Measuring ‘expected satiety’ in a range of common foods using a method of constant stimuli

Abstract

Humans have expectations about the satiety that is likely to develop after consuming particular foods. These expectations are potentially important, because they may influence decisions about meal size. Despite this, very little is known about the basis on which satiety expectations are formulated. This work introduces a methodology (based on a method of constant stimuli) that quantifies differences in expectations across foods. In Experiment 1 (N = 52) and Experiment 2 (N = 76) we compared expectations across 4 and 18 common foods, respectively. We discovered that a considerable mismatch occurs between satiety expectations and the energy content of foods (e.g., 200 kcal of pasta and 894 kcal of cashew nuts are expected to deliver equal satiety). This difference may reflect physical or macronutrient characteristics of these foods – energy-dense and high-fat foods have significantly lower ‘ratios of expected satiety.’ We also found a highly significant relationship between food familiarity and expected satiety (r = 0.86, p < 0.001), suggesting that expected-satiety judgements are learned. Across experiments, we were able to confirm both the reliability and robustness of our empirical approach. Future use of this methodology is discussed, both in relation to our understanding of portion-size decisions and its application more generally.

Introduction

Researchers with interests in short-term controls of meal size have tended to use ‘amount eaten’ as their dependent measure of interest. For example, participants may be given a large portion of food (larger than they can consume) and then told to eat until they no longer wish to continue. In these paradigms, intake is found to be influenced by a range of extrinsic factors, e.g., food variety (Norton, Anderson, & Hetherington, 2006; Raynor & Epstein, 2001), palatability (Yeomans, Lee, Gray, & French, 2001), distraction (Mitchell & Brunstrom, 2005), and serving size (Rolls, Roe, & Meengs, 2006), and by endocrine and neural controls that signal when to terminate a meal (Kissileff & Vanitallie, 1982; Woods, Benoit, Clegg, & Seeley, 2004). However, across these paradigms food tends to be presented very differently. For example, in some studies participants consume food directly from a large bowl containing more food than they can eat. In others, a large portion of food is offered and participants ‘self serve’ by transferring food onto a separate plate. This ‘self serving’ behaviour probably has greater ecological validity because very few foods are consumed directly from a large container of food (e.g., fondue). When self serving, meal size is likely to be governed by the amount of food that is initially placed on a plate (Wansink, Painter, & North, 2005), and by any subsequent additional servings associated with choosing ‘seconds.’ Thus, when eating behaviour follows this pattern, meal size is largely determined by one or more brief periods of cognitive activity, during which decisions about portion size are made (Brunstrom, 2007).

Pilgrim and Kamen (1963) explored predictors of food consumption in a large-scale study involving military personnel. Across a range of foods, perceived “fillingness” was the best predictor of food choice, better than the macronutrient composition or even the palatability of the foods. On this basis, it would seem likely that by understanding expectations about satiety and satiation we may gain insight into decisions about meal size. In this paper we begin this process by introducing a procedure for quantifying and exploring ‘expected satiety.’

Previously, expectations have been assessed in only two studies. Both suggest that humans are well able to express expectations about the consequences of consuming food. In one study ratings of ‘filling’ were elicited after consuming a mouthful of food (Green & Blundell, 1996). In the other, participants were shown photographs of foods and then estimated and rated how long they expected the foods would appease hunger (de Graaf, Stafleu, Staal, & Wijne, 1992). Both of these approaches have their advantages. In particular, measures are taken quickly and they are easily implemented. However, it is not clear that people can reliably estimate the time course of the satiety that results when particular foods are consumed, and ratings of this kind are subject to known sources of bias (Poulton, 1979).

Our approach to the problem of measuring satiety expectations relies on a relatively bias-free methodology known as the ‘method of constant stimuli.’ This classical psychophysical technique has been implemented in a variety of contexts and is commonly used by researchers of human sensory perception. In the context of expected satiety it is ideal, because it is widely regarded as being highly sensitive and because it can quantify differences in expectations across foods. In our version, one food of fixed and known energy content is displayed on a computer screen. Next to this ‘standard’ picture a different food is displayed.

On each trial, the amount of this second ‘comparison’ food changes, and the participant is asked to indicate which of the two foods will prevent them from feeling hungry for the longest period. After a sufficient number of trials, it is possible to plot the probability that the standard will be selected over the range of comparison values. Probit analysis is used to fit a sigmoid function from which a ‘point of subjective equality’ (PSE) can be derived. The PSE represents the point at which the standard is likely to be selected 50% of the time. This value is important, because it indicates the amount of the comparison (i.e., energy) that is expected to be equally as filling as the standard. Fig. 1 shows some hypothetical data and associated analyses: for a 200-kcal standard, the PSE with a (different) comparison food is 175 kcal. Hence, calorie for calorie; the comparison is expected to be 1.143 times as filling as the standard (200/175). Importantly, by making a set of systematic comparisons between a common standard and range of comparison foods it is possible to derive precise expected-satiety scores that can be compared across a range of foods.

Another major advantage of this approach is that it is possible to measure a person's sensitivity or ability to discriminate between foods based on expectations. The psychophysical function in Fig. 1 can have a steep or a shallow slope. A steep slope indicates that a respondent consistently discriminated between small changes in the value of the comparison around the PSE. A shallow slope suggests the converse. Together, the PSE and the gradient can be used to derive a useful measure of discrimination that is referred to as a ‘Weber fraction.’ A Weber fraction is a measure of sensitivity (a Just Noticeable Difference) that has been scaled according to the intensity or magnitude of a background stimulus (in this case the PSE). Weber fractions can be calculated by taking the difference between the PSE at 75% and the PSE at 50% (as shown in Fig. 1) and dividing this difference by the PSE at 50% (WF = (PSE_50% − PSE_75%)/PSE_50%).

Experiment 1 tested an implementation of this psychophysical procedure using a small number of common foods. In Experiment 2 we explored expected satiety across a much larger range of foods. In so doing, we sought to identify key features of foods that predict expectations about satiety.