Serving sizes and plate sizes

ANOVA
linear regression
Do people eat more food if their meals are served on larger plates? A randomized experiment with demographic and psychological control variables.
Author

Alex Reinhart

Published

September 1, 2019

Data files
Data year

2019

Motivation

Psychologists and food scientists have, in recent years, produced a great deal of research on a simple question: Do people eat more food if their meals are served on larger plates?

This question is of real practical importance. The ways food is packaged and presented undoubtedly affect how we eat it, and with obesity rates high across the Western world, simple ways to reduce overconsumption of food could have major public health benefits. Perhaps people overconsume food because they tend to eat most of what’s on their plates, leading them to eat more when the plate is large; if so, general advice to buy smaller plates could have a nontrivial effect.

Several experiments suggested this hypothesis is true, and that reducing plate size would help. But some of these experiments were conducted by Cornell University’s Food and Brand Lab, and starting in 2017 researchers started noticing serious problems with the lab’s publications, including miscalculated test statistics, inconsistencies in results, and incorrect descriptions of experimental procedures; since then, many of the lab’s papers have been retracted entirely, and the lab’s director, Brian Wansink, resigned from Cornell. It is not clear which conclusions can be trusted and which cannot.

This study sought to settle the question by conducting a large, carefully designed experiment, whose design and analysis was preregistered so it could not be manipulated once data was collected. In the study, participants were recruited and told they were participating in an experiment “examining the impact of time of the day on a range of mental processes,” and that they happened to be scheduled for a lunchtime session. They were told not to eat anything for three hours before the session; upon arriving, they completed various survey tasks, then were invited to serve themselves from a food trolley. The size of plate they were provided (small or large) was randomly assigned, and the exact quantity of food they consumed was measured (by measuring what was left in the trolley and on their plate). The session was also video recorded by hidden cameras, so researchers could count how many bites the participants took and the rate at which they ate.

Data

Each row of the dataset is one participant in the study. Beyond quantities of food consumed and the rate at which it was consumed, some basic demographic information was collected, as well as several standard psychological tasks.

Note: A value of -999 means that value was not recorded; a value of -888 on a survey question means the participant answered that they would “rather not say.”

Data preview

plate-size.csv

Variable descriptions

Variable Description
Date Date this participant participated
ID Participant ID number
plate.size 0 = small plate, 1 = large plate
kcal.consumed Total number of calories consumed
g.consumed Mass of food the participant ate (grams)
g.served Mass of food the participant put on their plate (grams)
g.remaining Mass of food left on participant’s plate (grams)
servings Number of times the participant served themself food
meal.duration Time from the first bite to the last bite (minutes)
n.bites Number of bites taken
bite.size Average bite size (total food mass / number of bites, grams)
eat.rate Time taken per bite (meal duration / number of bites)
bite.rate Number of bites per minute
highest.qual Highest educational qualification. 1 = none, 2 = up to 4 GCSEs, 3 = 5 or more GCSEs or 1 A-level, 4 = 2 or more A-levels, 5 = bachelor’s degree, 6 = post-graduate degree, 7 = rather not say.
highest.qual.split Highest educational qualification discretized
income Total household income per year, before tax (self-reported). 1 = up to £4,499; 2 = up to £6,499; 3 = up to £7,499; 4 = up to £9,499; 5 = up to £11,499; 6 = up to £13,499; 7 = up to £15,499; 8 = up to £17,499; 9 = up to £24,999; 10 = up to £29,999; 11 = up to £39,999; 12 = up to £49,999; 13 = up to £74,999; 14 = up to £99,999; 15 = over £100,000
income.med.split 0 = total household income below median; 1 = total household income above median.
imd Index of Multiple Deprivation of the area where the participant lives. Measures income, unemployment, education, crime, and other factors related to socioeconomic status and the living environment. Small values mean greater deprivation.
imd.med.split 0 = IMD below median value; 1 = IMD above median value.
ssrt Stop Signal Task response time (milliseconds). A task intended to measure participants’ response inhibition and impulsivity; slow response times indicate poor impulse control.
BIS-TOTAL Total score from the Barratt Impulsiveness Scale. Higher scores indicate greater impulsiveness.
BIS-Q1 BIS item 1: ‘I plan tasks carefully.’ 1 = almost always/always; 2 = often; 3 = occasionally; 4 = rarely/never.
BIS-Q2 BIS item 2: ‘I do things without thinking.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q3 BIS item 3: ‘I make-up my mind quickly.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q4 BIS item 4: ‘I am happy-go-lucky.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q5 BIS item 5: ‘I don’t pay attention.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q6 BIS item 6: ‘I have racing thoughts.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q7 BIS item 7: ‘I plan trips well ahead of time.’ 1 = almost always/always; 2 = often; 3 = occasionally; 4 = rarely/never.
BIS-Q8 BIS item 8: ‘I am self controlled.’ 1 = almost always/always; 2 = often; 3 = occasionally; 4 = rarely/never.
BIS-Q9 BIS item 9: ‘I concentrate easily.’ 1 = almost always/always; 2 = often; 3 = occasionally; 4 = rarely/never.
BIS-Q10 BIS item 10: ‘I save regularly.’ 1 = almost always/always; 2 = often; 3 = occasionally; 4 = rarely/never.
BIS-Q11 BIS item 11: ‘I squirm at plays or lectures.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q12 BIS item 12: ‘I am a careful thinker.’ 1 = almost always/always; 2 = often; 3 = occasionally; 4 = rarely/never.
BIS-Q13 BIS item 13: ‘I plan for job security.’ 1 = almost always/always; 2 = often; 3 = occasionally; 4 = rarely/never.
BIS-Q14 BIS item 14: ‘I say things without thinking.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q15 BIS item 15: ‘I like to think about complex problems.’ 1 = almost always/always; 2 = often; 3 = occasionally; 4 = rarely/never.
BIS-Q16 BIS item 16: ‘I change jobs.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q17 BIS item 17: ‘I act on impulse.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q18 BIS item 18: ‘I get easily bored when solving thought problems.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q19 BIS item 19: ‘I act on the spur of the moment.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q20 BIS item 20: ‘I am a steady thinker.’ 1 = almost always/always; 2 = often; 3 = occasionally; 4 = rarely/never.
BIS-Q21 BIS item 21: ‘I change residences.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q22 BIS item 22: ‘I buy things on impulse.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q23 BIS item 23: ‘I can only think about one thing at a time.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q24 BIS item 24: ‘I change hobbies.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q25 BIS item 25: ‘I spend or charge more than I earn.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q26 BIS item 26: ‘I often have extraneous thoughts when thinking.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q27 BIS item 27: ‘I am more interested in the present than the future.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q28 BIS item 28: ‘I am restless at the theater or lectures.’ 1 = rarely/never; 2 = occasionally; 3 = often; 4 = almost always/always.
BIS-Q29 BIS item 29: ‘I like puzzles.’ 1 = almost always/always; 2 = often; 3 = occasionally; 4 = rarely/never.
BIS-Q30 BIS item 30: ‘I am future oriented.’ 1 = almost always/always; 2 = often; 3 = occasionally; 4 = rarely/never.
line.task.correct Number of correct answers to a line orientation task, out of 24 possible. Measures ability to orient objects in space.
gender Gender. 1 = male, 2 = female, 3 = other, 4 = rather not say.
age Age (years)
ethnicity Ethnicity. 1 = white; 2 = mixed white and black, Asian, Caribbean, or other; 3 = Asian; 4 = Black; 5 = other; 6 = rather not say.
ethnicity.dicot Ethnicity dicotomized. 0 = white, 1 = non-white.
BMI Body mass index
hunger How hungry the participant reported being before lunch. 100 point scale. 0 = not at all, 100 = very much.
fullness How full the participant reported being before lunch. 100 point scale.
like.food How much the participant reported liking the food provided in the study. 100 point scale.
fasting Whether the participant fasted for 3 hours before the study, as requested (yes/no).

Questions

  1. The simplest question is also the one that directly answers the study’s main question: Did participants given a larger plate eat more, on average, than those given a smaller plate?
  2. Perhaps the effect of plate size varies depending on impulsivity (measured by the Barratt Impulsiveness Scale and the Stop Signal Task) or socioeconomic status (measured by income, education, and the Index of Multiple Deprivation). One could imagine, for example, that people who regularly struggle to get enough food might be more affected by the desire to eat everything on their plate. Conduct an analysis to see if the effect of plate size differs depending on these factors.
  3. The study was designed to have the sample size needed to get 80% statistical power to detect a difference of d = 0.5 standard deviations in the amount of food consumed between the two groups. The power analysis did not determine the sample size needed to detect if the effect of plate size differs depending on other factors. Use simulation to determine the power of the study to detect such differences. Is it capable of reliably detecting them?

References

Daina Kosīte, Laura M. König, Katie De-loyde, Ilse Lee, Emily Pechey, Natasha Clarke, Olivia Maynard, Richard W. Morris, Marcus R. Munafò, Theresa M. Marteau, Paul C. Fletcher & Gareth J. Hollands (2019). “Plate size and food consumption: a pre-registered experimental study in a general population sample,” International Journal of Behavioral Nutrition and Physical Activity 16 no. 75. https://doi.org/10.1186/s12966-019-0826-1

Data available from the Open Science Framework: https://osf.io/k74cu/