R C Lippke S Courneya K S Birkett N amp Sigal R J (2008) Physicalactivity and social cognitive theory A test in a population sample of adults withtype 1 or type 2 diabetes Applied Psychology 57(4) 628ndash643

Rhodes R E amp Nigg C R (2011) Advancing physical activity theory A review andfuture directions Exercise and Sport Sciences Reviews 39(3) 113ndash119

Rogers R W (1975) A protection motivation theory of fear appeals and attitudechange The Journal of Psychology 91(1) 93ndash114

Schwarzer R (1992) Self-efficacy in the adoption and maintenance of health behaviorsTheoretical approaches and a new model Washington DC Hemisphere PublishingCorp

Schwarzer R Schuumlz B Ziegelmann J P Lippke S Luszczynska A amp Scholz U(2007) Adoption and maintenance of four health behaviors Theory-guidedlongitudinal studies on dental flossing seat belt use dietary behavior and physicalactivity Annals of Behavioral Medicine 33(2) 156ndash166

Seo M amp Ilies R (2009) The role of self-efficacy goal and affect in dynamicmotivational self-regulation Organizational Behavior and Human Decision Processes109(2) 120ndash133

Taylor N Conner M amp Lawton R (2012) The impact of theory on the effectivenessof worksite physical activity interventions A meta-analysis and meta-regressionHealth Psychology Review 6(1) 33ndash73

Thiagarajah K Fly A D Hoelscher D M Bai Y Lo K Leone A et al (2008)Validating the food behavior questions from the elementary school SPANquestionnaire Journal of Nutrition Education and Behavior 40(5) 305ndash310

Urbig D amp Monsen E (2012) The structure of optimism ldquoControllability affectsthe extent to which efficacy beliefs shape outcome expectanciesrdquo Journal ofEconomic Psychology 33(4) 854ndash867

Weinstein N D (2007) Misleading tests of health behavior theories Annals ofBehavioral Medicine 33(1) 1ndash10

White S M Woacutejcicki T R amp McAuley E (2012) Social cognitive influences onphysical activity behavior in middle-aged and older adults The Journals ofGerontology Series B Psychological Sciences and Social Sciences 67 P18ndashP26

170 AL Larsen et alAppetite 84 (2015) 166ndash170