The Roles of Behavioral and Implementation Intentions in Changing Physical Activity in Young Children With Low Socioeconomic Status

Transcription

1 Journal of Sport & Exercise Psychology, 2010, 32, Human Kinetics, Inc. The Roles of Behavioral and Implementation Intentions in Changing Physical Activity in Young Children With Low Socioeconomic Status Christopher J. Armitage and Christine A. Sprigg University of Sheffield There is a dearth of research examining physical activity in children aged 6 10 years with low socioeconomic status, despite the fact there is good reason to suspect this is a critical period when physical activity habits are created. Physical activity and theory of planned behavior variables were measured at three time points, and children (N = 77) randomized to the experimental condition were additionally asked to form an implementation intention. Intention was a potent mediator of the past behavior future behavior relationship and the implementation intention intervention significantly increased physical activity compared with the control condition. The findings suggest that physical activity can be increased in children aged 6 10 years with low socioeconomic status and that implementation intentions might enhance the effectiveness of children s physical activity programs. Keywords: low socioeconomic status, young children, physical activity, implementation intentions, theory of planned behavior Although the relationships between low socioeconomic status, low physical activity and poorer health are widely accepted in adults (e.g., Stamatakis, 2006), the corresponding relationships in children and adolescents are more complex. For example, Kelly et al. (2006) showed no differences between high and low socioeconomic status groups on objective measures of physical activity in two samples of children aged around 4.2 years (study 1) and 5.6 years (study 2), and Brodersen, Steptoe, Boniface, and Wardle s (2007) 5-year longitudinal study of physical activity in children found that most differences between... SES groups were present at age 11 years, and did not evolve over the teenage years (p. 14). The implication is that the period between ages 6 and 10 years may be critical to the establishment of positive physical activity habits in children with low socioeconomic status and that interventions specifically targeted at this group need to be tested. Armitage is with the Department of Psychology, University of Sheffield, Sheffield, United Kingdom. Sprigg is with the Institute of Work Psychology, University of Sheffield, Sheffield, United Kingdom. 359

2 360 Armitage and Sprigg Interventions to Increase Physical Activity in Children Salmon, Booth, Phongsavan, Murphy, and Timperio s (2007) systematic review identifies 42 separate interventions designed to increase physical activity in children (defined as 4- to 12-year-olds) who were tested using experimental or quasiexperimental designs, and conclude that the most effective interventions were carried out in school settings. However, with very few exceptions, the interventions consisted of multiple potential active ingredients (e.g., multiple social cognition change, increased teacher contact) that were evaluated against passive control groups (e.g., standard physical education lessons, usual care, measurement only). Both factors mean it is difficult to isolate the active ingredients of the interventions, even if they have been successful; for example, it is perhaps unsurprising if multiple sessions of physical education tend to be successful in increasing physical activity compared with passive control groups, but it is not always clear what elements of those sessions are exerting the observed effects (e.g., presenter characteristics, attitude change vs. norm change). More generally, there is a dearth of studies within this field that have focused specifically on the critical period identified above as 6 10 years old. The present study was designed with two broad goals: First, to identify psychosocial mediators of past physical activity on intention and future physical activity and so form the basis of theory-driven interventions, and, second, to test a brief intervention based on Gollwitzer s (1993) implementation intentions to promote behavior change. Mediators of the Past Behavior Future Behavior Relationship One approach to understanding better the low uptake and maintenance of physical activity has been to investigate the psychosocial antecedents of physical activity with a view to developing effective theory-driven campaigns (e.g., Marcus et al., 2000). Research on the theory of planned behavior (Ajzen, 1991) has dominated this field, as evidenced by the large number of reviews and meta-analyses, which consistently report strong relationships between theory of planned behavior variables and physical activity (e.g., Hagger, Chatzisarantis, & Biddle, 2002). Ajzen s (1991) theory of planned behavior is a model of the factors that motivate human behavior. Central to the theory of planned behavior is the intention construct, which is conceptualized as a summary of the motivation required to engage in a particular behavior. Underpinning intentions are three constructs: attitude, subjective norm and perceived control. Attitudes are global positive or negative evaluations of behavior, subjective norms are a summary of the social pressure from significant others to engage or not to engage in a behavior, and perceived control reflects the perceived ease or difficulty of performing a behavior and is regarded as being synonymous with Bandura s concept of self-efficacy, or confidence in one s own ability (see Ajzen, 1998, and Bandura, 1998). Attitudes, subjective norms, and perceived control are hypothesized to derive from beliefs about outcomes, significant referents, and specific facilitating/inhibiting factors, respectively (Ajzen, 1991), although these beliefs are to a large extent based on prior experience (e.g., Armitage, 2007). Ajzen (1991) regards the inclusion of past behavior in the model as a test of its sufficiency: The theory of planned behavior is sufficient to the extent that its variables are predictive of subsequent behavior controlling for past behavior. From

3 Physical Activity in Young Children With Low Socioeconomic Status 361 a more practical perspective, if theory of planned behavior variables can mediate the effects of past behavior on future behavior, they may represent targets for future interventions. The theory of planned behavior has been applied extensively and accounts for large proportions of the variance in health-related intentions and behaviors (e.g., Armitage & Conner, 2001). This consistent pattern of findings also emerges when applications of the theory of planned behavior to physical activity are considered separately (e.g., Hagger et al., 2002). For example, Hagger et al. (2007) showed that the theory of planned behavior was predictive of young people s physical activity intentions and subsequent physical activity, even across cultures. However, according to Rhodes, Macdonald, and McKay s (2006) review of the literature, there were just three theory of planned behavior studies that examined physical activity in preadolescent children. Even accounting for those articles published since Rhodes et al. (2006), just one has tested the theory of planned behavior in a sample that analyzes the critical period (i.e., children aged 6 10 years) identified above. In that study, Mummery, Spence, and Hudec (2000) used the theory of planned behavior to predict intentions to engage in physical activity in a sample of children from across Canada (socioeconomic status not assessed). Of particular interest to the present research are the children in the youngest of four age groups Mummery et al. tested. Using a sample of Canadian grade 3 school pupils with a mean age of 8.2 years (SD = 0.5), Mummery and colleagues showed that attitude, subjective norm, and perceived control were all significant predictors of intention to engage in physical activity, accounting for 35% of the variance. This would suggest that the theory of planned behavior can be applied in younger children and that further research is needed to establish whether theory of planned behavior variables can predict subsequent behavior in a sample in this age group with low socioeconomic status. The first aim of the present research was therefore to test whether the theory of planned behavior could be applied successfully in a sample of children aged 6 10 years with low socioeconomic status, and whether theory of planned behavior variables could mediate the past behavior-future behavior relationship, and so provide potential targets for future interventions. Implementation Intentions and Physical Activity Despite large relationships between variables from the theory of planned behavior and subsequent behavior, there is still a large proportion of the variance unaccounted for (e.g., Armitage & Conner, 2001). To account for this apparent discrepancy between motivation and behavior, several authors have argued that there are two stages motivational and volitional through which people pass before acting (e.g., Gollwitzer, 1993; Heckhausen, 1991). The motivational stage is concerned with increasing people s orientation toward enacting the behavior and culminates in the formation of a behavioral intention. The volitional stage culminates in the actual performance of the behavior in question and is important in translating motivational cognitions into action (e.g., Gollwitzer; Heckhausen). Thus, whereas the theory of planned behavior (Ajzen, 1991) provides clear guidance as to what motivates people (e.g., attitudes, norms), it is less clear about the ways in which motivation is translated into action (but see Araújo-Soares, McIntyre, & Sniehotta, 2009; Dombrowski & Luszczynska, 2009, for examples of research seeking to

4 362 Armitage and Sprigg expand the theory of planned behavior in this direction). Gollwitzer s (1993, 1999) concept of implementation intention is important in this regard. Implementation intentions are volitional strategies that work independently of motivation by ensuring decisions are acted upon. Thus, whereas behavioral intentions tap the extent to which individuals are motivated to act, implementation intention are regarded as important in translating that motivation into action (Gollwitzer, 1993, 1999). Consistent with this idea, implementation intentions are only effective when people are at least somewhat motivated to act (Armitage, 2006). Implementation intentions consist of two principal components: a critical situation and an appropriate behavioral response. In other words, implementation intentions are if-then plans that work by linking in memory a critical situation ( if ) with an appropriate behavioral response ( then ). The idea is that specifying where and when one will act ensures that the appropriate behavioral response will be triggered at the appropriate time and place in the future (Gollwitzer, 1993, 1999). There is now a large body of research attesting to the efficacy of implementation intention-based interventions for changing health behavior in field settings. In a recent review, Gollwitzer and Sheeran (2006) identified 94 independent tests of implementation intentions (including laboratory tests) that yielded a medium to large effect size (d =.65). From a methodological perspective, implementation intentions offer several key advantages to the kinds of physical activity interventions that are described above. First, whereas many physical activity interventions take an omnibus approach by delivering simultaneously several different types of intervention, research into implementation intentions is characterized by studies that manipulate implementation intentions alone, meaning that implementation intentions are unambiguously causing behavior change. Second, implementation intentions are effective but brief meaning they are inexpensive and can be administered to large populations without first having to screen participants, and without having to target or tailor the materials. To date, numerous published experimental studies have shown positive effects of implementation intentions for increasing physical activity (e.g., Gollwitzer & Sheeran, 2006), although none have tested their effects in preadolescent children. Thus, the second aim of the present research was to test whether an intervention based on the concept of implementation intentions could increase physical activity in a sample of children aged 6 10 years with low socioeconomic status. Consistent with best practice in the field, the intervention was tested against an equivalently active control group and the data were subject to intention to treat analysis. Rationale and Hypotheses The research reviewed above provides the following rationale for the current study. First, interventions to increase physical activity in children with low socioeconomic status might best be targeted between 6 and 10 years before physical activity differentials due to socioeconomic status become established. Second, very little research has examined children aged between 6 and 10 years per se either in respect to applying the theory of planned behavior, testing physical activity interventions or in employing implementation intentions. Third, the research conducted on implementation intentions in the context of physical activity has suffered some limitations. It was predicted that: (a) theory of planned behavior variables would mediate the

5 Physical Activity in Young Children With Low Socioeconomic Status 363 effects of past physical activity on (i) intention, and (ii) future physical activity; and (b) a brief intervention based on implementation intentions would be successful in increasing physical activity in young children with low socioeconomic status. Participants Method The initial sample consisted of 77 children aged 6 10 years recruited from three classes (each with a single teacher) in an inner city primary school in England. The school and its catchment area were based in the top 10% deprived areas in the UK, based on income; employment; health and disability; education, skills and training; barriers to housing and services; crime; and living environment (Sheffield City Council, 2007). The study was given ethical approval by the appropriate Internal Review Board. Teachers acting in loco parentis gave consent; pupils were informed that they were free to choose whether to participate and that they could withdraw themselves or their data at any time. After three phases of data collection, attrition (due to absence from school) resulted in a full dataset consisting of 71 children (Figure 1). However, all analyses were conducted using intention to treat, meaning that any missing data (applying to six cases only) were treated as no changers ( last observation carried forward ). The sample consisted of 39 boys and 38 girls, with an average age of 8 years (M = 8.06, SD = 1.63). At power =.80, the sample size can detect a large effect size at α =.01 (Cohen, 1992). Design A randomized controlled design was used, and the children were randomly allocated to either the experimental (n = 39) or control (n = 38) condition on the basis of coin tosses. Physical activity and theory of planned behavior variables were measured at three time points. Time 1 acted as a baseline and physical activity and theory of planned behavior variables were measured at the same time. Two weeks later (Time 2), the children completed measures of theory of planned behavior constructs and were exposed to the intervention (or a control manipulation) before playtime and then completed measures of physical activity after playtime. Time 3 came one month later when physical activity and theory of planned behavior variables were measured after playtime. Procedure Before being asked to complete the questionnaires, the children were informed that they were taking part in a study that was looking at their participation in physical activity, that they would be asked to complete three questionnaires, and that the experimenter would like them to do lots of running around at playtimes. Piloting showed that the children interpreted the term running around as general physical activity. To ensure this was the case, the children were reminded at each time point that running around included various activities, including skipping, jumping and physical games (e.g., tag) as well as running.

6 364 Figure 1 Flow diagram of children s progress through the phases of the experiment. Note that the data were analyzed according to intention to treat.

7 Physical Activity in Young Children With Low Socioeconomic Status 365 The children were removed from their classroom in groups of four and taken to a quiet room away from the other children to complete the questionnaires. Memberships of these groups of four children were randomized to minimize potential demand characteristics (i.e., the children s interpretations of the experiment affecting their responses) or experimenter expectancy effects. The questionnaires were completed under the guidance of the experimenter, who read each question aloud to the groups and explained the meaning of the response options as appropriate. This basic procedure was repeated at each time point. At Time 2, children in the experimental condition had printed at the bottom of their theory of planned behavior questionnaire the implementation intention: if it is playtime, then I will run around as much as possible, which they said aloud three times. Although it is relatively unusual for implementation intentions to be formed in this way in field studies on adults, in laboratory studies, it is commonplace for participants to be provided with implementation intentions by the experimenter. For example, in Webb and Sheeran (2004, experiment 1): All participants were told to repeat their plans a couple of times (p. 410), on the grounds that it ensures quality and consistency across participants. Thus, we regard our procedure as a good way to ensure that the children were actually forming implementation intentions as opposed to making vague plans that do not adhere strictly to the definition of an implementation intention. Several studies now show that adhering closely to the if-then format of an implementation intention is very important (see Chapman, Armitage, & Norman, 2009), whereas simply asking people to plan to change their behavior has no effect compared with asking them to form an implementation intention, defined as linking a critical situation with an appropriate behavioral response (e.g., Armitage, 2008a; Armitage & Arden, 2010). Moreover, a recent study demonstrated no difference between whether an implementation intention was provided by the experimenter or whether it was generated solely by the participant (Armitage, 2009a). The control group did not receive an implementation intention but were asked to say aloud three times it s good to run around as much as possible. For both groups, the experimenter ensured that each child said the phrase aloud three times. After testing was completed the children went out to play. To minimize potential demand characteristics, the experimenter did not observe the children in the playground. Measures All theory of planned behavior measures were assessed on 5-point Likert-type scales similar to those used in Armitage s (2008b, study 1) study with 10- to 11-year-old school children. However, piloting indicated that the numbers typically used on such scales were confusing, particularly to the younger children, and so the children were asked to choose faces that best represented their response. The facial expressions on the simple line-drawing faces ranged from sad, somewhat sad, neutral, somewhat happy and happy, which were recoded from 1 to 5. Intention was measured using three items: I will run around as much as possible in the next playtime will not will, I am going to run around as much as possible in the next playtime am not am, and How likely is it that you will run about as much as possible in the next playtime? not likely likely. The measure had high internal reliability at each time point: alphas =.79,.84, and.79, respectively.

8 366 Armitage and Sprigg Attitude was measured using three items: Running around as much as possible in the next playtime will make me... sad happy, Running around as much as possible in the next playtime is... silly clever, and Running around as much as possible in the next playtime is... bad good. At Time 1, the alpha coefficient was.76, the same measure at Time 2 had an alpha coefficient of.71, at Time 3 the alpha coefficient was.65. Responses to three items were averaged to tap subjective norm: People who are important to me think I should not should run around as much as possible in the next playtime, People who are important to me would not like like me to run around as much as possible in the next playtime, and People who are important to me want me to run around as much as possible in the next playtime not likely likely. The alpha coefficients were.69,.81, and.66 at Times 1, 2, and 3, respectively. Perceived control was measured with, I am able to run around as much as possible in the next playtime: I am not I am, How much do you see yourself as able to run around as much as possible in the next playtime? not at all lots, and Do you feel you are able to run around as much as possible in the next playtime? I am not able I am able. Alpha coefficients indicated acceptable internal reliability at each time point:.84 (Time 1),.80 (Time 2), and.67 (Time 3). Physical activity was measured using self-report, rather than accelerometers because the aim of the study was to develop a brief intervention with minimal demand characteristics, yet accelerometers require at least 4 days of monitoring in preadolescents to achieve satisfactory reliability (Trost, Pate, Freedson, Sallis, & Taylor, 2000). Given that there are currently no self-report measures of physical activity in younger children, the present measure was designed for the purposes of the current study. The design of the physical activity measure was informed by Ajzen s (1991) principle of compatibility to ensure it mapped as closely as possible onto the theory of planned behavior measures and onto the instructions of the experimenter. Thus, physical activity was assessed with three items: I ran around as much as possible in the last playtime: I did not I did, I did run around as much as possible at the last playtime: I did not I did, and I was able to run around as much as possible at the last playtime: I was not able I was able. The measure had good internal reliability at each time point, as indicated by high alpha coefficients:.91 (Time 1),.84 (Time 2), and.83 (Time 3). Randomization Check Results MANOVA, with condition (experimental vs. control group) as the independent variable and age, gender, physical activity at Time 1, and theory of planned behavior variables at Time 1 as the dependent variables were used to check the randomization procedure. The multivariate test and all the univariate tests were nonsignificant, F univariate (1, 77) = , ps =.44 to.99, η p2 s <.01, indicating that randomization was successful. This means that, before the implementation intention manipulation, children randomized to the experimental and control conditions were engaging in similar levels of physical activity and were equally motivated to increase their physical activity.

9 Physical Activity in Young Children With Low Socioeconomic Status 367 Predictive Validity of the Theory of Planned Behavior Table 1 presents the zero-order correlations between theory of planned behavior variables and physical activity at each time point. Although there were some strong intercorrelations among theory of planned behavior variables within a time point (e.g., intention and perceived control at Time 2), all were significantly less than perfect (i.e., two standard errors below 1.0; see Netemeyer, Johnston, & Burton, 1990), demonstrating discriminant validity. The theory of planned behavior measures showed very good test retest reliabilities, r =.65 to.83, p <.01, with the exception of subjective norm, where the test retest correlations were r =.31, p <.01 (Time 1 to Time 2); r =.62, p <.01 (Time 1 to Time 3); and r =.55, p <.01 (Time 2 to Time 3). Although we were anticipating changes in physical activity over time, the test retest reliabilities for our measure of physical activity was still adequate, r =.59 to.68, p <.01. Consistent with predictions derived from the theory of planned behavior, attitude, subjective norm and perceived control were all strongly and positively correlated with intention; and intention and perceived control were both strongly correlated with behavior (Table 1). The ability of the theory of planned behavior to mediate the effects of past physical activity was tested in two hierarchical regression analyses. The first hierarchical regression analysis is reported in Table 2 and uses Time 1 intention as the dependent variable. For this analysis, the experimental and control groups were collapsed because the randomization check showed that the two groups did not differ significantly on any of the variables tested in Table 2. Past physical activity (Time 1) was entered on Step 1 and accounted for 26% of the variance in intention. The addition of theory of planned behavior variables on Step 2 significantly augmented the model, which explained 71% of the variance in intention. On the second step, attitude, subjective norm and perceived control were significant predictors of intention, but the beta weight for the measure of past physical activity was nonsignificant. The analyses presented in Table 2 suggest that the theory of planned behavior variables might mediate the effects of past behavior on intention. This was tested formally using the bootstrapping procedures outlined in Preacher and Hayes (2008), and all potential mediators of the past behavior intention relationship were tested. The basis for these analyses is that the indirect effect of prior physical activity on intention is the product of the paths between Time 1 physical activity and potential mediator (i.e., attitude, subjective norm, perceived control), and between potential mediator and intention. However, such indirect effects are not normally distributed, meaning that bootstrapping is necessary (Preacher & Hayes). This involves resampling random subsets of data to gain a nonparametric approximation of the sampling distribution of the product of the past behavior mediator and mediator intention paths. The analyses presented here are based on 1,000 resamples, although repeating the analyses with more resamples made no difference to the findings. Consistent with the data presented in Table 2, the 95% confidence intervals associated with each of the theory of planned behavior variables did not include zero (95% CI attitude =.001,.179; 95% CI subjective norm =.047,.306; 95% CI perceived control =.088,.465) meaning that attitude, subjective norm, and perceived control significantly (p <.05) albeit narrowly in the case of attitude mediated the effect of past behavior on intention.

10 Table 1 Zero-Order Correlations Between Measured Variables Variables Time 1 1. Attitude 2. Subjective Norm.40** 3. Perceived Control.49**.63** 4. Intention.54**.68**.78** 5. Physical Activity.36**.41**.44**.51** Time 2 6. Attitude.73**.47**.66**.72**.58** 7. Subjective Norm.48**.31**.46**.54**.33**.53** 8. Perceived Control.53**.57**.73**.81**.45**.75**.61** 9. Intention.46**.46**.69**.79**.48**.72**.62**.83** 10. Physical Activity.33**.30**.38**.50**.59**.56**.42**.50**.47** Time Attitude.74**.43**.62**.70**.59**.76**.54**.64**.59**.51** 12. Subjective Norm.48**.62**.65**.65**.57**.62**.55**.58**.55**.55**.67** 13. Perceived Control.56**.73**.78**.82**.55**.78**.56**.77**.70**.52**.72**.82** 14. Intention.49**.68**.67**.83**.46**.69**.43**.73**.65**.46**.66**.65**.82** 15. Physical Activity.37**.38**.42**.44**.60**.46**.47**.37**.39**.68**.50**.58**.51**.36** **p <

11 Physical Activity in Young Children With Low Socioeconomic Status 369 Table 2 Hierarchical Regression Analysis Predicting Intention to Engage in Physical Activity From Baseline Physical Activity and Theory of Planned Behavior Variables Step and Variable β R 2 ΔR 2 Step 1.26**.26 Physical Activity (Time 1).51** Step 2.71**.45 Physical Activity (Time 1).13 Attitude.15* Subjective Norm.27** Perceived Control.48** *p <.05, **p <.01. The theory of planned behavior was further evaluated using subsequent behavior as the dependent variable (Table 3). Again, data from the experimental and control conditions were pooled and so the effects of condition were statistically controlled at each step of the analysis because it was anticipated that the intervention would produce changes in the dependent variable. Consistent with this prediction, condition accounted for 10% of the variance (p <.01) in subsequent behavior, meaning that the children in the experimental condition exercised more at followup than did those in the control condition. Comparable with the analyses reported above, past physical activity was entered on the second step, and accounted for an additional 35% of the variance in subsequent physical activity. Theory of planned behavior variables were entered on step 3, and intention emerged as a significant predictor, alongside condition and past physical activity in the final model, which accounted for 50% of the variance in subsequent physical activity. The analyses presented in Table 3 imply that intention may have mediated the effect of past behavior on future behavior, but all potential mediators were assessed simultaneously using Preacher and Hayes s (2008) bootstrapping procedure for testing multiple mediators, as outlined earlier. Consistent with the data presented in Table 3, only intention did not have zero in the 95% confidence interval for the indirect effect of past behavior on future behavior (95% CI =.018,.373). Thus, intention significantly mediated the past behavior future behavior relationship. However, because past behavior remains a significant predictor of future behavior, it implies that although intention is a potent mediator of the past behavior future behavior relationship, there are likely to be other mediators of this relationship (see Baron & Kenny, 1986). However, these data do imply that an intervention designed to facilitate the translation of intentions into action might be valuable in changing behavior.

12 370 Armitage and Sprigg Table 3 Hierarchical Regression Analysis Predicting Time 2 Physical Activity Using Time 1 Physical Activity and Theory of Planned Behavior Variables While Controlling for the Effects of Experimental Condition Step and Variable β R 2 ΔR 2 Step 1.10**.10 Condition.31** Step 2.45**.35 Condition.31** Physical Activity (Time 1).59** Step 3.50**.05 Condition.30** Physical Activity (Time 1).48** Intention.32* Perceived Control.04 Attitude.05 Subjective Norm.14 Note. Condition is coded 1 = control condition; 2 = experimental condition. *p <.05, **p <.01. Effects of the Implementation Intention Intervention The effects of the implementation intention intervention on theory of planned behavior variables and physical activity were tested initially using a series of repeated-measures ANOVAs. Condition (experimental vs. control) was the betweenparticipants factor, time (Time 1 vs. Time 2 vs. Time 3) the within-persons factor, and physical activity and the theory of planned behavior variables were the dependent variables. The data in Table 4 show there was just one significant condition time interaction, for physical activity. This was decomposed using a series of ANOVAs. First, separate within-persons ANOVAs for the experimental and control conditions were used to test the simple within-persons effects. Simple contrasts using Time 1 as the reference category were used to clarify whether any effects observed at Time 2 were sustained to Time 3. Second, between-participants ANCOVAs were used to examine the effects of condition on physical activity Times 2 and 3, controlling for physical activity at Time 1. Consistent with the patterns of means reported in Table 4, there was no change in physical activity over time for children in the control condition, F(2, 36) = 1.76, p =.19, η p 2 =.09, but significant improvements in the experimental condition, F(2, 37) = 6.35, p <.01, η p 2 =.26. Simple contrasts showed that both Time 2 physical activity, F(1, 38) = 10.91, p <.01, η p 2 =.22, and Time 3 physical activity, F(1, 38) = 12.68, p <.01, η p 2 =.25, differed significantly from Time 1. ANCOVAs (controlling for Time 1 physical activity) confirmed that physical activity differed significantly between the two conditions at both Time 2, F(1, 76) = 13.27, p <.01, η p 2 =.15, and Time 3, F(1, 76) = 17.78, p <.01, η p 2 =.19.

13 Table 4 Effects of the Implementation Intention Intervention on Physical Activity and Theory of Planned Behavior Variables Across Time Time 1 Time 2 Time 3 F a Dependent Variables M SD M SD M SD df = 2, 74 Physical Activity 7.01** Control group, n = Experimental group, n = Attitude 0.98 Control group, n = Experimental group, n = Subjective Norm 0.89 Control group, n = Experimental group, n = Intention 1.45 Control group, n = Experimental group, n = Perceived Control 0.04 Control group, n = Experimental group, n = a Refers to the condition time interaction. **p <

14 372 Armitage and Sprigg Given that theory of planned behavior variables did not differ across time, the implication is that the effects of the implementation intention intervention were unmediated by theory of planned behavior variables. Discussion Between the ages of 6 and 10, there appears to be a critical period during which socioeconomic differentials in physical activity arise and are established (e.g., Brodersen et al., 2007; Kelly et al., 2006), yet there is a dearth of research that has looked specifically at this age group. The present study was designed to examine social cognitive predictors of physical activity in younger children with low socioeconomic status and test the ability of an implementation intention-based brief intervention to increase physical activity in this group of children. There were three key findings. First, the theory of planned behavior provided a very good account of physical activity intentions and behavior, crucially showing that theory of planned behavior variables significantly mediated the effects of prior physical activity on subsequent physical activity. Most notably, behavioral intention significantly mediated the past behavior-future behavior relationship, thereby partly explaining why implementation intention interventions might be successful in this domain. Second, a brief intervention based on implementation intentions significantly increased physical activity in young children with low socioeconomic status. Interestingly, the strength of the effect increased marginally over time, such that there were larger differences between experimental and control group at Time 3 than there were at Time 2. Third, the effects of the intervention on physical activity occurred independently of theory of planned behavior variables, suggesting that motivational and volitional processes are independent. Mediators of the Past Behavior Future Behavior Relationship The present findings map onto previous work that has applied the theory of planned behavior to predict physical activity, and the magnitude of the zero-order correlations are comparable with relevant meta-analyses (e.g., Hagger et al., 2002). The main difference between the present research and prior research was that perceived control was not a significant predictor of subsequent behavior whereas in general it is (see Hagger et al.). The implication is that the present population may differ in some important respects from samples that are more commonly tested. It would be valuable to discover in future work whether low socioeconomic status, the relative youth of the sample, or a combination of the two was responsible for the weak effects of perceived control. After all, Ajzen (1991) argues that the relative predictive power of different components of the theory of planned behavior is likely to differ across topics and samples. It is notable that the mediational effects clearly point to a route for promoting physical activity. Consistent with Ajzen s (1991) conceptualization of the theory of planned behavior, the effects of past behavior on intention were almost fully mediated through attitude, subjective norm and perceived control. In turn, the effects of past behavior were at least partially mediated through intention. There are two main implications of this finding: (a) it is notable that the effects of past behavior on future behavior were not mediated through attitude and subjective norm, which

15 Physical Activity in Young Children With Low Socioeconomic Status 373 is consistent with Ajzen s (1991) idea that attitudes and subjective norms should only predict intention, and (b) a combination of interventions targeting people s perceptions of their past behavior (e.g., Armitage, 2007) and theory of planned behavior variables should increase the effectiveness of interventions to increase physical activity in these kinds of samples. The implementation intention intervention in addition to past behavior and intention also contributed to the prediction of subsequent physical activity. Not only is this consistent with other studies that have successfully applied implementation intentions to physical activity (e.g., Armitage & Arden, 2010), but extends previous findings to young children with low socioeconomic status. Thus the present findings speak to how robust and replicable implementation intentions are for changing behavior. However, the present research tested just one implementation intention and it would be valuable to conduct further research into what should form the content of implementation intention manipulations. Given that there seems to be little difference between implementation intentions that are generated by the individual and those that are provided by the experimenter (Armitage, 2009a), there is great scope for researchers to develop and refine implementation intentions that will maximize their impact on physical activity or behavior in general. For example, some recent work has shown that implementation intentions based on concepts from Prochaska and DiClemente s (1983) transtheoretical model were successful in increasing physical activity in adults with low socioeconomic status (Armitage & Arden, 2010). However, it is notable that the inclusion of theory of planned behavior variables did not reduce the past behavior-future behavior relationship to zero, meaning that further mediators of this relationship are yet to be found. Another important avenue for further research is therefore to identify additional variables that will mediate the effects of past behavior on future behavior. Implementation Intentions In field settings, participants are typically encouraged to form their own implementation intentions in an open-ended format using standard manipulations (e.g., Gollwitzer, 1993). Although this approach has proved successful in terms of changing behavior, it is limited because it has proven difficult to code the content of people s implementation intentions meaning that it has not been possible to identify particularly effective implementation intentions. In contrast, participants in laboratory research are most often provided with an implementation intention by the experimenter and asked to say it to themselves (e.g., Gollwitzer & Sheeran, 2006). Again, this technique has proved successful in terms of observed differences, but it is difficult to verify that the instructions have been carried out successfully. For the purpose of the present field research in young children with low socioeconomic status, the procedure was amended to draw on the strengths of both the field and laboratory approaches (see Armitage, 2009b). Specifically, the children were provided with an implementation intention but were asked to say it aloud, meaning that the experimenter could assure fidelity to the manipulation. Given that the present findings map closely onto relevant meta-analyses (e.g., Gollwitzer & Sheeran), the implication is that the format of the implementation intention maybe less important than the mental act of linking critical situation with appropriate behavioral response (see Armitage, 2009a).

16 374 Armitage and Sprigg Potential Limitations It is important to note a number of possible limitations that could temper enthusiasm for the promising findings obtained in this group of young children with low socioeconomic status. First, the principal dependent measure was a novel measure of physical activity, developed for the purposes of the current study because no measures of physical activity participation exist for children aged 10 years and younger. However, there was some preliminary evidence to support its reliability and validity, the measure possessed (a) good internal reliability at each time point; (b) good test retest reliability across the time points; (c) good face validity, established in piloting; (d) good predictive validity, as established using theory of planned behavior measures as independent variables; and (e) the measure was sensitive to the implementation intention manipulation. Nonetheless, it would be valuable to further establish the reliability and validity of the measure. Second, physical activity was self-reported to avoid intrusive baseline assessments using accelerometers (e.g., Trost et al., 2000) raising the possibility that reporting biases might account for the present findings. However, there is reason to be cautiously optimistic on this point because it is notable that the effect size associated with the intervention was actually greater at Time 3 than at Time 2, despite the fact that the demand characteristics were presumably greater at Time 2. It is also not clear why reporting biases should differ between self-reports of behavior and self-reports of cognition, providing further confidence in the present findings. A third potential limitation was the postintervention follow-up, which was 1 month but would ideally have followed the children over several years. However, exploratory early trials are important to build up evidence toward full trials as they can improve understanding of the way the intervention works, provide effect size estimates and test theory; moreover, there is some evidence that habits can established within weeks rather than months (e.g., Armitage, 2005). Finally, because implementation intentions work by heightening cue accessibility and strengthening cue-response links (Gollwitzer & Sheeran, 2006), and such measures are not currently available outside the laboratory, were unable to test the factors that mediated the effects of the intervention nor conduct a manipulation check. Conclusions This is the first study to have tested a brief implementation intention-based intervention to increase physical activity in young children. Crucially, the children were in the critical 6 10 age group at which socioeconomic differentials in health are likely to crystallize (e.g., Brodersen et al., 2007; Kelly et al., 2006). The findings show that implementation intentions were successful in increasing physical activity across a sustained period of time and were consistent with the idea that there are distinct motivational and volitional phases in human behavior. Further research is required to ensure these effects can be sustained in the longer term to see whether interventions at an early age can prevent socioeconomic differentials in physical activity from arising. Acknowledgments We would like to thank Sam Beresford for her help collecting and entering the data.

17 Physical Activity in Young Children With Low Socioeconomic Status 375 References Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, Ajzen, I. (1998). Models of human social behavior and their application to health psychology. Psychology & Health, 13, Araújo-Soares, V., McIntyre, T., & Sniehotta, F.F. (2009). Predicting changes in physical activity among adolescents: The role of self-efficacy, intention, action planning and coping planning. Health Education Research, 24, Armitage, C.J. (2005). Can the theory of planned behavior predict the maintenance of physical activity? Health Psychology, 24, Armitage, C.J. (2006). Evidence that implementation intentions promote transitions through the stages of change. Journal of Consulting and Clinical Psychology, 74, Armitage, C.J. (2007). Changing past behavior without means of a time machine: Effects on future behavioral decisions. Social Cognition, 25, Armitage, C.J. (2008a). A volitional help sheet to encourage smoking cessation: A randomized exploratory trial. Health Psychology, 27, Armitage, C.J. (2008b). Cognitive and affective predictors of academic achievement in schoolchildren. British Journal of Psychology, 99, Armitage, C.J. (2009a). Effectiveness of experimenter-provided and self-generated implementation intentions to reduce alcohol consumption in a sample of the general population: A randomized exploratory trial. Health Psychology, 28, Armitage, C.J. (2009b). Is there utility in the transtheoretical model? British Journal of Health Psychology, 14, Armitage, C.J., & Arden, M.A. (2010). A volitional help sheet to increase physical activity in people with low socioeconomic status: A randomised exploratory trial. Psychology & Health. doi: / Armitage, C.J., & Conner, M. (2001). Efficacy of the theory of planned behaviour: A metaanalytic review. British Journal of Social Psychology, 40, Bandura, A. (1998). Health promotion from the perspective of social cognitive theory. Psychology & Health, 13, Baron, R.M., & Kenny, D.A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, Brodersen, N.H., Steptoe, A., Boniface, D.R., & Wardle, J. (2007). Trends in physical activity and sedentary behaviour in adolescence: Ethnic and socioeconomic differences. British Journal of Sports Medicine, 41, Chapman, J., Armitage, C.J., & Norman, P. (2009). Comparing implementation intention interventions in relation to young adults intake of fruit and vegetables. Psychology & Health, 24, Cohen, J. (1992). A power primer. Psychological Bulletin, 112, Dombrowski, S., & Luszczynska, A. (2009). The interplay between conscious and automatic self-regulation and adolescents physical activity: The role of planning, intentions, and lack of awareness. Applied Psychology, 58, Gollwitzer, P.M. (1993). Goal achievement: The role of intentions. European Review of Social Psychology, 4, Gollwitzer, P.M. (1999). Implementation intentions: Strong effects of simple plans. American Psychologist, 54, Gollwitzer, P.M., & Sheeran, P. (2006). Implementation intentions and goal achievement: A metaanalysis of effects and processes. Advances in Experimental Social Psychology, 38, Hagger, M.S., Chatzisarantis, N.L.D., Barkoukis, V., Wang, J.C.K., Hein, V., Pihu, M., et al. (2007). Cross-cultural generalizability of the theory of planned behavior among young people in a physical activity context. Journal of Sport & Exercise Psychology, 29, 1 19.

18 376 Armitage and Sprigg Hagger, M.S., Chatzisarantis, N.L.D., & Biddle, S.J.H. (2002). Meta-analysis of the theories of reasoned action and planned behavior in physical activity: An examination of predictive validity and the contribution of additional variables. Journal of Sport & Exercise Psychology, 24, Heckhausen, H. (1991). Motivation and action. New York: Springer. Kelly, L.A., Reilly, J.J., Fisher, A., Montgomery, C., Williamson, A., McColl, J.H., et al. (2006). Effect of socioeconomic status of objectively measured physical activity. Archives of Disease in Childhood, 91, Marcus, B.H., Dubbert, P.M., Forsyth, L.H., McKenzie, T.L., Stone, E.J., Dunn, A.L., et al. (2000). Physical activity behavior change: Issues in adoption and maintenance. Health Psychology, 19(Suppl.), Mummery, W.K., Spence, J.C., & Hudec, J.C. (2000). Understanding physical activity intention in Canadian school children and youth: An application of the theory of planned behavior. Research Quarterly for Exercise and Sport, 71, Netemeyer, R.G., Johnston, M.W., & Burton, S. (1990). Analysis of role conflict and role ambiguity in a structural equations framework. Journal of Applied Psychology, 75, Preacher, K.J., & Hayes, A.F. (2008). Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behavior Research Methods, 40, Prochaska, J.O., & DiClemente, C.C. (1983). Stages and processes of self-change in smoking: Toward an integrative model of change. Journal of Consulting and Clinical Psychology, 51, Rhodes, R. E., Macdonald, H. M., & McKay, H. A. (2006). Predicting physical activity intention and behaviour among children in a longitudinal sample. Social Science and Medicine, 62, Salmon, J., Booth, M.L., Phongsavan, P., Murphy, N., & Timperio, A. (2007). Promoting physical activity participation among children and adolescents. Epidemiologic Reviews, 29, Stamatakis, E. (2006). Obesity, eating and physical activity. In M. Bajekal, V. Osborne, M. Yar, & H. Meltzer (Eds.), Focus on health (pp ). London: Palgrave Macmillan. Trost, S.G., Pate, R.R., Freedson, P.S., Sallis, J.F., & Taylor, W.C. (2000). Using objective physical activity measures with youth: How many days of monitoring are needed? Medicine and Science in Sports and Exercise, 32, Webb, T.L., & Sheeran, P. (2004). Identifying good opportunities to act: Implementation intentions and cue discrimination. European Journal of Social Psychology, 34, Manuscript received: July 23, 2009 Revision accepted: December 21, 2009