Choking Under Pressure: Multiple Routes to Skill Failure

Abstract

Poor performance in pressure-filled situations, or "choking under pressure," has largely been explained by two different classes of theories. Distraction theories propose that choking occurs because attention needed to perform the task at hand is coopted by task-irrelevant thoughts and worries. Explicit monitoring theories claim essentially the opposite-that pressure prompts individuals to attend closely to skill processes in a manner that disrupts execution. Although both mechanisms have been shown to occur in certain contexts, it is unclear when distraction and/or explicit monitoring will ultimately impact performance. The authors propose that aspects of the pressure situation itself can lead to distraction and/or explicit monitoring, differentially harming skills that rely more or less on working memory and attentional control. In Experiments 1-2, it is shown that pressure that induces distraction (involving performance-contingent outcomes) hurts rule-based category learning heavily dependent on attentional control. In contrast, pressure that induces explicit monitoring of performance (monitoring by others) hurts information-integration category learning thought to run best without heavy demands on working memory and attentional control. In Experiment 3, the authors leverage knowledge about how specific types of pressure impact performance to design interventions to eliminate choking. Finally, in Experiment 4, the selective effects of monitoring-pressure are replicated in a different procedural-based task: the serial reaction time task. Skill failure (and success) depends in part on how the performance environment influences attention and the extent to which skill execution depends on explicit attentional control.

Full text

Choking Under Pressure: Multiple Routes to Skill Failure
Marci S. DeCaro
Miami University and Vanderbilt University Robin D. Thomas
Miami University
Neil B. Albert
Spencer Foundation and The University of Chicago Sian L. Beilock
The University of Chicago
Poor performance in pressure-filled situations, or “choking under pressure,” has largely been explained
by two different classes of theories. Distraction theories propose that choking occurs because attention
needed to perform the task at hand is coopted by task-irrelevant thoughts and worries. Explicit monitoring
theories claim essentially the opposite—that pressure prompts individuals to attend closely to skill
processes in a manner that disrupts execution. Although both mechanisms have been shown to occur in
certain contexts, it is unclear when distraction and/or explicit monitoring will ultimately impact perfor-
mance. The authors propose that aspects of the pressure situation itself can lead to distraction and/or
explicit monitoring, differentially harming skills that rely more or less on working memory and
attentional control. In Experiments 1–2, it is shown that pressure that induces distraction (involving
performance-contingent outcomes) hurts rule-based category learning heavily dependent on attentional
control. In contrast, pressure that induces explicit monitoring of performance (monitoring by others) hurts
information-integration category learning thought to run best without heavy demands on working
memory and attentional control. In Experiment 3, the authors leverage knowledge about how specific
types of pressure impact performance to design interventions to eliminate choking. Finally, in Experi-
ment 4, the selective effects of monitoring-pressure are replicated in a different procedural-based task:
the serial reaction time task. Skill failure (and success) depends in part on how the performance
environment influences attention and the extent to which skill execution depends on explicit attentional
control.
Keywords: pressure, category learning, working memory, attention, serial reaction time task
People often find themselves in high-stakes situations where
performing their best carries implications for future opportunities
and success. Whether it is a high school student taking the SAT, a
golfer playing to make the cut for the PGA tour, or a violinist
auditioning for an orchestra, high-level performance in important
situations is crucial for advancement in most facets of life. In these
types of high-stakes situations, the desire to perform as well as
possible is thought to create performance pressure (Baumeister,
1984; Hardy, Mullen, & Jones, 1996; Beilock & Carr, 2001).
Unfortunately, in both real-world (e.g., Dandy, Brewer, & Tott-
man, 2001; Davis & Harvey, 1992; Dohmen, 2008; Forgas, Bren-
nan, Howe, Kane, & Sweet, 1980; Heaton & Sigall, 1989; Paulus,
Shannon, Wilson, & Boone, 1972) and laboratory situations (e.g.,
Beilock, 2008; Beilock & Gray, 2007), this pressure to attain
performance success often causes people to perform below their
actual abilities. The term choking under pressure describes this
phenomenon. Choking is not just poor performance. Rather, it is
performing more poorly than expected, given one’s skill level, in
situations where performance pressure is at a maximum (Beilock
& Gray, 2007).
Understanding why choking occurs is important for devising
training regimens to alleviate it. Yet investigations into un-
wanted skill failure can do a lot more. Understanding skill
failure and success under pressure may shed light on the sim-
ilarities and differences in the cognitive control structures un-
derlying a diverse set of skills, ranging from math problem
solving to golf putting. Moreover, by uncovering the mecha-
nisms governing pressure-induced failure, we can also further
our understanding of how emotional and motivational factors
combine with memory and attention processes to impact skill
learning and performance. An understanding of how the per-
formance environment alters cognitive processes not only ad-
vances our understanding of the “choking under pressure” phe-
nomenon specifically but also provides insight into related
situations in which performance inadvertently falters, ranging
from test anxiety (Ashcraft, 2002; Wine, 1971) to the threat of
conforming to a negative stereotype (i.e., stereotype threat;
Steele, 1997).
This article was published Online First May 16, 2011.
Marci S. DeCaro is now in the Department of Psychology and Human
Development at Vanderbilt University; Robin D. Thomas, Department of
Psychology; Neil B. Albert, Department of Psychology; Sian L. Beilock,
Department of Psychology.
This research was supported by National Science Foundation CAREER
Grant DRL-0746970 to Sian L. Beilock and National Science Foundation
Grant BCS-0544688 to Robin D. Thomas. The authors thank Krista Carl-
son for her assistance with response modeling.
Correspondence concerning this article should be addressed to Marci S.
DeCaro, Department of Psychology and Human Development, 230 Apple-
ton Place, Peabody #552, Vanderbilt University, Nashville, TN 37203.
E-mail: marci.decaro@vanderbilt.edu
Journal of Experimental Psychology: General © 2011 American Psychological Association
2011, Vol. 140, No. 3, 390– 406 0096-3445/11/$12.00 DOI: 10.1037/a0023466
390

Mechanisms of Skill Failure
Why does poor performance sometimes occur in high-pressure
situations? Two different theories have been proposed to answer
this question. Distraction theories, originating from work in aca-
demic testing situations, propose that high-pressure situations
harm performance by diverting individuals’ attention to task-
irrelevant thoughts, such as worries about the situation and its
consequences (Beilock & Carr, 2001; Lewis & Linder, 1997;
Wine, 1971). Pressure essentially creates a dual-task environment
in which situation-related worries compete with the attention
needed to execute the task at hand.
Attention is a key component of working memory (Engle, 2002),
a short-term memory system responsible for actively maintaining
a limited amount of task-relevant information while inhibiting
irrelevant information (Engle, 2002; Miyake & Shah, 1999). Ac-
cording to distraction theories, because high-pressure situations
coopt attentional resources, tasks that rely heavily on working
memory should be most negatively impacted under pressure. This
is exactly what has been found (Beilock & DeCaro, 2007; Gim-
mig, Huguet, Caverni, & Cury, 2006; Markman, Maddox, &
Worthy, 2006). For example, Beilock, Kulp, Holt, and Carr (2004)
demonstrated that math problems heavily dependent on working
memory (i.e., requiring the online maintenance and manipulation
of intermediate problem steps) were solved less accurately in a
high-pressure test compared with a low-pressure test. In contrast,
math problems that were highly practiced and thus could be
directly retrieved from long-term memory (Logan, 1988), circum-
venting demanding computations in working memory, were per-
formed just as well in low- and high-pressure situations.
Although there is evidence that pressure induces failure by
distracting attention away from skill execution, a very different
class of theories has also been put forth to explain skill failure as
well. Explicit monitoring or skill-focus theories suggest that pres-
sure increases self-consciousness about performing correctly,
which in turn leads performers to focus their attention on skill
execution to ensure an optimal outcome (Beilock & Carr, 2001).
Explicit attention to step-by-step processes is thought to disrupt the
learning and execution of proceduralized processes that normally
run outside of conscious awareness (Baumeister, 1984; Beilock,
Bertenthal, McCoy, & Carr, 2004; Beilock & Carr, 2001; Kimble
& Perlmuter, 1970; Langer & Imber, 1979; Masters, 1992).
Support for explicit monitoring theories is found primarily in
proceduralized skills ranging from golf putting (Beilock & Carr,
2001; Lewis & Linder, 1997; Masters, 1992), to hockey dribbling
(Jackson, Ashford, & Norsworthy, 2006), to baseball batting (R.
Gray, 2004). For instance, R. Gray (2004) examined how expert
baseball players batted in a baseball simulator in both low-pressure
and high-pressure conditions. Gray found an increase in batting
errors and movement variability under high pressure, relative to
the low-pressure situation. It is notable that this pressure-induced
batting failure was accompanied by improvement in the players’
ability to judge the direction their bat was moving during skill
execution. Specifically, when batters were asked to judge whether
their bat was traveling upward or downward at the moment an
intermittent tone sounded, these judgments were more accurate
under pressure. These results suggest that, under pressure, the
batters focused explicitly on the components of the batting skill.
This focus of attention led to more accurate judgments about bat
direction but disrupted proceduralized batting processes and hurt
hitting performance overall.
Thus, distraction and explicit monitoring theories of choking
under pressure pose very different mechanisms of skill failure.
Whereas distraction theories suggest that pressure harms perfor-
mance by shifting attention and working memory resources away
from execution, explicit monitoring theories suggest that pressure
shifts too much attention toward skill processes and procedures.
How can pressure do both? One possibility is that pressure coopts
working memory when individuals are performing demanding
cognitive tasks, whereas it induces attention to skill processes
during proceduralized motor skill execution. But it seems odd to
suggest that a high-pressure situation would exert different effects
simply depending on whether one is holding a pencil or a baseball
bat in one’s hands.
We believe the answer to the above question lies in aspects of
the performance environment itself. High-pressure situations may
actually involve multiple components and thus exert multiple
effects, leading to distracting thoughts, explicit monitoring, or
even both, depending on specific elements of the stress imposed.
Whether or not performance will fail, and how this failure will
come about, depends on aspects of the pressure situation and the
attentional demands of the task being performed. In the current
work, we examine whether different types of pressures exert
distinct effects and, moreover, whether these pressures carry dif-
ferent implications for skill success and failure based on the type
of task being performed.
The Pressure Situation
Although most investigations of performance under pressure
have largely ignored the makeup of the pressure situation itself,
real-world pressure situations (and the laboratory pressure manip-
ulations that imitate them) have multiple elements. Individuals
might be watched by a teacher, audience, or video camera; people
may try to obtain a personally important title, scholarship, or
monetary reward; or people may be out for a high test score. Even
though these distinct elements of a high-pressure situation may
elicit feelings of pressure and anxiety, a closer look reveals subtle
differences among them.
For instance, the pressure of being watched by others—which
we refer to as monitoring pressure—may increase attention to skill
processes and procedures, particularly when one’s performance is
evaluated in some manner. This notion is supported by work in
social psychology showing that the presence of an audience, video
camera, or mirror increases self-consciousness or self-awareness
(e.g., Carver & Scheier, 1978; Davis & Brock, 1975; Duval &
Wicklund, 1972; Geller & Shaver, 1976). On the other hand,
pressure induced by offering an incentive if a certain outcome is
achieved—which we refer to as outcome pressure—may serve to
shift performers’ focus of attention to the situation and its conse-
quences. A greater metacognitive awareness of the performance
situation may lead people to simulate different outcome possibil-
ities or think about how they are measuring up during performance
(e.g., “I haven’t even gotten one right so far”; Sarason, 1972, p.
411), diverting attention away from skill execution.
A given pressure situation may therefore differentially empha-
size outcome pressures or monitoring pressures. In many high-
pressure situations, aspects of both may be present (e.g., Beilock,
391
MECHANISMS OF SKILL FAILURE

Kulp, et al., 2004; Beilock & Carr, 2005; Beilock & DeCaro, 2007;
R. Gray, 2004; Hardy et al., 1996), simultaneously disrupting
working memory availability and directing what attention that
remains in ways that are counterproductive. Thus, pressure may
lead to skill failure in multiple ways, depending in large part on
features of the performance environment itself.
We are aware of no previous research that has attempted to tease
apart the pressure situation to explore whether multiple pressure
elements exert systematically different effects on performance.
However, a close examination of one recent study offers some
preliminary support for our ideas. Markman et al. (2006) investi-
gated performance under pressure on category-learning tasks (i.e.,
rule-based and information-integration category tasks) that differ
in their reliance on attentional control. For rule-based category-
learning tasks, individuals must discover an easily verbalizable
rule to categorize stimuli into two groups (e.g., all items with a
particular feature belong to Category A). Doing so involves ex-
plicitly testing various hypotheses about category membership and
therefore relies on working memory and executive attention, sim-
ilar to many of the demanding cognitive tasks (e.g., math problem
solving) explored in the pressure literature to date (DeCaro,
Thomas, & Beilock, 2008; Waldron & Ashby, 2001; Zeithamova
& Maddox, 2006). Information-integration category-learning
tasks, on the other hand, rely on the integration of multiple stim-
ulus dimensions at a predecisional stage. Categories are similarity-
based, and categorization is thought to rely on procedural learning
(i.e., accumulating stimulus–response associations) that occurs
largely outside of attentional control. This task representation is
said to be similar to that of the complex sensorimotor skills (e.g.,
golf putting) studied under pressure (Maddox & Ashby, 2004).
In their work, Markman et al. (2006) looked at both rule-based
and information-integration category learning under pressure. The
authors found that rule-based category learning was worse under
high-pressure compared with low-pressure conditions. However,
the opposite was found for information-integration category learn-
ing: Performance was actually better under high-pressure than
low-pressure conditions. These findings can be readily explained
with the distraction theory of choking: Pressure consumed the
working memory and attention needed for hypothesis testing,
which in turn harmed rule-based category learning. In contrast,
information-integration category learning, relying instead on
similarity-based responses elicited by the procedural learning sys-
tem, benefited when attention was diverted from performance (for
related work in complex sensorimotor skills, see Beilock, Carr,
MacMahon, & Starkes, 2002).
In light of explicit monitoring theories, however, the latter
findings are unexpected: Why doesn’t information-integration cat-
egory learning fail, via pressure-induced explicit monitoring? Ac-
cording to the above hypothesis concerning multiple types of
pressure, the answer has to do with the pressure situation itself—
Markman et al. (2006) only employed an outcome-based incentive
scenario. Individuals were told that, by exceeding a performance
criterion, they could earn money for both themselves and a partner.
There was no aspect of monitoring pressure (e.g., social evalua-
tion, video camera) that we propose would lead to explicit moni-
toring. Had monitoring pressure been present, we anticipate that
information-integration category learning would falter. Such a
result would provide evidence for multiple routes to skill failure
under pressure.
Current Experiments
Across four experiments, we examine whether different types of
pressures exert distinct effects and, moreover, whether these pres-
sures carry different implications for skill success and failure
based on the control structure of the task being performed. Because
the category-learning domain uniquely affords the opportunity to
solicit attention-demanding learning processes in one task and less
attending-demanding processes in a different task, while holding
all other aspects of the learning situation invariant (e.g., same
stimuli, same general learning paradigm), we start in Experiment 1
by showing that categorization tasks similar to the ones used by
Markman et al. (2006) can differentially fall prey to distraction and
explicit monitoring. In Experiment 2 we examine whether outcome
pressure hurts rule-based categorization, in line with distraction
theories of pressure (as did Markman et al., 2006), and monitoring
pressure hurts information integration-integration categorization,
in accordance with explicit monitoring theories of pressure (sim-
ilar to work in proceduralized sensorimotor skills). In Experiment
3 we leverage findings of the first two experiments to examine
how pressure-induced performance decrements can be alleviated
in both types of categorization tasks. Finally, in Experiment 4 we
extend these findings to proceduralized performance in another
domain, examining whether monitoring pressure, as opposed to
outcome pressure, harms skilled performance on the serial reaction
time task (SRTT; Nissen & Bullmeyer, 1987).
Experiment 1
Individuals completed both rule-based and information-
integration category-learning tasks in a baseline single-task con-
dition followed by one of two types of secondary task conditions:
(a) a distracting secondary task designed to divert attention away
from category learning or (b) an explicit monitoring secondary
task intended to prompt attention to the step-by-step components
of learning.
If rule-based category learning relies on attention and working
memory to discover and maintain different hypotheses about cat-
egory membership (Waldron & Ashby, 2001), then it should take
longer to learn this type of category structure when simultaneously
performing a distracting versus explicit monitoring secondary task.
The opposite should occur for information-integration category
learning, where optimal performance is not thought to occur via
attention-demanding hypothesis testing but by processes that make
little demand on attention and working memory (Maddox &
Ashby, 2004). Learning this type of category structure should be
unaffected by a distracting secondary task but impaired when
performing an explicit monitoring secondary task. Such findings
would set the stage in Experiment 2 to test whether there are
systematically different types of pressures and to explore how
these pressures might exert their impact.
Method
Participants. Undergraduate students (N⫽103) at a large
U.S. Midwestern university served as participants (Mage ⫽19.1
years, SD ⫽2.7). Participants had no reported colorblindness.
Individuals were randomly assigned to a distracting secondary task
condition (n⫽46) or an explicit monitoring secondary task
condition (n⫽57).
392 DECARO, THOMAS, ALBERT, AND BEILOCK

Procedure. After giving informed consent, participants com-
pleted the category-learning task individually on the computer (see
below for details). Individuals were instructed to place each stim-
ulus into either Category A or Category B by pressing one of two
marked keys on the computer keyboard. Following each catego-
rization selection, immediate feedback was displayed, with the
word “correct” or “incorrect” appearing directly below the stimu-
lus, until the individual pressed the spacebar to continue to the next
trial. The screen then went blank for a 1,500-ms intertrial interval.
Once participants reached a learning criterion of eight correct
categorization trials in a row or a 200-trial maximum, they exited
that particular category structure (Waldron & Ashby, 2001). They
were then given a rest period during which they were informed that
they would be given a new category structure. Only individuals
who successfully reached this learning criterion prior to the 200-
trial maximum in the single-task baseline block (see below) were
included in the current work. This allowed us to examine the
impact of the experimental manipulations on performance for only
those individuals who demonstrated that they were able to perform
these types of categorization tasks in the first place.
All participants performed the same four category structures,
separated into two blocks. Within each block, individuals saw two
different types of category structures: one rule-based category (R)
and one information-integration category (I; see below). Thus, four
global orders were possible (i.e., RI RI, IR IR, RI IR, IR RI) and
counterbalanced across participants.
1
The specific category struc-
ture (e.g., which rule-based task came first) was randomly selected
without replacement across participants, and the categorization
stimuli within each category structure were randomly selected with
replacement across participants.
The first block served as a single-task baseline. Before the
second block, participants read instructions for either the distract-
ing or explicit monitoring secondary tasks, depending on their
assigned condition. Then the second block was performed concur-
rently with one of the two secondary tasks. After the second block
was complete, individuals completed a series of questionnaires and
were thanked and debriefed.
Category learning task. Stimuli were adapted from Waldron
and Ashby (2001). Each was a square with one or two symbols
embedded within it. Sixteen stimuli were constructed by taking the
factorial combination of four dimensions, two levels each: square/
background color (yellow or blue), embedded symbol shape (circle
or square), symbol color (red or green), and number of embedded
symbols (one or two). The four category structures each used all 16
stimuli but differed in the mapping from stimuli to responses.
Rule-based categories had one relevant dimension, affording an
easily verbalizable rule (e.g., “If the embedded symbol is red,
choose Category A; if the symbol is green, choose Category B”).
Because previous studies (e.g., Waldron & Ashby, 2001) found no
differences in performance depending on the dimension selected to
be relevant, symbol color was randomly chosen as the relevant
dimension for one rule-based structure and symbol shape for the
other.
Information-integration categories involved three relevant di-
mensions and one irrelevant dimension (randomly determined as
background color for one information-integration structure and
number of embedded symbols for the other). The three relevant
dimensions were labeled X,Y, and Z, and each binary value of
these three dimensions was randomly assigned either a ⫺1ora⫹1
(e.g., a red symbol ⫽–1, and a green symbol ⫽⫹1). Stimuli were
then categorized according to the following rule: If value(X)⫹
value(Y)⫹value(Z)⬎0, classify as Category A; otherwise
classify as Category B (Waldron & Ashby, 2001). Because this
rule would be very difficult to derive verbally over a series of
individual learning trials, it is more likely that these dimensional
values are integrated at a predecisional stage, presumably without
access to conscious awareness (Ashby & Maddox, 2005).
Secondary tasks. A letter-monitoring task served as the dis-
tracting secondary task. At the beginning of each trial, a 200-ms
fixation point (a plus sign in the center of the screen) was replaced
by a randomly selected alphanumeric letter appearing for 2,000
ms. Individuals were instructed to press the space bar if the
displayed letter was an “S.” If the letter was not an “S,” they were
instructed to do nothing (i.e., a go/no-go task). The letter “S” was
shown more often than any other individual letter (37.5% of the
time if all 200 trials were completed).
After either the spacebar was pressed or the 2,000-ms time
interval had passed, the word “correct” or “incorrect” replaced the
letter stimulus for 1,000 ms, providing immediate feedback for the
secondary task. Feedback was used to emphasize the importance of
the letter-monitoring task for the participant. The screen then went
blank for 1,000 ms, and the categorization trial resumed as de-
scribed above. Only participants who maintained 90% accuracy on
the letter- monitoring task were included to ensure that participants
were allocating attention to both the primary and secondary tasks.
A confidence-judgment task served as the explicit monitoring
secondary task. This task was intended to induce individuals to
explicitly monitor the component steps of the categorization pro-
cess—how they went about deciding the stimuli should go in a
particular category. At the beginning of each trial, participants
were shown a categorization stimulus along with a confidence-
rating prompt. They were asked to first think about how they were
going to categorize the stimulus and then rate how confident they
were in their category selection, by typing the number correspond-
ing to their confidence rating on a scale ranging from 1 (not at all
confident)to7(extremely confident). After selecting their confi-
dence rating, the prompt disappeared, and individuals categorized
the stimulus as usual.
Questionnaires. Following the category-learning task, partic-
ipants rated how important they felt it was to perform at a high
level during the last two sets of category-learning tasks (i.e., the
secondary task block), on a scale ranging from 1 (not at all
important)to7(extremely important). We only included partici-
pants who responded at the midpoint or higher in our analyses.
Because performance pressure, by definition, only occurs when
individuals feel it is important to perform their best (Baumeister,
1984), reporting at least moderate task importance is often used as
a criterion for study participation in experiments exploring the
choking phenomenon (Beilock & Gray, 2007). Because we imple-
ment this criterion in Experiments 2–4, where pressure is explic-
itly manipulated, for consistency we implement it in Experiment 1
as well. Last, individuals completed a brief demographics ques-
tionnaire and were thanked and debriefed.
1
In both Experiments 1 and 2, preliminary analyses indicated that order
did not moderate the key Condition ⫻Category Structure interaction.
393
MECHANISMS OF SKILL FAILURE

Results and Discussion
Categorization performance. Of primary interest was how
many categorization trials individuals performed before reaching
the learning criterion of eight correct categorization trials in a row.
Trials to criterion were log transformed because of a positive skew
in the distribution (Tabachnick & Fidel, 1996), which is common
for category-learning tasks (e.g., Waldron & Ashby, 2001; DeCaro
et al., 2008). These data are included in Appendix A for all three
experiments. However, because we were interested in understand-
ing how adding a secondary task changes category-learning per-
formance, our dependent variable was a difference score created
by subtracting Block 1 (single-task baseline) log-transformed trials
to criterion from Block 2 (secondary task block) log-transformed
trials to criterion. This was done separately for the rule-based and
information-integration category structures to examine the impact
of the secondary tasks on each type of category structure. Thus,
higher scores on this difference measure indicate worse category-
learning performance (i.e., taking more trials to learn the catego-
ries) during the secondary task block compared with the single-
task baseline.
These difference scores were submitted to a 2 (Category Struc-
ture: rule-based, information-integration) ⫻2 (Secondary Task
Type: distracting, explicit monitoring) analysis of variance
(ANOVA), with the last factor between subjects. There was no
main effect of category structure or secondary task condition
(Fs⬍1). However, there was a significant Category Structure ⫻
Secondary Task Condition interaction, F(1, 101) ⫽4.78, p⫽.03,
MSE ⫽.26.
To understand this interaction, we examined each category
structure separately. As can be seen in Figure 1, for rule-based
category learning, the difference in learning from the baseline to
distracting secondary task block was significantly larger than zero
(M⫽.17, SE ⫽.07), t(45) ⫽2.21, p⫽.03. Rule-based category
learning, heavily reliant on attention and working memory for
optimal performance, was worse when individuals were distracted.
These results are consistent with previous research using an
attention-demanding dual task during rule-based category learning
(i.e., Waldron & Ashby, 2001; Zeithamova & Maddox, 2006).
Simultaneously performing two attention-demanding tasks typi-
cally leads to worse performance than if either task was performed
alone (Baddeley & Logie, 1999). In contrast, rule-based category
learning was not impacted in the explicit monitoring secondary
task—the difference between baseline and explicit monitoring
rule-based learning blocks was essentially zero (M⫽–.03, SE ⫽
.07), t(56) ⫽–.48. Asking individuals to think explicitly about
each rule-based categorization judgment and rate their confidence
in their selection did not impact performance.
In terms of information-integration category learning, the dif-
ference from the single-task baseline to the explicit monitoring
secondary task block was significantly greater than zero (M⫽.17,
SE ⫽.07), t(56) ⫽2.47, p⬍.02. Information-integration category
learning was worse when individuals were asked to make confi-
dence judgments by explicitly attending to the steps of the cate-
gorization process. Such findings are consistent with those of
Maddox, Love, Glass, and Filoteo (2008), who demonstrated that
additional feedback that draws attention to explicit aspects of
categorization can impair information-integration category learn-
ing. These results also resemble findings in complex sensorimotor
skills (Wulf, 2007). Secondary tasks requiring individuals to mon-
itor the processes of performance (e.g., attending to the hands or
feet) disrupt skilled field-hockey dribbling (Jackson et al., 2006),
soccer dribbling (Beilock et al., 2002), and baseball batting (R.
Gray, 2004). Finally, the distracting letter-monitoring task had no
impact on information-integration category learning (M⫽.06,
SE ⫽.08), t(45) ⫽.78. This finding is consistent with work with
proceduralized sensorimotor skills as well. Performing a secondary
tone-monitoring task during hockey or soccer dribbling does not
negatively impact well-learned skill execution (Beilock et al.,
2002; Jackson et al., 2006).
Categorization strategies. We also examined the strategies
individuals used to learn the information-integration categories in
the two secondary task conditions. We focused specifically on the
information-integration task for two reasons. First, although indi-
viduals are generally thought to learn the information-integration
categories to criterion by relying on the integration of multiple
category dimensions at an implicit predecisional stage (Ashby &
Maddox, 2005; Waldron & Ashby, 2001), previous work using the
same category-learning task has shown that, in some circum-
stances, people can successfully learn these categories to a crite-
rion of eight correct categorization trials in a row by employing
simple, one-dimension rules (e.g., “Categorize all items with a
green symbol as Category A”; DeCaro, Carlson, Thomas, &
Beilock, 2009; Tharp & Pickering, 2009). Such simple rules, as
well as more complex explicit rules involving two or three stim-
ulus dimensions, can lead to a relatively high degree of accuracy
(75%–87.5% of the stimuli), and therefore may provide alternative
ways in which participants can attain the learning criterion. How-
ever, using explicit rules can also prevent responding based on the
more optimal, procedural, system and therefore could also lead to
poorer performance. Thus, by modeling strategies, we hoped to
shed light on how individuals were solving the information-
integration task.
Second, if individuals are more likely to pay explicit attention to
performance in the explicit monitoring secondary task condition
relative to a single-task baseline condition, then one might also
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
Category Structure
Distracting Secondary Task
Explicit Monitoring Secondary Task
Rule-B ased In formation -Integratio n
Log Trials to Criterion (Test Minus Baseline)
Figure 1. Mean trials to criterion (log transformed) difference score (test
minus baseline) as a function of category structure and secondary task type.
Error bars represent standard errors.
394 DECARO, THOMAS, ALBERT, AND BEILOCK

expect them to rely more heavily on explicit rules. In contrast,
people given a distracting secondary task should not use explicit
rules more (and may even use them less) than they do during a
single-task baseline condition. Thus, modeling strategy data al-
lowed us to specifically test our prediction that the secondary task
conditions would alter how individuals allocated attention to per-
formance (for response modeling details, see Appendix B).
Figure 2 displays the degree to which participants changed their
strategies during information-integration category learning from
the single-task baseline to each secondary task condition (second-
ary task minus baseline); a higher number represents a greater
proportion of strategy use in the secondary task condition relative
to the single-task condition. As shown in Figure 2, participants in
the explicit monitoring task condition used a greater proportion of
both one-dimension (M⫽.07, SE ⫽.04), t(56) ⫽2.52, p⬍.02,
and two-dimension rules (M⫽.03, SE ⫽.01), t(56) ⫽2.39, p⫽
.02. In addition, the more individuals implemented these rule-
based strategies, the more trials they took to learn the information-
integration task (r
1-Dim.
⫽.89, p⬍.001, r
2-Dim.
⫽.79, p⬍.001).
Participants in the explicit monitoring condition were also less
likely to implement the “optimal” strategy (i.e., categorizing the
stimuli as set up by the experimenters, at 100% accuracy; M⫽
–.11, SE ⫽.04), t(56) ⫽–2.51, p⬍.02. Greater use of the optimal
strategy was associated with taking fewer trials to learn the
information-integration task (r⫽–.93, p⬍.001). No other
changes in response strategies were found for the explicit moni-
toring condition. Moreover, participants in the distracting second-
ary task condition did not change any strategies relative to base-
line.
Thus, explicit monitoring leads to shifts in the strategies used to
perform the information-integration categorization task. Moreover,
because participants in the explicit monitoring condition showed a
greater decrement than participants in the distracting secondary
task condition, these strategy data are also consistent with the idea
that using explicit verbal rules on the information-integration task
hurts learning. Individuals used these rules more often in the
explicit monitoring secondary task condition and performed more
poorly.
The response-modeling data also support the idea that the opti-
mal strategy is indeed less attention-demanding and likely proce-
dural in nature. Use of the optimal strategy decreased in the
skill-focus secondary task condition—the same condition in which
other explicit strategies increased. In addition, the use of the
optimal strategy did not decrease in the distracting secondary task
condition—if the optimal strategy depended heavily on attention
(e.g., a complex rule-plus-exception strategy), then one would
expect the use of this strategy to decrease under distraction. Taken
together, these results provide support for the less attention-
demanding nature of the optimal information-integration strategy.
In conclusion, rule-based category learning was harmed when
individuals were required to perform a distracting secondary task.
In contrast, information-integration category learning was harmed
when individuals performed a task that prompted attention to
execution. These results set the stage to test whether different
high-pressure conditions elicit effects analogous to the distracting
and explicit monitoring secondary tasks.
In Experiment 2, we examine the effects of outcome pressure
versus monitoring pressure on rule-based and information-
integration category learning. If outcome pressure coopts attention
and working memory, then we should find the same performance
pattern as the distracting secondary task condition in Experiment 1.
And to the extent that monitoring pressure increases attention to
performance, category-learning results should parallel those
seen in the explicit monitoring secondary task condition in
Experiment 1.
Experiment 2
Individuals learned rule-based and information-integration cat-
egories under a low-pressure baseline condition followed by either
a low-pressure control condition or one of two high-pressure
conditions. In the outcome-pressure condition, individuals were
told that a 20% performance improvement from baseline would
earn both themselves and a partner a monetary reward. As outlined
in the introduction, we predicted that outcome-pressure would lead
individuals to worry or ruminate about the consequences of their
performance, distracting them from the task at hand. If so, rule-
based category learning should suffer, but information-integration
category learning should not. In the monitoring-pressure condition,
participants were watched and videotaped during the category-
learning task and told that the footage would be viewed by other
students and researchers. We predicted that monitoring pressure
would lead individuals to focus on what others are watching—the
step-by-step task processes they are performing. If so, then
information-integration category learning should suffer relative to
baseline whereas rule-based category learning should not.
Method
Participants. Undergraduate students (N⫽130) at the same
university as in Experiment 1 participated in Experiment 2 (M
age ⫽19.13 years, SD ⫽1.12). Participants had no reported
-.25
-.20
-.15
-.10
-.05
.00
.05
.10
.15
.20
.25
Optimal 1-Dim. 2-Dim. 3-Dim.
Strategy Type
Proportion Strategy Use (Test Minus Baseline)
Dist racti ng Secondary Task
Expl icit Monitoring Secondary Task
Figure 2. Differences in proportion of strategy use (test minus baseline)
during information-integration category learning as a function of each
strategy type (optimal, one-dimension rule, two-dimension rule, and three-
dimension rule) and secondary task condition. Error bars represent standard
errors.
395
MECHANISMS OF SKILL FAILURE

colorblindness. Individuals were randomly assigned to a low-
pressure control condition (n⫽47), an outcome-pressure condi-
tion (n⫽43), or a monitoring-pressure condition (n⫽40).
Procedure. The procedure was exactly the same as in Exper-
iment 1, except the last two category structures (Block 2) were
performed under either a low-pressure control condition or one of
two types of pressure conditions, rather than with secondary tasks.
Low-pressure control condition. After Block 1 (single-task
baseline) was completed, the experimenter returned to the testing
room and explained to participants in the low-pressure control
group that they would continue to perform a category-learning task
and instructed them to try to do their best. The experimenter then
left the room, and participants completed the last block of the
category-learning task.
Outcome-pressure condition. Participants in the outcome-
pressure condition were given a high-pressure scenario before
continuing to Block 2. Specifically, the experimenter explained
that the computer had been calculating a score based on the
participant’s categorization accuracy in the first block, and a 20%
improvement in this score during the next sets of categories would
earn the participant an additional $10 at the end of the study. The
experimenter explained, however, that the study was actually
about teamwork, and both the participant and a “partner” must
improve their scores to earn the money. The partner, participants
were told, had already completed the experiment and improved his
or her score, leaving it up to the present participant to do well for
both individuals to be rewarded. Of course, if their categorization
accuracy did not improve, they were told that neither the partici-
pant nor his or her partner would receive the bonus. After explain-
ing the stakes, the experimenter left the room, and the participant
completed Block 2. At the end of the experiment, participants were
fully debriefed, including the fact that their partner was actually
fictitious, and participants were given the money regardless of
their performance.
Monitoring-pressure condition. Before beginning Block 2,
the experimenter informed participants in the monitoring-pressure
condition that their performance during the next sets of categories
would be videotaped, for students and professors at the university
to watch how people perform this skill. In addition, participants
were told that the footage may also be used in a film about the
basic skills of category learning funded for nationwide distribution
to researchers and psychology classes. The experimenter set up the
camera about1mtotheleft of the participant, so that the
participant and the computer screen were in view, and stayed
behind the camera watching individuals’ performance during the
category learning task. When Block 2 was completed, the exper-
imenter turned off the camera and faced it away from the partic-
ipant. After the experiment was completed, participants were fully
debriefed concerning the purpose of the study and were reassured
that, in fact, no one would be watching the tapes of their perfor-
mance.
Questionnaires. As in Experiment 1, all individuals com-
pleted a question regarding how important it was to them to
perform at a high level on the last two sets of category-learning
tasks. Next, individuals rated how much pressure they felt to
perform at a high level, ranging from 1 (very little performance
pressure)to7(extreme performance pressure). Last, participants
completed a brief demographics questionnaire.
Results and Discussion
Pressure ratings. Ratings of performance pressure were sig-
nificantly higher for individuals in the outcome-pressure (M⫽
4.95, SE ⫽.20) and monitoring-pressure (M⫽5.15, SE ⫽.15)
conditions, compared with the low-pressure control group (M⫽
4.29, SE ⫽.20), t(89) ⫽2.31, p⫽.02, and t(86) ⫽3.27, p⬍.01,
respectively. The outcome-pressure and monitoring-pressure
groups did not differ in their pressure reports, t(81) ⫽.76. Thus,
both pressure conditions served to elevate feelings of performance
pressure, to comparable levels, above that of the low-pressure
control group.
Categorization performance. The dependent measure was
again the number of trials taken to learn the categories to the
criterion of eight correct trials in a row, log transformed. As in
Experiment 1, a difference score was obtained by subtracting
Block 1 (single-task baseline) trials to criterion from Block 2
(high-pressure or control block) trials to criterion. This computa-
tion was done separately for the rule-based and information-
integration category structures to examine the relative impact of
the pressure conditions on each type of category structure. Again,
higher scores indicate worse category-learning performance (i.e.,
taking more trials to learn to criterion) during Block 2 relative to
the single-task baseline.
This difference score was examined as a function of category
structure and pressure condition. A 2 (Category Structure: rule-
based, information-integration) ⫻3 (Pressure Condition: low-
pressure control, outcome-pressure, monitoring-pressure) mixed
ANOVA revealed no main effects of category structure (F⬍1) or
pressure condition (F⬍1). However, a significant Category
Structure ⫻Pressure Condition interaction was found, F(2, 127) ⫽
4.75, p⫽.01, MSE ⫽.21.
As shown in Figure 3, the low-pressure control condition did not
alter either rule-based (M⫽.04, SE ⫽.05), t(46) ⫽1.14, ns,or
information-integration (M⫽.03, SE ⫽.08), t(27) ⫽.51, ns,
category learning—the changes from baseline (i.e., the difference
scores) were essentially zero.
-.4
-.3
-.2
-.1
.0
.1
.2
.3
.4
Category Structure
Control
Outcome-Pressure
Monitoring-Pressure
Rule-Based Information-Integration
Log Trials to Criterion (Test Minus Baseline)
Figure 3. Mean trials to criterion (log transformed) difference score (test
minus baseline) as a function of category structure and pressure type. Error
bars represent standard errors.
396 DECARO, THOMAS, ALBERT, AND BEILOCK

Under high-pressure conditions, rule-based category learning
was hurt by outcome pressure (M⫽.16, SE ⫽.06), t(42) ⫽2.93,
p⬍.01, but not monitoring pressure (M⫽.03, SE ⫽.06), t(39) ⫽
.40. Outcome pressure was created by informing individuals that a
monetary reward, for both themselves and another person, was
contingent on their performance. Given that rule-based category
learning only suffered in this pressure condition, and only via a
distracting secondary task in Experiment 1, these findings suggest
that attention was diverted away from performance during the
outcome-pressure condition. This result is consistent with the
distraction theory of choking under pressure.
The opposite pattern was found for information-integration cat-
egory learning, which was unaffected by outcome pressure (M⫽
–.11, SE ⫽.08), t(42) ⫽–1.20, ns, but harmed by monitoring
pressure (M⫽.19, SE ⫽.08), t(39) ⫽2.51, p⬍.02. In the
monitoring-pressure condition, individuals were watched by an
experimenter who videotaped their performance, nominally so that
others could watch how people learn new categories. Information-
integration category learning was only hurt by this type of pres-
sure, and the findings were analogous with the explicit monitoring
secondary task condition in Experiment 1. These findings are
consistent with the explicit monitoring theory of choking under
pressure. It appears that, under the watchful eye of others, indi-
viduals focus more explicitly on the steps of the skill being
performed. This skill monitoring disrupts performance that oper-
ates best outside of explicit attentional control.
Categorization strategies. As in Experiment 1, we also
examined the relative use of explicit rules during information-
integration category learning between Block 1 (low-pressure base-
line) and Block 2 (high-pressure or low-pressure control) by com-
puting difference scores (Block 2 minus Block 1) for the
proportion of optimal, and one-, two-, and three-dimension strat-
egies (see Appendix B). As shown in Figure 4, participants used
more two-dimension (M⫽.04, SE ⫽.02), t(39) ⫽2.14, p⬍.04,
and three-dimension (M⫽.03, SE ⫽.01), t(39) ⫽2.45, p⬍.02,
strategies under monitoring pressure. In addition, the more indi-
viduals used these rule-based strategies, the more trials they took
to reach the learning criterion (r
2-Dim.
⫽.83, p⬍.001; r
3-Dim.
⫽
.69, p⬍.001). The optimal strategy was also used less often (M⫽
–.11, SE ⫽.05), t(39) ⫽–2.25, p⫽.03, and was associated with
quicker learning (r⫽–.92, p⬍.001). Although somewhat higher,
one-dimension rule use did not significantly increase under mon-
itoring pressure (M⫽.05, SE ⫽.03), t(39) ⫽1.69, p⫽.10.
Outcome pressure had little impact on response strategies. One-
dimension rules tended to be used less often (M⫽–.06, SE ⫽.03),
t(42) ⫽–1.90, p⫽.06, which, in conjunction with the unimpaired
performance seen in this condition, aligns with the idea that
performance on the information-integration task is better when
individuals do not use these rules. No other strategies changed
significantly under outcome pressure. Strategies remained un-
changed in the low-pressure control condition as well. These
results are consistent with the prediction that monitoring pressure
increases explicit attention to the processes of category learning,
leading to an increase in the use of explicit rule-based strategies.
When people adopt explicit strategies to perform a task that
operates best outside of explicit attentional control, performance
suffers.
In conclusion, different types of tasks failed—and thrived—
under different types of pressure situations. These findings under-
score the idea that performance situations can be composed of
various elements that impact working memory and attentional
control in multiple ways. Revealing predictable relationships be-
tween pressure type and task type not only allows us to integrate
disparate theories of skill failure but it also enables us to develop
systematic ways to aid performance under pressure. We do this in
Experiment 3.
Experiment 3
In Experiment 3, we specifically focused our investigation on
those pressure/task combinations shown to be most detrimental to
performance in Experiment 2: Rule-based category learning under
outcome pressure, and information-integration category learning
under monitoring pressure. We added the distracting and explicit
monitoring secondary tasks (from Experiment 1) to these pressure/
task combinations as a means to demonstrate that knowledge of (a)
how a particular type of pressure impacts performance and (b) the
control structure of a task can be used to inoculate people against
choking under pressure.
If both outcome pressure and a distracting secondary task serve
to divert attention and working memory from performance, then
rule-based category learning should be impaired when both of
these conditions are combined. However, if an explicit monitoring
secondary task that prompts people to attend closely to the steps of
performance is added to rule-based categorization under outcome
pressure, then performance decrements should be counteracted (at
least in part).
On the other hand, if both monitoring pressure and an explicit
monitoring secondary task act to enhance attention toward perfor-
mance, then information-integration category learning should be
harmed when these conditions occur simultaneously. However, a
distracting secondary task that continually diverts attention away
-.25
-.20
-.15
-.10
-.05
.00
.05
.10
.15
.20
.25
Optimal 1-Dim. 2-Dim. 3-Dim.
Strategy Type
Proportion Strategy Use (Test Minus Baseline)
Outcome-Pressure
Monitoring-Pressure
Control
Figure 4. Differences in proportion of strategy use (test minus baseline)
during information-integration category learning as a function of each
strategy type (optimal, one-dimension rule, two-dimension rule, and three-
dimension rule) and pressure condition. Error bars represent standard
errors.
397
MECHANISMS OF SKILL FAILURE

from task procedures may prevent monitoring pressure from taking
such a toll. Demonstrating the validity of the above predictions
would not only provide further support for multiple types of
pressure but it would also provide useful clues for counteracting
pressure’s negative effects. Thus, we can demonstrate not only
how certain triggers in the performance environment impact atten-
tion but also how we can redirect attention and reclaim perfor-
mance.
Method
Participants. Undergraduate students (N⫽37) at the same
university as in Experiments 1 and 2 served as participants (M
age ⫽18.97 years, SD ⫽.73). Participants had no reported
colorblindness. Individuals were randomly assigned to either the
outcome-pressure/rule-based category-learning condition (n⫽15)
or the monitoring-pressure/information-integration category-
learning condition (n⫽22).
Procedure. Participants were introduced to the same
category-learning task as in Experiments 1 and 2, with a few
exceptions. In the current experiment, each participant completed
either three rule-based or three information-integration category
structures. Recall that the stimuli consisted of four total dimen-
sions: symbol color, symbol shape, background color, and number
of embedded symbols. For each rule-based structure, the salient
dimension for each set in Experiment 3 was randomly determined,
without replacement, from these four possibilities. For each
information-integration structure, the irrelevant dimension was
randomly selected from the four total dimensions.
Participants first completed a low-pressure single-task base-
line set, followed by two high-pressure sets. The two high-
pressure manipulations were exactly the same as in Experiment
2. In the outcome-pressure condition, participants were in-
formed by the experimenter that they could earn $10 for them-
selves and a partner if they improved their categorization ac-
curacy by 20%. Participants in the monitoring-pressure
condition were monitored by the experimenter, who stood be-
hind a video camera under the auspices that the film could be
watched by students and professors at their university and
across the country.
Each set performed under high pressure was also performed
concurrently with one of the two secondary tasks used in Exper-
iment 1: a distracting and an explicit monitoring secondary task, in
counterbalanced order. As in Experiment 1, the distracting task
was a letter-monitoring task, in which participants viewed a letter
between every categorization trial and pressed the space bar on the
computer keyboard if the letter was an “S.” The explicit monitor-
ing task was a confidence-rating task, in which individuals were
asked to rate how confident they were that the category they were
about to select was correct on a 7-point scale ranging from 1 (not
at all confident)to7(very confident).
Following the categorization task, participants were given the
same importance, pressure, and demographics questionnaires as in
Experiment 2. Last, they were thanked and debriefed.
Results and Discussion
Pressure ratings. As in Experiment 2, participants rated
feelings of performance pressure as equivalent between the two
high-pressure groups: outcome pressure (M⫽5.07, SE ⫽.41),
monitoring pressure (M⫽4.96, SE ⫽.31; F⬍1). These ratings
were very similar to those reported by the pressure groups in
Experiment 2.
Categorization performance. As in the previous two exper-
iments, the number of trials taken to learn rule-based and
information-integration categories to criterion (eight correct in a
row) was measured for each set and log transformed, and differ-
ence scores were computed by subtracting the low-pressure set
baseline score from each high-pressure set score. These scores
were examined in a 2 (Category Structure/Pressure Condition:
rule-based/outcome-pressure, information-integration/monitoring-
pressure) ⫻2 (Secondary Task Condition: distracting, explicit
monitoring) mixed ANOVA, with secondary task condition
within-subjects. This analysis revealed no main effects of cate-
gory structure/pressure condition (F⬍1) or secondary task
condition (F⬍1). But a Category Structure/Pressure Condi-
tion ⫻Secondary Task interaction was found, F(1, 35) ⫽7.22,
p⫽.01, MSE ⫽.09.
As shown in Figure 5, rule-based categories were learned more
slowly during a combination of outcome pressure and distracting
secondary task conditions, relative to a low-pressure, single-task
baseline—the difference score was significantly greater than zero
(M⫽.22, SE ⫽.12), t(14) ⫽2.35, p⬍.04. However, when
rule-based category learning was performed with an explicit mon-
itoring secondary task under outcome pressure, learning was no
longer significantly impacted by pressure (M⫽.10, SE ⫽.11),
t(14) ⫽1.24, ns. Instructing individuals to think about how they
were going to categorize each stimulus counteracted outcome
pressure effects.
Information-integration category learning was slowed during
simultaneous monitoring pressure and explicit monitoring second-
ary tasks relative to baseline performance (M⫽.24, SE ⫽.09),
t(21) ⫽2.20, p⬍.04. However, information-integration category
learning was not harmed under monitoring pressure when learning
was coupled with a distracting secondary task designed to redirect
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
Category Structure
Distracting Secondary Task
Explicit Monitoring Secondary Task
Rule-Based
(Outcome-Pressure)
Information-Integration
(Monitoring-Pressure)
Log Trials to Criterion (Test Minus Baseline)
Figure 5. Mean trials to criterion (log transformed) difference score (test
minus baseline) as a function of secondary task type and category structure/
pressure type (rule-based tasks were performed under outcome pressure;
information-integration tasks were performed under monitoring pressure).
Error bars represent standard errors.
398 DECARO, THOMAS, ALBERT, AND BEILOCK

the focus of attention away from the online steps of performance
(M⫽–.01, SE ⫽.10), t(21) ⫽–.09, ns.
Categorization strategies. Response strategies for
information-integration category learning were consistent with
these findings. As shown in Figure 6, in the explicit monitoring
secondary task condition (under monitoring pressure), participants
increased their use of explicit one-dimension rules (M⫽.08, SE ⫽
.04), t(21) ⫽2.39, p⬍.03. Moreover, greater use of these rules
was associated with worse category learning (r⫽.84, p⬍.001).
Participants in this condition also used the optimal strategy less
(M⫽–.14, SE ⫽.07), t(21) ⫽–2.06, p⫽.05, and greater use of
the optimal strategy was associated with better learning (r
.
⫽–.91,
p⬍.001). No other significant strategy changes were found for
this or the distracting secondary task condition. It appears that
combining monitoring pressure with a secondary task that in-
creases attention to the task increases use of explicit, verbal strat-
egies. However, adding a distracting secondary task helps prevent
the increased use of these suboptimal strategies.
These findings enable us to dissociate further the attentional
mechanisms at play in these secondary task and pressure types by
demonstrating their opposing effects. Crossing situations that dis-
tract with those that enhance attention toward a skill seems to
counteract the negative impact that would otherwise occur. These
findings also offer promise for interventions aimed to alleviate
performance pressure. When a performance situation tends to
distract a performer from an attention-demanding task, perhaps a
method to redirect attention back to the steps of performance will
help (cf. DeCaro, Rotar, Kendra, & Beilock, 2010). On the other
hand, when a performance situation leads individuals to focus
explicitly on the component processes of a proceduralized skill, an
intervention designed to mildly distract performers may prove
beneficial.
Experiment 4
In Experiments 1–3, we established that skill failure depends on
both aspects of the performance environment and the attentional
demands of the task being performed. In Experiment 4, we con-
ducted a final experiment to replicate and extend our findings to
another skill domain. In particular, some people may be concerned
about the comparability of the information-integration task used in
Experiments 1–3 with sensorimotor skills that have formed the
basis of much of the choking literature to date. Specifically,
performance of the information-integration task relies on the pro-
cedural learning system (i.e., is largely procedural at the outset of
learning this task; Ashby & Maddox, 2005), whereas sensorimotor
skills typically become less reliant on explicit attentional control
and more proceduralized with increasing practice (Anderson,
1982). Thus, in Experiment 4 we investigate whether monitoring
pressure selectively disrupts a skill that has become largely pro-
ceduralized with practice, much like the well-learned skills that are
often examined under performance pressure. Because of their
similar reliance on the procedural system, we expect to see the
same pattern of performance as with the information-integration
category-learning task.
Participants were trained on the SRTT, in which they learned to
press four keys on the computer keyboard in response to shifting
probes on the computer screen. Unknown to participants, often-
times these probes repeated a regular sequence (not unlike, for
example, a simple sequence on a piano keyboard). Like other
sensorimotor skills, the SRTT is believed to rely initially on both
attention-demanding (declarative) and nondemanding (procedural)
processes to learn the associations between subsequent key-presses
(Robertson, 2007). However, over time (even within the course of
a single learning session), the demand on attention decreases, and
pressing the keys in sequence relies more on procedural processes,
largely outside of explicit attentional control (Nissen & Bulle-
meyer, 1987).
Because the SRTT is a well-studied sensorimotor task that, like
other sensorimotor skills, increases in reliance on the procedural
system with increasing skill (Anderson, 1982), we expected per-
formance to be affected in much the same way as the information-
integration task used in Experiments 1–3. Specifically, after train-
ing on the SRTT, we examined performance in either a low-
pressure control condition or one of the two high-pressure
conditions used in Experiments 2 and 3 (i.e., outcome pressure and
monitoring pressure). We expected worse SRTT performance un-
der monitoring pressure relative to both outcome pressure and the
low-pressure control condition.
Method
Participants. Participants were right-handed undergraduate
students (N⫽65) at the same university as in Experiments 1–3 (M
age ⫽19.15 years, SD ⫽.87). Individuals were randomly assigned
to either the outcome-pressure condition (n⫽24), monitoring-
pressure condition (n⫽21), or low-pressure control condition
(n⫽20).
Procedure.
SRTT training. Participants were asked to complete the
SRTT by placing their dominant right hand on four horizontally
adjacent marked keys on the computer keyboard. Participants were
-.25
-.20
-.15
-.10
-.05
.00
.05
.10
.15
.20
.25
Optimal 1-Dim. 2-Dim. 3-Dim.
Strategy Type
Proportion Strategy Use (Test Minus Baseline)
Dist racting S econdary Task
Expl icit Monitori ng Secondary Task
Figure 6. Differences in proportion of strategy use (test minus baseline)
during information-integration category learning as a function of each
strategy type (optimal, one-dimension rule, two-dimension rule, and three-
dimension rule) and secondary task/monitoring-pressure condition. Error
bars represent standard errors.
399
MECHANISMS OF SKILL FAILURE

told that they should press one of the four keys every time they see
a square appear on the screen and that the location of the square on
the screen would indicate which key to press. Individuals were told
to press the keys as quickly as possible and that the square would
remain on the screen until the correct key was pressed. Following
instructions, participants completed two 50-trial practice blocks in
random order, to acquaint themselves with the key-pressing pro-
cedure. Next, participants completed eight training blocks, during
which a 12-item second-order conditional sequence (1-4-3-1-2-4-
2-3-2-1-3-4) was repeated 12 times (eight times in the introductory
block). Before and after each training block were 24 random trials
(48 in the introductory block). Thus, during training, participants
completed a total of 1,536 trials (1,104 sequence trials and 432
random trials). As is commonly done with the SRTT, participants
were never told that they were learning a sequence, to support the
procedural nature of this skill.
SRTT test. Following SRTT training, all participants com-
pleted a test block including eight repetitions of the sequence,
preceded and followed by 48 random trials (for a total of 192
trials). The test block was completed under either a low-pressure
control condition or one of two high-pressure conditions, all of
which were nearly identical to those in Experiment 2. Participants
in the low-pressure control condition were informed that they
would be completing the same task they had been doing and asked
to try to do their best. Participants in the outcome-pressure con-
dition were informed that they needed to improve both their speed
and accuracy by 20% in order to earn $10 for themselves and a
partner. Participants in the monitoring-pressure condition were
videotaped and told that the footage may be watched by students
and professors interested in research on basic skill learning.
Data analysis. All incorrect responses were removed from
analyses, and five participants were excluded from analyses be-
cause they had less than 80% accuracy during training (one in the
control condition, two in the outcome-pressure condition, and two
in the monitoring-pressure condition). Any reaction time (RT)
greater than three standard deviations from a participant’s median
for the trial type (sequence or random) was also removed. Because
we were interested in comparing differences in performance across
conditions on the test block, we computed a measure of relative
skill (rSRT; Galea, Albert, Ditye, & Miall, 2010): the difference
between medians for the random and sequence trials, divided by
the median of the random trials ([random ⫺sequence] / random).
This skill score (rSRT) accounts for individual differences in time
to respond to random stimuli and can range from 0 (indicating that
random and sequence trials were performed at the same speed; i.e.,
no learning of the sequence) to 1 (indicating the highest possible
amount of sequence learning, which is approached as time to
respond to the sequence approaches zero). To ensure that carryover
effects from the preceding trial type did not influence results (e.g.,
taking inordinately longer to respond to random trials immediately
following a series of sequence trials could inflate the differences
between these trial types), only the second half of the trials from
the random and sequence blocks were included in this skill score
(Brown & Robertson, 2007).
Questionnaires. Following the SRTT, participants were
given the same importance and pressure questionnaires as in
Experiments 2–3. They also completed the State version of the
State-Trait Anxiety Inventory (STAI; Spielberger, Gorsuch, &
Lushene, 1970) as an additional measure of felt anxiety during the
test block. Finally, they completed a brief demographics question-
naire and were thanked and debriefed.
Results and Discussion
Perceptions of pressure. Reports of state anxiety (STAI)
differed between low- and high-pressure conditions. Compared
with the low-pressure control condition (M⫽36.56, SE ⫽2.55),
participants reported feeling more anxiety in the outcome-pressure
condition (M⫽41.71, SE ⫽2.36, 95% confidence interval [CI] ⫽
36.99–46.44, d⫽0.45) and the monitoring-pressure condition
(M⫽41.83, SE ⫽2.55, CI ⫽38.73–46.94, d⫽0.56), which did
not differ. Although they were in the expected direction, ratings of
performance pressure did not significantly differ between the high-
pressure conditions (outcome-pressure M⫽4.96, SE ⫽.28;
monitoring-pressure M⫽5.26, SE ⫽.27) and the low-pressure
control condition (M⫽4.68, SE ⫽.27, CI ⫽4.09–5.28).
SRTT performance. We first verified that participants
learned the sequence by examining their median RT on test-block
sequence trials compared with random trials. As expected, partic-
ipants overall were faster to respond to sequence trials during the
test block (M⫽418.40, SE ⫽7.43) than to random trials (M⫽
458.22, SE ⫽7.45), t(59) ⫽–9.08, p⬍.001, indicating that they
were indeed skilled at this task.
We next compared performance on the test block across the
three conditions, using the skill score described above (rSRT).
Overall, condition had a marginal effect on skill, F(2, 59) ⫽2.98,
p⬍.06. As shown in Figure 7, planned contrasts revealed that
participants in the monitoring-pressure condition (M⫽.05, SE ⫽
.02) performed worse on the test block than participants in the
control condition (M⫽.10, SE ⫽.02), t(57) ⫽–2.00, p⫽.05, and
the outcome-pressure condition (M⫽.10, SE ⫽.02), t(57) ⫽
–2.24, p⬍.03. The latter two conditions did not differ, t(57) ⫽
.17, ns. Again, higher skill scores as measured by the rSRT
indicate better SRTT performance during the test block. Thus, we
see that different types of pressure had selective effects on the
trained SRTT—a sensorimotor skill that operates largely outside
of conscious awareness (Nissen & Bullemeyer, 1987). The pres-
0
0.05
0.1
0.15
Condition
Skill Sc ore
Control Outcome-
Pressure
Monitoring -
Pressure
Figure 7. Serial Reaction Time Task skill score as a function of pressure
type. Higher skill scores represent better task performance during the test
block. Error bars represent standard errors.
400 DECARO, THOMAS, ALBERT, AND BEILOCK

sure associated with being monitored by others significantly
harmed performance relative to a low-pressure control condition,
whereas pressure associated with a monetary outcome for oneself
and another person did not affect performance.
These findings are consistent with those of information-
integration category learning in Experiments 2–3, a task also
believed to rely on procedural control processes. Thus, the selec-
tive effects of monitoring pressure appear to extend beyond the
category learning domain to performance of a complex sensori-
motor skill that is thought to become proceduralized with practice.
Again, we see that, by understanding the processes supporting
different skills, we can not only better understand how these skills
might fail but also in what circumstances.
General Discussion
Situation-induced performance decrements occur across a vari-
ety of skill domains under a diverse set of situations. However, not
everyone fails under what might typically be considered a high-
pressure situation. In the current work, we examined when and for
what reasons people fail versus succeed under pressure. By exam-
ining tasks that differ in their reliance on working memory and
attentional control under two different types of high-pressure sit-
uations, we found evidence that pressure can impair performance
in multiple ways.
Pressure to attain a particular performance-based outcome
harmed a skill that relies on working memory and attention (i.e.,
rule-based category learning) but not skills less dependent on
executive control (i.e., information-integration category learning
and skilled SRTT performance). These findings suggest that out-
come pressure coopts attention and working memory resources,
perhaps with distracting worries or ruminations about perfor-
mance. Thus, the distraction mechanism of choking under pressure
may be most likely to occur when attention-demanding tasks are
performed under pressure that emphasizes an important score,
grade, or monetary reward.
Pressure involving performance monitoring (e.g., by another
person and a video camera) disrupted information-integration cat-
egory learning and SRTT performance but not rule-based category
learning. Online evaluation of performance by others may lead
individuals to attend to step-by-step skill execution because they
expect others are doing so as well. Thus, explicit monitoring
theories of choking under pressure are also supported by our
findings. Under the pressure of being watched and evaluated by
others, skills that operate best outside of close attentional control
(e.g., procedural skills) suffer from a performer’s enhanced atten-
tion to performance processes.
Given these findings, both distraction and explicit monitoring
theories of choking under pressure seem to be correct. Whether
attention is diverted from and/or enhanced toward the task at hand
depends in large part on characteristics of the performance situa-
tion one is facing. Moreover, whether performance fails because of
this situation depends also on the attentional demands of the task
being performed. Our findings offer an important step toward
reconciling these seemingly disparate theories of skill failure. With
the multiple mechanisms of performance failure more clearly
delineated, we can not only better understand when and how
choking under pressure may occur, but we are also in a better
position to design interventions to alleviate the negative effects of
performance pressure (Baumeister & Showers, 1986).
In Experiment 3 we revealed one way to mitigate the negative
impact of pressure, by setting up secondary tasks that counteract
pressure’s harmful impact. We asked individuals under outcome
pressure to complete rule-based category learning concurrently
with a secondary task involving explicit monitoring. Whereas
outcome pressure seems to divert attention and working memory
away from performance, frequently prompting people to think
explicitly about the processes they are using to perform the skill
inoculated them against the ill effects of stress (cf. DeCaro et al.,
2010). The opposite intervention helped in a monitoring-pressure
situation. Mildly distracting people during performance made
information-integration category learning immune to pressure’s
negative effects. This type of distraction may have kept individuals
from overthinking a skill that operates best with little explicit
control.
Classifying Performance Situations
The current work demonstrates that the performance environ-
ment can influence how a skill is performed. Moreover, by better
understanding when distraction and/or explicit monitoring may
come about under stress, we are in a better position to link
performance under pressure to other high-stress situations where
performance may go awry.
Multifaceted high-pressure situations. The two high-
pressure situations studied in the current work were designed to
isolate pressures related to concerns over performance outcomes
versus performance process. However, high-pressure situations are
often composed of both of these elements, such as when an athlete
vies for an important title (and possible monetary award) in front
of a large audience. Under these type of combined pressure situ-
ations, participants perform more poorly on attention-demanding
math skills (Beilock & Carr, 2005; Beilock & DeCaro, 2007;
Beilock, Kulp, et al., 2004) and perform worse at proceduralized
golf putting and baseball batting skills (Beilock & Carr, 2001; R.
Gray, 2004). One might wonder, however, how this can be the case
if these two types of pressure exert seemingly opposite effects.
One possibility is that both types of pressure actually impact
performance at the same time. For example, working memory
capacity could be reduced in these situations, but what attention is
still available may be explicitly devoted to step-by-step control.
Thus, attention-demanding skills (e.g., difficult math problem
solving) would suffer from the reduction in working memory
capacity, whereas proceduralized skills (e.g., well-learned senso-
rimotor skills) would be impaired by the increase in explicit
monitoring.
An alternative possibility is that the impact of one type of stress
may actually serve to lessen the impact of the other. Much like the
results of Experiment 3, in which a secondary task that impacts
attention in the opposite way as pressure reduces pressure’s impact
(e.g., an explicit monitoring secondary task performed under out-
come pressure), people could become less distracted under out-
come pressure if they are also facing the pressure of being explic-
itly monitored (and vice versa).
A third possibility is that people are most affected by the aspect
of the pressure situation that is most salient to them. If an indi-
vidual is most concerned about earning high marks, for example,
401
MECHANISMS OF SKILL FAILURE

then he or she may suffer more from distraction than explicit
monitoring, even if someone is watching. Although the current
work has dissociated aspects of the pressure situation that differ-
entially impact attentional control, whether and how these atten-
tional mechanisms interact in multifaceted pressure situations re-
mains an important empirical question.
Stereotype threat. As with the high-pressure situations in-
vestigated in the current work, other situations in which a person-
ally important ability in a domain is questioned, such as conditions
of stereotype threat, also often lead to performance decrements
(Steele, 1997). Introducing a negative stereotype about one’s
social group impairs performance across a variety of domains,
including proceduralized skills (e.g., Beilock, Jellison, Rydell,
McConnell, & Carr, 2006; Chalabaev, Sarrazin, Stone, & Cury,
2008) and attention-demanding tasks (e.g., Beilock, Rydell, &
McConnell, 2007; Schmader & Johns, 2003). There is evidence
for both distraction and explicit monitoring mechanisms of skill
failure under stereotype threat, yet the literature is inconclusive
regarding when each impacts performance (see Schmader, Johns,
& Forbes, 2008; Steele, Spencer, & Aronson, 2002, for a review).
Drawing from the current findings, one might speculate that
aspects of the stereotype-threat situation may influence whether
attention is diverted from and/or enhanced toward skill perfor-
mance. Stereotype threat is typically brought about in the labora-
tory by highlighting the diagnosticity of a test regarding one’s
aptitude in a domain or intelligence in general (Steele et al., 2002).
On some occasions, performance is also measured and watched by
another person (e.g., Beilock et al., 2006; Chalabaev et al., 2008).
The emphasis on a performance outcome may lead attention and
working memory to be disrupted during performance, whereas
explicit monitoring may play a bigger role when a stereotype is
made salient and then monitored by others. Thus, as in the current
work, the mechanisms of skill failure in stereotype-threat situa-
tions may also depend on aspects of the performance situation
itself, in addition to the type of task being performed. Viewing
stereotype threat in terms of the situational impact on working
memory and attentional control may offer new insight into how
(and when) stereotype threat will impact performance and shed
light on new interventions for counteracting its impact (e.g., setting
up secondary tasks to refocus attention optimally).
Test anxiety. Related work on test anxiety has suggested that,
like distraction theories of choking under pressure, performance
anxiety interferes with working memory processes critical for
successful test performance (e.g., Ashcraft, 2002; Ashcraft & Kirk,
2001; Eysenck & Calvo, 1992; Hayes, Hirsch, & Mathews, 2008;
Rapee, 1993; Wine, 1971). Test-anxious individuals have more
worries and intrusive thoughts, particularly in situations with im-
portant contingencies, such as reward, test, or ego-involving situ-
ations (e.g., “This is an intelligence test”; Eysenck & Calvo, 1992,
p. 421; Ikeda, Iwanaga, & Seiwa, 1996). The precise nature of
these distractions is typically thought to be of worry over evalu-
ation and concern over the level of performance relative to that
required (Eysenck & Calvo, 1992; Sarason, 1972). Such self-
imposed pressure is much like the outcome pressure in the current
work—test-anxious individuals become distracted by the potential
outcomes of performance and are most negatively impacted on
tests relying heavily on attention and working memory (e.g.,
inductive reasoning; Calvo, 1985). This body of literature can thus
be easily connected with the present work, adding the idea that
certain individual differences (i.e., test anxiety) may lead some
people to be more sensitive to performance-based outcomes than
others.
Regulatory focus. Other work has linked the pressure and
regulatory focus literatures, suggesting that individuals in high-
pressure situations become sensitive to the potential for losses in
the environment (i.e., prevention focus), as opposed to being in
low-pressure situations where people are simply compensated for
participating in the research study (i.e., promotion focus). Worthy,
Markman, and Maddox (2009) posited that this regulatory focus
would interact with the rewards structure of the task, where a gains
rewards structure entails accumulating points for correct re-
sponses, and a losses rewards structure involves losing fewer
points for correct compared to incorrect responses.
Worthy et al. (2009) found that rule-based category learning was
best in regulatory fit conditions (i.e., high-pressure with a losses
reward structure, low-pressure with a gains reward structure) and
worse in regulatory mismatch conditions (i.e., high-pressure with
a gains rewards structure, low-pressure with a losses reward struc-
ture). Information-integration category learning showed the oppo-
site pattern (i.e., worse performance in fit than mismatch condi-
tions). Worthy et al. suggested that regulatory fit may lead to a
sense of “feeling right,” which increases confidence in perfor-
mance, whereas regulatory mismatch decreases such confidence,
potentially leading to anxiety that reduces working memory re-
sources.
Worthy et al.’s (2009) findings coincide with some aspects of
the current work, namely by showing that pressure can lead to
failure or success with different types of tasks because of the
availability of attentional resources during performance. Specifi-
cally, using Worthy et al.’s framework, one might presume that our
outcome-pressure condition (the same pressure manipulation as
Worthy et al.) induced a prevention focus, interacting with our
“gains” reward structure to invoke a regulatory mismatch. This
regulatory mismatch coopts working memory and attention, con-
sistent with the distraction theory of choking under pressure.
However, our monitoring-pressure condition produced the op-
posite effect, possibly because it lead to a promotion focus instead.
According to Worthy et al. (2009), this promotion focus, coupled
with the gains reward structure of our task, should not impair
rule-based categorization but, instead, harm information-
integration performance, which is exactly what we found. We
admit that the above classification seems a bit arbitrary, as some-
one given a monitoring-pressure condition could just as easily
wish to avoid a negative evaluation as they could desire to earn
high regard by those watching. Similarly, outcome pressure could
prompt individuals to either attend to the potential monetary and
social gains or to the potential losses. It is notable that the stereo-
type threat literature is also inconsistent regarding whether stereo-
type threat leads to promotion or prevention focus (e.g., Chalabaev
et al., 2008; Grimm, Markman, Maddox, & Baldwin, 2009; Keller
& Dauenheimer, 2003; Seibt & Förster, 2004). Nonetheless, if
prevention focus is more likely to come about in situations that
highlight a performance-contingent outcome, whereas monitoring
pressure is more likely to elicit promotion focus, then the findings
of the current experiments fit nicely with the work on regulatory fit
under stress. Our work also extends this previous research by
demonstrating that explicit monitoring may occur in regulatory fit
conditions, whereas distraction may occur in regulatory mismatch
402 DECARO, THOMAS, ALBERT, AND BEILOCK

situations, and these differences in attentional control may, in turn,
be what drives performance success versus failure. Future research
addressing these issues is needed.
Conclusion
The current work demonstrates that attention can be diverted or
enhanced because of such factors as whether a personally impor-
tant performance-based incentive is at stake or whether perfor-
mance is monitored by other people. Moreover, such changes in
attentional control have different effects on skills that rely more or
less on this important cognitive resource. By denoting certain
aspects of the pressure environment that may lead individuals to
focus on the process of performance versus the outcome of per-
formance, we are not only in a position to predict when perfor-
mance will fail or succeed but we can also provide interventions to
help mitigate the possibility of failure. Thus, this work enables us
to better understand performance failure—and ways to prevent
it—across a variety of skill types and situations, from a student
taking a math test to an expert on the playing field.
This work joins a recent body of literature seeking to understand
the interplay between cognition, motivation, and emotion (e.g.,
Beilock, 2008; J. R. Gray, 2004; J. R. Gray, Braver, & Raichle,
2002; Grimm et al., 2009). The present findings align with, and
potentially inform, research across several related areas (e.g., ste-
reotype threat, test anxiety, regulatory focus). By focusing on the
specific cognitive mechanisms by which pressure can exert its
impact, we can begin to cut across such domains, working toward
an overarching theory of when performance will fail versus suc-
ceed under stressful situations.
References
Anderson, J. R. (1982). Acquisition of cognitive skill. Psychological Re-
view, 89, 369–406. doi:10.1037/0033-295X.89.4.369
Ashby, F. G., & Maddox, W. T. (2005). Human category learning. Annual
Review of Psychology, 56, 149–178. doi:10.1146/annurev
.psych.56.091103.070217
Ashcraft, M. H. (2002). Math anxiety: Personal, educational, and cognitive
consequences. Current Directions in Psychological Science, 11, 181–
185. doi:10.1111/1467-8721.00196
Ashcraft, M. H., & Kirk, E. P. (2001). The relationships among working
memory, math anxiety, and performance. Journal of Experimental Psy-
chology: General, 130, 224–237. doi:10.1037/0096-3445.130.2.224
Baddeley, A., & Logie, R. H. (1999). Working memory: The multiple-
component model. In A. Miyake & P. Shah (Eds.), Models of working
memory: Mechanisms of active maintenance and executive control (pp.
28–61). New York, NY: University Press.
Baumeister, R. F. (1984). Choking under pressure: Self-consciousness and
paradoxical effects of incentives on skillful performance. Journal of
Personality and Social Psychology, 46, 610–620. doi:10.1037/0022-
3514.46.3.610
Baumeister, R. F., & Showers, C. J. (1986). A review of paradoxical
performance effects: Choking under pressure in sports and mental tests.
European Journal of Social Psychology, 16, 361–383. doi:10.1002/
ejsp.2420160405
Beilock, S. L. (2008). Math performance in stressful situations. Current
Directions in Psychological Science, 17, 339–343. doi:10.1111/j.1467-
8721.2008.00602.x
Beilock, S. L., Bertenthal, B. I., McCoy, A. M., & Carr, T. H. (2004). Haste
does not always make waste: Expertise, direction of attention, and speed
versus accuracy in performing sensorimotor skills. Psychonomic Bulletin
& Review, 11, 373–379. doi:10.3758/BF03196585
Beilock, S. L., & Carr, T. H. (2001). On the fragility of skilled perfor-
mance: What governs choking under pressure? Journal of Experimental
Psychology: General, 130, 701–725. doi:10.1037/0096-3445.130.4.701
Beilock, S. L., & Carr, T. H. (2005). When high-powered people fail:
Working memory and “choking under pressure” in math. Psychological
Science, 16, 101–105. doi:10.1111/j.0956-7976.2005.00789.x
Beilock, S. L., Carr, T. H., MacMahon, C., & Starkes, J. L. (2002). When
paying attention becomes counterproductive: Impact of divided versus
skill-focused attention on novice and experienced performance of sen-
sorimotor skills. Journal of Experimental Psychology: Applied, 8, 6–16.
doi:10.1037/1076-898X.8.1.6
Beilock, S. L., & DeCaro, M. S. (2007). From poor performance to success
under stress: Working memory, strategy selection, and mathematical
problem solving under pressure. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 33, 983–998. doi:10.1037/0278-
7393.33.6.983
Beilock, S. L., & Gray, R. (2007). Why do athletes “choke” under pres-
sure? In G. Tenenbaum & R. C. Ecklund (Eds.), Handbook of sport
psychology (3rd ed., pp. 425–444). Hoboken, NH: Wiley.
Beilock, S. L., Jellison, W. A., Rydell, R. J., McConnell, A. R., & Carr,
T. H. (2006). On the causal mechanisms of stereotype threat: Can skills
that don’t rely heavily on working memory still be threatened? Person-
ality and Social Psychology Bulletin, 32, 1059–1071. doi:10.1177/
0146167206288489
Beilock, S. L., Kulp, C. A., Holt, L. E., & Carr, T. H. (2004). More on the
fragility of performance: Choking under pressure in mathematical prob-
lem solving. Journal of Experimental Psychology: General, 133, 584–
600. doi:10.1037/0096-3445.133.4.584
Beilock, S. L., Rydell, R. J., & McConnell, A. R. (2007). Stereotype threat
and working memory: Mechanisms, alleviation, and spill over. Journal
of Experimental Psychology: General, 136, 256–276. doi:10.1037/0096-
3445.136.2.256
Brown, R. M., & Robertson, E. M. (2007). Off-line processing: Reciprocal
interactions between declarative and procedural memories. The Journal
of Neuroscience, 27, 10468–10475. doi:10.1523/JNEUROSCI.2799-
07.2007
Calvo, M. G. (1985). Effort, aversive representations and performance in
test anxiety. Personality and Individual Differences, 6, 563–571. doi:
10.1016/0191-8869(85)90005-4
Carver, C. S., & Scheier, M. F. (1978). Self-focusing effects of disposi-
tional self-consciousness, mirror, presence, and audience presence. Jour-
nal of Personality and Social Psychology, 36, 324–332. doi:10.1037/
0022-3514.36.3.324
Chalabaev, A., Sarrazin, P., Stone, J., & Cury, F. (2008). Do achievement
goals mediate stereotype threat? An investigation on females’ soccer
performance. Journal of Sport & Exercise Psychology, 30, 143–158.
Dandy, J., Brewer, N., & Tottman, R. (2001). Self-consciousness and
performance decrements within a sporting context. The Journal of Social
Psychology, 141, 150–152. doi:10.1080/00224540109600540
Davis, D., & Brock, T. C. (1975). Use of first person pronouns as a
function of increased objective self-awareness and performance feed-
back. Journal of Experimental Social Psychology, 11, 381–388. doi:
10.1016/0022-1031(75)90017-7
Davis, M. H., & Harvey, J. C. (1992). Declines in major league batting
performance as a function of game pressure: A drive theory analysis.
Journal of Applied Social Psychology, 22, 714–735. doi:10.1111/j.1559-
1816.1992.tb00999.x
DeCaro, M. S., Carlson, K. D., Thomas, R. D., & Beilock, S. L. (2009).
When and how less is more: Reply to Tharp and Pickering. Cognition,
111, 397–403. doi:10.1016/j.cognition.2009.03.001
DeCaro, M. S., Rotar, K. E., Kendra, M. S., & Beilock, S. L. (2010).
Diagnosing and alleviating the impact of performance pressure on math-
403
MECHANISMS OF SKILL FAILURE

ematical problem solving. Quarterly Journal of Experimental Psychol-
ogy, 63, 1619–1630. doi:10.1080/17470210903474286
DeCaro, M. S., Thomas, R. D., & Beilock, S. L. (2008). Individual
differences in category learning: Sometimes less working memory ca-
pacity is better than more. Cognition, 107, 284–294. doi:10.1016/
j.cognition.2007.07.001
Dohmen, T. J. (2008). Do professionals choke under pressure? Journal of
Economic Behavior & Organization, 65, 636–653. doi:10.1016/
j.jebo.2005.12.004
Duval, S., & Wicklund, R. A. (1972). A theory of objective self-awareness.
New York, NY: Academic Press.
Engle, R. W. (2002). Working memory capacity as executive attention.
Current Directions in Psychological Science, 11, 19–23. doi:10.1111/
1467-8721.00160
Eysenck, M. W., & Calvo, M. G. (1992). Anxiety and performance: The
processing efficiency theory. Cognition & Emotion, 6, 409–434. doi:
10.1080/02699939208409696
Forgas, J. P., Brennan, G., Howe, S., Kane, J. F., & Sweet, S. (1980).
Audience effects on squash players’ performance. Journal of Social
Psychology, 111, 41–47. doi:10.1080/00224545.1980.9924271
Galea, J. M., Albert, N. B., Ditye, T., & Miall, R. C. (2010). Theta-burst
TMS to the dorsolateral prefrontal cortex facilitates the consolidation of
procedural skills. Journal of Cognitive Neuroscience, 22, 1158–1164.
doi:10.1162/jocn.2009.21259
Geller, V., & Shaver, P. (1976). Cognitive consequences of self-awareness.
Journal of Experimental Social Psychology, 12, 99–108. doi:10.1016/
0022-1031(76)90089-5
Gimmig, D., Huguet, P., Caverni, J. P., & Cury, F. (2006). Choking under
pressure and working memory capacity: When performance pressure
reduces fluid intelligence. Psychonomic Bulletin & Review, 13, 1005–
1010.
Gray, J. R. (2004). Integration of emotion and cognitive control. Current
Directions in Psychological Science, 13, 46–48.
Gray, J. R., Braver, T. S. & Raichle, M. E. (2002). Integration of emotion
and cognition in the lateral prefrontal cortex. Proceedings of the Na-
tional Academy of Sciences, USA, 99, 4115–4120.
Gray, R. (2004). Attending to the execution of a complex sensorimotor
skill: Expertise differences, choking, and slumps. Journal of Experimen-
tal Psychology: Applied, 10, 42–54. doi:10.1037/1076-898X.10.1.42
Grimm, L. R., Markman, A. B., Maddox, W. T., & Baldwin, G. C. (2009).
Stereotype threat reinterpreted as a regulatory mismatch. Journal of
Personality and Social Psychology, 96, 288–304. doi:10.1037/a0013463
Hardy, L., Mullen, R., & Jones, G. (1996). Knowledge and conscious
control of motor actions under stress. British Journal of Psychology, 87,
621–636.
Hayes, S., Hirsch, C., & Mathews, A. (2008). Restriction of working
memory capacity during worry. Journal of Abnormal Psychology, 117,
712–717. doi:10.1037/a0012908
Heaton, A. W., & Sigall, H. (1989). The “championship choke” revisited:
The role of fear and acquiring a negative identity. Journal of
Applied Social Psychology, 19, 1019–1033. doi:10.1111/j.1559-
1816.1989.tb01236.x
Ikeda, M., Iwanaga, M., & Seiwa, H. (1996). Test anxiety and working
memory system. Perceptual and Motor Skills, 82, 1223–1231.
Jackson, R. C., Ashford, K. J., & Norsworthy, G. (2006). Attentional focus,
dispositional reinvestment, and skilled motor performance under pres-
sure. Journal of Sport & Exercise Psychology, 28, 49–68.
Keller, J., & Dauenheimer, D. (2003). Stereotype threat in the classroom:
Dejection mediates the disrupting effect on women’s math performance.
Personality and Social Psychology Bulletin, 29, 371–381. doi:10.1177/
0146167202250218
Kimble, G. A., & Perlmuter, L. C. (1970). The problem of volition.
Psychological Review, 77, 361–384. doi:10.1037/h0029782
Langer, E. J., & Imber, L. (1979). When practice makes imperfect: Debil-
itating effects of overlearning. Journal of Personality and Social Psy-
chology, 37, 2014–2024. doi:10.1037/0022-3514.37.11.2014
Lewis, B., & Linder, D. (1997). Thinking about choking? Attentional
processes and paradoxical performance. Personality and Social Psychol-
ogy Bulletin, 23, 937–944. doi:10.1177/0146167297239003
Logan, G. D. (1988). Toward an instance theory of automatization. Psy-
chological Review, 95, 492–527. doi:10.1037/0033-295X.95.4.492
Maddox, W. T., & Ashby, F. G. (2004). Dissociating explicit and
procedural-learning based systems of perceptual category learning. Be-
havioural Processes, 66, 309–332. doi:10.1016/j.beproc.2004.03.011
Maddox, W. T., Love, B. C., Glass, B. D., & Filoteo, J. V. (2008). When
more is less: Feedback effects in perceptual category learning. Cogni-
tion, 108, 578–589. doi:10.1016/j.cognition.2008.03.010
Markman, A. B., Maddox, W. T., & Worthy, D. A. (2006). Choking and
excelling under pressure. Psychological Science, 17, 944–948. doi:
10.1111/j.1467-9280.2006.01809.x
Masters, R. S. W. (1992). Knowledge, knerves and know-how: The role of
explicit versus implicit knowledge in the breakdown of a complex motor
skill under pressure. British Journal of Psychology, 83, 343–358.
Miyake, A., & Shah, P. (1999). Models of working memory: Mechanisms
of active maintenance and executive control. New York, NY: University
Press.
Nissen, M. J., & Bullemeyer, P. (1987). Attentional requirements of
learning: Evidence from performance measures. Cognitive Psychology,
19, 1–32. doi:10.1016/0010-0285(87)90002-8
Paulus, P. B., Shannon, V. C., Wilson, D. L., & Boone, T. D. (1972). The
effects of operator presence on gymnastic performance in a field situa-
tion. Psychonomic Science, 29, 88–90.
Rapee, R. M. (1993). The utlisation of working memory by worry. Behav-
iour Research and Therapy, 31, 617–620. doi:10.1016/0005-
7967(93)90114-A
Robertson, E. M. (2007). The serial reaction time task: Implicit motor skill
learning? Journal of Neuroscience, 27, 10073–10075. doi:10.1523/
JNEUROSCI.2747-07.2007
Sarason, I. G. (1972). Test anxiety and the model who fails. Journal of
Personality and Social Psychology, 22, 410–413. doi:10.1037/
h0032893
Schmader, T., & Johns, M. (2003). Converging evidence that stereotype
threat reduces working memory capacity. Journal of Personality and
Social Psychology, 85, 440–452. doi:10.1037/0022-3514.85.3.440
Schmader, T., Johns, M., & Forbes, C. (2008). An integrated process model
of stereotype threat effects on performance. Psychological Review, 115,
336–356. doi:10.1037/0033-295X.115.2.336
Seibt, B., & Förster, J. (2004). Stereotype threat and performance: How
self-stereotypes influence processing by inducing regulatory foci. Jour-
nal of Personality and Social Psychology, 87, 38–56. doi:10.1037/0022-
3514.87.1.38
Spielberger, C. C., Gorsuch, R. L., & Lushene, R. (1970). State-Trait
Anxiety Inventory. Palo Alto, CA: Consulting Psychology Press.
Steele, C. (1997). A threat in the air: How stereotypes shape intellectual
identity and performance. American Psychologist, 52, 613–629.
Steele, C. M., Spencer, S. J., & Aronson, J. (2002). Contending with group
image: The psychology of stereotype and social identity threat. Advances
in Experimental Social Psychology, 34, 379–440. doi:10.1016/S0065-
2601(02)80009-0
Tabachnick, B. G., & Fidel, L. S. (1996). Using multivariate statistics.
New York, NY: HarperCollins.
Tharp, I. J., & Pickering, A. D. (2009). A note on DeCaro, Thomas, and
Beilock (2008): Further data demonstrate complexities in the assessment
of information-integration category learning. Cognition, 111, 411–415.
doi:10.1016/j.cognition.2008.10.003
Waldron, E. M., & Ashby, F. G. (2001). The effects of concurrent task
interference on category learning: Evidence for multiple category learn-
404 DECARO, THOMAS, ALBERT, AND BEILOCK

ing systems. Psychonomic Bulletin & Review, 8, 168–176. doi:10.3758/
BF03196154
Wine, J. (1971). Test anxiety and direction of attention. Psychological
Bulletin, 76, 92–104. doi:10.1037/h0031332
Worthy, D. A., Markman, A. B., & Maddox, W. T. (2009). What is
pressure? Evidence for social pressure as a type of regulatory focus.
Psychonomic Bulletin & Review, 16, 344–349. doi:10.3758/
PBR.16.2.344
Wulf, G. (2007). Attentional focus and motor learning: A review of 10
years of research. E-Journal Bewegung und Training, 1, 4–14.
Zeithamova, D., & Maddox, W. T. (2006). Dual-task interference in
perceptual category learning. Memory & Cognition, 34, 387–398.
Appendix A
Table A1
Mean Trials to Criterion (Log-Transformed) in Experiments 1 and 2
Experiment Condition
Category
structure
Block 1: Baseline
M(SD)
Block 2: Manipulation
M(SD)
1 Distracting secondary task RB 1.23 (.32) 1.40 (.38)
II 1.70 (.38) 1.76 (.35)
Explicit monitoring secondary task RB 1.28 (.34) 1.25 (.31)
II 1.54 (.35) 1.71 (.36)
2 Control RB 1.21 (.24) 1.24 (.24)
II 1.65 (.35) 1.68 (.36)
Outcome pressure RB 1.23 (.31) 1.39 (.33)
II 1.66 (.41) 1.55 (.37)
Monitoring pressure RB 1.30 (.37) 1.32 (.27)
II 1.59 (.36) 1.77 (.34)
Note. RB ⫽rule-based; II ⫽information-integration.
Table A2
Mean Trials to Criterion (Log-Transformed) in Experiment 3
Condition
Category
structure
Baseline
M(SD)
Distracting secondary task
M(SD)
Explicit monitoring secondary task
M(SD)
Outcome pressure RB 1.22 (.19) 1.45 (.34) 1.32 (.32)
Monitoring pressure II 1.58 (.38) 1.57 (.36) 1.82 (.34)
Note. RB ⫽rule-based; II ⫽information-integration.
(Appendices continue)
405
MECHANISMS OF SKILL FAILURE

Appendix B
Response Strategies and Information-Integration Category Learning
We examined the extent to which participants’ responses
matched four different strategy types during information-
integration category learning. Participants could use the “optimal”
strategy, categorizing the stimuli as established by the experiment-
ers (at 100% accuracy), most likely by using a procedural learning
strategy. Or participants could use explicit, rule-based strategies to
try to learn the information-integration categories. Such explicit
strategies could involve either one dimension (e.g., items with a
green symbol belong to Category A), two dimensions (e.g., items
with a blue background or a square symbol belong to Category A),
or three dimensions (e.g., items with a blue background color, or
items with green squares, belong to Category A). One- and two-
dimension strategies can result in 75% accuracy, and three-
dimension strategies can lead to 87.5% accuracy. We found three
possible strategies for each dimension, for a total of 10 possible
strategies (including the optimal strategy; see also DeCaro, Carl-
son, Thomas, & Beilock, 2009). One could also postulate a number
of very complex rule-plus-exception strategies that lead to 100%
accuracy, but these cannot be dissociated from the optimal strat-
egy. As we discuss in Experiments 1–3, the evidence suggests that
the optimal strategy indeed reflects a less attention-demanding
strategy.
We attempted to ascertain the strategies individuals used to
learn the information-integration categories by examining the re-
sponses given on each learning trial and comparing these with the
predicted responses from each possible strategy. To model these
strategies, we divided the 200 possible trials into 20 blocks of 10
trials each and determined the strategy that matched the most
responses given within each block. If two (or more) strategies
accounted for the maximum number of responses within a given
trial, then each was given an equal weight for that block (e.g., .5
for two strategies). If participants exited the task after reaching the
learning criterion, the remaining blocks were scored as matching
the optimal strategy. The number of response agreements across all
blocks was summed for each strategy and converted to proportions
for analysis.
To examine whether individuals relied more on explicit strate-
gies in secondary task and/or high-pressure conditions than at
baseline, we created a difference score. The proportion of each
strategy type during the single-task and/or low-pressure baseline
(depending on the experiment) was subtracted from the proportion
of each strategy type during the secondary task and/or pressure
condition. Results are presented in Figures 2, 4, and 6 and dis-
cussed in the text.
Received June 12, 2009
Revision received February 4, 2011
Accepted February 4, 2011 䡲
406 DECARO, THOMAS, ALBERT, AND BEILOCK