Professional Documents
Culture Documents
Postdecision Changes in The Desirability of Alternatives
Postdecision Changes in The Desirability of Alternatives
ALTERNATIVES1
JACK W. BREHM"
University of Minnesota
re-evaluate the alternatives in order to reduce not significant, while the same difference in the
the dissonance. Examination of Corrected Info condition (.68) is significant at the 5 per
Change in Dissonance scores in Table 1 reveals cent level. However, the second-order differi-
a significant reduction of dissonance in all but ence for the Info and No Info conditions is not
the Low Diss-Info condition. It may also be statistically significant. Thus, the expectation
noted that reduction of dissonance is accom- that there would be a greater difference in re-
plished both by raising the desirability of the duction of dissonance between High and Low
chosen alternative and by lowering the desira- Diss conditions in the Info condition is sup-
bility of the unchosen alternative. ported only by trends in the data.
According to Hypothesis 2, the magnitude The effect of ownership. Previous studies have
of the dissonance and consequent pressure to found that when a person is given an object he
reduce it is greater the more closely the alter- tends subsequently to see it as more desirable.
natives approach equal desirability. Thus the This may be called the effect of ownership. The
reduction of dissonance should be greater in present results may perhaps be regarded as
the High than in the Low Diss conditions. simply a reflection of this phenomenon. The
Data in Table 1 further indicate that in the No Corrected Rating Changes for the unchosen
Info condition the Corrected Change in Dis- alternative in each experimental condition pro-
sonance for High Diss (—.92) shows greater vide pertinent data. It is clear that changes in
reduction than that for Low Diss (—.62), desirability of the unchosen alternative are due
though the difference is not statistically sig- to choice rather than ownership. In Table 1 it
nificant. In the Info condition the Corrected may be seen that the Corrected Rating
Change score for High Diss (—.79) shows Changes of the unchosen object in the Low
greater reduction than that for Low (—.11) Diss conditions are for No Info, — .24 and for
and is significant at the 5 per cent level of con- Info, .00, neither of which is a significant
fidence. Thus, the data clearly support Hy- change. For the High Diss conditions these
pothesis 2. figures are —.66 and —.41, respectively, and
The effect of new cognitive elements. According are statistically significant. It is therefore evir
to Hypothesis 3, exposing a person to new dent that desirability ratings of the unchosen
relevant cognitive elements, at least some of alternative not only show a decrease, but also
which are consonant, facilitates the reduction reflect the difference between High and Low
of dissonance. The magnitude of the reduction Diss conditions in the pressure to reduce dist-
should therefore be greater in the Info than in sonance. At least this part of the present re^
the No Info condition. suits is thus not attributable to ownership.
It will be noted from Table 1 that the Cor- It still might be true, however, that part or
rected Change in Dissonance scores show no all of the gain in desirability of the chosen alter-
greater reduction of dissonance in the Info than native is due to ownership rather than choice.
in the No Info conditions. In fact, the amount To check this possibility, some 5s were simply
in High Diss-No Info is greater than that in given an object that corresponded in initial
High Diss-Info, and similarly, that in Low rating to the chosen alternative of the choice
Diss-No Info is greater than that in Low conditions. If the effect of ownership were to
Diss-Info, though neither of these differences account for any part of the gain in desirability
is significant. The expectation that there would of the chosen alternative, then this gift object
be greater reduction of dissonance in the Info should also increase in desirability. The row
than in the No Info condition is, therefore, not labeled "Gift" in Table 1 presents the mean
supported. changes in desirability ratings for this condi-
The second expectation concerning the effect tion. It is dear from the Corrected Rating
of new cognitive elements was that the differ- Change of .00 that none of the gain in desira-
ence in reduction of dissonance between High bility of the chosen alternative may be attrib-
and Low Diss conditions would be greater in uted to the effect of ownership.
the Info than in the No Info condition. It will Changes in desirability and avoidance of con-
be observed in Table 1 that the difference in flict. A plausible alternative interpretation of
Corrected Change in Dissonance between High the various desirability changes so far reported
and Low Diss in the No Info condition (.30) is should be considered. According to this view
388 JACK W. BREHM
a person tends to avoid conflict as unpleasant. TABLE 2
Thus, if equally desirable alternatives are of- MEAN RATING CHANGES (IGNORING REGRESSION) FOH
fered, conflict is created, leading to attempts to ALTERNATIVES WITH COGNITIVE OVERLAP
reduce it. It is clear that the conflict could be No Info Info
reduced by changing the desirability of the
alternatives in order to make them less equal Low Diss
in desirability. One would expect the chosen N 8 11
alternative to increase and the unchosen alter- Chosen .37 -.14
native to decrease in desirability. Furthermore, Unchosen 1.26 .79
Change in dissonancef + .89 + .93
these changes would be proportional to the High Diss
amount of conflict, which in turn would vary N 7 4
with the equality of desirability of the alter- Chosen .59 -.70
natives. Consequently, one would expect Unchosen .00 -.25
Change in dissonance -.59 + .45
greater changes in the High Diss conditions
than in the Low Diss conditions. The predic- t A minus sign indicates decrease in dissonance.
tions are so far consistent with the obtained
results. be compared with those from the regular ex-
However, there is an interesting distinction perimental choices, which had less cognitive
between "the consequences of avoidance of overlap. But first it will be pertinent to ex-
conflict" and "reduction of dissonance." The amine data relevant to the amount of conflict
amount of dissonance is a direct function of the experienced by persons having these two types
proportion of relevant elements which are dis- of choice.
sonant. It follows that the amount of disso- All 5s were asked, after the experiment was
nance is limited by the proportion of relevant completed, if there was any conflict in making
elements which are common to the alterna- the choice, and if so, how much. Their re-
tives. If all relevant elements are contained by sponses were categorized by the experimenter
both alternatives, no dissonance is created by as "none," "little," "moderate," or "high."'
a choice between them. For example, choosing To obtain category frequencies large enough
between identical automobiles would create for a chi-square test, the categories were re-
little or no dissonance. Thus it may be said duced to two: "none" and "some" reported
that, other things being equal, the greater the conflict. It was then found that of all subjects
amount of overlap of cognitive elements, the who had a high dissonance choice, 63 per cent
less is the resulting dissonance. But conflict of those choosing between alternatives without
arises from an inability to determine which cognitive overlap and 88 per cent of those
alternative to choose, i.e., from approximately choosing between alternatives with cognitive
equal tendencies to choose both alternatives. overlap, reported "some" conflict. This differ-
Increasing the similarity of the choice objects ence is significant at the 6 per cent level by chi
will not necessarily increase a person's ability square. Of those who had a low dissonance
to choose between them. While conflict remains choice, 47 per cent of those choosing between
high, dissonance created by making the choice alternatives without overlap, and 25 per cent
would be relatively low. of those choosing between alternatives with
In the course of the experiment 30 5s were overlap, report "some" conflict. The latter
inadvertently given a choice between relatively relationship is in the opposite direction to the
similar objects, and the data for this type of first but is not statistically significant. It may
choice were analyzed separately." The choices be concluded that those who choose between
designated as having large overlap of cognitive nearly equally desirable objects with large
elements were the following: a choice between overlap of cognitive elements experience more
any two of the coffee-maker, the toaster, and conflict than those choosing between dissimilar
the grill; or a choice between the art book and 7
The classification by E, who knew whether or not
the silk-screen reproduction. The changes in the alternatives were nearly equal in desirability, may
desirability from this type of choice may then well be biased in respect to the High versus Low Diss
manipulation. However, £ did not expect to separate
6
The author is indebted to Dr. Festinger, who first out choices involving cognitive overlap so there is no
noted this difference. reason to suspect a bias in regard to this variable.
POSTDECISION CHANGES IN DESIRABILITY OF ALTERNATIVES 389
objects. If the changes in desirability in this native will be dissonant. Thus a state of dis-
experiment are due to avoidance of conflict, sonance and pressure to reduce it are created.
then, one would expect greater changes where Examination of the possible ways in which dis-
the alternatives are similar. On the other hand, sonance may be reduced was limited in the
if the changes are due to a pressure to reduce present study to tendencies to re-evaluate the
dissonance, one would expect them to be less choice alternatives after the decision.
where the alternatives are similar. Female 5s were asked to rate each of eight
In Table 2 may be found the uncorrected articles on desirability, choose between two of
mean changes hi desirability ratings for the them, and then rate each of the articles again.
chosen and unchosen alternatives which had In addition, some 5s were exposed to a mixture
cognitive overlap. These may be compared of good and bad information about the choice
with the uncorrected rating changes in Table 1. alternatives after the choice was made.
Corrections for regression are not necessary The results supported the prediction that
since the important comparisons are between choosing between alternatives would create
the two High Diss conditions and between the dissonance and attempts to reduce it by mak-
two Low Diss conditions. It will be seen that ing the chosen alternative more desirable and
the changes hi the direction of reducing disso- the unchosen alternative less desirable. A sec-
nance (or avoiding conflict) are small or non- ond prediction, that dissonance and conse-
existent. For all those with a high dissonance quent attempts to reduce it would be greater
choice, 27 per cent of those choosing between the more nearly the choice alternatives ap-
similar and 59 per cent of those choosing proached equality, also received support. The
between dissimilar alternatives show rating third prediction, that exposure to new informa-
changes in the direction of reducing dissonance tion containing at least some consonant ele-
or avoiding conflict. This difference is signifi- ments would facilitate reducing dissonance,
cant, by an exact test, at the 11 per cent level. did not receive clear support. A control condi-
There is thus some evidence that choices be- tion ruled out the possibility that the obtained
tween alternatives with overlapping cognitive increase in desirability of the chosen alterna-
elements create less tendency to change the tive was due to ownership. Finally, some of the
desirability of the alternatives in the expected data consistent with "dissonance theory" were
direction. Since it has already been seen that found not to be consistent with traditional
such choices are accompanied by more, rather "conflict theory."
than less, conflict, it appears that these changes
in desirability reflect reduction of dissonance REFERENCES
rather than avoidance of conflict. 1. ADAMS, D. K. Conflict and integration. /. i'ers.,
1954, 22, 548-556.
SUMMARY AND CONCLUSIONS 2. FESTINGER, L. The relation between cognition and
action. Paper read at Symposium on Cogniiion,
The present experiment was designed to Boulder, Colo., May, 1955.
examine some of the consequences of making a 3. LEWIN, K. Field theory in social science. New York:
decision. Specific predictions about the conse- Harper, 1951.
quences were based on a theory by Festinger. 4. MARTIN, A. H. An experimental study of the factors
According to this theory, when a person and types of voluntary choice. Arch. Psychol.,
1922, 22, No. 51.
chooses one of two alternatives, all of the items
of information which favor the unchosen alter- Received July 15,1955.