You are on page 1of 14

J Happiness Stud (2012) 13:985–998

DOI 10.1007/s10902-011-9302-6

RESEARCH PAPER

Overall Hedonic Evaluations and Evaluation of Specific


Moments from Past Relationships and High School Days

Simon Kemp • Zhe Chen

Published online: 12 October 2011


 Springer Science+Business Media B.V. 2011

Abstract Two studies were conducted in which student participants gave overall hedonic
evaluations of past relationships or their high school days, and, additionally, gave similar
evaluations of the best, most memorable, worst, initial, and final moments from them. They
also evaluated a typical moment and the percentages of good and bad times in these
experiences. Such results from some individuals were given to other participants, who were
also asked to make overall evaluations. Overall, the overall evaluations of the original
participant related weakly rather than strongly to the evaluations of the specific moments.
Generally, the other participants’ overall evaluations converged on those of the original
recalling participants, but their overall evaluations of the past relationships were more
positive than either their own initial evaluations or those of the original, recalling
respondents. In sum, overall evaluations of these long, diverse experiences were not well
predicted from valuations of the specific moments we investigated. Hedonic evaluations of
past experiences may be influenced by appraisals of their success.

Keywords Hedonic evaluation  Peak-end  Autobiographical memory

People often look back at an extended period of their life and characterise it as happy or
unhappy, awful or marvellous. So, for example, Churchill (1930, p. 13) remarked of his
preparatory school (St James’s): ‘‘How I hated this school, and what a life of anxiety I lived
there for more than 2 years … I counted the days and the hours to the end of every term,
when I should return home from this hateful servitude.’’ This paper investigates some
aspects of the way people make summary evaluations of two kinds of past experiences,
past relationships and school days.
There is now a reasonable body of research that has focussed on the way in which
summary evaluations of brief past experiences are arrived at. Redelmeier and Kahneman

S. Kemp (&)  Z. Chen (&)


Psychology Department, University of Canterbury, Private Bag 4800, Christchurch 8140, New Zealand
e-mail: Simon.Kemp@canterbury.ac.nz
Z. Chen
e-mail: Zhe.Chen@canterbury.ac.nz

123
986 S. Kemp, Z. Chen

(1996) asked patients undergoing a painful medical treatment to give moment by moment
accounts of the pain they were experiencing and after the procedure to rate the total amount
of pain they had experienced. They found the total unrelated to the duration of the pro-
cedure (which varied somewhat between participants) and reasonably well-predicted from
the most painful moment (peak) and the final moment (end).
The findings that an overall hedonic evaluation is reasonably well-predicted from the
average of the peak and end moments of the experience—the peak-end rule—and is largely
independent of the duration of the experience have frequently been reported since (e.g.
Ariely and Loewenstein 2000; Ariely and Zauberman 2000; Fredrickson and Kahneman
1993; Schreiber and Kahneman 2000). Moreover, the findings have been recorded for
pleasant as well as unpleasant experiences (Fredrickson and Kahneman 1993). On the other
hand, a number of complicating factors have emerged. For example, whether or not the
experience becomes more or less pleasant during its course is sometimes important (Ariely
and Loewenstein 2000). Overall, however, Frederickson and Kahneman’s (1993) claim
that memories for events appear to consist of a few stills rather than a continuous record
has held up well.
Most of the research has been carried out in laboratory or clinical settings and has
obtained continual ratings from participants at the time of the recalling experiences.
Unsurprisingly, then, most research has dealt with rather brief events. However, it is also
obviously important to find out how people arrive at summaries of longer and more
emotionally varied events, even though here the opportunity to obtain moment by moment
accounts of people’s moods or emotions might be more limited. Kemp et al. (2008) asked
students who were going on holiday for periods up to 2 weeks to send daily text messages
rating the happiness they had felt in the best, worst, present, and average moments of the
previous day. On returning, they were asked to rate the overall happiness of the experience
as well as the time course of their hedonic experience. Overall, the results indicated that
little actual hedonic experience was remembered and that holiday duration was neglected
in the evaluations, but the summary evaluations did not appear to be based exclusively or
principally on the peak and end moments.
Kemp et al’s (2008) study sacrificed moment by moment detail about hedonic values for
examination of longer lasting and more varied experiences than those studied in the earlier
experimental studies. In the two studies reported below, we studied still longer and more
varied experiences but obtained no data at all during the period of the actual experiences.
Instead data about the overall evaluation of the experiences was obtained in the same
questionnaire as data about specific events from the experiences. Thus, all the data were
retrospective. Although obviously this methodology has drawbacks compared to obtaining
hedonic valuations as they actually occur, we did not see how such methods (e.g. Brewer
1988; Burt 1992; Thompson 1982) could be applied to the kinds of experience we consider
here.
In general, if we ask someone to rate the hedonic value of a past moment, we would
expect the rating to be different to what might have been obtained at the time. Research
shows that generally both positive and negative affect tend to fade with time but normally
negative affect fades more rapidly (e.g. Walker et al. 1997). Peak and end moments seem
to be subject to a similar process of hedonic fading to other moments (Kemp et al. 2008).
However, remembered affect has also been found to be influenced by the participant’s
mood, goals and situation at the time of recall (e.g. Burt et al. 1998; Conway 2005;
Teasdale et al. 1980; Walker et al. 2003). Thus, if a participant gives a rating now of how
she felt during a moment 6 months in the past, it is not possible to make a good estimate of
how she actually might have felt then by extrapolating backwards.

123
Hedonic Evaluations of Past Experiences 987

On the other hand, if the overall hedonic evaluation of an experience were based on a
small sample of moments from that experience, it is not clear a priori whether the hedonic
evaluation should be based on the remembered affect from these moments or the affect
experienced originally. Either seems possible. For example, people might keep a kind of
running average during the experience, and this average is then recalled later as a summary
of the experience. Note that the running average might still only sample a few moments,
and that it, too, might be subject to fading or influence from the participant’s mood or
situation at the time of recall. Alternatively, the overall evaluation may be made at the time
of recall, in which case it would be based on the sample of moments from the experience
that is available, and the overall evaluation is likely to incorporate changes in the
remembered affect of these moments. Effectively, the methodology we use here focusses
on this second possibility.
Both overall hedonic evaluations and recollections of the remembered affect from
specific moments of that experience are likely to be at least partly reconstructed (Rubin
1995). For example, someone remembering a romantic dinner with a partner in the early
stages of a relationship might recall herself as very happy then because it is the kind of
experience that is generally thought to be happy. Fredrickson and Kahneman (1993) found
that watching waves break on a beach was more pleasant at the time than watching a puppy
play with a flower, but the latter was rated more highly in retrospect. This result may reflect
participants thinking that the idea of the puppy was pleasanter. Overall, Kemp et al’s
(2008) participants tended to recall their holidays as happier than they were at the time.
Homesick students probably recall their life at home as happier than it probably was (Burt
et al. 1998).
Cojuharenco and Ryvkin (2007)and Kemp et al. (2008) both found that the different
summary measures of the moments (e.g. the average moment evaluation and the average of
the peak and end moments) are often highly correlated, and thus it is not easy to tell which
are the critical moments. A further complication is that for an emotionally diverse expe-
rience it is not clear whether the peak moment should be taken as the best or the worst
moment. Thus, in the studies reported below we collected information about a number of
plausible specific moments.
We also distinguish between the hedonic value attached to specific moments, and the
hedonic values which would appear to be based on summing over a variety of moments,
such as the percentage of good times someone recalls from an experience or the hedonic
value of a typical moment of the experience. One might expect that the typical moment
would be formed from a number of similar moments, and that the memory for it would be
repisodic, to use Neisser’s (1981) term. Such summative variables have low claims to
causal status in determining the overall evaluations. They might themselves be measures of
the value of the past relationship, or even determined from the overall evaluation. We
included them largely as yardsticks for the predictive power of the specific moments and
other variables.
In the following studies we asked samples to evaluate either a past relationship that had
lasted at least 2 months or their schooldays. We chose these experiences because they are
relatively long, because they are likely to have been emotionally diverse, because they are
very common, and because they are usually thought to be important. Moreover, the
experiences may produce different patterns of evaluation over time. Generally, we
expected the end of a relationship to be recalled as unhappy, and the end of school as
pleasant.
As is clear from the preceding paragraphs, we cannot presume that people’s memory for
the specific moments of either experience is always accurate or causally prior to the overall

123
988 S. Kemp, Z. Chen

evaluations. Two features of our methods were introduced to get extra information about
the causal role of the remembered moments.
Firstly, participants were asked to provide two overall evaluations, one preceding recall
of the specific moments and one afterwards. We asked the participants to rate not only how
they had felt in the specific moments but also to answer various questions about these
moments using measures developed by Rubin et al. (2003). These measures (for example,
asking how clearly a participant can see the moment in her mind) were originally devised
mainly to assess vividness of recall but here they were intended principally to focus the
participant on the remembered moment. (The recalled vividness of an experience does not
provide a good measure of the accuracy of someone’s memory. E.g., Sheen et al. 2001;
Kemp and Burt 2006). We expected that requiring this focus would produce systematic
changes from the first to the second evaluation.
Secondly, we took a subsample of the completed questionnaires, edited them to remove
the overall evaluations and personal identifiers, and gave one each to a sample of wholly
new ‘‘other’’ participants, who were then asked to make overall hedonic evaluations of the
recalling participants’ experience both before they were given the information about the
recalling participants’ specific moments and afterwards. (For similar studies, see, e.g.,
Cojuharenco 2007; Kemp et al. 2003.) The critical information obtained from the others’
results comes from comparing their initial overall evaluations (which should reflect only
general knowledge) with the final ones. If the other participants’ estimates move towards
the evaluations of the recalling participants after they have read through the specific
moment and other data, such a result would indicate that the overall evaluations of the
recalling participants are also based on evaluations of these data.
To summarise, we examined overall evaluations of two kinds of lengthy past experi-
ences, romantic relationships and high school days. The critical questions that we asked
were:
1. How well do the overall evaluations relate to the evaluations of specific moments from
those past experiences, and which of the moments are of most importance?
2. Are the specific moments better determinants of the overall evaluations than more
summative measures such as the estimated percentage of good and bad times in the
overall experience or the hedonic value of a typical moment?
3. Are the overall evaluations much affected by the duration of the experience, by the
time elapsed since the experience, and by the participant’s life satisfaction at the time
of recall?
4. Do the other participants produce evaluations that are like those of the recalling
participants after they have seen the responses made by a recalling participant?
5. Are the two kinds of past experiences evaluated in the same way?

1 Study 1 (Past Relationships)

1.1 Method

1.1.1 Participants

A total of 200 participants was recruited from around the University of Canterbury, many
as they were awaiting participation in an unrelated psychological experiment. These par-
ticipants are referred to below as recalling participants. The key criterion for participation

123
Hedonic Evaluations of Past Experiences 989

was having ‘‘ended a romantic relationship with someone in the last 5 years, if that
relationship lasted for at least 2 months, and if that relationship ended at least 1 month
ago’’. Participants who had ended more than one such relationship were asked to answer
the questionnaire for the most recent one. Participants completed the questionnaire in
privacy and were asked to place it in an envelope and seal it before giving it to a
researcher. No great difficulty was encountered in finding the participants, but initial
analyses showed a number of them (N = 28) featured major inconsistencies or missing
data in their responses, and their data were excluded from further analyses. Criteria for
exclusion were, for example, if the worst moment was rated more positively or the best
moment more negatively than the average moment. Of the remaining 172 recalling par-
ticipants, 52 were men, and 122 were women. The median age was 22 years, with a range
from 18 to 43.
A further 43 ‘‘other’’ participants were later recruited. They were matched for gender
with the first 43 (36 were female) recalling participants with consistent data. The median
age of these participants was 23 years with a range from 18 to 58. They are referred to
below as others. Each participant, in both Studies 1 and 2, was paid NZ$5 for taking part in
the study.

1.1.2 Questionnaire

The recalling participant questionnaire began with a request for summary details about the
participant’s current life satisfaction, age and sex, and the relationship. Then the partici-
pants were asked to remember a number of different moments from the relationship.
Finally, a few more summary details were requested. At the outset participants were
requested not to turn back to previous pages of the questionnaire.
The initial summary questions asked the recalling participant’s rating of ‘‘how you feel
about your life as a whole these days’’, answered on a scale from 0 (completely awful) to
10 (completely marvellous), and a rating of how they felt about the last ended relationship
on the same scale. They were also asked for the month and year in which the relationship
ended and its duration in years and months.
In the next six pages the participants were asked to remember (in unvarying order and
one page per moment) the initial moment of the relationship, the worst moment, a typical
moment, the end moment, the most memorable moment, and the best moment. Eight
statements were presented for each moment. In succession, participants read:
As I remember the moment, I am travelling back to the time when it happened.
As I remember the moment, I am travelling back to the location where it happened.
As I remember the moment, I feel as though I am reliving it.
As I remember the moment, I can hear it in my mind.
As I remember the moment, I can see it in my mind.
As I remember the moment, I can feel now the emotions that I felt then.
The emotions that I felt at the moment were very intense.
The [initial etc] moment was ….
The first seven statements were adapted from those used by Rubin et al. (2003). In the
present study they served mostly to focus the participants on the moments, although they
also measure the vividness of their recall of them. All statements were rated on a scale
from 0 to 10. The first seven statements used endpoints of ‘‘not at all’’ (0) and ‘‘definitely’’
(10). The last statement used endpoints of ‘‘completely awful’’ (0) and ‘‘completely
marvellous’’ (10), as in the summary ratings.

123
990 S. Kemp, Z. Chen

On the last page of the questionnaire, recalling participants were asked to estimate the
percentage of good times, neutral, and bad times of the relationship, with a total of 100%.
Finally, the participants rerated their feelings about the ended relationship using the same
scale as at the outset.
Other participants also completed a questionnaire. This began by telling them that their
basic task was to assess a past relationship that another person had had. Others made an
initial assessment of the recalling participant’s evaluation of his or her past relationship,
worded: ‘‘On a scale of 0 to 10’’, where 0 indicates ‘‘completely awful’’ and 10 indicates
‘‘completely marvellous’’, how do you think an average person would rate a romantic
relationship that occurred in the last 5 years, that lasted for at least 2 months, and that
ended at least 1 month ago?’’
Other participants then read through the questionnaire answered by one of the recalling
participants. Note, however, firstly, that the recalling participant’s two relationship eval-
uations were removed and, secondly, that copies only of the original questionnaires were
used (to avoid the remote possibility of a particular number circling style being recogni-
sed). Finally, others valued the recalling participant’s relationship with the wording: ‘‘Now
that you have finished reading the questionnaire, on a scale of 0–10, please rate how that
person would have rated his or her last relationship.’’

1.2 Results

1.2.1 Initial and Second Overall Relationship Evaluations

Table 1 shows average values of the two relationship evaluations by the recalling and other
respondents. There was no difference between the initial and second rating of how
recalling participants felt about the relationship (t(171) = 1.05, ns), thus the process of
recollecting the different moments produced no change to the overall evaluation, and
Pearson correlations were high between the first and second relationship evaluation
(r = .77, p \ .05). The others’ initial evaluation of the recalling participants’ relationships
was similar to that of the participants themselves (t(42) = 1.19, ns). (Note that these
estimates were made before reading any information, and thus presumably reflect only the
other participants’ intuitions of how people generally evaluate past relationships). How-
ever, after reading the moment evaluations and other information, this evaluation rose
significantly and substantially both in comparison with the others’ own first estimates
(t(42) = 6.73, p \ .001) and in comparison with the recalling participants’ second esti-
mates (t(42) = 4.16, p \ .001). The others’ second relationship evaluation correlated 0.49
and 0.42 (both p \ .05) with the matched recalling participants’ first and second overall
relationship evaluation. (The others’ initial evaluations did not correlate significantly with
either of these measures.)

Table 1 Average (and SD) of the relationship evaluations for both recalling participants (both the complete
sample and only those matched to the others) and other participants from Study 1
Recalling (All) Recalling (Matched) Others

Relationship evaluation (First) 5.3 (2.1) 4.8 (2.3) 4.3 (1.6)


Relationship evaluation (Second) 5.2 (2.5) 4.9 (2.6) 6.5 (1.7)
Evaluations on a scale from 0 (completely awful) to 10 (completely marvellous)

123
Hedonic Evaluations of Past Experiences 991

1.2.2 Average Values and Correlations with Overall Evaluations

Table 2 shows the mean life satisfaction value, the average time elapsed since the rela-
tionship end, the average duration of the relationship and average evaluations of the six
moments, as well as the average percentages of time that the recalling respondents felt had
been good and bad in the relationship. Clearly there are large, if unsurprising, differences
in how participants felt about the different moments (F(5, 855) = 547.7, p \ .001). The
table also shows Pearson correlations between the two relationship evaluations and all of
these variables. Evaluations of the relationship were most strongly related to the evaluation
of the typical moment, and the percentages of good and bad times recalled. Unsurprisingly,
those remembering a higher percentage of good times and a lower percentage of bad times
evaluated the relationship more positively. Interestingly, given the low overall evaluations,
the average percentage of good times was considerably higher than that of the bad times.
Note that, contrary to the peak-end hypothesis, the end moment did not correlate signifi-
cantly with the overall evaluations, although both the worst and best moments did.
The others’ initial evaluation of the relationship was not significantly correlated with
any other recalling participant measures, but their second (informed) rated relationship
evaluations did correlate with a number of the measures provided by the recalling par-
ticipants. (See the last column of Table 2) . The correlations show that the others did base
their relationship evaluations on the information they were given about the recalling
participants. Note, however, that the pattern of correlation is somewhat different to that
shown by the recalling participants. Significance testing of the differences between
recalling participants’ and others’ correlations indicated that others’ evaluations were more
strongly related to the percentages of good and bad times than the recalling participants’
first or second evaluations were, but no other differences were significant (two-tailed,
p \ .05).

Table 2 Average and standard deviation (SD) of the evaluations for the six recollected moments and
percentage of good and bad times from Study 1
Mean SD Initial Second Others
r r r

Life satisfaction 7.3 1.7 .11 .09 .23


Months from relationship end 16 15 .02 .02 -.23
Relationship duration (months) 16 15 .10 .10 -.20
Moment evaluations
Initial 7.5 1.9 .17* .20* .31*
Worst 1.5 1.5 .24* .24* .28
Final 2.2 2.2 .13 .14 .24
Memorable 7.9 2.5 .22* .30* .42*
Best 8.7 1.3 .17* .28* .47*
Typical 6.7 2.0 .45* .53* .38*
% Good time 58 23 .48* .51* .75*
% Bad time 24 19 -.40* -.46* -.83*
Shown are Pearson correlations (r) of these measures with the two recalling participants’ and the others’
second overall relationship evaluations
Moment valuations and life satisfaction on scales from 0 (completely awful) to 10 (completely marvellous)
* p \ .05, two-tailed

123
992 S. Kemp, Z. Chen

Table 3 Results of ordinary least squares regression (b and R2) predicting initial and second overall
relationship evaluation with summative independent variables excluded (Model 1) or included (Model 2)
Model 1 Model 2

Initial b Second b Initial b Second b

Life satisfaction .07 .04 .07 .05


Time from relationship end .05 .06 .05 .06
Relationship duration .13 .10 .15* .12
Initial moment .13 .10 .01 -.02
Worst moment .24* .25* .11 .10
Final moment .04 .05 .04 .05
Most memorable moment .14 .17 -.03 -.01
Best moment .04 .16 -.07 .04
Typical moment .29* .33*
% Good time .27* .22*
% Bad time -.08 -.12
Overall R2 .14* .20* .33* .40*
Change in R2 (Model 1 to Model 2) .19* .20*
n = 171
* p \ .05, two-tailed

1.2.3 Regression Results

Table 3 shows the results of regressing the recalling participants’ overall evaluations on
the independent variables. Results of two different regressions are shown for each overall
evaluation. In Model 1, the independent variables are current life satisfaction, time elapsed
since the end of the relationship, the duration of the relationship, and the evaluations of 5
of the 6 moments. Model 2 adds the percentage of good and bad times and the typical
moment evaluation.
A number of results are evident from the table. First, the second evaluations are better
predicted than the initial ones, indicating that the questionnaire itself influenced the way
that the recalling respondents evaluated their past experiences. Second, when the sum-
mative variables are included, the percentage of good times recalled and the valuation of
the typical moment are strong predictors of the overall evaluations. Note, too, that R2 is
markedly higher when these variables are included. Third, when the summative variables
are excluded, the best predictor is the worst moment. Fourth, in no regression is the value
of the final moment a significant predictor.1

2 Study 2 (High School Days)

This study was generally similar to Study 1 but the participants answered questions relating
to high school.

1
We also conducted a number of analyses with measures of vividness of the specific moments but these
were found to have little relationship to the overall evaluations and we omit their description. Similar
analyses were conducted with similar results in Study 2.

123
Hedonic Evaluations of Past Experiences 993

2.1 Method

Two hundred participants were recruited in a similar way to Study 1. To be eligible, par-
ticipants had to have attended high school, and to have finally left at least 1 month before the
study. One hundred and ninety-two participants passed the consistency check on their results.
The resulting sample contained 55 male and 137 female participants, and had a median age of
21 years with a range from 17 to 43 years. The questionnaire was essentially identical to that
used for the relationship study except that participants were asked the questions about their
high school experience. Layout and order of questions were unchanged. If they had attended
more than one high school, they were asked to recall the most recent one.
Subsequently, 45 other participants were recruited who were matched for gender to the
first 45 recalling participants with consistent data. Their median age was 21 years with a
range from 19 to 55. Their questionnaire was organised in exactly the same way as the
others’ questionnaire from Study 1 except that information about the recalling participants’
high school experience was provided, and corresponding judgements called for.

2.2 Results

2.2.1 Initial and Second Overall High School Evaluations

The analysis follows the same sequence as that for the results section of Study 1. Table 4
shows the average overall evaluations of the high school experience. There was no sig-
nificant difference between the initial and second evaluations by the recalling respondents
(t(191) = 0.51, ns, and there was a high correlation between the two (r = .84, p \ .05).
Others’ second evaluation, after reading the recalling participant’s questionnaire, increased
significantly over their first (t(44) = 2.58, p \ .05). Interestingly, both evaluations proved
to be significantly lower than the corresponding evaluations from the recalling participants
(initial, t(44) = 2.93, p \ .01; second, t(44) = 2.56, p \ .05). The others’ second high
school evaluation correlated 0.51 and 0.66 (both p \ .05) with the matched recalling
participants’ first and second overall high school evaluation; the others’ initial evaluations
did not correlate significantly with either of the recalling participants measures. Signifi-
cance testing of the differences between recalling participants’ and others’ correlations
showed that others’ evaluations were more strongly related to the valuations of the final
and most memorable moment than those of the recalling participants, but other differences
were not statistically significant (two-tailed, p \ .05).

2.2.2 Average Values and Correlations with Overall Evaluations

Table 5 shows large differences in how the participants felt about the six different
moments (F(5, 995) = 345.0, p \ .001). Again, the average percentage of good times
recalled was considerably larger than the percentage of bad times. Evaluations of high

Table 4 Average (and SD) of the high school experience evaluations for the recalling participants (both the
complete sample and only those matched to the others) and other participants from Study 2
Recalling (all) Recalling (matched) Others

High school (first) 7.3 (2.0) 7.3 (1.7) 6.4 (1.2)


High school (second) 7.2 (2.3) 7.6 (2.1) 7.0 (1.4)
Evaluations on a scale from 0 (completely awful) to 10 (completely marvellous)

123
994 S. Kemp, Z. Chen

Table 5 Mean and standard deviation (SD) of the evaluations for the six recollected moments from high
school and percentage of good and bad times from Study 2
Mean SD Initial Second Others
r r r

Life satisfaction 7.3 1.7 .22* .23* .32*


Months from high school end 31 44 -.14 -.11 -.38*
High school duration (months) 60 13 .10 .01 .28
Moment evaluations
Initial 5.5 2.1 .24* .26* .37*
Worst 2.2 1.7 .23* .28* .06
Final 7.2 2.2 .12 .12 .45*
Memorable 8.0 2.3 .17* .21* .53*
Best 8.8 1.3 .16* .20* .44*
Typical 6.8 1.8 .37* .43* .53*
% Good time 57 23 .63* .68* .66*
% Bad time 20 17 -.67* -.69* -.63*
Shown are Pearson correlations (r) of these measures with the two recalling participants’ and the others’
second overall high school evaluations
Evaluations and life satisfaction on scales from 0 (completely awful) to 10 (completely marvellous)
* p \ .05, two-tailed

school were significantly correlated with evaluations of the most memorable moment, best
moment, initial moment, worst moment, and current life satisfaction. However, the highest
correlations are between the overall evaluations and the valuations of the typical moment,
and estimated percentages of good and bad times. The third column of Table 5 shows
correlations with other data obtained from the recalling participants. Once again, the
number of significant correlations shows that the others were considerably influenced by
the data they saw, and there were no significant correlations between the others’ initial high
school evaluation and any of the recalling participants’ data.

2.2.3 Regression Results

Table 6 shows the results of regressing overall high school evaluations on the independent
variables. Again, the summative variables proved to be the best predictors, and, indeed,
when they are included in the regressions they are the only statistically significant pre-
dictors. When they are excluded, the valuation of the worst moment is the best single
predictor, but the level of current life satisfaction is a significant predictor of the initial
overall evaluation and both life satisfaction and the value of the best moment are signif-
icant predictors of the second evaluation. Once again, the second overall evaluations are
slightly better predicted by the independent variables than the initial ones.

3 Discussion

The two studies produced a complex variety of results relating to how people evaluate
extended past experiences. The discussion is ordered according to the five critical questions
which concluded the introduction, followed by brief consideration of some general issues.

123
Hedonic Evaluations of Past Experiences 995

Table 6 Results of ordinary least squares regression (b and R2) predicting initial and second overall high
school evaluation with summative independent variables excluded (Model 1) and included (Model 2)
Model 1 Model 2

Initial b Second b Initial b Second b

Life satisfaction .17* .15* .09 .07


Time from relationship end -.11 -.10 -.01 .00
Relationship duration .09 .00 -.06 -.04
Initial moment .12 .10 .06 .04
Worst moment .24* .30* .05 .10
Final moment .02 .00 -.07 -.08
Most memorable moment .01 .03 -.07 -.05
Best moment .11 .21* -.01 .09
Typical moment .10 .14*
% Good time .26* .31*
% Bad time -.45* -.41*
Overall R2 .17* .21* .54* .61*
Change in R2 (Model 1 to Model 2) .37* .40*
n = 191
* p \ .05, two-tailed

3.1 How Do Specific Moments Relate to Overall Evaluations?

Both studies revealed significant correlations between the overall evaluations and the
hedonic values that were assigned to some of the specific moments. Thus, overall evalu-
ations of even lengthy and diverse experiences such as these may be influenced by specific
moments, particularly the moment recalled as the worst. The comparisons between the
regression results for the initial and second evaluations also indicate that this influence is
somewhat strengthened if the participant focuses on the specific moments before being
asked to make the overall evaluation. On the other hand, other results from the regression
analyses, where for both studies no specific moment has a significant independent effect in
Model 2 and the variance accounted for in Model 1 is rather low, indicate that this
influence is limited. Certainly, one strong version of the peak-end rule—that the overall
evaluation is largely determined by the remembered peak (whether good or bad) and the
remembered end moment—fails for both of these experiences. It is noteworthy that the
hedonic value of the final moment—the ‘‘end’’ part of the peak-end rule—had particularly
low correlations with the overall evaluations in both studies, a result which is similar to
that reported by Kemp et al. (2008). This convergence of results suggests that the end of a
lengthy experience is often simply not important in its overall evaluation.

3.2 Summative Variables and Specific Moments

Both the correlational and the regression analyses indicate that the summative variables—
the value of the typical moment and the estimates of percentages of good and bad times—
were better predictors of the overall evaluations than the specific moments or the other
variables examined. As remarked earlier, this result should not be taken as a strong
indication that summative variables have a real causal role in determining the overall

123
996 S. Kemp, Z. Chen

evaluations, but it does underscore the conclusion that the specific moments have only a
weak influence.

3.3 Time and Satisfaction Relationships with Overall Evaluations

Neither the time that elapsed since the end of the experience nor the duration of the
experience had much effect on the evaluations of either experience, although it is inter-
esting that the duration of the relationship related significantly but weakly to its initial but
not its second overall evaluation. Thus, the previous reported independence of overall
evaluation from the duration of the experience (e.g. Fredrickson and Kahneman 1993;
Kemp et al. 2008; Redelmeier and Kahneman 1996) is not completely replicated here.
Current life satisfaction related moderately strongly to both the initial and second high
school evaluations.

3.4 Do the Others’ Evaluations Resemble Those of the Recalling Participants?

As Tables 2 and 5 show, the other participants were clearly influenced by their reading of
the original questionnaire responses. Indeed, often the others’ overall evaluations were
more highly correlated with this information than those of the recalling participant. Thus, it
is unsurprising to find that in both studies the second overall evaluations by the others
correlated quite highly with the overall evaluations of their matched recalling participants.
Moreover, in Study 2 the average overall others’ evaluations moved closer to the recalling
respondents’ high school evaluations after they had read the information. On the other
hand, in Study 1 precisely the opposite result was obtained: After obtaining the infor-
mation, the others’ evaluations, which were initially quite close to those of the recalling
participants, were considerably more positive than those of the original participants. An
implication of this surprising result is that whatever determines the low evaluation of past
relationship experiences is not readily apparent in the information that we requested from
the recalling participants and the others read.

3.5 Are Evaluations Similar for Past Relationships and High School Days?

Comparison of the two studies indicates both similarities and differences in the way that
past relationships and high school days are evaluated. A comparison of Tables 2 and 5
shows that, with the exception of the final moment, the hedonic values of the specific
moments were quite similar. More surprising, perhaps, is the similarity of the typical
moment and the percentages of good and bad time in each. Of the specific moments, the
hedonic value of the worst moment correlates best with the overall evaluations for both
types of experience. The regression results show similar consequences of focussing on the
specific moments (the R2 change from initial to second evaluation), a similar dominance of
the summative variables as predictors, and, perhaps most importantly, when the summative
variables are excluded, the remaining variables do about equally well (or poorly) in pre-
dicting the overall evaluations of past relationships and high school experiences.
One difference comes from the relationship of the overall evaluations to current life
satisfaction. In brief, life satisfaction correlates with the evaluation of high school but not
the evaluation of past relationships. Why this should be so is not clear, but one possibility
is that the high school experience might be seen as contributing to one’s present success
and thus satisfaction with one’s life.

123
Hedonic Evaluations of Past Experiences 997

Probably the most important difference in the recall of the two experiences, however, is
simply that past relationships are evaluated less positively than one’s high school days. As
we have just remarked, this difference probably reflects some difference between the two
experiences that we did not measure. There are a number of possibilities but one likely
source of difference in evaluation is that the recalling participants saw their past rela-
tionship—all of which were over at the time of the study—as basically a story of failure.
On the other hand, those who recalled their high school days could reasonably look back on
a story of success: (Remember that in both studies all of our participants were university
students). This and the difference with respect to life satisfaction underline a point made in
the introduction—that evaluation is determined partly by the goals and situation of the
person making it, as a good deal of previous research indicates (e.g. Conway 2005;
Teasdale et al. 1980; Walker et al. 2003).

3.6 General Considerations

The research reported here shows that the hedonic evaluation of relatively lengthy expe-
riences such as relationships or high school days can be weakly but not strongly predicted
from people’s evaluations of a few remembered moments from that past experience. It is
possible that the prediction would be stronger if actual evaluations of such moments were
available from the time as they were experienced, but it is difficult to see how such
information could be gained. One might also question whether in fact a very different
outcome could be expected. The research suggesting the power of the peak-end rule was
conducted mostly with relatively brief experiences that were relatively homogenous:
painful but not overlong medical procedures (Redelmeier and Kahneman 1996) or brief
film clips (Fredrickson and Kahneman 1993). One would not really expect exactly the
same rules to apply to the longer and much more diverse experiences that we investigated.
On the other hand, it is quite possible that the deeper principles found from this earlier
research do apply to the evaluations of long, diverse experiences. It is possible, extrapo-
lating a little from Frederickson and Kahneman’s idea of evaluations being based on a few
stills rather than a continuous record, that evaluations of long experiences are based on a
limited sample of events. Perhaps, too, the sample of events might change depending on
the context of reviewing the experience. For example, it is possible that in the future one or
two of the recalling participants from Study 1 might resume the apparently now finished
past relationships. Might this not change their evaluation of the previous time together?
Similarly, a different evaluation of high school days might be obtained from people who
did not go on to university. It would also be interesting to see how middle-aged or elderly
people remembered their first relationship or school days.

References

Ariely, D., & Loewenstein, G. (2000). When does duration matter in judgment and decision making?
Journal of Experimental Psychology: General, 129, 508–523.
Ariely, D., & Zauberman, G. (2000). On the making of an experience: The effects of breaking and com-
bining experiences on their overall evaluation. Journal of Behavioral Decision Making, 13, 219–232.
Brewer, W. F. (1988). Memory for randomly sampled autobiographical events. In U. Neisser & E. Winograd
(Eds.), Remembering reconsidered: Ecological and traditional approaches to memory (pp. 21–90).
Cambridge: Cambridge University Press.
Burt, C. D. B. (1992). Reconstruction of the duration of autobiographical events. Memory & Cognition, 20,
124–132.

123
998 S. Kemp, Z. Chen

Burt, C. D. B., Strongman, K. T., & Constanzo, C. L. (1998). Memorial distortions and homesickness
following relocation. Australian Journal of Psychology, 50, 106–113.
Churchill, W. S. (1930). My early life: A roving commission. New York: Charles Scribner.
Cojuharenco, I. (2007). Lay intuitions about overall evaluations of experiences. Judgment and Decision
Making, 2, 40–47.
Cojuharenco, I., & Ryvkin, D. (2007). Evaluations of experiences lived across time: Reconciling empirical
findings with folk beliefs (Unpublished manuscript).
Conway, M. A. (2005). Memory and the self. Journal of Memory and Language, 53, 594–628.
Fredrickson, B., & Kahneman, D. (1993). Duration neglect in retrospective evaluations of affective epi-
sodes. Journal of Personality and Social Psychology, 65, 45–55.
Kemp, S., & Burt, C. D. B. (2006). Memories of uncertain origin: Dreamt or real? Memory, 14, 87–93.
Kemp, S., Burt, C. D. B., & Furneaux, L. (2008). A test of the peak-end rule with extended autobiographical
events. Memory & Cognition, 36, 132–139.
Kemp, S., Burt, C. D. B., & Sheen, M. (2003). Remembering dreamt and actual experiences. Applied
Cognitive Psychology, 17, 577–591.
Neisser, U. (1981). John Dean’s memory: A case study. Cognition, 9, 102–115.
Redelmeier, D. A., & Kahneman, D. (1996). Patients’ memories of painful medical treatments: Real-time
and retrospective evaluations of two minimally invasive procedures. Pain, 66, 3–8.
Rubin, D. C. (Ed.). (1995). Remembering our past: Studies in autobiographical memory. Cambridge:
Cambridge University Press.
Rubin, D. C., Schrauf, R. W., & Greenberg, D. L. (2003). Remembering, reliving, and believing autobio-
graphical memories: Inter- and intra-individual analyses. Memory & Cognition, 31, 887–901.
Schreiber, C. A., & Kahneman, D. (2000). Determinants of the remembered utility of aversive sounds.
Journal of Experimental Psychology: General, 129, 27–42.
Sheen, M., Kemp, S., & Rubin, D. (2001). Twins dispute memory ownership: A new false memory
phenomenon. Memory & Cognition, 29, 779–788.
Teasdale, J. D., Taylor, R., & Fogarty, S. J. (1980). Effects of induced elation-depression on the accessibility
of memories of happy and unhappy experiences. Behavior Research and Therapy, 18, 339–446.
Thompson, C. P. (1982). Memory for unique personal events: The roommate study. Memory & Cognition,
10, 324–332.
Walker, W. R., Skowronski, J. J., & Thompson, C. P. (2003). Life is pleasant—and memory helps keep it
that way. Review of General Psychology, 9, 203–210.
Walker, W. R., Vogl, R. J., & Thompson, C. P. (1997). Autobiographical memory: Unpleasantness fades
faster than pleasantness over time. Applied Cognitive Psychology, 11, 399–413.

123

You might also like