Background

Children’s appetite, dietary intake and related phenomena, including eating behaviours, are profoundly relevant to the health and wellbeing of individuals, families, and societies. Presently, there are concerns about the state of children’s diets and weights in many countries [2,3,4,5], which are increasing children’s risk of acquiring chronic diseases later in life [4, 6]. Children’s diets and weight partly stem from children’s eating behaviours and related constructs (herein called “eating behaviours”). Indeed “eating behaviours” has emerged as one of the core components of research and understanding of children’s dietary food intake. A plethora of models and theories has emerged from a range of disciplinary perspectives that are directed to understanding elements and processes in relation to appetite, eating and their regulation. This has contributed to research on eating behaviours, especially in relation to the development of overweight/obesity [7,8,9,10,11,12,13,14,15,16,17]. These models and theories are relevant across the age-span, including for children. In parallel, efforts to measure children’s eating behaviours has continued apace. The measures have included questionnaires/self-reports as well as behavioural protocols, including laboratory-based observations. While the research on children’s eating behaviour is extensive, to advance knowledge and understanding, there is a need for more attention to be directed to the theoretical and conceptual foundations and associated definitions of measures [1].

Confusion or lack of clarity about the theoretical basis of children’s eating behaviours can contribute to poor construct definition. This, in turn can lead to difficulties in operationalisation and measurement of the eating behaviours and constructs, and therefore uncertainty in the interpretation of findings. The consequences could be many, including that research findings are less clear, the behaviour or construct of interest might not be measured, replication could be difficult and attempts to modify the target behaviour could fail. Almost 20 years ago, when discussing the field of children’s emotion regulation, Bridges et al. [18] argued that underlying the measurement of emotion regulation (implicit or explicitly stated) is a theoretical conceptualization of the physiological, behavioural and cognitive processes. They noted that there was a wide range of measures used and not enough emphasis on linking the measures with definitions of constructs and associated underlying processes. The domain of children’s eating behaviours currently seems to face some of the same issues. In particular, there seems to be limited identification or articulation of underlying theoretical frameworks.

What could be expected from a theoretical framework for children’s eating behaviours? It would probably begin with background assumptions or principles such as that children’s eating behaviours are partly biologically based, but subject to environmental influences and that the behaviours change or develop with age. A theoretical framework could provide for several overlapping domains or constructs of eating behaviours. Whether these domains or constructs are amenable to a single theoretical framework is a moot point. It is more likely there will be separate theoretical approaches to different eating behaviours and constructs. In each case, these approaches might involve separate concepts and definitions, with different assumptions and basic principles. Each would describe and explain different eating behaviour phenomena.

At present there does not appear to be either an overarching theoretical framework for children’s eating behaviour, or a set of well-articulated frameworks for each of the separate domains of eating behaviour. To help advance scholarship about the theoretical bases of children’s eating behaviours, the aims of the present review are to (a) highlight the importance and role of conceptual and theoretical foundations for children’s eating behaviour, (b) discuss the conceptual and theoretical foundations of some of the main current eating behaviours, and (c) discuss recent advances in theories and models of appetite, eating and their regulation and the associated implications for construct development, measurement and therefore understanding of children’s eating behaviour. The eating behaviours included as exemplars in the review are satiety responsiveness, food responsiveness/enjoyment of food, appetite self-regulation, food delay of gratification, eating in the absence of hunger, caloric compensation, food fussiness and food neophobia. Throughout, we were interested in processes in the development of the common questionnaire and behavioural measures, especially in relation to the role of theory in the formulation of constructs and associated measures, and then in whether and how these constructs are supported by recent theoretical models and related evidence. The review is partly informed by scholarship on the role of conceptual and theoretical foundations in social and developmental science.

Methods

Age range

The focus in the review is children aged ~ 0–12 years. These are ages during which research has examined the origins and development of eating behaviours and where there is a continuity in measures. This age period provides a substantial body of scholarship to address the three aims of the review.

Defining “eating behaviours” and related constructs

There is ambiguity in both terminology and definitions about “eating behaviours” and related constructs in this age group. Further, there are several potentially relevant conceptual models, and many relevant constructs (c.f. the list of 33 appetite self-regulation-related constructs listed in [19]). Studies of children’s “eating behaviours” have considered what, how much, when, how, and why foods and beverages are consumed but they rarely provide a definition of “eating behaviours”. Children’s eating has typically been conceived in terms of individual appetitive traits such as food responsiveness, food fussiness or satiety responsiveness [20,21,22,23]. It is also possible to conceive of many of the eating behaviours or appetitive traits in terms of underlying processes. For example, food responsiveness could be seen as a process involving the perception of food cues together with the initiation of hedonic responses (liking, preference, appeal) and the engagement of liking or wanting and possibly inhibitory controls. In addition, eating behaviours have been conceived as skills, such as being able to inhibit responses to palatable food, to make decisions about portion sizes or to regulate eating rates. In addition, consistent with the trait-state distinction in relation to personality characteristics such as anxiety [24] or impulsivity [25, 26], child eating behaviours such as eating in the absence of hunger (EAH) or other behavioural measures have been considered to be more state-like [27, 28]. Finally, some constructs involve multiple components or processes. Examples include appetite self-regulation (ASR), EAH, caloric compensation, delay of gratification (DoG) and fussiness/food neophobia, as is evident from the multiple measures used in research on each of these constructs in the discussion below.

The literature we examined shows that “eating behaviours” are complex and multidimensional and have been variously defined. Generally, they are assumed to be constitutionally based, and therefore influenced by underlying biological processes, including genetic and epigenetic factors [11, 20, 29, 30]. However, they are also shaped by environmental factors [29], although this process could be bidirectional [10, 31]. Broadly, eating behaviours have been assumed to cover food selection, consumption and regulation of eating (what, when, why), hedonics, food reward value, food avoidance/approach, reactions to food cues, rate of eating, food and eating attitudes and beliefs/cognitions, habits. This broad understanding of eating behaviours informed our selection of constructs, theoretical models and measures to include in the review.

Selection of children’s eating behaviours and constructs

For the body of the review, we chose eating behaviour measures and constructs that have been prominent in research in recent decades. We drew on integrative reviews of eating behaviours in recent decades (e.g., [27, 32,33,34,35,36,37,38] and scans of the research literature to assure that the most prominent measures were included. The selection was not intended to be exhaustive, rather to enable a discussion of prominent and representative measures. The selection involved questionnaire measures such as the Children’s Eating Behaviour Questionnaire (CEBQ) and their associated constructs, and behavioural measures such as EAH, caloric compensation, DoG and measures of food fussiness. We first searched for theoretical and conceptual foundations in the rationale and explanations for the questionnaires and behavioural measures in the original or early publications about the measures. Second, we searched more recent publications that have used the measures for comments and suggestions about theoretical foundations, including via interpretations of findings.

Results

Conceptual and theoretical foundations of behavioural constructs

In this section, we first draw on scholarship in psychological and developmental science about the importance of the conceptual and theoretical foundations of behavioural constructs, where it has been argued that these constructs should be rooted in a coherent theoretical perspective [18, 39, 40]. This could be by first theoretically deriving dimensions of the construct, as Rothbart et al. did in the case of child temperament [41]. The associated argument is that behaviours to be measured should be clearly defined, their theoretical underpinnings and scope articulated, and a description provided of how they have been operationalised. Eating behaviours are theoretical constructs devised to facilitate the description and analysis of behaviour associated with diet and eating. In this sense, they parallel other constructs in psychology and child development, such as temperament, emotion regulation, and social withdrawal. The meaning and measurement of these constructs has been enhanced by efforts to provide an account of their theoretical foundations [39, 41, 42].

With respect to questions about the role of theoretical foundations, child temperament is an instructive source of comparison with children’s eating behaviour. The conceptualisation of child temperament has a history of several decades, during which prominent proponents of different conceptualisations and associated methods and measures of child temperament emerged. Traditionally there are four main theoretical approaches to temperament (including Goldsmith, Buss & Plomin, Rothbart and Thomas & Chess) (see the roundtable discussion chaired by Goldsmith in 1987 [43] for an introduction as well as the chapter by Shiner and DeYoung [44]). In addition, Kagan and colleagues [45, 46] focused on temperamentally inhibited or uninhibited children.

Each of these theorists and approaches to temperament articulated a theoretical basis as the foundation of construct definition and measurement. In turn, each of the theoretical approaches is associated with questionnaire tools and laboratory-based or behavioural strategies for the measurement of the articulated theoretical constructs, and these have continued to be developed and refined. This has engendered research to compare the convergent and discriminant validity of different temperament questionnaires for young children to identify commonalities and conceptual boundaries [47]. A productive aspect of the richness of theoretical models and measurement strategies in temperament is that it has generated efforts to identify and agree about core dimensions of temperament [48, 49]. Individual scholars face a similar task of identifying significant dimensions of temperament when conceptualizing and designing research, such as the case of food neophobia where negative emotionality (fearfulness or withdrawal) has been determined as the aspect of temperament most relevant [50].

Scholarship on temperament has benefited from the attention to theoretical models and foundations in at least three ways: (1) questions about the construct and its dimensions have been enriched, enhanced, and enabled by the theoretical underpinning (as evident in the roundtable discussion, for example), (2) the development and evaluation of measurement instruments and procedures have been assisted by the theoretical context provided, and (3) the theoretical models have contributed to the interpretation and application of research results, including in areas such as children’s food neophobia [50]. Similar comments could be made about the benefits for scholarship of theoretical models in the field of self-regulation, where there has been a proliferation of models [51].

In contrast to this approach in the field of child temperament, as we outline below, child “eating behaviours” and related constructs seem generally to have origins in research about outcomes such as healthy diets or weight gain, obesity, or food avoidance. Often, as we argue below, the articulation of underpinning theoretical foundations has not appeared to be a priority: rather, the potential practical or applied significance of the behaviour seems to have been paramount. However, we argue that now is the time for greater consideration to be given to the underlying theoretical foundations of these constructs for both practical and theoretical benefits.

Recent discussions of social and emotional development have highlighted the importance of theoretical and conceptual foundations for measurement. Darling-Churchill and Lippman [52], for example, underscore the importance of conceptualization in order to identify the constructs in relation to social and emotional development. Central to the advancement of measurement in their view was clarity about the subdomains of social and emotional development and their associated measures. Comparable comments about measurement in other areas of psychology were made by Flake and Fried [40] who argue that even after decades of research and thousands of studies on depression, concerns remain about depression as a construct and its associated measures, and that this has limited advancements in the field. There seem to be similar issues in relation to children’s eating behaviours and related constructs where numerous subdomains of eating behaviours have been proposed, many without an articulation of their theoretical or conceptual origins and this limits both measurement and interpretation.

Meta-theories and theoretical frameworks

In addition to underlying theoretical perspectives, it is also helpful if constructs and measurement are linked and informed by broad theoretical frameworks [53] or meta-theories [54, 55]. There seems to be some consensus in developmental psychology around overlapping guiding frameworks variously labelled as a transactional model, a biopsychosocial approach, or a bioecological model [11]. We [10, 11] and others [56,57,58] have suggested that in relation to the development of children’s eating behaviour and obesity, a biopsychosocial approach (i.e., involving the interconnection of biological, psychological and socio-environmental factors) could also provide a useful general theoretical framework to guide the articulation of definitions, generate predictions, help in the design of measurements and assist in the unification of approaches and results. As outlined above when discussing definitions of children’s eating behaviour and related constructs, there is broad agreement, but not to a degree that could be described as a coherent or overarching theoretical framework. We suggest below that it might be necessary for separate frameworks for different domains of children’s eating behaviours, such as food approach behaviours, food fussiness, and appetite self-regulation.

Boundary problems in construct definitions

Boundary problems have been identified in psychological and developmental science as contributing to difficulties in construct definition and measurement. For example, problems in construct definition in relation to social and emotional development have included the “jingle and jangle fallacies” [59]. The jingle fallacy refers to the use of a single term to refer to a number of constructs while the use of different terms to refer to the same construct is the jangle fallacy. The jingle and jangle fallacies exist at least partly due to inadequacies in identifying boundaries between constructs and building measurement tools on inadequate theoretical foundations. In relation to construct boundaries, there is also the likely problem of what Darling-Churchill & Lippman [52] referred to as conceptual clutter in their discussion of early social-emotional development. Clutter arises from the lack of agreement about the definition of separate domains or dimension of social-emotional development, or in our case children’s eating behaviour. As we argue below, in the case of children’s eating behaviour, it is possible that constructs such as “food responsiveness”, “satiety responsiveness”, “eating in the absence of hunger”, “caloric compensation”, and “delay of gratification” are examples of the jingle fallacy as well as conceptual clutter.

Conceptual and theoretical foundations of selected eating behaviours and related constructs

In this section we discuss conceptual and theoretical foundations of a selection of prominent child eating behaviours and constructs. We begin with the more commonly used questionnaire/parent report measures and their associated eating behaviours together with the specific traits of satiety responsiveness and food responsiveness. The construct of appetite self-regulation is then examined, followed by the behavioural protocols of eating in the absence of hunger, food delay of gratification and caloric compensation. Finally, we discuss food neophobia/food fussiness.

Questionnaire/parent-report measures and their eating behaviours

The CEBQ [60] and the parallel eating Baby Eating Behaviour Questionnaire (BEBQ) [61] have gained wide acceptance in research on children’s eating behaviour. Other relevant questionnaires include the Children's Dutch Eating Behaviour Questionnaire (DEBQ-C) [62, 63], the child and adolescent version of the Three-Factor Eating Questionnaire (CTFEQr17) [64] and the Child Self-Regulation in Eating Questionnaire [65].

The CEBQ and BEBQ were founded on ideas about genetic influences on obesity occurring through eating behaviours [66]. In constructing the questionnaire, Wardle and colleagues [60] selected six eating style constructs that they identified from behavioural and questionnaire research as contributing to overweight/obesity, as well as interviews with parents and extrapolation from adult theory and evidence on eating. Items were generated to assess the six constructs derived from the literature that they labelled as food responsiveness, external eating, appetite/enjoyment of food, satiety sensitivity, slowness in eating, fussiness and emotional overeating together with four additional constructs identified from the parent interviews, namely emotional undereating, appetite for drinks, social eating and distractibility. Items were culled after analysis of responses from three samples of families with young children. Factor analyses were used to confirm the empirical coherence of the individual subscales and structure of the overall questionnaire. Three points about the development of the CEBQ are relevant to the present discussion: (a) the constructs were primarily empirically derived, (b) the designers provided limited discussion or justification for the labels they used for each of the constructs, and (c) the constructs have subsequently been almost reified, with limited interrogation against recent theoretical developments. We argue, however, that a critical examination of the constructs is necessary to advance scholarship on children’s eating behaviours. Below we examine the constructs of food responsiveness and satiety responsiveness in relation to more recent theory and evidence.

The development of the BEBQ was described by the authors as being designed to measure four appetitive traits “that are thought to be important for weight” (61 p. 389). The constructs and items were based on the scales from the CEBQ, the literature on milk-feeding, and interviews with mothers. The Child Self-Regulation in Eating Questionnaire was developed from a previous review of the literature on child feeding and weight status and from qualitative data from a pilot study [65]. The development of the CTFEQr17 drew on the adult version of the Three-Factor Eating Questionnaire and used interviews with children and adolescents to gauge their understanding of the adult items and to create appropriate items accordingly. The original adult version was developed by selecting items from two existing questionnaires about restraint and latent obesity plus new items based on clinical experience. These questionnaires, then, appear not to stem in the first instance from theoretical conceptions of eating behaviours.

Results from subsequent studies utilising the CEBQ and BEBQ suggest that the constructs measured with these tools could suffer from limited construct definition, arising from overlap among the subscales. For example, the satiety responsiveness and slowness in eating subscales tend to correlate strongly, as do the food responsiveness and enjoyment of food subscales, although there is some variation across studies [67, 68]. Testing in different populations has at times largely confirmed the factor structure of the original CEBQ [69], but studies with diverse samples from a range of countries [70,71,72,73,74,75,76,77] have revealed poorer fits or variations from the original factor structure, suggesting that they may not reflect common underlying processes or that there are theoretical weaknesses in the conceptualisation of the constructs. Oyama et al. [71] could not replicate the factor structure of the CEBQ or BEBQ in a Samoan sample, and speculated that the wording of the questions could at least partly explain cross-cultural differences, but also that the appetitive traits measured by the CEBQ may suffer from being theoretically indistinct. It is also possible that there are cultural or individual differences in word meanings or images generated by the questions, a potential difficulty associated with questionnaires [78].

The DEBQ [63] was reportedly based on “psychosomatic theory”, which was related to emotional eating, “externality theory”, which was related to external eating (eating in response to food-related stimuli) and the theory of restrained eating that implicates loss of contact with internal signals of hunger and satiety. All three theories were selected because of their relevance to the development and maintenance of obesity. The authors selected items from three existing questionnaires to measure three constructs: restrained, emotional and external eating. Factor analysis over several studies was used to refine the items for each construct. In this case, the DEBQ appeared to draw on an underlying theoretical framework.

Satiety responsiveness

Most of the CEBQ or BEBQ subscales have individually been the subject of research, with the satiety responsiveness subscale being an important focus [79,80,81]. In the case of the CEBQ and BEBQ, the theoretical foundation of the satiety responsiveness subscale is somewhat unclear. The CEBQ subscale combines items about both satiation (e.g., my child leaves food on his/her plate at the end of a meal, my child gets full up easily) and satiety (e.g., my child cannot eat a meal if he/she has had a snack just before, and my child has a big appetite). Satiation has been defined in terms of the processes that bring a meal to an end and satiety as involving postingestive process that inhibit further eating (e.g., via feelings of fullness and suppression of hunger) and impacts the frequency of eating [82]. The fact that this CEBQ/BEBQ subscale is a component of “food avoidance” and is correlated with subscales such as food fussiness and slowness in eating [34, 60, 83], further complicates efforts at conceptual clarity and suggests that the jingle fallacy could be relevant here.

Food responsiveness/enjoyment of food

Articulating the conceptual boundaries and theoretical distinctness between other constructs in the CEBQ and BEBQ is also difficult, especially in the correlated subscales of “food responsiveness” and “enjoyment of food”. On inspection, the items in these subscales appear to overlap and are open to questions related to the jingle and jangle fallacies as well as being subject to a level of conceptual clutter. For example, items for the food responsiveness subscale include “my child’s always asking for food” and “given the choice, my child would eat most of the time”, while items for the enjoyment of food subscale include “my child enjoys eating” and “my child loves food”. As noted earlier, there does not seem to have been an explanation for the terms chosen as labels for these constructs. Theoretically and conceptually, it would help if there was an account of the separation and overlap of these subscales with other related constructs such as hedonic hunger/hedonic eating (eating for pleasure), food reward, food reward sensitivity, the power of food, the reinforcing value of food and food cue responsiveness.

Appetite self-regulation

Appetite self-regulation (ASR) has been identified as a core construct in relation to children’s weight gain and obesity [13, 27, 37, 84, 85]. ASR has been argued as pertaining to responding to hunger, satiation, and satiety cues [13, 86]. These processes are easy to define in a general way, but difficult to measure [87]. Further, they are implicated in ASR, but are not themselves ASR [37]. Scholarship is at present engaged in efforts to clarify the construct and its measurement [37]. The diversity of approaches is illustrated by the fact that Frankel et al. [88] defined and measured ASR by the “satiety responsiveness”, “enjoyment of food” and “food responsiveness” subscales of the CEBQ. In contrast, Monnert-Patris et al. [89] developed a parent-report questionnaire drawing on the concepts of “eating in the absence of hunger” and “caloric compensation”. Tan & Hollub [65] also developed a parent-report measure of ASR. It contained a disparate collection of eight items covering different presumed elements of ASR that were selected based on earlier research on parent feeding attitudes and practices [90]. In relation to theoretical and conceptual underpinnings, none of these scales appears to have been based on an articulated or agreed conceptualisation of ASR.

In recent years, there have been several new approaches to the conceptualisation of ASR. This includes the bottom-up, top-down model [37, 91, 92] also described as a dual processing model [93] and a similar model described by Reigh et al. [94]. In these models, ASR is conceptualised as involving bottom-up approach or avoidance reactions together with top-down regulatory control. Attention has been directed to conceptualising and measuring both the bottom-up processes, including hedonic responses to food and food cues [95,96,97], aversive or avoidant reactions to food as in food fussiness and food neophobia [36, 98], and the top-down regulatory processes. In the latter case, there have been investigations of the role of inhibitory control [99,100,101,102].

The complexity of the conceptualisation and measurement of ASR is illustrated by continuing debates and the use of a diversity of measures of inhibitory control and impulsivity [25, 103,104,105,106,107,108], where it is acknowledged that they themselves are multidimensional constructs. Bennett et al. [109] argued from their results that relationships between impulsivity and eating behaviour could be measure- and respondent- dependent. In addition to inhibitory control and impulsivity, it is instructive to consider the relevance of the theoretical foundations of general self-regulation to ASR. For example, Gagne et al. [110] outline a model with self-regulation and emotion-related self-regulation having foundations in Effortful Control (EC) and Executive Function (EF). There is a body of scholarship that attempts to locate aspects of the theoretical foundations of ASR in EC and EF [19, 58, 99, 111]. In this way, the conceptualisation and measurement of ASR can be informed by established domain-general models of self-regulation and other fundamental processes in child development.

Food Delay of Gratification (DoG)

Food delay of gratification is now a widely used measure in research on ASR [19, 112,113,114]. The most frequently used measure is a choice delay task, which emerged from the research of Mischel and colleagues on self-control in the 1950s and 1960s [115,116,117]. Here a child is offered a choice between an immediate but smaller food reward and a delayed but larger food reward [117]. Marshmallows were frequently used as the food reward, but a variety of other palatable foods has been used. An alternative is the sustained delay task [118, 119] in which the child makes an implicit choice and must sustain the choice to receive the reward later. The origins of the choice DoG task were in efforts by Mischel and colleagues to develop a laboratory assessment of children’s self-control. The traditional assumption was that a tendency to delay reflects increased self-control [118].

Over subsequent decades, increased attention has been directed at theoretical underpinnings of the task, to the interpretation/significance of children’s performance and to the possible long-term significance for child development. This scholarship has yielded several theoretical possibilities and insights that raise questions about what the DoG tasks actually measure. In early research, Mischel and colleagues [116] examined cognitive and attentive mechanisms in DoG performance and other covariates. Watts and colleagues [120, 121] more recently questioned whether DoG performance is a measure of self-control and about the contribution of cognitive ability. They discuss whether DoG can be conceived as a unique construct. Giuliani & Kelly [112] investigated relationships between performance on several cognitive tasks and DoG. In this case, these relationships were stronger for a tongue delay task in which the snack was placed on the child’s tongue, and the child needed to wait for a signal to eat it versus the traditional choice delay procedure. Duran & Grissmer [118] examined relationships between delay and measures of EF. Contrary to expectations, immediate gratification was related to better EF and to better school-related outcomes. They argued that immediate gratification could be adaptive among some children in some contexts (the children in their sample were mostly Black and from low-income families). Watts et al. [120] identify additional complexities associated with processes linking DoG to later outcomes.

There seem to be three common and divergent theoretical interpretations of DoG performance: (a) delay for the larger reward arises from increased sensitivity to the reward value of food, (b) choice for the immediate reward arises from greater impulsivity, and (c) delay for the larger reward is due to better top-down regulatory control [37]. Suor et al. [122] have also argued that there is a lack of understanding of processes contributing to children’s delay abilities.

There are two further possible complications about theoretical interpretations of DoG performance. The first is that there could be differences in underlying processes and outcomes associated with food versus non-food DoG tasks [19]. A second complexity has been expressed by Hughes et al. [86]. They argued that food DoG is not a measure of ASR, satiation, satiety, or energy balance regulation, and therefore is not a measure of ASR. Overall, from the early origins of the DoG task with a relatively practical purpose, variations of the measure have emerged, and the field has spawned a rich field of scholarship about possible theoretical underpinnings. However, if food DoG is included under the umbrella of eating behaviour, there is still a need for advances in theoretical aspects of the measure. This could be elucidated somewhat by attention to the abilities or factors that underpin performance, processes that are involved in the task itself, and processes associated with linkages between DoG performance and developmental outcomes.

Eating in the absence of hunger

The early development and use of the EAH protocol is especially associated with the work of Fisher and Birch [123,124,125,126] where the emphasis was on links between parents’ restrictive access to palatable foods and children’s consumption of those foods. The EAH protocol was used as a method of assessing this association. A theoretical rationale at the time was that restrictive/controlling feeding practices limit children’s opportunities to exercise/learn self-control and that these practices thereby impede the development of ASR. Another suggestion at the time was that EAH could be indicative of children’s responsiveness to environmental cues taking precedence over children’s own fullness. From an original purpose of investigating the possible effects of restrictive feeding on the consumption of palatable food, in the subsequent two decades, the EAH protocol has evolved to the point of being claimed as representing “objectively measured appetite regulation” [127]. But there is presently no agreed definition or conceptualisation of appetite regulation or ASR to support a claim such as this.

Indeed, the “eating in the absence of hunger” (EAH) construct and measures are also examples where there appear to be limitations in theoretical and conceptual foundations. In recent years, the EAH protocol has been conceptualised variously as measuring “satiety responsiveness”, “disinhibited eating”, “satiety”, “satiation”, as well as “food cue responsivity”, “hedonic eating”, along with a combination of two (“satiation” and “food cue responsivity”). Possible links with constructs such as the “relative reinforcing value of food” are also relevant [128]. While EAH is usually assumed to be about eating beyond satiation [35, 129], other researchers have conceptualised it as being about having eaten to satiety [130, 131]. Francis et al. [132] argued that EAH involves over-riding satiation and satiety cues. There seems to be some uncertainty amongst researchers about the role of satiation and satiety in EAH. Further, it is evident that EAH is not itself a measure of either satiation or satiety. EAH seems to reflect aspects of poor ASR and is possibly related to disinhibited eating [80, 126, 133]. It could be that children who do not EAH are more sensitive to satiation cues and/or less responsive to food cues. Taken together, these points suggest that the conceptualisation of the EAH construct would benefit from further clarification.

In relation to the validity of the EAH protocol, there is some evidence of associations between the CEBQ scales and results from the EAH protocol [134]. More usual, however, is for there to be limited associations between the EAH protocol and questionnaire and self-report measures of eating behaviour (e.g., from the CEBQ) or ASR. These results prompt questions about the conceptual foundations of the EAH protocol and interpretations of results [27, 28, 135,136,137]. Furthermore, while there is agreement about the general procedures for the EAH protocol, there are variations across publications in the details. The possible implications of these variations for the measurement and interpretation of EAH is unclear. For example, there are variations in relation to the content of the prior meal, the time between the meal and subsequent snack provided and in the specific components of the snack, and the offered alternative (e.g., toys, activities, books) [37]. Madowitz et al. [137] suggested that more research is needed to determine which aspects of EAH are being measured via questionnaires and the EAH protocol, implying that the two procedures may be measuring different constructs. With regard to relationships to diet and health outcomes, there is reasonably consistent evidence about the predictive validity of measures from the EAH protocol with respect to relationships with Body Mass Index (BMI) or overweight/obesity and increased energy intake [138,139,140].

It is apparent, then, that EAH is a multi-dimensional set of behaviours and processes with several possible theoretical foundations. However, there is no coherent agreement about it conceptually or what it measures. Differences between questionnaire/report and behavioural measures add to uncertainty about the construct. It is unclear whether EAH could reflect higher levels of the hedonic value of food (the appeal, preference, or reward value of the food) or poorer capacities for inhibitory control [141, 142] or some combination of these. Overall, EAH is a widely accepted and used protocol with evidence of links to children’s energy intakes, weight gain, and BMI. Nevertheless, it is also a protocol that is in search of a theoretical foundation. Until there is more conceptual clarity, including about construct boundaries and underlying mechanisms, there will remain some confusion about what the protocol measures, how to interpret results, and how to position it in relation to other constructs such as ASR.

Caloric compensation

The term “caloric compensation” is used for a collection of protocols that developed or emerged from early efforts to provide children with a defined preload, with subsequent intake measured to determine whether or to what extent the subsequent intake compensated for the preload [143]. This was measured in terms of whether or how much the intake was adjusted or increased in response to the caloric content of the preload [143]. Since its inception, there have been multiple variations on the “caloric compensation” protocol, aligned to a variety of purposes. These purposes include whether or how the child adjusts to portion size, to a prior meal, from day-to-day, and according to the characteristics of the preload (e.g., nutrient content, energy density, and food form) [94, 144,145,146,147,148,149]. This research also shows that there have been differences in the time delay between preload and meal, and the number and characteristics of the food provided (preload and subsequent meal) including sensory qualities and palatability. In short, there is no standardised procedure or protocol. This adds to complexities associated with the conceptual or theoretical underpinning of the construct.

Underlying the set of compensation procedures are questions about what factors influence children’s ability to sense and respond to hunger and satiety cues, whether children can “perceive” calories [143] and, what the mechanisms are that are involved in compensation. It is generally assumed that ASR is somehow a component of compensation, however, it is evident that compensation does not directly measure ASR, nor is it a measure of satiety [37] although it could help in identifying general phenotypes such as low or high satiety phenotypes [150]. Presently, researchers appear to be confident about being able to measure compensation and relate this to prior intake, yet the reasons for the level of compensation and factors contributing to the compensation are less clear.

An early impetus for the caloric compensation protocol appears to have been questions about children’s abilities to adjust their intake based on caloric density, and possible implications of this ability for weight gain/adiposity [143]. Gradually, theoretical underpinnings are emerging, often tied either to characteristics of the prior intake (e.g., energy density or food form) or to possible compensation mechanisms or processes. As noted above, discrepant outcomes between different measures of eating behaviour constructs such as self-report versus observational measures continue to raise questions not only about construct validity, but also whether different measures capture different constructs. Similar questions could be raised about variations in measurement protocols for caloric compensation with implications for construct definition and measurement.

Fussiness and food neophobia

When Lumeng and Fisher [1] called for researchers to address the conceptualization of eating behaviours, they cited “fussiness” as one example where there is confusion in construct definitions and measurement. They noted that “food fussiness”, “food neophobia”, “food selectivity”, and “food rejection” are all related or overlapping concepts. Indeed, “food fussiness” is a term that appears to suffer from the jangle fallacy. A widely accepted definition of “food fussiness” is avoidance of both new and familiar foods [151], with “food neophobia” (rejection of new foods) a subcomponent of “food fussiness” [98, 152]. There are many terms that refer to the concept of “food fussiness” including “food pickiness”, “food avoidance”, “food refusal/rejection”, “choosy eating”, “selective eating”, “faddy eating”, and “finicky eating”, and a number of associated definitions [36, 38, 153,154,155]. There is also a variety of assessment tools used to measure these concepts ranging from single item parent reported perceptions to observational measures and validated questionnaires, yet the theoretical foundations of the various measures are rarely articulated.

Dovey et al. [152] highlighted the need for consistent measurement of “food fussiness”, noting that the use of measures with poor reliability and validity would lead to “further confusion and problematic theoretical interpretation” (p. 188). There remain challenges associated with measuring “food fussiness” arising from the diversity of measures [156]. Rejection of new foods is considered as an evolutionarily natural behaviour for a food safety purpose to avoid the toxic and harmful [151], however, the rejection of familiar foods does not have a clear underlying process or theoretical foundation as a component of “food fussiness” [155]. While the two constructs are theoretically distinct and are predicted by different factors, they are also often highly correlated [152], a challenge Rioux and colleagues recognised and attempted to address with the development of a “food rejection scale” [157]. A broad theoretical framework that acknowledges biological, psychological and social factors in food liking and preferences could be helpful in conceptualising and measuring fussy eating and associated constructs.

Discussion

Since the development of the measures of children’s eating behaviour and related constructs discussed here, contributions to research and theory have progressed; especially in relation to the development of theoretical models of eating behaviour that can contribute to the conceptualization and measurement of children’s eating behaviour and related constructs. Here we discuss how three of these models could help.

One theoretical advancement has been in the development of biopsychosocial models of children’s eating behaviours. For example, Chawner & Hetherington [57] outline an integrated (biopsychosocial) model for the behaviour of liking and consuming vegetables. Anzman-Frasca et al. [158] adapted Gottlieb’s [159] theory of probabilistic epigenesis (also a biopsychosocial model) to children’s food preference and behaviour. Both models highlight the need to focus on definitions and measurement of liking and food preferences aspects of children’s eating behaviour with biopsychosocial and psychophysiological dimensions of these behaviours providing the theoretical foundations. Biopsychosocial models highlight additional components of children’s eating behaviours. For example, visual, tactile and flavour exposure, behavioural and biological responses to visual, tactile and flavour experiences, positive and negative food experiences, and willingness to taste. Also, cultural contexts and practices in children’s food and eating have been identified in biopsychosocial models. These models assist in the conceptualisation of eating behaviours, for example, as states and/or traits, as developmental and as context-dependent behaviours.

A second area relates to several recent theoretical developments and associated models that derive from neuropsychological and health neuroscience perspectives, which appear to make helpful contributions to the conceptualisation and measurement of children’s eating behaviours. The theoretical developments from this perspective have come in several areas, including homeostasis, hedonics and food reward, ASR and inhibitory control, and satiety. This work integrates psychological aspects of eating behaviour and neurological/biological processes. The presentation of a health neuroscience perspective and model set out by Lowe et al. [160, 161] illustrates the advances in this area and how this theoretical perspective can inform the conceptualisation and measurement of children’s eating behaviour. They draw on Erikson et al. [162] who describe the health neuroscience perspective as one in which the brain influences, and is influenced by, physical health. In the Lowe et al. model, physical health is especially related to consumption of calorie-dense food and obesity. Lowe et al. [161] point out that there has been considerable focus on the reward system and heightened responsivity to food cues in neurobehavioural models. They suggest that emphasis also needs to be placed on prefrontal cortex structure and functionality and associated consequences for executive functions (EF), especially inhibitory control.

There is now a collection of research, models and theoretical developments from a neuroscience and neuropsychology perspective in relation to children’s eating behaviour. These include theory and research about the hedonic system, the neurobiology of food reward circuitry, cognitive control circuitry and associated attention to inhibitory control [95, 96, 101, 163,164,165,166,167]. The hedonic component of eating behaviour has also been separated into the individual elements of liking and wanting [168,169,170]. There has been considerable attention to the role of inhibitory control, EF and impulsivity in children’s eating behaviour together with associated neural mechanisms [105, 132, 142, 169, 171,172,173,174,175,176,177,178]. Theoretical models such as proposed by Lowe et al. [161] and the associated literature suggests a framework of children’s eating behaviour and related constructs that would include: consumption of calorie-dense food, behavioural and inhibitory control, reward sensitivity, EF (including EF impairments), food cue sensitivity, food habituation, evaluation of food nutritional value, impulsivity and food decision-making processes. An implication is that measurement would include biological, psychophysiological and behavioural methods.

A third relevant theoretical framework that could inform the conceptualisation of children’s eating behaviours is the Satiety Cascade, which was proposed in 1987 by Blundell and colleagues [179]. It is described as a psychophysiological framework that combines physiological, behavioural and psychological processes [82, 166, 180,181,182]. The Satiety Cascade is a theoretical framework that involves sensory and cognitive processes in the pre-ingestion phase, satiation associated with the post-ingestion phase and satiety associated with the post-absorptive phase. It links eating motivations and behaviours to cognitive and physiological processes across the three phases. There is an extensive literature on biomarkers of satiation and satiety [183] and an associated literature on the assessment of satiation and satiety using biological and behavioural measures [82, 150, 166, 180, 181, 184, 185] as well as brain imaging following exposure to food-related cues [169].

Overall, it is apparent that the evidence together with the theoretical models accumulating from biopsychosocial, neuroscience and neuropsychology perspectives plus the satiety cascade provide a foundation for advances in the conceptualization and measurement of eating behaviour in children. The focus of this evidence and theoretical models seems to have been in relation to (a) hedonics and food reward, (b) inhibitory control or regulatory processes, (c) impulsivity and (d) satiation and satiety. There is now scope for advances in construct definition and measurement strategies in each of these areas that include self-report/questionnaire, behavioural, psychophysiological, and neural measures of children’s eating behaviour and related constructs.

In the case of hedonics and food reward, Cheon et al. [186] provide a conceptual model for sweetness hedonics together with measurement implications involving multiple behavioural dimensions. The biological aspects of food hedonics/food reward also yields ideas for conceptualisation and measurement of eating behaviour, possibly with an emphasis on taste, flavour and odour liking [165]. Inhibitory control and regulatory processes together with the contrasting trait of impulsivity [100, 187, 188] have been increasingly emphasized in the conceptualization and measurement of ASR [37] and as being related to weight gain and obesity in children [189,190,191]. There is an opportunity for scholarship on inhibitory control and impulsivity to contribute further to the conceptualisation and measurement of children’s eating behaviour. It is apparent that there is scope for an increased integration of theory and evidence from these areas into efforts to conceptualise and measure children’s eating behaviour.

Future Directions

We argued the need for a renewed examination of the core constructs and measures that were included in the body of the present review. This will necessitate attention to theoretical underpinnings, construct definitions and then operationalisation. At the same time, there is a need for the development of an overarching theoretical framework that enables the linking of different eating behaviours and constructs [53]. There are some possible strategies to assist advances in conceptual and theoretical bases of children’s eating behaviours and related constructs.

As an initial step, it is important to recognise that children’s eating behaviour involves the dynamic interaction of multiple behavioural, psychophysiological, psychological, and affective systems in different environments. Conceptualisation and theoretical underpinning should encompass these multiple systems and consider them in context. A comparable “roundtable discussion” to that chaired by Goldsmith [43] on temperament and involving major theorists and approaches to children’s eating behaviour, including emerging approaches, could be helpful in guiding future work on conceptualisation and theoretical underpinnings.

Person-centred analyses of children’s eating behaviours could be helpful. In this approach, instead of treating eating behaviours as single constructs, researchers conceptualize eating behaviours in terms of phenotypes that include patterns or collections of behaviours related to differences in eating behaviours/phenotypes in sub-groups of children. This approach assumes that the person is an integrated totality whose behaviours are interwoven and interacting over time [192, 193]. It is a data driven and exploratory approach [193]. Its contribution at present seems mainly to suggest the presence of different phenotypes that involve a combination or pattern of individual eating behaviours and eating behaviour trajectories. For example, a phenotype of food avoidance trending towards low food approach in infancy, or a phenotype of high and continuing food approach in infancy [194] or a phenotype of dysregulated behaviour (low inhibitory control and high impulsivity) together with higher food approach and lower food avoidance in childhood [102]. The person-centred approach therefore moves the conceptualisation of eating behaviour from individual variables to integrated combinations of variables and trajectories [102, 193, 194]. It places an emphasis on the dynamic interplay between behaviours rather than on individual eating behaviours. We have noted that the contribution of this approach to the conceptualisation of eating behaviour will be enhanced through the development and inclusion of a wider variety and number of individual behaviours included in person-centred analyses [193].

Concept/content mapping could help reduce the conceptual clutter around the theoretical foundations of children’s eating behaviours. Content mapping draws on an expert panel to sort and group relevant constructs together with the relevant terminology and definitions identified from a systematic search [195]. Concept mapping also involves an expert panel in grouping concepts identified from published literature a priori and placing them in named constructs with definitions. Multi-dimensional scaling and hierarchical cluster analysis is then used to arrive at clusters of constructs [196]. Similar strategies with individual eating behaviours could assist in the identification of the main constructs or subdomains of eating behaviour and relationships among them, together with associated definitions.

An additional strategy to increase a focus on the theoretical foundations of children’s eating behaviours could be for journal editors and reviewers to place more emphasis on authors articulating theoretical bases. Finally, efforts at theoretical reviews to identify how theories have been applied and the associated supporting evidence could be helpful [197].

Conclusions

There still seems to be force in Lumeng & Fisher’s (1) argument that research on children’s eating behaviours should direct more attention to theoretical underpinnings, conceptualisation and measurement. The main eating behaviours that we discussed here seem to have had their origins in relatively practical or applied questions (such as can children “perceive” calories in the case of caloric compensation or whether environmental cues might override children’s sense of fullness in the case of EAH, or how parents perceive their children’s eating in the case of the CEBQ). The measures have served the field well and created a rich body of research. However, the foundation of measurement in science is dependent on the clarity of the theoretical bases and the associated construct definitions. It is in these areas where advances would benefit the field qua science. The implication is that advances in construct definition, operationalisation, and measurement and all that follows (more valid and reliable evidence, clearer targets for intervention and better assessment of intervention outcomes) are dependent on greater regard to the theoretical foundations. Practical outcomes of science, such as contributing to more healthy eating in children, are in turn dependent on the clarity (interpretability) and validity of the associated evidence.