Joumal of Experimental Psychology: © 2013 American Psychological Association

Learning. Memory, and Cognition 0278-7393/14/$l2.00 DOI: 10.1037/aOO34887

2014, Vol. 40. No. 2. 544-554

Base Rates: Both Neglected and Intuitive

Gordon Pennycook Dries Trippas and Simon J. Handley

University of Waterloo Plymouth University

Valerie A. Thompson
University of Saskatchewan

Base-rate neglect refers to the tendency for people to underweight base-rate probabilities in favor of
diagnostic information. It is commonly held that base-rate neglect occurs because effortful (Type 2)
reasoning is required to process base-rate information, whereas diagnostic information is accessible
to fast, intuitive (Type 1) processing (e.g., Kahneman & Frederick, 2002). To test this account, we
instructed participants to respond to base-rate problems on the basis of "beliefs" or "statistics," both
in free time (Experiments 1 and 3) and under a time limit (Experiment 2). Participants were given
problems with salient stereotypes (e.g., "Jake lives in a beautiful home in a posh suburb") that either
conflicted or coincided with base-rate probabilities (e.g., "Jake was randomly selected from a sample
of 5 doctors and 995 nurses for conflict; 995 doctors and 5 nurses for nonconflict"). If utilizing
base-rates requires Type 2 processing, they should not interfere with the processing of the
presumably faster belief-based judgments, whereas belief-based judgments should always interfere
with statistics judgments. However, base-rates interfered with belief judgments to the same extent
as the stereotypes interfered with statistical judgments, as indexed by increased response time and
decreased confidence for conflict problems relative to nonconflict. These data suggest that base-
rates, while typically underweighted or neglected, do not require Type 2 processing and may, in fact,
be accessible to Type 1 processing.

Keywords: dual-process theories, base-rate neglect, conflict detection, intuition

Supplemental materials: http://dx.doi.org/10.1037/a0034887.supp

Much research has demonstrated that preconceived notions theorists that some degree of slow and deliberative "Type 2"
about representativeness can influence probability judgments (Bar- processing is required for prior probabilities to enter into judg-
bey & Sloman, 2007; Kahneman & Tversky, 1973; Tversky & ment, indicating that a disparity in the ease with which base-rates
Kahneman, 1974). Eor example, given a description of a randomly and stereotypes are processed is the source of base-rate neglect
sampled male U.S. citizen described as being shy and introverted, (e.g., Bonner & Newell, 2010; De Neys & Glumicic, 2008; Eer-
one might judge that he is more likely to be a librarian than a reira, Garcia-Marques, Sherman, & Sherman, 2006; Kahneman &
farmer, even though there are roughly 20 times more farmers than Erederick, 2002). This perspective is theoretically grounded in
librarians in the United States (Kahneman, 2011). These types of dual-process theory, which postulates a fundamental cognitive
"biased" judgments are thought to occur because representative- difference between autonomous "Type 1" processes and working
ness cues an intuitive response that is difficult to override (Kah- memory dependent "Type 2" processes (Evans, 2008; Evans &
neman & Erederick, 2002; Kahneman & Tversky, 1973; Tversky Stanovich, 2013; Kahneman, 2003; Sloman, 1996; Stanovich,
& Kahneman, 1974). It is also frequently assumed by dual process 2004; Thompson, 2013). In this work we adapt an instruction
manipulation from a deductive reasoning paradigm (Handley,
Newstead, & Trippas, 2011) to test this "analytic base-rate" ac-
count of base-rate neglect.
This article was published Online First November 11, 2013.
Gordon Pennycook, Department of Psychology, University of Waterloo,
Waterloo, Ontario, Canada; Dries Trippas and Simon J. Handley, Depart-
Base-Rate Neglect
ment of Psychology, Plymouth University, Plymouth, England; Valerie A. Consider the following example (De Neys & Glumicic, 2008,
Thompson, Department of Psychology, University of Saskatchewan, Sas-
adapted from Kahneman & Tversky, 1973):
katoon, Saskatchewan, Canada.
This research was funded, in part, by a Discovery Grant from the Natural In a study 1,000 people were tested. Among the participants there
Engineering and Research Council of Canada to Valerie Thompson. Gor- were 995 nurses and five doctors. Jake is a randomly chosen partic-
don Pennycook and Dries Trippas contributed equally to this work. ipant of this study.
Correspondence concerning this article should be addressed to Gor-
don Pennycook, Department of Psychology, University of Waterloo, Jake is 34 years old. He lives in a beautiful home in a posh suburb. He
200 University Avenue West, Waterloo, ON, Canada, N2L 3GI. E-mail: is well spoken and very interested in politics. He invests a lot of time
gpennyco @ uwaterloo.ca in his career.
544
 BASE RATES ARE INTUITIVE 545
What is more likely? the opportunity to rethink their answer. This evidence suggests Tl
processes are used to evaluate both base-rates and stereotypes and,
(a) Jake is a nurse.
in cases where they conflict, T2 processing is required to resolve
(b) Jake is a doctor. the conflict in one direction over the other. Thus, the evidence
indicates that both base-rates and stereotypes can be processed
This very basic type of base-rate problem contains two conflict- quickly and that both are also processed analytically, contrary to
ing pieces of information: (a) a base-rate probability that suggests the assumption that belief information is only processed quickly
Jake is likely to be a nurse, and (b) a stereotypical description that via Tl processes and that utilizing base-rates requires slower, T2
suggests Jake is likely to be a doctor. Decades of research has processing.
demonstrated that people have a strong tendency to favor the
diagnostic information over the base-rates, leading to the term
base-rate neglect (Kahneman & Tversky, 1973; Tversky & Kah- Conflict Detection and Base-Rate Neglect
neman, 1974). Further, participants higher in cognitive abilities Researchers have manipulated base-rate problems so that the
(i.e., intelligence, working memory capacity, etc.) and/or more base-rates either conflict or coincide with the diagnostic informa-
disposed to analytic thought are more likely to use base-rate tion (see Appendix) to determine whether base-rates are entirely
information during reasoning (Pennycook, Cheyne, Barr, Koehler, neglected or, rather, if they have subtle effects on behavior when
& Fugelsang, 2013; Pennycook, Cheyne, Seli, Koehler, & Fugel- they conflict with diagnostic information. In a series of experi-
sang, 2012; but see Stanovich & West, 2008). ments, it has been shown that conflict problems elicit longer
It is frequently claimed and broadly accepted that diagnostic response times (De Neys & Glumicic, 2008; Thompson, Prowse
information, such as stereotypes, are preferred over base-rates Tumer, & Pennycook, 2011) and lower confidence (De Neys,
because the two types of information are processed very differ- Comheeke, & Osman, 2011; Thompson et al., 2011) relative to
ently (see Barbey & Sloman, 2007, for a review). Proponents of nonconflict problems. This interference as a result of conflict even
fuzzy-trace theory, for example, contend that base-rate neglect arises when participants give the stereotypical response, which De
occurs because diagnostic information cues intuitive gist-based Neys and colleagues have taken to suggest that humans are able to
associations in memory, whereas base-rates are processed via more at least implicitly recognize when they are being "biased" (see De
analytical verbatim-based reasoning (Reyna, 2004; Reyna & Neys, 2012, for a review).
Brainerd, 2008; Reyna & Mills, 2007; Wolfe & Fisher, 2013). While the research by De Neys and colleagues did not set out to
Such dual-process theories, including more general versions that investigate base-rate neglect per se, their findings were originally
nonetheless presume that base-rates, unlike diagnostic informa- interpreted under the "analytic base-rate" view (De Neys, Com-
tion, require analytic reasoning (e.g., Barbey & Sloman, 2007; De heeke, & Osman, 2011; De Neys & Franssens, 2009; De Neys &
Neys & Glumicic, 2008; Ferreira et al., 2006; Kahneman & Fred- Glumicic, 2008; De Neys, Vartanian, & Goel, 2008; Franssens &
erick, 2002), are supported by the finding that problem structure De Neys, 2009). As discussed recently by De Neys (2012), the
factors that allow base-rates to be more easily processed make issue of whether base-rates are accessible to Tl processing (and
base-rate neglect much less prominent (Barbey & Sloman, 2007; therefore do not require T2 processing) has implications for the
Brainerd, 2007; Reyna & Mills, 2007). For example, Barbey and nature of conflict detection, particularly as it applies specifically to
Sloman (2007) argued that frequency formats make the nested set base-rate neglect. If comprehending base-rates requires T2 pro-
nature of the problem more evident, which induces people to cessing, some level of analytic reasoning is therefore necessary for
substitute associative (Type 1) processing with rule-based (Type 2) conflicts between base-rates and stereotypes to be detected. How-
processing thereby leading to decreased levels of base-rate neglect. ever, participants were able to detect the conflict under a working
More generally, on this view participants tend to neglect or un- memory load, suggesting that conflict detection is highly efficient
derweight the base-rates because humans tend to forego Type 2 and relatively effortless (Franssens & De Neys, 2009), at least
(T2) processing; and those who are more willing and able to think when the base-rates are salient (Pennycook, Fugelsang, & Koehler,
analytically are more likely to override the intuitive stereotypical 2012). On the basis of this (among other studies). De Neys (2012)
response and employ the base-rate information. However, as has rejected the earlier stance that the detected conflict is between
pointed out by De Neys (2007), there is limited evidence that a Tl (stereotype) and a T2 (base-rate) response, and now, consis-
utilizing base-rates actually requires T2 processing. tent with Permycook and Thompson (2012) and Thompson et al.
Moreover, recent research by Pennycook and Thompson (2012) (2011), claims that the conflict is between the outputs of two sets
has called this account into question, suggesting instead that base- of Tl processes (see also De Neys & Bonnefon, 2013). Again,
rates may be evaluated using Type 1 (Tl) processing. Participants given that conflict detection is a general phenomenon that has been
evaluated problems similar to the above example; however, the established using a number of reasoning tasks (such as syllogisms.
base-rates were excluded for half of the participants. The base-rate De Neys & Franssens, 2009; De Neys, Moyens, & Vansteenwe-
problems did not elicit longer response times relative to cases gen, 2010; the conjunction fallacy. De Neys et al, 2011; Villejou-
where base-rates were excluded despite evidence that participants bert, 2009; ratio bias, Bonner & Newell, 2010; and the Cognitive
incorporated them into their judgments. This finding is inconsis- Reflection Test; De Neys, Rossi, & Houdé, 2013), De Neys's
tent with the claim that base-rates require slow, T2 processing. (2012) Tl-Tl conflict stance was not intended as a specific theory
Also, contrary to the assumption that stereotypes are always pro- of base-rate neglect, though it is consistent with the perspective of
cessed in a primarily intuitive manner, participants were just as Pennycook and Thompson (2012; see also Pennycook, Fugelsang,
likely to change their answer in the direction of the stereotype as & Koehler, 2012; Thompson et al., 2011) and inconsistent with the
they were to change it in the direction of the base-rate when given "analytic base-rate" view that is prominent in the field (e.g..
546 PENNYCOOK, TRIPPAS, HANDLEY, AND THOMPSON

Barbey & Sloman, 2007; Eerreira et al., 2006; Kahneman & Experiment 1
Erederick, 2002).
By analogy to Handley et al. (2011), we presented people with
a series of base-rate neglect problems under either "belief or
Research Paradigm "statistics" instructions. Belief instructions required participants to
respond according to their "knowledge of what they think to be
As with base-rate problems, participants are able to reliably
true in the world"; these instructions were designed to cue re-
detect confiict between logical validity and conclusion believabil-
sponding on the basis of the stereotypical description while ignor-
ity inherent in relatively simple syllogisms (De Neys & Eranssens,
ing the base-rate. Conversely, under statistics instructions, partic-
2009; De Neys et al, 2010). Perhaps unsurprisingly, then, a
ipants were told to assume that their prior beliefs about the world
parallel debate has arisen in the deductive reasoning literature
are not necessarily relevant and that they should concentrate on the
where some authors have claimed that people have an intuitive
actual probability that something will happen. Participants were
sense of logical validity (e.g., Morsanyi & Handley, 2012; but see
therefore led to respond on the basis of the base-rate while ignor-
Klauer & Singmann, 2012). Eollowing those who have used in-
ing the stereotypical description for the statistics instructions. The
structional manipulations to investigate the relationship between
goal of the instruction manipulation was to lead participants to
deductive and inductive reasoning (Heit & Rotello, 2010; Rotello
focus on one of the two categories of information in the base-rate
& Heit, 2009), Handley, Newstead, and Trippas (2011) instructed
problems (i.e., base-rates or stereotypes). Tbe analytic base-rate
participants to judge both the logical validity and the believability
view outlined above predicts that participants should be less in-
of a series of simple reasoning problems. The instances of interest
clined to incorporate base-rates into their judgments when given
were problems in which logic and belief conflicted. The common
the opportunity to respond according to beliefs because they
view of belief bias is very similar to that of base-rate neglect,
should already have a representation of the problem derived from
namely, that Tl processes produce belief-based judgments that are
a highly accessible representativeness heuristic (Kahneman, 2003).
difficult to overturn via analytic T2 processing (e.g., Newstead,
Pollard, Evans, & Allen, 1992; Sa, West, & Stanovich, 1999).
Contrary to this account, participants were quicker and more Method
accurate at making logic judgments than belief judgments. More Participants. A total of 48 undergraduate students volun-
important, belief-logic confiict actually interfered more with belief teered to participate in exchange for course credit (23 from the
judgments during reasoning than with logic judgments. That is, University of Saskatchewan, Canada and 25 from Plymouth Uni-
inhibiting logical considerations appears more difficult than inhib- versity, England). Of the participants, 25 were female and 23 were
iting belief-based considerations for simple propositional reason- male. The mean age was 21 years {SD = 4).
ing problems. It was concluded that (a) making logically valid Materials and procedure. The participants received a total of
judgments does not necessarily require T2 processing, and (b) 24 base-rate problems similar to those used by De Neys and
making belief-based judgments does not exclusively rely on Tl Glumicic (2008). Twelve of the problems were taken from Thomp-
processing. son et al. (2011; previously adapted from De Neys & Glumicic,
In the current work, we test the hypothesis that utilizing base- 2008) and 12 were novel. The stereotypes for the 12 novel prob-
rates requires T2 processing by adapting this instruction manipu- lems were developed from adjective typicality ratings for a series
lation for base-rate problems. Whereas logic and belief confiict for of professions taken using data from an earlier study (Neilens,
syllogisms, base-rate probabilities and stereotypes (belief) confiict Handley, & Newstead, 2009). Stereotypical descriptions of com-
for base-rate problems. According to the analytic base-rate view, parable length to those used by De Neys and Glumicic were
information about beliefs enter into judgment autonomously via developed by using the adjective that was rated to be the most
Tl processes and should therefore interfere with the slower, more diagnostic for each profession (see online supplemental materials
deliberate T2 processing required to evaluate the base-rates. In for full set of problems). A pilot test using nine graduate student
addition, because the base-rates should be processed more slowly, volunteers from Plymouth University ensured that all descriptions
they should not interfere with judgments based on the stereotype. for both old and new items were stereotypical for the intended
Thus, if instructed to respond according to "belief," participants group and nondiagnostic for the complementary group (see results
should be able to make probability estimates that accord strongly below).
with the stereotypes, respond quickly, and be highly confident Participants were presented with two types of problems as
regardless of whether base-rates and stereotypes conflict. In con- defined by the relationship between the stereotypical description
trast, when instructed to respond according to "statistics," partic- and the base-rates (see Appendix); (a) base-rates and stereotype
ipants should have difficulty inhibiting the intuitive diagnostic suggested the same response (nonconflict), and (b) base-rates and
information, leading to slower response time, decreased confi- stereotype suggested different responses (conflict). In addition, we
dence, and less accurate probability estimates for conflict prob- created four neutral items in which the description was not diag-
lems. However, if base-rates can be comprehended by Tl pro- nostic of either group. These were used as practice items. Two
cesses, responding under statistics instructions should be no more versions of the stimulus set were created wherein each stereotyp-
difficult than responding under belief instructions and the conflict ical description matched the larger group (nonconflict) or the
between the two pieces of information should make these prob- smaller group (conflict) an equal number of times. Tbree base-rate
lems difficult, regardless of instructions. This should lead to longer probability ratios were presented equally often; 995/5, 996/4,
response times, lower confidence, and less accurate probability 997/3. Extreme ratios were used to remain consistent with the
estimates for conflict relative to nonconflict problems for both problem set that was used in earlier base-rate researcb (e.g.. De
belief and statistics judgments. Neys & Glumicic, 2008; De Neys et al., 2008; Pennycook, Eugel-
 BASE RATES ARE INTUITIVE 547
sang, & Koehler, 2012, 2013; Pennycook & Thompson, 2012; consistency of the raters was on the high end of acceptable (Cron-
Thompson et al., 2011). Participants were asked to estimate the bach's a = .79).
probability that the individual belonged to one of the two groups. Scoring. The data in the conflict condition were recoded so
The groups that were asked about were counterbalanced such that a high score always indicated a correct response given the
that, for each question, the large group was asked about half the instructions. This facilitated easy interpretation of the data. For
time, and the small group was asked about half of the time, instance, if there were 3 lawyers and 997 farmers in the sample and
allowing us to use the same description in both the nonconflict and the participant was questioned about the probability of being a
conflict conditions (e.g., in the doctor problem presented earlier, lawyer while under statistics instructions, a low probability esti-
changing the base-rate to 995 doctors and 5 nurses turns it into a mate would indicate a response in line with the instructions. On the
nonconflict problem). Each participant received half the items in other hand, if the participant was questioned about the farmers, a
their "nonconflicf form and half in their "conflict" form; the items high probability estimate would indicate a correct response. Under
were counterbalanced across participants such that each item ap- belief instructions, on the other hand, if the personality description
peared equally often in each form. Problem order was randomized was stereotypical for the lawyers and the participant was asked
for each participant. After each judgment, participants were asked about the fanners under belief instructions, a low probability
to rate their confidence on a scale from 1-7, ranging from not judgment would indicate compliance to the instructions. Finally,
confident at all to very confident. still under belief instmctions, if the participant was questioned
The instructions were manipulated within participant, such that about the lawyers, a high probability judgment would be correct.
they responded on the basis of belief half the time and on the basis To put all the trials on the same scale for comparison, we recoded
of statistics half the time: the probability estimates for the relevant cases by subtracting them
from 100 so that high values always indicated compliance with the
For each problem, you will be cued to answer either according to instmctions (hereafter referred to as "correct" probability esti-
beliefs or according to statistics. When instructed to answer according
mates). Examples for each case can be found in the Appendix.
to your beliefs, this means you must answer according to your
knowledge of what you think to be true in the world. For example, if Power analysis. We performed power analysis using More-
you met someone on the street who is dressed in very ragged clothing Power 6.0 (Campbell & Thompson, 2012) to ensure power was
and asking for money, it is a good bet that such a person is homeless. adequately high to detect the crucial Conflict X Instructions in-
If you were to be asked the probability that such a person is homeless, teraction. Handley et al. (2011) conducted five experiments with
you would want to give a high probability beeause, based upon our a very similar methodology and reported an average effect size for
knowledge of the world, people who dress in ragged clothing and ask the interaction of TI^ = .23 (range: .14-.33). We calculated the
for money on the street are usually homeless.
power to detect an effect with a conservative assumed effect
In contrast, when instructed to answer according to statistics, this
size of Tip = .15 for an alpha level of .05, given our sample. In
means you must assume that your prior beliefs about the world aren't the current experiment, power was acceptable (.81) given the
necessarily relevant. Instead, you should concentrate on the actual sample of 48 participants.
probability that something will happen. For example, if you knew that Probability estimates. We analyzed the probability estimates
only a small percentage of people in a city were homeless, then you as scored to reflect compliance with the instmctions (hereafter
would want to give a low probability because, based on statistics, only referred to as accuracy; see above) using a 2 (Congmency: non-
a small percentage of people are homeless. conflict vs. conflict) X 2 (Instmctions: belief vs. statistics)
repeated-measures analysis of variance (ANOVA). Probability
The instructions manipulation was counterbalanced across par-
estimates were less accurate for conflict (M = 65.8) than noncon-
ticipants, such that the problems presented under belief instmc-
flict (M = 81.3) trials, F(l, 47) = 43.4, p < .001, -q^ = .48. No
tions to half the participants were presented under statistics in-
other effects approached significance(all Fs < 1.5, all ps > .22).
structions to the other half, and vice versa. Participants were
In other words, conflict problems produced interference, regard-
prompted with "BELIEF" or "STATISTICS" at the bottom of the
less of whether people were instmcted to judge on the basis of
screen for each problem in a randomized order.
belief or statistics.
Latency. We transformed the response times (RTs) using the
Results natural logarithm—RTs in Table 1 are in original units. The
resulting values were submitted to a 2 (Congmency: nonconflict
Pilot test. Nine volunteers who did not participate in the vs. conflict) X 2 (Instmctions: belief vs. statistics) repeated-
actual experiment were presented our 24 descriptions. For each measures ANOVA. A main effect of instmctions was found,
description participants were given a 7-point Likert-type scale and indicating that participants made faster judgments under statistics
asked to rate whether the information was more stereotypical for instmctions (M = 12.73 s) than under belief instmctions (M =
the intended target group (= 7) or a group for which the descrip- 14.05 s), F(l, 47) = 6.24, p = .016, i\f, = .12. There was no main
tion was not intended (= 1). A score of 4 indicated that the effect of congmency, f( 1,47) = 2.55,p = .12,11^= .05, although
information was not particularly diagnostic for either group. We the trend is in the expected direction (i.e., longer RTs for conflict
tested whether the personality descriptions were rated as more relative to nonconfiict problems). The interaction did not approach
diagnostic of the intended group than for the complementary group significance (F < 1, p > .41).
using a one-sample t test comparing the average rating per partic- Confidence ratings. We performed a 2 (Congruency) X 2
ipant with the scale midpoint of four. The descriptions were (Instmctions) repeated-measures ANOVA on the confidence rat-
significantly more diagnostic of the target groups than the nontar- ings data. There was a main effect of congmency, wherein people
gets (M = 5.79, SD = 0.50), i(8) = 22.2, p < .001. The intemal were less confident for judgments where stereotype and base-rate
548 PENNYCOOK, TRIPPAS, HANDLEY, AND THOMPSON

Table 1
Mean Probability Estimates, Response Times, and Confidence Ratings as a Function of Congruency and Instruction for Experiment 1

Nonconflict Conflict
Belief Statistics Total Belief Statistics Total
Variable M SEM M SEM M SEM M SEM M SEM M SEM

Probability estimates 79.3 1.9 83.3 2.3 81.3 1.6 63.9 3.0 67.7 4.0 65.8 2.3
Response time (seconds) 13.7 0.9 12.2 0.9 12.9 0.8 14.4 1.1 13.2 0.9 13.8 0.9
Confidence (out of 7) 5.6 0.1 5.8 0.1 5.7 O.I 5.3 0.1 5.3 0.1 5.3 0.1

information were in confiict {M = 5.28) than when both cues Metbod

suggested the same response (M = 5.68), F(l, 47) = 19.2, p <
.001, r\j = .29. A marginally significant interaction between Participants. Eighty undergraduates from the University of
congruency and instructions was also found, F(l, 47) = 2.9, p = Saskatchewan participated in exchange for course credit. There
.098, T|p = .06. Follow-up comparisons confirmed that participants were 33 male and 47 female participants (M = 25, SD = 8).
were more confident in their judgments for nonconflict than for Materials and procedure. The materials and procedure were
conflict trials for both belief, i(47) = 3.04, p = .004, and logic identical to those used in Experiment 1, with the exception of the
instmctions, r(47) = 3.71, p = .001 (see Table 1). time limit. Participants were instmcted to respond within 5 s. After
the time limit had passed, the problem text would go bold and red
indicating that the participants had to respond immediately.
Discussion
We found a strong effect of congruency on accuracy, but no Results
main effect of the instruction manipulation. This finding suggests
tiiat processing the stereotypes interfered with making judgments Missing data. Participants responded within the time limit for
on the basis of the base-rates but also that the base-rates interfered 82% of the trials on average. We removed those participants who
with making judgments on the basis of the stereotypes. The results did not score within the time limit for at least half the trials {n =
from the confidence ratings converged on the same finding. At a 10). The remaining 70 participants responded within the imposed
minimum, this suggests that inhibiting the stereotypical informa- time limit for 89% of the trials. We excluded all responses that
tion is no more difficult than inhibiting the base-rate probability. were not given within the time limit. Missing values were replaced
Somewhat surprisingly, statistics judgments were performed more by the cell mean (Handley et al., 2011).
quickly than belief judgments without any loss of confidence, or Power analysis. We calculated the power to detect a large
without affecting the ability to comply with the instmctions when effect (T)P = .15) for an alpha level of .05 and given the current
making probability judgments. These results are clearly inconsis- sample size {n = 80 for the original sample but 70 for the actual
tent with the view that analytic processing is required for base- analysis). Power was high (>.93) for both cases.
rates to affect judgment. Rather, both base-rates and stereotypes Probability estimates. The probability estimates were re-
appear to be accessible to intuitive processing. scored in the manner described for Experiment 1. We submitted
the estimates to a 2 (Congruency: nonconflict vs. conflict) X 2
(Instmctions: statistics vs. belief) repeated-measures ANOVA.
Experiment 2
The analysis resulted in a main effect of congmency with more
In Experiment 1, the probability and the confidence data sup- accurate probability estimates for nonconflict (M = 80.2) than
ported an account of base-rate neglect that allows for intuitive conflict {M = 60.3) items, F(l, 69) = 104.4,/? < .001, j\j = .60
processing of base-rate probabilities. Although the latency data (see Table 2). No other effects reached significance (all Fs < 2, all
were less straightforward in terms of the congruency effect, over- ps > .17).
all, it was clearly not consistent with the view that utilizing Latency. All RTs were transformed using the natural loga-
base-rates requires analytic processing. In Experiment 2, we used rithm—RTs in Table 2 are in original units. We submitted the
a response deadline paradigm as a means to perform a stronger test resulting values to a 2 (Congmency: nonconflict vs. conflict) X 2
of both accounts of base-rate neglect by limiting T2 processing. (Instmctions: statistics vs. belief) repeated-measures ANOVA.
Limiting the time available to respond has been shown to interfere The analysis resulted in a main effect of congmency showing that
with analytic processing in deductive reasoning (Evans & Curtis- participants responded faster to nonconflict (M = 3.69 s) than to
Holmes, 2005; Schroyens, Schaeken, & Handley, 2003). If both conflict (M = 3.80 s) trials, F(l, 69) = 6.10,p = .016, T)^ = .08.
base-rate and stereotypes are accessible to Tl processing, base- No other effects approached significance (all Fs < 1, all ps > .50).
rates will interfere with stereotypes (and vice versa), as in El, Confidence ratings. The confidence ratings were submitted
regardless of whether participants are asked to respond according to a 2 (Congruency: nonconflict vs. conflict) X 2 (Instmctions:
to belief or statistics and regardless of the time deadline. In statistics vs. belief) repeated-measures ANOVA. The analysis re-
contrast, the analytic base-rate view suggests that participants sulted in a main effect of congruency, showing that people were
should have difficulty responding according to the "more analytic" more confident responding to nonconflict (M = 5.53) than to
base-rate information, particularly when under time presstire. confUct (M = 5.31) trials, F(l, 69) = 12.4, p = .001, -q^ = .15. No
 BASE RATES ARE INTUITIVE 549

Table 2
Mean Probability Estimates, Response Times, and Confidence Ratings as a Function of Congruency and Instruction for Experiment 2

Nonconflict Conflict
Belief Statistics Total Belief Statistics Total
Variable M SEM M SEM M SEM M SEM M SEM M SEM
Probability estimates 79.9 1.7 80.4 2.0 80.1 1.5 63.9 2.7 56.6 3.4 60.3 1.7
Response time (seconds) 3.68 0.07 3.70 0.07 3.69 0.07 3.81 0.08 3.79 0.08 3.80 0.07
Confidence (out of 7) 5.5 0.1 5.5 0.1 5.5 0.1 5.3 0.1 5.3 0.1 5.3 0.1

Other effects approached significance {Es < I, ps > .70). See were cycled through in-tum. Participants entered probability esti-
Table 2 for the means and standard deviations. mates in a text box and gave confidence ratings on a 7-point scale
(as above). Items were presented on separate pages, although
Discussion alongside confidence ratings.

As in Experiment 1, we found a strong effect of congruency on

accuracy and confidence data regardless of the time deadline, but Results
no main effect of the instruction manipulation. Participants also
took longer to respond for conflict relative to no confiict problems, Missing data. Given that participants were not continually
regardless of instruction. As above, base-rates interfered with kept on task by a trial-by-trial instruction prompt, as in Experi-
belief judgments—a finding that is not consistent witii tiie claim ments 1 and 2, we included a number of measures to ensure that
that comprehending base-rates requires slower, T2 processes, only participants who followed the belief/statistics instructions
whereas beliefs are processed using faster, Tl processes. More- were included in the analysis. Thus, participants were excluded
over, base-rates had robust effects on behavior even when partic- from analysis for a number of reasons. Specifically, two pariici-
ipants were instructed to respond according to beliefs under a time pants were excluded because they had missing data, three indicated
deadline intended to minimize T2 processing. that they had seen similar problems before in previous studies, one
person indicated that he or she was not fluent in English, and six
people failed a "check" question included to ensure that partici-
Experiment 3
pants were paying attention.' We also included three additional
In the previous two experiments we manipulated instruction as follow-up questions to ensure that participants followed the belief/
a within-participant factor. While the constant reminder about statistics instructions. Of the remaining participants, eight re-
which type of information to focus on may have kept participants sponded "yes" when asked, "Did you forget at any point during the
on task, it is also possible that switching from belief-based re- task to respond according to statistics?" (or "beliefs" depending on
sponding to statistics-based responding (and vice versa) may have condition), and six responded "no" when asked, "Did you read all
led to increased interference as a function of response confiict. In of the information for each problem carefully?" Finally, and most
other words, requiring participants to alternate use of base-rates important, participants were given an open-ended text box and
and stereotypes within the same experiment may have increased asked the following question directly after completing the 24
the probability of confiict detection regardless of instruction. It is base-rate problems (i.e., prior to the above check questions): "At
also possible that the act of switching in and of itself could create the beginning of this study, you were asked to respond to the
interference. To rule out these possible confounds we manipulated problems using probability estimates based on what?" In total, 27
instruction as a between-participants factor in an online study. of the remaining participants (28.7%) failed to correctiy answer
this question and were excluded on that basis.^ Thus, 67 partici-
Method pants (31 statistics condition, 36 beliefs condition) remained in the
final sample. None of the remaining participants had identical IP
Participants. We recruited 120 participants online via Ama- addresses.
zon Mechanical Turk. There were 67 male and 53 female pariic-
Power analysis. We calculated the power to detect a large
ipants (M„g^ = 36, SD = 13.8).
effect {vij = .15) for an alpha level of .05 and given tiie retained
Materials and procedure. Items from Experiment 1 were
sample size of 67 participants. As in the previous two Experi-
presented to participants in a single random order. Participants
ments, power to detect the interaction between congruency and
were presented with either statistics or belief instructions based on
instruction was sufficientiy high (.92).
whether their birthday falls on the first or second half of the year,
counterbalanced across participants. This was done because the
online format requires each condition to be run in-tum. Thus, ' Participants were given a list of hobbies following a demographics
participants filtered into the crucial two between-participants con- questionnaire and asked "Below is a list of hobbies. If you are reading these
ditions at roughly equivalent times. Congruency was also counter- instructions please write 'I read the instructions' in the 'other' box."
balanced across participants such that each problem was a conflict ^ We permitted some minimal leniency here. For example, "probability"
or "percentages" were common answers in the statistics condition and
problem for half of the participants and a nonconflict problem for "knowledge of the world" and "experience" came up in the belief condi-
the other half. This yielded four between subjects conditions that tion.
550 PENNYCOOK, TRIPPAS, HANDLEY, AND THOMPSON

Probability estimates. We submitted the probability ratings Why Are Base-Rates Neglected?
recoded to signify correct responding to a 2 (Congruency; non-
Using a binary choice format, only roughly 18-24% of respon-
conflict vs. conflict) X 2 (Instructions; statistics vs. belief) mixed
dents typically choose tbe base-rate option over the stereotypical
ANOVA. The analysis resulted in a main effect of congruency
response for problems similar to the ones used here (De Neys &
with higher probability estimates for nonconflict (M = 88.6) than
Glumicic, 2008; Pennycook, Eugelsang, & Koehler, 2012). If
conflict (M = 64) items, F(l, 65) = 59.47,p < .001, •^j, = .48 (see
base-rates are accessible to intuitive processing, why are they
Table 3). No other effects reached significance(all Fs < 1.5, all
routinely underweighted or, in some cases, neglected completely?
ps > .23).^
One view is that base-rates are neglected because they require
Confidence ratings. The confidence ratings were submitted
effortful analytic reasoning and that responses based upon explicit
to a 2 (Congruency; nonconflict vs. conflict) X 2 (Instructions;
cues of this kind are preempted by intuitive judgments that draw
statistics vs. belief) mixed ANOVA. The analysis resulted in a upon readily available stereotypes (e.g.. De Neys & Glumicic,
main effect of congruency, showing that people were more confi- 2008; Eerreira et al., 2006; Kahneman & Erederick, 2002). We
dent responding to nonconflict (M = 5.83) than to conflict (M = suggest instead that participants typically respond according to
5.26) trials, F(l, 65) = 42.24, p < .001, -q^ = .39. No other effects stereotypes rather than base-rates not because the former is pri-
approached significance (Fs < 1, ps > .35).^* See Table 3 for the marily intuitive and the latter is primarily analytic, but because the
means and standard deviations. former is simply more salient than the latter in the base-rate
problems that are typically constructed. Whereas the base-rates
represent a single source of information, the descriptions that we
Discussion
presented contained three to five pieces of detailed information
The results fi"om Experiment 1 were replicated in a between- that reinforced the targeted stereotype (see supplemental materials
participants design despite the fact that we were very conservative for items). If we assume that the likelihood and strength of a
in our analysis, including only the participants who passed a particular intuitive output is affected by contextual factors such as
number of "check" questions (representing 55.8% of the sample). this, it is no surprise that diagnostic information is typically fa-
These data rule out the possibility that switching from belief- vored over base-rate probabilities. This context-dependent view of
based to statistics-based judgments (and vice versa) explain the base-rate use is consistent with recent research which shows that
results from Experiments 1 and 2. We also note that the prob- the percentage of base-rate responses for problems with rich
ability estimates and confidence ratings for the between- sources of diagnostic information (as used here) tends to be very
participants online sample experiment (Table 3) closely parallel low (24%; Pennycook, Eugelsang, & Koehler, 2012). However, an
the data from the within-participant university sample experi- alternative set of problems where the descriptions merely consisted
ments (Tables 1 and 2). of a set of five personality traits (e.g., orderly, organized, precise,
practical and realistic) along witb extreme base-rates, resulted in a
majority (59%) of base-rate responses.
General Discussion How does this relative salience account explain existing data
The finding that base-rates are underweighted or neglected is that seems to support the claim that utilizing base-rates requires T2
commonly explained by noting the compelling intuitive nature of processing? Specifically, high capacity participants respond more
stereotypical or diagnostic information and asserting a failure to often according to tbe base-rates than lower capacity individuals
adjust the response by integrating the base-rates via analytic rea- (Pennycook et al., 2013; but see Stanovich & West, 2008) and the
soning processes (e.g.. De Neys & Glumicic, 2008; Eerreira et al., mere willingness to engage analytic reasoning processes also pre-
2006; Kahneman & Erederick, 2002). This assumption is consis- dicts base-rate responding (Pennycook, Cheyne, et al., 2012).
tent with the finding that reasoners who are more able and/or Consistent with the conclusions presented above, we speculate that
willing to engage in T2 processing are more likely to utilize the observed differences may be attributed to strategy choices.
base-rate information when making judgments (e.g., Pennycook et Under this explanation, the overall preference for diagnostic in-
al., 2013). However, the data presented here indicate that, at the formation likely has more to do with cognitive miserliness (i.e.,
very least, extreme base-rates are accessible to rapid, Tl process- nonanalytic cognitive style) and less to do witb competence (i.e.,
ing. Specifically, both belief and statistical judgments were made absent or ineffective cognitive abilities) because, as our data indi-
difficult by conflict, even when the possibility for T2 processing cate, it does not appear to be a processing deficit that leads to
was reduced by imposing a time deadline. Einally, participants base-rate neglect. In other words, we suggest that the base-rates are
were just as fast and confident when responding according to typically recognized by all but that individuals who are willing to
statistics as when making judgments based on belief. These results engage T2 processing consider them more explicitly. More miserly
all converge on the same conclusion; Base-rates probabilities (at participants may focus on the more detailed stereotypical informa-
least the extreme ones used in our study) are accessible to Tl
processing and can influence beliefs just as beliefs can influence ^ As would be expected, the excluded participants gave probability
statistics judgments. Thus, the conflict observed between base- estimates that were (moderately) less in-line with the instruction relative to
rates and stereotypes may result from a conflict of Tl outputs, the included participants, i(106) = 1.72, SE = 2.95, p = .089.
rather tban a conflict between a Tl and T2 output (De Neys, 2012; '' The excluded participants also had lower confidence for conflict (M =
5.40) relative to nonconflict (M = 5.85) problems, f ( l , 40) = 20.35, p <
Pennycook, Eugelsang, & Koehler, 2012; Pennycook & Thomp- .001, -qf, = .34. Without the instruction manipulation, the task for the
son, 2012; Thompson et al., 2011). excluded participants would be akin to a standard base-rate neglect study.
 BASE RATES ARE INTUITIVE 551

Table 3
Mean Probability Estimates and Confidence Ratings as a Function of Congruency and Instruction for Experiment 3

Nonconflict Conflict
Belief Statistics Total Belief Statistics Total
Variable M SEM M SEM M SEM M SEM M SEM M SEM
Probability estimates 87.7 1.7 89.4 2.3 88.5 1.4 60.6 3.9 67.5 4.8 63.8 3.1
Confidence (out of 7) 6.0 O.I 5.7 0.2 5.8 O.I 5.4 0.2 5.2 0.2 5.3 0.1

tion because it leads to a stronger intuitive response that comes to are accessible to Tl processing is a violation of the definition of Tl
mind more fluently, leading to an increased feeling of rightness processing as evolutionarily old, contextualized, and isolated to
(Thompson et al., 2011, 2013). Participants who are more willing implicit knowledge (for overviews of dual-process theories, see
to engage analytic reasoning are simply more likely to weigh the Evans, 2008, 2009; Evans & Frankish, 2009; Evans & Stanovich,
base-rates against the stereotypes, given that the more salient 2013; Sloman, 1999; Stanovich, 2009; Stanovich & West, 2000).
diagnostic information remains the default, this leads to decreased However, as recently discussed by Evans and Stanovich (2013),
levels of stereotypical responding. This process may also be facil- factors such as contextualization are typical correlates of Tl or T2
itated by the possibility that more analytic individuals are better processes, but not defining features. Evans and Stanovich have
able to detect conflicts between cognitive outputs, leading to isolated the use of working memory and the processes of cognitive
increased deliberative weighing of the conflicting responses decoupling and mental simulation as defining features of T2 pro-
against each other (Pennycook et al., 2013). This account is con- cessing. In contrast, Tl processes are considered autonomous and
sistent with the finding that analytic cognitive style is more pre- operate independently of working memory. Thus, although base-
dictive of base-rate responding than cognitive ability (Pennycook, rates may not be intuitive in the sense that they cue a cognitive
Cheyne, et al., 2012, 2013; Stanovich & West, 2008), though this output via gist-based associations in memory (Reyna, 2004, 2012;
finding needs to be replicated using full IQ scales. It is also Reyna & Brainerd, 2008), it is possible that the use of base-rates
consistent with the finding that "neutral" base-rate problems (i.e., in judgment has been practiced enough by adulthood that they are
items that lack stereotypical content) are associated with lower part of an "encapsulated knowledge base" that is accessible to Tl
"feelings of rightness" than problems with stereotypical content, processing (Stanovich, West, & Toplak, 2011). This potentially
regardless of conflict status (Thompson et al., 2011). We note, in explains how a simple but nonetheless abstract probabilistic con-
addition, that interpretation of the individual difference data is also cept can be processed via Tl processes.
complicated by the fact that the normatively correct response to The finding that base-rates interfered with purportedly belief-
typical base-rate problems according to Bayes' theorem is to based judgments suggests that they entered into judgment auton-
combine the diagnostic and base-rate probabilities, a strategy that omously; i.e., the execution of the Tl process was mandatory once
does not appear to arise very often for conflict problems (Penny- the stimulus was encountered (Evans & Stanovich, 2013; Thomp-
cook & Thompson, 2012). son, 2013). However, it is not neces.'iarily true that the autonomous
Another finding that appears to support the claim that utilizing instantiation of the rule "if stimulus then response" is qualitatively
base-rates requires T2 processing is that base-rate neglect has been different from more deliberative T2 processes (Kruglanski, 2013;
shown to increase for conflict problems when participants were Kruglanski & Gigerenzer, 2011 ). This forms part of a larger debate
given a working memory load, suggesting a decrease in base-rate that certainly will not be resolved here (see Evans & Stanovich,
processing concurrent with a disruption of T2 processing (Frans- 2013; Keren, 2013; Kmglanski, 2013; Osman, 2013; Thompson,
sens & De Neys, 2009). However, Franssens and De Neys (2009) 2013). However, our data do speak to this debate. Specifically, the
also demonstrated that recall of base-rates was not affected by the finding that conflict selectively increases response time is parsi-
working memory load, indicating that the base-rates were pro- moniously explained in a dual-process framework as an increase in
cessed despite the decrease in base-rate responding. Moreover, as T2 processing that is required to resolve the conflict between two
previously discussed, giving participants an opportunity to rethink cognitive outputs. The difficulty that participants apparently have
an initial response leads to a roughly equal proportion of answer deciding between the base-rates and stereotypes (Pennycook &
changes toward the base-rates as away from them (Pennycook & Thompson, 2012) is not at all surprising given the claim that
Thompson, 2012). These findings indicate that while T2 process- cognitive decoupling is a primary function of (typically) more
ing is not likely required for base-rates to enter into judgment, T2 effortful T2 processing (Stanovich, 2009, 2011). Our data indicate
processing may nonetheless be necessary to decide between the that participants must actively inhibit the conflicting source of
two sources of conflicting information. information to solve the problem, a process that seems to be
fundamentally different from the one that engendered the output in
the first place.
Dual-Process Theory and Intuitive Base-Rates
Given that our data are inconsistent with the standard dual-
process explanation of base-rate neglect, one might wonder why
Additional Implications
we have nonetheless framed our results in terms of dual-process These data have implications for research outside of base-rate
theory. One could argue, for example, that the claim that base-rates neglect. In Experiment 2, participants had lower confidence and
552 PENNYCOOK, TRIPPAS, HANDLEY, AND THOMPSON

slower response time for conflict relative to nonconflict problems De Neys, W., & Bonnefon, J.-F. (2013). The 'whys' and 'whens' of
despite the short deadline. This indicates that the participants in individual differences in thinking biases. Trends in Cognitive Sciences.
our sample were able to rapidly detect the conflict between base- Advance online publication. doi:10.1016/j.tics.2013.02.001
rates and stereotypes (De Neys & Glumicic, 2008; De Neys et al., De Neys, W., Comheeke, S., & Osman, M. (2011). Biased but in doubt:
Conflict and decision confidence. PLoS ONE, 6, el5954. doi:10.137l/
2008, 2011). This, along with evidence that conflict detection for
journal.pone.0015954
base-rate problems is not disrupted by working memory load
De Neys, W., & Franssens, S. (2009). Belief inhibition during thinking:
(Franssens & De Neys, 2009), suggests that the detection of
Not always winning but at least taking part. Cognition, 113, 45-61.
conflict during reasoning is an effortless and potentially automatic doi:10.1016/j.cognition.2009.07.009
process (De Neys, 2012). Future research is required to further De Neys. W., & Glumicic, T. (2008). Conflict monitoring in dual process
investigate this claim. theories of thinking. Cognition, 106, 1248-1299. doi:10.1016/j
That base-rates and stereotypes appear to have been encoded .cognition.2007.06.002
and activated in parallel, a likely prerequisite for rapid conflict De Neys, W;, Moyens, E., & Vansteenwegen, D. (2010). Feeling we're
detection, is also consistent with the parallelism assumption of biased: Autonomie arousal and reasoning conflict. Cognitive, Affective,
fuzzy trace theory (e.g., Reyna, 2012). We note, however, that & Behavioral Neuroscience, 10, 208-216. doi: 10.3758/CABN. 10.2.208
although fuzzy-trace theory holds that verbatim and gist-based De Neys, W., Rossi, S., & Houdé, O. (2013). Bats, balls, and substitution
representations are encoded and processed in parallel, base-rate sensitivity: Cognitive misers are no happy fools. Psychonomic Bulletin
& Review. Advance online publication.
neglect is nonetheless explained by positing that base-rates are
De Neys, W., Vartanian, O., & Goel, V. (2008). Smarter than we think:
processed differently than stereotypes (i.e., diagnostic information
When our brains detect that we are biased. Psychological Science, 19,
cues gist-based associations in memory, whereas base-rates are
483-489. doi: 10.1111/j.14679280.2008.02113.x
processed via verbatim-based reasoning; see Wolfe & Fisher, Evans, J. S. T. (2008). Dual-processing accounts of reasoning, judgment,
2013). Our data do not support the claim that base-rates and and social cognition. Annual Review of Psychology, 59, 255-278. doi:
stereotypes are processed in fundamentally different ways. How- 10.1146/annurev.psych.59.103006.093629
ever, it may be the case that extreme base-rate probabilities pre- Evans, J. S. T. (2009). How many dual process theories do we need: One,
sented in frequency format fall somewhere in-between the oppo- two, or many? In J. Evans & K. Frankish (Eds.), In two minds: Dual
site poles of verbatim and gist representations, thereby leading to processes and beyond (pp. 33-54). Oxford, England: Oxford University
very subtle processing differences that require altemative methods Press. doi:10.I093/acprof:oso/9780199230167.003.0002
and measures to discem. Here, too, further research is required. Evans, J. S. T. (2012). Dual-process theories of deductive reasoning: Facts
and fallacies. In K. Holyoak & R. G. Morrison (Eds.), The Oxford
handbook of thinking and reasoning (pp. 115-133). Oxford, England:
Conclusion Oxford University Press. doi:10.1093/oxfordhb/9780199734689.013
.0008
The tendency for researchers to make inferences about the type
Evans, J. S., & Curtis-Holmes, J. (2005). Rapid responding increases belief
of reasoning process that has occurred based solely on the type of
bias: Evidence for the dual-process theory of reasoning. Thinking &
response that was given is widespread (Evans, 2012). Above all Reasoning, II, 382-389. doi:10.1080/13546780542000005
else, the data presented here demonstrate the difficulty with this Evans, J. S. T., & Frankish, K. (Eds.). (2009). In two minds: Dual
approach. Specifically, the claim that stereotypes are "intuitive" processes and beyond. Oxford, England: Oxford University Press, doi:
and base-rates are "analytic" does not hold up to serious scrutiny. 10.1093/acprof :oso/9780199230167.001.0001
This finding, paired with similar results in the domain of deductive Evans, J. S. T., & Stanovich, K. E. (2013). Dual-process theories of higher
reasoning (Handley et al., 2011), provide empirical evidence that cognition: Advancing the debate. Perspectives on Psychological Sci-
reinforces the point that attempts to infer a reasoning process ence, 8, 223-241. doi:10.1177/1745691612460685
solely from a response should be treated with caution. Ferreira, M. B., Garcia-Marques, L., Sherman, S. J., & Sherman, J. W.
(2006). Automatic and controlled components of judgment and decision
making. Journal of Personality and Social Psychology, 91, 797-813.
References doi: 10.1037/0022-3514.91.5.797
Barbey, A. K., & Sloman, S. A. (2007). Base-rate respect: From ecological Franssens, S., & De Neys, W. (2009). The effortless nature of conflict
rationality to dual processes. Behavioral and Brain Sciences, 30, 241. detection during thinking. Thinking & Reasoning, 15, 105-128. doi:
doi: 10.1017/SO140525X07001653 10.1080/13546780802711185
Bonner, C , & Newell, B. R. (2010). In conflict with ourselves? An Handley, S. J., Newstead, S. E., & Trippas, D. (2011). Logic, beliefs, and
investigation of heuristic and analytic processes in decision making. instruction: A test of the default interventionist account of belief bias.
Memory & Cognition, 38, 186-196. doi:10.3758/MC.38.2.186 Joumal of Experimental Psychology: Learning, Memory, and Cogni-
Brainerd, C. J. (2007). Kissing cousins but not identical twins: The de- tion, 37, 28-43. doi:10.1037/a0021098
nominator neglect and base-rate respect models. Behavioral and Brain Heit, E., & Rotello, C. M. (2010). Relations between inductive reasoning
Sciences, 30, 257-258. doi:10.1017/S0140525X07001689 and deductive reasoning. Journal of Experimental Psychology: Learn-
Campbell, J. I. D., & Thompson, V. (2012). MorePower 6.0 for ANOVA ing, Memory, and Cognition, 36, 805-812. doi:10.1037/a0018784
with relational confidence intervals and Bayesian analysis. Behavior Kahneman, D. (2003). A perspective on judgment and choice: Mapping
Research Methods, 44, 1255-1265. doi:10.3758/sl3428-012-0186-0 bounded rationality. American Psychologist, 58, 697-720. doi: 10.1037/
De Neys, W. (2007). Nested sets and base-rate neglect: Two types of 0003-066X.58.9.697
reasoning? Behavioral and Brain Sciences, 30, 260-261. doi:10.1017/ Kahneman, D. (2011). Thinking, fast and slow. New York, NY: Farrar,
S0140525X07001719 Straus and Giroux.
De Neys, W. (2012). Bias and conflict: A case for logical intuitions. Kahneman, D., & Frederick, S. (2002). Representativeness revisited: At-
Perspectives on Psychological Science, 7, 2 8 - 3 8 . doi:10.1177/ tribute substitution in intuitive judgment. In T. Gilovich, D. Griffln, &
1745691611429354 D. Kahneman (Eds.), Heuristics and biases (pp. 49-81). New York,
 BASE RATES ARE INTUITIVE 553
NY: Cambridge University Press. doi:10.1017/CBO9780511808098 Behavioral and Brain Sciences, 30, 278-280. doi: 10.1017/
.004 S0140525X07001872
Kahneman, D., & Tversky, A. (1973). Psychology of prediction. Psycho- Rotello, C. M., & Heit, E. (2009). Modeling the effects of argument length
logical Review, 80, 237-251. doi:10.1037/h0034747 and validity on inductive and deductive reasoning. Journal of Experi-
Keren, G. (2013). A tale of two systems: A scientiflc advance or a mental Psychology: Leaming, Memory, and Cognition, 35, 1317-1330.
theoretical stone soup? Commentary on Evans & Stanovich (2013). doi:I0.1037/a0016648
Perspectives on Psychological Science, 8, 257-262. doi:10.1177/ Sa. W., West, R., & Stanovich, K. (1999). The domain specificity and
1745691613483474 generality of belief bias: Searching for a generalizable critical thinking
Klauer, K. C , & Singmann, H. (2012). Does logic feel good? Testing for skill. Joumal of Educational Psychology, 9/, 497-510. doi: 10.1037/
intuitive detection of logicality in syllogistic reasoning. Journal of 0022-0663.91.3.497
Experimental Psychology: Learning, Memory, and Cognition. Advance Schroyens, W., Schaeken, W., & Handley, S. (2003). In search of counter-
online publication. doi:l0.1037/a0030530 examples: Deductive rationality in human reasoning. The Quarterly
Kruglanski, A. W. (2013). Only one? The default interventionist perspec- Journal of Experimental Psychology. A, Human Experimental P.sychol-
tive as a unimodel: Commentary on Evans & Stanovich (2013). Per- ogy, 56, 1129-1145. doi: 10.1080/02724980245000043
spectives on Psychological Science, 8, 2 4 2 - 2 4 7 . doi: 10.1177/ Sloman, S. (1996). The empirical case for two systems of reasoning.
I745691613483477 Psychological Bulletin, ¡¡9. 3-22. doi:10.1037/0033-2909.119.1.3
Kruglanski, A. W., & Gigerenzer, G. (2011). Intuitive and deliberative Stanovich, K. E. (2004). The Robot's rebellion: Finding meaning in the
judgments are based on common principles. P.'ivchological Review, ¡18. age of Darwin. Chicago, IL: The University of Chicago Press, doi:
97-109. doi: 10.1O37/a0020762 10.7208/chicago/9780226771199.001.0001
Morsanyi, K., & Handley, S. J. (2012). Logic feels so good—I like it! Stanovich, K. E. (2009). Is it time for a tri-process theory? Distinguishing
Evidence for intuitive detection of logicality in syllogistic reasoning. the reflective and algorithmic mind. In J. S. T. Evans & K. Frankish
Journal of Experimental Psychology: Leaming. Memory, and Cogni- (Eds.), In two minds: Dual processes and beyond (pp. 55-88). Oxford,
tion. 38, 596-616. doi:10.1037/a0026099 England: Oxford University Press. doi:10.1093/acprof:oso/
Neilens, H., Handley, S., & Newstead, S. E. (2009). Effects of training and 9780199230167.003.0003
instruction on analytic and belief-based reasoning. Thinking & Reason- Stanovich, K. E. (2011). Rationality and the reflective mind. New York,
ing, ¡5, 37-68. doi:l0.1080/13546780802535865 NY: Oxford University Press.
New.stead, S., Pollard, P.. Evans, J., & Allen, J. (1992). The source of belief Stanovich, K. E., & West, R. F. (2000). Individual differences in reasoning:
bias effects in syllogistic reasoning. Cognition, 45, 257-284. doi: Implications for the rationality debate. Behavioral and Brain Sciences,
10.1016/0010-0277(92)90019-E 23, 645-665. doi:10.1017/S0140525X00003435
Osman, M. (2013). A case study: Dual-process theories of higher cogni- Stanovich, K. E., & West, R. F. (2008). On the relative independence of
tion: Commentary on Evans & Stanovich (2013). Perspectives on Psy- thinking biases and cognitive ability. Joumal of Personality and Social
chological Science, 8. 248-252. doi:10.1177/1745691613483475 Psychology, 94. 672-695. doi:10.1037/0022-3514.94.4.672
Pennycook, G., Cheyne, J. A., Barr, N.. Koehler, D. J., & Fugelsang, J. A. Stanovich, K. E., West, R. F., & Toplak, M. W. (2011). The complexity of
(2013). Cognitive style and religiosity: The role of conflict detection. developmental predictions from dual process models. Developmental
Memory & Cognition. Advance online publication. doi:10.3758/sl3421- Review, 31, 103-118. doi:l0.1016/j.dr.2011.07.003
013-0340-7 Thompson, V. A. (2013). Why it matters: The implications of autonomous
Pennycook, G., Cheyne, J. A., Seli, P., Koehler, D. J., & Fugelsang, J. A. processes for dual-process theories: Commentary on Evans & Stanovich
(2012). Analytic cognitive style predicts religious and paranormal belief. (2013). Perspectives on Psychological Science, 8, 253-256. doi:
Cognition, 123, 335-346. doi:10.1016/j.cognition.2012.03.003 10.1177/1745691613483476
Pennycook, G., Fugelsang, J. A., & Koehler, D. J. (2012). Are we good at Thompson, V. A., Prowse Turner, J. A., & Pennycook, G. (2011). Intuition,
detecting conflict during reasoning? Cognition, 124, 101-106. doi: reason, and metacognition. Cognitive Psychology, 63, 107-140. doi:
10.1016/j.cognition.2012.04.004 10.1016/j.cogpsych.2011.06.001
Pennycook, G., & Thompson. V. A. (2012). Reasoning with base-rates is Thompson, V. A., Prowse Turner. J., Pennycook, G., Ball, L., Brack, H.,
routine, relatively effortless, and context dependent. Psychonomic Bul- Ophir, Y., & Ackerman, R. (2013). The role of answer fluency and
letin & Review. ¡9. 528-534. doi: 10.3758/s 13423-012-0249-3 perceptual fluency as metacognitive cues for initiating analytic thinking.
Reyna, V. F. (2004). How people make decisions that involve risk: A Cognition, 128, 237-251. doi:10.I0I6/j.cognition.2012.09.012
dual-processes approach. Current Directions in Psychological Science. Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heu-
13. 60-66. doi:10.1111/j.0963-7214.2004.00275.x ristics and biases. Science, 185, 1124-1131. doi:10.1126/science.l85
Reyna, V. F. (2012). A new intuitionism: Meaning, memory, and devel- .4157.1124
opment in fuzzy-trace theory. Judgment and Decision Making, 7, 332- Villejoubert, G. (2009). Are representativeness judgments automatic and
359. rapid? The effect of time pressure on the conjunction fallacy. Proceed-
Reyna, V. F., & Brainerd, C. J. (2008). Numeracy, ratio bias, and denom- ings of the Annual Meeting of the Cognitive Science Society, 30. 2980-
inator neglect in judgments of risk and probability. Learning and Indi- 2985.
vidual Differences, IS, 89-107. doi:10.1016/j.lindif.2007.03.011 Wolfe, C. R., & Fisher, C. R. (2013). Individual differences in base rate
Reyna, V. F., & Mills, B. (2007). Converging evidence supports fuzzy- neglect: A fuzzy processing preference index. Leaming and Individual
trace theory's nested sets hypothesis, but not the frequency hypothesis. Differences, 25, 1-11. doi:l0.1016/j.lindif.2013.03.003

{Appendix follows)
554 PENNYCOOK, TRIPPAS, HANDLEY, AND THOMPSON

Appendix
Example Problem

Table Al

High base-rate Low base-rate

High diagnosücity [CONGRUENT, CELL #1] [INCONGRUENT, CELL #2]

In a study 1,000 people were tested. Brannon is a randomly In a study 1,000 people were tested. Brannon is a randomly
chosen participant of this study. chosen participant of this study.
Among the participants there were 995 accountants and 5 Among the participants there were 5 accountants and 995
street artists. street artists.
Brannon is 29 years old. He is very good with numbers but Brannon is 29 years old. He is very good with numbers but
is shy around people. He spends much of his time is shy around people. He spends much of his time
working. working.
What is the probability that Brannon is an accountant? What is the probability that Brannon is an accountant?
Correct according to statistics ~ 100 Correct according to statistics ~ 0
Correct according to beliefs ~ 100 Correct according to beliefs ~ 100
Correct according to base-rate ~ 100 Correct according to base-rate ~ 0

Low diagnosticity [INCONGRUENT, CELL #3] [CONGRUENT. CELL #4]

In a study 1,000 people were tested. Brannon is a randomly In a study 1,000 people were tested. Brannon is a randomly
chosen participant of this study. chosen participant of this study.
Among the participants there were 5 accountants and 995 Among the participants there were 995 accountants and 5
street artists. street artists.
Brannon is 29 years old. He is very good with numbers but Brannon is 29 years old. He is very good with numbers but
is shy around people. He spends much of his time is shy around people. He spends much of his time
working. working.
What is the probability that Brannon is a street artist? What is the probability that Brannon is a street artist?
Correct according to statistics ~ 100 Correct according to statistics ~ 0
Correct according to beliefs ~ 0 Correct according to beliefs ~ 0
Correct according to base-rate ~ 100 Correct according to base-rate ~ 0

Received March 4, 2013

Revision received September 6, 2013
Accepted September 9, 2013
Copyright of Journal of Experimental Psychology. Learning, Memory & Cognition is the
property of American Psychological Association and its content may not be copied or emailed
to multiple sites or posted to a listserv without the copyright holder's express written
permission. However, users may print, download, or email articles for individual use.