Are Current Health Behavioral Change Models Helpful in Guiding Prevention of Weight Gain Efforts?


  • Tom Baranowski,

    Corresponding author
    1. Children's Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, Houston, Texax
      Children's Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, 1100 Bates Street, Room 2038, Houston, TX 77030-2600. E-mail:
    Search for more papers by this author
  • Karen W. Cullen,

    1. Children's Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, Houston, Texax
    Search for more papers by this author
  • Theresa Nicklas,

    1. Children's Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, Houston, Texax
    Search for more papers by this author
  • Deborah Thompson,

    1. Children's Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, Houston, Texax
    Search for more papers by this author
  • Janice Baranowski

    1. Children's Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, Houston, Texax
    Search for more papers by this author

Children's Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, 1100 Bates Street, Room 2038, Houston, TX 77030-2600. E-mail:


Effective procedures are needed to prevent the substantial increases in adiposity that have been occurring among children and adults. Behavioral change may occur as a result of changes in variables that mediate interventions. These mediating variables have typically come from the theories or models used to understand behavior. Seven categories of theories and models are reviewed to define the concepts and to identify the motivational mechanism(s), the resources that a person needs for change, the processes by which behavioral change is likely to occur, and the procedures necessary to promote change. Although each model has something to offer obesity prevention, the early promise can be achieved only with substantial additional research in which these models are applied to diet and physical activity in regard to obesity. The most promising avenues for such research seem to be using the latest variants of the Theory of Planned Behavior and Social Ecology. Synergy may be achieved by taking the most promising concepts from each model and integrating them for use with specific populations. Biology-based steps in an eating or physical activity event are identified, and research issues are suggested to integrate behavioral and biological approaches to understanding eating and physical activity behaviors. Social marketing procedures have much to offer in terms of organizing and strategizing behavioral change programs to incorporate these theoretical ideas. More research is needed to assess the true potential for these models to contribute to our understanding of obesity-related diet and physical activity practices, and in turn, to obesity prevention.


Obesity is a highly prevalent problem in the United States (1). Dramatic increases in prevalence started ∼20 years ago (2). If this trend cannot be reversed or at least halted, high levels of obesity will naturally lead to increases in the prevalence of heart disease, some cancers, type II diabetes, and other chronic diseases (3), with enormous economic and personal costs (4,5). As a result, obesity prevention has become an international priority (6).

The energy balance equation proposes that increasing adiposity is the net result of inadequate energy expenditure for the energy being consumed. On the energy expenditure side, whereas physiological and metabolic abnormalities of energy expenditure (e.g., reduced resting energy expenditure) may account for some of the obesity, most seems to be caused by inadequate physical activity. Although high levels of intake of certain nutrients (e.g., dietary fat) may be more likely to result in obesity because of metabolic factors, obesity has been associated with a number of behaviors; for example, it is positively associated with soft drink consumption (7) and is inversely associated with eating breakfast (8) and increased fruit and vegetable consumption (9). Thus, this paper focuses on physical activity and dietary behaviors.

Behavioral or social science theories or conceptual models provide the basis for understanding these behaviors. The mediating variable model has been proposed as a framework both for designing interventions and for understanding how interventions work to promote change in diet and physical activity behaviors (10,11). Mediating variables are in a cause-effect sequence between an intervention and an outcome. From this perspective, programs or interventions result in behavioral change due to changes in mediating variables (Figure 1). Changes in the mediating variables result in changes in the behavior, which result in changes in the desired physiological and anthropometric outcomes. The mediating variables are the influences on the behaviors of interest and come from the theoretical or conceptual models of behavior. Interventions are targeted at changing the variables from these selected models. Interventions are more likely to be effective if the mediating variables are strongly related to the behaviors of interest and if procedures for manipulating these mediating variables in desired directions are available (10,11). Thus, theoretical models of eating and physical activity behaviors provide possible mediating variables that are the foundation for effective behavioral obesity prevention programs. The term mediator or mediating variable can also be used to analyze how some of these variables are in cause-effect sequences from other cognitive or environmental effects to behaviors.

Figure 1.

Mediating variable model applied to obesity prevention.

Sometimes intervention programs work in one group [e.g., boys (12) or girls (13)] but not in the other, or at one level of a variable (e.g., among those at higher risk), but not at other levels (14). The variable that defines these groups or levels of different outcomes are called effect moderators, or moderating variables. It is possible that the theories to be discussed in this paper predict behavior in some groups (e.g., some ethnic groups, some education levels), but not others. The variables within which the models predict differently would also be called moderating variables.

The act of eating, whether it is a meal or a snack, is complex, involving many decisions, from whether a meal should be eaten to when to stop (Figure 2). Although these decisions may vary in how much people think about them (cognitive elaboration) for each eating event, the decisions are effectively made by the time of the eating event. Environmental, psychosocial, behavioral, and biological variables influence these decisions (15,16). In turn, each of these decisions has been demonstrated to affect some aspect of dietary intake. For example, not eating breakfast is associated with reduced energy intake (17), the foods served and the supersizing of portions at fast-food restaurants tend to lead to higher energy intakes (18), being physically active before lunch tends to increase intake of foods at school lunch (19), and so forth. The relationships of earlier behavioral decisions to consumption are often mediated by later steps; for example, eating at a fast-food restaurant affects what is eaten (e.g., the fat content). Changes in each of these steps has the potential to substantially modify dietary intake. Furthermore, several of these steps could be used strategically to modify intake. For example, one could have a small late afternoon snack, such as yogurt or a piece of fruit, to minimize intake at dinner (by preventing hunger pangs at dinner time) or eat more fruit and vegetables to minimize the intake of higher-fat alternatives.

Figure 2.

Decisions in an eating event.

The same kinds of decisions need to be made in regard to physical activity, with equivalent sequential dependencies (Figure 3). For example, one must make an initial decision to be active, which is somewhat different from eating, because although we all must eat to sustain life, we do not have to be active at an intensity or duration that provides health benefits. The decision to be active for health benefits is likely influenced by similar environmental, psychosocial, behavioral, and biological factors (20), for example, who one is with, what they are doing, and whether are they encouraging or discouraging activity. Genetic factors could influence this decision as well, because some people seem to find activity more rewarding than others. With the initial decision made, decisions then need to be made in regard to location, social support, start time, type of activity, and duration. Ecological factors (e.g., proximity to a park and friendship patterns) seem to be important here. (Ecological means the physical or social environment.) The decision to stop an activity could be cognitive (e.g., a person may have only limited time available for an activity or may have the confidence to continue walking, despite some discomfort), personality related (a person may have tried too hard and wore him- or herself out), or biological (e.g., people with low levels of fitness cannot exercise for long).

Figure 3.

Decisions in a physical activity event.

A variety of behavioral theories or conceptual models that can be related to decisions in behavioral events have been proposed. An effective intervention program would use the model-specified procedures to increase a person's motivation and resources for change along the specified processes. Different procedures may be more important at different decisions in the behavioral change process.

Some variables are moderators of the effect of a program (Figure 1). For example, some programs have been shown to promote change among girls but not boys. Thus, the sex of a child was a moderator of program effectiveness. One commonly used behavioral change procedure, goal setting, was found to be effective when children did not prefer fruit, but not when children preferred fruit (21). In this case, a child's preference for fruit was a moderator of the effectiveness of goal setting. Variables may moderate program effectiveness anywhere along the chain from program design to anthropometric change outcomes (Figure 1).

This paper reviews some of the more common behavioral conceptual models that can be used in obesity prevention program development and assesses how strongly the model is related to diet or physical activity behavior. The review of each conceptual model addresses four issues:

  • Why would a person want to change his or her behavior? That is, what is the person's motivation to change?
  • What are the personal or other resources that a person needs to change the behavior?
  • What are the processes by which behavioral change is likely to occur, and what decisions are made in performing a behavior?
  • What procedures encourage change in these mediators and, in turn, in behavior?

This review of models is historically organized. This broad overview cannot be considered a thorough analysis of all the strengths and weaknesses of each model but is an introduction to and a preliminary assessment of the potential of each model for use in obesity prevention.

Knowledge-Attitude-Behavior Model

Knowledge at some level, in some form, is a logical prerequisite to the intentional performance of health-related behaviors. The kind(s) and level of knowledge and how that knowledge is related to behavior have not been elucidated. Different types of knowledge could influence all the decisions in the eating and physical activity events (Figures 2 and 3).

The Knowledge-Attitude-Behavior (KAB)1 model has been proposed as a way of explaining the role of knowledge. The KAB model proposes that behavior changes gradually. As knowledge accumulates in a health behavior domain, changes in attitude are initiated. Over some period of time, changes in attitude accumulate, resulting in behavioral change.

The change in attitude (a construct little specified in this public health literature) seems to be the motivational force. Attitude could be some simple set of valenced (pluses and minuses) beliefs about a behavioral mechanism or something more complex. This model assumes that a person is rational and has been called the Theory of Enlightened Self-Interest (22). There has been substantial concern, however, that most people in most situations do not exhibit what would be considered objectively “rational” behavior (23). Attitudes could influence all the decisions in the eating and physical activity events.

The primary resource in this model seems to be the accumulation of knowledge. At some point, this accumulation cascades into changes in attitudes, behaviors, or both. The process by which behavioral change occurs in KAB models has not been specified. The most common procedure for promoting change by use of this model has been the provision of information, most often in the form of school curricula. In most programs described in the literature, program-specified knowledge is assessed, knowledge deficits are documented, and didactic education programs are prescribed (24), with no further attention given to behavioral change processes or procedures. For example, parents who had less accurate knowledge of dietary principles were more likely to have overweight children (25). An unstated implication was that this knowledge deficit was a cause of the obesity and could be remediated, thereby minimizing the percentage of overweight children. No research, however, has demonstrated that knowledge-based intervention programs lead to behavioral change (26).

Scientific support for the knowledge component of KAB models is weak. Within complex predictive models with very large samples, measures of knowledge were weakly related to physical activity behavior but not to changes in physical activity behavior (27). A problem has been that the concept of knowledge is not well specified. For example, in regard to diet, knowledge could include awareness that the consumption of certain foods or nutrients is related to specific health outcomes, the means of identifying the food sources of nutrients or the geographic or historic sources of foods, how to prepare certain foods, how to prepare certain foods in the context of select cuisines, or how to prepare certain foods in ways that people tend to enjoy eating them, and so forth. These types of knowledge are rarely clearly delineated or specified in any particular measure or dietary intervention program. In these specific forms, the knowledge variables begin to morph into constructs in other models. Although some of these types of knowledge could be motivating—for example, knowledge that the consumption of certain foods prevents a chronic illness could motivate a person to consume those foods—it would most likely be that the motivation would occur only if the person was concerned about that illness and perceived some personal vulnerability. Knowledge of how to prepare new foods so that people like them, how to purchase and store foods to maximize their healthfulness, or how to purchase the most healthful foods at fast-food or other restaurants may also promote behavioral change. This “how-to” form of knowledge, however, becomes difficult to distinguish from “skills,” an important concept in other conceptual models. It has been shown that knowledge partially mediates a relationship between goal setting and self-efficacy but is not related to a change in the behavior (28). Thus, knowledge may be integrated into larger conceptual frameworks to help provide some understanding of the process or mechanism of change, but increasing knowledge by itself does not seem to be useful in promoting behavioral change.

Attitudes are the frequent focus of basic social psychology research. Basic social psychology researchers, however, are primarily interested in the relationships among attitudes and the influences of behaviors on attitudes rather than the relationships from attitudes to behavior (29). There has been some recent attention to comprehensive measures of “attitudes” in regard to specific foods, such as red meat (30). Although the factors resulting from the analysis of such attitude items are interesting, the reliabilities in terms of internal consistencies for most of the factors were modest to low (suggesting that the factors were overextracted), the factors resembled specific constructs in other conceptual models (e.g., social norms), and the correlations for models that used these attitude factors to predict the intake of specific foods (e.g., red meat, fruit, and vegetables) were mostly low (<15% of the variance). The one exception was prediction of the intake of red meat (r2 = 0.29). The variable accounting for most of the predictable variance was “appreciation” (r2 = 0.20), which resembles “preference” in other literature on correlates of dietary intake (30).

We all know people who are highly self-controlled and who, when they become convinced of the value of a behavior (i.e., accumulate knowledge leading to a positive attitude), marshal their personal resources and otherwise do what is necessary to change their behavior in the desired direction. This anecdotal information suggests that knowledge by itself can lead to behavioral change among the “right” people. However, this is clearly operative in only very limited subsets of people (31); it has not been clearly determined who these people are, nor has it been determined what other resources (personal or environmental) these people might need to possess. The processes by which increased knowledge results in change or perhaps whether the increased knowledge reflects the behavioral changes made have also not been determined.

Thus, the KAB model, by itself, seems to be inadequate as a means of understanding or promoting dietary or physical activity-related behavioral change. The concepts of knowledge and attitude need to be more clearly specified conceptually and related to other variables within an overall process of change. The development of knowledge and attitude scales for each decision in the eating or physical activity event (Figures 2 and 3) may be a way of providing the needed specificity. If knowledge is revealed to be a key influence on behavioral change, procedures to change knowledge need to be more clearly specified within the context of effectively promoting behavioral changes in diet and physical activity.

Behavioral Learning Theory

In the first part of the 20th century and extending into the 1960s, the dominant paradigm for understanding behavior was learning theory. Although there were many versions of behavioral learning theory (BLT), operant conditioning was the most common. According to operant conditioning, behaviors are performed in response to stimuli, and the frequency of occurrence of the behavior after a stimulus increases if the behavior is reinforced (e.g., a child will be more physically active when mother says it is time to play if mother also gives the child a piece of candy in response to the child's activity) (32). Reinforcers that have a stronger effect on behavior are considered to have more reinforcing value. Some believed that reinforcements worked because they reduced physiological drives (e.g., hunger or thirst) (33). Within this framework, there is no need for cognitions or thoughts as explanations of behavior. (This is a very simplified version of potentially complex issues.) Learning models were first developed to explain animal behavior (34) and were later applied to human behavior (35). Within BLT, the motivation to perform a behavior is an aversive physiological drive, for example, hunger reduction. Personal or other resources are not an issue within BLT. The process of behavioral change occurs such that when a stimulus occurs, behaviors randomly occur. Reinforced behaviors (also called responses) are more likely to occur again, when the stimulus reoccurs. Within an individual, stimulus-response associations occur in memory, and these associations increase the likelihood of similar responses to the same stimuli in the future. The procedure for changing behavior is to gain control of the stimuli and reinforcers in a person's life and reinforce only desired behaviors or to present only the stimuli already linked to desired behaviors.

A modern version of BLT that has been applied to obesity is the Behavioral Economics model (36). As suggested by economics, behavior is the result of benefits and costs. Benefits are interpreted as reinforcers. The reinforcing value of behaviors or the outcomes of those behaviors differ among people. Obese people obtain more reinforcing value from food than nonobese people (37). The ability to wait longer to earn a larger reinforcer instead of taking a smaller reinforcer immediately is called self-control (which is the opposite of being impulsive). People who were dieting were found to be more impulsive in obtaining food reinforcers (38).

In a similar vein, physical activity was found to be more reinforcing among nonobese people, least reinforcing among very obese people, and reinforcing at a middle level for moderately obese individuals (39). The distance to a preferred physical activity reduced the reinforcing value of the preferred activity (40). Thus, obese people tend to find behaviors that lead to obesity (e.g., eating more and higher-fat foods and being more sedentary) more reinforcing than behaviors that lead to energy balance or energy deficit.

Research on ways in which the Behavioral Economics model could be used to prevent obesity would include finding ways to supplement the reinforcing value of low-energy foods and high-energy physical activity among those who find high-energy foods and sedentary behaviors highly reinforcing. This could be done by providing other reinforcers (e.g., money), finding balances between more and less reinforcing behaviors, or reducing the reinforcing value of undesired behaviors, for example, by increasing the distance. For example, increased physical activity was obtained by providing reinforcers for decreasing sedentary activity (41). Application of these procedures requires a highly trained individual who is exceptionally focused on controlling behavior. Some, but probably not all, parents may be able to do this. These procedures could be applied to every decision in the eating or physical activity event (Figures 2 and 3). Public health physical environmental changes to promote activity (e.g., designing a neighborhood with sidewalks and parks) could also influence the reinforcement value of a behavior.

Health Belief Model

The Health Belief Model (HBM) was the first conceptual model of behavior developed with a concern for public health issues (42). The primary HBM constructs include perceived susceptibility (a person's perceived risk for contracting an illness or health condition of concern to the researchers), perceived severity [a person's perception of the personal impact (clinical or social) of contracting the illness], perceived benefits (a person's perception of the good things that could happen from undertaking specific behaviors, especially in regard to reducing the threat of the disease), perceived barriers (a person's perception of both the difficulties in performing the specific behaviors of interest and the negative things that could happen from performing those behaviors), cues to action [the environmental events (e.g., learning that a parent had a heart attack), bodily events (e.g., aches or pains), or stories in the media that trigger perceptions of susceptibility], and in later versions of HBM, self-efficacy (a person's belief or confidence that he or she can perform a specific behavior). Early applications of HBM focused on single preventive behaviors, such as participation in a tuberculosis screening test, whereas later applications extended HBM to lifestyle behaviors, which required extended (life-long) behavioral changes.

The primary motivation to change within HBM is the level of perceived threat or the risk of a specific condition (also called “readiness to act”), that is, a combination of perceived seriousness and susceptibility. Theoretically, statistical multiplicative interactions between measures of these two concepts should be related to the behavior of interest, but few have been detected.

The primary resource for change within HBM is self-efficacy. People with greater levels of self-efficacy, or confidence, will more likely engage in a specific behavior, persist until they get it right, and maintain the behavior.

The process by which behavioral change occurs is something like the following. A person obtains certain cues, such as a television program on heart disease, that stimulate or exacerbate the person's perceived threat of the disease by either influencing perceived seriousness or susceptibility, or both. More powerful cues, cues with more personal relevance (e.g., a parent has a heart attack), or the accumulation of cues stimulates the perceived threat to some threshold at which the person decides to take action. Which action a person selects is the result of the perceived benefits, the perceived barriers that could be overcome, the costs that would be avoided, and the confidence (self-efficacy) that the person has in his or her ability to perform the alternative behavior. This conceptualization suggests that individuals select or decide from among alternative behaviors to minimize the perceived threat. The research, however, usually focuses on doing, or not doing, a single specific behavior.

Little research has addressed cues to action, a possible procedure for promoting change, in part because they may be so unpredictable or ephemeral (42). Each cue in a long diverse list of possible cues was rated on its potential to motivate changes in health behaviors, although the cues were not related to perceived threat or behavioral change (43). Internal cues, that is, feeling better physically or mentally after beginning to take action, were rated as the most likely to prompt action (43). The next influence reported to be the most likely to prompt action was the receipt of information from a personal physician based on tests that the person underwent (43). However, individuals who received their blood cholesterol levels from a screening were no more likely to change their diet than those who did not receive their blood cholesterol levels; among those with normal levels, those receiving feedback were less likely to change their behaviors to lower their levels than those who did not receive feedback (a possible reassurance effect) (44). This suggests that people may not accurately rate the importance of a cue to prompt behavior. Contrary to expectations from HBM, young adults who learned that a family member experienced a heart attack or stroke were not more likely to initiate weight loss or physical activity (45).

High proportions of adult middle-aged women from rural primary care clinics (who were objectively at substantial risk of cardiovascular disease) did not perceive themselves to be susceptible to cardiovascular disease (26%) or hypercholesterolemia (36%) (46). Among women perceiving their susceptibility to cardiovascular disease, the perception of two or more barriers to a low-fat diet was substantially related to the percentage of calories from fat in their diets and was more weakly related to their saturated fat and dietary fiber intakes per 1000 kcal (46). Self-efficacy for eating a lower-fat diet was not significantly related to the percentage of calories from fat and dietary fiber per 1000 kcal, but self-efficacy in choosing a low-fat meal in a restaurant was related to the percentage of calories from fat and dietary fiber per 1000 kcal. Perceived susceptibility to cardiovascular disease or high blood cholesterol levels was not related, however, to an intention to change their diets (46), as might be expected from the theoretically predicted relationships.

There is separate, but conceptually related, literature on perceived risk (47,48). There is substantial disagreement on how to project objective disease risk, and there are substantial discrepancies between subjective perceptions of disease risk and the objective estimates (47,48). Attempts to align subjective perceptions with objective estimates of disease risk have not been successful (49). In addition, perceived risks have been only weakly related to corresponding preventive behaviors (47). Risk communication has evolved through several stages but still has not achieved the ability to communicate the minimization of risk (50).

Fear-based communications were initially investigated as means by which behavioral change could be promoted by affecting perceived susceptibility and seriousness (51). Fear-based communications did effectively influence individuals’ perception of a threat, and the individuals selected a behavior to reduce the threat according to both the efficacy of the response (whether the behavior reduces the threat) and self-efficacy (whether the person is confident that he or she can perform the behavior) (52). A meta-analysis of this literature revealed that fear is a modestly effective behavioral change procedure (52).

The Theory of Protection Motivation (TPM), which is similar to HBM, has been proposed (53). Within TPM, a susceptibility by seriousness interaction term in regard to a person's motivation to act was detected in a within-subjects analysis, but not a between-subjects analysis (54), and the patterns of the interaction varied substantially among people. When exercise was perceived as an effective means of reducing the threat of colon cancer, people were more motivated to exercise (55). However, the threat component did not predict intentions or behavior, leaving the authors to conclude that there are limitations to fear campaigns, and effectiveness was most likely to be derived from self-efficacy and response efficacy communications (56).

Whereas it is an intuitively appealing model, many of the predictions from HBM have not been confirmed. To be useful, research with HBM and related models (Perceived Risk Model, Risk Communication Model, Fear Appeals Model, and TPM) needs to establish a perceived seriousness of and susceptibility to obesity (in regard to a broad range of medical, personal, and social outcomes), the cues to action, what behaviors are perceived to minimize risks from obesity, the benefits of the various behaviors, the barriers to performing the various behaviors, and the self-efficacies from performing the various behaviors. These concepts can be applied to each decision in the eating or physical activity event, except when to start or stop the behavior. The same relationships must be assessed among both obese and nonobese individuals in a variety of populations (people of different sexes, ethnic groups, and age groups). Procedures that affect these mediating variables and that result in the desired corresponding behavioral changes must be identified. Because children and adolescents tend to perceive themselves as immortal, the concepts may not be very useful among these age groups. HBM may become more useful as the generally perceived seriousness of and susceptibility to obesity become more severe in the society at large.

Social Cognitive Theory

Social Cognitive Theory (SCT) has been the one model most commonly used in the design of nutrition education interventions (26). Developed from social learning theory, SCT offers a comprehensive framework for understanding health-related behaviors and changing them (57). The SCT concept of reciprocal determinism proposes that behavior is a function of aspects of the environment and of the person, all of which are in constant reciprocal interaction.

The primary personal concepts of SCT for understanding behavior include skills (the ability to perform the behavior when desired), self-efficacy (the confidence that one can perform a specific behavior under a variety of circumstances), and outcome expectancies (the outcomes likely to occur from performing the behavior). Key environmental variables include modeling (learning how to do a behavior by watching someone do it and receiving reinforcement for it) and availability (whether food or physical activity equipment is present in an environment for consumption or use) (58). The primary concepts of SCT for changing behavior revolve around the ability to control one's own behavior: self-control (59). One can achieve self-control by setting specific behavioral change goals, monitoring one's own behavior through the process of change, rewarding one's self when goals are attained, and engaging in problem solving and decision making when goals are not attained to find more effective ways to attain initial goals or set new more attainable goals (57,60).

The primary motivational variable in SCT is outcome expectancies: people desire to achieve positive outcomes and avoid negative outcomes. The primary resources in SCT are skills and self-efficacy to perform the behaviors. Because skills are difficult to measure (and thereby difficult to study), a focus on self-efficacy has displaced an early emphasis on skills. Although in SCT the processes by which behavioral change occurs have not been adequately elucidated (61), the process is something like the following. When the attractiveness of the new behavior exceeds the negatives of that behavior and perhaps some threshold is reached, the person is motivated to try the new behavior. Whether the person tries the new behavior depends on the perceived self-efficacy for doing the new behavior. Early success at performing the behavior enhances the self-efficacy to do the behavior, thereby increasing the probability that the new behavior will be performed. One can learn self-control skills, a behavioral change procedure, in which increasingly more difficult behavioral change goals are set, progress toward goal attainment is monitored, and success in goal attainment is rewarded by either external incentives or an internal sense of accomplishment. If a goal is not attained, problem-solving and decision-making procedures can be used to increase the likelihood that the old goal will be attained or a new, more achievable goal will be set (57,62).

Various studies have evaluated outcome expectancies and self-efficacy related to diet, with the following results. Men perceived greater health outcomes (outcome expectancies) from eating a variety of foods than women (63). Self-efficacy predicted change in dietary risk from an intervention (64). Outcome expectancies and action self-efficacy were predictors of intention to initiate a low-fat diet, whereas intention and coping self-efficacy were predictors of eating a low-fat, high-fiber diet (65). Among adults, self-efficacy was more strongly related to intention to perform healthy eating practices than outcome expectancies (66). Modeling affected dietary intake among college students, irrespective of the level of the students’ hunger (67). Among third-grade children, however, food preferences (an immediate outcome expectancy) were the primary correlate of fruit, juice, and vegetable intake, with only a weak correlation with self-efficacy (68,69). Alternatively, self-efficacy and outcome expectancies were primary predictors of a quality of diet index among fourth graders (70).

Enjoyment can be considered an outcome expectancy of physical activity (71). More active boys and girls had higher self-efficacies, but only more active girls had higher outcome expectancies (72). Three components of self-efficacy for physical activity, but not outcome expectancies, were the most significant correlates of physical activity among teens and preteens (73). In a complex model accounting for many competing influences, self-efficacy was the most significant predictor of physical activity for 2 consecutive years (27). A structural equation model with SCT constructs accounted for 55% of the variance in physical activity among college students (74). Self-efficacy affected physical activity, primarily indirectly, through self-regulation practices (74). People increased their self-efficacy more from performing an act than from seeing a model do it or from someone persuading them (75).

Obese individuals had lower levels of moderate physical activity and lower levels of vigorous physical activity than nonobese individuals (76); self-efficacy for physical activity was lower among obese individuals, but there were no differences in outcome expectancies between obese and nonobese individuals (76).

Goal setting has been a common intervention procedure in dietary and physical activity behavioral change programs (60). In a comprehensive review of the dietary intervention literature, goal setting was identified as one of the few intervention procedures consistently associated with changes resulting in the consumption of more fruits, juices, and vegetables and lower levels of fat among adults (77). Goal setting and self-monitoring combined affected the intake of dietary fiber among college students, primarily through effects on knowledge and self-efficacy (28). Little such evidence exists for children (60). Recent findings from studies with children revealed that enhancement of dietary change by goal setting was a complex function of whether children preferred the food and how much they ate at the baseline (60). Goal setting enhanced fruit consumption among both those who did not like fruit and those who ate more fruit (60).

A substantial amount of literature exists on issues related to SCT. A comprehensive review of diverse concepts of “control” proposed that self-efficacy covers only the confidence that a person can perform a behavior and needs to be expanded to include the personal belief that performing that behavior achieves a goal (which was important to the person) (78). Similar diversity exists in the concepts of “self-control” and “self-regulation” (38,79,80,81).

SCT offers a diversity of concepts for explaining behavior and procedures for promoting change. Many of these concepts have been predictive of dietary and physical activity behaviors. Dietary and physical activity change programs based on the concepts have resulted in some changes. Research should be conducted to apply the explanatory concepts of SCT to each decision in an eating or physical activity event (Figures 2 and 3) within the context of obesity prevention; for example, how can these concepts be adapted to be predictive of targeted behaviors among those at risk for becoming obese. Procedures that consistently and substantially change those mediating variables among those at high risk of obesity must be developed and tested. Field trials of the variables that promote change in community groups should then be developed and implemented and then evaluated in terms of the extent to which they change behavior and prevent obesity and how those changes can be accounted for by the concepts and procedures of SCT. To date, the poor predictiveness of these concepts for understanding diet and physical activity among children (82) is of substantial concern. It is not clear whether the concepts have not been appropriately applied; they are too cognitive or cerebral to capture the behaviors of children, or the measures are too unreliable. Further research with SCT and children should emphasize decisions in the eating and physical activity events over which children exert the most control, and thereby, the cognitive variables could be expected to be predictive. Younger children may not exercise much control over their diet or physical activity. Perhaps environmental variables [e.g., parenting (83) or availability (58)] offer the most promise with younger children.

Theory of Reasoned Action or Theory of Planned Behavior

Perhaps the most innovative research on correlates of diet and physical activity has been conducted in the area of the Theory of Reasoned Action (TRA) or the Theory of Planned Behavior (TPB). TRA was originally formulated to explain the relation between attitudes and behavior (84). TRA proposed that people are more likely to perform a behavior when they intend to perform that behavior. The level of intention to perform a behavior is higher among those who have a more positive attitude and more of a subjective norm toward the behavior. The attitude toward the act, in turn, is an interactive function of the strength of the person's beliefs about what will happen as a result of doing the behavior (similar to SCT's outcome expectancies and HBM's pros) and the strength of the extent to which the person positively or negatively values those outcomes. A person's subjective norm, in turn, is an interactive function of the strength of the person's beliefs about whether specific people want them to do the behavior (or not) and the strength of the person's desire to please or otherwise comply with those people. Attitudes and subjective norms linearly combine to cause intention, and intention predisposes an individual to perform a behavior within the context of other influences (85).

A limitation of TRA has been that some behaviors are not under a person's control (e.g., healthier food choices may not be available at the stores where a person shops). TPB expanded TRA by proposing that intention is also influenced by perceived behavioral control (86). Perceived behavioral control, in turn, is an interactive function of control beliefs (i.e., whether there are factors that facilitate or inhibit performance of the behavior) and perceived power (i.e., the strengths of each factor to facilitate or inhibit the behavior) (85). Perceived behavioral control is believed to moderate the relationship of intention to behavior, that is, intention will convert to behavior when perceived behavioral control is high. Perceived behavioral control added a small, but consistent, amount of predictiveness to both intention and behavior (87).

The motivating factors within TRA and TPB are the positive or negative values of the outcomes of the behavior and the desire (or lack of desire) to comply with the expectations of the important people in the person's life. The resource necessary to perform the behavior is perceived behavioral control. Among those with high levels of behavioral control, high levels of intention result in behavioral change, but not otherwise. The authors of TRA and TPB do not deal with the process of performing the behavior or behavior change, nor do they specify the procedures that change those mediating variables and that in turn result in behavioral change.

A meta-analysis revealed that TPB accounted for 41% of the variance in intentions and 34% of the variance in behavior across a variety of health behaviors (88), which is good for psychosocial variables (82). For dietary behaviors, attitude predicted most of the variability, with progressively lower proportions predicted by perceived behavioral control and subjective norms (89). The level of predictiveness dropped dramatically when somewhat more objective measures of the behavior were used (90,91). The subjective estimates, which were more highly predicted, revealed more error, suggesting the models were predicting error variability (91). Among adolescents, TPB accounted for only 7% of the variance in fruit and vegetable intake (92) but 35% of the variance in soft drink intake (93). A meta-analysis of TRA and TPB in regard to physical activity supported the model constructs, but prior behavior accounted for appreciable amounts of the variance previously accounted for by TRA and TPB constructs (94). Among adolescents, TPB accounted for only 6% of the variance in moderate to vigorous physical activity when it was objectively measured (95). A variety of modifications and extensions to TPB have been proposed, for example, procedures to enhance assessment of belief salience, past behavior (or habit), moral norms, self-identity, affect, and others (96). Much of the current TPB research involves these possible enhancements (97).

TRA and TPB have typically been applied to what was eaten, but they could be applied to all decisions in an eating event (Figure 2). To relate TPB to obesity, investigators must identify the several diet and physical activity behaviors most likely related to increased adiposity within specific target groups and then develop the many components of the model in regard to each of those behaviors. Stronger ties to obesity might be obtained by specifying a concern for obesity as an outcome in the attitudinal outcomes and in the normative expectations. This, however, would violate the usual prescription in this literature of allowing target groups to specify these outcomes. Perhaps as the general populace begins to understand the public health importance of obesity, research participants will generate these items. Many investigators have great confidence in TPB and thereby believe that applying it to obesity will substantially pay off (88).

To deal with the problem of obesity, one author who has published in the TRA and TPB tradition formulated a theory of goal striving (98), wherein “desire” was inserted before intention and “trying” was inserted before behavior, and incorporated goal attainment after behavior (99). The reasons for losing or maintaining weight were analyzed and shown to differ substantially between men and women (99). These models show promise in predicting obesity related behaviors.

The lack of attention to procedures for changing components of TPB somewhat limits enthusiasm for using TPB as the foundation for intervention programs. Although attitude change procedures (29) could be used to change components of the TPB model, little work has focused on these issues within the context of diet, physical activity, and obesity.

Transtheoretical Model and Stages of Change

The Transtheoretical Model (T), as its name suggests, was originally introduced as an integration of theories and concepts from clinical psychology (100). As in clinical psychology, the focus of T is on promoting change in behavior, but several of the constructs employed (e.g., pros, cons, self-efficacy) imply a model for understanding behavior. Most of the early T research concerned smoking, an addictive behavior (100). Although T did not originate the concept of stages of change (SOC) (101), a major contribution of T has been heavy emphasis on the extent to which behavioral change occurs in stages and the explanatory concepts used to show differences in influence across stages (102). How many stages adequately capture change is subject to dispute (103,104); however, the most common set of stages include precontemplation (not thinking about change or suppressing thoughts of change), contemplation (considering change but taking no action), planning or preparation (anticipating making efforts to change and considering what behavior one will do), action (actually engaging in efforts to change), and maintenance (expending effort to retain the changes made during action). It is not clear whether changed behaviors become so habituated that one no longer expends effort to maintain them, which would require a last stage. Because the desired behaviors in our society are so many, so fluid (e.g., sodium and fat restriction in the 1980s and early 1990s to increased levels of fruit, juice, and vegetable consumption in the late 1990s), and unstable (i.e., should fat reduction be to 30% or 25% of total kilocalories), a final stage of little effort to maintain change does not seem possible for diet or physical activity (contrary to the research on smoking where no smoking is the only acceptable alternative).

Besides SOC, the primary T constructs include decisional balance or pros and cons of performing a behavior (similar to the outcome expectancies from SCT or the attitude to the act from TPB, these are the positive or negative outcomes from performing a behavior), self-efficacy (the same as self-efficacy from SCT, the confidence to perform a behavior), and processes of change (the factors that encourage or facilitate behavior change, of which there are up to 13, depending on the study).

The pros and cons of behavioral change are the motivational mechanism in T: people change their behavior to attain desired ends and to avoid undesired ends. The resources within T include self-efficacy and the processes of change. Behavioral change, as a process, is initiated by changes in cognitions, that is, the pros and cons, across the first three stages. In much research, the perception of the benefits of the new behavior exceeds the costs of the existing behavior between the first two stages (102), although this crossing may occur between the second and third stages for diet (105) and physical activity (106). Self-efficacy is likely the key factor in the action stage. Factors that maintain behavior have not been clearly specified.

The T procedures for changing behavior are likely embodied in what are called the “processes of change.” How to manipulate these processes has not been clearly delineated. There has been much attention to “tailoring” within T. In tailoring, the investigator identifies each person's initial level of pro, con, or self-efficacy, prioritizes which should or can be targeted, and provides messages that attempt to change the belief (usually attempting to increase the perceived pros of the behavior or self-efficacy to perform the behavior). The procedures for stating or framing these messages to maximize the desired effect(s) have not been elucidated and seem to be conducted on an intuitive basis.

How best to stage people has been problematic for dietary behaviors because people are often not aware of how much or even the sources of fat in their diet (82). The percentages of the population in each stage of fruit, juice, and vegetable consumption has varied by population and the algorithm used to assess the stage (107). The percentages of women in each stage, the age of the participants, and the education of participants tended to increase as stages advanced (107). The levels of fruit, juice, and vegetable consumption were highest in the action and maintenance stages (107,108); the level of dietary fat consumption was lowest in the maintenance stage and was more so among women than among men (108).

Although support has been obtained for applying T to physical activity, a number of anomalies have been identified, and a call has been made for more standardization of T constructs (109). Women were most likely (79.3%) to report being in the maintenance stage of physical activity, which is not consistent with other published literature on levels of activity (110). People can be in different stages in regard to different types of physical activity (outdoor aerobic activity vs. everyday commuting activity) (103). Although the relationships expected among the variables in T in regard to physical activity have been found in some studies, only 45% were supported in a large longitudinal survey (111). Hypothesized relationships concerning self-efficacy were the most likely to be confirmed (111). Stage-related differences in the processes of change were detected in 8 of 10 processes of change, but mean values were curvilinear (quadratic) for some processes for some groups, linear for others, and erratic for others (112). Neither experiential nor behavioral processes of change were related to increased levels of physical activity across stages, whereas behavior predicted self-efficacy, pros, and cons over time (113).

A stage-based intervention was more effective than a control intervention in promoting changes in fish and fruit and vegetable consumption but was not more effective than a KAB model of intervention plus a skills-based intervention (114). Predicted changes in mediating variables caused by the intervention occurred in regard to fish consumption, but not fruit and vegetable consumption; when they occurred, there was no difference between the KAB model plus skills- and stage-based interventions (114).

Methodological concerns have been raised in regard to SOC. The duration during which people were in the action stage varied substantially (in contrast to the maximum of 6 months usually specified in T), and people were not always aware of the foods and practices that contributed to their behavior, making it difficult to stage them (89). Several methodological issues have been raised with the research on T (115).

T has been applied to obesity. In one study, the SOC for weight control at the baseline were not related to weight loss over a 3-year period (116). Linear stage-related differences in levels of physical activity and self-efficacy were detected among adults in the overweight or obese categories (106). Among Mexican-American women, obesity treatment practices among clusters of individuals were interpreted to be consistent with SOC by use of a multi-item approach to staging (117).

Some conceptual clarity or consistency is needed in assigning people to stages. Currently, staging by algorithm produces different population clusters (with even different category names) from staging done by cluster analysis (117,118). More research is needed to apply T to each of the decisions in an eating or physical activity event (Figures 2 and 3) by comparing those who are at risk of being obese and those who are not, identifying differences in each by SOC, testing tailoring procedures among those at risk of obesity, and developing measures of processes of change appropriate to those at risk of obesity.

Ecological and Social Ecological Models

We all live in environments, called ecologies (119). Aspects of these environments are physical, which are often called ecological factors; other aspects of these environments include people, which are often called social ecologies. It has been documented for some time that environments directly affect health (120). Delineation of how the environment affects health has been more challenging because of the multilevel (e.g., regions, nations, states, cities, and neighborhoods), multistructural (e.g., physical environment, socioeconomic status, and social capital), multifactorial (e.g., diet, physical activity, smoking, and stress), and multi-institutional (e.g., local government, family, and local agency) nature of the influences and the interrelationships among these influences.

One of the paths by which the environment can affect health is through behavior: large supermarkets with fresher foods and lower prices are more likely to be located in middle-class neighborhoods (121); the availability of healthier food choices in grocery stores (122) was found to be related to the availability of healthier food choices in the home (123), which was found to be related to residents’ healthier food consumption patterns (58,124,125). Availability has been found to be related to dietary intake (126,127) and to obesity (126) in experimental studies as well. Similar paths exist for physical activity (122,127). A preliminary multilevel complex model of environmental influences on behavior integrated environmental and personal variables by taking into account cues or prompts from the environment, the behaviors of interest, personal factors in regard to that behavior, related social factors (e.g., the family and employer), related physical factors, time period, and feedback effects of the outcomes of the behavior on all the above (128).

The ecological and social ecological models have generally not included cognitive variables and thereby have no motivational variables. The resources necessary for behavioral change and the processes of behavioral change have not been clearly delineated. Possible processes include cues to prompt a behavior (129), facilitating or enhancing a behavior by having select foods or facilities available and accessible (58), and inhibiting or prohibiting a behavior by retarding or negating the availability or accessibility of foods or facilities (58). In addition, the attractiveness of a behavioral alternative could be made relatively more attractive (or unattractive) by the width and depth of alternatives available in the environment.

Procedures for change within an ecological framework have been discussed extensively (130) and could include legislation, policy changes (e.g., making only healthy choices available in vending machines) (131), ecologically sound design of neighborhoods, changes to the physical environment [e.g., adding physical activity-promoting playground equipment (132)], or placing signs to encourage the use of stairs (133). The possible social ecological processes have recently received extended discussion (134).

Analysis of the characteristics of the physical activity environment yielded three groups of factors: the home equipment, the neighborhood esthetics, and the convenience of facilities groups. All were associated with the number of minutes of walking per week and the amount of vigorous exercise (135). People were more likely to walk for exercise when the environment was perceived to be esthetically pleasing and convenient, but women also walked more when they had a pet (136). However, individual and social factors were stronger correlates of physical activity than access to recreational facilities (137). The predictions from ecological research may not always be obvious. Although children in neighborhoods of lower socioeconomic status perceived more hazards, they were also more likely to be physically active (138). Boys were more likely to be physically active when they were outside and had supervision, whereas girls were more likely to be active indoors without supervision (139). A predominantly environmental intervention that manipulated physical education at school increased physical activity among male but not female students (12).

With regard to eating behavior, children who ate the lunch provided through the National School Lunch Program (NSLP) ate a more nutritious lunch (140). Greater participation in NSLP was associated with smaller school size and a closed campus policy at lunch (141). The amount of time waiting in the lunch line was a disincentive to participation in NSLP (142). The availability of vending machines and snack bars in school lowered the levels of participation in NSLP (143). Decreasing the price of low-fat snacks in vending machines by 50% was associated with a 93% increase in their consumption (144). Moving from an elementary school where only the NSLP lunch was available to a middle school with a snack bar resulted in a 25% decrease in fruit and vegetable consumption (140).

The ecological model has generated substantial interest in light of the limited effectiveness of lifestyle change programs based on psychosocial variables (26,77). The ecological model has been applied conceptually to obesity in regard to both diet and physical activity (145,146), encompassing physical, economic, political, and sociocultural influences. Ecological and social ecological factors concern the “from where to obtain,” “where to do,” and “with whom to do” decisions in the eating and physical activity events (Figures 2 and 3). The ecological and social ecological models have much to offer obesity prevention (144,147). More research is needed on how environments enhance or constrain diet and physical activity. More conceptually refined models of how environments might affect behavior are necessary, such as whether they affect behavior directly or through as yet unspecified mediating variables. More research that manipulates the characteristics of the environment and assesses its impact on behavior is needed.

Social Marketing

Social marketing is not a model for understanding behavior but a set of procedures for promoting change in health-related behaviors. Based on the marketing principles used to sell products to consumers, social marketing has five key characteristics: 1) it attempts to change and maintain the voluntary behavior of target market members; 2) it does this by offering the target market members both a package of benefits from and a minimization of barriers to performing these behaviors; 3) the primary beneficiaries are the target market members rather than the marketers; 4) it promotes change by advocating for the target market member's self-interest (or the marketers’ perception of that interest); and 5) those who use the marketing are fulfilling their own interest as well (e.g., reducing future medical costs by preventing obesity) (148). Thus, social marketing formally incorporates the outcome expectancy, pros and cons, or benefits and barriers ideas from several of the theories discussed earlier.

The primary motivation in social marketing is perceived self-interest. The primary resource of the participant is not well specified, although some attention to self-efficacy is provided. The processes and procedures of change are that a social marketing company conducts an analysis. The target market is segmented into a small number of maximally homogeneous subgroups (market segments), and analyses are conducted to identify each of those submarket's perceived benefits and barriers. A marketing campaign is developed that attempts to prioritize the market segments to which a campaign should be targeted, presents a product that is perceived as valuable to the target market (and that is perceived as more valuable than its alternatives), minimizes barriers to the selection of the product, makes it maximally convenient to choose the product, and promotes the product (by providing information and using persuasion) in a way best understood by the target audience (148). During and after the campaign, members of the market process the information and elect to change to the new behavior because it is in their interest to do so.

Because of the nature of social marketing, there have been few experimental trials (with treatment and control groups with random assignment between groups) of social marketing campaigns. Furthermore, because social marketing has not ordinarily been done by academic investigators, there are few well-published social marketing campaigns for obesity, diet, or physical activity. Thus, it is hard to disentangle whether changes in these areas have been caused by a specific campaign or other secular changes occurring in the society. Despite this limitation, many of the ideas (e.g., market segmentation) and procedures (e.g., market assessment using focus group discussions) of social marketing have been incorporated into programs conducted by academic investigators to promote behavioral change, and they continue to offer much promise.

Interface with Biological Variables

The ability of psychosocial models to predict dietary behaviors (82) or physical activity behaviors (20) has been limited. Part of the problem has been the inherent complexity of the influences on these behaviors (149,150), the inherent unreliability of assessments of the target behaviors and of the theoretical constructs (151), and the lack of assessment of the influences of all factors related to making decisions to perform certain behaviors (Figures 2 and 3). Another limitation is that the influences of biological variables on many of these decisions are being ignored.

Biological variables have been demonstrated to influence a variety of health-related behaviors (152). A substantial literature on biological influences on dietary intake exists, although one on biological influences on physical activity does not. A three-tier system of physiological, metabolic, and neurotransmitter influences on dietary intake revealed the enormous complexity of these influences (16). A series of regulated systems in both excitatory (drive) and inhibitory (satiety) processes influence aspects of eating, and these processes work in parallel (16). Given the excitatory and inhibitory biological influences, one can propose situational, social, cognitive, emotional, and biological influences on the decisions in an eating event (Figure 4).

Figure 4.

Complex influences on decisions in eating a meal or snack.

Researchers in the biological domain seem to believe in a rigid determinism of biological influences on dietary intake. For example, genetic factors have been demonstrated to account for 65% of the variability in meal size (153), 44% of the variability in meal frequency (153), and 15% to 45% of the variability in food selection (154). It is not clear whether the biological influences account for variables that are also accounted for by psychosocial factors or whether the biological influences account for a separate set of variables, nor is it clear the extent to which the cognitive variables can override or countermand the biological influences (155). Possible issues for research combining psychosocial and biological variables are proposed in regard to leptin, genetic influences on food preference, and the selection of behaviors for change known to have the greatest biological impact.

Energy intake seems to be regulated to ensure that it is sufficient to meet energy deficits from energy expenditure, and leptin seems to play a role in this phenomenon (156). Leptin is a hormone produced by white fat tissue in direct proportion to the white fat mass (157). Leptin receptors in the hypothalamus inhibit food intake and increase energy expenditure (thermogenesis) (157). The pattern of experimental findings suggests that leptin affects intake more by inhibiting an excitatory pathway than by directly increasing satiety (16). The concept of leptin resistance has been advanced to account for the fact that obese people have the largest amounts of leptin (158). Leptin resistance seems to occur at a rate of ∼20 ng/mL (158). It would be of interest to conduct studies combining psychosocial and biological variables to test the relationship of self-efficacy to saying “no” to additional high-fat food (which may modify leptin resistance) among people who have been determined to be leptin resistant and those who are not, the extent to which environmental availability overrides leptin signals (16,158), and whether self-control procedures can be used to control dietary intakes among individuals with leptin resistance or proportionally low leptin levels (155).

Genetic factors have been shown to affect many aspects of dietary intake (159) and energy expenditure (160), but the precise mechanisms have yet to be delineated. Sensitivity to the bitter taste of certain vegetables has been known to be genetically determined for some time (161). People sensitive to these bitter tastes were less likely to enjoy such foods (162). Of interest would be research that determines whether people who are sensitive to the bitter tastes have lower self-efficacies and greater negative outcome expectancies for eating these foods, suggesting perhaps that biological phenomena may account for psychosocial variables. Alternatively, people develop tastes for bitter foods (e.g., coffee and wine). Thus, research should address the extent to which self-control procedures can enable a person to overcome this reluctance to eat the bitter foods. For example, can goal setting be used to enable people who are sensitive to the bitter taste and who do not like the vegetables consume more vegetables, or does goal setting work only among those who cannot taste the bitterness?

Another way to combine psychosocial and biological phenomena is for behaviorists to select targets for change that enhance the likelihood of obesity prevention. For example, a diet high in fruits and vegetables has been demonstrated to lower dietary fat levels and BMI (9). Increased water intake may enhance satiety (163) and displace caloric intake from soft drinks and sweetened fruit-flavored drinks (7). Some of the effects of nutrient consumption, however, may not be intuitively obvious. Although there is reason to believe that the consumption of a high-fat diet predisposes an individual to obesity (164), recent data suggest that people on a high-fat diet may compensate for the enhanced caloric intake by an increased basal metabolic rate and higher energy expenditure at night (165). The consumption of some fats (e.g., medium-chain fatty acids) may enhance satiety and energy expenditure and result in decreased fat stores and a lower body weight (166). Although the consumption of sweetened beverages has been associated with obesity in children (7), consumption of a highly concentrated sucrose solution resulted in less weight gain than consumption of an isocaloric fat or control diet (167). Physical activity may be particularly important for weight management because it burns calories in general and because it preferentially burns fat (168). Obesity prevention would be enhanced by more research elucidating the relationships among psychosocial, biological, and behavioral variables.


This paper has presented only a brief overview of several of the more commonly used models for understanding dietary and physical activity behaviors. A variety of other models have been proposed (169,170). This research and the models have limitations in regard to predicting behavior, developing and testing procedures to promote changes in the mediating variables specified by the theories, and conducting behavioral change interventions in the field.

Limitations of Research

The United States is rapidly becoming a country with a higher proportion of ethnic minorities than the white majority group. Obesity is more common among ethnic minorities, and the predisposition to obesity is detectable among children. Despite these factors, only small percentages of the theory-based research described here have been conducted with obese individuals or those at risk of obesity, ethnic minority groups, individuals of lower socioeconomic status, or children. Although some have reported a lack of differences in psychosocial variables by ethnic group (171), it seems likely that the values for individuals from different cultures will influence the mean values for many of these psychosocial variables, the relationships among these psychosocial measures will vary by ethnic group, and the low levels of literacy common in some of these groups will minimize the ability to obtain reliable measures and detect relationships. More research needs to be conducted with these groups to identify both the strengths and the limitations of these models in these groups.

At what ages these models apply and how these models can reliably be used to obtain measurements for children have not been specified. Early in childhood, children are under the strong influence of their parents and other adult caretakers (172). When and how much independence and self-control in regard to these behaviors emerge in childhood have not been specified. For example, the continued influence of parents has been reported in samples of college students (173). The competing influences of parents, peers, and personal control across ages have not been elucidated; and so the age at which these models, which are strongly based on the individual, apply is not clear.

Reliable measures are needed to detect relationships with other measures (174). Multiple-item measures of these psychosocial variables are needed to reliably assess the complex underlying concepts (82). Similar issues exist in regard to measuring dietary and physical activity behaviors, and these often revolve around multiple days of assessment (175). Some papers in the literature, however, use single- or two-item measures of constructs or overextract factors from factor analysis, leading to scales with insufficient numbers of items to reliably assess the underlying constructs (117). Even the multi-item scales had low reliabilities (176). Methodological shortcomings are of particular concern when an investigator reports the lack of a relationship, because the lack of a relationship may simply be because of poor measurement. Research that has not paid substantial attention to the reliability of the measures collected should not be published. The relationships obtained can be statistically corrected (approximated) for the unreliability of assessment (174). New models of measurement are needed to create and test more reliable and stronger measures of the constructs (177), and measures of behavior less susceptible to bias than self-reporting, on which studies persistently rely, are needed (178).

On occasion, investigators use very different procedures to measure a construct than those used by the originator, without conducting any validational research. Whereas there is a need to use brief scales to avoid respondent burden (117) and thereby avoiding alienating all citizens from participating in human research, simply dropping items from existing scales runs the risk of changing the meaning of the scale assessed (a validity issue) and of using an unreliable indicator of the construct, thereby making it more difficult to detect relationships with other variables. Investigators wishing to use substantially abbreviated numbers of items for a scale must engage in the necessary psychometric work to assure validity and reliability. Alternatively, lower respondent burden could be achieved by conducting more theoretically focused research and measuring a fewer number of scales. The latter preserves the measurement characteristics of the original scales and forces the investigator to more carefully consider the relationships to be tested.

Sampling theory proposes that samples are selected from populations to which the results can be generalized. One purpose of conducting research is to generalize back to populations. In much of this theory-based research, however, the population was never clearly specified, the selection or recruitment procedures were never clearly presented, and an assessment of possible participation bias was rarely done or reported. In addition, little has been explicated on whom the interventions have reached (179) or how representative the participants were. Attention needs to be devoted to these issues in future research.

The statistical tests used in these analyses varied from very simple two-group t tests or bivariate correlations to complex multiple-regression and structural equation models. Because some of the variables are intercorrelated and issues of mediation, moderation, suppressor effects, and interaction terms are likely, the most sophisticated statistical model appropriate to the data should be used. Human behavior is likely very complex; therefore, the presence of interaction terms seems to be likely (54,60), and these terms should be sought.

Most of the predictive research has been cross-sectional. In one of the few longitudinal studies on behavior and self-efficacy that has been conducted, the issue of direction of causality was raised: behavior caused self-efficacy rather than the reverse (113). This obviously calls for more longitudinal research, especially to establish the direction of causality of the variables. Furthermore, causality can best be established by using an experimental paradigm with random assignment of individuals to groups and attempts at manipulation of the underlying constructs. The literature in this area would also benefit from experimental basic research.

The targeted behaviors are complex. Much of the existing research takes one aspect of diet, for example, a food group (e.g., fruit and vegetables), or a factor related to a particular nutrient (e.g., the number of grams of fat in the diet or the percentage of calories from fat), and assesses differences or changes along that single dimension. Because there are upper and lower limits to calories in a diet, changes in consumption in one food or food group (e.g., fruits and vegetables) implies changes in other aspects of a diet (e.g., dietary fat and calories) (9). Future research needs to monitor these ancillary changes to detect both improvements and possible worsening of a diet. This is especially a problem in obese individuals, in whom a reduction in dietary fat does not necessarily mean a decline in the numbers of calories consumed and weight loss. Increasing vegetable consumption by adding fat in the form of butter, sauces, or gravies may not have the desired reduced adiposity effects. This more comprehensive dietary monitoring could be done with an index of overall dietary quality or by use of sentinel indicators important for obesity (e.g., total calories or the percentage of calories from fat).

There have been few tests of competing models to predict the targeted behaviors and thereby elucidate the predictiveness and use of the models. Many of the behaviors and the psychosocial variables vary by demographic characteristics (e.g., ethnicity, sex, age, socioeconomic status, and acculturation). Many of the tests of models have not taken demographic characteristics into account. It is possible that some of the predictiveness of the models in these studies was primarily caused by differences in predictors and dependent variables by demographics. All tests of models need to take demographic factors into account.

Limitations of the Theoretical Models

Each of the models or theories described here (except T and Social Marketing) has been proposed as a self-contained explanation of behaviors. The predictiveness of most of these models or theories of diet or physical activity behavior has been modest (r2 < 0.3) (82), and there has been no clear dominance of one of these models in terms of its ability to out predict the others. This suggests that there are substantial limitations to these models; although recently, research in the tradition of the TPB seems to be consistently exceeding this level and generating a variety of innovative constructs.

Much of the published literature predicting dietary or physical activity variables from one or more theories has involved taking a set of variables (usually associated with one theory, but sometimes with no theory at all) and conducting a regression analysis to assess the extent to which the variables predicted the behavior of interest. Whereas this was satisfactory in the early stages of understanding the behaviors, the major contributions henceforth will be made from critical tests of the assumptions of the theories, comparing the extent to which two theories predict the same behavior, and introducing new variables to extend the existing models to encompass new concerns. The limited predictiveness of most of the existing models suggests there is substantial room for growth of our knowledge.

It is possible that each model applies to different situations. For example, the fear-based nature of HBM may be most appropriate in patient populations—such as those diagnosed with obesity, in whom prevention of the complications of obesity is a key issue—but not among children who are blind to their vulnerabilities. A categorization of situations by types of illness and type of population may be valuable for delineation of when each model is appropriate.

Research with these models usually deals with one targeted behavior at a time (e.g., dietary fat consumption). However, people are faced with choices among behaviors (e.g., whether to have 100% fruit juice vs. a soft drink for a beverage after school or fruit or ice cream for dessert). The models should be expanded so that they pay attention to the choices that people face.

Emotions (depression, anxiety, arousal) are important aspects of behavior. Depression and anxiety may be particularly important aspects of obesity (180). Only investigators in the tradition of TPB have made substantial attempts to incorporate emotion-related variables (181).

Adherents of TPB have recognized several of these limitations and have explored a broad variety of expansions of their model (96). The movement in more basic behavioral science research, however, seems to be toward taking individual variables from more than one theory and testing their interrelationships and joint relationships with behavior (182). Perhaps investigators in the more applied behavioral science realms should also combine variables in regard to diet and physical activity. There have been several attempts to create overarching models of concepts from several of these theories (183), and these may be useful. Combinations of variables based on specific hypotheses about the behavior and informed by empirical tests would seem to be more useful in the longer term.

Given the great theoretical ferment among investigators in the tradition of TPB (87,94) and their formulation of innovative models for dealing with obesity and obesity-related behavioral change (98), it seems that the greatest benefits would be attained by pursuing research within the TPB tradition.

Procedures to Change Theoretical Variables

The interventions that have been reported have promoted change in the mediating variables by using intuitively reasonable procedures (e.g., providing information on the benefits of a behavior when a person may not believe that there are many benefits). Unfortunately, when possible mediators of intervention programs have been tested, most mediators specified by the investigators did not change, and those that did changed only a small amount (10,11). This suggests that intuitively reasonable procedures have little efficacy. Even goal-setting procedures among children did not seem to work in obvious ways (21). It would be most helpful, alternatively, to have a library of procedures that change the mediating variables, and that have been experimentally validated with various target groups. Thus, once the correlational research has identified the most influential causative variables, research should be refocused on developing and testing procedures for changing these variables, especially those that are the most effective among people at risk of obesity.

Another possible explanation for the low levels of efficacy of interventions is that changes in personal characteristics are not sufficient for behavioral change. Because environmental factors have been related to eating (124,125) and physical activity (137), it seems unlikely that interventions focused on changing the characteristics of individuals alone will result in substantial changes in diet or physical activity, with little resulting change in adiposity. Similarly, environmental variables were less related to behavior than personal characteristics were (137), and changes to environmental variables by themselves seem unlikely to result in behavioral change. Thus, a promising approach to behavioral change would seem to combine changes in the environment (e.g., enhanced availability of fruit and vegetables) with changes in individual characteristics (e.g., preferences for fruit and vegetables). Such dual-level interventions may attain synergy in promoting behavioral change.

Field Interventions

An early comprehensive review of nutrition education interventions suggested that interventions designed in light of behavioral theory were more likely to result in change than those not predicated on theory (26). A more recent review that applied screening criteria to the quality of the research to be included suggested that studies that were designed in light of theory were no more likely to result in behavioral change than those that were not so designed (77). This could be because of inadequacies in the theory (especially the strength of predictiveness); inadequate implementation of theory, especially in regard to procedures for promoting change in the constructs; or inadequate reporting of the theoretical basis of the interventions. More detail needs to be presented in professional reports of intervention. However, editors of medical journals, especially, do not like what they believe to be unnecessary reporting of detail. At a minimum, to advance the science, someone must specify what needs to be reported in papers with outcome results.


It is difficult to get external funding for large field trials of interventions. Thus, when such funding is obtained, funding should be included for process evaluation (179) and analyses of mediating and moderating variables (10,11). Process evaluation is the collection of data on a complex series of issues concerning the resources necessary to recruit participants, whether the program was implemented as designed, whether the target audience was reached and whether the target audience used the intervention, and whether higher levels of each were related to behavioral change (179). It is very important to assess these data, because they are components that help to determine whether the program made a difference in peoples’ lives. For example, attaining certain goals was not related to changes in fruit, juice, and vegetable consumption in children (21). In addition, certain recipes were not prepared; therefore, those recipes should not be provided in future programs (184).

Analyses of mediating variables provide information on whether change occurred by means of the theoretically prescribed mechanisms or processes (10,11). For example, analyses of mediating variables often reveal that 10% or less of the change in a behavior was caused by a particular set of mediators. This means that 90% was caused by unknown causes. If change is occurring without an identifiable mechanism, it will be difficult to take rational steps to further improve the intervention. Therefore, all field trials that receive funding should be held to a high standard to conduct analyses of mediating variables.

Analyses of moderating variables provide information on whether programs worked with some kinds of people but not others. For example, if it is consistently found that a program is not working with certain groups, alternative programs need to be developed for those other groups. Analyses of moderating variables should also be required of field trials of interventions that receive funding.


In some sense, behavioral research on nutrition and physical activity is in its infancy. More effective programs will not be designed until there is a better understanding of why people eat the foods they eat or participate in physical activity. The astounding advances in biological medical science today, in which pharmacotherapeutic agents are working at the level of cell receptors, is predicated on 40-plus years of intensive investment in understanding biological systems and processes. A similar investment must occur in the behavioral sciences. Substantial funding must be given to better understand basic behavioral mechanisms. More effective interventions will result when more is known about the major influences and procedures for the effective promotion of changes to these influences are available.

Although obesity is due in part to certain genetic predispositions and metabolic abnormalities, the huge increases in obesity in the past 20 years have been caused by behavioral and social ecological factors. Behavior- and ecology-based problems require behavior- and ecology-based solutions. Much interesting and important work has been done, but little has been conducted in regard to obesity prevention. The pursuit of explanatory and intervention research within the context of TPB (because of the theoretical advances), goal striving (because of the application to obesity), and social ecology (because of the apparent influences of ecological variables on diet, physical activity, and obesity) seem to offer the greatest promise for obesity prevention. Research on the relation of behavioral and biological influences on behavior and obesity also seems to be promising.


This work is a publication of the U.S. Department of Agriculture (USDA/ARS) Children's Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, Houston, Texas, and had been funded in part with federal funds from USDA/ARS under Cooperative Agreement 58-6250-6001. The contents of this publication do not necessarily reflect the views or policies of the USDA, nor does mention of trade names, commercial products, or organizations imply endorsement from the U.S. government.


  • 1

    Nonstandard abbreviations: KAB, Knowledge-Attitude-Behavior; BLT, Behavioral Learning Theory; HBM, Heath Belief Model; TPM, Theory of Protection Motivation; SCT, Social Cognitive Theory; TRA, Theory of Reasoned Action; TPB, Theory of Planned Behavior; T, Transtheoretical Model; SOC, stages of change; NSLP, National School Lunch Program.