Thinking about the bigger picture: Influence of holistic processing on the dishware size effect.

Abstract

Individuals vary in the extent to which they engage in holistic and analytic information processing styles. Holistic processing involves focusing on the interconnectivity and relatedness of items being evaluated, while analytic processing involves focusing on items being judged as discrete elements and independent of context. We examined the contribution of these basic processing styles to the dishware size effect, which proposes that food consumption patterns may be influenced by the size of the dishware (i.e., larger plates increase the amount of food consumed). We observed that participants self-served and consumed more food when using and eating from a larger plate (LP) compared with a smaller plate (SP) (p≤0.01). Importantly, participants who reported greater levels of holistic information processing related to attitudes towards contradictions and attention allocation exhibited smaller variations in portions of food self-served and consumed based on the dishware size used (SP vs. LP). These findings suggest that the susceptibility of individuals to the dishware size effect may be associated with an individual’s dispositional tendency to process information in a holistic (vs. analytic) manner.

Introduction

Research suggests that environmental factors have significant influences on an individual’s eating behaviour (McCrickerd & Forde, 2016). These influences were previously observed to include the social (Herman, Roth, & Polivy, 2003; Higgs, 2015), physical (Crabtree & Blannin, 2015; Westerterp-Plantenga, 1999) and food environments (McCrory et al., 1999; Rolls, Morris, & Roe, 2002). How food is arranged, packaged, or contained, can also serve as a microscale environmental factor that guides food intake patterns through contextual influences from the immediate platescape or tablescape (Rowley & Spence, 2018; Sobal & Wansink, 2007; Spence, Piqueras-Fiszman, Michel, & Deroy, 2014). One such factor that has garnered a significant amount of research attention is the influence of dishware size on eating behaviour (Holden, Zlatevska, & Dubelaar, 2016; Hollands et al., 2015; Robinson et al., 2014), which is based on the observation that people are guided by the size of dishware used in judgments about the appropriate amount to serve and consume. Specifically, this research has suggested that more food is served and consumed when larger (versus smaller) dishware is used. Interestingly, despite being extensively studied, evidence of the influence of dishware size on consumption levels remains unclear and inconsistent (Robinson et al., 2014)¹, and a subject of continued debate (Holden et al., 2016). For instance, Rolls et al (2007) reported that caloric intake was not influenced by the size of the dishware when comparing participants’ use of small-, medium- and large-sized plates, while other researchers found that portion size expectations (Sharp, Sobal, & Wethington, 2019), as well as portion selection and intake (Marchiori, Corneille, & Klein, 2012; van Kleef, Shimizu, & Wansink, 2012) had a positive relationship with dishware size – increasing and decreasing significantly when larger and smaller-sized dishware were used, respectively.

Prior research proposes several mechanisms to explain the dishware size effect. One explanation proposes that dimensions and sizes of dishware activate and guide an individual’s implicitly held norms regarding typical and appropriate portion size (Herman & Polivy, 2005; Marchiori et al., 2012). That is, when compared with smaller dishware sizes, larger dishware sizes suggest an expectation (norm) for greater amounts of food to be consumed. Another widely proposed mechanism, which may serve to inform portion related norms, relates to the information processing biases associated with the Delbouef illusion (Delboeuf, 1865; McClain et al., 2014; Petit, Velasco, & Spence, 2018; Van Ittersum & Wansink, 2012). The Delbouef illusion is described as a visual phenomenon that results in equal-sized circles being perceived as larger and smaller when surrounded by outer circles that are slightly larger (Figure 1A) and much larger (Figure 1B) in diameter, respectively. Possible explanations for these information processing biases include: i) the tendency to assimilate inner and outer shapes that are closer in size and proximity (Figure 1A), resulting in the inner shape being perceived as larger, and ii) the tendency to contrast inner and outer shapes with larger differences in size and proximity, resulting in the inner shape being perceived as smaller (Figure 1B) (Goto et al., 2007). In addition to the proposed influence on the dishware size effect, the Delbouef illusion has also been used to explain geometric illusions that may occur in everyday life, such as an individual’s appearance. Assimilative illusions induced by eyebrows that are positioned closer to the eyes and eye shadow make-up are reported to make eyes appear larger compared to a face with eyebrows positioned higher and without eye make-up (Morikawa, Matsushita, Tomita, & Yamanami, 2015). Consequently, given that the influence of the Delbouef illusion may be determined by the extent of assimilation or contrast between the inner and outer shapes, an individual’s tendency to perceive and process information in relation to contextual elements may be an important determinant of the susceptibility to the dishware size effect.

Figure 1 –

Delboeuf illusion. A: Inner circle surrounded by a slightly larger outer circle. B: Identically sized inner circle (with A) surrounded by a much larger outer circle - providing the illusion that the inner circle of B is smaller than the inner circle of A. The difference (%) in diameters of the outer circles of A and B is the same as that of the small (20 cm; C) and large plates (26 cm; D) (i.e. 30% larger) used in the present study.

Individual differences in cognitive and perceptual styles like motivation, information processing, and attention (Gruszka, Matthews, & Szymura, 2010; Revelle, Wilt, & Condon, 2011) offer a potential explanation of the equivocal findings observed in the dishware size and eating behaviour literature. There appear to be support suggesting holistic versus analytic information processing tendencies having a significant relationship with how one evaluates environments, at least visually. Holistic information processing involves the consideration of the relationship between entities and being mindful of contextual information when navigating an environment. Conversely, analytic information processing is proposed to involve the consideration of entities to be discrete elements and independent from their context (Nisbett, Peng, Choi, & Norenzayan, 2001). Indeed, in populations reported to exhibit more holistic processing styles (e.g., East Asian cultures), judgements of a focal object are shown to be more regularly considered together with the wider field or context compared with more analytically oriented populations (e.g., Western cultures) (Cheon, Tang, Chiao, & Tang, 2018). For example, Masuda and colleagues (2008) reported that compared with North Americans, Japanese were more likely to holistically incorporate contextual information when evaluating a scene, in this instance, emotions of other people in the background when judging the emotion of a focal person placed centrally. Notably, this difference in contextual sensitivity between holistically versus analytically inclined populations is also observed in a variety of environments at macro and micro levels; from patterns of attention whilst evaluating a “cityscape” to the discerning of relationships between elements in a underwater scene (Masuda & Nisbett, 2001; Miyamoto, Nisbett, & Masuda, 2006). This is also in line with self-reported ratings of holistic (versus analytic) orientation among East-Asian individuals that suggest an increase in inclination to pay attention to the whole (vs. the parts) and to link causality with the surrounding environment (Choi, Koo, & Choi, 2007). This tendency to incorporate information from the field amongst holistically oriented individuals is also observed in other measures designed to assess conceptually similar relationships between focal items and contexts. For example, in a study using the Framed-lined task (Kitayama, Duffy, Kawamura, & Larsen, 2003), which involves the reproduction of lines based on variations in dimensional information from the field, participants from a holistically inclined population (East Asians) were reported to be more accurate in reproducing lines relative to a reference. It should be noted though, results from studies employing the Navon Task (Navon, 1977), a measure that assesses global versus local cognitive processing (a form of context sensitive vs. insensitive processing styles) have yielded less consistent relationships with participants’ cultural backgrounds (Miyamoto, 2013).

Applied to the proposed mechanisms of information processing biases linked with the Delbouef illusion discussed above, an individual’s tendencies to perceive holistically or analytically may potentially predict the influence of dishware size on food selection and consumption outcomes. Hence, it could be argued that individuals who are oriented to process information more holistically may be more aware of dinnerware size variations and consequently be less susceptible to the Delbouef illusion - resulting in minimal influences on consumption behaviour. Conversely, individuals who process information analytically, who are less sensitive to the context and relationships between elements, may be less aware of variations in dinnerware size and be more susceptible to the assimilation of elements that underlies the Delboeuf illusion. In this case, participants may perceive there is “less” food on a larger plate, resulting in increased serving size and consumption (see conceptual diagram in Figure 2). However, while the line of reasoning presented above may potentially explain the relationship between holistic/analytic processing styles and the dishware size effect, one may likewise argue that being an analytically oriented individual could confer a “protective effect” against the dishware size effect. That is, by thinking analytically and viewing objects as discrete from other elements and context, analytically oriented individuals may focus on attributes of the focal object (i.e., the portion size of the food) and be less susceptible to any variations in dishware size (i.e., the context or related object). To our knowledge, this relationship between the Delbouef illusion and context-sensitive thinking styles like holistic versus analytic processing has not yet been directly examined.

Figure 2 –

Conceptual diagram for the proposed relationship between the variations in dishware size, holistic versus analytic information processing tendencies, and the dishware size effect.

Consequently, an examination of the role of these individual differences in processing styles may shed some light on the cognitive processes that may influence the relationship between dishware size and eating behaviour. The main objectives of the present study were twofold. First, we sought to examine potential effects of context-sensitive information processing styles (holistic vs. analytic) on the relationship between dishware size (small vs. large) and eating behaviours, and ii) identify whether these information processing styles may be asymmetrically more influential on eating behaviour when encountering a specific plate size (small vs. large).

We hypothesized that self-reported information processing tendencies would influence the relationship between dishware size and eating behaviours. Specifically, participants who reported higher levels of holistic (vs. analytic) processing were predicted to be less affected by dishware size – selecting and consuming similar portions when using both small and large dishware. We also expected similar relationships with other measures that reflect a tendency to focus on focal items relative to relationships between focal items and contexts (i.e., global vs. local processing on the Navon Task, and holistic vs. analytic information processing on the Framed-Line Task). Furthermore, we conducted exploratory moderation analyses to determine whether holistic (vs. analytic) processing styles may be especially associated with eating behavior when participants used a specific plate size (e.g., large vs. small), although we did not have any specific hypotheses on which plate size would be more susceptible to influence from these thinking styles.

Methods

Participants

One hundred and seventeen participants (44 men; age 22 ± 2 years; BMI 21.8 ± 3.5 kg/m²) were recruited from a Singaporean university community for the present study. All participants were ethnically Asian (113 Chinese, 2 Indian, 1 Malay, one participant’s ethnicity was not captured). Participants were recruited through personal referrals and printed and electronic advertisements. Two participants could not consume the test meal due to self-reported allergies and were therefore excluded from analysis. The sample size required was based on calculations from previous research (Gregersen et al., 2008), which reported a between-subjects and repeated-measures study design requires 33 and 17 participants, respectively, to detect differences in caloric intake of ad-libitum (non-restricted) meals based on an alpha level of 0.05 and statistical power of 0.80. Previous studies examining the effect of dishware size on eating behaviour had sample sizes that ranged between 10 to 45 participants (repeated measures design) and between 57 to 85 participants (between measures design) (Robinson et al., 2014). Participants were compensated with $5 and a self-served ad-libitum lunch, consumed during the experiment, for completing each of the two study sessions. The study protocol was approved by the university’s institutional review board (IRB) and written consent obtained from all participants.

Experimental design

To test our hypotheses, participants were recruited to two experimental conditions, where they either served themselves food from a buffet style lunch onto a small plate (SP condition) or a large plate (LP condition). Recruited participants had the choice to sign up for both experimental sessions (SP and LP conditions) or one (either SP or LP) from six counterbalanced lunch time slots which were made available over two days for the experiment scheduled 4 weeks apart. Participants who signed up for a SP session on day 1 were only allowed to sign up for a LP condition session on day 2 and vice versa if they started off with a LP session on day 1. Participants were naïve to the experimental conditions of each session and were told the second session was used to confirm outcomes from their first session. Eighty-four participants (27 men; age 22 ± 2 years; BMI 22.1 ± 3.5 kg/m², Overweight/Obese BMI > 25.0 n=14, Underweight BMI <18.5 n=10) completed one experimental session (44 participants completed one SP session and 40 participants one LP session), and 31 participants (17 men; age 22 ± 2 years; BMI 21.2 ± 3.6 kg/m², Overweight/Obese BMI > 25.0 n=3, Underweight BMI <18.5 n=7) completed both the SP and LP experimental sessions.

Experimental Session

Participants were required to fast for two hours prior to the lunch experimental session. At lunch, participants arrived at the lab, registered and were briefed as a group about the session. Participants were provided a cover story that the study was examining social influences on health outcomes. The goal of studying the effect of plate size on food intake was not explicitly mentioned. Importantly, no participant reported or commented on the size of dishware used during the sessions. After written informed consent was obtained, participants rated current feelings of appetite using a visual analogue scale. Next, participants completed (in the following order) the Framed-Line Test (FLT), the Global-Local Reaction Time Measure (GLRTM) which is a modified Navon letter task, and two questionnaire-based assessments: the Analysis-Holism Scale (AHS) and the Self Construal Scale (SCS). Following these measures, participants were provided ad-libitum access to a lunch time test-meal. After participants finished their meal, they answered the same appetite rating questions from the start of the session. Participants were also asked at this point to hedonically rate the meal. Participants were debriefed at the end of their last session. Details of each measure can be found in the Constructs and Measures subsection below.

Constructs and Measures

Appetite Ratings

Participants’ feelings of hunger, fullness, satiety, desire to eat, and prospective food consumption were assessed using 100-point (mm) visual analogue scales (Blundell et al., 2010; Flint, Raben, Blundell, & Astrup, 2000).

Framed-Line Test

The Framed-Line Test (FLT) (Kitayama et al., 2003) is a widely used pen and paper task that evaluates the ability of individuals to visually incorporate or ignore contextual information (Kitayama, Park, Sevincer, Karasawa, & Uskul, 2009; Na et al., 2010; Zhang, Fung, Stanley, Isaacowitz, & Zhang, 2014). Previously reported reliability scores based on correlational measures of errors made within each task were moderate (r<.60) (Kitayama et al., 2009). The FLT initially presents participants with a square containing a vertical line extending from the middle of the inside top horizontal length. Then, participants were presented with another blank square of the same or different dimension and required to draw a vertical line that is either identical to the line initially presented (absolute task; AT) or in proportion to the height of the current square (relative task; RT). Five different frame combinations consisting of a reference frame with vertical line and a blank frame, excluding an example at the start of the test, were presented in total. A smaller error, represented by the difference in lengths between the line drawn by participants and the reference line, indicated each participant’s information processing bias corresponding to the task (AT vs. RT). A composite FLT score was computed based on the difference in error scores between the absolute and relative tasks.

Global-Local Reaction Time Measure

The Global-Local Reaction Time Measure (GLRTM) is a computer-based modified version of the widely used Navon Letter Task (Navon, 1977) that assesses global-local processing tendencies. The previously reported test-retest reliability of the global and local reaction time measures of the Navon Task, the main variable assessed for the current study, was reasonably high, r = .66 and r = .83 respectively) (Dale & Arnell, 2013). For this task, following the presentation of a fixation cross (“+”) in the centre of the screen for 500ms, participants were randomly (counterbalanced) presented one of eight composite letters that included four “global” letters – large target letters H or L composed of smaller letters F and T, and four “local” letters – large letters F and T composed of smaller target letters H or L. Participants were instructed to keep their index fingers on the letters H and L on the keyboard and to respond as quickly as they could to the stimuli displayed on the screen. This involved participants pressing the “H” key if the stimulus contained the letter H or the “L” key if it was the letter L. Overall, participants were exposed to 48 trials (24 global, 24 local trials), of which the first 8 were practice trials. Shorter average reaction time in response to the respective targets, global reaction time (GRTM) or local reaction time (LRTM), indicated each participant’s predominant information processing tendency. A composite GLRTM score is computed to be the difference in GRTM and LRTM scores. GLRTM scores were missing or not available for 1 participant from the within-subject analysis (n = 30) and 6 participants from the between-subjects analysis (n = 78).

Analysis-Holism Scale

Analytic versus holistic thinking styles of participants were assessed using the 24 item Analysis-Holism Scale (AHS; Cronbach α = 0.74) (Choi et al., 2007). The subscales that together make up the AHS include the Causality (example item “Everything in the universe is somehow related to each other”; Cronbach α = 0.71), Attitude Toward Contradictions (example item “It is more desirable to take the middle ground than go to extremes; Cronbach α = 0.69), Perception of Change (example item “Every phenomenon in the world moves in predictable directions”; Cronbach α = 0.58) and Locus of Attention (example item “It is more important to pay attention to the whole than just its parts; Cronbach α = 0.56) subscales. Participants rated their agreement for each item on a 7-point Likert scale, which were averaged to create a composite score of AHS. Higher AHS scores indicate increased tendency to i) pay attention to the whole versus the parts (Attention subscale); ii) link causality with surrounding environment and situations (Causality subscale); iii) perceive entities and events as dynamic and malleable (Perception of Change subscale); and iv) favour naïve dialecticism or tolerance of contradictions (Contradiction subscale). Data for the AHS was missing for 1 participant from the between-subjects analysis (n = 83).

Self-Construal Scale

The orientation of participants’ interdependent and independent self-construal, the extent to which individuals define themselves interdependent or independent of others, and think relationally (vs. independently) in social contexts was measured using the 24 item Self Construal Scale (SCS) (Singelis, 1994). The SCS is made up of two subscales; i) the interdependent (INT) subscale (12 item; Cronbach α 0.73) – example of an item “It is important to maintain harmony in my group” and ii) the independent (IND) subscale (12 item; Cronbach α 0.69) – example of an item “I act the same way no matter who I am with”. Participants rated their agreement for each item on a 7-point Likert scale, with higher scores indicating an increased degree to how they viewed themselves (i.e., interdependent or independent). A composite SCS score is computed to be the difference in INT and IND scores.

Servings and consumption

Participants were provided ad-libitum access to a lunchtime test-meal of Yang Chow fried rice - a popular, commonly consumed local dish that consists of rice wok-fried together with egg, cut up chicken, vegetables, and shrimp. The test meal was prepared and served ready to eat in buffet trays on the days of the experiment by a professional catering company. Participants were informed that they could serve themselves any amount of fried rice they wanted and were also explicitly told they could have more than one serving if they wished. Importantly, participants served themselves separately, largely out of the view of the other participants and consumed their meal alone, in private cubicles with high partitions. Depending on the experimental condition, participants served the food onto either a white side plate 20 cm in diameter (418±14g) (Dinera series, IKEA©, Delft, Netherlands) for the SP condition or a white dinner plate 26 cm in diameter (744±29g) (Dinera series, IKEA©, Delft, Netherlands) for the LP condition (Figure 2). The food, together with the plate, was weighed (using a DS-673 digital weighing scale, DIGI, Shanghai, China) immediately after participants served themselves and after completion of their meal. Weighing was done covertly as participants were distracted by a researcher who requested them to answer appetite rating questions while their food was weighed out of sight. Leftover food, if there was any, on each participant’s plate along with the weight of the empty plate were also recorded to derive separate measures of the: i) amount of food the participant served themselves at the buffet line, and ii) amount of food actually consumed. After participants finished their meal, they answered the same appetite rating questions from the start of the session. Participants were also asked to rate on a 7-point agreement Likert scale how appealing (4.73 ± 1.06) and tasty (4.73 ± 1.06) the YCFR served in the present study was.

Participants i) served themselves separately, largely out of the view of the other participants (buffet table was set up in front of the test room), ii) ate alone in cubicles that had high partitions that provided privacy and iii) were only presented with a specific size of plate (SP or LP) for each session. That is, each experimental session only used one specific size of plate. Following the session, the caloric density of the fried rice (~193 kcal/100g) was determined via a commercially available near-infrared calorie-analysis device (Calorie Answer, Joy World Pacific, Aomori, Japan) (Accuracy: Reported calorie content ~ 4% deviant from known stated values, Reproducibility: Reported average of 1.7% relative standard deviation) (Lau, Goh, Quek, Lim & Henry, 2016).

Statistical Analysis

As the study had i) participants who completed either only a SP or LP experimental session, and ii) participants who completed both SP and LP, several different analyses on the study’s main outcome variables of portions of food selected and consumed were conducted. Firstly, a repeated measures analysis of covariance (ANCOVA), accounting for baseline hunger, was conducted to compare the effect of SP vs. LP on the amount of food served and consumed (kcal) by participants that completed both experimental sessions (n=31). Secondly, a between-subjects ANCOVA, accounting for baseline hunger, was conducted for responses from participants who completed only one session (SP: n=44, LP: n=40). Finally, a between-subjects ANCOVA, accounting for baseline hunger, was conducted on combined data for participants who completed at least one experimental session. That is, data of participants who completed only one session and participants who completed both sessions were pooled together based on experimental condition (SP: n=75, LP: n=71). This final supporting analysis is presented in the Supplementary Materials.

Next, Pearson partial correlation analyses, accounting for baseline hunger, were conducted to determine whether the cognitive style measures, Analysis-Holism Scale (AHS) and its subscales, Global-Local reaction time measure (GLRTM), Framed-Line Test (FLT), and Self-Construal Scale (SCS), administered in the present study were associated with eating behaviour outcomes (differences in portions served and consumed between plate size conditions). We also used General Linear Modeling (GLM) to test whether individual differences in the cognitive style measures were more strongly associated with eating behaviour (portions served or consumed) in one plate size condition over the other among the participants that completed both conditions. Specifically, this was to examine whether the relationship between information processing styles and the dishware size effect was mostly driven by changes in response to a particular size of dishware (large or small). The above analyses were conducted, controlling for baseline hunger. As with the ANCOVAs conducted on food portions selected and consumed discussed earlier, both correlation and moderation analyses were also conducted on pooled data and are presented in the Supplementary Materials.

Finally, prior research has suggested inconsistent findings on the role of body weight as a potential predictor of susceptibility to the dishware size effect, with findings reporting that plate size affected expected food intake of normal weight (but not overweight) participants (Peng, 2017), while other research has suggested no relationship between weight status and response to plate size (Shah, Schroeder, Winn, & Adams-Huet, 2011). We conducted exploratory analyses that controlled for BMI as a covariate to examine the effect of LP versus SP conditions on food portions self-served and consumed.

While statistical significance was set at the typical value of α ≤ 0.05, Bonferonni adjustments (α/4) to account for multiple hypotheses testing on the four main context-sensitive thinking styles measured (AHS, GLRTM, FLT, and SCS) resulted in an α of ≤ .013. Statistical analyses were conducted using Statistical Package for Social Sciences (SPSS version 23, IBM Corp., Armonk, N.Y., USA).

Results

Plate-size effect on eating behaviour outcomes

A within subject repeated measures ANCOVA (n=31), revealed that participants served themselves more, F(1,29) = 8.51, p = .007, ${\eta }_p^2=0.23$, and consumed more, F(1,29) = 9.57, p = .004, ${\eta }_p^2=0.25$, of the ad-libitum test meal when using LP compared with SP (Figure 3A). Yet this relationship between plate size and portion served, F(1,29) = 0.61, p = .44, ${\eta }_p^2=0.02$, and consumed, F(1,29) = 0.82, p = .37, ${\eta }_p^2=0.03$, was no longer significant when participant BMI was entered into the model as a covariate. A between-subjects ANCOVA (n=84) showed no significant differences in the calories participants served themselves, F(1,81) = 2.37, p = .13, ${\eta }_p^2=0.03$, and consumed, F(1,81) = 3.49, p = .065, ${\eta }_p^2=0.04$, between conditions (SP, n=44 vs. LP, n=40). Although non-significant, the pattern of behavior was similar to the within-subjects condition (Figure 3B) ². The outcomes presented above controlled for baseline hunger levels.

Figure 3 –

Average amount of test meal (yang chow fried rice) that participants served themselves and consumed in the small plate vs. large plate condition, controlling for hunger for within subject (repeated measures) analysis (3A) and between subject analysis (3B).

^#Significantly different from small plate condition (p ≤ 0.01). Standard errors are represented by the error bars attached to the columns.

Cognitive style measures

On the between subject analysis, participants’ mean scores of the thinking style measures were not significantly different (p > .05) between the SP and LP experimental conditions, except for the relational task (RT) of the Framed-Line Task, F(1,81) = 4.30, p = .04, ${\eta }_p^2=0.05$(Table 1).

Table 1 –

Means and standard deviations (SD) of the thinking style measures and respective subscales from between-subjects analysis.

	All Subjects Mean ± SD	SP Condition Mean ± SD	LP Condition Mean ± SD
AHS	4.86 ± 0.41	4.85 ± 0.32	4.87 ± 0.50
Causality	5.14 ± 0.60	5.15 ± 0.61	5.13 ± 0.60
Attitude Toward Contradiction	4.88 ± 0.69	4.89 ± 0.61	4.87 ± 0.77
Perception of Change	4.46 ± 1.01	4.33 ± 0.85	4.40 ± 1.16
Locus of Attention	5.04 ± 0.64	5.01 ± 0.57	5.08 ± 0.71
GLRTM (GRTM minus LRTM) (msecs)	−18 ± 105	−34 ± 112	2 ± 93
GRTM (msecs)	707 ± 129	706 ± 118	707 ± 143
LRTM (msecs)	725 ± 132	741 ± 123	706 ± 142
FLT (RT minus AT) (mm)	−0.62 ± 8.88	−0.23 ± 7.87	−1.22 ± 10.33
RT (mm)	9.59 ± 7.41	8.01 ± 5.55	11.35 ± 8.77
AT (mm)	8.98 ± 7.12	7.77 ± 7.21	10.32 ± 6.85
SCS (INT minus IND)	0.32 ± 0.64	0.35 ± 0.70	0.30 ± 0.57
INT	5.05 ± 0.83	5.17 ± 0.54	4.92 ± 1.05
IND	4.73 ± 0.87	4.82 ± 0.71	4.62 ± 1.01

SP Small plate condition LP Large plate condition AHS Analysis-Holism Scale GLRTM Global-Local Reaction Time Measure GRTM Global Reaction Time Measure LRTM Local Reaction Time Measure FLT Framed-Line Task RT Relative Task AT Absolute Task SCS Self-Construal Scale INT Interdependent subscale IND Independent subscale. Means were not significantly different (p = >.05) between the two experimental conditions (SP vs. LP), except for RT (p = .04).

Correlational Analyses

The AHS subscales of, Attitude Toward Contradictions and Locus of Attention, were negatively correlated with the difference between conditions (LP minus SP condition) in the number of calories participants self-served and consumed during the test meal controlling for hunger (Table 2)². Notably, Attitude Toward Contradictions was also negatively correlated with the number of calories participants self-served, r(28) = −.39, p = .03, and actually consumed r(28) = −.40, p = .03, in the LP condition (adjusting for baseline hunger). Yet, the relationship between the AHS (overall composite scores and all subscales) and the number of calories participants self-served and consumed was not observed in the SP condition, p > .05. There were no other correlations observed between composite AHS scores or any other subscales and eating behaviour in either the SP or LP conditions, p > .05. Furthermore, no significant correlations were observed between GLRTM, FLT, SCS and the eating behaviour outcome measures assessed in the present study (see Table 2 and supplemental correlational analyses²).

Table 2 –

Partial correlation coefficients from within-subject analyses of the Analysis-Holism Scale (AHS) with eating behaviour outcomes – difference of the amount of ad-libitum lunch self-served and consumed (kcal) by participants between conditions (large plate minus small plate condition), controlling for hunger.

	Self-Served (kcal) (n=31)	Consumed (kcal) (n=31)
AHS	−.32	−.35
Causality	.15	.12
Attitude Toward Contradiction	−.39*	−.40*
Perception of Change	.06	.06
Locus of Attention	−.38*	−.40*
GLRTM (GRTM minus LRTM)	−.12	−.12
GRTM	.11	.11
LRTM	.19	.19
FLT (RT minus AT)	−.21	−.20
RT	.11	.11
AT	−.14	−.12
SCS (INT minus IND)	−.29	−.29
INT	−.29	−.29
IND	.04	.04

AHS Analysis-Holism Scale GLRTM Global-Local Reaction Time Measure GRTM Global Reaction Time Measure LRTM Local Reaction Time Measure FLT Framed-Line Task RT Relative Task AT Absolute Task SCS Self-Construal Scale INT Interdependent subscale IND Independent subscale.

^*p ≤ 0.05.

Participant BMI was not correlated with the difference in calories self-served, r(28) = −.06, p = .76, or consumed, r(28) = −.08, p = .69, between conditions (LP minus SP). BMI was also not associated with calories self-served or consumed within the LP or SP conditions, p > .05.

Moderation Analyses

For amount of calories participants self-served, GLM revealed interactions between plate size conditions (SP or LP) and the AHS subscales of Attitudes towards Contradictions, F(1, 28) = 4.98, p = .03, ${\eta }_p^2=0.15$, and Locus of Attention, F(1, 28) = 4.60, p = .04,${\eta }_p^2=0.14$. The interaction between condition and overall AHS score was not significant, F(1, 28) = 3.09, p = .09, ${\eta }_p^2=0.10$. Likewise, for amount of calories participants actually consumed, GLM revealed interactions between plate size condition and the subscales of Attitudes towards Contradictions, F(1, 28) = 5.35, p = .03, ${\eta }_p^2=0.16$, and Locus of Attention, F(1, 28) = 5.30, p = .03,${\eta }_p^2=0.16$. The interaction between condition and AHS overall score was not significant, F(1, 28) = 3.85, p = .06,${\eta }_p^2=0.12$. Yet, for all these interactions, there were no significant differences in the strength of the relationship between these components of the AHS and eating behavior (portions self-served or consumed) across the two plate size conditions, p > .05.

Discussion

The present study demonstrated partial support for a dishware size effect, such that the same participants self-served and consumed more calories when handling larger plates compared to smaller plates (within-subjects/repeated-measures analysis), but not when distinct groups of participants handling plates of differing sizes were compared (between-subjects analysis). This finding is consistent with conclusions of a meta-analysis (Holden et al., 2016) that found that dishware size only had a significant effect on food intake in certain situations; largely, when the food was self-served, and participants were unaware that they were in an eating behaviour experiment, as per the present research. The susceptibility of the dishware size effect to these types of situational and individual variations may have also contributed to the effect being observed only when participants were compared with themselves (within-subjects comparisons) across the SP and LP conditions.

However, it is also notable that our findings are in contrast some other studies that relied on within-subject study designs to test the dishware size effect. Rolls and colleagues (2007) showed that in a study that examined the ad-libitum food intake of a buffet-style meal when using different sized plates (small vs. medium vs. large), no significant effect of plate-size on consumption was observed. Compared to only 21% of participants in the present study that had a second serving (no one had more than two servings) in the SP condition, participants in Rolls and colleagues’ study were reported to make significantly more trips to the buffet line in the small plate condition compared with medium and large plate conditions. The tendency for participants in the SP condition to refrain from self-serving additional portions in our study could be related to self-presentation or normative concerns related to the presence of other participants in the room. Although we sought to minimize the direct attention of others when participants self-served and consumed the food by having the buffet tray set up separately from where participants were seated at cubicles and having participants sit alone at individual cubicles to consume the food, it may be possible that participants still experienced some normative concerns against retuning to serve themselves multiple portions.

Considering the role of individual differences in eating behaviour to significantly influence food selection and consumption (French, Epstein, Jeffery, Blundell, & Wardle, 2012), the present research set out to primarily examine whether differences in an individual’s information processing style may play an important role in the relationship between dishware size and eating behaviour. In support of our hypotheses, the present study’s outcomes suggest that one’s orientation towards analytic or holistic cognition has a notable influence on how variations in dishware size can affect subsequent caloric selection and consumption. Specifically, we observed that AHS subscales of Attitudes towards Contradictions and Locus of Attention were negatively correlated with differences between LP and SP conditions on self-served and consumed food intake. That is, participants who exhibited greater holistic (relative to analytic) cognition tendencies displayed lower variation in portion size selection and food intake patterns when using different sized dishware (SP vs. LP). Importantly, this effect appeared to be attributed to the tendency for participants exhibiting greater holistic thinking to select and consume smaller portions when using larger dishware, suggesting an adjustment of food intake (by appropriately decreasing portion size) associated with holistic thinking. Further supporting this notion, this relationship between AHS scores and portion size was stronger in the LP condition compared to SP condition (significant interaction of condition and AHS) when moderation analyses was conducted pooling together the participants who completed just one and both conditions of the study (see Supplementary Materials). Conversely, these findings also suggest that greater endorsement of analytic cognition tendencies may be due to greater variation in eating behaviour outcomes when using different sized dishware (SP vs. LP). Providing further insight into the specific characteristics that may be responsible for the main effects of the AHS presented above, we also found ratings for the subscales Attitudes towards contradictions and Locus of attention to be significantly correlated with eating behaviour outcomes.

The present study provides a novel demonstration of basic information processing and cognitive styles influencing the magnitude of food portions selected and consumed amid variations in dishware size. The finding that individuals with increased holistic (decreased analytic) cognitive tendencies may be less prone to food portion selection and consumption discrepancies based on dishware size, is consistent with our proposed notion that the inclination to avoid going to extremes, recognise the relatedness of objects and view objects in context (key aspects of holistic orientation) (Choi et al., 2007; Nisbett et al., 2001) may make individuals less susceptible to the Delboeuf illusion - the mechanism proposed to be driving the dishware size effect (McClain et al., 2014; Van Ittersum & Wansink, 2012). Essentially, this increased awareness of variations in dishware size compared to food placed on those dishes is proposed to allow holistically oriented individuals to make judgements and necessary cognitive adjustments regarding appropriate food portion sizes. Supporting this line of reasoning, individuals who displayed eating behaviour reflecting lower influence by the dishware size effect had higher agreement with subscales of Attitudes towards contradictions (i.e., preference for dialecticism and avoidance of extremes) and Locus of attention (i.e., emphasis on paying attention to the whole rather than its parts) (Choi et al., 2007). This may suggest that these individuals preferred not to “fill the dishware up” even when it increased in size (avoiding extremes) and appreciated that they were using larger-sized dishware and adjusted food intake accordingly (awareness of the contextual features).

It is important to note though that the discussion presented above assumes that the dishware size effect is primarily explained by the Delbouef illusion. Interestingly, our findings may also be applied to other proposed explanations of the dishware-size effect. For instance, holistically-oriented individuals may be more acutely aware of variations in dishware size and consequently be less susceptible to reliance on implicitly held norms regarding expected portion sizes, such as greater portion sizes being normative for larger dishware (Herman & Polivy, 2005). Supporting this notion, manipulations of dishware size (large vs. small plates) were reported to have no observable influence on eating behaviour in East Asians (a population that places relatively greater emphasis on holism) compared with participants from Western cultures (Peng et al., 2017). However, it should be noted that this study only involved images of foods presented on different plate sizes and self-reported eating behaviour ratings, as opposed to the actual serving and consumption of food as per the present study.

While relationships between subscales of the AHS and variations in eating behavior resulting from dishware size were observed, no significant relationships between the other measures associated with context sensitive or relational processing styles and eating behaviour outcomes were observed in the present study. While used in the literature to examine constructs related to elements of holistic/analytic thinking (i.e. global/local processing, interdependent/independent self-concepts), scores from the GLRTM, FLT and SCS were not found to be associated with AHS scores or food selection and consumption patterns of the present study. The GLRTM was not associated with behaviors across conditions of the study despite prior findings suggesting that global versus local processing styles may also affect food-relevant perceptions (Lewis, Seeley, & Miles, 2009). Potential explanations for the lack of outcomes associated with these measures may be due to the AHS and SCS measuring seemingly similar, yet ultimately distinct constructs. Previous research has also showed non-significant correlations between these measures (Choi et al., 2007), and proposed that context-sensitive thinking in social versus non-social may reflect distinct processes (Wong, Wyer, Wyer, & Adaval, 2021). Additionally, slight variations in how the GLRTM and FLT were administered in the current study compared to prior studies may have further contributed to a lack of robust relationships between these measures and outcomes of interest. While these measures were administered together with other assessment tools and on a group of participants in the present study, they were the sole assessment tool administered and with a 1:1 experimenter/participant ratio in previous research (Kitayama et al., 2003; Navon, 1977).

A supplementary finding was that the effect of plate size on the amount of food self-served and consumed was no longer observed when participant BMI was controlled in analyses. Prior studies comparing normal weight and overweight participants on the dishware size effect have suggested inconsistent roles of body weight in this phenomenon (Peng, 2017, Shah, et al., 2011). Our findings suggest that body weight may contribute to participants’ susceptibility to the dishware size effect but given the lack of any associations between BMI and portion selection/consumption behavior across the two plate size conditions, it remains unclear how body weight may be regulating this process. Future studies systematically designed to establish the influence of BMI would be a promising direction for future research into the dishware size effect.

Potential limitations apply to the current research. Firstly, the study’s eating behaviour outcomes were based on and restricted to a single food item: Yang Chow Fried Rice (YCFR). Nonetheless, palatability has been shown to be an important predictor of food selection and intake (Sorensen, Moller, Flint, Martens, & Raben, 2003), and participants’ ratings of YCFR in the present study indicate it was an appropriate test food item based on overall neutral ratings from participants on how appealing and delicious it was (see Results section for ratings). Secondly, while participants were required to fast for two hours prior to the lunch experimental session, breakfast intake was not standardized and participants were instead told to follow their normal breakfast routine. This limitation was accounted for when analysing the present study’s eating behaviour outcomes by controlling for hunger levels at the start of the experimental session. Thirdly, as dishware weight has been shown to influence consumption behaviour, with contents of heavier dishware expected to be more filling (Piqueras-Fiszman & Spence, 2012), the difference in weight of the dishware in the experimental conditions (SP vs. LP) may play a modulating role in the present study’s outcomes. However, this influence appears to be minimal considering one would expect this to result in participants serving themselves less in LP compared to SP, which would be contrary to the dishware size effect. That said, future research on the dishware size effect would benefit from systematically accounting for potential confounding influences from differences in weight between large and small dishware (e.g., presenting different sized dishware with identical weights). Next, given the associative nature of the present investigation, we were not able to examine the causal relationship between holistic cognition and the dishware size effect. Finally, limitations of a non-randomized recruitment process, relatively modest sample-size and statistical power, and multiple-hypothesis testing across several measures of thinking styles should be considered when interpreting the findings of the study. Future studies with strictly randomised and adequately powered samples where processing styles are experimentally manipulated prior to exposure to variances in dishware size would be required to causally confirm the “protective” role of holistic information processing on the dishware size effect. Of interest, it has been recently reported that holistic thinkers may be susceptible to maladaptive eating behaviour when indulgent food is presented alongside an occasion-setting background (i.e. at the movie theatre, airport) (Hildebrand, Harding, & Hadi, 2019). A complementary topic for research would be examining the relationship between holistic information processing and contextual influences in the broader environment that eating behavior is situated in.

In conclusion, the present study’s main findings suggest that the tendency of an individual’s eating behaviour to be influenced by the size of the dishware may be associated with individual differences (and potentially cultural differences) in holistic relative to analytic information processing orientations. These findings shed some light on cognitive characteristics and traits that may contribute to the relationship between dishware size and consumption behaviour. Consequently, our results may contribute explanations to the mixed findings in the literature on the dishware size effect and offer initial insights into how basic information processing orientations may interact with the built environment, in this case how food is laid out and presented (Rowley & Spence, 2018; Spence et al., 2014) to influence how consumers navigate the food environment. An important implication of the present study is that future research involving eating and food related built environments may have to consider the influence of individual differences in basic information processing styles, such as holistic versus analytic information processing tendencies.

Highlights.

• One’s sensitivity to contextual information may be applied to eating behaviour

• Dishware size proposed to have an influence on food consumption

• Larger dishes proposed to increase food intake (dishware size effect)

• Holistic processing was associated with less susceptibility to dishware size effect