The Psychological Record

https://doi.org/10.1007/s40732-021-00475-w

ORIGINAL ARTICLE

The Gym Membership Purchase Task: Early Evidence Towards

Establishment of a Novel Hypothetical Purchase Task
Jeremiah Brown 1,2 & Wendy Donlin Washington 3 & Jeffrey S. Stein 1,2 & Brent A. Kaplan 4

Accepted: 31 May 2021

# Association for Behavior Analysis International 2021

Abstract
Hypothetical purchase tasks are a widely used tool to determine the reinforcing value of commodities, especially commodities
which are difficult to deliver experimentally. Amazon Mechanical Turk users (N = 375) completed two novel hypothetical
purchase tasks (quantity of purchase and probability of purchase) and other measures to estimate how much an individual values
the opportunity to exercise in a gym. We examined correlations between demand indices generated by each task and measures
related to physical activity. In addition, we compared rates of systematic and nonsystematic responding between the two tasks.
Exploratory analyses of demand indices and measures of physical activity suggest initial evidence of construct validity for each
task. When accounting for an order effect, the probability of purchase task generated significantly lower rates of nonsystematic
responding compared to the quantity of purchase task (2.87% vs. 14.6%, respectively). We discuss how these results may inform
improved construction of future hypothetical purchase tasks and implications for using likelihood and quantity purchase tasks.

Keywords Behavioral economics . Hypothetical purchase task . Value of physical activity . Exercise . Demand

Sedentary individuals are at increased risk of heart disease and Americans older than 15 years of age engaged in some exer-
heart attack, high blood pressure, high cholesterol, being over- cise or sport activity on a typical day (Bureau of Labor
weight or obese, diabetes, and stroke, and are at higher risk of Statistics, 2016). This failure to meet recommended guidelines
dying from cardiovascular disease (Pandey et al., 2016; suggests that, for many individuals, the value of physical ac-
Warren et al., 2010). However, according to the Centers for tivity may be relatively lower compared to the value of sed-
Disease control (CDC, n.d.), about 76% of American adults in entary alternatives. This may not be surprising given seden-
2018 did not engage in enough physical activity per week to tary activities often require little effort and may be more im-
meet the CDC’s current physical activity guidelines (i.e., mediately reinforcing. However, methods for direct estima-
150 min of moderate-intensity or 75 min of vigorous- tion of how individuals’ value physical activity are currently
intensity aerobic physical activity, and 2 days of full-body limited to in-lab procedures (Barkley et al., 2009; Flack et al.,
muscle strengthening activities of at least moderate intensity; 2017). Development of widely disseminable and accessible
Piercy et al., 2018). From 2011 to 2015, only 18% of methods may be critical in the study of motivating factors
underlying physical activity and assessing the effects of inter-
ventions designed to increase such motivation.
* Jeremiah Brown The hypothetical purchase task (HPT) is a self-report ques-
[email protected] tionnaire that has been used extensively to measure behavioral
economic demand when the commodity would be prohibitive-
1
Center for Transformative Research on Health Behaviors, Fralin
ly difficult or unethical to deliver (e.g., Jacobs & Bickel, 1999;
Biomedical Research Institute at VTC, Roanoke, VA, USA Murphy et al., 2005; Murphy & MacKillop, 2006; Strickland
2
Department of Human Nutrition, Foods, and Exercise, Virginia
et al., 2020). In the most common case, the HPT first describes
Polytechnic Institute and State University, Blacksburg, VA, USA hypothetical conditions regarding the imagined scenario, then
3
Department of Psychology, University of North Carolina,
asks participants to indicate the quantities of a commodity
Wilmington, NC, USA they would like to purchase across a range of prices (Roma
4
Department of Family and Community Medicine, University of
et al., 2017). The features of the demand curve generated by
Kentucky, Lexington, KY, USA HPTs and the parameters calculated while modeling, known
 Psychol Rec

as demand indices, provide insights regarding the value of the a single purchase as the framework to measure consumption
commodity under study. Demand intensity (Qo) is the amount (in addition to the traditional quantity framework) may gener-
demanded (or probability of consumption) by the individual ate meaningful data for commodities that are typically pur-
when the commodity is free or very close to free. The rate of chased infrequently or in single units as well as for other
change by which an individual decreases their consumption in commodities. However, the authors note that probability de-
face of increasing costs can be conceptualized as the sensitiv- mand indices were always greater than quantity demand indi-
ity of consumption to price changes and is known as the rate of ces including essential value (up to 10 times greater) and ad-
change in elasticity (α; Gilroy et al., 2020; Roma et al., 2017). vised caution when making direct comparisons between de-
These quantity-based HPTs have been widely used to mand indices generated by each task type.
estimate the degree to which an individual values health- In the development of novel HPTs to estimate valuation of
relevant commodities (e.g., alcohol, cigarettes, obesogenic physical activity, similar explorations of convergent validity
foods, anabolic-androgenic steroids; Few et al., 2012; Pope between demand indices and validated clinical scales is an
Jr. et al., 2010), to identify individuals who respond to important early step. The degree to which demand indices
treatment (Madden & Kalman, 2010; Snider et al., 2016; correlate with theoretically relevant, real-world behavior pro-
Stein et al., 2018), and to examine the influence of inter- vides information for between-task comparisons. Demand for
ventions (Epstein et al., 2010; Grace et al., 2015; Murphy alcohol, commonly measured using quantity-based hypothet-
et al., 2015). However, no studies to our knowledge have ical alcohol purchase tasks (APTs; Kaplan et al., 2018), cor-
applied HPT methods to estimate valuation of exercise. relates with a wide range of theoretically consistent behaviors
This may be because, unlike consumption of alcohol or or participant characteristics, including heavier self-reported
other commodities examined frequently using HPTs, indi- weekly drinking (MacKillop et al., 2009) and frequency of
viduals do not engage in discrete quantities of physical heavy-drinking episodes (Murphy & MacKillop, 2006). In
activity. Instead, exercise occurs over varying durations addition, close correspondence has been observed between
of time and with varying frequency. Moreover, like many estimated consumption and actual consumption evaluated
health-related behaviors, exercise may occur infrequently, using the APT and cigarette purchase tasks (Amlung et al.,
is vulnerable to satiation, and would be impossible to ob- 2012; MacKillop et al., 2008; Wilson et al., 2016). These
serve and deliver consequences in natural or contrived en- associations identify behaviors for which the task may be
vironments. As a result, exercise may be less amenable to predictive, strengthening the utility of the purchase task and
assessment using an HPT framework in which participants suggesting convergent validity. Indeed, demonstrations of
report the quantity of a given commodity they would like convergent validity between established measures and de-
to purchase. In the example of gym membership, most mand indices have been observed in alcohol, cigarette, and
individuals purchase one gym membership at a time, yet indoor-tanning HPTs (MacKillop et al., 2008; Reed et al.,
pay recurring (e.g., monthly or yearly) membership fees to 2016). No data has yet explored similar convergent validity
gain access to an environment in which it is easier to en- with physical activity.
gage in physical activity. It is unclear, however, if an adap- In addition to comparisons with real-world behaviors, one
tion of the HPT using a quantity purchased framework will characteristic that may provide insight regarding how well a
adequately measure demand for gym membership. task is suited to the specific commodity is the degree to which
Recent probability HPTs (Roma et al., 2016) ask partici- the task results in nonsystematic response patterns. Stein et al.
pants to indicate the probability that they would purchase a (2015) described a method to identify nonsystematic re-
commodity on a scale of 0 to 100 rather than indicating unit sponses using a three-criteria algorithm operating under two
consumption. Roma et al. (2016) examined the effects of price assumptions: a reduction in consumption will occur from the
density and purchase task type (probability or quantity) on lowest price to the highest price (i.e., the Law of Demand),
measures of behavioral economic demand for six commodi- and that the change in consumption across price increases will
ties. Three of the commodities were considered “big ticket be directionally consistent. The rate at which a task generates
items” (e.g., refrigerators, vacation packages, fine-dining nonsystematic responding may be indicative of the interpret-
meals), and the remaining three were identified as “small tick- ability of the task or important characteristics the study popu-
et items” (e.g., hamburgers or sandwiches, rolls of toilet paper, lation and can serve as an additional comparator between
pay-per-view movies). Both task types allow researchers to quantity of purchase and probability of purchase tasks. Thus,
plot consumption (or probability thereof) on the y-axis and the present study sought to: (1) explore possible relationships
price on the x-axis, which may then be modeled using nonlin- between demand indices and variables theoretically related to
ear regression (Hursh & Silberberg, 2008). The authors found demand for gym membership or access, and (2) compare
both probability and quantity HPTs generated consumption which variant of an HPT would generate higher rates of sys-
data well-described by Hursh and Silberberg’s (2008) expo- tematic responding for a “big ticket” commodity novel to
nential model of demand and suggest that using probability of behavioral economic demand analysis (i.e., gym
Psychol Rec

membership). To this end, a survey containing two variations between 0 and 100 for each price. The probability task prices
of an HPT for gym membership (i.e., quantity, probability of were $0.01, $10, $15, $20, $25, $30, $35, $40, $45, $50, $60,
purchase) and measures of self-reported exercise behavior, the $75, $100, $150, $250, $500, $1,000. The quantity task prices
Exercise Addiction Inventory (EAI; Terry et al., 2004) and were $0.01, $0.50, $1, $1.5, $2, $2.5, $3, $ 3.5, $4, $4.5, $5,
Global Physical Activity Questionnaire (GPAQ; Armstrong $7.5, $10, $25, $50, $100, $200. The probability purchase
& Bull, 2006), were administered using an online task stated:
crowdsourcing service. Demand indices generated from each
task were compared to scores on measures of exercise A gym opens near your place of residence that is open
addiction and physical activity status; rates of nonsystematic 24 hours a day and has all of the equipment, services,
and systematic responding were calculated according to the and utilities that you want to use. Assume that there are
Stein et al. (2015) criteria and compared between both varia- no other gyms available for you to join, you have no
tions of the task. access to other gyms or exercise equipment, and that
you have the same income and savings that you have
now. What is the likelihood that you sign up for a month
Method of membership at this gym at the given price? Use a
value of 0 if you would never consider joining this
Participants gym. Use a value between 0–100 to indicate the likeli-
hood that you would join at the given price. Use a value
A total of 375 adults living in the United States were recruited of 100 to indicate that you would always sign up at the
using Amazon Mechanical Turk (mTurk; Strickland & given price.
Stoops, 2019). To be eligible to complete the survey, partici-
pants were required to be at least 18 years old, geographically The quantity purchase task stated:
located in the United States, and to have an approval rate of
85% or greater for previous mTurk surveys. Participation was A gym opens near your place of residence that is open
voluntary; participants could withdraw from the study at any 24 hours a day and has all of the equipment, services,
time, and consent was implied by respondents’ choice indicat- and utilities that you want to use. Assume that there are
ing agreement to participate. The procedures were determined no other gyms available for you to join, you have no
be exempt from full review by the University of North access to other gyms or exercise equipment, and that
Carolina-Wilmington Institutional Review Board under cate- you have the same income and savings that you have
gory (2): educational tests, surveys, interviews, or observa- now. The following questions will ask how many times
tions. Data were collected on March 3, 2018. you would pay to use the gym in the next 30 days if each
visit cost a different price. Please use a value between 0
Procedure and 100 to indicate the number of times you would pay
to use the gym in the next 30 days for each price.
Participants were compensated $0.25 for completing a 20-min
survey of demographic information, current exercise habits in The Exercise Addiction Inventory (EAI) measures an indi-
and outside of gyms or health clubs, and measures pertaining vidual’s likelihood of self-reporting an exercise addiction, al-
to physical activity and exercise addiction (EAI, GPAQ). though the Diagnostic and Statistical Manual of Mental
Finally, participants completed two hypothetical gym- Disorders, Fifth Edition, does not officially recognize exercise
membership purchase tasks (GMPTs) presented in random addiction (Terry et al., 2004). The EAI asks questions
order. concerning the role of exercise in the following domains of
addiction: salience, conflicts, mood modification, tolerance,
Materials withdrawal symptoms, and relapse (Griffiths, 1996).
Participants indicated their agreement with six statements
The probability purchase task asked the probability of pur- using a 5-point Likert scale. The EAI identifies individuals
chasing a month of gym membership at a given price, whereas who are at risk for exercise addiction (a score of 24 or higher),
the quantity purchase task asked how many times the individ- have some symptoms (a score of 13–23), or have no symp-
ual would go to the gym each month if the cost of each visit toms (a score of 0–12).
were a given price. Each purchase task was presented on a The Global Physical Activity Questionnaire (Armstrong &
single page. The page began with the description of the gym Bull, 2006), developed by the World Health Organization in
and the assumptions regarding the membership, followed by a 2002, is used to measure self-reported physical activity in
list of each available price in ascending order. For both tasks, adults. The GPAQ asks 16 questions to estimate an individ-
participants were required to manually enter a number ual’s physical activity levels across work, transportation, and
 Psychol Rec

leisure time, as well as their sedentary behaviors (Cleland log-unit range in price occurs. Nonsystematic trend variations
et al., 2014). The GPAQ classifies individuals into three ac- include nonnegligible reductions in consumption as price in-
tivity groups, based on satisfying weekly physical activity creases, no changes in consumption as price increases, and
recommendations: highly physically active, moderately phys- increases in consumption as price increases.
ically active, and inactive. The reported physical activity Nonsystematic bounce refers to data sets in which the con-
scores have been demonstrated to show moderate correlations sumer indicates increases or “jumps” in consumption greater
with pedometers and accelerometers, though the GPAQ ap- than 25% of initial consumption. Stein et al. (2015) recom-
pears to be valid only for monitoring physical activity and mend a criteria detection limit of 0.1, which will flag re-
change in physical activity. The GPAQ may lack validity sponses that include jumps at 10% or more of the total number
when measuring sedentary behaviors and change in sedentary of price points. Reversals from zero refers to individuals indi-
behaviors (Cleland et al., 2014; Hoos et al., 2012). In addition, cating zero consumption at two prices, and later indicating
the GPAQ has been shown to be valid when self-administered nonzero consumption at a higher price. Reversals may be
(Chu et al., 2015). considered a special case of bounce; it may be presumed that
To measure current exercise habits in and outside of gyms or after prices have risen high enough to completely suppress
health clubs, participants were asked, “In a typical week, how consumption, demand should no longer be indicated at subse-
many minutes do you spend exercising at a gym, fitness center, quent higher prices. However, reversals may occur due to
or health center?” (gym minutes); “In a typical week, how participant error in responses or when the order of prices has
many times do you visit a gym, fitness center, or health center been randomized. Thus, a response is flagged when zero con-
to exercise?” (gym visits); “In a typical week, how many mi- sumption across two or more prices is followed by any rever-
nutes do you spend exercising outside of a gym, fitness center, sals from zero.
or health center?” (outside minutes); and “How much money Data sets were considered nonsystematic according to the
(in USD) do you pay per month for membership to your gym, Stein et al. (2015) criteria outlined above (i.e., the default
fitness center, or health club? Please do not include start up or settings used to detect nonsystematic responding in
initial membership fees” (membership price). beezdemand); however, some additional criteria were applied
due to the uniqueness of the commodity. In the quantity
GMPT, responses at $0.01 price (i.e., the lowest price
Data Analysis assessed) exceeding 62 visits per month were considered non-
systematic; this value was chosen to capture individuals with
R software Version 3.6.2 was used for all data analysis and high rates of gym usage behavior (62 would capture an indi-
figure creation (R Core Team, 2019). The R package vidual who uses the gym twice each day within a 31-day
beezdemand, which utilizes the Stein et al. (2015) algorithm period) while excluding unrealistic levels of responding
and assumptions, was used to identify nonsystematic re- (e.g., 100 or 90 visits per month at zero price). In both tasks,
sponses on both GMPTs and for curve fitting (Kaplan et al., responses of zero consumption at all prices (i.e., no demand)
2019). The three-criterion algorithm includes an expression of were considered systematic. It is possible that some individ-
relative change scores (calculated using the log-unit reduction uals may find no value in gym membership; these responses
in consumption relative to the log-unit change in price), the should not be considered a shortcoming of the purchase task
number of local price-to-price increases (greater than 25% of type. To compare rates of nonsystematic responding between
initial consumption) relative to the total number of prices (i.e., tasks, Fisher’s exact test was used.
bounce), and the absolute number of reversals from zero. The We used the exponentiated model of demand (Koffarnus
relative change expression is as follows: et al., 2015) to fit individual curves (i.e., two-stage approach;
logQ1 −logQn Kaplan et al., in press) on each hypothetical gym membership
ΔQ ¼ ð1Þ purchase task. Values of the span parameter, k, were deter-
logPn −logP1
mined by subtracting the average consumption (in log10 units)
where ΔQ is the relative change in the amount of the com- at the highest price from average consumption (in log10 units)
modity purchased, Q1 is the amount of the commodity pur- at the lowest price and adding 0.5 (daily k = 2.51; probability k
chased at the first available price, Qn is the amount of the = 2.85). We estimated α and Q0 as free parameters. For the
commodity purchased at the last available price, P1 is the first probability task, we constrained Q0 to a maximum of 100
price, and Pn is the last price. The calculated relative change during the fitting process, as the task itself limits probability
score (ΔQ) is then compared to an experimenter–set criterion of consumption from 0 to 100. Demand curves depicting av-
detection limit (X); ΔQ values lower than X would be labeled erage consumption (i.e., fit-to-group approach using all avail-
as nonsystematic. Stein et al. (2015) suggests an X value of able data) for each task were created using similar modeling
0.025; this enables the algorithm to detect data as nonsystem- methods (i.e., determining k values). Only systematic re-
atic if less than a 0.025 log-unit reduction in consumption per sponses were included in group demand curves.
Psychol Rec

Spearman’s rank correlations were used to identify possi- Table 2 Spearman correlations between quantity task demand indices
and exercise-related measures
ble relations between individual demand indices, current ex-
ercise habits in and outside of the gym, membership price, and Variable 1 2 3 4 5 6 7
scores on the EAI and GPAQ. Only demand indices generated
from systematic responses were included in correlations. 1. Alpha (α)
Because these analyses were exploratory in nature, type I error 2. Q0 (derived) -.41**
correction was not utilized. Outliers of demographic values 3. Gym mins. -.42** .31**
(i.e., responses 3 standard deviations or more above or below 4. Gym visits -.49** .38** .91**
the mean on height or income; or unrealistically low re- 5. Outside mins. .13* -.10 -.26** -.24**
sponses, such as height less than 40 in or weight less than 6. EAI -.21** .22** .34** .38** .07
75 lb) were removed from the demographics table (Table 1) 7. GPAQ -.12* .14* .23** .20** .28** .40**
but not removed from demand analyses or correlations be- 8. Mem. price -.45** .23** .62** .64** -.27** .30** .08
cause rates of nonsystematic purchasing data for demographic
Note. * indicates p < .05. ** indicates p < .01
outliers closely resembled that of nonoutlying responses.
Probabilities of demographic outliers and nonoutliers generat-
ing nonsystematic responses on GMPTs are presented. All task demand indices) and Table 3 (probability task demand
tests were considered significant at the .05 alpha level. indices). In both tasks, expected correlations between demand
indices were observed (e.g., strong negative correlations be-
tween α (rate of change in elasticity) and Q0 (model-derived
demand intensity)). In addition, significant correlations were
Results observed between model-derived demand indices, EAI scores,
GPAQ rankings, gym min, gym visits, and membership price.
Sample demographics are presented in Table 1; demographic These correlations were all positive, except for correlations
outliers for some variables have been removed from this table. involving α, as lower scores indicate a lower sensitivity to
The results of the Spearman’s rank correlations between de- changes in price and therefore higher value for the commod-
mand indices, the EAI and GPAQ, exercise habits, and real- ity. The exponentiated equation resulted in good model fits
world gym membership price are depicted in Table 2 (quantity when applied to group-level responding in the quantity (R2 =
0.32, α = .00434, Q0 = 17.88) and probability tasks (R2 = 0.46,
α = .0000641, Q0 = 95.3; see Fig. 1). When fit to each partic-
Table 1 Demographic information describing the sample collected ipants’ demand curve individually, median R2 = 0.91 (IQR:
using mTurk. N = 375 0.84–0.95) and median R2 = 0.82 (IQR: 0.73–0.92) for the
Variable Means (SDs) Min–Max quantity and probability tasks, respectively.
On the probability GMPT, 97.3% (n = 365) of the 375 total
Age 38.04 (12.8) 18–82 responses were considered systematic according to the Stein
% Female 59.7 - et al. (2015) criteria; 9 failed due to violations of trend and 1
BMIa 27.36 (6.76) 15.9–54.9 due to a reversal from zero. On the quantity GMPT, 75.3% (n
b
Income $47,531.14 ($35,392.9) $0–$220,000 = 284) of the 375 total responses were considered systematic.
Mem. Price $16.7 ($43.48) $0–$500 Of the 91 nonsystematic responses, 20 failed due to violations
Gym Minutes/week 103.3 (155.31) 0–1,200 of either trend (17), or reversals from zero (3; one participant
Visits 2.6 (8.01) 0–120 both violated the trend and reversal criteria); 71 were consid-
EAI 16.87 (4.93) 6–30 ered nonsystematic due to initial reported consumption of
GPAQ greater than 62 visits per month. Results of the Fisher’s exact
High n = 179 - test indicated that the probability task generated significantly
Moderate n = 78 - lower rates of nonsystematic responding than the quantity task
Inactive n = 118 - (2.7% vs. 24.7% respectively, p < .001).
Fisher’s exact test was used to compare differences in the
Note: Mem. Price denotes self-reported actual gym membership price frequency of generating nonsystematic responding on the
among participants who currently pay for membership. Minutes denotes
the self-reported number of minutes spent exercising in a gym in a typical
quantity GMPT while accounting for the order in which the
week. Visits denotes the self-reported number of visits to a gym in a GMPTs were completed. Participants who took the quantity
typical week. a n = 351 after removing outliers for height and weight. task first (n = 196) generated nonsystematic responding on the
b
n = 373 after removing outliers for incomes. Outliers were considered quantity task at a lower rate than those who took the quantity
responses greater or less than 3 standard deviations above or below the
mean, or unrealistically low values of height (less than 40 in) or weight
task second (n = 179; 14.6% vs. 58.4%, respectively, p <
(less than 75 lb). .001). A similar order effect was not observed among
 Psychol Rec

indicating that participants who took the quantity task first

generated nonsystematic responding at a higher rate than par-
ticipants who took the probability task first (n = 25; 14.6% vs.
n = 5; 2.87%, respectively, p < .001, see Fig. 2).
Of the total 375 responses, 26 were identified as demo-
graphic outliers based on values of height, weight, or income.
Of the 349 responses that were not considered outliers, 342
(97.9%) were found to provide systematic responses on the
likelihood task, whereas 265 (75.9%) provided systematic re-
sponses on the daily task—closely matching the rates of non-
systematic responding in the full sample including demo-
graphic outliers. Of the 26 individuals identified as demo-
graphic outliers, 23 responses on the likelihood task were
considered systematic (88.5%), whereas 19 responses on the
daily task were considered systematic (73%). Table 4 displays
rates of systematic responding on each task by demographic
outliers and nonoutliers.

Discussion

The goal of the study was to explore correlations between de-

Fig. 1 Demand curves depicting average probability of purchase (i.e., 1- mand indices and exercise-related measures and examine the
month membership; top panel) and average quantity purchased (i.e., num- impact of purchase task type on systematicity in an infrequently
ber of visits per month; bottom panel) at each price. Note. Error bars purchased commodity novel to operant demand analyses. Initial
denote 1 standard error of the mean
analyses indicated directionally expected correlations between

nonsystematic responses on the probability GMPT between

those who took the probability task first (n = 179) and the
probability task second (n = 196; 2.87% and 2.62%, respec-
tively, p > .05). In order to compare rates of systematic and
nonsystematic responding without the influence of the order
effect, participants who completed the quantity task first (n =
196) were compared with participants who took the probabil-
ity task first (n = 179). Significant differences in rates of non-
systematic responding between the groups was observed,

Table 3 Spearman correlations between probability task demand

indices and exercise-related measures

Variable 1 2 3 4 5 6 7

1. Alpha (α)
2. Q0 (derived) -.40**
3. Gym mins. -.38** .27**
4. Gym visits -.36** .26** .89**
5. Outside mins. .09 -.04 -.25** -.26**
6. EAI -.12* .18** .38** .39** .01
7. GPAQ -.15** .09 .24** .21** .26** .37**
8. Mem. price -.34** .20** .59** .63** -.24** .28** .08
Fig. 2 Percent of responses flagged as nonsystematic in each task,
Note. * indicates p < .05. ** indicates p < .01 separated by the order in which participants completed each GMPT
Psychol Rec

Table 4 Crosstabulation of
demographic outliers and Systematic Demand Nonsystematic Demand Row Totals
nonsystematic responses
Likelihood Task
Demographic Non-outliers 342 (91%) 7 (2%) 349 (93%)
Demographic Outliers 23 (6%) 3 (1%) 26 (7%)
Column Totals 365 (97%) 10 (3%) 375 (100%)
Quantity Task
Demographic Non-outliers 265 (71%) 84 (22%) 349 (93%)
Demographic Outliers 19 (5%) 7 (2%) 26 (7%)
Column Totals 284 (76%) 91 (24%) 375 (100%)

Note: Percentages outside column and row totals sum to 100 within each task type

demand indices from both tasks and exercise-related measures. The exponentiated model of demand provided an excellent
In addition, the probability task generated significantly lower fit to average consumption data among systematic responses
rates of nonsystematic responding than the quantity task; how- in each task. We observed lower estimates of α in the proba-
ever, it was unclear if these findings are the result of an order bility purchase task, suggesting a decreased sensitivity to
effect or of an aspect of the purchase task type. To control for the changes in price. This discrepancy, which has been observed
possible influence of an order effect, a comparison was conduct- previously (Roma et al., 2016), may be a result of differences
ed involving only participants who took each task first; responses in k values or a fundamental difference in the nature of the
on the probability task remained significantly less likely to be question posed by each task. Participants may indicate 100%
nonsystematic. However, this analysis effectively reduced the probability of consumption at many prices far below the real-
sample size of the probability and quantity GMPT groups by world market price, creating longer inelastic sections of the
half. These data support Roma et al.’s (2016) conclusion that probability demand curves than the quantity demand curves.
both task types may be used to obtain systematic responses and For example, participants were asked to indicate how many
are therefore useful when measuring demand for big ticket items; visits to the gym they would make in 1 month across varying
however, there may be some items (e.g., commodities obtained prices per visit in the quantity task, or to indicate their proba-
infrequently, vulnerable to satiation, or are not commonly pur- bility of purchasing a single month of membership at a given
chased in discrete quantities) for which questions about con- price in the probability task. Stated another way, the probabil-
sumption are less easily interpreted in a quantity-based format. ity purchase task asks participants to indicate their probability
Significant Spearman correlations between demand indices to make a single decision that has a set maximum cost (buy
and exercise-related measures provide initial evidence of con- gym membership to have this month at the given price),
vergent validity for both versions of the GMPT. Beyond ex- whereas the quantity purchase task asks participants to indi-
pected correlations between mathematically related demand cate how frequently they would a series of decisions (how
indices within each task, demand indices from each task also many visits to the gym this month at the given price) without
correlated with self-reported exercise behavior (e.g., gym mi- a set maximum cost (i.e., the participant could decide to visit
nutes, gym visits), indicating that individuals who report the gym 5 times for $5 each visit, or 10 times for $5 each
higher frequency of visits and longer duration of visits place visit). Thus, for individuals who value gym membership, it
greater value on gym membership or the opportunity to visit a is unsurprising that higher probabilities of consumption
gym. When comparing correlations between demand indices persisted at prices below market value (reflected by the ex-
and validated measures of exercise addiction and physical tended inelastic portions of the probability demand curve) in
activity (i.e., the EAI and GPAQ), only the quantity task re- the probability task, and that quantity purchased for consump-
sulted in significant correlations for all indices (i.e., α, Q0). tion in the quantity task (i.e., visits each month) was gradually
However, this may be a result of decreased variability in reduced as prices for each visit escalated.
model-derived Q0 in the probability GMPT due to the nature Some information regarding the challenges in adapting in-
of the task (i.e., many individuals would agree to have access frequently purchased commodities to quantity tasks may be
to their preferred gym for free) and from constraining model- gleaned by examining the Stein et al. (2015) criteria violated
estimates of Q0 to a maximum of 100 during model fitting. by participants in each task. In the probability task, all 10
Finally, significant correlations in both tasks were observed nonsystematic responses violated trend or reversal from zero
between the price individuals currently pay monthly for gym criteria; in the quantity task, only 20 of the 91 nonsystematic
membership with α and Q0. More work is needed to establish responses violated these criteria. The remaining 71 nonsys-
the utility of either GMPT, but these results may be considered tematic responses were deemed nonsystematic due to unrea-
a first step towards demonstrating construct validity. sonably high levels of initial consumption (i.e., using the gym
 Psychol Rec

more than 62 times in 1 month when each visit is free). Indeed, It is unclear what properties of the probability GMPT ren-
of these 71 responses, 67 participants indicated that they dered the task more likely to generate systematic responding
would go to the gym 100 times in 1 month if each visit were than the quantity GMPT. Because the tasks differed not only
free. It appears that the order effect had a significant impact on by task type but also by price points, our data has limited
systematicity through the frequency at which participants ability to answer this question. Though price density (i.e., 17
responded with 100 at zero price on the quantity task. Only price points) were kept the same between tasks, different price
14 participants responded in this manner when taking the points were chosen to reflect realistic prices based on the how
quantity task first, whereas 53 participants did when taking the nature of consumption changed with each task type.
the quantity task after the probability task. Our data suggests Matching the quantity task prices and the probability task
that participants who completed the quantity task after com- prices could have resulted in limited exposure to prices at
pleting the probability task may have failed to attend to the which consumption is likely to be sensitive, potentially reduc-
relevant stimuli preceding each task (i.e., the written instruc- ing the resolution of the demand curve for both GMPTs.
tions that differentiated the probability and quantity GMPTs). Using the probability pricing with the quantity task would
It is possible that some participants thought the quantity likely cause rapid drop off in consumption, whereas using
GMPT was a continuation of the probability GMPT but with the quantity pricing with the probability task would likely
the commodity offered at different prices, and thus responded cause high probability of purchase without substantial de-
accordingly (i.e., indicating a probability of 100 to purchase creases across early prices. When examining gym member-
membership at the given price, despite the question asking, ship in the quantity of purchase format used in this study, the
“use a value between 0 and 100 to indicate the number of consumption cannot be viewed in the same way as in the
times you would pay to use the gym in the next 30 days for probability of purchase task. The probability task asked par-
each price”). These results highlight the importance of ensur- ticipants to indicate their probability of purchasing 1 month of
ing that surveys using multiple HPTs present stimuli (e.g., gym membership at the given price. The quantity task asked
instructions, quizzes testing understanding) that increase the participants to indicate how many visits they would make to
probability that participants will discriminate between tasks. the gym in 1 month at a given price; it did not ask participants
In addition, setting an a priori maximum feasible limit on to indicate how many 1-month memberships they would pur-
possible consumption values and a corresponding message chase for use in the next month, as framing the task in this
informing the participant that their response is invalid or un- manner would presumably limit consumption to one or zero.
reasonable may also serve as a discriminative stimulus to re- We believe this framework for presenting the commodity of
view the instructions before completing the task. gym membership in a quantity task was necessary for com-
Even after accounting for the order effect, participants parison purposes but highlights a weakness of the quantity
who took the quantity task first generated nonsystematic framework over the probability framework, at least as applied
responding at a significantly higher rate than participants in the current study’s context of gym membership and visits.
who took the probability task first (n = 25; 14.6% vs. n = 5; Though many gyms offer day or drop-in passes, the quantity-
2.87%, respectively). One possibility is that the Stein et al. based purchase framework is unlike how gym memberships
(2015) criteria were developed to identify nonsystematic are purchased in the real-world market. The differences in the
responding for quantity-based HPTs and are not sensitive nature of the questions about consumption, and the lack of
to the idiosyncrasies of probability tasks. For example, the familiarity with the quantity of purchase framework, may
nature of the probability task restricts maximum consump- have contributed to the higher rates of nonsystematic re-
tion to 100, which may lead to decreased variance across sponses in the quantity GMPT.
lower than real-market prices in which many people would There are several limitations to consider when
be likely to indicate a high probability of consumption. interpreting the results of this study. The mTurk qualifica-
This restricted variance could lead to reduced opportunities tions used for this study are less stringent than those used
for probability responses to violate the bounce criteria, ar- in previous work in operant demand analyses. In particular,
tificially decreasing the rates of systematic responding; the present study used an 85% HIT approval rate, whereas
however, of the 10 nonsystematic responses on the proba- previous studies often used 90% or above (Amlung et al.,
bility task and 20 nonsystematic responses on the quantity 2019; Kaplan et al., 2017; cf. Morris et al., 2017). It is
task, none were flagged due to bounce violations, suggest- unclear how this lower approval criterion may have affect-
ing that the probability task did not enjoy an advantage in ed data quality regarding demand. Although we observed
this criterion over the quantity task. In addition, standard approximately 7% of the 375 participants reported demo-
deviations of probability of consumption across low to graphic values that met our outlier criteria, the degree to
moderate real-world market prices (i.e., $0.01–$50) varied which demand were systematic within each outlier/
between 29.37% and 39.62%, suggesting that the potential nonoutlier sample were similar. Future work should exam-
effects of a ceiling were not realized. ine the gym membership and other probability-based tasks
Psychol Rec

using standardized approval ratings and in-person samples. Availability of Data and Materials The datasets generated during and
analyzed during the current study are available from the corresponding
In addition, the different prices used in each task were
author on reasonable request.
chosen to enhance resolution of consumption/likelihood,
these differences complicate between-task comparisons.
Declarations The present study was conducted by the first author in
Future research could use the same pricing structure, eas- partial fulfilment of the requirements for the master’s degree in psychol-
ing the ability of researchers to normalize demand curves ogy at the University of North Carolina-Wilmington. Portions of this
of both tasks and examine the degree to which curves study were presented at the 2018 Society for the Quantitative Analysis
of Behavior Annual Meeting in San Diego, California.
differ. However, as Roma et al. (2016) reported, the
probability-based purchase task tends to produce higher
Conflict of Interest On behalf of all authors, the corresponding author
values of α even when equating price between tasks. states that there is no conflict of interest. This is an observational study.
Thus, comparisons of demand indices between tasks in The University of North Carolina-Wilmington determined this study to be
the present study may have been of limited utility. exempt from further review under category (2): educational tests, surveys,
interviews, or observations.
The present study provides the first successful adapta-
tion of the HPT to gym membership. Although demand
indices from both tasks correlated with measures related
to physical activity and exercise, the responses on the
References
probability of purchase task were significantly less likely
American Psychiatric Association. (2013). Diagnostic and statistical
to generate nonsystematic responses, even after account- manual of mental disorders (5th ed.). https://doi.org/10.1176/appi.
ing for an order effect. These findings suggest that some books.9780890425596.
aspect of the probability task was more easily interpreted Amlung, M., Acker, J., Stojek, M., Murphy, J. G., & MacKillop, J.
(2012). Is talk “cheap?” An initial investigation of the equivalence
by the sample, and we speculated that this was a result of
of alcohol purchase task performance for hypothetical and actual
the probability task appearing in a similar way to the real- rewards. Alcoholism, Clinical & Experimental Research, 36(4),
world market commodity. Although more research is 716–724. https://doi.org/10.1111/j.1530-0277.2011.01656.x.
needed to establish the utility of the GMPT, the relations Amlung, M., Reed, D. D., Morris, V., Aston, E. R., Metrik, J., &
observed between demand indices generated by both tasks MacKillop, J. (2019). Price elasticity of illegal versus legal cannabis:
A behavioral economic substitutability analysis. Addiction, 114(1),
with theoretically relevant exercise related measures are 112–118. https://doi.org/10.1111/add.14437.
promising early demonstrations of convergent and con- Armstrong, T., & Bull, F. (2006). Development of the World Health
struct validity. Previous research has observed lower rates Organization Global Physical Activity Questionnaire (GPAQ).
of weight loss among children who demonstrate high Journal of Public Health, 14(2), 66–70.https://doi.org/10.1007/
s10389-006-0024-x.
levels of relative reinforcing value of food and steep Barkley, J. E., Epstein, L. H., & Roemmich, J. N. (2009). Reinforcing
discounting of future food rewards (Best et al., 2012). In value of interval and continuous physical activity in children.
addition, research utilizing cigarette and alcohol purchase Physiology & Behavior, 98(1–2), 31–36. https://doi.org/10.1016/j.
tasks has shown measures of reinforcing efficacy to be physbeh.2009.04.006.
Best, J. R., Theim, K. R., Gredysa, D. M., Stein, R. I., Welch, R. R.,
correlated with higher levels of consumption and future
Saelens, B. E., Perri, M. G., Schechtman, K. B., Epstein, L. H., &
problematic usage (González-Roz et al., 2019; Murphy Wilfley, D. E. (2012). Behavioral economic predictors of over-
et al., 2015). In a similar vein, it may be that individuals weight children’s weight loss. Journal of Consulting & Clinical
who show lower valuation as measured by the GMPT Psychology, 80(6), 1086–1096. https://doi.org/10.1037/a0029827.
Bureau of Labor Statistics. (2016). Sports and exercise among
may engage in more sedentary behavior and may be less
Americans. Economics Daily. Retrieved June 28, 2021, from
receptive to exercise interventions. Finally, increased https://www.bls.gov/opub/ted/2016/sports-and-exercise-among-
engagement in physical activity could increase the americans.htm
reinforcing value of exercise. Carr and Epstein (2020) Carr, K. A., & Epstein, L. H. (2020). Choice is relative: Reinforcing value
called for more research to develop programs that focus of food and activity in obesity treatment. American Psychologist,
75(2), 139–151. https://doi.org/10.1037/amp0000521.
on increasing the reinforcing value of physical activity, Centers for Disease Control & Prevention (CDC). (n.d.). Trends in meet-
and the GMPT—due to its scalability and accessibility ing the 2008 physical activity guidelines, 2008–2018. https://www.
as a brief self-report tool—may be uniquely positioned cdc.gov/physicalactivity/downloads/trends-in-the-prevalence-of-
to measure changes in value following such programs. physical-activity-508.pdf
Chu, A. H., Ng, S. H., Koh, D., & Muller-Riemenschneider, F. (2015).
Future research will need to evaluate the relative contri- Reliability and validity of the self and interviewer-administered ver-
butions of Q0 and α in relation to actual exercise behavior sions of the global physical activity questionnaire (GPAQ). PLoS
and/or interventions. For example, individuals who are ONE, 10(9), 1–18. https://doi.org/10.1371/journal.pone.0136944.
highly sensitive to increasing costs may be encouraged Cleland, C. L., Hunter, R. F., Kee, F., Cupples, M. E., Sallis, J. F., &
Tully, M. A. (2014). Validity of the global physical activity ques-
to engage in cheaper, more accessible exercise modalities. tionnaire (GPAQ) in assessing levels and change in moderate-
vigorous physical activity and sedentary behavior. BMC Public
Health, 14, 12–55.
 Psychol Rec

Epstein, L. H., Dearing, K. K., & Roba, L. G. (2010). A questionnaire reinforcement from a behavioral economic demand curve.
approach to measuring the relative reinforcing efficacy of snack Psychopharmacology, 203(1), 33–40.
foods. Eating Behaviors, 11(2), 67–73. https://doi.org/10.1016/j. Madden, G. J., & Kalman, D. (2010). Effects of bupropion on simulated
eatbeh.2009.09.006. demand for cigarettes and the subjective effects of smoking.
Few, L. R., Acker, J., Murphy, C., & MacKillop, J. (2012). Temporal Nicotine & Tobacco Research, 12(4), 416–422. https://doi.org/10.
stability of a cigarette purchase task. Nicotine & Tobacco Research, 1093/ntr/ntq018.
14(6), 761–765. https://doi.org/10.1093/ntr/ntr222. Morris, V., Amlung, M., Kaplan, B. A., Reed, D. D., Petker, T., &
Flack, K. D., Johnson, L., & Roemmich, J. N. (2017). The reinforcing MacKillop, J. (2017). Using crowdsourcing to examine behavioral
value and liking of resistance training and aerobic exercise as pre- economic measures of alcohol value and proportionate alcohol re-
dictors of adult’s physical activity. Physiology & Behavior, inforcement. Experimental & Clinical Psychopharmacology, 25(4),
179(2017), 284–289. https://doi.org/10.1016/j.physbeh.2017.06. 314–321. https://doi.org/10.1037/pha0000130.
016. Murphy, J. G., & MacKillop, J. (2006). Relative reinforcing efficacy of
Gilroy, S. P., Kaplan, B. A., & Reed, D. D. (2020). Interpretation(s) of alcohol among college student drinkers. Experimental & Clinical
elasticity in operant demand. Journal of the Experimental Analysis Psychopharmacology, 14, 219–227. https://doi.org/10.1037/1064-
of Behavior, 114(1), 106–115. https://doi.org/10.1002/jeab.610. 1297.14.2.219.
González-Roz, A., Jackson, J., Murphy, C., Rohsenow, D. J., & Murphy, J. G., Correia, C. J., Colby, S. M., & Vuchinich, R. E. (2005).
MacKillop, J. (2019). Behavioral economic tobacco demand in re- Using behavioral theories of choice to predict drinking outcomes
lation to cigarette consumption and nicotine dependence: A meta- following a brief intervention. Experimental & Clinical
analysis of cross-sectional relationships. Addiction, 114(11), 1926– Psychopharmacology, 13, 93–101. https://doi.org/10.1037/1064-
1940. https://doi.org/10.1111/add.14736. 1297.13.2.93.
Grace, R. C., Kivell, B. M., & Laugesen, M. (2015). Predicting decreases Murphy, J. G., Dennhardt, A. A., Martens, M. P., Yurasek, A. M.,
in smoking with a cigarette purchase task: Evidence from an excise Skidmore, J. R., MacKillop, J., & McDevitt-Murphy, M. E.
tax rise in New Zealand. Tobacco Control, 24(6), 582–587. (2015). Behavioral economic predictors of brief alcohol intervention
Griffiths, M. (1996). Behavioral addiction: An issue for everybody? outcomes. Journal of Consulting & Clinical Psychology, 83(6),
Employee Counselling Today: The Journal of Workplace 1033–1043. https://doi.org/10.1037/ccp0000032.
Learning, 8, 19–25. Pandey, A., Salahuddin, U., Garg, S., Ayers, C., Kulinski, J., Anand, V.,
Hoos, T., Espinoza, N., Marshall, S., & Arredondo, E. M. (2012). Mayo, H., Kumbhani, D. J., de Lemos, J., & Berry, J. D. (2016).
Validity of the global physical activity questionnaire (GPAQ) in Continuous dose-response association between sedentary time and
adult Latinas. Journal of Physical Activity & Health, 9, 698–705. risk for cardiovascular disease: A meta-analysis. JAMA Cardiology,
Hursh, S. R., & Silberberg, A. (2008). Economic demand and essential 1(5), 575. https://doi.org/10.1001/jamacardio.2016.1567.
value. Psychological Review, 115, 186–198. https://doi.org/10. Piercy, K. L., Troiano, R. P., Ballard, R. M., Carlson, S. A., Fulton, J. E.,
1037/0033-295X.115.1.186. Galuska, D. A., George, S. M., & Olson, R. D. (2018). The physical
Jacobs, E. A., & Bickel, W. K. (1999). Modeling drug consumption in the activity guidelines for Americans. JAMA, 320(19), 2020. https://doi.
clinic using simulation procedures: Demand for heroin and ciga- org/10.1001/jama.2018.14854.
rettes in opioid-dependent outpatients. Experimental & Clinical Pope Jr., H. G., Kean, J., Nash, A., Kanayama, G., Samuel, D. B., Bickel,
Psychopharmacology, 7, 412–426. W. K., & Hudson, J. I. (2010). A diagnostic interview module for
Kaplan, B. A., Reed, D. D., Murphy, J. G., Henley, A. J., Reed, F. D. D., anabolic-androgenic steroid dependence: Preliminary evidence of
Roma, P. G., & Hursh, S. R. (2017). Time constraints in the alcohol reliability and validity. Experimental & Clinical
purchase task. Experimental & Clinical Psychopharmacology, Psychopharmacology, 18(3), 203–213. https://doi.org/10.1037/
25(3), 186–197. https://doi.org/10.1037/pha0000110. a0019370.
Kaplan, B. A., Foster, R. N. S., Reed, D. D., Amlung, M., Murphy, J. G., R Core Team. (2019). R: A language and environment for statistical
& MacKillop, J. (2018). Understanding alcohol motivation using the computing. R Foundation for Statistical Computing. https://www.
alcohol purchase task: A methodological systematic review. Drug & R-project.org/.
Alcohol Dependence, 191, 117–140. https://doi.org/10.1016/j. Reed, D. D., Kaplan, B. A., Becirevic, A., Roma, P. G., & Hursh, S. R.
drugalcdep.2018.06.029. (2016). Toward quantifying the abuse liability of ultraviolet tanning:
Kaplan, B. A., Gilroy, S. P., Reed, D. D., Koffarnus, M. N., & Hursh, S. A behavioral economic approach to tanning addiction. Journal of
R. (2019). The R package beezdemand: Behavioral economic easy the Experimental Analysis of Behavior, 106, 93–106. https://doi.org/
demand. Perspectives on Behavior Science, 42(1), 163–180. https:// 10.1002/jeab.216.
doi.org/10.1007/s40614-018-00187-7. Roma, P. G., Hursh, S. R., & Hudja, S. (2016). Hypothetical purchase
Kaplan, B. A., Franck, C. T., McKee, K., Gilroy, S. P., & Koffarnus, M. task questionnaires for behavioral economic assessments of value
N. (in press). Applying mixed-effects modeling to behavioral eco- and motivation. Managerial & Decision Economics, 37, 306–323.
nomic demand: An introduction. Perspectives on Behavior Science. https://doi.org/10.1002/mde.2718.
https://doi.org/10.1007/s40614-021-00299-7. Roma, P. G., Reed, D. D., DiGennaro Reed, F. D., & Hursh, S. R. (2017).
Koffarnus, M. N., Franck, C. T., Stein, J. S., & Bickel, W. K. (2015). A Progress of and prospects for hypothetical purchase task question-
modified exponential behavioral economic demand model to better naires in consumer behavior analysis and public policy. The
describe consum ption data. Experimental & Clinical Behavior Analyst, 40(2), 329–342. https://doi.org/10.1007/s40614-
Psychopharmacology, 23(6), 504–512. https://doi.org/10.1037/ 017-0100-2.
pha0000045. Snider, S. E., LaConte, S. M., & Bickel, W. K. (2016). Episodic future
MacKillop, J., Murphy, J. G., Ray, L. A., Eisenberg, D. T. A., Lisman, S. thinking: Expansion of the temporal window in individuals with
A., Lum, J. K., & Wilson, D. S. (2008). Further validation of a alcohol dependence. Alcoholism: Clinical & Experimental
cigarette purchase task for assessing the relative reinforcing efficacy Research, 40(7), 1558–1566. https://doi.org/10.1111/acer.13112.
of nicotine in college smokers. Experimental and Clinical Stein, J. S., Koffarnus, M. N., Snider, S. E., Quisenberry, A. J., & Bickel,
Psychopharmacology, 16, 57–65. https://doi.org/10.1037/1064- W. K. (2015). Identification and management of nonsystematic pur-
1297.16.1.57. chase task data: Toward best practice. Experimental & Clinical
MacKillop, J., Murphy, J. G., Tidey, J. W., Kahler, C. W., Ray, L. A., & Psychopharmacology, 23, 377–386. https://doi.org/10.1037/
Bickel, W. K. (2009). Latent structure of facets of alcohol pha0000020.
Psychol Rec

Stein, J. S., Tegge, A. N., Turner, J. K., & Bickel, W. K. (2018). Episodic Theory, 12, 489–499. https://doi.org/10.1080/
future thinking reduces delay discounting and cigarette demand: An 16066350310001637363.
investigation of the good-subject effect. Journal of Behavioral Warren, T. Y., Barry, V., Hooker, S. P., Sui, X., Church, T. S., & Blair, S.
Medicine, 41(2), 269–276. https://doi.org/10.1007/s10865-017- N. (2010). Sedentary behaviors increase risk of cardiovascular dis-
9908-1. ease mortality in men. Medicine & Science in Sports & Exercise,
Strickland, J. C., & Stoops, W. W. (2019). The use of crowdsourcing in 42(5), 879–885.
addiction science research: Amazon Mechanical Turk. Experimental Wilson, A. G., Franck, C. T., Koffarnus, M. N., & Bickel, W. K. (2016).
& Clinical Psychopharmacology, 27(1), 1–18. https://doi.org/10. Behavioral economics of cigarette purchase tasks: Within-subject
1037/pha0000235. comparison of real, potentially real, and hypothetical cigarettes.
Strickland, J. C., Campbell, E. M., Lile, J. A., & Stoops, W. W. (2020). Nicotine & Tobacco Research, 18(5), 524–530. https://doi.org/10.
Utilizing the commodity purchase task to evaluate behavioral eco- 1093/ntr/ntv154.
nomic demand for illicit substances: A review and meta-analysis.
Addiction, 115(3), 393–406. https://doi.org/10.1111/add.14792.
Publisher’s Note Springer Nature remains neutral with regard to jurisdic-
Terry, A., Szabo, A., & Griffiths, M. (2004). The exercise addiction
tional claims in published maps and institutional affiliations.
inventory: A new brief screening tool. Addiction Research &