Content uploaded by Jiao Zhang
Author content
All content in this area was uploaded by Jiao Zhang
Content may be subject to copyright.
1
䉷2003 by JOURNAL OF CONSUMER RESEARCH, Inc. ●Vol. 30 ●June 2003
All rights reserved. 0093-5301/2004/3001-0001$10.00
Medium Maximization
CHRISTOPHER K. HSEE
FANG YU
JIAO ZHANG
YAN ZHANG*
A medium—for example, points or money—is a token people receive as the im-
mediate reward of their effort. It has no value in and of itself, but it can be traded
for a desired outcome. Experiments demonstrate that, when people arefaced with
options entailing different outcomes, the presence of a medium can alter what
option they choose. This effect occurs because the medium presents an illusion
of advantage to an otherwise not so advantageous option, an illusionof certainty
to an otherwise uncertain option, or an illusion of linearity to an otherwise concave
effort-outcome return relationship. This work has implications for how pointsinflu-
ence consumer choice and how money influences human behavior.
Wealth is evidently not the good we are seek-
ing; it is merely useful and for the sake of
something else. (Aristotle,Nicomachean
Ethics)
In order to achieve a desired outcome, we usually have
to exert effort. Often, however, the immediate payoff of
our effort is not the outcome we actually care about. Instead,
it is merely an instrument, a token, which has no value in
itself but can be traded for the outcome we care about. In
other words, it is merely a “medium” between our effort
and the desired outcome, as in the following:
Effort rmedium routcome.
Media are present in many decision contexts. For ex-
ample, when a frequent-flyer-program member flies, he ac-
cumulates miles. Miles, however, are not what he really cares
about. They are merely a medium that he can trade for free
travel (a desired outcome) later on. When a loyalty-program
member purchases a product, she may earn points, but the
*Christopher K. Hsee is professor of behavioral science and marketing at
the University of Chicago Graduate School of Business, Chicago, IL 60637
(chris.hsee@gsb.uchicago.edu). Fang Yu is Ph.D. candidate, University of
Chicago School of Business (fang.yu@gsb.uchicago.edu). Jiao Zhang is
Ph.D. candidate, University of Chicago Graduate School of Business
(jiao.zhang@gsb.uchicago.edu). Yan Zhang is research associate at China
Europe International Business School, Shanghai, China (zronnie@
ceibs.edu). Address correspondence to Christopher K. Hsee. Many individuals
have provided helpful comments on this research. In particular, the authors
thank Pankaj Aggarwal, Danny Kahneman, Yuval Rottenstreich, Suzanne
Shu, Jack Soll, Dick Thaler, Stijn van Osselaer, George Wu, and the JCR
editor, associate editor, and reviewers for their helpful comments on drafts
of this article. Funding for this project is provided by the University of
Chicago Graduate School of Business, the National Science Foundation, and
China Europe International Business School.
points are not what she really wants. They are just a medium
that she can redeem for a gift (a desired outcome) later on
(see, e.g., Kivetz and Simonson 2002; Van Osselaer, Alba,
and Manchanda 2001). When children in schools that adopt
a token economy complete a task, they may receive chips
(Kazdin 1982), but the chips are not what the children really
want. They are just a medium that the children can trade
for a piece of candy or some other desired good later on.
More importantly, the money we earn from work is also a
medium. Thus, the potential implication of research on me-
dium is not medium; it is extra large.
Since the medium is inherently worthless, people should
skip it and base their decision solely on the effortroutcome
relationship. However, the effort routcome relationship is
typically not directly given, and one has to infer it from the
two typically given relationships: the relationship between
effort and the immediate payoff in medium (effort rme-
dium) and the exchange relationship between medium and
outcome (medium routcome). In order to effectively skip
the medium, one has to pay equal attention to these two
types of relationships.
Our tenet is that people often fail to fully skip the medium
and they maximize not just the effort routcome return but
also the effort rmedium return. We refer to the pursuit of
the effort rmedium return as medium maximization. For
example, suppose that a person could choose either a less-
effortful action that gives her level of medium , which
M
1
corresponds to outcome , or a more effortful action that
O
1
gives her level of medium , which corresponds to out-
M
2
come . Her decision of which action to take will be in-O
2
fluenced not only by how much is better than butOO
21
also by how much is greater than . In other words,MM
21
if is sufficiently greater than , then, even if is noMMO
212
2JOURNAL OF CONSUMER RESEARCH
better than , she may still choose the more effortful actionO
1
yielding and .MO
22
The remainder of the article is organized as follows. We
first review the relevant literature and then define what we
refer to as a medium effect. The main body of the article
focuses on three specific types of medium effect and presents
empirical findings for each type. We conclude with a dis-
cussion of theoretical and applied implications.
RELEVANT LITERATURE
A medium is similar to what learning theorists refer to
as a conditioned reinforcer. Vast literature on learning sug-
gests that a neutral stimulus can acquire a reinforcement
value through association with a primary reinforcer and can
change behaviors even after the primary reinforcer is re-
moved (e.g., Armus 1982; Boysen et al. 1996; Bugelski
1938; Herrnstein 1964; Mazur 1995; Williams and Dunn
1991). A medium, such as points, may have acquired its
reinforcement value via the extensive learning history ex-
tending back to one’s childhood. For example, more points
on a child’s exam are usually followed by more praise from
parents. As a consequence, points may have acquired a pos-
itive value and remain attractive even when the primary
reinforcers (e.g., praise from parents) are no longer present.
Our medium maximization idea is also inspired by prior
research suggesting psychological myopia. Psychological
myopia here refers to a tendency in decision makers to focus
on information immediately related to their choice or judg-
ment and to ignore other (e.g., background) information. A
prototypical example is money illusion, a phenomenon that
has intrigued economists and psychologists alike for many
years (Fisher 1928; Kahneman, Knetsch, and Thaler 1986;
see also Fehr and Tyran 2001; Shafir, Diamond, and Tversky
1997). Money illusion is the finding that, in times of infla-
tion, people overlook inflation rate information and base
their judgment of a financial outcome on its nominal value
rather than on its inflation-adjusted real value. For example,
people find a 10% salary increase in times of a 12% inflation
more satisfying than a 1% salary deduction in times of no
inflation, even though the latter is better in real monetary
terms. The money illusion reflects respondents’ tendency to
focus on the face value of the event to be judged and to
ignore the background exchange relationship between the
face value and the real outcome.
The same myopia seems to underlie medium maximi-
zation. As we discussed earlier, in decisions involving a
medium, the two pieces of information decision makers typ-
ically have are (a) the effort rmedium relationship and (b)
the medium routcome relationship. The former is about
the immediate reward of their choice, and the latter isabout
the exchange rate between medium and outcome (very much
like the exchange rate between nominal value and real value
in money illusion). In order to effectively cancel the influ-
ence of a medium, decision makers should give equal con-
siderations to the two relationships. But psychological my-
opia implies that decision makers will be sensitive to the
effort rmedium relationship but relatively insensitive to
the medium routcome relationship. As a result, they will
fail to fully cancel the influence of the medium.
Psychological myopia is a ubiquitous phenomenon and
is implied in many other, seemingly unrelated, studies. One
example is the research on proxy attributes versus funda-
mental attributes (e.g., Keeney 1980; 1994; Keeney and
Raiffa 1976). A proxy attribute is an indirect, and often
more available, index of a more fundamental attribute—a
factor with which the decision maker is more concerned.
For example, the concentration of pollutants in the air is a
proxy attribute for a more fundamental attribute—the health
consequence of pollution to humans. Keeney and Raiffa
(1976) argued that the use of proxy attributes may lead to
systematic biases and recommended that decision analysts
use fundamental attributes instead. Subsequent research by
Fischer et al. (1987) found that decision makers who were
presented with proxy attributes did not spontaneously trans-
late them into fundamental attributes and gave the proxy
attributes more weight than warranted by expected utility
theories. Consistent with medium maximization, this finding
reflects the tendency to focus on the immediate outcomes
and overlook the more fundamental outcomes.
Another study showing insensitivity to fundamental out-
comes is from the ultimatum games literature. Kagel, Kim,
and Moser (1996) found that, when players bargained over
chips with different exchange rates, their perception of fair-
ness was more focused on the distribution of the chips than
the distribution of the final outcomes.
In their work on loyalty programs, Van Osselaer et al.
(2001) have documented myopic maximization of intrinsi-
cally worthless points. In one of their simulated airline
choice studies, the authors asked research participants to
imagine that they could receive a free ticket if they had
earned a certain number of points from an airline and to
choose among several airlines that awarded points in dif-
ferent temporal sequences. The respondents tended to
choose the airline that awarded the most points at the time
of their choice, even though doing this was normatively
suboptimal.
Another example of psychological myopia is the pseudo-
certainty effect. Kahneman and Tversky (1984) found that
most respondents preferred a 20% chance at $45 over a 25%
chance at $30, but when the same problem was framed in
terms of a two-phase problem where one had a 25% chance
to pass the first phase and, if one passes the first phase, one
can choose between an 80% chance at $45 or a 100% chance
at $30, most people choose the latter. This result reflects
two psychological processes. First, certainty (100%) has a
particularly strong appeal. Second, and more pertinent to
the present work, respondents in the contingent-probability
condition based their decision myopically on the probabil-
ities directly associated with the choice options (80% vs.
100%) and overlooked the background probability of 25%.
Another example of psychological myopia in the context of
a two-phase problem is the recent work by Sood, Rotten-
streich, and Brenner (forthcoming) on decisions leading to
other decisions. For example, when consumers who are
MEDIUM MAXIMIZATION 3
shopping for a camera are asked at which camera store they
would buy the camera, instead of being directly asked which
camera they would buy, the consumers show high sensitivity
to the characteristics of the stores (e.g., number of options
available) and low sensitivity to the qualities of theavailable
cameras, which are the actual outcomes.
Another form of psychological myopia is identified in a
set of recent studies by Leclerc, Hsee, and Nunes (2003).
The studies show that consumers tend to favor a high-rank-
ing product in a low-ranking category over a low-ranking
product in a high-ranking category, holding the acutual qual-
ity of the products constant. This effect presumably occurs
because people evaluate the product relative to the the local
category rather than to all the products across all the cat-
egories. Leclerc et al. refer to this phenomenon as “narrow
focusing.”
More generally, Read, Loewenstein, and Rabin (1999)
suggest that, in making decisions involving multiple phases,
people often fail to broadly assess all the consequencestaken
together and myopically focus on the most immediately
available consequences. They refer to this phenomenon as
“narrow bracketing,” and they use it to explain a wide range
of research findings, including preferences for temporal se-
quences (e.g., Loewenstein and Prelec 1993; Read et al.
1999), risk aggregation (e.g., Benartzi and Thaler 1995; Tha-
ler et al. 1997), mental accounting (Thaler 1985, 2000),
attribute evaluability and joint-separate evaluation reversals
(e.g., Hsee 1996; Hsee et al. 1999), and taste change (e.g.,
Herrnstein and Prelec 1992; Heyman 1996). In this research,
we study this psychological myopia in the form of a medium
effect. Furthermore, we identify factors that moderate this
effect.
THE MEDIUM EFFECT
We define a medium effect as the difference in decision
between two consequentially equivalent conditions, one
with a medium (which we will refer to as the medium con-
dition) and one without (which we will refer to as the control
condition). For ease of discussion, let us consider the fol-
lowing simple medium and control conditions: In the me-
dium condition, people are faced with two options (option
1 and option 2), each corresponding to a certain level of
effort, a certain amount of medium, and a certain outcome.
They are informed of the relationship between effort and
medium and the relationship between medium andoutcome,
but not the relationship between effort and outcome. The
control condition is the same as the medium condition, ex-
cept that there is no medium and people are directly in-
formed of the relationship between effort and outcome.
Which option will people choose in each condition? For
ease of discussion, let E
1
and E
2
denote the disutility of
effort associated with option 1 and option 2, respectively;
O
1
and O
2
denote the desirability of the outcomes; and M
1
and M
2
denote the levels of medium.
In the control condition, we assume that people will sim-
ply compare the relative difference in effort between the
options with the relative difference in outcome between the
options. Thus, their likelihood of choosing one option over
the other, say, option 2 over option 1, can be modeled
roughly as
OE
22
L(control) p⫺.(1)
OE
11
The reason for using ratios rather than absolute differences
is that people usually judge the advantage of one option
over another in relative terms (Shafir, Osherson, and Smith
1993).
In the medium condition, our previous discussion on my-
opia and medium maximization suggests that people’s de-
cision will be influenced not only by the final outcome but
also by the medium. Thus, their likelihood of choosing op-
tion 2 over option 1 can be modeled roughly as
MOE
222
()
L medium pw+(1⫺w)⫺,(2)
MOE
111
where the weight w, ranging between zero and one, indicates
the degree to which the decision maker is influenced by the
medium.
1
A comparison between equation 1 and equation 2 yields
the following propositions. First, the presence of a medium
may lead people to make different choices. That is,
L(medium) may be different from L(control). We shall refer
to this difference as the medium effect. Second, the mere
presence of a medium is not sufficient to produce a medium
effect. As equations 1 and 2 show, whether the presence of
a medium leads to a medium effect depends on whether
is different from . If it is not, then equation 2M/MO/O
21 21
will reduce to equation 1, and there will be no medium
effect. We will elaborate on these propositions when pre-
senting the experiments.
THREE CASES OF THE MEDIUM EFFECT
The medium effect manifests itself in different situations.
In this article we focus on three types of situations that we
believe are representative of many real-world decisions in-
volving media. The first case is about a choice between a
less effortful action and a more effortful action, where the
more effortful action does not yield a much better outcome
than the less effortful one. Here, the medium may give the
more effortful action an illusion of advantage and lead more
people to choose it. The second case is about options whose
outcomes are uncertain. Here, the medium may give those
options an illusion of certainty and increase their attrac-
tiveness. The last case concerns situations where the mar-
ginal return of effort diminishes over time, that is, the return
function of effort is concave. Here, a medium may create
an illusion of linearity and lead people to exert more effort.
1
These equations are not meant to be precise models; they are only
approximations.
4JOURNAL OF CONSUMER RESEARCH
Of these three cases, the first is the most basic, and the latter
two are extensions.
MEDIUM EFFECT CASE 1: THE
ILLUSION OF ADVANTAGE
This section concerns situations where people choose be-
tween two options (say options 1 and 2), and the advantage
of option 2 over option 1 in medium is large but the ad-
vantage in actual outcome is small. In other words,
is large but is small.M/MO/O
21 21
This type of situation is reminiscent of many real-world
decisions where moving from a less effortful action to a
more effortful action (e.g., from a less demanding job to a
more demanding one) may be highly advantageous in the
medium (e.g., money) but not so advantageous in the final
outcome (e.g., overall happiness).
In this type of situation, the medium gives option 2, which
is originally not very advantageous to option 1, an illusion
of advantage, and thereby will increase people’s likelihood
of choosing option 2. This prediction can be easily derived
from equations 1 and 2. Because L(medium) is a function
of both and , and L(control) is a function ofM/MO/O
21 21
only , it is obvious that when is greater thanO/OM/M
21 2 1
(and wis not equal to zero), L(medium) is greaterO/O
21
than L(control). The above analysis leads to the following
hypothesis:
H1a: When faced with two options (option 1 and op-
tion 2), where is greater than , peo-M/MO/O
21 21
ple will be more likely to choose option 2 in the
medium condition than in the control condition.
The effect described in hypothesis 1a will be referred to as
a medium effect. If equations 1 and 2 are correct, then the
following hypothesis will also hold:
H1b: The medium effect could be “turned off” if the
ratio is made smaller and closer to theM/M
21
ratio.O/O
21
Hypothesis 1b sets a boundary condition for hypothesis 1a,
suggesting that the mere presence of a medium is not suf-
ficient to generate a medium effect.
We now report three studies. Studies 1 and 2 are pilot
studies, and they tested only hypothesis 1a. Study 3 tested
both hypotheses 1a and 1b.
STUDY 1 (TASK/ICE CREAM)
Overview
The study involved a choice between a short task and a
long task, each corresponding to a different flavor of ice
cream as the reward (outcome). The amount of ice cream
was kept constant (1 gallon). There were two between-sub-
ject conditions, control and medium. The relationships be-
tween task and outcome in these conditions can be sum-
marized symbolically as follows:
Control condition:
Short task (6 minutes) rvanilla ice cream.
Long task (7 minutes) rpistachio ice cream.
Medium condition:
Short Task (6 minutes) r60 points rvanilla ice cream.
Long Task (7 minutes) r100 points rpistachio ice
cream.
Notice that this design fits the characteristics of the sit-
uation discussed above, namely, . Here,M/M1O/O
21 21
or 1.67. In a pretest, we found that mostM/Mp100/60
21
students did not consider pistachio ice cream to be better
than vanilla ice cream. Hence, it is reasonable to assume
that is smaller than .O/OM/Mp1.67
21 2 1
Method
Ninety-six students from a midwestern university were
approached on campus and told that they could participate
in an experiment in the following week and receive 1 gallon
of Haagen Dazs ice cream in return. They then filled out a
questionnaire, ostensibly designed to gather information
from potential participants. In the control condition, the re-
spondents were told about the following week’s experiment
that they could choose between two similar tasks, one re-
quiring six minutes and one seven minutes; that the flavor
of the ice cream they would receive depended on which
task they chose; and that they would receive vanilla ice
cream, if they chose the short task, and pistachio ice cream,
if they chose the longer task.
The medium condition was the same as the control con-
dition, except that a medium was introduced: respondents
were told that they would receive 60 points, if they chose
the short task or 100 points, if they chose the longer task,
and that, with 50–99 points, they would receive vanilla ice
cream and with 100 or more points they would receive
pistachio ice cream. They were also told that the points had
no other value and were nontransferable.
After having made their choices, respondents were also
asked which type of ice cream they liked more. This question
was designed to test whether the medium influenced judg-
ments of the actual outcomes.
Results and Discussion
As summarized in figure 1, the result supports hypothesis
1a: more respondents chose the long task in the medium
condition than in the control condition
2
(xp7.43,p!
. We consider this as preliminary evidence for medium.01)
maximization.
A possible alternative explanation for the finding is that
respondents in the medium condition may have used points
to infer the relative desirability of the ice creams and there-
fore chose the pistachio ice cream, which required more
points. Had this been the case, then, in the subsequent ice
cream preference question, those respondents should also
have expressed a greater liking for pistachio ice cream than
MEDIUM MAXIMIZATION 5
FIGURE 1
STUDY 1 RESULTS
N
OTE
.—The percentage of participants choosing the long task is greater in the
medium condition than in the control condition, although the percentage ofpar-
ticipants preferring the outcome of the long task (pistachio ice cream) remains
the same across the two conditions.
those in the control condition. But this is not what we found.
As figure 1 shows, there was no difference. In both con-
ditions, the respondents liked vanilla ice cream more. It
seems that the presence of a medium may lead people to
choose a longer task and end up with a less preferred
outcome.
STUDY 2 (BANKING)
Overview
Study 2 replicates study 1 in a marketing-relevant context.
It also includes a poststudy questionnaire to ensure that the
respondents understood the instructions. The decision in the
study was to choose between two branches of a bank at
which to buy a certificate of deposit (CD). The outcome
was a prize offered by the bank for buying the CD. There
were two possible prizes: $100 now or $150 three years
from now when the CD matured. The study had two be-
tween-subject conditions, which can be summarized as fol-
lows:
Control condition:
Branch 1 (5 minutes away) r$100 now.
Branch 2 (6 minutes away) r$150 three years from
now.
Medium condition:
Branch 1 (5 minutes away) r100 points r$100 now.
Branch 2 (6 minutes away) r150 points r$150 three
years from now.
Again, this design matches the characteristics conducive
to a medium effect, namely, that . Here,M/M1O/O
21 21
pthe attractiveness ofM/Mp150/100 p1.5; O/O
21 21
$150 in three years/the attractiveness of $100 now. In a
pretest, most students preferred $100 now to $150 in three
years. Thus, is less than one and smaller thanO/O
21
M/Mp1.5.
21
Method
Respondents were 236 students from a midwestern uni-
versity in the United States and an east coast university in
China. Those in the control condition were asked to imagine
that they had $5,000 cash and planned to buy a three-year
$5,000 CD at one of two nearby branches of a large bank;
one branch was five minutes away, and the other was six
minutes away. In a one-day promotion, the bank was of-
fering a prize for those purchasing a $5,000 CD. The re-
spondents were told that, if they bought the CD at the branch
five minutes away, the bank would give them $100 im-
mediately, and if they bought the CD at the branch six
minutes away, the bank would give them $150 three years
later when the CD matured. They were asked to assume that
they only had enough money to buy one such CD. Their
task was to decide at which branch to buy the CD.
The medium condition was identical to the control con-
dition, except that a medium (points) was introduced be-
tween choice of branch and type of outcome. Respondents
were told that, if they bought the CD at the branch five
minutes away, the bank would give them 100 points, and
if they bought the CD at the branch six minutes away, the
bank would give them 150 points. They were also told that,
if they had 100 points, the bank would give them $100
immediately, and if they had 150 points, the bank would
give them $150 three years later when the CD matured. The
points were said to have no other value. The decision was
whether to buy the CD at the five-minute-away branch that
would give them 100 points or the six-minute-away branch
that would give them 150 points.
To test whether the respondents understood the instruc-
tions, we asked the following questions after they had made
their choice. The first question (for both the control and the
medium conditions) asked how many prizes one could pos-
sibly receive from the bank (the correct answer was one).
The second question (also for both the control and the me-
dium conditions) asked the respondents to recall the prize
they would receive if they bought the CD at the branch six
minutes away; the correct answer was $150 three years
later). The third question (only for the medium condition)
asked whether the points had any other use besides allowing
one to receive a cash prize; the correct answer was no). Of
the 236 respondents, 23 respondents answered one or more
of the questions incorrectly or did not answer all the ques-
tions, and they were excluded from the analysis. (Including
these respondents in the analysys does not significantly
change the results.)
Results and Discussion
Again, we found a medium effect. The proportion of re-
spondents choosing the six-minute-away branch was sig-
6JOURNAL OF CONSUMER RESEARCH
nificantly higher in the medium con-
2
(xp18.66,p!.01)
dition (44%) than in the control condition (17%).
This study also rules out several potential alternative ex-
planations for the medium effect of study 1 (Tasks/Ice Cream
study). One is that the respondents may have misunderstood
the instructions. This explanation is not likely for this study
because the result is based only on respondents who an-
swered all the poststudy questions correctly. Another alter-
native explanation is that points may have served as a cue
to the desirability of the outcomes. Again, this explanation
does not easily apply to study 2 because it is unlikely that
people need to infer the desirability of “$100 now” or “$150
in three years” from points. A third alternative explanation
for the medium effect in study 1 is that points may have
served as a cue for the importance of the tasks. Because the
long task in study 1 awarded more points than the short
task, the respondents may have considered the long task to
be more important and therefore chose that task. This ex-
planation is less viable for study 2. Study 2 is a scenario
study, and, in the scenario, the points were merely a means
of promotion by the bank. It is unlikely that respondents
used these points to infer the relative importance of their
tasks: to bank at five-minute-away branch or six-minute-
away branch.
STUDY 3 (SNICKERS/ALMOND JOY)
Overview
Study 3 extends studies 1 and 2 in two nontrivial direc-
tions. First, it included three medium conditions, with dif-
ferent ratios. This feature was designed to test hy-M/M
21
pothesis 1b, that concerning the boundary condition of the
medium effect. Second, unlike the first two studies, where
people with more points could not trade for the reward for
fewer points, in study 3, people with more points could
choose the reward for fewer points. This feature was de-
signed to make the study more realistic. In real life, people
with more medium typically have more options of outcomes.
To hold this feature constant across all conditions, we in-
troduced the flexibility not only to the medium conditions
but also to the control condition.
This study contained the following four conditions:
Control condition:
Short task (20 minutes) r1 pound of Snickers bars.
Long task (25 minutes) rchoice of 1 pound of Snickers
bars or 1 pound of Almond Joy bars.
Medium condition 1 (the 60/100 condition):
Short task (20 minutes) r60 points r1 pound of
Snickers bars.
Long task (25 minutes) r100 points rchoice of 1
pound of Snickers bars or 1 pound of Almond Joy bars.
Medium condition 2 (the 60/61 condition):
Short task (20 minutes) r60 points r1 pound of
Snickers bars.
Long task (25 minutes ) r61 points rchoice of 1
pound of Snickers bars or 1 pound of Almond Joy bars.
Medium condition 3 (the color medium condition):
Short task (20 minutes) ra brown ticket r1 pound
of Snickers bars.
Long task (25 minutes) ra blue ticket rchoice of 1
pound of Snickers bars or 1 pound of Almond Joy bars.
Notice that the final rewards associated with the tasks
were held constant across all the conditions. The differences
among the conditions lay in whether there was a medium
and what the medium was. In the control condition, there
was no medium, and was relatively small. Here, O
1
O/O
21
was the desirability of Snickers bars and O
2
was the desir-
ability of a choice between Snickers bars and Almond Joy
bars. In a pretest ( ), we found that most (71%) stu-np81
dents preferred Snickers bars to Almond Joy bars. Therefore,
the relative advantage of having the option of Almond Joy
bars was small, and would not be much greater thanO/O
21
one. In the condition, the ratio was rather60/100 M/M
21
large ( ). Thus, based on hypothesis 1a, we expected100/60
more people to choose the long task in this condition than
in the control condition.
The remaining two medium conditions were designed to
test hypothesis 1b, according to which the medium effect
could be turned off by manipulating . In theM/M60/61
21
condition, the ratio was reduced to , close toM/M61/60
21
one. In the color medium condition, the medium wassimply
colored tickets, and was either meaningless or one.M/M
21
Thus, we predicted that the medium effect found in the
original medium condition would fade or vanish in the latter
two medium conditions.
Method
This study was conducted in the dining hall of a southern
university in the United States with 174 unpaid students.
The method was similar to that used in study 1. The choices
were to complete a 20-minute survey or a 25-minute survey.
The outcome for completing the short task was to receive
one pound of Snickers bars and that for the long task was
to choose either 1 pound of Almond Joy bars or 1 pound
of Snickers bars.
Other than what is described above, the control condition
and the condition were basically the same as the60/100
control and the medium conditions in study 1, respectively.
The other two medium conditions were identical to the
condition, except that 60 points and 100 points were60/100
replaced with 60 points and 61 points, respectively, in the
condition, and with a brown ticket and a blue ticket,60/61
respectively, in the color medium condition.
After having made their decisions, participants were asked
which type of chocolates they liked more. This question was
MEDIUM MAXIMIZATION 7
FIGURE 2
STUDY 3 RESULTS
N
OTE
.—The percentage of participants choosing the long task is greater in the medium condition (where the medium ratio between the long and the short60/100
tasks is high) than in any of the other conditions, although the percentage of participants preferring the outcome unique to the long task (Almond Joy) is no greater in
that condition than in any of the other conditions.
designed to replicate the observation in study 1 that the
medium did not affect preferences for the outcomes.
Results and Discussion
The results are summarized in figure 2. As predicted in
hypothesis 1a, more people signed up for the long task in
the 60/100 condition than in the control condition
2
(xp
. As predicted in hypothesis 1b, the medium18.8, p!.001)
effect evaporated in the other two medium conditions
( in both cases), thus leaving the condition as
2
x!1 60/100
the only one that was significantly different from the control.
Further analyses reveal that the condition differed60/100
not only from the control but also from the other two me-
dium conditions from the con-
2
(xp13.47, p!.001, 60/61
dition, and from the color medium
2
xp12.81, p!.001,
condition). These results not only replicate the medium ef-
fect but also show that the medium effect did not arise from
the mere presence of the medium but from the illusion of
advantage the medium creates.
The chocolate preference data collected at the end of the
study revealed no significant differences between the control
and any of the medium conditions, a result consistent with
what we found in study 1. This study mimics real-life sit-
uations in which people can either exert less effort or more
effort where the real advantage of exerting more effort is
to have more options for reward, not more rewards. Nor-
matively, what people ought to do is consider whether the
extra options are any better than the existing ones, and if
they are not, then there is no need to make the extra effort.
But what this experiment shows is that, when a medium is
introduced, people may focus on the medium, and, if the
payoff in medium from the extra effort seems relatively
large, people will exert the extra effort, although the extra
options in the reward set are of no extra value to them.
MEDIUM EFFECT CASE 2: THE
ILLUSION OF CERTAINTY
This case is concerned with a choice between two options,
where the outcome of one option is certain, the outcome of
the other is uncertain, and the levels of medium in both
options are certain. We propose that the medium in this case
will create an illusion of certainty to the uncertain option
and therefore increase its attractiveness. Specifically, weof-
fer the following hypothesis:
H2: When faced with two choice options, one with a
risky outcome, one with a sure outcome, and both
with certain levels of medium in the medium con-
dition, more people will choose the option with a
risky outcome in the medium condition than in
the control condition.
The following study tests this hypothesis.
8JOURNAL OF CONSUMER RESEARCH
STUDY 4 (STOCK/CASH)
Overview
The choice in this study was whether to quit a temporary
job now, with a sure outcome, or to quit it in the following
month, with a risky outcome. The outcome was money, and
the medium was stock. The share price of the stock was $1
now and would be either $0 or $2 in the following month.
The two conditions of the study can be summarized as fol-
lows:
Control condition:
Quitting now r$1,000.
Quitting in a month r$0 or $6,000 with even chance.
Medium condition:
Quitting now r1,000 shares of stock r$1,000.
Quitting in a month r3,000 shares of stock r$0 or
$6,000 with even chance.
We predicted that more people would work for the extra
month in the medium condition than in the control condition
because the medium (stock) transformed the original risky
decision between a sure monetary payoff and an uncertain
monetary payoff into a seemingly riskless decision between
two apparently sure stock payoffs.
Method
Respondents (105 students from a southern university and
a midwestern university in the United States) were asked to
imagine that they had worked as a temporary assistant in a
company for one month and that they had the option either
to terminate the job now or to work one more month. For
ease of exposition, let us first consider the medium condi-
tion. Respondents were told that they would only receive
stock as their compensation and that they had to redeem
their shares as soon as they received them. Respondents
were further told that the stock value was $1/share at the
present time and would be either $0 or $2 per share with
equal chance a month later and that if they terminated their
job now, they would receive a total of 1,000 shares ofstock,
and if they worked one more month, they would receive a
total of 3,000 shares.
In the control condition, there was no mention of the
stock. The respondents were directly informed of the mon-
etary consequence of their choices: the monetary figures
were simply calculated from the corresponding stock values
in the medium condition. Thus, the control and the medium
conditions were identical in terms of the final outcome. In
both conditions, the dependent variable was whether to quit
now or to work one more month.
Results and Discussion
As expected, significantly more people (
2
xp10.30, p!
) opted to work one more month in the medium condition.001
(61%) than in the control condition (30%). In support of
hypothesis 2, the medium (stock) indeed increased the pop-
ularity of the risky option. Presumably, the medium (stocks)
endowed the risky option an illusion of certainty.
In real life, the outcome of an activity is often uncertain.
What our findings suggest is that the introduction of a me-
dium can create an illusion of certainty and thereby increase
willingness to engage in the activity. For example, theactual
benefit of physical exercises (e.g., jogging) is often uncer-
tain. however, if people are awarded points every time they
jog, they will probably be more motivated to perform the
exercise. Indeed, some fancy heart-rate monitors can present
joggers with feedback of how many calories they have
burned. To some extent, the calorie readings serve as a
medium, which makes people feel that for every unit of
effort they exert, they are achieving a guaranteed reward—a
jump in the reading, even though the actual out-
come—benefit to their health—remains uncertain.
Thaler and Shefrin (1981) hypothesized that keeping track
of one’s calorie consumption has a self-control function and
can lead people to eat less. This hypothesis is probably the
flip side of the heart-rate-monitor effect discussed above.
Just as telling people how many calories they have burned
can motivate them to exercise, telling people how many
calories they have gained can demotivate them to eat. In
both cases, the calorie count serves as a medium that pro-
vides a sense of certainty to an otherwise uncertain outcome.
MEDIUM EFFECT CASE 3: THE ILLUSION
OF LINEARITY
The two cases discussed so far are about choices between
discrete options. The case to be discussed here is about when
to terminate a continuous activity. In particular, the present
case involves two properties. First, people continuously ex-
ert effort (or endure pain or incur cost) over time, and their
decision is about when to terminate the ongoing process.
Second, the marginal outcome of the activity diminishes
over time. In other words, the outcome is a concave function
of time. Again, this situation is representative of manyreal-
world decisions, where the marginal return of effort dimin-
ishes over time.
Based on the medium maximization hypothesis, we pro-
pose the following hypothesis:
H3a: If the outcome of an ongoing aversive activity
is a concave function of time but the payoff of
the medium of the activity is a linear function
of time, then people in the medium condition
will engage in the activity longer than those in
the control condition.
Here, the medium generates an illusion of linearity.
However, according to the medium maximization notion,
the effect postulated in hypothesis 3a does not arise from
the mere presence of the medium but from the fact that the
medium payoff is more linear than the outcome. It implies
that, if the medium payoff pattern is the same as the outcome
MEDIUM MAXIMIZATION 9
FIGURE 3
RELATIONSHIP AMONG TIME, POINTS, AND M&M’S IN STUDY 5
N
OTE
.—In all the conditions, the M&M’s payoff is a concave function of time. In the control condition, this concave relationship is transparent. In the linear medium
condition, this concave relationship is masked by a linear relationship between points and time. In the concave medium condition, this concave relationship is again
transparent.
payoff pattern, the medium effect would vanish. This leads
to the following hypothesis:
H3b: If the payoff of the medium of the activity is the
same concave function as the outcome, then there
will be no medium effect.
The following study tested both hypotheses 3a and 3b.
STUDY 5 (NOISE/M&M’S)
Overview
The decision investigated in this study was when to quit
listening to a noise emitted from headphones. The reward
for listening to the noise was M&M’s. The medium was
points. The study consisted of three between-subject con-
ditions—control, linear medium, and concave medium. Fig-
10 JOURNAL OF CONSUMER RESEARCH
FIGURE 4
STUDY 5 RESULTS
N
OTE
.—The amount of time for which participants are willing to endure the
noise is greater in the linear medium condition than in either the control or the
concave medium condition.
ure 3 presents a graphic summary of the three conditions.
Note that the ultimate relationship between M&M’s (out-
come) and time (effort) was identical in all three conditions:
it was concave.
Participants were informed of these relationships prior to
the noise episode. While listening to the noise, they saw a
readout of the number of M&M’s (in the control condition)
or of the number of points (in the two medium conditions)
changing on a computer screen and could stop the noise at
any time.
We had two predictions. First, people in the linear-me-
dium condition would endure the noise longer than the con-
trol condition participants because the linear medium en-
dowed the otherwise concave M&M’s payoff function an
illusion of linearity. Second, people in the concave-medium
condition would not differ significantly from those in the
control condition because the pattern of the medium here
maintained the original M&M’s payoff pattern.
Method
Participants (100 students from a large university in
China) were asked to listen to an increasingly loud noise
emitted from a set of earphones, and in exchange they would
receive M&M’s plus a fixed payment of 5 yuans (approx-
imately 60 cents). M&M’s were popular among college stu-
dents there and were expensive relative to their living
expenses.
Participants were run individually, each seated in frontof
a computer and wearing a pair of earphones. Once the noise
started, the participant could terminate it at any time by
pressing a key. The dependent variable was how long the
participant would endure the noise before terminating it. The
longer they endured the noise, the more M&M’s they would
receive. If the participant did not terminate the noise by the
150th second, the computer automatically stopped.
In the control condition, before the noise started, partic-
ipants were told that, for the first 10 seconds they listened
to the noise, they would earn 10 M&M’s; for the second
10 seconds, nine M&M’s; for the third 10 seconds, eight
M&M’s; and so forth. During the noise episode, they saw
on the computer screen the number of M&M’s they had
earned.
In the linear medium condition, before the noise started,
participants were told that, for every second they listened
to the noise, they would earn one point, and that, for the
first 10 points, they would earn 10 M&M’s; for the second
10 points, nine M&M’s; for the third 10 points, eight
M&M’s; and so forth. During the noise episode, they saw
on the computer screen the number of points they had
earned. As will be elaborated upon later, the prediction is
that respondents in this condition would be willing to endure
the noise longer than those in the control condition.
In the concave medium condition, prior to the noise ep-
isode, participants were told that, for the first 10 seconds
they listened to the noise, they would earn 10 points; for
the second 10 seconds, nine points; for the third 10 seconds,
eight points; and so forth; and that, with every point, they
would receive one M&M.
After the experiment, participants were asked how many
M&M’s they thought they would receive for the last 10
seconds they had listened to the noise. This question was
designed to see whether the respondent’s judgment matched
the actual reward rate, as had been recorded by the computer.
We consider a moderately inaccurate judgment acceptable
because the inaccuracy may have resulted from the difficulty
of remembering the duration of the noise or the difficulty
of inferring the return rate from the readings on the computer
screen. But if the judgment was too far from reality, which
we consider as more than Ⳳ3 from the correct answer, the
respondent may have misunderstood the instructions. Nine
respondents fell into this category and were excluded from
analysis. Cutting the data at Ⳳ2orⳲ3 does not significantly
change the results.
Results and Discussion
The results (both the means and the medians) are sum-
marized in figure 4. As predicted in hypothesis 3a, respon-
dents in the linear medium condition endured the noise
longer than those in the control condition (tp1.98,pp
, two-tailed test; one-tailed test). As predicted.05 p!.05,
in hypothesis 3b, respondents in the concave medium con-
dition did not differ significantly from those in the control
condition ( NS, one-tailed test or two-tailed test). Fur-t!1,
ther analysis indicated that the linear medium condition dif-
fered not only from the control condition but also from the
concave medium condition, though the effect was only mar-
ginally significant ( , two-tailed test;tp1.70,pp.09 p!
one-tailed test). A planned contrast analysis that as-.05,
signed a weight of 2 to the linear medium condition and a
MEDIUM MAXIMIZATION 11
weight of ⫺1 to the other two conditions yielded a signif-
icant result ( ), which further con-F(1,88) p4.61, p!.05
firmed the prediction that willingness to endure the noise
was greatest in the linear medium condition.
Study 5 extends the previous studies to continuous be-
haviors and shows that the presence of a medium can indeed
increase willingness to exert effort when the payoff function
of the outcome is concave. It also demonstrates that the
medium effect is not due to the mere presence of the medium
but, rather, is due to the linear payoff pattern of the medium.
If the payoff pattern of the medium is as concave as the
outcome, there will be no medium effect.
Theoretically, neither the linear pattern of the medium nor
the concave pattern of the outcome is a necessary condition
for the medium effect. As long as the medium has a different
payoff pattern than the outcome, there should be a medium
effect.
Ecologically, however, the combination of a linearpattern
of the medium and a concave pattern of the outcome is
significant because it resembles many real-world work-
money relationships, where (a) the amount of money people
earn increases linearly as they work harder or longer, but
(b) the utility function of the money is concave.
A SYNTHESIS OF THE THREE CASES
In this section, we examine the three cases. In the first
case, a medium makes a not so advantageous option appear
more advantageous. In the second case, a medium trans-
forms a risky choice into a seemingly riskless one. In the
last case, a medium turns a concave payoff relationship into
a seemingly linear one.
Despite the apparent differences, the second and the third
cases are variations of the first. In each case, the medium
effect occurs because is greater than . To dem-M/MO/O
21 21
onstrate this, consider the Stock/Cash study (study 4), which
demonstrates the illusion-of-certainty effect. In that study,
shares of stock, and shares ofMp1,000 Mp3,000
12
stock; and , whereOpv($1,000), Opw(50%)v($6,000)
12
vis the prospect theory value function and wis the prospect
theory probability weighting function. Because the prospect
theory value function is concave in the gain domain and w
(50%) is typically no greater than 0.5 (e.g., Kahneman and
Tversky 1979), it follows that .
2
This ex-M/M1O/O
21 21
plains why people were more willing to work the extra
month in the medium condition than in the control condition.
The same analysis can be applied even to the Noise study
(study 5) involving continuous behavior. Let and de-OO
12
note the numbers of M&M’s earned at time and time ,tt
12
respectively , and and denote the number of(t1t)MM
21 1 2
points earned at and , respectively.
3
Then, for anytt t1
12 2
2
Note that even if we were to subject and to prospect theory’s valueMM
12
function transformation, it would still hold that v(M)/v(M)p
21
v(3,000)/v(1,000) 1O/Opw(50%)v($6,000)/v(1,000).
21
3
Strictly speaking, and should denote the desirability of the numberOO
12
of M&M’s at and , but the number of M&M’s and its desirablity arett
12
highly correlated within the range permitted in this study. Even if the two
variables are not perfectly correlated, the basic analysis here will still hold.
, the ratio in the linear medium condition10 seconds M/M
21
was always greater than the ratio , as well as than theO/O
21
in the concave medium condition. This explains whyM/M
21
the participants were more willing to endure the noise in
the linear medium condition than in the other two conditions.
These analyses illustrate the underlying similarity of the
illusion-of-certainty and the illusion-of-linearity cases to the
illusion-of-advantage case.
GENERAL DISCUSSION
People often make decisions in situations where the im-
mediate outcome is a medium. In this article, we have dem-
onstrated a systematic difference between choice with a me-
dium and choice without a medium, and we have identified
several boundary conditions of this effect. In this section,
we discuss potential topics for future research and practical
and theoretical implications of research on medium.
Potential Topics for Future Research
Research on medium can potentially encompass many
interesting topics. The current research covers only a small
subset of these topics, but we hope that it is seminal and
that, along with other recent works in the area (e.g., Kivetz
and Simonson 2002; Van Osselaer et al. 2001), it will gen-
erate interest for future research. The following are some
potential topics. First, this article has focused only on three
types of medium effects. We believe that a medium can
exert its influence through other channels. For example, sup-
pose that the actual outcome of a behavior has to be deliv-
ered long after the occurrence of the behavior. Then, if a
medium is introduced and is delivered right after the oc-
currence of the behavior, it is likely to have a positive effect.
This effect can be called the illusion of immediacy. Also,
controlling for the actual outcome, a stream of medium that
is dispensed in an ascending temporal sequence is likely to
produce a different effect than a stream of medium that is
dispensed in a descending temporal sequence (e.g., Van Os-
selaer and Alba 2000; see also Ariely 1998; Hsee and Abel-
son 1991; Loewenstein and Prelec 1993). This effect can
be called the illusion of trend. In general, a medium is likely
to have an effect on willingness to exert effort if the medium
alters the perceived return of the effort.
Second, although we have defined medium narrowly as
an instrument used to trade for something else, our findings
may be generalized to a more broadly defined medium—any
proxy representation for a more fundamental value. For ex-
ample, a student’s GRE score can be considered as a medium
for her chance of being admitted to a desired graduate pro-
gram (outcome) and a professor’s number of publications
can be considered as a medium for his contribution to the
field (outcome). Just as a narrowly defined medium can
create illusions of advantage, certainty, or linearity for the
outcome, so can a proxy variable create such illusions for
the more fundamental value.
Third, future research may investigate other moderators
of the medium effect than what this research has identified.
12 JOURNAL OF CONSUMER RESEARCH
Generally, factors that lead people to focus their attention
on the medium phase will enhance the medium effect. One
such factor is cognitive load. Cognitive load will limit peo-
ple’s mental capacity to see beyond the immediately avail-
able layer and thereby increase the influence of themedium.
Another such factor is the mental effort needed to figure
out the amount of medium for each option. The more effort
it requires, the more people will focus their attention on the
medium phase, and the greater will be the medium effect.
A study we recently conducted lent preliminary support to
this idea. The study involved one control condition (without
points) and two medium conditions (with points). In all
conditions, there were three choice options, and the corre-
sponding outcomes were similarly attractive. In the two me-
dium conditions, the three options entailed different numbers
of points. In one of the medium conditions (the easy-medium
condition), which option corresponded to how many points
was transparently given. In the other medium condition (the
difficult-medium condition), which option corresponded to
how many points was not readily given; the research par-
ticipants had to follow a specified calculation rule to figure
out the relationship by themselves. Two results emerged
from this study. First, those in the easy-medium condition
were more likely to choose the option with the most number
of points than were those in the control condition. This is
merely a replication of the regular medium effect. Second,
confirming our speculation about effort and the medium
effect, the tendency to choose the option with the most
number of points was even greater in the difficult-medium
condition than in the easy-medium condition. It seems that
people who exert effort to deal with the intricacies of media
are more susceptible to the influence of the media and more
oblivious to the final outcomes.
It should be noted that the moderators discussed above
are different from the moderators in the Snickers/Almond
Joy study (whether the medium ratio was or60/100
) and in the Noise/M&M’s study (whether themedium60/61
return was linear or concave). To appreciate the difference,
let us revisit our definition of the medium effect. We have
defined the medium effect as the difference in choice be-
tween the medium and the control conditions, namely, the
difference between
MOE
222
()
L medium pw+(1⫺w)⫺(2)
MOE
111
and
OE
22
L(control) p⫺.(1)
OE
11
These equations suggest that the medium effect can be mod-
erated by two types of factors: those influencing the size of
and those influencing the size of w. The moderatorsM/M
21
in the Snickers/Almond Joy study and in the Noise/M&M’s
study concern . On the other hand, the factors weM/M
21
proposed in this section—attention and cognitive load—
concern w. Since the current research has studied M/M
21
more extensively, future research should focus more on w.
Finally, future research should shed more light on the
psychological process underlying the medium effect. Our
preferred explanation of the effect is psychological myopia;
it is most consistent with the diverse literatures reviewed
earlier. According to that explanation, the effort rmedium
relationship is more directly related to one’s decision than
the medium routcome relationship, and therefore exerts a
greater impact.
We now entertain two other explanations. One also attrib-
utes the medium effect to the excessive influence of theeffort
rmedium relationship, but it attributes the excessive influ-
ence not to myopia but to people’s preference for objective
over subjective information (see Hsee et al. [forthcoming] for
evidence for this preference). Typically, which option brings
more medium (the effort rmedium relationship) seems more
objective and unequivocal than which medium level brings
the better outcome (the medium routcome relationship);
therefore it exerts more influence. However, this explanation
seems incompatible with the results of the Noise/M&M’s
study, in which the number of M&M’s is as unequivocal as
the number of points. Another explanation attributes the me-
dium effect merely to people’s preference for more medium.
Obviously, the first two explanations also rely on the as-
sumption that people prefer more medium but that they com-
bine that preference with myopia and preference for objective
information. The last explanation treats the preference for
more medium as a sufficient cause for the medium effect.
This explanation seems most parsimonious, but it also seems
too general and amorphous to be satisfactory. Regardless of
its ultimate explanation, the medium effect, we believe, is an
interesting and fertile topic, and it entails both practical and
philosophical implications, which we discuss next.
Implications
The practical implications of the current research are ob-
vious. For instance, it can help organizations devise token
reinforcement systems to boost worker motivation. It can
also assist marketers in designing point-awarding frequency
programs to enhance consumer loyalty. Frequency programs
have become ubiquitous in recent years, and researchers
have identified many factors that contribute to their popu-
larity (e.g., Kivetz and Simonson 2002; Van Osselaer et al.
2001). The current work provides an additional perspective:
marketers can stimulate purchasing behavior by manipulat-
ing the way the medium in such a program—for example,
points or credits—is distributed.
More importantly, the current work delivers broader phil-
osophical implications. Although in this article we have
treated things such as candies and money as “outcomes,” they
are actually also media. Arguably, the ultimate outcome of
any action is affective experience—broadly defined, happi-
ness. Anything between our behavior and happiness is a me-
dium. This proposition leads to some interesting corollaries:
First, there are typically multiple layers of media between
MEDIUM MAXIMIZATION 13
our behavior and the ultimate outcome—happiness. For ex-
ample, when a frequent-flyer-program member flies, he ac-
cumulates miles. He then redeems the miles for a free airline
ticket. With the ticket, he takes a vacation. He then derives
happiness from the vacation. Here, miles, the ticket, and
even the vacation can all be construed as media, asillustrated
below:
Flying rmiles rticket rvacation rhappiness.
As another example, consider a student who uses the
money she has earned from work to purchase a camera, then
takes photos, and then derives joy from looking at the pho-
tos. Here, the money, the camera, and even the photos are
all media:
Work rmoney rcamera rphotos rhappiness.
Indeed, we are surrounded by media—multiple layers of
media. If you are “digitally correct,” you may say that life
is a “multimedia system.”
Second, although this research has focused primarily on
the relationship between narrowly defined media (e.g.,
points) and their immediate outcomes (e.g., candies), its
implications are much broader and are applicable to the
relationship between any layer of medium in the “multi-
media system” and the ultimate outcome, happiness. Social
scientists have observed that people in industrialized coun-
tries have been working harder and harder and accumulating
more and more wealth, yet their happiness has notincreased
appreciably (e.g., Diener and Biswas-Diener 2002; Easterlin
2001; Frank 1999; Frey and Stutzer 2002; Schor 1993; Veen-
hoven 1993). Although these observations have received
numerous interpretations (e.g., Brickman and Campbell
1971; Scitovsky 1992; see also Kahneman, Diener, and
Schwarz 1999), the current research provides a new per-
spective. Typically, the longer or harder we work, the more
money and material goods we can receive. In other words,
the relationship between work and wealth/material goods is
usually riskless and linear. However, the relationship be-
tween money/material goods and ultimate happiness is usu-
ally uncertain and concave. What the current research sug-
gests is that money/material goods can serve as a medium
that creates illusions of certainty and linearity. Thus, people,
who focus on these media are likely to overwork and ov-
eraccumulate money/material goods.
In this article, we have portrayed the pursuit of an in-
herently worthless medium as if it were a mistake. But it
may not be. The accumulation of a medium, especially when
it requires effort, may engender a sense of accomplishment
and self-efficacy (e.g., Bandura 1982) and generate joy in
and of itself. It may also enhance the enjoyment of the
reward, as the reward may be perceived as a self-gift for
the accomplishment (e.g., Mick and DeMoss 1990). If these
conjectures hold, then it is not irrational to pursue media.
For many people, happiness comes not only from the end
of the effort rmultimedia rhappiness chain but also from
amid the multimedia phases. In a way, the multimedia chain
is like an onion. An onion is a multilayer entity. When a
person peels off a layer, his action can be considered as a
means to accessing the heart of the onion. But there is not
really a heart. If people resist the temptation to enjoy each
layer as they peel inward, then there will be nothing left to
enjoy once they reach the center. To enjoy an onion is to
enjoy its layers. To enjoy life is to enjoy its media.
[David Glen Mick served as editor and Joel Huber served
as associate editor for this article.]
REFERENCES
Ariely, Dan (1998), “Combining Experiences over Time: The Ef-
fects of Duration, Intensity-Changes, and On-Line Measure-
ments on Retrospective Pain Evaluations,” Journal of Be-
havioral Decision Making, 11 (1), 19–45.
Armus, Harvard L. (1982), “Secondary Reinforcement Strength
and Acquisition Reinforcement Schedule,” Bulletin of the Psy-
chonomic Society, 20 (3), 129
Bandura, Albert (1982), “Self-Efficacy Mechanism in Human
Agency,” American Psychologist, 37 (January), 122–147.
Benartzi, Shlomo and Richard H. Thaler (1995), “Myopic Loss
Aversion and the Equity Premium Puzzle,” Quarterly Journal
of Economics, 110 (February), 73–92.
Boysen, Sarah T., Gary G. Berntson, Michelle B. Hannan, and
John T. Cacioppo (1996), “Quantity-Based Interference and
Symbolic Representations in Chimpanzees (Pan troglo-
dytes),” Journal of Experimental Psychology: Animal Behav-
ior Processes, 22 (January), 76–86.
Brickman, Philip and Donald T. Campbell (1971), “Hedonic Rel-
ativism and Planning for the Good Society,” in Adaptation
Level Theory: A Symposium, ed. M. H. Appley, New York:
Academic Press, 287–302.
Bugelski, Richard (1938), “Extinction with and without Sub-goal
Reinforcement,” Journal of Comparative Psychology,26
(March), 121–134.
Diener, Ed and Robert Biswas-Diener (2002), “Will Money In-
crease Subjective Well-Being?” Social Indicators Research,
57, 119–169.
Easterlin, Richard (2001), “Income and Happiness: Towards aUni-
fied Theory,” Economic Journal, 111, 465–484.
Fehr, Ernst and Jean-Robert Tyran (2001), “Does Money Illusion
Matter?” American Economic Review, 91 (December),
1239–1262.
Fischer, Gregory W., Nirmala Damodaran, Kathryn B. Laskey, and
David Lincoln (1987), “Preferences for Proxy Attributes,”
Management Science, 33 (September), 198–214.
Fisher, Irving (1928), The Money Illusion, New York: Adelphi.
Frank, Robert (1999), Luxury Fever: Why Money Fails to Satisfy
in an Era of Excess, New York: Free Press.
Frey, Bruno S. and Alois Stutzer (2002), “What Can Economists
Learn from Happiness Research?” Journal of Economic Lit-
erature, 40 (June), 402–435.
Herrnstein, Richard (1964), “Secondary Reinforcement and Rate
of Primary Reinforcement,” Journal of Experimental Analysis
of Behavior, 7 (January), 27–36.
Herrnstein, Richard and Drazen Prelec (1992), “Melioration,” in
Choice over Time, ed. George Loewenstein and Jon Elster,
New York: Russell Sage Foundation, 235–264.
Heyman, Gene M. (1996), “Resolving the Contradictions of Ad-
14 JOURNAL OF CONSUMER RESEARCH
diction,” Behavioral and Brain Sciences, 19 (December),
561–610.
Hsee, Christopher K. (1996), “The Evaluability Hypothesis: An
Explanation for Preference Reversals between Joint and Sep-
arate Evaluations of Alternatives,” Organizational Behavior
and Human Decision Processes, 67 (September), 247–257.
Hsee, Christopher K. and Robert P. Abelson (1991), “The Velocity
Relation: Satisfaction as a Function of the First Derivative of
Outcome over Time,” Journal of Personality and Social Psy-
chology, 60 (March), 341–347.
Hsee, Christopher K., George F. Loewenstein, Sally Blount, and
Max H. Bazerman (1999), “Preference Reversals between
Joint and Separate Evaluation of Options: A Review and The-
oretical Analysis,” Psychological Bulletin, 125 (September),
576–590.
Hsee, Christopher K., Jiao Zhang, Fang Yu, and YihengXi (forth-
coming), “Lay Rationalism and Inconsistency between Pre-
dicted Experience and Decision,” Journal of Behavioral De-
cision Making.
Kagel, John H., Chung Kim, and Donald Moser (1996), “Fairness
in Ultimatum Games with Asymmetric Information and
Asymmetric Payoffs,” Games and Economic Behavior, 13
(March), 100–110.
Kahneman, Daniel, Ed Diener, and Norbert Schwarz (1999),Well-
Being: The Foundations of Hedonic Psychology, New York:
Russell Sage Foundation.
Kahneman, Daniel, Jack Knetsch, and Richard Thaler (1986),
“Fairness as a Constraint on Profit-Seeking: Entitlements in
the Market,” American Economic Review, 76 (September),
728–741.
Kahneman, Daniel and Amos Tversky (1979), “Prospect Theory:
An Analysis of Decision under Risk,” Econometrica, 47
(March), 263–291.
——— (1984), “Choices, Values, and Frames,” American Psy-
chologist, 39 (4), 341–350.
Kazdin, Alan E. (1982), “The Token Economy: A Decade Later,”
Journal of Applied Behavioral Analysis, 15 (3), 431–445.
Keeney, Ralph L. (1980), Siting Energy Facilities, New York:
Academic Press.
——— (1994), “Creativity in Decision Making with Value-Fo-
cused Thinking,” Sloan Management Review, 35 (Summer),
33–41.
Keeney, Ralph L. and Howard Raiffa (1976), Decisions with Mul-
tiple Objectives: Preferences and Value Tradeoffs, New York:
Wiley.
Kivetz, Ran and Itamar Simonson (2002), “Earning the Right to
Indulge: Effort as a Determinant of Customer Preferences
towards Frequency Program Rewards,” Journal of Marketing
Research, 39 (May), 155–170.
Leclerc, France, Christoper K. Hsee, and Joseph Nunes (2003),
“Best of the Worst or Worst of the Best?” Working paper,
Graduate School of Business, University of Chicago, Chicago,
IL 60637.
Loewenstein, George F. and Drazen Prelec (1993), “Preferences
for Sequences of Outcomes,” Psychological Review, 100 (Jan-
uary), 91–108.
Mazur, James E. (1995), “Conditioned Reinforcement and Choice
with Delayed or Uncertain Primary Reinforcers,” Journal of
the Experimental Analysis of Behavior, 63 (March), 139–150.
Mick, David G. and Michelle DeMoss (1990), “Self-Gifts: Phe-
nomenological Insights from Four Contexts,” Journal of Con-
sumer Research, 17 (December), 322–332.
Read, Daniel, George Loewenstein, and Matthew Rabin (1999),
“Choice Bracketing,” Journal of Risk and Uncertainty,19
(December), 171–197.
Schor, Juliet B. (1993), The Overworked American: The Unex-
pected Decline of Leisure, New York: Basic Books.
Scitovsky, Tibor (1992), Joyless Economy: The Psychology of Hu-
man Satisfaction, New York: Oxford University Press.
Shafir, Eldar, Peter Diamond, and Amos Tversky (1997), “Money
Illusion,” Quarterly Journal of Economics, 112 (May),
341–373.
Shafir, Eldar, Daniel N. Osherson, and Edward E. Smith (1993),
“The Advantage Model: A Comparative Theory of Evaluation
and Choice under Risk,” Organizational Behavior and Human
Decision Processes, 55 (August), 325–378.
Sood, Sanjay, Yuval Rottenstreich, and Lyle Brenner (forthcom-
ing), “On Decisions That Lead to Decisions: Direct and De-
rived Evaluations of Preference,” Journal of Consumer
Research.
Thaler, Richard H. (1985), “Mental Accounting and Consumer
Choice,” Marketing Science, 4 (Summer), 199–214.
——— (2000), “Mental Accounting Matters,” in Choices, Values
and Frames, ed. Daniel Kahneman and Amos Tversky, Cam-
bridge: Cambridge University Press.
Thaler, Richard H. and Hersh M. Shefrin (1981), “An Economic
Theory of Self-Control,” Journal of Political Economy, 89
(April), 392–410.
Thaler, Richard H., Amos Tversky, Daniel Kahneman, and Alan
Schwartz (1997), “The Effect of Myopia and Loss Aversion
on Risk Taking: An Experimental Test,” Quarterly Journal
of Economics, 112 (May), 647–661.
Van Osselaer, Stijn M. J. and Joseph W. Alba (2000), “Consumer
Learning and Brand Equity, Journal of Consumer Research,
27 (June), 1–16.
Van Osselaer, Stijn M. J., Joseph W. Alba, and PuneetManchanda
(2001), “Loyalty Programs and Process in Mediated Intertem-
poral Choice,” Working paper, Graduate School of Business,
University of Chicago, Chicago, IL 60637.
Veenhoven, Rutt (1993), Happiness in Nations: Subjective Appre-
ciations of Life in 56 Nations, Rotterdam: Erasmus University.
Williams, Ben A. and Roger Dunn (1991), “Preference for Con-
ditioned Reinforcement,” Journal of the Experimental Anal-
ysis of Behavior, 55 (January), 37–46.
