The Psychological Record (2019) 69:415–424

https://doi.org/10.1007/s40732-019-00340-x

ORIGINAL ARTICLE

On Four Types of Devaluation of Outcomes Due to Their Costs:

Delay, Probability, Effort, and Social Discounting
Wojciech Białaszek 1 & Paweł Ostaszewski 1 & Leonard Green 2 & Joel Myerson 2

Published online: 22 March 2019

# The Author(s) 2019

Abstract
Discounting refers to decreases in the subjective value of an outcome with increases in some attribute of that outcome. The
attributes most commonly studied are delay and probability, with far less research on effort and social discounting. Although
these attributes all represent costs that reduce subjective value, it is as yet unclear how the extent to which they do so is related at
the individual level. Accordingly, the present study examined the degree to which individual participants discounted hypothetical
monetary rewards on each of four discounting tasks in which the delay, probability, effort, and number of people with whom the
money was to be shared were manipulated. At the group level, larger amounts were discounted less steeply than smaller amounts
when delay and effort were varied, whereas larger amounts were discounted more steeply when probability and number of people
were varied. At the individual level, the correlational pattern was examined using exploratory factor analysis. A six-factor
structure (with separate factors for delay and effort, and two factors each for social and probability discounting) described the
relations among indifference points. At a more molar level, a two-factor structure, which corresponded to the direction of the
observed magnitude effects, described the relations among area-under-the-curve measures of discounting in the eight conditions
resulting from crossing two monetary amounts with the four cost factors. We conclude that despite sharing some similarities,
individual and group differences in discounting involving the different types of costs reflect mostly separate processes and traits.

Keywords Delay discounting . Probability discounting . Effort discounting . Social discounting . Magnitude effect . Factor
analysis

Consider a situation in which you could receive something & Raineri, 1992) noted that outcomes lose value (i.e., their value
you want after a delay but its delivery is not certain. is discounted) when they are delayed, their receipt is not certain,
Furthermore, you are required to engage in some effort to or if they are to be shared with others. Later studies added a
obtain the outcome, and once it is received, you have to share fourth cost factor to discounting, namely the effort needed to
it with others. Such a situation occurs when we have to wait to obtain a reward (Białaszek, Marcowski, & Ostaszewski, 2017;
receive money; its receipt is not always certain, sometimes we Mitchell, 1999, 2003; Ostaszewski, Bąbel, & Swebodziński,
have to exert some effort to receive it, and we often have to 2013; Sugiwaka & Okouchi, 2004). A question of both theoret-
share it with other people. Each of these four cost factors, ical and practical importance is whether these cost factors engage
delay, risk, effort, and a social component of sharing with the same or different underlying processes and traits (Myerson,
others, decreases the subjective value of the outcome. Green, Hanson, Holt, & Estle, 2003). To answer this question,
Such decreases in subjective value have been studied in the the present study takes two approaches. The first approach ex-
context of discounting (Rachlin, 2006). Rachlin (1993; Rachlin amines the effects of experimental manipulations based on the
assumption that if a specific manipulation affects two dependent
variables in the same way, then the same processes may be
* Wojciech Białaszek involved in both cases, but that if that manipulation affects the
wbialaszek@swps.edu.pl two dependent variables differently, then different processes may
be involved. The second approach examines individual differ-
1
SWPS University of Social Sciences and Humanities, Chodakowska ences and the intercorrelations among the different discounting
19/31, 03-815 Warszawa, Poland types, using factor analysis to determine the number of
2
Washington University, St. Louis, MO, USA discounting factors.
416 Psychol Rec (2019) 69:415–424

Discounting can be measured using indifference points to 2013; Kirby, 1997; Thaler, 1981). In this example, the cost
estimate the subjective value of a reward. That is, one can factor, delay, reduces the relative subjective value of a larger
estimate the amount of reward that a person judges approxi- reward less than it reduces the relative subjective value of a
mately equal in value to that of another (usually larger) reward smaller reward. The magnitude effect has been observed not
amount, the acquisition of which depends on some cost factor. only with respect to delayed monetary payoffs but also with
In the case of delay discounting, for example, one may deter- other delayed outcomes including health (Chapman, 1996)
mine the smallest amount of immediate reward exchangeable and consumable rewards as diverse as candy and vacations
for a larger, delayed reward. If the delayed outcome is, say, PLN (Estle, Green, Myerson & Holt, 2007; Raineri & Rachlin,
150 (i.e., 150 new Polish złotys) in one month and the imme- 1993). Much less is known about the magnitude effect in
diate one is also PLN 150, the typical choice will be the imme- effort discounting, but it appears to be similar to that in delay
diate reward. If the immediate amount is systematically de- discounting. For example, Ostaszewski et al. (2013; see also
creased, however, at some point the participant may prefer the Białaszek et al., 2017) found that larger reward amounts were
delayed alternative. It is important to note that this point may be discounted less steeply than smaller amounts on tasks in
different for every participant: some may switch their choice to which the amount of physical effort was varied as well as on
the delayed alternative when the immediate amount is only tasks that varied in the amount of cognitive effort required.
slightly smaller (e.g., PLN 145), indicating shallow A reverse magnitude effect, in which larger amounts of re-
discounting, whereas others may continue to choose the imme- ward are discounted more steeply than small amounts, is consis-
diate reward until it is decreased below PLN 50, indicating they tently observed in probability discounting (Green, Myerson, &
discounted the larger amount steeply because it was delayed Ostaszewski, 1999; Myerson, Green, & Morris, 2011). A reverse
and they chose not to wait, even for the additional PLN 100. magnitude effect also is observed with social discounting: larger
Steep discounting of delayed rewards is associated with an rewards to be shared with others lose a greater proportion of their
array of maladaptive behaviors, and with addiction in partic- value as the number of people with whom the reward is to be
ular. A meta-analysis by MacKillop et al. (2011; see also shared increases (Ostaszewski & Osiński, 2011) or as their social
Amlung, Vedelago, Acker, Balodis, & MacKillop, 2017) distance increases (e.g., from friends to strangers; Rachlin &
found that individuals with an addiction discounted delayed Jones, 2008a). Thus, the effects of the reward amount appear to
rewards more steeply than control groups. Discounting of divide these types of discounting into two categories, depending
probabilistic rewards, however, is less predictive. Some stud- on whether larger amounts are discounted more or less steeply
ies report that substance use and abuse is related to greater than small amounts.
probability discounting (Reynolds, Richards, Horn, &
Karraker, 2004; Yi, Carter, & Landes, 2012), whereas others
have failed to find such a relationship (Reynolds, Karraker,
Horn, & Richards, 2003; Takahashi, Ohmura, Oono, & Correlation Analyses and Factor Analytic
Radford, 2009). Very few studies have examined the relations Approach
between addiction and effort or social discounting, although a
study of methamphetamine users found that they showed The majority of studies investigating the relations among dif-
steeper social discounting, which involves sharing rewards, ferent types of discounting have focused on delayed and prob-
than controls (Yi et al., 2012). Further research is needed, abilistic rewards, perhaps initially because time and risk are
and such research should include not just studies examining dimensions that were strongly linked theoretically. If one di-
the different types of discounting separately, but also studies mension or one trait (e.g., impulsivity) underlies the other,
that examine more than one type of discounting in order to then delay and probability discounting should be strongly cor-
determine the relations among the effects of different cost related. On the contrary, researchers have typically found ei-
factors. We believe that assessing the extent to which an indi- ther no or weak correlations between rates of delay and prob-
vidual’s tendency to discount delayed rewards predicts their ability discounting (Holt, Green, & Myerson, 2003; Mitchell,
tendency to discount effortful rewards, for example, is essen- 1999; Myerson et al., 2003; Ohmura, Takahashi, Kitamura, &
tial if we want to explain the traits and processes underlying Wehr, 2006; Shead & Hodgins, 2009). The literature on the
problematic behaviors. relation between other types of discounting is relatively
sparse, although a bit more is known about social discounting
than about effort discounting (cf. Mitchell, 1999). Rachlin
Amount-Dependent Discounting (2006) noted parallels between delay and social discounting,
and indeed, some studies report a positive correlation between
The best-known example of a magnitude effect is the finding them (Rachlin & Jones, 2008a) as well as correlations among
that larger delayed reward amounts are discounted less steeply delay, social, and probability discounting (Jones & Rachlin,
than smaller ones (e.g., Green, Myerson, Oliveira, & Chang, 2009), although these were rather modest.
Psychol Rec (2019) 69:415–424 417

It is perhaps surprising that only a few studies have exam- Participants were provided a 24-page response booklet that
ined the factor structure of different discounting tasks, and was used to determine 24 indifference points (4 discounting
those concerned delay and/or probability discounting. Green types, 2 reward amounts, and 3 cost values), one per page. A
and Myerson (2013) showed that delay discounting and prob- heading at the top of each page specified the nature of the
ability discounting of monetary payoffs and consumable choice alternatives (e.g., “gain immediately or in 6 months”).
goods loaded on two separate factors linked not to the nature Below the heading were two columns. In the left column,
of the rewards but to the type of discounting. In contrast, other amounts corresponding to potential indifference points were
research suggests that individual differences in probability presented in descending order, beginning with the actual
discounting (Terrell, Derenne, & Weatherly, 2014) and delay amount of the to-be-discounted reward and ending with
discounting (Weatherly, Terrell, & Derenne, 2010) may be PLN 0 in 30 decrements. Sample headings for the three other
outcome-specific. Although factor analysis has been relatively discounting tasks are “gain for sure or with a 45% chance,
neglected to date in studies of discounting, it is commonly “gain to be kept for yourself or to be shared with five
used to study individual differences in personality and seems strangers,” and “gain without an effort or after climbing to
particularly well suited to address the issue of how many dif- the 11th floor.” Note that, following Rachlin (1993) and
ferent traits are involved in the various types of discounting. Rachlin and Raineri (1992), the social discounting cost was
defined as the number of people with whom a reward was to
be shared.
Aims and Scope of the Present Study All reward amounts and cost values across the 24 experi-
mental conditions are presented in Table 1. The four types of
There are multiple parallels among at least some different discounting tasks were presented in a counterbalanced order,
types of discounting. There are consistencies between magni- and within each task, the two amount conditions (small or
tude effects both between delay and effort discounting (where large) of the to-be-discounted reward were counterbalanced.
discounting decreases with amount), and between probability For each amount in each discounting task, three indifference
and social discounting (where the opposite is true). The lack of points were estimated, one for each cost value. The cost values
direct comparisons among the same participants engaged in always were presented in an ascending order (e.g., 1 month,
all four types of discounting, however, led us to conduct the then 6 months, and then 2 years). Different reward amounts
present study, the main purpose of which was threefold: (1) to were used in each discounting task to make the experimental
determine how amount of reward affects delay, probability, session less monotonous for participants and to reduce the
effort, and social discounting; (2) to investigate correlations likelihood of repetitive responses.
among the four types of discounting; and (3) to explore the
underlying factor structure of discounting and its implications Data Analysis
for individual differences in choice and decision making.
To answer the question of whether the four discounting types
reflect the same underlying mechanisms and traits or involve
Method separate mechanisms and traits, we used a three-step ap-
proach. First, we examined whether the different types of
Participants discounting are similarly affected by the amount of reward.
Second, we analyzed the intercorrelations among the four
We recruited 160 participants from the local community in discounting types. Third, we performed exploratory factor
accordance with university ethics committee regulations. analyses in order to determine whether the associations among
Participants (82 males and 78 females) ranged in age from the various discounting conditions can be described by a set of
21 to 60 years (M = 35.74; SD = 10.45). more fundamental variables (factors). Two factor analyses
were performed, one on the indifference points (expressed as
Procedure
Table 1. Amounts of reward and costs for each discounting task
After signing an informed consent form, participants pro-
Discounting Small Large Small Medium Large
vided basic demographic information and then proceeded
task reward reward cost cost cost
to the discounting tasks. The discounting tasks used a
fixed-choice procedure (Rachlin, Raineri, & Cross, 1991) delay PLN 200 PLN 40,000 1 month 6 months 2 years
to estimate indifference points, where an indifference point probability PLN 150 PLN 30,000 98% 45% 3%
is the amount of money available without any cost that is effort PLN 100 PLN 20,000 3rd floor 11th floor 40th floor
subjectively equal in value to the undiscounted amount social PLN 50 PLN 10,000 2 people 5 people 12 people
accompanied by a specified cost.
418 Psychol Rec (2019) 69:415–424

proportions of the to-be-discounted amount) and the other on smaller rewards were discounted less steeply. Following
the areas under the curve (AuCs; Myerson, Green, & Cohen (1988), these effect sizes can be interpreted as large
Warusawitharana, 2001). Indifference points were estimated for delay and effort discounting, medium for probability
as the last amount of immediate, certain, effortless, or un- discounting, and small for social discounting.
shared reward that a participant chose in each condition.

Correlations among Discounting Types

Results Moderate positive correlations (.54 to .59; all ps < .001) were
observed between the AuCs for different amount conditions
Amount-Dependent Discounting within the same discounting type (e.g., probability PLN 150
and PLN 30,000: r = .55). All other correlations between
To determine whether amount of reward has the same effect different discounting conditions were lower, implying greater
on different types of discounting, we performed t-tests for similarity within types of discounting than between types.
dependent samples on the AuC values for the small and large As may be seen in Table 2, half of the correlations among
reward conditions (shown in Figure 1). These tests revealed the AuCs for different discounting conditions were signifi-
significant magnitude effects for delay (t(159) = 12.11; p < cant, and half were not. It should be noted that delay and effort
.001; d = 0.96) and effort discounting (t(159) = 12.94; p < discounting were significantly correlated at all four combina-
.001; d = 1.02) in which smaller rewards were discounted tions of reward amounts (small–small, large–large, and the
more steeply, and significant reverse magnitude effects for two small–large combinations), and for probability and social
probability (t(159) = 7.64; p < .001; d = 0.60) and social discounting, three of the correlations among amount condi-
discounting (t(159) = 4.56; p < .001; d = 0.36) in which tions were significant, suggesting that two separate factors

Fig. 1 Mean Areas under the Curve (AuC) for small and large reward present for probability and social discounting. Error bars represent
amounts for each of the four types of discounting. A magnitude effect is ±1standard error of the mean
present for delay and effort discounting, and a reverse magnitude effect is
Psychol Rec (2019) 69:415–424 419

Table 2. Intercorrelations among AuC measures

delay delay probability PLN 150 probability effort effort social social
PLN 200 PLN 40,000 PLN 30,000 PLN 100 PLN 20,000 PLN 50 PLN 10,000

delay 1
PLN 200
delay .59** 1
PLN 40,000
probability 0.05 0.04 1
PLN 150
probability -0.03 -0.09 .55** 1
PLN 30,000
effort .19* .38** -0.01 -0.02 1
PLN 100
effort .17* .35** -.23** -.21** .55** 1
PLN 20,000
social .20* 0.08 .31** 0.13 -.16* -0.04 1
PLN 50
social .23** 0.10 .25** .24** -0.12 -0.02 .54** 1
PLN 10,000

Note: Correlations between the two reward amounts of the same discounting task appear in bold
*p < .05; **p < .001

may underlie individual differences in discounting at this level was .692, exceeding the threshold of .5, and Bartlett’s test of
of analysis. In contrast, delay and probability discounting con- sphericity was significant (Χ2(276) = 2148.70; p < .001).
ditions were never significantly correlated. The other combi- Overall, the six factors seen in Table 3 explained 58.33% of
nations of conditions from different types of discounting (i.e., the variance. All factor loadings above .32 are bolded in the
delay and social, effort and social, effort and probability) pro- pattern matrix, following Tabachnick and Fidell’s (2007) rec-
duced intermediate results (one or two significant ommendation that only loadings higher than that should be
correlations). interpreted. The first factor explained 18.56% of the variance,
and the following five factors explained 16.97%, 7.91%,
Factor Analyses 6.34%, 4.71%, and 3.83%, respectively. The first factor is
composed of all six conditions of the delay discounting task
Factor analyses were conducted at two levels, one based on (i.e., all three delays crossed with the small and large
the eight AuCs just discussed, and the other based on the 24 amounts). The second and sixth factors correspond to social
indifference points corresponding to the different cost factor discounting of small and large amounts, respectively. The
conditions. To determine the number of factors in exploratory third and fifth factors represent different probability
factor analysis of the indifference points, we relied on discounting conditions, with small and large amount condi-
Velicer’s (1976) minimum average partial method (MAP), tions loading on each factor, and the fourth factor is composed
which is based on mean partial correlations between variables of all six conditions of the effort discounting task. These anal-
(O’Connor, 2000). This method identified a six-factor solution yses show high consistency in the loading of different types of
as best; the smallest mean square of partial correlation was discounting on separate factors as well as clear discriminabil-
achieved in the sixth step and equaled 3.472*10-2. Similar ity between factors. It is interesting that delay and effort
conclusions follow from Cattell’s (1966) scree plot analysis, discounting each formed a factor that included both reward
which suggested a six- to eight-factor solution, and Kaiser’s amounts, whereas different amount conditions of social
(1974) criterion (eigenvalue greater than 1.0), which resulted discounting and probability discounting loaded on separate
in a seven-factor solution. factors (see Table 3).
The factor analysis of indifference points was conducted In addition, an exploratory factor analysis was performed
using the principal axis extraction method with Oblimin rota- on the AuC measures in order to explore the structure of
tion and an initial value of delta = 0, not imposing any artificial discounting factors at this higher level. This analysis, howev-
ad-hoc correlations between factors, as recommended by er, may have limitations. It is important to note that the MAP
Fabrigar, Wegener, MacCallum, and Strahan (1999). test (Velicer, 1976) did not provide a coherent solution (i.e., it
Oblique rotation takes into account that the factors do not have could not reliably identify a clear factor structure for the AuC
to be orthogonal and, in theory, could be correlated. The measures), which might be because there was more unsystem-
Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy atic than systematic variance (O’Connor, 2000). Therefore,
420 Psychol Rec (2019) 69:415–424

Table 3. Factor structure of the

indifference points obtained from Factor Condition Factor
the four discounting tasks. Factors (reward amount and cost)
are presented in descending order 1 2 3 4 5 6
of the variance accounted for.
Within factors, factor loadings 1. Delay PLN 200, 6 months .815 -.021 -.046 -.123 -.005 .142
also are presented in descending PLN 40.000, 2 years .784 -.009 .077 .084 .076 -.046
order
PLN 200, 2 years .739 -.044 .025 -.074 -.036 .164
PLN 200, 1 month .736 .042 -.129 -.028 .056 .035
PLN 40.000, 6 months .727 .023 .008 .207 -.002 -.106
PLN 40.000, 1 month .568 .034 .057 .179 -.093 -.137
2. Social-A PLN 50, 5 people .027 .887 .005 .013 -.027 .137
PLN 50, 12 people .017 .817 .070 .003 .030 .122
PLN 50, 2 people -.033 .680 .029 -.009 -.016 .090
3. Probability-A PLN 150, 3% .029 .114 .886 -.033 -.123 .003
PLN 30.000, 3% -.084 -.117 .646 -.012 .072 .152
PLN 150, 45% .012 .252 .584 -.048 .231 -.067
4. Effort PLN 20.00, 40th floor -.022 -.042 -.137 .801 -.052 .185
PLN 100, 11th floor .044 -.098 .154 .760 -.052 -.034
PLN 20.000, 11th floor -.025 .101 -.227 .708 .004 .024
PLN 100, 40th floor .080 -.221 .218 .591 .044 .030
PLN 100, 3rd floor -.012 .133 .008 .545 .038 -.085
PLN 20.000, 3rd floor .177 .002 -.091 .390 .034 -.086
5. Probability-B PLN 30.000, 98% -.080 -.074 -.160 .056 .866 .076
PLN 30.000, 45% -.074 -.051 .250 -.036 .597 .175
PLN 150, 98% .154 .105 .058 -.023 .564 -.165
6. Social-B PLN 10.000, 5 people .098 .152 .041 -.003 -.011 .842
PLN 10.000, 12 people .054 .139 .009 .068 -.005 .764
PLN 10.000, 2 people .063 .145 .077 -.024 .071 .442

Note: Factor loadings over .32 appear in bold; letters A and B refer to factors that do not include all six reward and
cost conditions of a discounting task

we based our decision on how many factors to retain on the discounting task. The second factor was composed of the
simplest criterion, the change in the eigenvalue with each AuCs for both amount conditions of the probability and social
additional factor, as represented in Cattel’s scree plot. After discounting tasks (see Table 4). There were no cross loadings
the first two eigenvalues (2.201 and 2.104), the third was between the two factors. It is interesting that the first factor
1.187, and the fourth was 0.873. Given the large reduction involves types of discounting that showed strong magnitude
in eigenvalues and the correlational results described previ- effects, and the second factor involves types of discounting
ously, we decided to retain a two-factor solution. that showed reverse magnitude effects, although these were
Again, the basic assumptions were met (the KMO equaled somewhat weaker. The general finding of both factor analyses
0.595), and Bartlett’s test of sphericity was significant (Χ2(28) is that the different types of discounting are not reducible to a
= 323.655; p < .001). The average communality after extrac- single process or trait, consistent with the different effects of
tion was .390 (with a minimum of .242 and maximum of reward amount on discounting and the patterns of correlations
.610). Despite the low communality for the large reward con- among the variables.
dition of probability discounting (i.e., .242), we decided to
include this data in further analyses because of their explor-
atory nature and the theoretical importance of decisions in- Discussion
volving large, probabilistic rewards.
The two-factor solution for the AuC data explained 39.02% The goal of the present study was to examine similarities and
of the variance (20.23% by the first factor) and provided an differences among four different types of discounting (i.e.,
interpretable solution. The first factor was composed of the delay, probability, effort, and social) defined by different cost
AuCs for both amount conditions of the delay discounting factors that can each result in devaluation of rewards. It is
task and the AuCs for both amount conditions of the effort important to note that the findings indicate that more than
Psychol Rec (2019) 69:415–424 421

one mechanism underlies the discounting of large and small respectively). The other two tasks showed less consistency, with
rewards, and more than one trait underlies individual differ- the indifference points from each task loading on two separate
ences on the various types of discounting. factors (see Table 3). Although one is tempted to interpret the fact
With respect to the underlying mechanism(s), we would note that the small amount indifference points from the social
that the amount of reward had different effects depending on the discounting task loaded on one factor and the points from the
type of task. Consistent with previous findings, small delayed large amount conditions loaded on another, the fact that the in-
rewards were discounted more steeply than large ones, whereas difference points from the probability discounting task loaded on
large probabilistic rewards were discounted more steeply. The separate factors without regard for reward amount suggests that
effects of amount on effort discounting paralleled those with further research with larger samples and different procedures for
delay discounting, and the effects of amount on social estimating indifference points may be warranted before engaging
discounting, although weaker, paralleled those with probability in too much speculation.
discounting. It is interesting that analysis of the AuCs revealed As noted, exploratory factor analyses of the AuCs, which are
that tasks that showed similar amount effects loaded on the same effectively weighted averages of the indifference points from
factors (i.e., delay and effort on Factor 1, and probability and each amount condition of each task, thereby potentially reducing
social on Factor 2; see Table 4). The present finding of a corre- measurement error, revealed a much simpler structure at the mo-
spondence between type of discounting and factor loading has lar level, although, again, the structure is inconsistent with a
not been reported previously, but it fits neatly with the previous single underlying trait or mechanism. And again, further research
finding that whereas delay and probability discounting load on with larger samples and different estimation procedures would be
separate factors, the discounting of losses, which does not show desirable, as would the use of confirmatory factor analysis in
amount effects, loaded on yet a third factor (Mejía-Cruz, Green, future studies, particularly because of the implications of the
Myerson, Morales-Chainé, & Nieto, 2016). Taken together, these present findings for treatment interventions. Although individual
results suggest that amount effects may reflect something more differences in discounting rates are related to various maladaptive
fundamental about the different types of discounting than just behaviors (Madden & Bickel, 2010; MacKillop et al., 2011), the
their susceptibility to the effect of amount. existence of multiple factors implies that an intervention that
Exploratory factor analyses were conducted at both more mo- successfully targets one type of discounting does not guarantee
lecular and more molar levels (i.e., indifference points and AuCs, a similar change in other types of discounting. Nevertheless, the
respectively). Although different factors were observed in these existence of two factors at the more molar AuC level of analysis
analyses, in both cases the results are clearly inconsistent with a is encouraging with respect to transfer at least between delay and
single factor (e.g., impulsivity) underlying all types of effort discounting on the one hand, and between probability and
discounting. In part, the results of the analysis of indifference social discounting on the other.
points may be thought of as assessing the internal consistency One concern, however, is that compared to the delay and
of the four tasks, and the results revealed that two of the probability discounting tasks, the social discounting and effort
discounting tasks, delay and effort, showed higher consistency discounting tasks used here may not be as representative, which
than the others, with all six indifference points from the delay might affect transfer. For example, one can exert cognitive or
discounting task loading on one factor, and all six points from the physical effort (Białaszek et al., 2017; Ostaszewski et al., 2013),
effort discounting task loading on another (i.e., Factors 1 and 4, or even both kinds of effort at once. Likewise, in social
discounting, when the reward is shared with other people, these
others can be relatives, friends, or strangers, and their social dis-
Table 4. Factor structure of AuC measures obtained from the four tance as well as their number affects the degree of discounting
discounting tasks. Factor 1 corresponds to delay and effort discounting,
and Factor 2 corresponds to social and probability discounting. Within (Rachlin & Jones, 2008b; Osiński, Ostaszewski & Karbowski,
factors, factor loadings are presented in descending order 2014). Therefore, research is needed to examine the degree to
which the present findings transfer to other instances of these
Factor 1 Factor 2
types of discounting. At the same time, other aspects of the pres-
AuCdelay PLN 40,000 .776 .140 ent tasks may facilitate transfer. For example, effort discounting
AuCdelay PLN 200 .586 .278 often involves delay discounting, as in the scenario studied in the
AuCeffort PLN 20,000 .577 -.237 present experiment. Climbing more stairs typically requires more
AuCeffort PLN 100 .538 -.172 time. Although this is in some sense a confound, it is nonetheless
AuCsocial PLN 10,000 .117 .630 a characteristic of many situations that people find themselves in
AuCsocial PLN 50 .087 .608 every day, particularly those involving iterative tasks, and one that
AuCprobability PLN 150 -.070 .591
might be exploited in efforts to reduce maladaptive behavior.
AuCprobability PLN 30.000 -.148 .461
The present findings suggest that individual differences in
different types of discounting reflect largely separate factors or
Note: Factor loadings over .32 appear in bold traits, but it also is possible to capture the different processes
422 Psychol Rec (2019) 69:415–424

involved using a single model. For example, Rachlin (1993) Jimura et al., 2011; for a review, see Green & Myerson,
proposed a three-component model of discounting in which 2013), and it has been suggested that probability discounting
the subjective value of an outcome is based on three processes: is also domain-specific (Terrell et al., 2014). We need to know
delay, probability, and social discounting. This model may be whether someone who is a steep discounter when making
extended by adding additional discounting processes. For ex- decisions involving one type of cost factor and kind of out-
ample, incorporating effort, Rachlin’s original model becomes: come will also be a steep discounter when the decisions in-
A volve other cost factors and commodities.
V¼ ð1Þ Regardless of how many traits are involved in different
ð1 þ kDÞð1 þ hΘÞð1 þ sN Þð1 þ lEÞ
types of discounting, we think it is necessary to study
where V is the subjective value of a reward of amount A, and the combinations of different discounting cost factors, rather
letters D, Θ, N, E stand for delay, probability (as odds against, than studying each one in isolation. After all, everyday
or Θ), the number of people with whom the reward is to be choice situations often involve outcomes that are not only
shared, and the measure of effort, respectively. The correspond- delayed and probabilistic but also require effort to obtain,
ing discounting parameters are represented by k, h, s, and l. (It and once obtained, they may be shared with others. For
also is possible, of course, that different exponents may be this reason alone, not to mention the theoretical signifi-
associated with the different expressions in the denominator, cance of the interactions and correlations among the types
as in Vanderveldt, Green, & Myerson, 2015.) of discounting, and regardless of the number of mecha-
It should be emphasized that the fact that different types of nisms and traits involved, these situations demand atten-
discounting can be combined into a single equation to de- tion in future studies.
scribe the interactions among different cost factors does not
mean that a single mechanism underlies the different types or Acknowledgements We thank members of the Behavioral Economics
that they involve the same or correlated traits. The opposite Research Group (BERG) at SWPS University of Social Sciences and
Humanities for their discussions. The research was funded by the
amount effects found with delayed and probabilistic rewards National Science Centre Poland research grant 2015/19/D/HS6/00770,
are evidence that different mechanisms are involved in at least and preparation of the manuscript was supported in part by the National
these two types of discounting, and there also is evidence that Institute on Aging of the National Institutes of Health under Award
they involve separate traits. Jones and Rachlin (2009) found Number R01AG058885.
that delay, probability, and social discounting were positively
correlated, and Mitchell (1999) reported weak but positive Compliance with ethical standards
correlations between delay, probability, and effort
Conflict of interest The authors declare that they have no conflict of
discounting. However, using multiple measures of delay and interest.
probability discounting, others have found that they load on
separate factors (Green & Myerson, 2013; Mejía-Cruz et al., Ethical approval All procedures performed in studies involving human
2016), and they also loaded on separate factors in the present participants were in accordance with the ethical standards of the institu-
study, indicating that these two types of discounting involve tional and/or national research committee and with the 1964 Helsinki
declaration and its later amendments or comparable ethical standards.
separate traits. In any case, it should be noted that positive
correlations between delay and probability discounting are
the opposite of what would be expected if both reflected a Open Access This article is distributed under the terms of the Creative
Commons Attribution 4.0 International License (http://
single impulsivity trait, because whereas steep delay creativecommons.org/licenses/by/4.0/), which permits unrestricted use,
discounting could be construed as reflecting impulsivity, steep distribution, and reproduction in any medium, provided you give appro-
probability discounting corresponds to risk aversion rather priate credit to the original author(s) and the source, provide a link to the
than risk taking (Green & Myerson, 2010). Creative Commons license, and indicate if changes were made.
To examine the relations among different types of
discounting, the present study used hypothetical monetary re-
wards as the outcome in each case so as not to confound type
of discounting with kind of reward. However, the relations
References
among the discounting of different rewards are also of consid-
erable interest and bear on the question of how many mecha- Amlung, M., Vedelago, L., Acker, J., Balodis, I., & MacKillop, J. (2017).
nisms and traits underlie the phenomena collectively termed Steep delay discounting and addictive behavior: a meta-analysis of
“discounting.” Although some researchers have highlighted continuous associations. Addiction, 112(1), 51–62.
what is domain-general in delay discounting, as revealed by Białaszek, W., Marcowski, P., & Ostaszewski, P. (2017). Physical and
cognitive effort discounting across different reward magnitudes:
positive correlations among discounting of different outcomes Tests of discounting models. PloS One, 12(7), e0182353.
(e.g., Odum, 2011), others have pointed out that delay Cattell, R. B. (1966). The scree test for the number of factors.
discounting is often domain-specific (Chapman, 1996; Multivariate Behavioral Research, 1(2), 245–276.
Psychol Rec (2019) 69:415–424 423

Chapman, G. B. (1996). Temporal discounting and utility for health and O’Connor, B. P. (2000). SPSS and SAS programs for determining the
money. Journal of Experimental Psychology: Learning, Memory, & number of components using parallel analysis and Velicer’s MAP
Cognition, 22(3), 771–791. test. Behavior Research Methods, Instruments, & Computers, 32(3),
Cohen, J. (1988). Statistical power analysis for the behavioral sciences 396–402.
(2nd ed.). Hillsdale: Lawrence Erlbaum Associates. Odum, A. L. (2011). Delay discounting: Trait variable? Behavioural
Estle, S. J., Green, L., Myerson, J., & Holt, D. D. (2007). Discounting of Processes, 87(1), 1–9.
monetary and directly consumable rewards. Psychological Science, Ohmura, Y., Takahashi, T., Kitamura, N., & Wehr, P. (2006). Three-
18(1), 58–63. month stability of delay and probability discounting measures.
Fabrigar, L. R., Wegener, D. T., MacCallum, R. C., & Strahan, E. J. Experimental & Clinical Psychopharmacology, 14(3), 318–
(1999). Evaluating the use of exploratory factor analysis in psycho- 328.
logical research. Psychological Methods, 4(3), 272–299. Osiński, J. T., Ostaszewski, P., & Karbowski, A. (2014). Social
Green, L., & Myerson, J. (2010). Experimental and correlational analyses discounting rate is negatively correlated with fluid intelligence.
of delay and probability discounting. In G. J. Madden & W. K. Personality & Individual Differences, 59, 44–49.
Bickel (Eds.), Impulsivity: The behavioral and neurological science Ostaszewski, P., Bąbel, P., & Swebodziński, B. (2013). Physical and
of discounting (pp. 67–92). Washington, DC: American cogniti9ve effort discounting of hypothetical monetary rewards.
Psychological Association. Japanese Psychological Research, 55(4), 329–337.
Green, L., & Myerson, J. (2013). How many impulsivities? A discounting Ostaszewski, P., & Osiński, J. T. (2011). Social discounting of monetary
perspective. Journal of the Experimental Analysis of Behavior, 99(1), 3– rewards: The effects of amount and social relationship. European
13. Psychologist, 16(3), 220–226.
Green, L., Myerson, J., Oliveira, L., & Chang, S. E. (2013). Delay Rachlin, H. (1993). The context of pigeon and human choice. Behavior
discounting of monetary rewards over a wide range of amounts. and Philosophy, 21(2), 1–17.
Journal of the Experimental Analysis of Behavior, 100(3), 269–281. Rachlin, H. (2006). Notes on discounting. Journal of the Experimental
Green, L., Myerson, J., & Ostaszewski, P. (1999). Amount of reward has Analysis of Behavior, 85(3), 425–435.
opposite effects on the discounting of delayed and probabilistic out- Rachlin, H., & Jones, B. A. (2008a). Social discounting and delay
comes. Journal of Experimental Psychology: Learning, Memory, & discounting. Journal of Behavioral Decision Making, 21(1), 29–43.
Cognition, 25(2), 418.
Rachlin, H., & Jones, B. A. (2008b). Altruism among relatives and non-
Holt, D. D., Green, L., & Myerson, J. (2003). Is discounting impulsive?
relatives. Behavioural Processes, 79(2), 120–123.
Evidence from temporal and probability discounting in gambling and
Rachlin, H., & Raineri, A. (1992). Irrationality, impulsiveness, and selfish-
non-gambling college students. Behavioural Processes, 64(3), 355–367.
ness as discount reversal effects. In G. F. Loewenstein & J. Elster
Jimura, K., Myerson, J., Hilgard, J., Keighley, J., Braver, T. S., & Green, L.
(Eds.), Choice over time (pp. 93–118). New York: Russell Sage
(2011). Domain independence and stability in young and older adults’
Foundation.
discounting of delayed rewards. Behavioural Processes, 87(3), 253–
Rachlin, H., Raineri, A., & Cross, D. (1991). Subjective probability and delay.
259.
Journal of the Experimental Analysis of Behavior, 55(2), 233–244.
Jones, B. A., & Rachlin, H. (2009). Delay, probability, and social
discounting in a public goods game. Journal of the Experimental Raineri, A., & Rachlin, H. (1993). The effect of temporal constraints on
Analysis of Behavior, 91(1), 61–73. the value of money and other commodities. Journal of Behavioral
Kaiser, H. F. (1974). An index of factorial simplicity. Psychometrika, Decision Making, 6(2), 77–94.
39(1), 31–36. Reynolds, B., Karraker, K., Horn, K., & Richards, J. B. (2003). Delay
Kirby, K. N. (1997). Bidding on the future: Evidence against normative and probability discounting as related to different stages of ad-
discounting of delayed rewards. Journal of Experimental olescent smoking and non-smoking. Behavioural Processes,
Psychology: General, 126(1), 54–70. 64(3), 333–344.
MacKillop, J., Amlung, M. T., Few, L. R., Ray, L. A., Sweet, L. H., & Reynolds, B., Richards, J. B., Horn, K., & Karraker, K. (2004).
Munafò, M. R. (2011). Delayed reward discounting and addictive Delay discounting and probability discounting as related to
behavior: a meta-analysis. Psychopharmacology, 216(3), 305–321. cigarette smoking status in adults. Behavioural Processes,
Madden, G. J., & Bickel, W. K. (2010). Impulsivity: The behavioral and 65(1), 35–42.
neurological science of discounting. Washington, DC: American Shead, N. W., & Hodgins, D. C. (2009). Probability discounting of gains
Psychological Association. and losses: Implications for risk attitudes and impulsivity. Journal of
Mejía-Cruz, D., Green, L., Myerson, J., Morales-Chainé, S., & Nieto, J. the Experimental Analysis of Behavior, 92(1), 1–16.
(2016). Delay and probability discounting by drug-dependent co- Sugiwaka, H., & Okouchi, H. (2004). Reformative self-control and
caine and marijuana users. Psychopharmacology, 233(14), 2705– discounting of reward value by delay or effort. Japanese
2714. Psychological Research, 46, 1–9.
Mitchell, S. H. (1999). Measures of impulsivity in cigarette smokers and Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics.
nonsmokers. Psychopharmacology, 146, 455–464. Boston: Allyn & Bacon.
Mitchell, S. H. (2003). Discounting the value of commodities according Takahashi, T., Ohmura, Y., Oono, H., & Radford, M. (2009). Alcohol use
to different types of cost. In N. Heather & R. E. Vuchinich (Eds.), and discounting of delayed and probabilistic gain and loss.
Choice, behavioral economics, and addiction (pp. 339–357). Neuroendocrinology Letters, 30(6), 749–752.
Oxford: Elsevier. Terrell, H. K., Derenne, A., & Weatherly, J. N. (2014). Exploratory and
Myerson, J., Green, L., Hanson, J. S., Holt, D. D., & Estle, S. J. (2003). confirmatory factor analyses of probability discounting of different
Discounting delayed and probabilistic rewards: Processes and traits. outcomes across different methods of measurement. American
Journal of Economic Psychology, 24(5), 619–635. Journal of Psychology, 127(2), 215–231.
Myerson, J., Green, L., & Morris, J. (2011). Modeling the effect of reward Thaler, R. (1981). Some empirical evidence on dynamic inconsistency.
amount on probability discounting. Journal of the Experimental Economics Letters, 8(3), 201–207.
Analysis of Behavior, 95(2), 175–187. Vanderveldt, A., Green, L., & Myerson, J. (2015). Discounting of monetary
Myerson, J., Green, L., & Warusawitharana, M. (2001). Area under the rewards that are both delayed and probabilistic: delay and probability
curve as a measure of discounting. Journal of the Experimental combine multiplicatively, not additively. Journal of Experimental
Analysis of Behavior, 76(2), 235–243. Psychology: Learning Memory, & Cognition, 41(1), 148–162.
424 Psychol Rec (2019) 69:415–424

Velicer, W. F. (1976). Determining the number of components from Yi, R., Carter, A. E., & Landes, R. D. (2012). Restricted psychological
the matrix of partial correlations. Psychometrika, 41(3), 321– horizon in active methamphetamine users: future, past, probability,
327. and social discounting. Behavioural Pharmacology, 23(4), 358.
Weatherly, J. N., Terrell, H. K., & Derenne, A. (2010). Delay discounting
of different commodities. Journal of General Psychology, 137(3), Publisher’s Note Springer Nature remains neutral with regard to jurisdic-
273–286. tional claims in published maps and institutional affiliations.