Child Meal Microstructure and Eating Behaviors: A Systematic Review

Abstract

Observational coding of children’s eating behaviors and meal microstructure (e.g., bites, chews) provides an opportunity to assess complex eating styles that may relate to individual differences in energy intake and weight status. Across studies, however, similar terms are often defined differently, which complicates the interpretation and replication of coding protocols. Therefore, this study aimed to compile methods of coding meal microstructure in children. To limit bias and ensure a comprehensive review, a systematic search was conducted in January, 2021 across three databases (PubMed, PsychInfo, Web of Science) resulting in 46 studies that coded at least one meal-related behavior in healthy children (i.e., no medical/psychological disorders) who were able to self-feed (i.e., no spoon-, breast-, or bottle-feeding). While the majority of studies had good interrater reliability, the details reported about study foods and the clarity of the definitions used for behavioral coding varied considerably. In addition to reported microstructure behaviors, a non-exhaustive review of individual differences was included. While few studies reported individual differences related to age and sex, there was a larger literature related to weight status that provided evidence for an ‘obesogenic’ style of eating characterized by larger Bites, faster Eating and Bite Rates, and shorter Meal Durations. However, some studies may not have been optimally designed or powered to detect individual differences because they did not set out a priori to examine them. Based on this systematic review, best practices for the field are recommended and include reporting more details about foods served and coded eating behaviors to improve reproducibility. These suggestions will improve the ability to examine patterns of individual differences across studies, which may help identify novel targets for intervention.

1. Introduction

Observational approaches to assessing children’s eating behaviors at a meal provide an opportunity to capture specific traits (e.g., eating speed, bite size) that may be related to overall consumption and weight status. These methods allow for characterization of complex and dynamic patterns of eating behavior within a meal, which have been shown to differ by weight status and disordered eating in adults (Kissileff & Guss, 2001; Ohkuma et al., 2015; Westerterp-Plantenga, 2000). Similarly, observational studies in children have shown that faster eating rates and larger bite sizes are characteristic of an ‘obesogenic’ eating style that is associated with greater consumption and obesity (Fogel et al., 2017b, 2017a). Characterizing patterns of behaviors within a single meal or eating episode originated from studies in animals that provided insight on the behavioral or physiological mechanisms of food intake control (Davis, 1989; Kissileff & Guss, 2001). These characteristics behaviors, such as licks, bites, chews, and swallows were referred to as meal ‘microstructure’ by John D. Davis (1989). In addition to the behavioral coding of discrete eating behaviors (e.g., bites or chews), the term ‘microstructure’ also encapsulates cumulative intake curves that are assessed through the continuous measurement of food intake at a meal (Kissileff et al., 1980). Given evidence that individual differences in meal microstructural patterns relate to the susceptibility for overconsumption (Fogel et al., 2017b; Langlet et al., 2017) and obesity (Berkowitz et al., 2010; Fogel et al., 2017a; Llewellyn et al., 2008), we aimed to systematically review methodological approaches to assessing meal microstructure in children.

Despite the benefits of using observational approaches to assess meal microstructure in children, there are also challenges. Although the foundation laid by animal models has led to a fairly consistent set of behaviors that are considered to reflect meal microstructure in humans (e.g., ‘bites’, ‘chews’), studies differ in their approach to defining and coding these behaviors. Since the same behavioral terms are often used (e.g., ‘bites’), differences in behavioral definitions and coding protocols across studies may not be immediately obvious. For example, the term ‘bite’ has been defined as any food touching the mouth (e.g., Fernandez et al., 2018) or as food being chewed and swallowed (e.g., Drabman et al., 1977). These differences may contribute to discrepancies in findings across studies, therefore, a systematic approach to compiling the existing behavioral definitions used to code meal microstructure is needed to understand the diversity of methods in the field and establish best practices for studies moving forward.

While prior reviews have summarized the general practices for observational coding of eating behavior in children (e.g., Pesch & Lumeng, 2017), the primary aim of this study was to systematically review the literature to compile methods of coding meal microstructure in children. Incorporating suggestions established by Pesch and Lumeng (2017), this study characterized aspects of study design such as study location (e.g., laboratory, home, etc), population (e.g., age, sex), and meal characteristics (e.g., meal partners, foods served, time of day). Additionally, to gather insight on the terminology used to define microstructure behaviors across studies, we compiled and compared the reported definitions and interrater reliability across studies. Lastly, we reported any individual differences in meal microstructure related to sex, age, or weight status to identify whether differing behavioral definitions contributed to any discrepancies across studies. Together, this review compiled the methods used to assess meal microstructure in children and established best practices for behavioral researchers in this field.

Methods

The systematic review followed the 2020 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist (Page et al., 2021), which is included in Supplementary Materials.

2.1. Systematic Literature Search

2.1.1. Literature Search.

In order to identify studies that coded child meal behaviors, the following search terms were applied to PubMed (from 1960 to 2020), PsychInfo (from 1978 to 2020) and Web of Science (from 1956 to 2020): meal, eating, food intake, eating in the absence of hunger, EAH, eating rate, bite size, satiation, energy compensation, short-term compensation, microstructure, universal eating monitor, intake curve, bite* in conjunction with the terms video, observational coding, or coding. Searches were limited to include only studies conducted in humans and articles written in English or translated in English (see Supplementary Materials for exact searches entered into each database). In order to identify relevant studies from the search, articles were excluded in five distinct steps: 1) removal of all duplicates across the three databases; 2) review of titles; 3) review of abstracts; 4) review of references to identify any missed articles; and 5) final review of all full-text journal articles (Figure 1). All steps of this process were independently completed by two research assistants (authors MC and ED), who had 80% or greater agreement on inclusion and exclusion decisions. Any disagreements were resolved by author ALP.

Figure 1.

Search results following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines

2.1.2. Exclusion Criteria for Titles.

Studies were excluded if it was clear from the title that the article: 1) was not published in English and a translated version could not be located; 2) was focused at the cellular or molecular level; 3) was not a peer-reviewed or empirical article (e.g., manual or dataset); 4) was clearly outside the scope of the review (e.g., tick or dog bites); 5) was a non-human sample (e.g., panda eating behavior); or 6) did not involve coding of child eating behavior (e.g., medical imaging of swallowing).

2.1.3. Exclusion Criteria for Abstracts.

In addition to the established exclusion criteria for titles, articles were excluded if it was clear from the abstracts that the study population was: 1) not pediatric (i.e., older than 18-years-old); 2) not healthy (e.g., diabetes, eating disorder, developmental disorder, etc.); or 3) the entire sample was not able to self-feed (e.g., included participants who were spoon-, bottle- or breast-fed).

2.1.4. Inclusion Criteria for Articles.

Full-texts of articles were reviewed for all the established exclusion criteria in addition to the following inclusion criteria. Studies must have included at least one child behavior coded during an eating paradigm (e.g., meal, snack, eating in the absence of hunger) that met the following criteria: 1) the behavior was relevant to eating or the meal context (e.g., bites, meal duration, distraction); 2) the behavior was autonomous (i.e., child could complete without help); and 3) the behavior did not require parental involvement (e.g., positive maternal interaction would not be included because it would require coding of both mother and child behavior).

2.2. Coding and Data Extraction

Once the final set of studies was determined, all articles were reviewed in detail and the following information was extracted: 1) sample characteristics (e.g., average age, sex distribution, country of origin); 2) quality of food reporting and behavioral definitions; 3) meal or food characteristics (e.g., time of day, foods served, location, meal patterns); 4) behavioral coding protocols (e.g., live versus video coding, reliability statistics, definitions of child behaviors); and 5) any reported association between child behaviors and child sex, age, or weight status. This was completed by authors OR and MC and was double checked for accuracy by author ALP.

2.2.1. Development of Study Quality Rating Scales.

Prior to coding and data extraction, a rating scale was developed to determine the quality of reporting for study foods which ranged from 1 (most detailed) to 5 (least detailed; Table 1). The best score (1 – most detailed) was based on recommendations for reporting foods in studies of ingestive behavior (Hetherington & Rolls, 2018). The remaining categories were defined by decreasing the required specificity of food reporting with the lowest score (5 – least detail) including no details on the foods consumed (see Table 1; see Supplementary Table S1 for reported foods served in each study). Following the recommendations for food reporting (Hetherington & Rolls, 2018), reporting of specific food brands was also noted (see Table 3).

Table 1.

Criteria for Ranking the Reporting of Study Foods

Score	Criteria	Examples
Most Detail	Specifies all foods served and at least one of the following: grams, calories, energy density	mac and cheese 120 grams, grapes 50 grams, etc.
1
2	Specifies all the foods served without reference to grams, calories, or amount served	apple, cheese, bread, tortilla
3	Describes meal or eating paradigm without specifying all food components	sandwich, fruit, percent of meals including bread
4	Provides protocol instructions that describe which food items should be present during the meal but does not confirm or report which items were actually present	parents instructed to include a fruit and vegetable in meal
5	Reference to meal or eating paradigm without any detail of the food served	school lunch or regular dinner
Least Detail

Table 3.

Study Characteristics Micrstructure Behavior

Study Characteristics					Meal Characteristics				Food Reporting		Microstructure Behavior
First Author (year)	Country	N	Male (%)	Age (SD) or Range	Meal	Provided Foods	Location	Others Present at Meal	Brands Reported	Description Score	N Reported (Computed)	aDefinition Avg. Score
1. Addessi (2005)	USA	27	26%	2.5 – 5.2 yr	Before Lunch	Study	School	Adult model	Yes	1	2	4
2. Akland (2000)	USA	6	66%	1.8 (0.8) yr	Any	Study	Home and childcare	Research Staff	No	3	2	3
3. Asta (2016)	USA	209	51%	21 – 33 mo	Lunch	Study	Home	Mother	Yes	1	b0	NA
4. Barkeling (1991)	Sweden	43	62%	11 yr	Lunch	Study	Laboratory	Alone	No	1	1 (2)	5
5. Bergmeier (2016)	Australia	34	43%	3 (0.8) yr	Lunch	Home	Home	Mother	No	5	b0	NA
6. Berkowitz (2010)	USA	61	50%	4 and 6 yr	Lunch	Study	Laboratory	Parent	No	1	2 (2)	4.5
7. Briones (2018)	USA	228	50%	71 (8) mo	NSF	Study	Laboratory SEP	Mother	Yes	1	b0	NA
8. Cooper (2004)	England	116	53%	56 (4) mo	Lunch	Home	Homes	Mother	No	5	b0	NA
9. Czaja (2011)	Germany	74	43%	10.8 (1.9) yr	Dinner	Home	Home	Family	No	5	2 (1)	3
10. Domoff (2017)	USA	220	54%	4.3 (0.5) yr	Dinner	Home	Home	Mother	No	5	b0	NA
11. Drabman (1977)	USA	120	NSF	1st – 6th grade	Lunch	School	School	Cafeteria peers	No	5	3 (1)	3
12. Drabman (1979)	USA	60	NSF	1.6 – 6 yr	Lunch	School	School	Cafeteria peers	No	5	3 (1)	3.3
13. Epstein (1976)	USA	6	50%	7 yr	Lunch	School	School	Cafeteria peers	No	3	2 (2)	2.5
14. Fernandez (2018)	USA	226	48%	71 (9) mo	NSF	Study	Laboratory SEP	Mother	Yes	1	2	2
15. Fisher (2003)	USA	35	49%	4.0 (0.5) yr	Lunch	Study	Laboratory or class	4–5 study peers	No	1	1 (1)	5
16. Fisher (2007)	USA	75	59%	5.6 (0.5) yr	Dinner	Study	Laboratory	3–4 study peers	No	1	1 (1)	5
17. C1Fogel (2017a)	Singapore	386	52%	4.5 (0.2) yr	Lunch	Study	Buffet in Laboratory	Mother	Yes	1	6 (6)	2.3
18. C1Fogel (2017b)	Singapore	386	52%	4.5 (0.2) yr	Lunch	Study	Buffet in Laboratory	Mother	Yes	1	5 (1)	1.8
19. C1Fogel (2018)	Singapore	195	51%	4.5 (0.2) yr	Lunch	Study	Buffet in Laboratory	T1Mother	Yes	1	5 (6)	1.8
20. C1Fogel (2019)	Singapore	255	50%	5.9–6 yr	Lunch	Study	EAH in Laboratory	T2Alone	Yes	1	2 (1)	2
21. Geller (1981)	USA	S148 S28	S150% S2NSF	3rd – 5th grade	S1Before Lunch S2Lunch	School	School	Alone	No	5	S12 (1) S23	S14.5 S23.7
22. Gisel (1988)	Canada	98	49%	6.1 (0.2) yr	Before Lunch	Study	School	Research Staff	No	2	2 (1)	1.5
23. Israel (1983)	USA	59	46%	8 – 10 yr	Lunch	School	School	Cafeteria peers	No	5	4	4.75
24. Israel (1985)	USA	60	40%	7.8 – 12.3 yr	Lunch	School	School	Cafeteria peers	No	5	3 (1)	5
25. Johnson (2018)	USA	109 *twins	47%	59 (18) mo	Lunch	Study	Buffet in Laboratory	Mother	No	1	1	3
26. Keane (1981)	USA	20	NSF	5th grade	Lunch	Study	School	Alone	No	1	3	2.3
27. Klesges (1983)	USA	14	50%	24 (9) mo	Dinner	Home	Home	Family	No	5	3	4.25
28. Klesges (1986)	USA	30	50%	30.5 mo	Dinner	Home	Home	Family	No	5	2	5
29. Koivisto (1994)	Sweden	50	50%	5.1 (1.5)yr	Dinner	Home	Home	Family	No	5	2	4.5
30. Kremer-Sadlik (2015)	USA, France	16	NSF	7–11 yr	Dinner	Home	Home	Family	No	3	1	2
31. Laessle (2001)	USA	80	55%	10.5 (1.7) yr	NSF	Study	Taste-test in Laboratory	Alone or mother	No	1	1 (2)	UEM
32. Langlet (2017)	Sweden	41	46%	16.7 (.4) yr	Lunch	School	School	Cafeteria peers	No	5	3 (1)	UEM
33. Llewellyn (2008)	England	254	40%	11.2 (0.6) yr	Lunch or Dinner	Study	Home	Twin	No	3	3 (3)	2
34. Luchini (2017)	USA	24 26	46% 54%	3–5 yr	Lunch	Study	Home Childcare	Parents Peers	No	1	1	5
35. Lumeng (2006)	USA	71	55%	5.1 (1.1)yr	NSF	Study	Laboratory	Mother	Yes	1	1	5
36. Marston (1976)	USA	32	NSF	6 – 14 yr	Lunch	School	School	Alone or peers	No	5	4 (1)	4.75
37. McKenzie (1991)	USA	S140 S219	S137% S2NSF	S14 – 8 yr S24 – 9 yr	Lunch	Home/School	Home/School	Parent or Family	No	5	1	5
38 Moding & Fries (2020)	USA	60	50%	24.7 (6.8) mo	All meals & snacks	Home	Home	Family	No	5	1	5
39. O’Connor (2009)	USA	144	51%	4.8 (0.7) yr	Dinner	Home	Home	Family	No	5	b0	NA
40. Pesch (2018)	USA	50	52%	72 (8) mo	NSF	Study	Laboratory SEP	Mother	Yes	2	1	1
41. Rendall (2020)	England	67	58%	3 (1) yr	NSF	Study	Home	Mother	Yes	1	1	2
42. C2Saltzman (2018)	USA	74	50%	41–51 mo	NSF	Home	Home	Family	No	5	1	1
43. C2Saltzman (2019)	USA	109	52%	21 (3) mo	Dinner	Home	Home	Family	No	5	1	1
44. Sharps (2015)	England	38	0%	15.4 (1.9) yr	Lunch	Study	Laboratory	Parents	Yes	1	1 (1)	4
45. Tan (2018)	USA	91	57%	T127 (1) mo T233 mo	Lunch + EAH	Home/Study	EAH at home	Lunch: Family EAH: Alone	No	4	3	3.5
46. Tovar (2016)	USA	48	50%	3.3 (1.1)yr	All meals & snacks	Home	Childcare	Childcare provider; peer NSF	No	5	b0	NA

Avg: Average; EAH: eating in the absence of hunger; M: male; mo: months; NSF: not specified; SEP: standardized eating protocol; UEM: universal eating monitor; yr: years.

^aaverage score for reported definitions of meal microstructure behaviors.

^bOther, non-microstructure behaviors are reported in Supplementary Table S5.

^{C1, C2} Studies that share overlapping samples of participants due to using the same cohort are marked ^C1GUSTO cohort and ^C2STRONG kids cohort

^{S1, S2} Studies with multiple independent, relevant samples of participants are marked ^S1study 1 and ^S2study 2

^{T1. T2} Studies with multiple relevant timepoints for data collection of the same sample are marked ^T1Time 1 and ^T2Time 2

Similarly, a rating scale was developed to determine the quality of reported definitions of microstructural behaviors. Since methods of coding meal microstructure in children have not been previously compiled, the rating scale for behavioral definitions relied on the definitions included in the final set of studies. The most detailed definitions represented the best score (1 – clearest) with the remaining categories defined by decreasing specificity of reported deviations with the lowest score (5 – least clear) reflecting definitions that simply restated the term (e.g., defining Bites by ‘bites’; see Table 2). After the categories were defined, were evaluated by authors OR and AP for whether there was enough detailed to execute and/or reproduce the behavioral coding. The best score (1 – clearest) was thorough enough for the reader to execute and reproduce behavioral coding. In contrast, the lowest score did not provide enough detail to execute or reproduce behavioral coding.

Table 2.

Criteria for Ranking the Reporting of Child Eating Behaviors Definitions

Score		Executable	Reproducible	Example of Level of Detail Required
Clearest 1	Very Detailed Definition	Yes	Yes	Taking multiple bites nonstop (i.e., nibbling) counted as two bites; bites must be separated by period of no food touching mouth; etc.
2	Detailed Definition	Yes	No	Taking of solid food into the mouth, subsequently chewed and not spat out; must chew and swallow
3	Moderately Detailed Definition	Somewhat	No	Taking of solid food into the mouth followed by a chew
4	Broad Definition	No	No	Placing food in mouth
Least Clear 5	No definition	No	No	Bite

Both the food reporting and behavior definition quality scales were developed by authors OR and MC (nutrition research assistants, B.S.) under the guidance of authors ALP (psychologist, PhD) and KLK (nutritional scientist, PhD). Ratings were completed by authors OR and MC with any disagreements resolved by author ALP. For microstructure behaviors, each reported behavior was independently rated for the definition provided with the average score across behaviors was computed for each study (Table 3). Scores and definitions for each behavior are listed in Supplementary Tables S2–3.

2. Results

3.1. Search Results (Figure 1)

In total, there were 4,618 titles after collating all search results across the three databases. There were 980 duplicates removed, which resulted in 3,638 titles. After applying the exclusion criteria to titles, 2,396 were excluded. The remaining 1,242 article abstracts were then reviewed with the full set of exclusion criteria, resulting in the exclusion of 1,106 articles. To ensure that no articles were missed in the search, the reference lists of the remaining 136 articles were reviewed for any titles that were not in the original search. This resulted in an additional 50 articles identified from the reference lists. In the final step, the full texts of 186 articles were reviewed for both the exclusion criteria and the inclusion criteria. After excluding 140 articles, the final set of consisted of 46 articles (Table 3).

3.2. Study Characteristics (Table 3)

Since the primary aim of the current systematic review was to compile methods of coding meal microstructure in children, all studies that met inclusion and exclusion criteria (see section 2.1 Systematic Literature Search) were included regardless of the primary aims study. Overall, 18 (39%) studies had primary aims related to meal microstructure behaviors while 17 (37%) had primary aims that focused on social influences or interactions during the meal (e.g., child-parent interactions). To see the aims of each study included in the review, see Supplementary Table S1.

In addition to varying in primary aims, the included studies differed in location and meal context. The final set of studies were conducted across 9 different countries with the majority conducted in the United States (n = 31, 67%). Studies were conducted in four different settings: schools, the child’s home, childcare centers, and laboratories. All studies published before 1990 (n = 8, 17%) were either school- or home-based with the earliest laboratory-based study conducted in Sweden in 1991 (i.e., Barkeling et al., 1991). Overall, there was roughly equal number of laboratory- (n = 15, 33%) and home-based (n = 17, 37%) studies and fewer school- (n = 11, 24%) and childcare-based (n = 3, 7%) studies.

3.2. Participant Characteristics (Table 3)

The number of participants included in studies ranged from 6 to 386 children. Home- and laboratory-based studies had larger sample sizes (Home: mean n = 96; Laboratory: mean n = 149) than school- or childcare-based studies (School: mean n = 41; Childcare: mean n = 48). While one study limited participants to just females (i.e., Sharps et al., 2015), the average sex distribution was 48.4% male and was consistent across study locations. The average age of children ranged from almost 2-years-old to 16-years-old. Childcare- and home-based studies had younger participants (Childcare: mean = 3.3 years; Home: mean = 4.5 years) than laboratory- or school-based studies (Laboratory: mean = 6.6 years; School: mean = 8.3 years). Therefore, while sex distribution did not differ by location, children participating in laboratory- and school-based studies tended to be older than children participating in home- and childcare-based studies.

3.3. Meal Characteristics (Table 3)

Most studies specified a time for the meal or eating protocol with 53% using a lunch meal (n = 24) and 20% a dinner meal (n = 9). Fewer studies used pre-lunch snacks (n = 2, 4%) or allowed the observation to occur at any meal (n = 3, 7%). Sixteen percent (n = 7) provided no information on meal or snack timing. Regardless of where the meal occurred, it was uncommon for the child to eat alone (n = 3, 7%). The majority of studies (n = 25, 54%) had children eat with at least one family member with 28% of studies specifically requesting the mother to be present (n = 13). In 22% of studies the participant ate in the presence of their peers at school or childcare (n = 8, 17%) or with other children enrolled in the laboratory study (n = 2, 4%). Therefore, the majority of the studies examined child eating behavior outside the laboratory (i.e., school or home; n = 30, 65%) during lunch when the child was eating with their family or peers.

One of the primary ways home- or school-based studies differed from laboratory-based studies was the ability to experimentally control the food served. Of the studies conducted outside a laboratory, 70% (21 out of 30 studies) did not control the type or amount of food served. Rather, children consumed a ‘typical’ family meal at home or a ‘standard’ cafeteria meal at school. In contrast, 100% of the laboratory-based studies controlled the amount and type of food served. Despite having control over the foods provided, the quality of food reporting for studies that provided meal foods ranged from 1 to 4 with only half the studies reporting which brands of food were used (n = 12). Therefore, regardless of where data collection was conducted, there was inconsistent reporting of meal- and food-related information across studies. The foods reported for each study are summarized in Supplementary Table S1.

3.4. Behavioral Coding

There were differences in the mode of behavioral observation based on the location of the study. Almost two-thirds of the studies (n = 28) used videos to record the meal or eating protocol with the majority of these conducted either at home (n=13, 46%) or in the laboratory (n=11, 39%). The remaining studies coded eating behaviors during live observation of the meal (n = 14, 30%) or used a universal eating monitor (UEM) to continuously measure meal weight (n = 3, 7%). The seventy-one percent of studies that used live observation implemented alternating windows of observation (e.g., 10- to 30-seconds) and recording (5- to 30-seconds; n = 10), which means the entire meal was not coded continuously. Less common was continuous live coding of behaviors (n = 4, 29%) or coding only pre-determined segments of the meal (n = 1, 7%; e.g., first 5 bites, 3-minute segment). Studies that used video recordings coded behaviors continuously with the exception of one study that used point-coding every 10 seconds (i.e., behavior was either occurring or not occurring at each point; Addessi et al., 2005). Coding methods are detailed by study in Supplementary Table S1.

3.4. Meal Microstructure Behavior Definitions

Any behaviors that directly contributed to or characterized the process of energy intake were included in our definition of meal microstructure. These included point behaviors, which are behaviors that do not have a duration and are measured using frequency (e.g., sips, bites, chews), and duration behaviors, which are measured as the length of time the behavior is occurring (e.g., oral exposure time, meal duration). Additionally, the calculation of any proportions, averages, or rates of microstructure behaviors were included as computed behaviors. When multiple articles were included on the same longitudinal study (i.e., GUSTO cohort), behavioral definitions were assumed to be consistent across articles unless otherwise noted. The overall quality of definitions provided is described for each microstructure behavior with breakdown by study location provided in the Supplementary Materials. Additionally, because they are outside of the scope of the current review, a summary of any non-microstructure behaviors that were coded as part of the reviewed studies is included in Supplementary Table S5.

3.4.1. Point Behaviors.

Across articles, the following point microstructure behaviors were coded: 1) bites, 2) chews, 3) sips, 4) swallows, and 5) general point behaviors related to meal intake (n = 2), referred to as General Eating Behaviors. All microstructure and general eating behaviors had good reported interrater reliability (see Supplementary Materials). Examples of point behavior definitions for each rating score are presented in Table 4 with detailed definitions for each study presented in Supplementary Table S2.

Table 4.

Microstructure Point Behavior Definition Scores and Examples

Study Index from Table 3	Score	Example Definition
Bites
15, 16, S121, 23, 24, 35, and 36	5	Bite
6, 12, S221, 25, 27, 29, and 44	4	Placing/taking food into mouth
1, 11, 14, and 26	3	Any taking of solid food into the mouth followed by a chew
13, C117 – 20, 41, and 45	2	A piece of food being cut off, not spat out, and subsequently chewed and swallowed
9, 33, and 40	1	Discernable bites of solid food; multiple, nonstop bites (nibbling) was two bites. Crumbs only counted if the quantity of food could have been eaten with a fork
Chews
23 and 24	5	Chew
11 and 12	4	Any masticatory movement of the jaw
C117 – 20	3	Movement of jaws after bite and results in swallowing
22	2	One downward-and-upward movement of the chin
Sips
18, 23, 24, and 36	5	Sip or Drink
45	4	Drink crossed lips
13, S221, and 26	3	Contact and then removal of drinking utensil to lips
11 and 12	2	Vessel or straw touch the lips followed by a swallowing motion
Swallows
C117 – 20	2	End of chewing followed by a movement in the esophagus
General Eating Behavior
37	5	Ingests food at meal (yes/no)
30	2	Scale: 0 (did not touch) - 3 (ate a second helping)

Note: see Supplementary Table S2 to view definitions for each study

^{C1, C2} Studies that share overlapping samples of participants due to using the same cohort are marked

^C1GUSTO cohort

^{S1, S2}Studies with multiple independent, relevant samples of participants are marked ^S1study 1 and ^S2study 2

3.4.1.1. Bites.

A total of 29 (63%) studies coded Bites, making it the most commonly coded microstructure behavior. With the exception of two studies that used a UEM, the studies were almost evenly split between video (n = 16, 55%) and live observation (n = 11, 38%). The majority of these studies were conducted in the laboratory (n = 11, 38%) or school (n = 12, 41%), with only 6 conducted at home (20%). Twenty-four percent (n = 7) of studies did not provide a definition beyond the term Bite or Mouthful (score of 5), 41% (n = 12) provided a definition that was not sufficient to execute coding (score of 3 or 4), and 31% (n = 9) provided a detailed definition that could be used to code the behavior (score of 1 or 2).

3.4.1.2. Chews.

Of the 9 (20%) studies that coded Chews, 4 (44%) were laboratory-based studies from the longitudinal GUSTO cohort and 5 (56%) were school-based. All of the GUSTO cohort studies and one school-based study used video recordings while the remaining four school-based studies used live observation. Twenty-two percent (n = 2) of studies did not provide a definition beyond the term ‘Chew’ (score of 5), 56% (n = 5) provided a definition that was not sufficient to execute coding (score of 4 or 3), and 1 study (11%) provided a detailed definition that could be used to execute coding (score of 2).

3.4.1.3. Sips.

A total of 10 studies (22%) coded Sips. All the school-based studies that coded sips (n=8, 80%) used live observation while the remaining two studies used video (20%; Home: n = 1; Laboratory: n = 1). Forty percent (n = 4) of studies did not provide a definition beyond the term Sip or Drink (score of 5), 40% (n = 4) provided a definition that was not sufficient to execute coding (score of 4 or 3), and 20% (n = 2) provided a detailed definition that could be used to execute coding (score 2).

3.4.1.4. Swallows.

All four studies (9%) that coded swallows were laboratory-based studies from the GUSTO cohort. The definitions of Swallow from these studies (Table 4) received a score of 2, indicating the definition had enough detail to execute coding.

3.4.2. Duration Behaviors.

Across articles, the following duration microstructure behaviors were coded: 1) latency to ingestion, 2) oral exposure time, 3) active mealtime, and 4) meal duration (Table 5). All duration behaviors had good reported interrater reliability (see Supplementary Materials). Examples of duration behavior definitions for each rating score are presented in Table 5 with detailed definitions for each study presented in Supplementary Table S3.

Table 5.

Microstructure Duration Behavior Definitions Scores and Examples

Study Index from Table 3	Score	Example Definition
Latency to Ingestion
1	5	Latency to ingestion
14	1	Time from start of meal to the first bite
Oral Exposure Time
C117 – 19	1	Total time food was in mouth (from bite to swallow)
Eating Duration
6, 27, and 28	5	Active mealtime, time eating
32	1	Clean eating time, removing time for visits to buffet or drink refills
Meal Duration
2, 4, 27–29, 34, 36, 38, and 40	5	Duration of meal
9, S1&S221, 32, and 33	4	Absolute length of meal
13, C117–19, 22, 26, C242, C243, and 45	1	from the first bite or food/drink first touched lips until the final swallow

Note: see Supplementary Table S3 to view definitions for each study

^{C1, C2} Studies that share overlapping samples of participants due to using the same cohort are marked

^C1GUSTO cohort and ^C2STRONG kids cohort

^{S1, S2}Studies with multiple independent, relevant samples of participants are marked ^S1study 1 and ^S2study 2

3.4.2.1. Latency to Ingestion.

Only two studies (4%) coded Latency to Ingestion, both of which used video recording. Of these two studies, the laboratory-based study provided a definition sufficient to reproduce coding (score of 1; i.e., Fernandez et al., 2018), while the school-based study provided no definition beyond the term Latency to Ingestion (score of 5; i.e., Addessi et al., 2005).

3.4.2.2. Oral Exposure Time.

The three studies (7%) that coded Oral Exposure Time were from the laboratory-based GUSTO cohort, used video recording, and provided a definition that could be used to replicate coding (score of 1).

3.4.2.3. Eating Duration.

A total of 4 studies (9%) coded Eating Duration. Three were school-based studies (75%), of which one used a UEM and two used live observation. The remaining study was a laboratory-based study that used video recording. While the study that used the UEM provided a definition that could be used to reproduce coding (score = 1; i.e., Langlet et al., 2017), the remaining 3 studies (75%) did not provide a definition beyond the use of phrases like Eating Duration or Active Mealtime (score = 5).

3.4.2.4. Meal Duration.

Twenty-three studies (50%) coded meal duration. Sixty-one percent of the studies used video recordings (n = 14), 30% used live observation (n = 7), and 9% used the UEM (n = 2). Most of the studies were done in the home (n = 9, 39%) or school (n = 7), with fewer conducted in laboratory (n = 4, 17%) or childcare settings (n = 2, 9%). Thirty-nine percent of studies (n = 9) did not provide a definition beyond the term Meal Duration (score of 5), 22% (n = 5) provided a definition that was insufficient to execute coding (score of 4), and 39% (n = 9) of studies provided a definition that could be used to reproduce coding (score of 1).

3.4.3. Computed Behaviors.

There was a total of eight computed microstructure behaviors reported across studies: 1) bite size, 2) bite rate, 3) oral exposure per bite, 4) chew rate, 5) sip rate, 6) eating rate, 7) deceleration or change in eating rate, and 8) active meal time. Definitions and for computed behaviors are presented in Table 6.

Table 6.

Microstructure Calculated Behavior Formulas

Study Index from Table 3	Formula
Bite Size
15, 16, C117, C119, 31, and 32	$\frac{\mathit{\text{Total}}\mathit{\text{Intake}},\mathit{\text{g}}}{\mathit{\text{Number}}\mathit{\text{of}}\mathit{\text{Bites}}}$
23 and 36	Size of mouthful was recorded on a scale using one (smallest) to five (largest)
33	Estimate of the number of bites taken to eat a sandwich quarter: small ≥ 7 bites, average 4–6 bites, or large 1–3 bites
Bite Rate
6, 13, 44	$\frac{\mathit{\text{Number}}\mathit{\text{of}}\mathit{\text{Bitess}}}{\mathit{\text{Meal}}\mathit{\text{Duration}},\mathit{\text{min}}}$
9, S121, 33	$\frac{\mathit{\text{Number}}\mathit{\text{of}}\mathit{\text{Bitess}}}{\mathit{\text{Eating}}\mathit{\text{Duration}},\mathit{\text{min}}}$
Oral Exposure per Bite
C117, C119	$\frac{\mathit{\text{Oral}}\mathit{\text{Exposure}}\mathit{\text{Time}},\mathit{\text{min}}}{\mathit{\text{Number}}\mathit{\text{of}}\mathit{\text{Bites}}}$
Chew Rate
C117, C119	$\frac{\mathit{\text{Number}}\mathit{\text{of}}\mathit{\text{Chews}}}{\mathit{\text{Total}}\mathit{\text{Intake}},\mathit{\text{g}}}$
C119	$\frac{\mathit{\text{Number}}\mathit{\text{of}}\mathit{\text{Chews}}}{\mathit{\text{Oral}}\mathit{\text{Exposure}}\mathit{\text{Time,s}}}$
11 12 24 and 36	$\frac{\mathit{\text{Number}}\mathit{\text{of}}\mathit{\text{Chews}}}{\mathit{\text{Number}}\mathit{\text{of}}\mathit{\text{Bites}}}$
22	$\frac{\mathit{\text{Meal}}\mathit{\text{Duration}},\mathit{\text{s}}}{\mathit{\text{Number}}\mathit{\text{of}}\mathit{\text{Chews}}}$
Sip Rate
13	$\frac{\mathit{\text{Number}}\mathit{\text{of}}\mathit{\text{Sips}}}{\mathit{\text{Observation}}\mathit{\text{Time}},\mathit{\text{min}}}$
Eating Rate
C117 – 20	$\frac{\mathit{\text{Intake,g}}}{\mathit{\text{Oral}}\mathit{\text{Exposure}}\mathit{\text{Time}}}$
4, 31, 33	$\frac{\mathit{\text{Intake,g}}}{\mathit{\text{Eating}}\mathit{\text{Duration}}}$
6	$\frac{\mathit{\text{Intake}},\mathit{\text{kCal}}}{\mathit{\text{Meal}}\mathit{\text{Duration}}}$
Change in Eating Rate
4	$\frac{1\mathit{\text{st}}\mathit{\text{half}}\mathit{\text{intake}},\mathit{\text{g}}-2\mathit{\text{nd}}\mathit{\text{half}}\mathit{\text{intake}},\mathit{\text{g}}}{\mathit{\text{Total}}\mathit{\text{Intake}},\mathit{\text{g}}}$
33	Bite Rate (bites/min) for each quarter of the meal
32	quadratic coefficient (k) of cumulative intake curve (i.e., y = kx2 + lx)
31	Area under the cumulative intake curve $\left(\frac{g}{s^2}\right)$
Active Mealtime
C1 17, C119,	$\frac{\mathit{\text{Oral}}\mathit{\text{Processing}}\mathit{\text{TIme}},\mathit{\text{min}}}{\mathit{\text{Meal}}\mathit{\text{Duration}}}$

g: grams, kCal: kilocalories; min: minutes, s: seconds

^C1GUSTO cohort; studies share overlapping samples of participants

^S1Studies with multiple independent, relevant samples of participants are marked ^S1study 1

3.4.3.1. Bite Size.

There were nine studies (20%) that estimated Bite Size. Most of these studies were done either in the laboratory (n=4, 44%) or school (n=4, 44%), with only one conducted at home (n=1, 11%). Six studies calculated the grams per Bite (67%) and three coded Bite Size observationally (33%; Table 6).

3.4.3.2. Bite Rate.

Of the six studies (13%) that calculated Bite Rate, five were done at school (83%) and one was conducted in the laboratory (17%). All studies calculated Bite Rate as the number of bites per minute. However, studies differed in which duration was used in the denominator (Table 6). Meal Duration was most commonly used in the denominator to calculate Bite Rate, followed by Eating Duration, and length of observation (Table 6).

3.4.3.3. Oral Exposure per Bite.

Oral Exposure per Bite reflects the average duration of each bite, which differs from the more commonly calculated Bite Rate. Only two studies (4%) coded Oral Exposure per Bite, both of which were laboratory-based studies from the GUSTO cohort (Table 6).

3.4.3.4. Chew Rate.

Eight studies (17%) calculated Chew Rate, five of which were school-based (63%) and three of which were laboratory-based (38%). The calculation of Chew Rate varied the most across studies (Table 6). The two studies from the GUSTO cohort calculated two estimates of Chew Rate: 1) number of Chews per gram of food consumed, which provides estimates of the amount of oral processing (i.e., chews per gram of food); and 2) number of Chews per Oral Exposure (sec), which reflects the speed of Chews (i.e., chews per time food is in mouth). In contrast, the remaining studies calculated the number of Chews per Bite (n = 4; 50%) and duration of each chew by dividing Meal Duration by number of Chews (n = 1, 13%).

3.4.3.5. Sip Rate.

Only one (2%) school-based study calculated Sip Rate, which was computed as the number of Sips per minute using observation duration as the denominator (Table 6).

3.4.3.6. Eating Rate.

Eight studies (17%) calculated eating rate. The majority of these studies were conducted in the laboratory (n=7, 88%) with only one study conducted at home (12%). Two of the laboratory-based studies used a UEM (25%) while the remaining studies all used video recordings (75%). The four studies from the GUSTO cohort calculated the number of grams consumed per Oral Exposure Time, while the remaining four studies calculated the energy consumed (kCal) per Eating Duration or Meal Duration.

3.4.3.7. Change in Eating Rate.

Four studies (9%) computed Change in Eating Rate. The three studies (75%) that used a UEM (laboratory: n = 2; school: n = 1) examined change in intake (grams) across the meal while the one home-based study (25%) examined change in Bite Rate (i.e., bites/minute). While two studies (50%) split the meal into segments to compare eating rate across the meal, the other two (50%) considered the trajectory of the cumulative intake curve from the UEM to calculate area under the curve or the change in slope (i.e., quadratic term).

3.4.3.8. Active Mealtime.

Two (4%) laboratory-based GUSTO cohort studies calculated the proportion of time spent eating by dividing Oral Exposure Time by Meal Duration (Table 6). Rather than reflecting total Eating Duration (sometimes termed Active Eating or Active Mealtime), this reflects the percent of time over the course of the meal that food is in the mouth.

3.5. Microstructure Behavior Associations with Child Characteristics

While studies were not required to examine associations with individual differences in child characteristics to be included in the current review, any reported associations were extracted in order to examine whether differences in microstructure behavior definitions contribute to discrepancies across studies. Overall, 41% (n = 19) of studies reported individual differences by sex (n = 6), age (n = 4), or weight status (n = 15). Too few studies examined associations with sex (n = 6) or age (n = 4) to determine whether differences in behavioral definitions contributed to discrepant findings across studies. Therefore, the full pattern of associations with sex and age are presented in Supplementary Table S4 and fully described in text in Supplementary Materials.

A total of 15 studies (33%) reported associations between meal microstructure behaviors and weight status (BMIz, obesity status) or adiposity (abdominal adiposity, percent body fat), collectively referred to as weight status throughout (Table 7). Higher weight status was consistently associated with shorter Eating Durations (Geller et al., 1981; Keane et al., 1981; Klesges et al., 1983; Llewellyn et al., 2008), larger Bite Sizes (Fogel et al., 2017a), faster Bite Rates (Geller et al., 1981; Keane et al., 1981; Llewellyn et al., 2008), and faster Eating Rates (Barkeling et al., 1991; Fogel et al., 2017b, 2017a; Laessle et al., 2001; Llewellyn et al., 2008). Additionally, when looking prospectively, faster Eating Rates and Bite Rates, and shorter Eating Durations at 4 years of age were associated with higher odds of having overweight (BMI>85^th percentile) at 6 years of age (Berkowitz et al., 2010). In contrast, weight status was not associated with number of Sips (Fogel et al., 2017a), Oral Exposure Time (Fogel et al., 2017a), Meal Duration (Barkeling et al., 1991; Klesges et al., 1983), or Active Mealtime (Fogel et al., 2017a). Together, this supports the premise of an ‘obesogenic’ style of eating that was robust to differences in study protocols.

Table 7.

Associations Between Microstructure Behaviors and Child Weight Status or Adiposity

	Behavior	Study	Association
	Bites	11. Drabman (1977) and 12. Drabman (1979)	More Bites in children with OW
Point Behaviors		33. Llewellyn (2008) and 40. Pesch (2018)	More Bites in children with OW/OB
		17. C1Fogel (2017a) and 24. Israel (1985)	NS association with weight status
		25. Johnson (2018) and 35. Lumeng (2006)	NS association with BMIz
		17. C1Fogel (2017a) and 25. Johnson (2018)	NS association with adiposity
	Chews	11. Drabman (1977)	Fewer Chews in children with OW
		17. C1Fogel (2017a) and 24. Israel (1985)	NS association with weight status
		17. C1Fogel (2017a)	NS association with adiposity
	Sips	17. C1Fogel (2017a)	NS association with weight status
		17. C1Fogel (2017a)	NS association with adiposity
Duration Behaviors	Oral Exposure Time	17. C1Fogel (2017a)	NS association with weight status
		17. C1Fogel (2017a)	NS association with adiposity
	Eating Duration	21. Geller (1981), 26. Keane (1981),	Shorter Eating Duration in children with OW/OB
		27. Klesges (1983) and 33. Llewellyn (2008)
	Meal Duration	4. Barkeling (1991)	NS association with weight status
		27. Klesges (1983	NS association with BMI %tile
	Bites Size	17. C1Fogel (2017a)	Larger Bite Size in children with OW/OB
		17. C1Fogel (2017a)	Larger Bite Sizes associated with greater adiposity
	Bite Rate	21. Geller (1981) and 26. Keane (1981), and 33. Llewellyn (2008)	Faster Bite Rate in children with OW/OB
Calculated Behaviors	Chew Rate	17. C1Fogel (2017a)	NS association with weight status
		17. C1Fogel (2017a)	NS association with adiposity
		11. Drabman (1977) and 12. Drabman (1979)	Slower Chew Rate in children with OW/OB
	Eating Rate	4. Barkeling (1991), 17. C1Fogel (2017a), 18. C1Fogel (2017b), and 33. Llewellyn (2008)	Faster Eating Rate in children with OW/OB
		31. Laessle (2001)	Faster Eating Rate in children with OW/OB—ONLY when mother was present
		17. C1Fogel (2017a) and 18. C1Fogel (2017b)	Faster Eating Rate associated with greater adiposity
	Deceleration	4. Barkeling (1991) and 31. Laessle (2001) 33. Llewellyn (2008)	Less Deceleration in Eating Rate in children with OW/OB NS association with weight status
	Active Meal Time	17. C1Fogel (2017a)	NS association with weight status
		17. C1Fogel (2017a)	NS association with adiposity

%tile: percentile; BMIz: body mass index z score; g: grams, HW: healthy weight; kCal: kilocalories; NS: non-significant; OW/OB: overweight or obese

^C1GUSTO cohort; studies share overlapping samples of participants

Although weight status was consistently associated with some meal microstructure behaviors, there were also some inconsistent findings (Table 7). While some studies indicated children with obesity or overweight had more Bites (Drabman et al., 1977, 1979; Llewellyn et al., 2008; Pesch et al., 2018) and fewer Chews (Drabman et al., 1977) than children with healthy weight, others did not (Fogel et al., 2017a; Israel et al., 1985; Johnson, 2018; Lumeng & Burke, 2006). There was no clear pattern of differences in the definitions between the studies that found an association with weight status and those that did not. In contrast, inconsistent findings for Chew Rate depended on how it was calculated such that higher weight status was associated with fewer Chews per Bite (Drabman et al., 1977, 1979), but not the number of Chews per gram (Fogel et al., 2017a). Similarly, studies that characterized change in Eating Rate based on intake (grams) showed more acceleration (Laessle et al., 2001) and less deceleration (Barkeling et al., 1991) in children with overweight or obesity, while a study that looked at differences in Bite Rate across 4 segments of the meal saw no association with weight status (Llewellyn et al., 2008).

4. Discussion

This systematic review provides a comprehensive assessment of behavioral definitions used to study meal microstructure in children and highlights the need for more consistent and comprehensive reporting of definitions in future studies. The majority of studies reviewed measured more than one microstructure behavior, reported good interrater reliability, and used consistent terms to label the measured behaviors. However, the quality of the definitions reported varied considerably across studies. Additionally, there was inconsistent reporting of details related to the types and amounts of foods served during the protocols, even amongst studies done under controlled laboratory conditions. Despite the inconsistencies across studies, increased weight status among children was consistently associated with a style of eating defined by larger Bites, faster Eating and Bite Rates, and shorter Meal Durations. Collectively, this study sheds light on key discrepancies in how meal microstructure is defined, coded, and interpreted across the literature.

Bites was the most commonly coded microstructure behavior, but the terms used to define this behavior varied the most across studies. For example, definitions that were reported in moderate detail (e.g., score of 3) varied from only requiring food to touch the mouth or lips (e.g., Fernandez et al., 2018) to requiring food to be chewed (e.g., Drabman et al., 1977) in order for a bite to be coded. However, none of the studies that reported a moderate level of detail specified whether the food must be swallowed or consumed to be considered a bite. In contrast, more detailed definitions (e.g., score of 1 or 2) differed in whether or not food must be swallowed or consumed to be considered a Bite (e.g., Fogel et al., 2018; Rendall et al., 2020) or whether it counted as a bite if the food was spit out (e.g., Pesch et al., 2018). While these more thorough definitions provided a better guide to reproduce the coding of a Bite, they also highlighted many decision points where studies could differ. For example, some studies counted nibbles (i.e., a series of rapid bites) as a single Bite (i.e., Llewellyn et al., 2008) while others counted them as multiple Bites (e.g., Czaja et al., 2011). Studies also differed on whether there was a minimum amount of food required to be ingested to count as a Bite, with some studies counting licks of food (e.g., Pesch et al., 2018; Tan et al., 2018) while others required Bites to be large enough to be eaten with a fork (e.g., Czaja et al., 2011). Although it is clear that differences in definitions can contribute to differences in the number of Bites recorded, more work is needed to determine the implications of these findings on the overall interpretation of the studies.

In addition to point behaviors, there were differences in how duration behaviors were defined. Among the studies that provided detailed definitions for Meal Duration, there were differences in how meal initiation and termination were designated. For some studies, Meal Duration was defined based on time from the first bite to the last bite or swallow (e.g., Keane et al., 1981). In contrast, other studies used the presence of food on the table or in the room as a marker for the beginning of meals and the removal of food to mark the end of meals (e.g., Saltzman et al., 2018). When Meal Duration is defined from first to last Bite, it reflects the amount of the total time the child is engaged with the meal. In contrast, defining Meal Duration based on the presence of food in the child’s environment (e.g., testing room, home) reflects the duration of exposure to food, but not necessarily time spent engaged with the meal. Importantly, the latter definition includes Latency to First Bite in the duration while the former removes this variability from the duration estimate. In addition to Meal Duration, some studies coded Eating Duration, which reflects time spent actively eating (e.g., Berkowitz et al., 2010; Klesges et al., 1983; Langlet et al., 2017) and does not include periods of distraction during the meal (e.g., conversation). This is in contrast to Oral Exposure Time, which only considers the time food is in the mouth (e.g., Fogel et al., 2018). A key difference between Eating Duration and Oral Exposure is that Eating Duration includes time that children are engaged in preparatory behaviors such as cutting food or putting food on a fork or spoon. Since duration behaviors are used to calculate many microstructure behaviors (e.g., Eating Rate), differences in definitions across studies can influence the interpretation and reproducibility of computed behaviors as well.

Computed microstructure behaviors not only varied due to differences in how point (e.g. Bites) and duration behaviors (e.g., Meal Duration) were defined, but also due to differences in the calculations used. Bite Size was sometimes quantified as the number of grams per bite (e.g., Fisher et al., 2003; Fogel et al., 2018) and other times it was estimated based on qualitative assessments (i.e., scale 1-small to 5-large; Israel et al., 1983; Marston et al., 1976) or the number of bites to eat a particular food item (Llewellyn et al., 2008). In addition to Bite Size, the calculated rates (i.e., Bite Rate, Chew Rate, Eating Rate) varied in their choice of duration behavior. While Meal Duration and Eating Duration were most commonly used, the choice of which to use has implications for interpretation. For example, two children with identical Eating Rates when calculated using Eating Duration could have different Eating Rates when calculated using Meal Duration if one child is more distracted during the meal than the other. Therefore, when interpreting rates calculated using Meal Duration, it is not possible to determine if individual differences are due to differences in active ingestive behaviors (e.g., Bites) or differences in the extent to which the child is distracted due to the presence of other influences (e.g., peers, media presence, etc.). In contrast, rates computed using Eating Duration can be more specifically attributed to differences during active eating periods. Greater clarity in the definitions of computed behaviors, including the microstructure behaviors they depend on, will improve the ability to interpret individual differences and generalize findings across studies.

Despite the variability in definitions used across studies, there was evidence for a consistent ‘obesogenic’ eating style that was associated with increased child weight status. Overall, shorter Eating Durations, larger Bite Sizes, and faster Bite and Eating Rates were associated with increased child weight status both concurrently and prospectively. Weight status was not, however, associated with number of Sips, Oral Exposure Time, Meal Duration, or Active Mealtime. In contrast, there were inconsistent findings for the association between higher weight status and number of Bites and Chews, Chew Rate, and how Eating Rate changes across the meal. While there were no distinct differences in the definitions of Bites and Chews between studies that showed significant associations with weight status and those that did not, there were patterns in the reporting of calculated behaviors. For example, weight status was associated with fewer Chews per Bite (Drabman et al., 1977, 1979), but not associated with Chew Rate when calculated as Chews per gram (Fogel et al., 2017a). This contradiction may be due to the fact that Chews per gram may more influenced by the foods served and therefore may be less sensitive to individual differences in eating style than Chews per Bite. In addition, weight status was associated with greater acceleration and less deceleration in Eating Rate over the meal when Eating Rate was characterized using intake (grams; Barkeling et al., 1991; Laessle et al., 2001). However, there was no association with weight status when Eating Rate was characterized as change in Bites per minute across 4 meal segments (Llewellyn et al., 2008). Together, these studies suggest that while children with higher weight status increase their rates of intake (grams per minute or second) as the meal progresses, their Bite Rate may not change over the course of the meal. Therefore, while there is evidence to support an ‘obesogenic’ eating style, future work is needed to understand impact of differing microstructure definitions on reported associations with individual differences in weight status.

While this study used a systematic search strategy to help reduce bias and ensure a comprehensive review of the literature, there are several limitations. First, this review does not take into account unpublished data or data that did not get published due to null results (i.e., the file drawer problem). While the primary purpose of the study was to review and catalog methods used in studies examining child meal microstructure, we did review findings related to individual differences in weight status. Therefore, we acknowledge that our review of individual differences was not exhaustive. Additionally, not all studies set out a priori to examine individual differences associated with meal microstructure. Therefore, it is possible that the studies we reviewed were not optimally designed or powered to detect individual differences. There was also inconsistent and incomplete information about the foods consumed in the studies, so it is not possible to determine the extent to which differences in food type influenced behavioral coding or were related to individual differences. Lastly, this study used a qualitative approach to rate the quality of the reported definitions, not the definitions or coding protocols that were actually implemented in the study. Given the good interrater reliability across studies, it is likely that the coding protocols and behavioral definitions were highly detailed in practice, even if that level of detail was not reported in the paper. Additionally, since this qualitative approach for rating the quality of definitions was informed by the available studies, future work is needed to validate and improve this rating approach.

Based on the results of the current systematic review, there are several recommendations for researchers in this field moving forward. Given that texture and food type can impact eating behaviors, it is important that studies examining child meal microstructure include detailed information on the types, amounts, brands, and preparation methods of foods served. While the ability to control the type and amount of food served differs by study location, efforts should be made to standardize how foods are reported to improve reproducibility. Similarly, in order to better interpret and generalize findings, full definitions should be made available for all microstructure behaviors. One challenge to this recommendation are manuscript formatting limitations (i.e., word and page limitations), in which case we recommend that full definitions are referenced in text and made publicly available through supplemental materials, appendices, or online resources (e.g., Open Science Framework, Zenodo, or lab website). In addition, we recommend that rate behaviors be calculated using both grams and energy consumed. This will help reconcile results between studies that use meals or foods that differ in energy density (Rolls, 2009, 2017). Additionally, while weight and volume of food consumed is thought to drive satiation, excess energy intake primarily drives weight gain (Rolls, 2017). Lastly, interpretations of calculated behaviors should include references to the behaviors involved. For example, greater food-oriented attention has been associated with greater meal intake (Lundquist et al., 2019), therefore, understating whether weight status is associated with Eating Rate only after meal engagement (i.e., Eating Duration) or throughout the entire meal period (i.e., Meal Duration) may help to further refine the description of an ‘obesogenic’ phenotype. Greater transparency of study foods and behavioral definitions will improve the ability to examine patterns of individual differences across studies and may help to identify modifiable, behavioral targets for prevention of excess weight in children.