One-reason decision making in risky choice? A closer look at the priority heuristic

Judgment and Decision Making, Vol. 3, No. 6, August 2008, pp. 457–462
One-reason decision making in risky choice? A closer look at the
priority heuristic
Benjamin E. Hilbig?
University of Mannheim
Abstract
Although many models for risky choices between gambles assume that information is somehow integrated, the re-
cently proposed priority heuristic (PH) claims that choices are based on one piece of information only. That is, although
the current reason for a choice according to the PH can vary, all other reasons are claimed to be ignored. However,
the choices predicted by the PH and other pieces of information are often confounded, thus rendering critical tests of
whether decisions are actually based on one reason only, impossible. The current study aims to remedy this problem by
manipulating the number of reasons additionally in line with the choice implied by the PH. The results show that par-
ticipants’ choices and decision times depend heavily on the number of reasons in line with the PH — thus contradicting
the notion of non-compensatory, one-reason decision making.
Keywords: priority heuristic, prospect theory, non-compensatory strategy, one-reason decision making, fast and frugal
heuristics, risky choice.
1 Introduction
tioned and substantial debates about the properties of the
PH (Birnbaum, 2008a; Brandstätter, Gigerenzer, & Her-
The adaptive toolbox metaphor, put forward by Gigeren-
twig, 2008; Johnson, Schulte-Mecklenbeck, & Willem-
zer and co-workers (Gigerenzer, 2001; Gigerenzer, Todd,
sen, 2008; Rieger & Wang, 2008) and, more generally,
& The ABC Research Group, 1999), implies that deci-
the plausibility of the fast-and-frugal-heuristics approach
sion makers possess and use a collection of simple rules
(Dougherty, Franco-Watkins, & Thomas, 2008; Gigeren-
of thumb — the so-called fast-and-frugal heuristics — to
zer, Hoffrage, & Goldstein, 2008) have arisen.
achieve very good results with very little effort. Since it
One serious caveat to studies investigating the PH lies
was originally formulated, the toolbox has been rapidly
in the selection of gambles used: The choice predicted by
growing and new heuristics are introduced almost regu-
the PH and the gamble favored by other pieces of infor-
larly. Despite Bröder’s (2003) criticism that “inventing
mation are often confounded. Thus, adherence rates to
more and more new heuristics may soon become futile
the PH, or modal choices as studied by Brandstätter et al.
if they are not seriously tested empirically” (p. 622) the
(2006) might be biased measures of whether participants’
adaptive toolbox was recently extended to preferential de-
decisions are truly based on one reason only, as claimed
cisions by means of the priority heuristic (PH, Brand-
by the PH. The current study aims to remedy this prob-
stätter, Gigerenzer, & Hertwig, 2006) — a simple lexi-
lem. First, the choice rule of the PH will be introduced
cographic rule for choices between gambles. Although
along with a description of the problem of confounded
the idea of adaptive decision making in choice is, in it-
information in gamble-pairs. Then, an experiment will
self, not novel (e.g. Payne, Bettman, & Johnson, 1993),
be reported in which the number of pieces of informa-
the PH represents a new development in this area.
tion confounded with the choice predicted by the PH was
Brandstätter et al. (2006) concluded that the PH out-
systematically varied to test whether participants actually
performs both normative and other heuristic models in
base their choices on one piece of information only.
predicting participants choices. Also, they claim that
it represents a process model of choice, describing the
1.1 The priority heuristic
sequence of steps taken by a decision maker’s cogni-
tive apparatus. Both claims have been seriously ques-
In the simple case of non-negative two-outcome gambles
comprising a minimum gain, a maximum gain, and ac-
?I thank Andreas Glöckner, Sebastian A. Markett, Jonathan Baron,
cording probabilities, the PH claims that the following
and two anonymous reviewers for insightful comments and helpful sug-
steps are taken by a decision maker: First, an aspira-
gestions. Address: Benjamin E. Hilbig, Center for Doctoral Studies
in Social and Behavioral Sciences, University of Mannheim, D-68131
tion level is computed which is 1/10 of the largest max-
Mannheim, Germany. Email: hilbig@psychologie.uni-mannheim.de.
imum gain (rounded to the nearest prominent number,
457

---

Judgment and Decision Making, Vol. 3, No. 6, August 2008
One-reason decision making in risky choice
458
Brandstätter et al., 2006). If the difference between the
PH would predict choice of gamble A since it comprises
minimum gains exceeds this aspiration level, the gamble
the more attractive probabilities (and since the minimum
with the larger minimum gain is chosen; thus, informa-
gains differ by less than 1/10 of the largest maximum
tion search is stopped after one piece of information has
gain). However, the minimum gains, the maximum gains,
been examined (henceforth PH1 case) and all probabil-
and the expected values (which are 2,850 and 1,430, re-
ities and maximum gains are ignored. If this is not the
spectively) all imply the same decision as the PH. Thus,
case, the probabilities (of the minimum gains) are consid-
choice of gamble A cannot imply that (only) the proba-
ered: should these differ by at least .10 the gamble with
bilities were considered in the decision process. By con-
the smaller probability (for the minimum gain) is cho-
trast, one can construct gamble-pairs for which the choice
sen. So, search is terminated after the second reason has
implied by the PH is not in line with any other piece of
been examined (thus labeled PH2 case) and a choice is
information. Deciding between gambles C (4,000; .40;
made ignoring all gains. Finally, if the probabilities yield
1,200; .60) and D (3,150; .50; 950; .50) is such an ex-
no such difference, the maximum gains are considered
ample: The PH would predict choice of gamble D in line
(PH3 case) and the gamble comprising the larger max-
with the probabilities (for the same reason as in the for-
imum gain is chosen. No trade-offs are made and thus
mer example). However, the minimum gains, the maxi-
there is no integration of information in the process.
mum gains, and the expected values (which are 2,640 and
Although the PH can be easily extended to multiple-
2,050, respectively) all imply choice of gamble C.
outcome and negative-outcome gambles, its niche is lim-
Since the PH claims that choices are always based on
ited by the following bounding conditions: The ex-
one piece of information only, the number of other rea-
pected values of the gambles may differ by a maximal
sons in line with the choice implied by the PH should be
ratio of 2:1, thus rendering choices adequately dif?cult
inconsequential. Stated bluntly, participants’ adherence
(Brandstätter et al., 2006, 2008). Moreover, cases of
to the PH should not depend on the number of additional
strict dominance are excluded (Brandstätter et al., 2006),
reasons in line the PH — since these additional reasons
even though such cases could be expected to be handled
are claimed to be ignored. So, in the above examples,
smoothly by any heuristic (Rieger & Wang, 2008). Taken
choice of gamble A over B should be just as likely as
together, these bounding conditions limit the PH’s ap-
choice of gamble D over C. By contrast, any strategy
plicability to less than 50% of a randomly generated set
which integrates different pieces of information would
of gamble-pairs as shown through simulation (Birnbaum,
predict that choices in line with the PH should increase
2008a). However, this limitation by no means rules out
with the number of reasons additionally in line with the
that the process predictions of the PH are adequate in
PH’s prediction.
those cases to which it is argued to apply.
Moreover, according to the PH, participants’ decision
As Brandstätter et al. (2006) demonstrate, the PH is
times should also not be affected by the number of rea-
successful at predicting majority choices in different sets
sons in line with the PH: in the above examples, decision
of gambles. Speci?cally, the authors show that it can pre-
makers should ?rst consider the minimum gains, then
dict modal choices better than quite a number of other
move on to the probabilities (since the minimum gains
models — including the most recognized: cumulative
do not differ suf?ciently in both cases), and upon doing
prospect theory (Tversky & Kahneman, 1992). However,
so stop search and make a choice. Thus, they should take
these results have been challenged and others have shown
equally long to choose A over B and D over C. Alter-
that cumulative prospect theory and the transfer of atten-
natively, one might claim that the choice between gam-
tion exchange model (e.g., Birnbaum, 2004) are likely
bles A and B should afford less time than choice between
to outperform the PH when more diagnostic gambles are
gambles C and D since in the former case all reasons im-
used and when competing models are allowed appropri-
ply the same decision (gamble A) whereas in the latter
ate parameter ?tting (Birnbaum, 2008a, 2008b; Glöckner
case the probabilities (speaking for gamble D) contra-
& Betsch, in press). In sum, whenever the PH and com-
dict the choice implied by all other pieces of information
peting models made different predictions, choices were
(gamble C).
mostly in line with the latter.
However, there is an inherent problem in some investi-
gations of whether participants adhere to the PH: Often,
2 Experiment
multiple pieces of information imply the same choice as
the PH. As a consequence, it is not always possible to
The predictions described above were tested in an ex-
conclude which piece of information — or which combi-
periment in which the number of reasons additionally in
nation of the latter — led to a given choice. The following
line with the choice implied by the PH was manipulated
example illustrates this problem: Considering gambles A
within participants. In the different gamble-pairs studied,
(4,000; .50; 1,300; .50) and B (3,900; .35; 1000; .65) the
which comprised the minimum gains, maximum gains,

---

Judgment and Decision Making, Vol. 3, No. 6, August 2008
One-reason decision making in risky choice
459
probabilities, and expected values as pieces of informa-
tion, either none, one, two, or three (all) reasons were
Table 1: Mean proportions of choices in line with the
additionally in line with the PH. That is, there were four
PH (standard deviations in parenthesis) for the four levels
levels of additional reasons in line with the PH which var-
of the number of additional reasons in line with the PH.
ied within participants.
t-statistic and Cohen’s d for the difference from chance
level (.50) for each of these means.
2.1 Materials and procedure
Number of
additional
Mean
Cohen’s
Gambles were randomly generated with maximum gains
t(40)
reasons in line (SD)
d
ranging from 1,000 to 5,000 (in steps of 50), minimum
with PH
gains taking values between 0 and 1,500 (also in steps of
50), and probabilities varying from 0 to 1 (in steps of .05).
None
.19 (.15)
–13.4*
2.1
Next, gambles were randomly paired and all pairs com-
One
.39 (.13)
–5.4*
0.8
prising dominance or ratios of expected values greater
than 2:1 were excluded. Consequently, all gamble-pairs
Two
.96 (.11)
26.2*
4.1
were within the PH’s niche as proposed by Brandstätter
Three (all)
.96 (.10)
28.4*
4.4
et al. (2006). Finally, 36 gamble-pairs were randomly
selected: 9 gamble-pairs for each of the four levels of
Note. * p < .001
reasons in line with the PH (none, one, two, and three,
respectively). Although there was no speci?c hypothesis
concerning the different PH cases (PH1, PH2 and PH3,
ni?cantly above chance level, t(40) = 13.7, p < .001, Co-
as described above) the number of these cases was held
hen’s d = 2.14. However, adherence rates differed sub-
constant across the four levels of reasons in line with the
stantially depending on the number of reasons in line with
PH. Thus, a four (levels of reason in line with the PH)
the PH: Table 1 shows the proportion of choices in line
by three (PH case — PH1, PH2, or PH3) matrix resulted,
with the PH separately for the four levels of the number
with a total of three gamble-pairs per cell. The gambles
of additional reasons in line with the PH.
used are listed in the accompanying data ?le.1
As can be seen, participants adhered to the PH signif-
The experiment was administered by means of a web-
icantly below chance level whenever none or one addi-
based questionnaire. First, participants were familiarized
tional reason was in line with the choice implied by the
with the structure of the gambles used (all two-outcome,
PH. By contrast, whenever two or three additional rea-
non-negative) and were instructed that their task was to
sons implied the same decision as the PH, choices were
choose which of two gambles they would prefer to play.
largely in line with its predictions. All effect sizes can
Then, after an exemplary choice task, all 36 choices be-
be considered to be large (Cohen, 1988). A repeated-
tween gambles were presented separately, one after the
measures ANOVA with adherence to the PH as dependent
other, in a predetermined randomized order which was
variable and the number of reasons in line with the PH as
the same for all participants. For each gamble the maxi-
independent variable con?rmed the differences between
mum gains, probabilities of maximum gains, minimum
these four levels with F(2.2, 86.9)2 = 391.6, p < .001, ?p²
gains, and probabilities of minimum gains were pre-
= .91. Thus, participants’ choices depended largely on
sented. The expected values were not presented since
the number of additional reasons in favor of the choice
this would most likely bias choices. Participants were
implied by the PH.3
instructed to respond speedily but to take the time they
Next, decision times were analyzed. For each partici-
needed to make their choices. They were also told that
pant the median decision time (excluding the ?rst of all
there were no correct or false responses.
36 decisions) was computed separately for the four lev-
41 participants (37 female) were recruited from an
els of the number of additional reasons in line with the
undergraduate-course in psychology at the University of
PH. As depicted in Figure 1, participants had the longest
Mannheim. Participants were aged 19 to 68 (M = 22.8
decision times (M = 9900ms, SE = 930ms) when one ad-
years, SD = 8.5) and received partial course credit for
2All degrees of freedom are Huynh-Feldt corrected.
their participation.
3The repeated-measures analysis could be questioned because the
same random order was used for all participants. Effects such as as-
similation and accommodation between adjacent choices might be con-
2.2 Results
founded with number of reasons. An additional analysis, using only
the means (across participants) of the 36 conditions yielded the same
Averaging across all cases, the proportion of choices in
result, however. In particular, with agreement proportion as the depen-
line with the PH was M = 63% (SD = 6%) which is sig-
dent variable, and predictors of number of reasons and order in which
each choice was presented (1–36), order had no signi?cant effect, and
1See http://journal.sjdm.org/vol.3.6.htm.
the coef?cient for number of reasons was .29 (t(31) = 8.89, p < .001).

---

Judgment and Decision Making, Vol. 3, No. 6, August 2008
One-reason decision making in risky choice
460
 
 
 ¡¢ ¡
¤¥¦ §¤
 §¤
§¥¥¥ ¨
 §¦
¤¢¨ ¥§¨
£
£
£
©
 ¢ ¥
 ¤ © 
   ©
¡ ¨¥¤¤ ¤
'
( %%%
£
£
©
 
 ©¨ ¡
¥¢§ ¢
¢ ¦
¤§ ¥¥¦§ ¤
¤ 
¡§  §©
£
£
£
Table 2: Additional priority heuristic models, order in
©
 ©© ©
¥¤© ¥©
¤ ¨
 ¦  ¡§¦ ¤
£
£
which pieces of information are sampled, and number
b
of participants (proportion in parenthesis) for whom each
 
¦    
 ¡¥¤¥
¤¥¢
¨ ¨¤ ¡ 

£
£
'
% %%%
ST

¦¦ ¥¤
 ¤¨ ¤
¦©¢
©¦¡© ¨¢
V

model ?tted best.
£
£
 
`
¥  §¡
 ¡¥¤¥
¢¤
¤¥¢¦ ¢¤
a
£
£
#
 
" "
¤¢ ¤©
¦ ¤¢ 
¨¢¤
©¤¤¦ ©¤
 
!
S
£
£
Model
Ordering
N (%) of participants
U
`
Y
U
WX
PH
U
V
&
%% %
i*
MinG, Prob., MaxG
1 (2.4)
R
ST
PHii
MinG, MaxG, Prob.
9 (22)
PHiii
Prob., MinG, MaxG
19 (46.3)
$ %% %
PHiv
Prob., MaxG, MinG
0
@ @A
) 0
0
1
2
3 1
2
4 5
4 67
22
8
9
PHv
MaxG, Prob., MinG
10 (24.4)
D
E
F FG HG
@
G
@ G
G H
)
0
0
0
Q
2
7
9
19
7
2
9
I
1
I
1
1
2
5
6
P
B C
PHvi
MaxG, MinG, Prob.
2 (4.9)
Figure 1: Median decision times across levels of the num-
Note. MinG = minimum gains, MaxG = maximum
ber of reasons in line with the PH (error bars represent
gains, Prob. = probabilities. * original PH model.
one standard error).
ditional reason was in line with the PH. Shorter decision
2.3 Additional analyses
times (M = 9100ms, SE = 790ms) were observed when
Although the analyses reported clearly reveal an in?u-
no other piece of information was in line with the PH.
ence of the number of reasons in line with the PH, a plau-
The shortest decision times occurred whenever two (M =
sible caveat needs to be addressed: possibly, individu-
8100ms, SE = 660ms) or all (M = 7700ms, SE = 370ms)
als differ in the heuristics they use. More speci?cally, it
other reasons predicted the same choice as the PH.
may be that all participants use some non-compensatory
A repeated-measures ANOVA with decision times as
priority-heuristic but that these heuristics differ in the or-
dependent variable and the number of reasons in line with
der in which pieces of information are considered. Thus,
the PH as independent variable con?rmed the difference
what may look like compensatory decision-making on
between these levels with F(1.6, 63.8) = 6.1, p < .001,
the aggregate level, may turn out to be a blend of dif-
?p² = .132. In sum, decision times were longest when-
ferent non-compensatory process at the individual level.
ever the different pieces of information con?icted most
To address this, all possible priority heuristics (with all
strongly (two vs. two); by contrast, decision times were
possible orderings) were modeled and the model ?tting
substantially shorter with an increasing number of rea-
each participants’ choices best5 was used for this partici-
sons implying the same choice.
pant (and will thus be denoted PH
Additionally, the analysis of decision times was re-
BEST in what follows).
Table 2 depicts these models along with the number of
peated including only those cases in which participants
participants for whom each model ?tted best. As can be
adhered to the PH. In line with the previous analysis, me-
seen, most participants’ choices were explained best by
dian decision times decreased from M = 11086ms (SE
a priority heuristic (PH
= 1370ms) for no additional reason in line with the PH
iii) which considered differences
in probabilities ?rst, followed by differences in minimum
to M = 7491ms (SE = 374ms) when all reasons sup-
gains.
ported the PH (with M = 9619ms, SE = 1095ms and M
= 7958ms, SE = 755ms, for one and two additional rea-
Next, the analyses concerning the impact of the num-
sons, respectively). This effect of the number of reasons
ber of additional reasons were repeated using the individ-
in line with the PH was again con?rmed by a repeated-
ually best-?tting priority heuristic for each participant.
measures ANOVA, F(1.5, 64.2) = 6, p = .005, ?p² = .151.
the relevant cases were selected, and the mean of these cases across par-
Stated simply, decisions in line with the PH afforded less
ticipants was computed for each case. This mean was then regressed on
time the more additional pieces of information supported
order and number of reasons. In this case, order had a strong effect, but
the PH-consistent choice.4
the effects of order and of number of reasons were both highly signi?-
cant: coef?cients of –.01 (for a change of one position in the ordering)
4Again, the repeated-measures analysis could be distorted by the use
and –.08 (per reason), t(32) = 3.73 and 3.35, p < .001 and p = .002,
of the same random order for all participants. Further, because of the re-
respectively.
striction to cases in which the PH model agreed with the response, very
5For differences in maximum gains the usual PH aspiration level
few participants contributed useful data to some of the choices. To in-
(of 1/10 of the larger maximum gain rounded to the nearest prominent
sure the best estimate, decision times were ?rst transformed logarithmi-
number) was used. If two models ?tted a participant’s choices equally
cally, which rendered their distribution approximately normal, and then
well the model from the better ?tting model category (minimum gains
they were centered on each participant’s log mean decision time. Then
?rst, probabilities ?rst, or maximum gains ?rst) was selected.

---

Judgment and Decision Making, Vol. 3, No. 6, August 2008
One-reason decision making in risky choice
461
Concerning choices, a repeated-measures ANOVA with
2006, 2008). The results obtained clearly contradict that
adherence to the PHBEST as dependent variable and the
choices are based on one piece of information only: Ad-
number of reasons in line with the PHBEST as independent
herence to the PH depended substantially on the number
variable revealed a signi?cant and large effect, F(1.8,
of reasons additionally in line with its predictions. When-
70.6) = 194.5, p < .001, ?p² = .829. Speci?cally, choices
ever few (none or one) additional reasons implied the
in line with the PHBEST increased from M = .27 (SE = .03)
same choice as the PH, adherence rates to the heuristic
when no additional reason was in line with the PHBEST to
were vastly below chance level. By contrast, whenever
M = .60 (SE = .02), M = .94 (SE = .01), and M = .96 (SE
two or three (all) additional reasons backed up the PH,
= .02), for one, two, or three additional reasons, respec-
choices were virtually always in line with its predictions.
tively. Thus, choices in line with the best-?tting priority
It can thus be concluded that good predictive performance
heuristic for each participant were again strongly in?u-
of the PH must be attributed to the fact that its prediction
enced by the number of additional reasons in line with
and the choices implied by other pieces of information
the PHBEST which con?icts with the claim of one-reason
are often confounded.
decision making.
The results with respect to decision times also con-
Likewise, median decision times for PHBEST showed
?icted with the predictions derived from the PH: Choices
the exact same pattern as in the previous analyses for
took longest whenever the different pieces of informa-
the original PH model. That is, decision times were M
tion con?icted most (two implying choice of one gam-
= 9371ms (SE = 896ms) for no additional reason in line
ble, the other two implying the opposite). If, by contrast,
with PHBEST , increased to M = 9977 (SE = 940ms) for
all pieces of information favored the same gamble, deci-
one additional reason, and then dropped to M = 8475ms
sions were made notably faster. Such differences, how-
(SE = 652ms) and M = 7673ms (SE = 367ms) for two and
ever, rule out that only one piece of information was de-
three additional reasons, respectively. This effect of the
cisive, as claimed by the PH. Rather, decision times were
number of reasons additionally in line with PHBEST was
well in line with the notion that different pieces of in-
con?rmed by a repeated-measures ANOVA, F(1.7, 68.8)
formation are integrated and that choices become eas-
= 4.9, p = .014, ?p² = .109. Likewise, and in line with the
ier (and thus faster) whenever all reasons favor one op-
analysis of the original PH model, decision times of only
tion. Likewise, decisions in line with the PH were per-
those cases in which participants adhered to the PHBEST
formed with increasing speed when the number of rea-
dropped linearly across the levels of additional reasons in
sons in line with the PH increased. Theoretical expla-
line with the PHBEST, again corroborated by a repeated-
nations outlining a (potentially compensatory) model ac-
measures ANOVA, F(1.9, 64.2) = 5.5, p = .007, ?p² =
counting for the reported patterns, however, must remain
.140. In sum, decision times for PHBEST comprised the
speculative at this point. Most obviously, the degree of
same pattern reported for the original PH model and thus
con?ict between different pieces of information seems to
contradicting a non-compensatory decision process.
feed into longer decision times. This could, for example,
be explained though a process which ?rst determines how
many reasons speak for each gamble before comparing
3 Discussion
options on certain attributes in case of con?ict. However,
future research advisably using process-tracing methods
The recently proposed and controversially debated pri-
(e.g. Johnson et al., 2008) is clearly needed since any
ority heuristic (PH, Brandstätter et al., 2006) represents
explanation herein will be post hoc and go untested.
a non-compensatory lexicographic strategy. As such, it
Additionally, a different best-?tting priority heuristic
claims that choices between gambles are based on the
for each individual was examined owing to the possibil-
consideration of one piece of information only. In the
ity that participants might all use non-compensatory pri-
current study, this notion was tested through disentan-
ority heuristics which differ in the order in which pieces
gling adherence to the priority heuristic from the num-
of information are considered. However, the results with
ber of other reasons implying the same choice as the PH.
respect to both choices and decision times again contra-
Strictly speaking, reasons additionally in line with the PH
dicted the claim of one-reason decision-making, since the
should be inconsequential since all other pieces of infor-
number of additional reasons in line with the prediction
mation are claimed to be ignored. For the same reason,
of the individually best-?tting model had a substantial im-
decision times should also not depend on these additional
pact.
pieces of information.
Note, however, that the analyses using the best-?tting
Both predictions were tested using a set of randomly
priority heuristic for each participant are post-hoc and
generated gambles which were all within the PH’s niche,
thus bear some limitations. Most importantly, they are
that is, excluding cases of dominance and ratios of
no longer based on an equal number of cases for each
expected values greater than 2:1 (Brandstätter et al.,
level of additional reasons within or across participants

---

Judgment and Decision Making, Vol. 3, No. 6, August 2008
One-reason decision making in risky choice
462
and should thus be interpreted with caution. However,
Bröder, A. (2003). Decision making with the “adaptive
the reported results consistently show that even using dif-
toolbox”: In?uence of environmental structure, intelli-
ferent priority heuristics for each individual the claim of
gence, and working memory load. Journal of Exper-
one-reason decision-making must be refuted.
imental Psychology: Learning, Memory, and Cogni-
Finally, it could be argued that the hypothesis and re-
tion, 29(4), 611–625.
sults presented herein are based on a rather literal imple-
Cohen, J. (1988). Statistical power analysis for the be-
mentation of the PH with ?xed aspiration levels (minimal
havioral sciences (2nd ed.). Hillsdale, NJ: Lawrence
differences). Consequently, one may claim that allowing
Earlbaum Associates.
these to vary would strongly increase the ?t of the PH.
Dougherty, M. R., Franco-Watkins, A. M., & Thomas,
Although I readily acknowledge this, it seems unlikely
R. (2008). Psychological plausibility of the theory of
that the effect of the number of additional reasons in line
Probabilistic Mental Models and the Fast and Frugal
with the PH – especially on decision times — would van-
Heuristics. Psychological Review, 115(1), 199–213.
ish if these aspiration levels were allowed to vary. Also,
Gigerenzer, G. (2001). The adaptive toolbox. In G.
letting aspiration levels vary would drastically increase
Gigerenzer & R. Selten (Eds.), Bounded rationality:
the complexity of the PH model which is exactly what its
The adaptive toolbox (pp. 37–50). Cambridge, MA:
proponents are aiming to avoid (Brandstätter et al., 2006,
The MIT Press.
2008).
Gigerenzer, G., Hoffrage, U., & Goldstein, D. G. (2008).
In sum, one of the PH’s central advantages — its for-
Fast and frugal heuristics are plausible models of
mulation as a process model — is turning out to be its
cognition: Reply to Dougherty, Franco-Watkins, and
downfall: different studies using different methods have
Thomas (2008). Psychological Review, 115(1), 230–
consistently shown that choices between gambles are not
237.
based on one piece of information and that the steps
Gigerenzer, G., Todd, P. M., & The ABC Research
claimed by the PH are not likely to be taken by deci-
Group. (1999). Simple heuristics that make us smart.
sion makers (Glöckner & Betsch, in press; Hilbig & Mar-
New York, NY: Oxford University Press.
kett, 2008; Johnson et al., 2008). Moreover, the heuris-
Glöckner, A., & Betsch, T. (in press). Do people make
tic’s predictive power is at least questionable (Birnbaum,
decisions under risk based on ignorance? An empiri-
2008a, 2008b; Glöckner & Betsch, in press) and turns
cal test of the priority heuristic against the cumulative
out to be rather poor whenever the PH is not backed up
prospect theory. Organizational Behavior and Human
by other pieces of information as shown in this article.
Decision Processes.
Hilbig, B. E., & Markett, S. A. (2008). On the priority
of the priority heuristic: critical tests of a fast and fru-
References
gal model for risky choice. Unpublished manuscript
(submitted).
Birnbaum, M. H. (2004). Causes of Allais common
Johnson, E. J., Schulte-Mecklenbeck, M., & Willemsen,
consequence paradoxes: An experimental dissection.
M. C. (2008). Process models deserve process data:
Journal of Mathematical Psychology, 48(2), 87–106.
Comment on Brandstätter, Gigerenzer, and Hertwig
Birnbaum, M. H. (2008a). Evaluation of the priority
(2006). Psychological Review, 115(1), 263–272.
heuristic as a descriptive model of risky decision mak-
Payne, J. W., Bettman, J. R., & Johnson, E. J. (1993). The
ing: Comment on Brandstätter, Gigerenzer, and Her-
adaptive decision maker. New York, NY: Cambridge
twig (2006). Psychological Review, 115(1), 253–260.
University Press.
Birnbaum, M. H. (2008b).
New tests of cumulative
Rieger, M. O., & Wang, M. (2008). What is behind the
prospect theory and the priority heuristic: Probability-
priority heuristic? A mathematical analysis and com-
outcome tradeoff with branch splitting. Judgment and
ment on Brandstätter, Gigerenzer, and Hertwig (2006).
Decision Making, 3(4), 304–316.
Psychological Review, 115(1), 274–280.
Brandstätter, E., Gigerenzer, G., & Hertwig, R. (2006).
Tversky, A., & Kahneman, D. (1992). Advances in
Making choices without trade-offs:
The priority
prospect theory: Cumulative representation of uncer-
heuristic. Psychological Review, 113(2), 409–432.
tainty. Journal of Risk and Uncertainty, 5, 297–323.
Brandstätter, E., Gigerenzer, G., & Hertwig, R. (2008).
Risky choice with heuristics: Reply to Birnbaum
(2008), Johnson, Schulte-Mecklenbeck, and Willem-
sen (2008), and Rieger and Wang (2008). Psychologi-
cal Review, 115(1), 281–289.

---