LINKING FOOD-RELATED DECISION-MAKING CONCEPTS 1
This is the pre-peer reviewed version of the following article: [Doucerain, M., & Fellows, L. K. (2012). Eating right: Linking food-related decision-making concepts from neuroscience, psychology, and education. Mind, Brain, and Education, 6(4), 206–219. doi:10.1111/j.1751-228X.2012.01159.x], which has been published in final form at:
Eating right: Linking food-related decision-making concepts from neuroscience, psychology and education.
Matthias Doucerain1,2 & Lesley K. Fellows2
1Harvard Graduate School of Education, Harvard University, Cambridge, MA 02138, USA
2Department of Neurology and Neurosurgery, McGill University, Montreal Neurological Institute, Montreal, Quebec, H3A 2B4, Canada
Abstract
This literature review uses four dimensions to classify and compare how food-related decision-making is conceptualized and experimentally assessed in neuroscience and other disciplines: (1) food-related decision-making other than the decision of what to eat that is part of each eating episode, (2) decision complexes other than the eating episode itself, (3) the evolution of food-related decision-making over time, and (4) the nature of food related decisions. In neuroscience in particular, food-related decision-making research has been dominated by studies exploring the influence of a wide range of factors on the final outcome, the type and amount of foods eaten. In comparison, the steps that are leading up to this outcome have only rarely been discussed. Neuroscientists should broaden their historically narrow conceptualization of food-related decision-making. Then neuroscience research could help group the numerous hypothesized influences for each of the decision complexes into meaningful clusters that rely on the same or similar brain mechanisms and that thus function in similar ways. This strategy could help researchers improve existing broad models of human food-related decision-making from other disciplines. The integration of neuroscientific and behavioral science approaches can lead to a better model of food-related decision-making grounded in the brain and relevant to design of more effective school and non-school lifestyle interventions to prevent and treat obesity in children, adolescents, and adults.
Keywords: decision making, eating, food choice, food intake, food selection, obesity
Obesity and its impact
The increase in obesity over the last several decades has been the subject of a vast literature (see e.g. Finucane et al., 2011; Flegal, Carroll, Odgen, & Curtin, 2010; Ogden, Carroll, Curtin, Lamb, & Flegal, 2010; Olds et al., 2011). The negative impact of this trend is well established, particularly for very high levels of BMI and for early life - and therefore likely long-term - exposure (Fontaine, Redden, Wang, Westfall, & Allison, 2003); as a corollary, an increase in the level of childhood obesity, a condition that is likely to persist into adulthood, is of particular concern to individuals and societies alike (Reilly et al., 2003; Lobstein, Baur, & Uauy, 2004). The consequences of obesity can be classified as physiological, psychological and economic: Physiological risks of obesity include increased rates of a wide range of medical conditions (see e.g. Dixon, 2010; Kopelman, 2000; Must et al., 1999; Pi-Sunyer, 1993). Psychological risks include stigmatization by others, resulting in an increasingly negative self-image in obese individuals (Dietz, 1998; Lobstein, Baur, & Uauy, 2004). Possibly driven by this psychological impact, obesity appears to be associated with lower income, lower marriage rates and lower socioeconomic status.
Educational relevance
The increasing prevalence of obesity and its impact on health and wellbeing are of particular relevance to education because adult eating patterns are established during childhood and adolescence (Dietz, 1997). Much eating during this period happens in schools, and both the foods available in school as well as the school social context strongly influence what children will eat (Schanzenbach, 2009; Vereecken, Bobelijn, & Maes, 2005). The most immediate and widespread type of risk in children is psychological in nature: Obese children are the subject of stereotyping by both teachers (Neumark-Sztainer, Story, & Harris, 1999) and peers (Janssen, Craig, Boyce, & Pickett, 2004). Obese children also miss more days of school than their normal-weight peers, a phenomenon also documented for children and adolescents with other chronic diseases (Schwimmer, Burwinkle, & Varni, 2003). Obesity is associated with lower grades, placement in special education or remedial classes, fewer years of schooling, and lower academic performance overall (Taras & Potts-Datema, 2005). The negative impact of obesity does not stop there: Even for students with equal credentials, obesity is associated with lower college acceptance rates (Lobstein, Baur, & Uauy, 2004).
Intervention landscape
Because of obesity’s prevalence and its massive impact, diverse interventions abound. Among existing interventions, surgery appears to be the most effective, but it is too invasive to become the treatment of choice for large parts of the population (Albrecht & Pories, 1999; Bruce & Mitchell, 2011; Sjöström, 2000). Pharmacological interventions are less invasive but less effective than surgery, and also have negative side effects (Cooke & Bloom, 2006; Finer, 2002). Large-scale environmental interventions such as changes in the food supply e.g. through taxation would address the likely causes of the obesity pandemic, but they are very difficult to implement politically (Glanz & Mullis, 1988; Story, Kaphingst, Robinson-O’Brien, & Glanz, 2008).
This leaves lifestyle interventions as the main choice for most societies. Unlike the other intervention types, lifestyle interventions are first and foremost educational interventions – their goal is to teach participants to live healthier lives. As such, many lifestyle interventions target children in schools (Gortmaker et al., 1999; Caballero et al., 2003; Neumark-Sztainer, Story, Hannan, & Rex, 2003), and they commonly employ educational tools such as health education curricula, nutritional games, and physical education classes. Unfortunately they often produce inconsistent results, and their effects tend not to persist in the long run (Jakicic et al., 2001; Lobstein, Baur, & Uauy, 2004).
The situation is not quite as dire as this assessment might suggest: we know that the effectiveness of lifestyle interventions is generally higher if they combine multiple approaches (Gonzalez-Suarez, Worley, Grimmer-Somers, & Dones, 2009), are theory-based (Bluford, Sherry, & Scanlon, 2007; Sharma, 2007), long-running (Lobstein, Baur, & Uauy, 2004), individualized (Brown & Summerbell, 2009; Stice, Shaw, & Marti, 2006), and if the focus is on prevention (particularly during childhood and in school and household settings) rather than treatment (Barlow, 2007). Indeed, the likely most promising way to improve interventions in a fragmented area of research such as food-related decision-making in humans is to compare and integrate existing effective interventions. However, if researchers and societies alike want to avoid a drawn-out and inefficient trial and error process towards intervention improvement, they face a number of considerable hurdles along this path to integration.
Hurdles to effective interventions
Comparing and integrating multiple interventions necessitates a range of comparable intervention characteristics (Opp & Wippler, 1990; Seipel, 1999). The theoretical basis underlying each intervention, if properly used, could provide these intervention characteristics. However, many interventions do not make their theoretical basis sufficiently explicit. Even if they do, they do not necessarily have the same object of investigation: While e.g. all interventions should be referring to some theory of changing food-related decision-making in humans, some only refer to a theory of human behavior change (not specific to eating) and others to a theory of food human decision-making (not particularly concerned with its change). Another hurdle is that theories underlying existing interventions are rarely general in nature; they are mostly specific. Both when comparing across general or specific theories, different levels of accuracy of individual theories as well as conflict between them can cause problems. Theory-comparisons involving specific theories introduce the additional problem-types of irrelevance (when theories do not address food-related decision-making that is problematic from the perspective of a given individual or group) and non-comparability (if two or more specific theories do not address at least some of the same food-related decision-making). Lastly, comparison and integration of interventions - even if compatible otherwise - can also be hampered by theories being expressed in different disciplinary ‘languages’ such as neuroscience or psychology and requiring (at times rather tedious) translation from one to the other.
Literature review goal
The present literature review aims to help overcome some of these hurdles to intervention integration by contributing to the development of an improved theory of food-related decision-making in humans. To do so, it reviews the fragmented and multidisciplinary literature on food-related decision-making in humans in a systematic way, extracting how food-related decision-making is conceptualized and experimentally assessed in individual publications as well as across contributing disciplines. To effectively deal with this multidisciplinary literature, the review selected one discipline – neuroscience – as its anchor, and compared its conceptualization and assessment to that of all others as a whole.
What is particularly exciting about this approach is that both neuroscience and behavioral science of food-related decision-making in humans have produced highly evolved theories that only partially overlap: The neuroscience literature has worked its way towards food-related decision-making from the gut up and is very concerned with the influences of particular neurochemicals (Berthoud, 2002). As such, it has a long history of informing surgical and pharmacological obesity interventions. However, so far it has rarely – if ever – been brought to bear in the lifestyle intervention arena even though much is known about e.g. the neuroscience of self-control and reward processing that could be employed quite readily.
Behavioral science, on the other hand, often informs lifestyle interventions. The behavioral science literature, in particular as exemplified by the keyword ‘food choice’, has started out by observing free-living humans and emphasizing a significantly wider range of influences on food intake (Buttriss et al., 2004). But while its list of influences is impressive, it is at times at risk of being perceived as no more than a laundry list for which mechanisms and interdependencies are far less well worked out than for the neuroscientific model. If successfully integrated, the neuroscientific and behavioral science approaches would result in a much improved model of food-related decision-making grounded in the brain that would support the design of the more effective school and non-school based lifestyle interventions to prevent and treat obesity in children, adolescents and adults that so many desire.
It should be noted that the disciplinary anchoring is not meant to imply primacy of neuroscience over other disciplines in understanding food-related decision-making in humans and battling obesity – far from it! The biggest rewards of this approach to theory integration will only be realized if any resulting insights into general theory development are translated back into the languages of the contributing disciplines to equally inform their research efforts.
Search Strategy
The literature review was conducted through an electronic search performed on Thomson Reuters’ Web of Knowledge. In the first step, all experimental neuroscience publications on food-related decision-making in humans, a total of just over 100 articles, were reviewed. In this and the following step, qualifiers are represented by a set of search terms (see table 1 below) that were jointly applied to either the title or topic fields. The results of this review of neuroscience publications on human decision-making were then contrasted with an equivalent review of experimental non-neuroscience publications on food-related decision-making in humans. Given the more than 2,000 hits of the unrestricted step two search and the intention to only expand on the existing detailed review in step one, (Times Cited) > 40 was used as an additional qualifier, resulting in a more manageable impact-weighted selection of around 150 additional publications for review.
Food:Title = ("Meal*" or "Food*" or "Eating" or "Obes*")
Decision-making:Title = ("Choice*" or "Selection*" or "Decision*" or "Judgment*")
Human:Topic = ("Human*" or "Men" or "Participant*" or "Patient*" or "Adult*" or "Child*" or "People*" or "Household*" or "Man" or "Women" or "Adolescent*" or "Student*" or "Parent*" or "Family")
Neuroscience:Topic = ("Brain*" or "Neuro*" or "Cortex" or "Cortical")
Table 1: Sets of search terms for each qualifier.
Publication selection
In the course of the search term selection process, a much wider list of search terms was considered for inclusion. For the given sets of search terms, each of the two searches produces a high proportion of relevant publications. An additional title, abstract, and if necessary full text level screening process ensured the elimination of all irrelevant papers nevertheless captured by these searches, such as the occasional papers not concerned with humans, papers which address decision-making by groups of agents, such as governments or corporations, rather than individuals, non-experimental papers, or papers exploring decision-making processes other than those directed at the consumption of food. However, this high relevance came at the cost of excluding search terms (and deciding on the application of search terms to the title rather than the broader topic level) which in addition to capturing proportionally many more irrelevant papers always also produced at least a few additional papers of relevance. To deal with this problem of non-inclusion at least in a partial manner, the database searches were supplemented with manual reviews of the reference lists of included relevant articles.
Analysis
The literature on food-related decision-making in humans is not a well-behaved or uniform literature: It is impossible to capture in a more or less complete form and without too many intrusions by a reasonable set of search terms. This is partially due to its multidisciplinarity, but the most important contributing factor is the complex and diverse nature of food-related decision-making itself.
Human food-related decision-making has been discussed using a range of different labels, including but not limited to ‘dietary choice’, ‘food choice’, ‘food selection’, or ‘nutrient selection’. The most broadly accepted – albeit rather restrictive – position seems to be that all of them are concerned with deciding what to eat. This position is exemplified by Buttriss’ et al. statement that food choice is defined as the “selection of foods for consumption” (Buttriss et al., 2004, p. 334). In its idealized form this decision-making process is taking place e.g. when we pick bite-size buffet items from a plate to put them into our mouth and eat. However, in the reality of most of our lives, settings where a fixed number of relatively discrete food choices exist and choosing and consuming co-occur in close temporal proximity and disconnected from other aspects of our lives represent only a fraction of eating situations overall. The majority of eating situations are significantly more complex (see figure 1 for a ‘simple’ example of this complexity) – and at least a subset of the literature reviewed here reflects this.
Figure 1: Food-related decision-making example. The figure shows the evolution over time of the decision to consume a bowl of cereal for breakfast on a given morning (specific eating episode), including the various production, acquisition, preparation, and clean-up decision complexes along with their interdependencies.
While there is no single best way to appropriately broaden the definition of food choice to include food-related decision-making processes more generally, a number of dimensions of complexification appear particularly promising. Four such dimensions will serve as a rubric to classify and compare how food-related decision-making is conceptualized and experimentally assessed in all publications reviewed:
(1) Food-related decision-making other than the decision of what to eat that are part of each eating episodes:
This dimension is evaluated by assessing to what extent any given publication explores or addresses any one of a number of individual decisions, including whether, where, when, with whom, how long, how, how much, and why.
(2) Decision complexes other than the eating episode itself:
This dimension is evaluated by assessing to what extent any given publication explores or addresses any one of a number of individual decision complexes, including production, acquisition, transport, preparation, serving, storing, digestion, and clean up.
(3) The evolution of food-related decision-making over time.
(4) The nature of food-related decision-making.
Food-related decision-making research in neuroscience
The usage of food-related decision-making vocabulary in neuroscience in humans is relatively rare – only around 100 among the tens of thousands of neuroscience publications concerned with food intake, food perception and related processes contain these words in their title or abstract according to our Web of Science search. Even though small, this group of articles is quite diverse in terms of neuroscientific methods used, and includes patient studies, neurochemical studies, and neuroimaging methods such as functional magnetic resonance imaging, positron emission tomography, and electroencephalography – it thus represents a rather well-balanced sample of the discipline at large. Among this small group of articles, food-related decision-making is generally equated with either food intake, food preferences, or some combination thereof. Both of these aspects are also important foci beyond the articles that refer to food-related decision-making in an explicit sense.
When food-related decision-making is interpreted to mean food intake, which is the case in the majority of reviewed neuroscience papers, researchers have the choice between measuring it directly through identifying and weighing all food items consumed by participants (see e.g. Blundell & Rogers, 1980; Born et al., 2010; Greenwood et al., 2005; Moller, 1986) or indirectly through questionnaires and self-reports (see e.g. Atkinson, Waggoner, & Kaiser, 1988; Breum, Moller, Andersen, & Astrup, 1996; Cohen, Yates, Duong, & Convit, 2011; Lammers et al., 1996; Pijl et al., 1991; Roberts, 2008) – and both methods are used with roughly equal frequency and sometimes jointly. Consumption measurement is without doubt the more precise of the two methods, but it also significantly increases study complexity and in the process creates eating scenarios which are likely to differ in fundamental ways from those typically encountered by participants – except in the case of institutionalized participants with externally controlled food provision. Self-reports of food intake can take the form of 24-hour or longer-delay recalls or of food frequency and diet history questionnaires (Block, 1982; Acheson, Campbell, Edholm, Miller, & Stock, 1980). They have been criticized for underreporting intake, particularly in overweight and obese participants (Schoeller, 1990), and for dependence on question wording, format and context (Schwarz, 1999), but they can be improved through the use of cross-checks and interview or evaluation by experienced dieticians.