Appetite 59 (2012) 364–371

Contents lists available at SciVerse ScienceDirect

Appetite

journal homepage: www.elsevier .com/locate /appet

Research report

Validating a behavioral economic approach to assess food demand: Effectsof body mass index, dietary restraint, and impulsivity q

Summar Reslan a,⇑, Karen K. Saules a, Mark K. Greenwald b

a Eastern Michigan University, Psychology Department, Ypsilanti, MI, United Statesb Wayne State University School of Medicine, Department of Psychiatry and Behavioral Neurosciences, Substance Abuse Research Division, Detroit, MI, United States

a r t i c l e i n f o a b s t r a c t

Article history:Received 23 August 2011Received in revised form 19 May 2012Accepted 21 May 2012Available online 31 May 2012

Keywords:Behavioral economicsFood choiceFood preferencesDietary restraintImpulsivityDelay discountingBody mass indexBMI

0195-6663/$ - see front matter � 2012 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.appet.2012.05.025

q Acknowledgements: This research was supportedMichigan University (EMU), Department of PsychologJoe Young, Sr. Funds (State of Michigan).⇑ Corresponding author.

E-mail address: shabhab1@emich.edu (S. Reslan).

Behavioral economic theory is a useful framework for analyzing factors influencing choice, but the major-ity of human behavioral economic research has focused on drug choice. The behavioral economic choiceparadigm may also be valuable for understanding food-maintained behavior. Our primary objective wastwo-fold: (1) Validate a human laboratory model of food-appetitive behavior, and (2) Assess the contri-bution of individual level factors that may differentially impact food choice behavior. Two studies wereconducted. In Study 1, female subjects (N = 17) participated in two consecutive food choice experimentalsessions, whereas in Study 2, female subjects (N = 21) participated in one concurrent food choice exper-imental session. During consecutive choice sessions (Study 1), demand for the more palatable food (i.e.,high-sugar/high-fat) was more inelastic than the less palatable (i.e., low-sugar/low-fat) option. Duringconcurrent choice sessions, demand for the more palatable food (i.e., high-sugar/high-fat) was moreinelastic for restrained vs. unrestrained eaters, and for those who were overweight vs. normal weight.Demand for both palatable and less palatable choices was more elastic for high-impulsive vs. low-impulsive subjects. These findings suggest that the behavioral economic framework can be usedsuccessfully to develop a human laboratory model of food-appetitive behavior.

� 2012 Elsevier Ltd. All rights reserved.

Introduction

Behavioral economic theory provides a framework for concep-tualizing and analyzing factors that influence choice. Behavioraleconomic choice paradigms, rooted in classic behavior analytictheory (Hursh, 1984), provide participants access to reinforcersvarying in accessibility, amount, or extent of work necessary to ob-tain them. By comparing response patterns, the relative reinforcingvalue of commodities can be gleaned and the influence of alterna-tives assessed. Behavioral economic studies of drug choice empha-size two fundamental principles that influence drug seekingbehavior (Bickel, Madden, & Petry, 1998; Vuchinich & Tucker,1988). First, consumption of a reinforcer (i.e., drug) tends to de-crease when constraints on access (e.g., price) are increased, andsecond, consumption is affected by the availability of concurrentalternative reinforcers. The same model may be applicable to thestudy of food choice behavior.

When given a choice between preferred snack foods or fruitsand vegetables, research suggests that individuals initially choose

ll rights reserved.

by funding from the Easterny, EMU Graduate School, and

the snack foods. However, as response cost to obtain the preferredfood (i.e., snacks) increases, choices shift to the less preferred butless costly alternative (Epstein & Leddy, 2006; Epstein, Salvy, Carr,Dearing, & Bickel, 2010; Goldfield & Epstein, 2002). In behavioraleconomic terms, this shift in food choice reflects an increase inthe elasticity (price-sensitivity) of demand for the preferred foodat higher response costs. Thus, in the present study, it was hypoth-esized that the reinforcing value of a food option will depend on itsunit price (UP [Hursh, 1984]; cost per unit amount), concurrentalternatives (e.g., sweet vs. less palatable food), and constraintson availability of these options. Increases in UP decrease consump-tion in a positive decelerating fashion. This relationship betweenUP and food consumption has been validated in laboratory animals(e.g., Collier, Johnson, Hill, & Kaufman, 1986; Madden, Dake, Mauel,& Rowe, 2005; Mathis, Johnson, & Collier, 1996).

Despite the value of behavioral economics in defining the rela-tionship between food UP and behavioral response, few studieshave been conducted with humans. Several studies have foundthat increases in response cost reduced food reinforcing value,yet most studies did not use a classical demand curve approach;rather, analyses of variance (ANOVAs) were used to block on fixedratio (FR) or variable ratio (VR) requirements (e.g., Giesen, Haver-mans, Nederkoon, Strafaci, & Jansen, 2008; Raynor & Epstein,2003; Temple, Chappel, Shalik, Volcy, & Epstein, 2008). In onerecent study, Yang and Chiou (2010) used this model with more

S. Reslan et al. / Appetite 59 (2012) 364–371 365

sensitive regression analysis (to estimate elasticity coefficients) toevaluate influences on beverage choice behavior. They noted thatlowering the cost of healthful alternatives and heightening con-cerns about less healthy alternatives shifted choice toward morehealthful beverage options (e.g., sugar-free green tea vs. sweet tea).

In the present investigation, a classical demand curve approachwas utilized to analyze food choice responding. Typically, a de-mand function describes the relationship between the quantityof a commodity consumed and its UP. This approach involves plot-ting consumption of the reinforcer (e.g., food) against its UP, whichyields a positively decelerating demand function (Hursh, 1980).The behavioral economic approach offers conceptual and analyticadvantages over conventional procedures. First, the influences ofseveral cost and benefit parameters can be parsimoniously inte-grated within a single UP metric (Bickel, DeGrandpre, Hughes, &Higgins, 1990). Second, choice/demand is predicted to be indepen-dent of the specific cost (numerator) and benefit (denominator)values that comprise the UP (principle of functional equivalence),which enables cross-study comparisons. Third, demand curvesprovide analytic and graphical results that are not intuitively obvi-ous from other response data such as breakpoints; furthermore,breakpoints depend on the specific values of the progressive ratioschedule (i.e., number of trials and relative steepness of responserequirements) and unit doses/amounts, resulting in findings thatcannot readily be compared across studies. Demand curves canbe adapted to study group choice behavior (instead of consump-tion amounts) when the participant only has one opportunity tochoose a commodity at each UP (e.g., Greenwald & Hursh, 2006).Under these conditions, one analyzes the percent of the group(i.e., consumer market) that chooses the commodity at each UP(known as ‘‘participation rate’’). The present study aimed to vali-date this approach to study food choice, and to examine microeco-nomic factors that may advance our understanding of palatablefood demand, analogous to prior findings with respect to humansubjects’ psychoactive drug demand (e.g., Greenwald, 2010; MacK-illop & Murphy, 2007; Madden & Bickel, 1999).

Several factors may differentially impact food choice behaviorand the reinforcing value of food including body mass index(BMI; Clark, Dewey, & Temple, 2010; Epstein et al., 2010; Epstein,Temple et al., 2007; Giesen, Havermans, Douven, Tekelenburg, &Jansen, 2010; Saelens & Epstein, 1996; Smith & Epstein, 1991;Temple, Legierski, Giacomelli, Salvy, & Epstein, 2008; Templeet al., 2009), level of dietary restraint (e.g., Giesen, Havermans, &Jansen, 2010; Giesen et al., 2008; Goldfield & Legg, 2006; Goldfield& Lumb, 2008; Houben, Roefs, & Jansen, 2010), and delay discount-ing (e.g., Christakou, Brammer, & Rubia, 2011; Rollins, Dearing, &Epstein, 2010). More specifically, the reinforcing efficacy of high-fat, high-carbohydrate, low-protein snack food is generally foundto be higher among obese relative to non-obese individuals (e.g.,Epstein, Temple et al., 2007; Saelens & Epstein, 1996; Templeet al., 2009), and, compared to unrestrained eaters, individuals highin dietary restraint tend to prefer unhealthful food over healthfulalternatives (i.e., high calorie snack vs. fruit; Houben et al., 2010).Delay discounting is a construct often used as a proxy measureof impulsivity (Perry, Larson, German, Madden, & Carroll, 2005),and greater discounting occurs when small immediate rewardsare chosen over larger delayed rewards (Bickel et al., 1998). Asimpulsivity is associated with binge eating (Dawe & Loxton,2004), it is not surprising that this variable differentially impactsfood reinforcement (Epstein, Dearing, Temple, & Cavanaugh,2008). The prevalence of overweight and obesity have increaseddramatically in recent decades (e.g., Flegal, Carroll, Ogden, & Cur-tin, 2010; Hedley et al., 2004), with recent estimates classifying66.3% of American adults as overweight or obese (Ogden et al.,2006). Using behavioral economic analysis to examine theseindividual difference factors contributing to food preferences and

reinforcement, and by extension, weight, may thus yield data re-lated to factors motivating food-related decision making, appeti-tive behavior, and weight gain.

The purpose of this investigation is two-fold. First, we aimed tovalidate a human laboratory model of food-appetitive behaviorusing a classical demand curve approach. To our knowledge, thisis the first investigation to use a human laboratory choice task toassess demand curves for two different food options. The secondaryaim was to assess the association between individual differencefactors that may alter food choice behavior. Two studies were con-ducted to accomplish these aims. The goal of Study 1 was to dem-onstrate that consecutive session demand curves could begenerated to significantly differentiate the reinforcing efficacy oftwo food options varying in palatability. It was hypothesized thata high-sugar/high-fat option (Dove Milk Chocolate™) would main-tain a greater reinforcing efficacy and inelastic behavior comparedto a low-sugar/low-fat option (Teddy Grahams™). Study 2 wasconducted to demonstrate that concurrent session demand curvescould be generated to significantly differentiate the reinforcingvalue of two food options, equated on mass, calories, and fat, anddiffering on only one major nutrient dimension (i.e., high-sugar/high-fat vs. low-sugar/high-fat options). Study 2 also assessed indi-vidual difference variables (BMI, dietary restraint, impulsivity) toevaluate whether they impact food choice responding in this par-adigm. These relationships were assessed to provide additionalevidence of validity for this model. Validity for this procedurewould be supported if food choice responding relates to BMI, re-straint, and impulsivity in ways that have been supported in theextant literature. While it was once again hypothesized that thehigh-sugar/high-fat option (Hershey Milk Chocolate Kiss™) wouldmaintain greater reinforcing efficacy and inelastic behavior com-pared to the high-fat/low-sugar option (Kraft Cheddar CheeseCubes™), BMI, dietary restraint, and impulsivity were expectedto moderate this relationship. In particular, both higher BMI andgreater dietary restraint were expected to be associated withgreater food-inelastic demand during experimental sessions forthe hypothesized more palatable (high-fat/high-sugar) food op-tion. Additionally, more extreme discounting (higher impulsivity)was expected to be associated with greater food-elastic demand,as individuals high in impulsivity are likely to desire smaller, moreimmediate outcomes. Individuals who choose more immediatefood rewards, however, may discount future problems associatedwith obesity (Epstein, Leddy, Temple & Faith, 2007).

Methods

Overview

In the two experiments reported here, candidates werescreened using online survey methodology, and those who met eli-gibility criteria were invited to the laboratory to complete either aconsecutive (Study 1) or concurrent (Study 2) food choice study inwhich they were asked to complete a computerized task to earnaccess to food varying in amount and type. The details of the twochoice procedures are described below, under Study 1 and Study2, respectively, but we first provide the general framework. Boththe screening and experimental phases of the project were re-viewed and approved by the Eastern Michigan University HumanSubjects Review Committee (IRB).

Recruitment

Email invitations were sent to undergraduate psychologyinstructors requesting that they distribute a 20–30 min screeningsurvey to students. A link to this screening survey was also posted

366 S. Reslan et al. / Appetite 59 (2012) 364–371

on a website that all undergraduate students could access andcomplete for extra course credit. The screening survey assesseddemographic information, food preferences, food allergies, nicotineuse, and the presence or absence of carpal tunnel syndrome. Thisscreening phase was used to ensure that those recruited for thelaboratory studies would be: female (to minimize gender-relateddifferences in eating behavior); non-pregnant (to avoid preg-nancy-related effects on food preference); non-smokers (to avoidchanges in appetite as a function of nicotine intake); without car-pal tunnel syndrome (because the laboratory task involved consid-erable computer ‘‘mouse’’-pressing); and endorsed liking (andnon-allergic to) the food items that would be offered in the exper-imental sessions. Individuals expressing interest in participating inthe laboratory-based portion of this investigation provided theircontact information at the end of the screening survey.

Experimental sessions

Informed consent was obtained from all participants before allexperimental sessions. To control for pre-session hunger, partici-pants were asked to refrain from eating two hours prior to the startof the experimental choice session. As all participants reportedadhering to the enforced 2 h food abstinence, no sessions wererescheduled. All experimental choice sessions were scheduled be-tween the hours of 1–6 p.m. to control for fluctuations in food pref-erences that occur as a function of the time of day.

Participants were seated in front of a computer where theexperimental choice task would be completed. A 3-h version of thiscomputerized choice progressive ratio (PR) task has been used pre-viously in behavioral economic drug studies (e.g., Greenwald,2010; Greenwald & Hursh, 2006; Greenwald & Steinmiller,2009); PR schedule parameters were adapted to create a 30-minsession version of this task in the present study (see below). Partic-ipants were informed that they would be completing a task that in-volved working for different foods, and that we would like them toplan to consume the food they earned following the completion ofboth sessions. Participants were encouraged to only work for foodthey would be willing to consume at the end of each study. If par-ticipants earned more than they could consume, they were not al-lowed to take the remaining food home.

Data analysis

The computer program for the behavioral economic choice pro-cedure automatically recorded each food choice and responses par-ticipants made. Demand curves were generated to examine therelative reinforcing efficacy of each food. Elasticity of each demandcurve was gleaned from analyzing the non-linear regressionparameter that measured rate of change in slope for group-percentchoice (post-session consumption) relative to changes in UP. Thedemand curve is inelastic when proportional decreases in choice/consumption are small relative to proportional increases in UP(i.e., slope >�1). Stated differently, the demand curve is inelasticwhen respondents are willing to defend their choice (i.e., find thecommodity more reinforcing) as price increases. The demand curvebecomes elastic when proportional decreases in choice/consump-tion become large relative to proportional increases in UP (i.e.,slope <�1; Hursh & Winger, 1995). That is, the demand curve iselastic when respondents are less willing to defend their choicefor a commodity at higher prices. The UP at which the slope ofthe demand curve equals -1 in log–log space is referred to as Pmax.The group-percent choice data (excluding ‘‘0’’ values, which areundefined in log10 transform) across the range of UPs are fit usingthe exponential demand equation: Y = log(L) ⁄ exp(�AX). In thisequation, parameter Y is group-percent choice; parameter L isthe group-percent food choice at the lowest UP; parameter X

is unit price; and parameter A is the rate of change in slope ofgroup-percent choice across UPs (elasticity).

GraphPad Prism v.4 was used to plot and test for differences indemand curves. When the level of food choice at the lowest UP con-verges to 100% (which was routinely observed in the present stud-ies), normalized demand calculations are used (i.e., L parameter isassigned a fixed value of 100 in the demand equation). As a result,the statistical test performed on the demand curves (e.g., whencomparing two groups or food types) is simplified to comparingonly the difference in the ‘‘A’’ parameter. The software providesthe mean (±1 SEM and 95% CI) A slope values of the two curves basedon goodness-of-fit criteria and performs an F-test. Pmax is derived di-rectly from the slope parameter (Pmax = 0.29 � A [see Greenwald &Hursh, 2006). Pmax is not statistically tested, but Pmax is reported in-stead of A because these values (unlike A values, which are small,usually <0.0001) can be readily visualized along the abscissa of UPvalues, such that Pmax is the UP value on the demand curve at thecurve’s ‘‘tipping point’’ or transition from inelastic to elastic de-mand. In Study 2, additional inferential statistics, described below,were used to evaluate the impact of individual difference variables(BMI, dietary restraint, impulsivity) on food choice behavior.

Experiment 1

Participants

This sample was comprised of 17 female undergraduate collegestudents from a Midwestern university. Participants were 67% Cau-casian and 33% African American, with a mean BMI of 23.33(SD = 3.72). Participants ranged in age from 18 to 43, with themajority (71%) in the 18–21 year old category.

Procedures

Two 11-trial experimental sessions were conducted over consec-utive 30-min intervals (i.e., 60 min total). Session order was alter-nated in a consecutive fashion, with nine participants receivingthe high-sugar/high-fat food condition (Dove Milk Chocolate™)first, and eight receiving this condition second. During each ses-sion, participants could work for units of food or do nothing. Theamount of food earned per trial was equated on mass, althoughcalories, sugar, and fat content varied. On each trial, the choicewas between one piece of Dove Milk Chocolate™ (7.5 g, 42 cal,4 g sugar, 3 g fat) or six Teddy Grahams™ (7.86 g, 32.5 cal, 1.98 su-gar, 1.02 fat). Participants were informed that, although amount offood earned per trial would be held constant, the response require-ment (i.e., number of computer ‘‘mouse’’ button presses) to earneach food on each subsequent trial would increase. Across the 11trials, number of responses increased exponentially (i.e., 5, 12, 33,100, 180, 340, 540, 835, 1220, 1660 and 2275 presses required toearn the same unit food amount); however, participants werenot specifically told the extent of increase in response requirementacross trials. This procedure was replicated for the second 30-minsession with the other food choice. After completing both experi-mental sessions, the combined units of food earned were providedand participants were asked to consume the entire amount earned.Delaying consumption after the first session was used to minimizefood satiation, which could alter responding in the second session.All participants consumed the entire amount of food earned priorto their departure. Participants were compensated $10 for theirparticipation.

Results

Group-percent food choice exhibited a positive deceleratingrelationship with UP, thus demonstrating sensitivity of the model

S. Reslan et al. / Appetite 59 (2012) 364–371 367

(see Fig. 1). At UPs >540 (i.e., response requirement on the 7th of 11trials), demand for the high-sugar/high-fat food became elastic. AtUPs >340 (i.e., response requirement on the 6th of 11 trials), de-mand for the low-sugar/low-fat food became elastic. Demand forthe high-sugar/high-fat food (Dove Milk Chocolate™) was mod-estly but significantly more inelastic than for the low-sugar/low-fat food (Teddy Graham™), Pmax = 663 vs. 494, respectively, F(1,17) = 6.45, p < 0.05. However, further analysis indicated a signifi-cant session order effect, i.e., demand for the food that was avail-able in the first session was more inelastic than food available inthe second session, (Pmax = 712 vs. 271), F(1, 17) = 132.30,p < 0.001, even though all food earned was not consumed untilafter completing both sessions.

Discussion

Study 1 validated that demand curves for independent foodchoices comparable in mass, but which differed in sugar and fatcontent, could be generated in consecutive choice sessions. Consis-tent with our initial hypothesis, demand curves differentiated therelative reinforcing efficacy of two food choices, as the high-sugar/high-fat option maintained more inelastic behavior (i.e.,greater reinforcing efficacy) compared to the low-sugar/low-fatoption. Results thus demonstrated that the demand curve for thehypothesized more palatable option (e.g., high-sugar/high-fat)could be generated in the direction expected. These encouragingfindings enhance our confidence that the behavioral economicframework can successfully be used to develop a human laboratorymodel of food-appetitive behavior that differs based on person-le-vel factors.

Results also suggested that when conducting consecutive choicesessions, it is important to consider the influence of session order,as this has the potential to confound findings. Despite our efforts tocontrol pre-experimental food intake and to delay consumption ofall earned food until after both sessions were completed, it appearsthat experimental sessions need to be spaced farther apart, or thatchoices should be assessed concurrently to avoid this confound.Alternatively, smaller units of food per trial should be offered insituations where food choice sessions are scheduled consecutively.Other notable limitations of Study 1 included the failure to includea ‘‘learning’’ trial prior to conducting the actual experimentalchoice sessions, and selecting food choices that differed on twodimensions, sugar and fat. When asked for feedback followingstudy completion, a small number of participants indicated thatthe directions for completing the experimental choice procedurewere slightly unclear.

Fig. 1. Demand for high-sugar/high-fat (Dove Milk Chocolate™) vs. low-sugar/low-fat (Teddy Grahams™) in consecutive choice sessions. Pmax values are providedabove the demand curve for the high-sugar/high-fat food and below the demandcurve for the low-sugar/low-fat option.

In Study 2, the objective was thus to establish the reinforcingefficacy of chocolate (i.e., palatable food choice) relative to a differ-ent hypothesized less-palatable food choice (i.e., cheddar cheese)in concurrent-food choice sessions, and to explore the influenceof individual difference factors such as BMI, dietary restraint, andimpulsivity on food choice responding to further validate thatthe behavioral economic approach is a sensitive way to detect dif-ferences in food choice. These design features and comparisonswere intended to improve sensitivity of the behavioral economicparadigm to detect differences between groups of interest. Also,in Study 2, a brief learning trial was administered to clarify any dis-crepancies prior to the experimental choice task. Finally, because itwas difficult to draw conclusions from Study 1 as to which foodingredient influenced participant responding, sugar or fat, in Study2, the two food options selected differed only in sweetness. Addi-tional parametric studies are necessary to compare how respond-ing differs based on other macronutrient domains (e.g., high-fatvs. low/no-fat).

Experiment 2

Participants

This sample included 21 female undergraduate college studentsfrom a Midwestern university. Participants were 62% Caucasian,24% African American, 10% Middle Eastern, and 5% Hispanic, witha mean BMI of 26.72 (SD = 5.99). BMI data are missing for one sub-ject who failed to provide an estimate of her height, and delay dis-counting data are missing for a different subject who failed tocomplete this questionnaire prior to the laboratory food choicetask. Participants ranged in age from 18 to 44 years, with a rela-tively even distribution across age categories.

Procedures

As in Study 1, participants were seated in front of a computerwhere the experimental choice task would be completed. Prior tothe experimental choice procedure, participants completed alearning trial (without eating earned food), to ensure they fullyunderstood the task. Participants then completed a single 11-trialcomputerized experimental session conducted over a 30-mininterval. The session involved 11 concurrent choices betweentwo food options, a high-fat/high-sugar food (Hershey Milk Choco-late Kisses™) or a high-fat/low-sugar food (Kraft Cheddar CheeseCubes™). Participants were told that they could work to earn dif-ferent foods, and that they should plan to consume the food theyearned upon completion. As in Study 1, participants could workfor units of food or do nothing. That is, they were not obligatedto earn all choices; thus, demand for the two food choices neednot be substitutable.

The amount of food earned per trial was equated on mass, cal-ories, and fat, although sugar content varied. On each trial, thechoice was between one piece of Hershey Milk Chocolate Kisses™(4.55 g, 16.7 cal, 1.33 g fat, 2.55 g sugar) or one Kraft CheddarCheese Cube™ (4.54 g, 15.0 cal, 1.38 g fat, 0 g sugar). Participantswere informed that, while the amount of food earned per trialwould be held constant, the response requirement (i.e., numberof ‘‘mouse’’ button presses) to earn each food on each subsequenttrial would increase. Across the 11 trials, the number of mousepresses increased exponentially (i.e., 5, 12, 33, 100, 180, 340, 540,835, 1220, 1660 and 2275 presses required to earn the same unitfood amount); however, participants were not told the extent ofthese increases in response requirements across trials. Of note,the two options compared had independent schedules (i.e.,responding for one option did not affect the response requirement

Table 1Correlation coefficients between independent variables.

1 2 3

Overweight status —Dietary restraint 0.19 (0.418) —Impulsivity �0.10 (0.673) �0.27 (0.270) —

Note: p values in parenthesis.

368 S. Reslan et al. / Appetite 59 (2012) 364–371

for the alternative). After the session, the combined units of foodearned were provided, and participants were asked to consumethe entire amount of food earned. All participants consumed theentire amount of food earned prior to their departure. Participantswere then compensated $10 for their participation.

Measures

ScreeningIn the online phase of Study 2, participants completed the same

screening as for Study 1, to assess study eligibility criteria, but alsocompleted the following measures of individual differences, en-abling us to evaluate the contributions of dietary restraint andimpulsivity on food choice behavior. Height and weight were as-sessed to enable evaluation of the influence of BMI on food choiceresponding.

Three Factor Eating Questionnaire (TFEQ; Stunkard & Messick, 1985)The TFEQ is a 51-item self-report questionnaire designed to as-

sess cognitive and behavioral components of eating. Comprised ofthree subscales (i.e., dietary restraint, disinhibition, and hunger),the TFEQ contains 36 dichotomously scored items (Yes/No), 14items scored on a 4-point scale, and one vertical rating item on a5-point scale. Of particular importance to the current investigationwas the influence of dietary restraint on food choice behavior. This21-item subscale of the TFEQ was specifically designed to measurethe extent of control one exerts over food intake with the goal ofinfluencing body weight (Stunkard & Messick, 1985). Higher scoresreflect higher levels of dietary restraint. Participants with a score of10 or greater on this subscale were classified as restrained eaters,which is a similar cut score employed in other investigationsassessing dietary restraint using the TFEQ (e.g., Brunstrom, Yates,& Witcomb, 2004).

Hypothetical food delay discountingSubjects completed a chocolate delay discounting task (1 Her-

shey Milk Chocolate Kiss™ now vs. 1 bag of 75 Chocolate Kisses™in [5–180] min). This instrument was created specifically for thisinvestigation. Because delay discounting varies by type of com-modity (Weatherly, Terrell, & Derenne, 2010), delay discountingwas assessed with respect to (hypothetical) choices between asmall amount of chocolate now vs. larger but delayed amounts ofthe same commodity later. Previous research has failed to detecta difference in responding when participants were offered hypo-thetical vs. real rewards (Lagorio & Madden, 2005; Madden et al.,2004), thus, the use of hypothetical rewards has been supportedas a valid procedure in the study of delay discounting. This delaydiscounting task was administered online and was meant to serveas a proxy measure of impulsivity. Using a median split, partici-pants were classified as high in impulsivity if they achieved a k va-lue of 3.55 or greater (see below, Data Analysis, for a detailedexplanation of k value calculations). Scoring below 3.55 resultedin classification as low in impulsivity. Despite online administra-tion, participants were provided with instructions typical of thoseused in delay discounting research (e.g., Odum & Rainaud, 2003).These instructions were as follows:

I am going to ask you to make some decisions about which rewardsyou would prefer. You will not receive the rewards you choose, but Iwant you to make your decisions as though you were really going toreceive the rewards you choose. You are going to be given the oppor-tunity to choose between different amounts of chocolate after differ-ent delays. There are 18 different questions, such as the following:Would you rather have 1 piece of chocolate NOW or have 1 bag ofchocolate in 180 minutes? Keep in mind, 1 bag is made up of 75pieces of chocolate. If you choose the option following the delay,

please imagine that you would be expected to wait in a laboratoryfor that specified amount of time before being given your reward.Once you receive your reward, however, you would be allowed toleave. You are able to choose only one of the following options.

Data analysis

Procedures to generate demand curves to examine food choiceresponding were the same as described above. In Study 2, however,to assess the influence of individual difference factors such as BMI,dietary restraint, and delay discounting (i.e., impulsivity) on foodchoice behavior, these variables were first dichotomized. All inde-pendent variables were dichotomized because the purpose of thisinvestigation was to generate demand curves for aggregate leveldata. Group curves allow for the analysis of the percentage of thegroup (i.e., consumer market) that chooses the commodity at eachUP, and provides UP information that is not intuitively obviousfrom the analysis of individual-level data.

Individuals with BMI P25 kg/m2 were classified as overweight,and those with BMI <25 kg/m2 classified as not overweight. Re-strained eaters were identified by a score of 10 or greater on thedietary restraint subscale of the TFEQ. Impulsivity classificationswere based on a median split of discounting rates (k values). Oftenreferred to as an impulsiveness parameter (Kirby, Petry, & Bickel,1999), k values were calculated using the standard hyperbolicfunction: V = A/(1 + kD). In this equation, parameter V is the pres-ent value of the immediate reward (i.e., 1 piece of chocolate);parameter A is the value of the delayed reward (i.e., 75 pieces ofchocolate); parameter D is the specified delay (ranging between5 and 180 minutes); and k is the free parameter left to determinediscounting rate (Kirby et al., 1999; Mazur, 1987). Higher k valuesindicate greater discounting of future rewards, thus, higher dis-counting rates (k values) correspond to heightened levels ofimpulsivity.

Results

The three independent variables (i.e., BMI, dietary restraint, andimpulsivity) were not significantly correlated (see Table 1).

During concurrent choice sessions, demand for the high-sugar/high-fat (Hershey Kiss™) and low-sugar/high-fat food (cheesecube) did not significantly differ for the overall sample (Pmax = 449vs. 388), but there were significant subgroup differences in concur-rent food choice. Notably, demand for the high-sugar/high-fat food(Fig. 2, top left panel) was more inelastic for overweight (n = 11)than non-overweight participants (n = 9), Pmax = 586 vs. 351,F(1,16) = 14.71, p < 0.01. In contrast, demand for the low-sugar/high-fat food (Fig. 2, top right panel) was more inelastic for non-overweight than overweight participants, Pmax = 577 vs. 339,F(1,17) = 5.06, p < 0.05. Chocolate relative to cheese choice differ-ences were significant only in participants who were overweight,Pmax = 586 vs. 339, F(1,16) = 19.22, p < 0.001, but fell short of signif-icance in those who were non-overweight, Pmax = 351 vs. 577,F(1,17) = 4.17, p = 0.06.

Demand for the high-sugar/high-fat food was significantly moreinelastic for Restrained (n = 12) than Unrestrained eaters (n = 9),

Fig. 2. Top panel: demand for high-sugar/high-fat (Hershey Milk Chocolate Kiss™) vs. low-sugar/high-fat (Kraft Cheddar Cheese Cube™) in concurrent choice sessions as afunction of BMI (High BMI = overweight [BMI P25 kg/m2], Low BMI = not overweight [BMI <25 kg/m2]). In the top left panel, Pmax values are provided above the demandcurve for the high BMI group and below the demand curve for low BMI group. In the top right panel, Pmax values are provided above the demand curve for the low BMI groupand below the demand curve for high BMI group. Middle panel: demand for high-sugar/high-fat (Hershey Milk Chocolate Kiss™) vs. low-sugar/high-fat (Kraft Cheddar CheeseCube™) in concurrent choice sessions among restrained and unrestrained eaters. In the middle left panel, Pmax values are provided above the demand curve for restrainedeaters and below the demand curve for those unrestrained eaters. In the middle right panel, Pmax values are provided above the demand curve for unrestrained eaters andbelow the demand curve for restrained eaters. Bottom panel: demand for high-sugar/high-fat (Hershey Milk Chocolate Kiss™) vs. low-sugar/high-fat (Kraft Cheddar CheeseCube™) in concurrent choice sessions as a function of impulsivity. Pmax values are provided above the demand curve for those classified as low in impulsivity and below thedemand curve for the high-impulsivity group.

S. Reslan et al. / Appetite 59 (2012) 364–371 369

Pmax = 632 vs. 297, F(1,15) = 27.36, p < 0.001 (Fig. 2, middle left pa-nel), whereas for the low-sugar/high-fat food, the opposite patternwas observed (i.e., more inelastic for Unrestrained than Restrainedeaters), Pmax = 278 vs. 711, F(1,15) = 11.91, p < 0.01 (Fig. 2, middleright panel). For Restrained eaters, demand was more inelasticfor chocolate than cheese, Pmax = 632 vs. 278, F(1,14) = 15.59,

p < 0.01; whereas, for Unrestrained eaters, demand for chocolatewas more elastic than cheese, Pmax = 297 vs. 711, F(1,16) = 21.99,p < 0.001.

Participants were classified from their chocolate delaydiscounting rates (k values) as High-Impulsive (n = 9) or Low-Impulsive (n = 11). Demand for the high-sugar/high-fat option

370 S. Reslan et al. / Appetite 59 (2012) 364–371

(Fig. 2, bottom left panel) was significantly more elastic for High-vs. Low-Impulsive subjects, Pmax = 407 vs. 546, F(1,17) = 5.89,p < 0.05. Similarly, demand for the low-sugar/high-fat option(Fig. 2, bottom right panel) was more elastic for High- vs. Low-Impulsive subjects, Pmax = 217 vs. 594, F(1,16) = 26.73, p < 0.001.For High-Impulsive subjects, demand for Hershey Kisses™ was sig-nificantly more inelastic than for cheese, F(1,14) = 14.73, p < 0.01;however, for Low-Impulsive subjects, demand did not significantlydiffer for chocolate vs. cheese, F(1,19) = 0.41, p = 0.53.

Discussion

Findings from Study 2 validate that demand curves for foodchoices comparable in mass, calories, and fat, but differing in sugarcontent, can be generated and differentiated. In particular, demandfor the high-sugar/high-fat food was more inelastic for restrainedvs. unrestrained eaters, and for those who were overweight vs. nor-mal weight, relative to the concurrent low-sugar/high-fat option.This observation is consistent with findings from animal researchthat show sugar content may enhance food reward value by acti-vating DA and endogenous opioid reward pathways (e.g., Avena,Rada, & Hoebel, 2008; Avena, Rada, Moise, & Hoebel, 2006). More-over, in laboratory studies of human food choice behavior, sugarcontent has also been demonstrated to be a property of food thatdrives reward processes (Epstein, Carr, Lin, & Fletcher, 2011).

Consistent with our initial hypothesis, significant group differ-ences in food demand were noted for individuals who are over-weight, restrained eaters, and those higher in impulsivity.Demand curves were more inelastic for chocolate (and more elasticfor cheese) for those who were overweight relative to their non-overweight counterparts, consistent with the notion that sweetfoods may be more reinforcing for obese individuals (e.g., Forslund,Torgerson, Sjostrom, & Lindroos, 2005). Among restrained eaters,demand for the more palatable food option (i.e., chocolate) wasmore inelastic than the non-palatable option (i.e., cheese), whilethe reverse pattern was observed for unrestrained eaters. Findingssupport the commonly held notion that attempts to restrict eatingmay specifically enhance desire/responding for more palatablefoods (e.g., Giesen et al., 2008; Giesen et al., 2010; Polivy, Coleman,& Herman, 2005). With respect to impulsivity, participants classi-fied as high in impulsivity were more price-sensitive (i.e., demandcurves were more elastic) to both reinforcers (i.e., chocolate andcheese); this finding aligns with the idea that impulsive subjectsdesire smaller, more immediate outcomes (Bickel et al., 1998). Dis-counting in the present study was defined with respect to choco-late, yet high (chocolate) impulsive subjects exhibited similardemand for both chocolate and cheese. Thus, the degree to whichthis task reflects a general vs. specific tendency is not yet clear. Gi-ven the current findings, it is likely that other important individualdifferences impacting food choice behavior exist, and future behav-ioral economic research should focus on better understandingthese relationships.

This study was not without limitations. In particular, the major-ity of the analyses used to assess individual difference characteris-tics and their influence on food choice behavior were exploratoryin nature. Although the selection of independent variables andtheir hypothesized association with food seeking behavior wasguided by extant literature (e.g., Clark et al., 2010; Giesen, Haver-mans, Douven et al., 2010; Giesen et al., 2008; Rollins et al.,2010; Saelens & Epstein, 1996), future research should replicatethese findings with a larger sample. It must be noted that becauseour sample consisted of only 21 participants, testing multiplehypotheses using (independent) subgroups of the same sampleelevates the risk for making a Type I error.

Next, although we purport that the food comparators used inthis investigation differed only on sweetness, the two food choices

also differed in texture and smell, which could have influencedtheir relative reinforcing value. Thus far, our efforts have focusedon the influence of nutrient content (e.g., sugar and fat) on thereinforcing value of food. However, other aspects of food (e.g., tex-ture, smell) may influence the reinforcing value of a commodity,and the behavioral economic approach offers an advantageousmethod for exploring these different elements which may influ-ence food choice responding. Finally, the only component of UPthat was varied in this initial investigation was response require-ment. Given that the UP metric can incorporate other cost/benefitcomponents (e.g., food unit amount, response requirement for thenon-palatable alternative), such parameters could be systemati-cally varied to evaluate whether food demand is functionallyequivalent, as has been found in animal laboratory studies (Hursh,1984). The relationships derived in this experiment provide sup-port for the validity of the behavioral economic approach to studyindividual differences that may influence food demand. Future re-search should assess whether those who exhibit greater demandutilizing this behavioral economic approach also consume morefood in an ad libitum eating task, to provide additional evidenceon the construct validity of this task.

Summary and concluding discussion

The primary purpose of this investigation was two-fold. We firstsought to validate a human laboratory model of food-appetitivebehavior, then to examine associations between food choiceresponding and individual differences such as BMI, dietary re-straint, and impulsivity using a classical demand curve approach.Findings suggest that more demand-inelastic (price-resistant)behavior was evident for the high-sugar/high-fat food (chocolate),however, the relative value of this option partly depended on thetype of choice procedure (i.e., consecutive or concurrent) and thefood comparator (i.e., low-sugar/low-fat or low-sugar/high-fat).Significant group differences in food demand elasticity were alsoinfluenced by BMI (overweight vs. non-overweight), dietary re-straint (restrained vs. unrestrained), and impulsivity (high vs.low). Earlier, two fundamental principles important to drug seek-ing behavior were outlined. That is, consumption of a drug tendsto decrease when constraints on access are increased and con-sumption is affected by the availability of concurrent alternativereinforcers. Together, these two studies provide evidence to sup-port that this same model is applicable to the study of fooddemand.

This investigation is the first human study of food choice behav-ior to utilize a classic behavioral economic approach to generatedemand curves for aggregate level data. Importantly, group curvesgive unit-price information that is not intuitively obvious fromother forms of analysis. Our results suggest that demand functionssystematically track the classical effect of UP on consumption but,perhaps more importantly, exhibit variations across differentconsumer subgroups (e.g., overweight vs. non-overweight). Simul-taneous analysis of food ‘‘purchasing’’ (a bi-tonic function of UP)and consumption (a monotonically decelerating function of UP)affords an innovative perspective that is not evident in simplefixed-operant or choice paradigms. The drawback of this approach,however, is that when participants only have a single opportunityto choose a food option at a given UP (i.e., no within-subjectreplications), as occurred here, Pmax values are not available atthe subject-level of analysis.

The behavioral economic approach to understanding foodchoice behavior and related outcomes has potential to advanceour understanding of factors that may be driving the obesity epi-demic and confounding individuals’ efforts to overcome weightproblems. However, as this initial project scheduled only a re-stricted number of comparisons among many possible food choice

S. Reslan et al. / Appetite 59 (2012) 364–371 371

situations that individuals may encounter, additional research iswarranted and necessary to better understand the microeconomicfactors that may be contributing to a macroeconomic phenome-non, the obesity epidemic. In conclusion, this is the first study todemonstrate that demand curves akin to those generated usingnon-linear regression methods in studies of drug choice can alsobe generated for food choice, further supporting the observationof commonalities across food and drug-seeking behavior (Davis &Carter, 2009). Future research should aim to apply this model toadvance our understanding of how certain types of choice behaviormay pose obstacles to overcoming eating and weight disorders.

References

Avena, N. M., Rada, P., & Hoebel, B. G. (2008). Evidence for sugar addiction.Behavioral and neurochemical effects of intermittent, excessive sugar intake.Neuroscience and Biobehavioral Reviews, 32(1), 20–39.

Avena, N. M., Rada, P., Moise, N., & Hoebel, B. G. (2006). Sucrose sham feeding on abinge schedule releases accumbens dopamine repeatedly and eliminates theacetycholine satiety response. Neuroscience, 139, 813–820.

Bickel, W. K., DeGrandpre, R. J., Hughes, J. R., & Higgins, S. T. (1990). Behavioraleconomics of drug self-administration. I. Functional equivalence of responserequirement and drug dose. Life Sciences, 47, 1501–1510.

Bickel, W. K., Madden, G. J., & Petry, N. M. (1998). The price of change. Thebehavioral economics of drug dependence. Behavioral Therapy, 29, 545–565.

Brunstrom, J. M., Yates, H. M., & Witcomb, G. L. (2004). Dietary restraint andheightened reactivity to food. Physiology & Behavior, 81(1), 85–90.

Christakou, A., Brammer, M., & Rubia, K. (2011). Maturation of limbic corticostriatalactivation and connectivity associated with developmental changes in temporaldiscounting. Neuroimage, 54, 1344–1354.

Clark, E. N., Dewey, A. M., & Temple, J. L. (2010). Effects of daily snack food intake onfood reinforcement depend on body mass index and energy density. TheAmerican Journal of Clinical Nutrition, 91, 300–308.

Collier, G. H., Johnson, D. F., Hill, W. L., & Kaufman, L. W. (1986). The economics ofthe law of effect. Journal of the Experimental Analysis of Behavior, 46(2), 113–136.

Dawe, S., & Loxton, N. J. (2004). The role of impulsivity in the development ofsubstance use and eating disorders. Neuroscience & Biobehavioral Reviews, 28(3),343–351.

Davis, C., & Carter, J. C. (2009). Compulsive overeating as an addiction disorder. Areview of theory and evidence. Appetite, 53(1), 1–8.

Epstein, L. H., Dearing, K. K., Temple, J. L., & Cavanaugh, M. D. (2008). Foodreinforcement and impulsivity in overweight children and their parents. EatingBehaviors, 9, 319–327.

Epstein, L. H., & Leddy, J. J. (2006). Food reinforcement. Appetite, 46(1), 22–25.Epstein, L. H., Leddy, J. J., Temple, J. L., & Faith, M. S. (2007). Food reinforcement and

eating. A multilevel analysis. Psychological Bulletin, 133(5), 884–906.Epstein, L. H., Carr, K. A., Lin, H., & Fletcher, K. D. (2011). Food reinforcement, energy

intake, and macronutrient choice. The American Journal of Clinical Nutrition, 94,12–18.

Epstein, L. H., Salvy, S. J., Carr, K. A., Dearing, K. K., & Bickel, W. K. (2010). Foodreinforcement, delay discounting, and obesity. Physiology & Behavior,100(2010), 438–445.

Epstein, L. H., Temple, J. L., Neaderhiser, B. J., Salis, R. J., Erbe, R. W., & Leddy, J. J.(2007). Food reinforcement, the dopamine D2 receptor genotype, and energyintake in obese and nonobese humans. Behavioral Neuroscience, 121(5),877–886.

Flegal, K. M., Carroll, M. D., Ogden, C. L., & Curtin, L. R. (2010). Prevalence and trendsin obesity among US adults, 1999–2008. JAMA, 303(3), 235–241.

Forslund, H. B., Torgerson, J. S., Sjostrom, L., & Lindroos, A. K. (2005). Snackingfrequency in relation to energy intake and food choices in obese men andwomen compared to a reference population. International Journal of Obesity, 29,711–719.

Giesen, J. C. A. H., Havermans, R. C., Douven, A., Tekelenburg, M., & Jansen, A. (2010).Will work for snack food. The association of BMI and snack reinforcement.Obesity, 18, 966–970.

Giesen, J. C. A. H., Havermans, R. C., Nederkoon, C., Strafaci, S., & Jansen, A. (2008).Working harder to obtain more snack foods when wanting to eat less. BehaviourResearch and Therapy, 47, 13–17.

Giesen, J. C. A. H., Havermans, R. C., & Jansen, A. (2010). Substituting snacks withstrawberries and sudokus. Does restraint matter? Health Psychology, 29(2),222–226.

Goldfield, G. S., & Legg, C. (2006). Dietary restraint, anxiety, and the relativereinforcing value of snack food in non-obese women. Eating Behaviors, 7,323–332.

Goldfield, G. S., & Lumb, A. (2008). Smoking, dietary restraint, gender, and therelative reinforcing value of snack food in a large university sample. Appetite,50(2–3), 278–289.

Goldfield, G. S., & Epstein, L. H. (2002). Can fruits and vegetables and activitiessubstitute for snack foods? Health Psychology, 21, 299–303.

Greenwald, M. K. (2010). Effects of experimental unemployment, employment andpunishment analogs on opioid seeking and consumption in heroin-dependentvolunteers. Drug and Alcohol Dependence, 111, 64–73.

Greenwald, M. K., & Hursh, S. R. (2006). Behavioral economic analysis of opioidconsumption in heroin-dependent individuals. Effects of unit price and pre-session drug supply. Drug and Alcohol Dependence, 85, 35–48.

Greenwald, M. K., & Steinmiller, C. L. (2009). Behavioral economic analysis of opioidconsumption in heroin-dependent individuals. Effects of alternative reinforcermagnitude and post-session drug supply. Drug and Alcohol Dependence, 104,84–93.

Hedley, A. A., Ogden, C. L., Johnson, C. L., Carroll, M. D., Curtin, L. R., & Flegal, K. M.(2004). Prevalence of overweight and obesity among US children, adolescents,and adults, 1999–2002. JAMA, 291(23), 2847–2850.

Houben, K., Roefs, A., & Jansen, A. (2010). Guilty pleasures. Implicit preferences forhigh calorie food in restrained eating. Appetite, 55, 18–24.

Hursh, S. R. (1980). Economic concepts for the analysis of behavior. Journal of theExperimental Analysis of Behavior, 34, 219–238.

Hursh, S. R. (1984). Behavioral economics. Journal of the Experimental Analysis ofBehavior, 42, 435–452.

Hursh, S. R., & Winger, G. (1995). Normalized demand for drugs and otherreinforcers. Journal of Experimental Analysis of Behavior, 64(3), 373–384.

Kirby, K. N., Petry, N. M., & Bickel, W. K. (1999). Heroin addicts have higher discountrates for delayed rewards than non-drug-using controls. Journal of ExperimentalPsychology: General, 128(1), 78–87.

Lagorio, C. H., & Madden, G. J. (2005). Delay discounting of real and hypotheticalrewards III. Steady-state assessments, forced-choice trials, and all real rewards.Behavioural Processes, 69, 173–187.

MacKillop, J., & Murphy, J. G. (2007). A behavioral economic measure of demand foralcohol predicts brief intervention outcomes. Drug and Alcohol Dependence, 89,227–233.

Madden, G. J., & Bickel, W. K. (1999). Abstinence and price effects on demand forcigarettes. A behavioral-economic analysis. Addiction, 94, 577–588.

Madden, G. J., Dake, J. M., Mauel, E. C., & Rowe, R. R. (2005). Labor supply andconsumption of food in a closed economy under a range of fixed- and random-ratio schedules. Tests of unit price. Journal of the Experimental Analysis ofBehavior, 83, 99–118.

Madden, G. J., Raiff, B. R., Lagorio, C. H., Begotka, A., Mueller, A., Hehli, D., et al.(2004). Delay discounting of potentially real and hypothetical rewards. Part II.Between- within- subject comparisons. Experimental and ClinicalPsychopharmacology, 12, 251–261.

Mathis, C. E., Johnson, D. F., & Collier, G. (1996). Food and water intake as functionsof resource consumption costs in a closed economy. Journal of the ExperimentalAnalysis of Behavior, 65, 527–547.

Mazur, J. E. (1987). An adjusting procedure for studying delayed reinforcement. InM. L. Commons, J. E. Mazur, J. A. Nevin, & H. Rachlin (Eds.), Quantitative analysesof behavior. The effect of delay and of intervening events on reinforcement value(Vol. 5, pp. 55–73). Hillsdale, NJ: Erlbaum.

Odum, A. L., & Rainaud, C. P. (2003). Discounting of delayed hypothetical money,alcohol, and food. Behavioral Processes, 64, 305–313.

Ogden, C. L., Carroll, M. D., Curtin, L. R., McDowell, M. A., Tabak, C. J., & Flegal, K. M.(2006). Prevalence of overweight and obesity in the United States, 1999–2004.JAMA, 295(13), 1549–1555.

Perry, J. L., Larson, E. B., German, J. P., Madden, G. J., & Carroll, M. E. (2005).Impulsivity (delay discounting) as a predictor of acquisition of IV cocaine self-administration in female rats. Psychopharmacology, 178(2–3), 193–201.

Polivy, J., Coleman, J., & Herman, C. P. (2005). The effect of deprivation on foodcravings and eating behavior in restrained and unrestrained eaters. InternationalJournal of Eating Disorders, 38, 301–309.

Raynor, H. A., & Epstein, L. H. (2003). The relative-reinforcing value of food underdiffering levels of food deprivation and restriction. Appetite, 40, 15–24.

Rollins, B. Y., Dearing, K. K., & Epstein, L. H. (2010). Delay discounting moderates theeffect of food reinforcement on energy intake among non-obese women.Appetite, 55, 420–425.

Saelens, B. E., & Epstein, L. H. (1996). Reinforcing value of food in obese and non-obese women. Appetite, 27, 41–50.

Smith, J. A., & Epstein, L. H. (1991). Behavioral economic analysis of food choice inobese children. Appetite, 17, 91–95.

Stunkard, A. J., & Messick, S. (1985). The Three-Factor Eating Questionnaire tomeasure dietary restraint, disinhibition and hunger. Journal of PsychosomaticResearch, 29(1), 71–83.

Temple, J. L., Bulkley, A., Badaway, R. L., Krause, N., McCann, S., & Epstein, L. H.(2009). Differential effects of daily snack food intake on the reinforcing value offood in obese and nonobese women. The American Journal of Clinical Nutrition,90(2), 304–313.

Temple, J. L., Chappel, A., Shalik, J., Volcy, S., & Epstein, L. H. (2008a). Dailyconsumption of individual snack foods decreases their reinforcing value. EatingBehaviors, 9, 267–276.

Temple, J. L., Legierski, C. M., Giacomelli, A. M., Salvy, S. J., & Epstein, L. H. (2008b).Overweight children find food more reinforcing and consume more energy thando nonoverweight children. The American Journal of Clinical Nutrition, 87(5),1121–1127.

Vuchinich, R. E., & Tucker, J. A. (1988). Contributions from behavioral theories ofchoice as a framework to an analysis of alcohol abuse. Journal of AbnormalPsychology, 92, 408–416.

Weatherly, J. N., Terrell, H. K., & Derenne, A. (2010). Delay discounting of differentcommodities. Journal of General Psychology, 137(3), 273–286.

Yang, C., & Chiou, W. (2010). Substitution of healthy for unhealthy beverages amongcollege students. A health-concerns and behavioral-economics perspective.Appetite, 54, 512–516.