This is the html version of the file http://pages.stern.nyu.edu/~jriis/Riis_et_al_2005-adaptation.pdf.
G o o g l e automatically generates html versions of documents as we crawl the web.

Google is neither affiliated with the authors of this page nor responsible for its content.
These terms only appear in links pointing to this page: ignorance of hedonic adaptation to hemo dialysis a 2
Ignorance of Hedonic Adaptation to Hemodialysis: A Study Using Ecological Momentary Assessment
Page 1
Ignorance of Hedonic Adaptation to Hemodialysis: A Study Using
Ecological Momentary Assessment
Jason Riis
University of Michigan
George Loewenstein
Carnegie Mellon University
Jonathan Baron
University of Pennsylvania
Christopher Jepson
University of Pennsylvania School of Medicine
Angela Fagerlin and Peter A. Ubel
University of Michigan and Veterans Affairs Medical Center, Ann Arbor
Healthy people generally underestimate the self-reported well-being of people with disabilities and
serious illnesses. The cause of this discrepancy is in dispute, and the present study provides evidence for
2 causes. First, healthy people fail to anticipate hedonic adaptation to poor health. Using an ecological
momentary assessment measure of mood, the authors failed to find evidence that hemodialysis patients
are less happy than healthy nonpatients are, suggesting that they have largely, if not completely, adapted
to their condition. In a forecasting task, healthy people failed to anticipate this adaptation. Second,
although controls understated their own mood in both an estimation task and a recall task, patients were
quite accurate in both tasks. This relative negativity in controls’ estimates of their own moods could also
contribute to their underestimation of the moods and overall well-being of patients.
In the same year that Brickman, Coates, and Janoff-Bulman
(1978) published the famously counterintuitive result that persons
with paraplegia are not that much less happy than lottery winners,
Sackett and Torrance (1978) demonstrated that there are other
serious health conditions that do not seem to be as badly experi-
enced by the people living with them as healthy people would
expect. For example, patients on dialysis rated their quality of life
as .56 (on a scale from 0 to 1), whereas healthy people estimated
that the quality of life of a patient on dialysis would be just .39.
Similar discrepancies have been demonstrated for other serious
health conditions (Boyd, Sutherland, Heasman, Tritcher, & Cum-
mins, 1990; Buick & Petrie, 2002) and for some less serious health
conditions (Baron et al., 2003).
Although the existence of the discrepancy is well established at
this point, its cause is not (Ubel, Loewenstein, & Jepson, 2003).
The most common explanation for the discrepancy is that it re-
flects bias on the part of healthy people—that is, that healthy
people think that health impairments are worse than they are.
Research in diverse domains has documented a general tendency
for people to underestimate their own and others’ speed of adap-
tation to negative as well as positive outcomes (Gilbert, Pinel,
Wilson, Blumberg, & Wheatley, 1998; Sieff, Dawes, & Loewen-
stein, 1999). A related mechanism is the focusing illusion
(Schkade & Kahneman, 1998; Wilson, Wheatley, Meyers, Gilbert,
& Axsom, 2000): When estimating the impact of any one life event
on their overall well-being, people tend to focus disproportionately
on aspects of life that would change as a result of the event and to
ignore aspects that would remain the same, thus exaggerating the
impact of the event (but see Ubel et al., 2001, and Baron et al.,
2003, for evidence that this latter account is less plausible).
Other kinds of explanations implicate the self-reports of chron-
ically ill or disabled patients themselves, rather than just the
judgments of healthy respondents. For a variety of reasons, pa-
tients may exaggerate their well-being, reporting a high quality of
life even though they experience more moments of misery and
fewer moments of joy than healthy people do. Patients may be in
the habit of exaggerating their reports of well-being for the benefit
of those who provide care for them (Diener, Suh, Lucas, & Smith,
1999), or they may simply need to avoid being perceived as
complainers so as to ensure continued support from family,
friends, and professional caregivers. They may also have devel-
oped a manner of coping with their situation whereby they tend to
focus on the more positive aspects of their experiences.
Jason Riis, Department of Psychology, University of Michigan; George
Loewenstein, Department of Social and Decision Sciences, Carnegie Mel-
lon University; Jonathan Baron, Department of Psychology, University of
Pennsylvania; Christopher Jepson, Department of Psychiatry, University of
Pennsylvania School of Medicine; Angela Fagerlin and Peter A. Ubel,
Division of General Internal Medicine, University of Michigan, and Vet-
erans Affairs Medical Center, Ann Arbor, Michigan.
The research was supported by National Institutes of Health Grant
AG16258 to Peter A. Ubel. We thank Barbara Fredrickson and Lisa
Feldman Barrett for discussion of methodological issues and Daniel
Kahneman, Norbert Schwarz, and Dylan Smith for commenting on a draft
of this article. We thank Linda Nyquist for assistance in recruiting control
subjects, and we thank the patients and staff at the participating dialysis
centers. We also thank Laura Damschroder, Christine Goldstein, Jennifer
Heckendorn, Susan Metosky, Molly Miklosovic, Jennifer Ream, Todd
Roberts, Brianna Sarr, Ryan Sherriff, and Penn Stabler for research
assistance.
Correspondence concerning this article should be addressed to Jason
Riis, who is now at the Center for Health and Wellbeing, Princeton
University, Princeton, NJ 08544. Email: jriis@princeton.edu
Journal of Experimental Psychology: General
Copyright 2005 by the American Psychological Association
2005, Vol. 134, No. 1, 3–9
0096-3445/05/$12.00 DOI: 10.1037/0096-3445.134.1.3
3
Page 2
Even if these motivational factors do not lead to exaggeration,
basic memory processes may have the same effect. In general,
summary reports of extended experiences tend to overweight cer-
tain salient features, such as the peak, end, or trend (for reviews,
see Ariely & Carmon, 2002; Kahneman, 1999, 2000), and they
tend to be insensitive to duration. Patients may experience fre-
quent, lengthy periods of moderately negative mood, but these
periods may not be well represented in an overall happiness
judgment if, say, the patients experience normal, positive peak
moods. Evidence suggests that the peak moods, however brief, will
have a disproportionate weight in an overall judgment (Kahneman,
Fredrickson, Schreiber, & Redelmeier, 1993), leading to exagger-
ated reports of well-being.
If, for any of these reasons, patients do exaggerate their well-
being, then the low well-being ratings that healthy people estimate
for patients could correctly reflect the difference in day-to-day
mood experienced by patients and healthy people. The discrepancy
in well-being ratings would reflect reality.
A related explanation for the discrepancy comes from evidence
that healthy people tend to underestimate their own past moods.
Thomas and Diener (1990) have suggested that people may gen-
erally tend to recall negative times more readily than positive
times. Indeed, they found that (healthy) people tend to underesti-
mate the percentage of time that they are in positive relative to
negative moods. Patients may not have this tendency; a coping
strategy that leads them to focus on positive aspects of life may
also make them less likely than healthy people to remember their
more negative moments. The tendency of healthy people to recall
more negative experiences could make them more likely than
patients to understate their own well-being. If healthy people
understate their own well-being, then they would be likely to
understate the well-being of other people (including patients) as
well, and this could contribute to the discrepancy.
The cause of the discrepancy matters. If healthy people misper-
ceive the quality of life associated with different health conditions,
they are likely to make suboptimal decisions both when it comes
to protecting their health and when it comes to deciding between
treatments. For example, someone suffering from a gastrointesti-
nal disorder who overestimates how miserable he or she would
be after having a colostomy would be likely to be too reluctant
to obtain one. The discrepancy could also lead policymakers
to misallocate scarce health care funds if different conditions
are subject to different degrees of bias. To make informed deci-
sions, laypersons and policymakers need to form accurate percep-
tions of how various health conditions would affect subjective
well-being.
Our goal in the research reported here is to test for the three
possible sources of the discrepancy discussed above: misestima-
tion of the impact of illness or disability by healthy people,
overstatement of mood experiences by patients, and understate-
ment of mood experiences by healthy people. This is accomplished
by measuring the moods of patients and controls and by comparing
these measured moods with various estimates and recollections of
mood. To measure mood, we use the method of ecological mo-
mentary assessment (EMA; Kahneman, 1999; Stone, Shiffman, &
DeVries, 1999).
In EMA, subjects are given personal digital assistants (PDAs;
e.g., Palm Pilots) to carry with them wherever they go for a period
of several days or more. The method is designed to minimize the
influence of biased recall. The PDA prompts the subject to answer
questions at random times throughout the day. In studies of well-
being, subjects are asked how they feel at that very moment.
Robinson and Clore (2002) found that subjects are faster to answer
questions about their immediate mood than about moods in the
past. Robinson and Clore argued that momentary mood reports are
thus less likely to reflect biases of episodic or semantic memory.
In this sense, momentary mood questions, asked repeatedly and in
the subject’s normal environment, are less likely to be edited by a
subject who wants to represent himself or herself in a certain way.
By asking people to make such momentary assessments several
times a day on a PDA, researchers can then aggregate these
assessments to get an overall picture of an individual’s experience
in a way that avoids the biases of the individual’s own recall or
aggregation process.
In this study, we compared mood estimates (made during inter-
views) and EMA mood responses (made over the course of a week
on a PDA) reported by a sample of chronically ill hemodialysis
patients and a sample of healthy controls. If the discrepancy in
well-being ratings occurs because patients tend to exaggerate their
own mood, then their mood estimates given at the beginning of the
week and their mood recollections made at the end of the week
should overstate their measured mood experience, as indicated by
their aggregated EMA responses. If this effect is large, then we
would expect the estimates and recollections of the patients to be
similar to those of the controls, but their measured mood should be
significantly lower. And the controls’ estimates of patients’ moods
may accurately reflect this lower mood.
Alternatively, if it is controls’ underestimation of their own
well-being that accounts for the discrepancy, then we would expect
their estimations and recollections of mood to underestimate their
measured mood. Patients, in contrast, may be quite accurate.
However, if patients have largely adapted to their condition,
then we would expect their aggregated EMA responses to be high
(higher scores indicating more positive mood), perhaps as high as
those of controls. If the discrepancy results from controls’ inability
to appreciate this adaptation, then the controls’ estimates of the
mood they would experience as hemodialysis patients should
understate the patients’ measured mood experience (as measured
by EMA).
Finally, to the extent that adaptation does occur, we hoped to
assess the extent to which patients have insights into its occur-
rence. We did so by asking them to estimate how happy they
would be if they had never had kidney trouble and had never
needed hemodialysis treatment. If patients have adapted and are
aware that they have adapted, then they should not expect good
health to have made them any happier than they are now. Also,
their estimates of “happiness if healthy” should not reflect a
higher level of happiness than that measured among healthy
nonpatients.
Method
Subjects
Subjects were 49 end-stage renal patients receiving hemodialysis treat-
ment three times per week and 49 healthy controls who were matched to
the patients on age, race, sex, and education. Subjects in 32 of the matched
4
RIIS ET AL.
Page 3
pairs were paid $30 to participate. The other subjects were paid $50 to
participate.
1
End-stage renal disease is a condition in which the kidneys fail to
perform their normal function of cleaning and filtering the blood. Treat-
ment consists of a procedure called hemodialysis in which a patient’s blood
is filtered through a machine. Most patients require treatment three times
per week for about 3 hr each time. Although discomfort and nausea are
possible, they are usually minor and the patient can read, write, talk, eat,
sleep, or watch TV during treatment. The patient’s lifestyle can include
most normal activities, including work, exercise, and leisure; however, the
patient will feel fatigued after missing treatment for several days. Also, the
patient must follow a strict diet that usually involves reducing sodium
intake, consuming relatively little meat, and drinking only small amounts
of fluids.
Patients were selected from a pool of 299 hemodialysis patients at nine
dialysis centers in the Ann Arbor, Michigan, area. Forty-five patients were
deemed ineligible because records or a preliminary interview showed that
they were either blind, deaf, or illiterate or that they did not speak English
or were experiencing dementia. The remaining patients were approached at
the centers while they were undergoing the treatment. After hearing the
description of the study, 90 of the remaining patients (35%) agreed to
participate.
Of the 90 patients who began the study, 82 completed all three phases
(described later). Sixty-nine of these patients responded to at least 40% of
the prompts during the EMA phase of the study. Healthy, matched controls
were sought for these 69 patients. Control subjects were recruited from
advertisements placed in a local newspaper. Individuals responding to the
advertisements were first screened in a telephone interview to ensure that
they did not have any major health conditions. Controls were sought to
match individual patients on sex, race, age (within 6 years), and years of
education (within 4 years). Fifty-six controls were recruited. Seven did not
complete all phases of the study (including 2 whose response rate during
the EMA phase of the study was less than 40%). Forty-nine were success-
fully matched to a patient.
Of the 49 matched pairs, 31 were women, 30 were Caucasian, 18 were
African American, and 1 was Hispanic. The average patient age was 49.3
years (SE
2.2) and the average control age was 49.0 years (SE
2.0).
The average education level was 14.1 years (SE
0.36) for the patients
and 15.0 years for the controls (SE
0.29). For the patients, the average
number of years on dialysis was 3.3 (SE
0.52). As intended, the controls’
self-reported health (M
2.2, SE
0.18) was better than that of the
patients (M
3.8, SE
0.16), as rated on a 5-point scale from 1
(excellent) to 5 (poor), t(29)
6.8, p
.001, d
1.7.
2
Procedure
Nonparticipant Questionnaire
To test for the possibility of selection bias among the patients, we gave
the nonparticipants (i.e., patients who were asked to participate but de-
clined) a questionnaire to assess their average mood. Specifically, they
were asked to rate their average mood during a typical week on a 5-point
response scale with the following response options: 2
very pleasant, 1
slightly pleasant, 0
neutral, 1
slightly unpleasant, and 2
very
unpleasant.
The questionnaire was given to 126 of the 164 nonparticipants (because
of an administration error, the questionnaire was not given to the first 38
nonparticipants). These patients were invited to complete the questionnaire
while they were at the center and undergoing treatment, and many reported
being simply too tired from the (concurrent) treatment to complete the
questionnaire. The response rate was nonetheless quite high (79%).
Three Stages
There were three stages to the study: (a) the entry interview, (b) the
EMA week, and (c) the exit interview. The mood estimation tasks used in
the entry and exit interviews were designed to allow direct comparison to
the EMA responses.
Entry Interview
During the entry interview, subjects first answered several sample ques-
tions on the PDA to familiarize themselves with how the screen could be
tapped to enter responses. They were also shown how the PDA could be
“put to sleep” so that it would not disturb them at night or at other
inappropriate times (e.g., while attending a movie).
Next, subjects completed a questionnaire in which they were asked to
estimate their typical mood. Subjects estimated the percentage of time
during a typical week that they would experience each of five mood levels.
The percentages were to add up to 100. The mood levels were the same
ones that would be offered as response options during the EMA session,
and they were the same ones that were used in the nonparticipant ques-
tionnaire (i.e., from 2, very pleasant, to 2, very unpleasant).
EMA
Subjects carried the PDAs for 7 days, beginning immediately after the
entry interview.
3
The first 7 pairs of subjects received PDAs scheduled to
beep randomly once within each 2-hr interval of the day (between 8:00 and
10:00, 10:00 and 12:00, etc.). The remaining 42 pairs received PDAs that
beeped once randomly within each 90-min period.
The first question was a single-item mood measure, instructing subjects
to think back to the mood they were feeling just before the PDA beeped
and to tap the button on the screen that best described that mood. Five
response buttons represented the same five mood levels described previ-
ously (i.e., from 2, very pleasant, to 2, very unpleasant).
There were nine additional mood measures for which subjects indicated
the extent to which they were experiencing specific emotions (happy,
joyful, pleased, enjoyment/fun, depressed/blue, unhappy, frustrated, angry/
hostile, worried/anxious) on a 0 to 6 scale anchored at not at all and
extremely much, respectively (cf. Thomas & Diener, 1990). Two additional
questions (using the same scale) asked subjects about the extent to which
they were feeling (a) pain or physical discomfort and (b) tired or fatigued.
Finally, on 10% of the prompts, an additional question was asked: On
these trials, this additional question was presented first, before the 12
questions described above. This was an overall life satisfaction question,
and because all other trials began with the momentary mood question, trials
beginning with the life satisfaction question opened with a brief introduc-
tion screen to emphasize the difference between the two kinds of questions.
The introduction screen said, “The next question will ask you about how
you feel about your life as a whole.” The question on the following
question screen was then simply, “How do you feel about your life as a
whole?” The seven response buttons corresponded to a 7-point scale
1
The increased payment was intended to increase recruitment rates, but
it had no effect.
2
Only the first 30 pairs were asked this health question. In an effort to
reduce participant burden, a questionnaire pertaining to health that was
administered at the end of the study was cut partway through the study.
This question was part of that questionnaire. Our recruiting methods were
similar for the remaining pairs, so there is no reason why this very large
difference in health would not also be observed for the remaining pairs.
3
The PDA was a Palm Pilot Model IIIxe. It ran a program called ESP
Blue, which was developed by Chip Jensen. The program can be down-
loaded at http://www.med.umich.edu/pihcd/esp/esp.htm. ESP Blue is
based on a program called ESP, which was developed at Boston College by
Lisa Feldman Barrett and programmed by Daniel J. Barrett. While the
program was running, users were not able to use the PDA’s other programs
or functions.
5
IGNORANCE OF HEDONIC ADAPTATION
Page 4
ranging from 3 to
3 and were anchored at very satisfied and very
unsatisfied, respectively.
Exit Interview
At the end of the EMA week, subjects made several mood estimates in
the same format as that used during the entry interview (i.e., estimates of
the percentage of time spent in each of the five mood levels). The
instructions for each estimate are described below.
Recall. Subjects estimated the percentage of time spent in each mood
level during the previous week (during which they had carried the PDA).
Typical. Subjects estimated the percentage of time spent in each mood
level during a typical week. This was identical to the estimation task given
at the entry interview.
Hemodialysis. Subjects were presented with a scenario describing the
experience of a hemodialysis patient. Patients and controls alike were
asked to imagine that they were the patient in the scenario. Controls were
asked to imagine that they had been hemodialysis patients for either 1 year
or for as long as their matched patient had been on hemodialysis (which-
ever was greater). All subjects then estimated the amount of time they
would spend in each mood level if their experience were the same as that
of the patient in the scenario.
Healthy. Patients estimated the percentage of time they would spend in
each mood level if they had never had kidney problems and had never
needed hemodialysis treatment.
Results
Analyses are reported for the 49 pairs of matched patients and
controls. All t tests are paired, unless otherwise specified.
4
Measured Mood: EMA Responses
The average EMA response to the overall mood question (on
the 2 to 2 scale) was 0.70 (SE
0.07) for the patients and 0.83
(SE
0.07) for the controls. This difference is not significant,
t(48)
1.4, p
.16. With this sample size and a critical value of
.05, we had the statistical power (.80) to detect a difference of 0.23
(on the 5-point scale) between the groups. It appears likely, then,
that if there is a difference in mood between the groups, the
difference is small. This supports the suggestion that the patients
have adapted quite well to their condition.
Further evidence of a lack of difference in mood comes from
responses to other EMA questions. For each subject, responses to
the four positive emotion questions (i.e., happy, joyful, pleased,
enjoyment/fun) were averaged, as were responses to the five
negative emotion questions (i.e., depressed/blue, unhappy, frus-
trated, angry/hostile, worried/anxious). The positive averages were
not significantly different between groups (M
3.1, SE
0.17,
for patients; M
3.3, SE
0.16, for controls), t(48)
1, nor were
the negative averages (M
0.9, SE
0.13, for patients; M
1.0,
SE
0.12, for controls), t(48)
1. There were also no significant
differences between groups on any of the nine individual emotions.
There were no significant differences between patients and
controls in their responses to the questions about pain (M
1.4,
SE
0.19, and M
1.1, SE
0.19, respectively), t(48)
1;
tiredness (M
2.1, SE
0.19, and M
2.1, SE
0.17,
respectively), t(48)
1; or overall life satisfaction (M
1.1, SE
0.18, and M
1.3, SE
0.16, respectively), t(45)
1.
The response rates for both groups were high. Patients re-
sponded to 72% of the PDA prompts (SE
2.1) for an average of
39 responses over the 7 days, whereas the controls responded to
78% (SE
2.1) for an average of 43 responses. These response
rates were significantly different, t(48)
2.3, p
.025, d
0.43;
however, EMA response rate was not correlated with average
EMA response, r(97)
.02, ns. Furthermore, when average EMA
response was regressed on EMA response rate, the group variable
(i.e., patient vs. control), and the interaction term, the interaction
term was not significant,
.001 (t
1). This does not support
the suggestion of a reporting bias whereby nonresponses occurred
during periods of better mood for one group than for the other.
For the patients, the number of years spent on hemodialysis was
not correlated with average EMA response, r(48)
.06, ns. All
patients had been on hemodialysis for at least 3 months, so
considerable time had already passed for adaptation to have oc-
curred. Furthermore, many patients would have had symptoms of
kidney disease for months or years before hemodialysis was re-
quired, so the period of adaptation to their health condition could
have been considerably longer than the months or years since they
began hemodialysis treatment.
Because our ultimate patient sample was not randomly selected,
we examined the possibility that patients were not representative,
with respect to mood, of the hemodialysis population. First, we
compared them with the nonparticipants from the dialysis centers.
That questionnaire was on the same 2 to
2 scale as the EMA
mood question. The average questionnaire response was 0.74
(SE
0.11), which is similar to the EMA average reported by the
49 matched patients (M
0.70, SE
0.07), t(146)
1. Further-
more, the 20 patients who responded to at least 40% of the EMA
prompts but who were not matched with a control subject were
very similar in EMA average to the 49 matched patients (M
0.69, SE
0.08, and M
0.70, SE
0.07, respectively), t(67)
1. Finally, of the 21 patients who did not complete the study, 17
had at least one EMA response, and their EMA average (M
0.69,
SE
0.12) was also similar to that of the 49 matched pairs,
t(64)
1. Among all patients, response rate was not correlated
with EMA average, r
.04, ns. None of these results support the
hypothesis that the matched patients were unrepresentative of the
hemodialysis population.
Average Mood Estimates
For each of the estimation tasks (i.e., entry, recall, typical,
hemodialysis, healthy), subjects estimated five mood percentages.
For some subjects, these percentages did not add up to 100, so they
were scaled accordingly. These scaled percentage estimates cor-
responded to a scalar estimated average mood, and those averages
are shown in Table 1. The averages were computed by dividing
each scaled percentage by 100 and multiplying the quotient by the
respective mood level (i.e., 2, 1, 0, 1, or 2). The sum of these
five values was the estimated average mood for that particular
estimation task.
Estimation of One’s Own Mood
For each subject, the difference between the entry estimate and
average EMA mood was computed. Patients’ entry estimates
slightly overestimated their EMA average mood (M
0.08, SE
4
None of the results change if, instead, the analyses compare the 49
controls with the full sample of 69 eligible patients (i.e., those who
responded to at least 40% of EMA prompts).
6
RIIS ET AL.
Page 5
0.08), although not significantly, t(48)
1.0, p
.32, whereas
controls did significantly underestimate their EMA average (M
0.16, SE
0.08), t(48)
2.1, p
.042. These estimation errors
were significantly different, t(48)
2.1, p
.042, d
0.44.
A similar pattern was observed for the recall estimates. Al-
though the patients’ recall estimates accurately reflected their
average EMA response (M
0.0, SE
.06), the controls showed
significant underestimation (M
0.23, SE
0 .07), t(48)
3.3,
p
.002. Again, these errors were different between groups,
t(48)
2.7, p
.009, d
0.49, suggesting that different processes
underlie patients’ and controls’ recall of their own moods.
5
An alternative explanation of these findings is that recall errors
are due not to memory differences but to biases in EMA reporting.
Because the subjects did miss some EMA prompts during the
week, an apparent recall error could occur if subjects accurately
recalled their actual mood and EMA was a biased representation of
actual mood. For such a bias to account for the data, the EMA bias
would have had to have been different for patients and controls,
with patients missing relatively more EMA prompts when they
were in bad moods and controls missing prompts when they were
in good moods. This interpretation seems unlikely given that, as
reported earlier, the relationship between response rate and aver-
age EMA response was not different for the two groups. Also,
response rate was not correlated with recall error, r(97)
.07,
ns, and when recall error was regressed on response rate, group,
and the interaction term, the interaction term was not significant,
.001 (t
1). This does not support the alternative explanation
attributing the recall error difference between patients and controls
to different biases in EMA response. The differences between
groups in recall error, then, can more readily be attributed to
differences in memory and evaluation processes.
In sum, there was little indication that patients exaggerate their
mood. In fact, their expectations and recollection were quite ac-
curate. However, the controls’ expectations were worse than their
measured experiences, and, consistent with prior research (Thomas
& Diener, 1990), their memories showed the same pattern.
Imagining Hemodialysis
For control subjects, the average estimate of mood while imag-
ining life under the hemodialysis scenario was 0.38 (SE
0.11).
This was a very large and significant underestimation of the EMA
average mood of the hemodialysis patients (M
0.70, SE
0.07),
t(48)
7.6, p
.001, d
1.7, and of the patients’ own estimation
of the hemodialysis scenario (M
0.63, SE
0.12), t(48)
5.6,
p
.001, d
1.3.
6
The controls’ estimates were significantly
negative, t(48)
3.43, p
.001, whereas the patients’ reported
experience was significantly positive, t(48)
9.95, p
.001. It is
important to note that patients’ mood estimates of a typical week
(at exit, M
0.61, SE
0.10) and of the hemodialysis scenario
(M
0.63, SE
0.12) were similar, t(48)
1, suggesting that the
scenario was a fair representation of their condition.
Mood if Healthy
For patients, the mood estimates when imagining that they had
never had kidney trouble or needed hemodialysis treatment (M
1.16, SE
0.09) were higher than their mood estimates for a
typical week (at exit, M
0.61, SE
0.10), t(48)
5.8, p
.001,
d
0.84, and higher than their EMA response average (M
0.70,
SE
0.09), t(48)
4.9, p
.001, d
0.82. Furthermore, their
healthy estimate was higher than the controls’ estimate of mood
during a typical week (at exit, M
0.67, SE
0.09), t(48)
4.0,
p
.001, d
0.79, and higher than the controls’ EMA response
average (M
0.83, SE
0.07), t(48)
3.1, p
.003, d
0.58.
These findings support the suggestion that patients are themselves
not aware of the extent to which they have adapted to their
condition.
Discussion
In this study, we sought to determine the source of the com-
monly observed discrepancy whereby quality-of-life ratings of
people who are sick or living with a disability are much higher
than healthy people expect. We did so by comparing measured and
estimated moods of hemodialysis patients and healthy controls.
Replicating earlier findings using different methods, we failed to
find evidence that patients experienced lower moods than healthy
controls did. Both patients and controls, however, predicted that
the difference in mood experienced under health versus illness
would be large. We also failed to find evidence that patients
exaggerate their mood, although we did find that healthy people
understate their own mood. Implications of these findings are
discussed below.
It appears that hemodialysis patients do, largely at least, adapt to
their condition. Although they report their health as being much
worse than that of healthy controls, they do not appear to be much,
if at all, less happy than people who do not have kidney disease or
any other serious health condition. The EMA procedure greatly
5
In an alternate test, EMA average was regressed simultaneously on
entry average and on group and then, separately, on recall average and on
group. In both cases, the group coefficient was significant,
.09, t(95)
2.0, p
.046, and
.09, t(95)
2.7, p
.008, respectively, consistent
with the results of the t tests. This rules out the possibility that the
significant t tests were artifacts of the use of difference scores. The
significant group coefficients also suggest that the lack of difference
between patients and controls in EMA average cannot be attributed to low
power.
6
For the hemodialysis estimate, 7 of the 49 control subjects were not
asked to imagine that they had had the condition for a period of time. They
were asked only to imagine that they were hemodialysis patients. When
these 7 controls are left out, the hemodialysis average for the 42 remaining
controls is 0.35 and the differences are still significant.
Table 1
Means of Measured and Estimated Mood for Hemodialysis
Patients and for Healthy Controls
Mood
Patients
(n
49)
Controls
(n
49)
Measured mood (average EMA response)
0.70
0.83
Estimated mood
Typical week (at entry)
0.78
0.67
Recall of EMA week
0.70
0.60
Typical week (at exit)
0.61
0.67
Imagining hemodialysis scenario
0.63
0.38
Imagining never having had kidney disease
1.16
Note. All means are on a 2 to 2 scale. EMA
ecological momentary
assessment.
7
IGNORANCE OF HEDONIC ADAPTATION
Page 6
reduced the likelihood of response biases. Subjects were asked
about their mood repeatedly, at different times of the day, through-
out the normal routine of their lives. If the patients really did spend
a great deal of time in a depressed mood, then this procedure
should have picked it up. The previously observed tendency of
healthy people to underestimate the reported quality of life of
people with various health conditions does seem to be due, in large
part, to their misperception of the extent to which people can adapt
to such conditions.
We cannot rule out the possibility of scale renorming, that is, the
possibility that the EMA response options meant something dif-
ferent to the patients than to the healthy controls. For example,
what a patient reports as a “very pleasant” feeling may be reported
as only “slightly pleasant” by a healthy person, because the pa-
tient’s standards may have lowered. We doubt very much, how-
ever, that this is the case. We have investigated the same possi-
bility in quality-of-life evaluations and, contrary to the scale
renorming hypothesis, have found that the discrepancy between
the evaluations of a particular condition by those who suffer from
that condition and by those who do not is actually greater when
quality of life is measured with scales that have well-defined
demarcations (Baron et al., 2003). We have also found the dis-
crepancy to persist even when scales that are unsusceptible to
recalibration are used (Lacey et al., 2004).
Headey and Wearing (1992) argued that people have a baseline
mood level to which they return after events move them tempo-
rarily above or below that baseline. Supporting this account are
findings from a twins study suggesting that genetic variation, not
variance in life circumstances, accounts for most of the variance in
well-being across individuals (Lykken & Tellegen, 1996). The
current finding is also consistent with this baseline account. Al-
though we do not have measures of well-being for people who
recently became sick, we do find that hemodialysis patients who
have endured illness and uncomfortable circumstances for months
or years are experiencing normal (or at least close to normal),
positive mood levels.
It is interesting that the hemodialysis patients themselves seem
unaware of the extent to which they have adapted. They believe
they would be happier if they had never been sick, yet they appear
to be incorrect in this belief, as they are already about as happy as
healthy people are. In imagining a life that had always been free of
illness, they may instead imagine the initial mood increase that
would follow the transition from their current state to one of good
health (Kahneman, 1999), and they may assume that the feeling
from such a transition would result not from the transition but from
the better quality of experience in the healthy state.
Healthy people are clearly unaware of the extent to which
adaptation to hemodialysis occurs. Their estimates of the moods
that they would experience if they were on hemodialysis were
much lower than the measured moods reported by the patients
actually on hemodialysis. In fact, healthy controls estimated neg-
ative average moods if they were on hemodialysis, whereas the
patients themselves actually reported positive average moods. This
is a rare case where people incorrectly estimate even the valence of
a different life circumstance (Wilson & Gilbert, 2003). The dis-
crepancy in well-being ratings reported by other researchers (e.g.,
Sackett & Torrance, 1978) thus does not seem to be a mere artifact
of different response processes used by patients and healthy people
when answering questions about their well-being. The surprisingly
high ratings that are often given by patients seem, at least in the
case of hemodialysis patients, to reflect a genuinely high frequency
and intensity of positive mood. To our knowledge, this is the first
study to show that healthy people grossly underestimate sick
people’s measured quality of emotional experience.
That said, we do find some evidence that part of the discrepancy
may be accounted for by differences in the manner in which
patients and healthy people make summary reports of well-being.
Consistent with prior research (Thomas & Diener, 1990), healthy
people tended to slightly underestimate their own average mood.
This was the case for the mood estimates (of a typical week) and
for the recall estimate of the past week. However, patients did not
underestimate their own average mood. It is possible that in coping
with their hardship, patients have developed a tendency to focus
more on positive experience. This may be a crucial part of the
adaptation process and should be an area of future investigation.
This difference in recall tendencies could account for some of
the discrepancy between patients’ self-reports of well-being and
healthy persons’ predictions of patients’ well-being. If healthy
people tend to recall past experiences as having been more nega-
tive than they were, then this should lead them to give slightly
deflated judgments of well-being. If their judgments of their own
well-being are deflated in this sense, then it is likely that their
estimations of others’ well-being would be deflated as well. This
may contribute to their underestimation of the well-being of pa-
tients. Still, although significant, the recall difference effect was
not large. Most of the discrepancy appears to be due to healthy
people simply not recognizing how positive the mood of hemodi-
alysis patients can be.
There are undoubtedly some circumstances to which people
cannot adapt (see Frederick & Loewenstein, 1999, for a discussion
of what conditions people do and do not adapt to), but people seem
to overestimate the range of circumstances falling into this cate-
gory. For most of us, it would take a lot more than we think to
make us permanently miserable. The current study provides what
is, to date, the most convincing demonstration of this fact. Healthy
people expect hemodialysis to lead to a much more miserable life
than it actually does. But this misperception will be a difficult one
to correct. Even hemodialysis patients who have themselves ex-
perienced adaptation seem not to appreciate the extent of their own
adaptation. Getting others to appreciate it will surely be more
difficult.
Concluding Comments
Ignorance of adaptation can have negative consequences for
decision making. It can cause individuals to opt for unnecessarily
risky surgeries and policymakers to invest in programs that have a
minimal impact on people’s well-being, possibly at the expense of
programs that really do prevent misery. This is not to say that
research and treatment of kidney disease should not continue to be
priorities. Indeed, hemodialysis treatments keep kidney patients
alive. But in making difficult policy decisions, consideration of the
moods experienced by patients may influence priorities between
serious conditions such as, for example, paraplegia and depression.
Further investigation of the relationship between mood and
retrospective reports of well-being is warranted. The relationship
seems to differ between different national populations. For exam-
ple, Oishi (2002) found that when American and Asian subjects
had similar levels of mood, the Americans tended to recall more
positive levels of mood than did the Asians (see also Riis,
8
RIIS ET AL.
Page 7
Schwarz, & Kahneman, 2004). Our finding that mood recall is
more accurate among hemodialysis patients than among healthy
patients is the first evidence that the relationship between mood
and retrospective reports may differ for different health popula-
tions as well. This line of research will be aided by developments
in EMA aimed at reducing subject burden (Kahneman, Krueger,
Schkade, Schwarz, & Stone, 2004) and at improving and validat-
ing its accuracy (Kahneman & Riis, in press).
When evaluating their quality of life and when making decisions
about how to improve that quality of life, people certainly think
about dimensions other than mood (Fredrickson, 2000). Meaning,
achievement, and identity are some of the other things that people
value, and these may be quite independent of mood. Healthy
people may fear illness not just because of its influence on mood
but because of its influence on these other dimensions. But insofar
as mood is an important dimension of quality of life, healthy
people’s apparent exaggeration of the influence of illness on mood
will lead to incorrect perceptions of how illness will influence
quality of life.
References
Ariely, D., & Carmon, Z. (2002). Summary assessment of experiences: The
whole is different from the sum of its parts. In G. Loewenstein, D. Read,
& R. Baumeister (Eds.), Time and decision: Economic and psycholog-
ical perspectives on intertemporal choice (pp. 323–349). New York:
Russell Sage Foundation.
Baron, J., Asch, D. A., Fagerlin, A., Jepson, C., Loewenstein, G., Riis, J.,
et al. (2003). Effect of assessment method on the discrepancy between
judgments of health disorders people have and do not have: A web study.
Medical Decision Making, 23, 422–434.
Boyd, N., Sutherland, H. J., Heasman, K. Z., Tritcher, D. L., & Cummings,
B. (1990). Whose utilities for decision analysis? Medical Decision
Making, 10, 58–67.
Brickman, P., Coates, D., & Janoff-Bulman, R. (1978). Lottery winners
and accident victims: Is happiness relative? Journal of Personality and
Social Psychology, 36, 917–927.
Buick, D. L., & Petrie, K. J. (2002). “I know just how you feel”: The
validity of healthy women’s perceptions of breast cancer patients receiv-
ing treatment. Journal of Applied Social Psychology, 32, 110–123.
Diener, E., Suh, E. M., Lucas, R. E., & Smith, H. L. (1999). Subjective
well-being: Three decades of progress. Psychological Bulletin, 125,
276–302.
Frederick, S., & Loewenstein, G. (1999). Hedonic adaptation. In D.
Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The founda-
tions of hedonic psychology (pp. 302–329). New York: Russell Sage
Foundation.
Fredrickson, B. L. (2000). Extracting meaning from past affective experi-
ences: The importance of peaks, ends, and specific emotions. Cognition
& Emotion, 14, 577–606.
Gilbert, D. T., Pinel, E. C., Wilson, T. D., Blumberg, S. J., & Wheatley,
T. P. (1998). Immune neglect: A source of durability bias in affective
forecasting. Journal of Personality and Social Psychology, 75, 617–638.
Headey, B., & Wearing, A. (1992). Understanding happiness: A theory of
subjective well-being. Melbourne, Victoria, Australia: Longman
Cheshire.
Kahneman, D. (1999). Objective happiness. In D. Kahneman, E. Diener, &
N. Schwarz (Eds.), Well being: The foundations of hedonic psychology
(pp. 3–25). New York: Russell Sage Foundation.
Kahneman, D. (2000). Evaluation by moments: Past and future. In D.
Kahneman & A. Tversky (Eds.), Choices, values, and frames (pp.
293–308). New York: Cambridge University Press and the Russell Sage
Foundation.
Kahneman, D., Fredrickson, B. L., Schreiber, C. A., & Redelmeier, D. A.
(1993). When more pain is preferred to less: Adding a better end.
Psychological Science, 4, 401–415.
Kahneman, D., Krueger, A. B., Schkade, D. A., Schwarz, N., & Stone,
A. A. (2004, December 3). A survey method for characterizing daily life
experience: The day reconstruction method. Science, 306, 1776–1780.
Kahneman, D., & Riis, J. (in press). Living, and thinking about it: Two
perspectives on life. In F. Huppert, B. Keverne, & N. Baylis (Eds.), The
science of well-being. Oxford, England: Oxford University Press.
Lacey, H. P., Fagerlin, A., Loewenstein, G., Smith, D., Riis, J., & Ubel,
P. A. (2004). Unbelievably happy? Exploring whether scale recalibra-
tion accounts for the happiness gap. Manuscript submitted for publica-
tion.
Lykken, D., & Tellegen, A. (1996). Happiness is a stochastic phenomenon.
Psychological Science, 7, 186–189.
Oishi, S. (2002). The experiencing and remembering of well-being: A
cross-cultural analysis. Personality and Social Psychology Bulletin, 28,
1398–1406.
Riis, J., Schwarz, N., & Kahneman, D. (2004). Some international differ-
ences in self-reported health are reality free: Implications for happiness.
Manuscript in preparation.
Robinson, M. D., & Clore, G. L. (2002). Belief and feeling: Evidence for
an accessibility model of emotional self-report. Psychological Bulletin,
128, 934–960.
Sackett, D. J., & Torrance, G. W. (1978). The utility of different health
states as perceived by the general public. Journal of Chronic Diseases,
31, 697–704.
Schkade, D. A., & Kahneman, D. (1998). Does living in California make
people happy? A focusing illusion in judgments of life satisfaction.
Psychological Science, 9, 340–346.
Sieff, E. M., Dawes, R. M., & Loewenstein, G. (1999). Anticipated versus
actual reaction to HIV test results. American Journal of Psychology, 112,
297–311.
Stone, A. A., Shiffman, S. S., & DeVries, M. W. (1999). Ecological
momentary assessment. In D. Kahneman, E. Diener, & N. Schwarz
(Eds.), Well being: The foundations of hedonic psychology (pp. 26–39).
New York: Russell Sage Foundation.
Thomas, D. L., & Diener, E. (1990). Memory accuracy in the recall of
emotions. Journal of Personality and Social Psychology, 59, 291–297.
Ubel, P. A., Loewenstein, G., Hershey, J., Baron, J., Mohr, T., Asch, D., &
Jepson, C. (2001). Do nonpatients underestimate the quality of life
associated with chronic health states because of a focusing illusion?
Medical Decision Making, 21, 190–199.
Ubel, P. A., Loewenstein, G., & Jepson, C. (2003). Whose quality of life?
A commentary exploring discrepancies between health state evaluations
of patients and the general public. Quality of Life Research, 12, 599–
607.
Wilson, T. D., & Gilbert, D. T. (2003). Affective forecasting. In M. Zanna
(Ed.), Advances in experimental social psychology (pp. 345–411). New
York: Elsevier.
Wilson, T. D., Wheatley, T., Meyers, J. M., Gilbert, D. T., & Axsom, D.
(2000). Focalism: A source of durability bias in affective forecasting.
Journal of Personality and Social Psychology, 78, 821–836.
Received December 18, 2002
Revision received October 27, 2004
Accepted October 27, 2004
9
IGNORANCE OF HEDONIC ADAPTATION