CURRENT DIRECTIONS IN PSYCHOLOGICAL SCIENCE

Progress in Executive-Function
Research
From Tasks to Functions to Regions to Networks
Adam R. Aron

University of California, San Diego

ABSTRACT—It has long been observed that damage to the including the famous case of Phineas Gage (the railroad worker
frontal cortex affects a person’s ability to control thought, whose frontal lobe was penetrated by a projectile), it was strik-
behavior, and emotion while sometimes leaving funda- ingly obvious that some ‘‘higher’’ cognitive power had been
mental processes such as vision, hearing, and long-term lost even when vision, audition, feeling, movement, speech, and
memory intact. Such observations have led theoreticians to long-term memory were intact.
suppose that a set of executive control functions exists, at The experimental study of executive functions began in ear-
the top of the hierarchy of mental processes. To study these nest when several now-classic tasks were administered to pa-
executive functions and their relation to the frontal cortex tients with frontal-lobe damage. Such tasks included the
and its subregions, researchers have long employed several Wisconsin Card Sort Test (WCST), the Stroop test, and tests of
now-classic cognitive tests in patients with brain damage. working memory and planning. The WCST requires subjects to
Yet until recently it has proved difficult to reliably localize override a tendency to sort cards according to a previously
the putative executive functions to discrete regions. This relevant category and instead to start sorting cards according to
article illustrates how recent progress in executive-func- a new, to-be-discovered, correct category. In the Stroop task,
tions research has been driven by the coupling of sop- subjects have to override the prepotent tendency to read a word
histicated neuroscience techniques with advances in in order to name the color in which the word is written (something
experimental psychology. Taking examples from recent that is more difficult because less habitual). Investigators
studies, it shows how experimental tasks may be decom- showed repeatedly that damage to various regions of the frontal
posed into cognitive components that can be localized to lobe produced behavioral impairment on such tests, particularly
discrete—but structurally connected—brain regions. for conditions requiring overcoming prepotent response ten-
What emerges is a new ontology for executive function in dencies, but the results were often inconsistent. Some of the
terms of which cognitive components exist and of how, and inconsistency was probably related to gross differences
when, they are recruited during task performance. among tasks. If such gross differences exist then different tasks
KEYWORDS—executive functions; frontal cortex may tap underlying psychological components in different
ways, and the relation to particular brain regions is difficult to
establish. It was thus supposed that the outlook for such research
Executive functions are higher-order cognitive functions that could be improved if executive functions could be decomposed
stand at the apogee of both evolutionary and mental develop- into constructs such as ‘‘initiating,’’ ‘‘sustaining,’’ ‘‘rule main-
ment. They enable us to formulate goals and plans; remember tenance,’’ ‘‘switching,’’ ‘‘inhibiting,’’ and ‘‘monitoring.’’ More
these goals across time; choose and initiate actions to help us refined tests were developed in both human and nonhuman
achieve these goals; and monitor and adjust our behavior, as animal experiments to try to map such components onto par-
necessary, until we complete or fail at them. The idea that such ticular sectors of the frontal cortex (reviewed by Robbins,
executive functions exist came from observing people with brain 1996).
damage, typically damage to their frontal lobes. In some cases, Nevertheless, 10 years ago an assessment of the state of
executive-function research found much to complain about
(Rabbitt, 1997). Although particular tasks had been broken
Address correspondence to Adam R. Aron, Department of Psychol-
ogy, University of California, San Diego, 9500 Gilman Drive, La Jolla, down into supposed components, it was unclear if these hypo-
CA 92093; e-mail: [email protected]. thetical processes had real construct value at the psychological

124 Copyright r 2008 Association for Psychological Science Volume 17—Number 2

 Adam R. Aron

or neural levels. It seemed that brain damage might affect tasks The ability to derive a good behavioral index of stopping has
designed to tap executive functions not because of effects on had strong implications for neuroscience research. It has facil-
functionally localized modules but because executive-function itated many studies (too numerous to detail here) in different
tasks are just somehow ‘‘more demanding’’ than non-executive- species and using different methods (see Aron et al., 2007 for
function tasks. Further, the pattern of data relating different review). Importantly, SSRT offers features uncommon among
executive functions to brain regions was still quite inconsistent. its sister variables in executive-function research: good within-
In this article, I aim to show that the intervening decade has subject reliability, good reliability across groups matched on
seen rapid progress in executive-function research that is driven age, good translational application in rodent and nonhuman
by two major, and strongly interacting, trends. The first trend is primate models, and good sensitivity for detecting effects of
the adoption of experimental-psychology paradigms that provide pharmacological manipulation and brain lesion.
measures of the putative cognitive functions that are more pre- My colleagues and I tried to find out which, if any, region of the
cise than ‘‘classical’’ neuropsychological tests. The new tests’ prefrontal cortex is critical for stopping by examining SSRT in
precision makes functional operationalization more reliable patients with brain damage. We found that the inferior frontal
within and across subjects and more sensitive to the effects of cortex (IFC) in the right hemisphere is critical for stopping (re-
brain damage and physiology. The second trend is the adoption viewed in Aron, 2007). In the initial study, damage to other
of cutting-edge neuroscience techniques in imaging, neuro- sectors of either the right or left frontal cortex seemed unim-
physiology, and brain stimulation. Together, these methods have portant, but recent work also points to an area at the top and
revived the possibility that individual executive functions are middle of the brain, the presupplementary motor area (preSMA;
rooted in particular neural systems. Yet the picture that emerges Floden & Stuss, 2006; Nachev, Wydell, O’Neill, Husain, &
is a complex one: of separate brain regions connected by direct Kennard, 2007); the relevance of this latter finding will be made
pathways for fast transmission of common information. To il- clear later in this article.
lustrate this approach, and to spell out the implications for Subsequent studies with functional magnetic resonance
experimental psychology, I offer several examples from studies imaging (fMRI) have shown, moreover, that the degree of acti-
of response control. vation in the right IFC predicts the speed of stopping (Aron,
Behrens, Smith, Frank, & Poldrack, 2007; Aron & Poldrack,
2006). This provides a neural correlate for a control process that
STOPPING AN INCIPIENT RESPONSE can be examined with fMRI when other kinds of stopping are
performed (eye, speech; see Xue, Aron, & Poldrack, 2008) and
Imagine you are about to press your foot down on your car’s when other kinds of control tasks are performed.
accelerator but, as you begin, a bicycle appears in front of your
car. Clearly you need to countermand the motor command. Like A CIRCUIT FOR BEHAVIORAL AND NEURAL
shifting in the WCST, and like the ability to overcome Stroop INHIBITION
interference, such stopping requires control over a prepotent
response. However, stopping an already-initiated (incipient) The foregoing shows that a particular sector of the right pre-
response is different from controlling other kinds of prepotent frontal cortex is important for stopping. Yet what is the signifi-
responses in a substantial way: When studied in the lab, ex- cance of this for our psychological understanding? One way to
perimentally, it allows a specific examination of the way in which answer this is to consider how the right IFC influences the in-
the controlling (stopping) process interacts with the initiated cipient motor command. In the earlier example of driving, if you
impulse (or ‘‘go process’’). A dependent measure is derived (see plan to press the accelerator with your foot and begin to activate
below) for the actual speed of the control process itself. This your muscles, by what mechanism can the putative frontal
speed-of-stopping measure has highly useful characteristics for command in the right IFC intercept the go command in the motor
brain research. system? Does it do this by suppressing the motor command itself
Measuring the speed of the stopping process became possible (and thereby inhibiting the agonist muscle), or does it activate an
with experimental paradigms such as the stop-signal task (Logan alternate representation (causing an antagonist muscle to block
& Cowan, 1984). In this task, subjects are instructed to respond the action)?
as fast as possible to imperative (go) stimuli and to do their best Although it has long been supposed that a cardinal role for the
to inhibit the incipient response when a stop signal subsequently frontal cortex is to inhibit and activate representations in pos-
occurs. If the delay between imperative stimulus and the stop terior cortical or subcortical regions (Miller & Cohen, 2001),
signal is short, subjects are likely to inhibit; if the delay is long, there have been few demonstrations of underlying mechanisms.
subjects are less likely to inhibit. By varying the delays and One way the stop process could intercept and block the go
employing a mathematical model of go and stop processes, it is process is via the basal ganglia. Much research has shown that
possible to estimate how quickly the subject can stop—the stop- initiating a motor response engages the so-called ‘‘direct path-
signal reaction time (SSRT; Logan & Cowan, 1984). way’’ of the basal ganglia, in which the planning regions of the

Volume 17—Number 2 125

 Executive Functions

Fig. 1. Circuitry between the cortex and basal ganglia (a–c) and the brain’s ‘‘stopping’’ or ‘‘braking’’
network discussed in the article (d). Filled heavy lines in the circuitry schematics are inhibitory connec-
tions; open arrows are excitatory connections. Information flows from the cortex through the basal ganglia
and back to the cortex — in the diagram this means from top to bottom and back to top. At rest (a), striatal
cells are quiescent, while pallidal neurons fire continually, thus suppressing thalamocortical output to the
primary motor cortex (where finger and other representations reside). The motor command (large arrow in
b) selectively excites a particular finger representation against the background of inhibition of irrelevant
responses. One way in which the incipient response could be stopped (shown in c) is via the fast, hyperdirect,
prefrontal-to-subthalamic route discussed in the article. A cutaway of the right hemisphere of the brain
(viewed from the front and side; d) reveals white matter tracts or ‘‘cables’’ (colored) that connect three
distant regions of the brain known to be important for controlling behavior. PreSMA 5 presupplementary
motor area; IFC 5 inferior frontal cortex; STN region 5 midbrain region consistent with subthalamic
nucleus. From ‘‘Triangulating a Cognitive Control Network Using Diffusion-Weighted Magnetic Resonance
Imaging (MRI) and Functional MRI,’’ by A.R. Aron, T.E. Behrens, S. Smith, M.J. Frank, & R.A. Pold-
rack, 2007, Journal of Neuroscience, 27 (p. 3746). Copyright 2007, Society for Neuroscience.
Reprinted with permission.

frontal cortex send activity to the putamen, which then projects pallidus, thus increasing inhibition of the thalamus and blocking
to the globus pallidus, then to the thalamus, then to primary basal ganglia output to the primary motor cortex (Fig. 1c).
motor cortex, and onward to the muscles (Fig. 1a,b). We found How might the prefrontal stopping command get to the STN in
activation in regions consistent with this direct pathway when the basal ganglia? We used a form of structural magnetic re-
subjects responded on go trials (Aron & Poldrack, 2006). On stop sonance imaging called diffusion weighted imaging to establish
trials, we found activation of a basal ganglia region in the vicinity that these regions are directly connected via a white-matter tract
of the subthalamic nucleus (STN), in addition to activation of the (Aron et al., 2007; Fig. 1d). Thus, the same frontal region that is
right IFC and preSMA (Aron & Poldrack, 2006). This is striking critical for stopping apparently sends axons into a region of the
because the STN has been conceived as a ‘‘stop button’’—its basal ganglia, the STN, that increases excitation of the globus
activity levels are altered in Parkinson’s disease, thus contrib- pallidus and so could block the go command. Could this
uting to rigidity (and tremor). The STN projects to the globus mechanism of control simply work by activating the antagonist

126 Volume 17—Number 2

 Adam R. Aron

representation? Probably not. A study that measured the activity works, a controller (i.e., a homunuculus/homuncula) needs to be
level of the cortical representation of muscles during a stop- posited. But if there is such a controller, then what controls it?
signal task showed that both the agonist muscle as well as an This leads to an infinite regress. Instead, it has been proposed
irrelevant muscle representation in the cortex were suppressed that a monitoring function could constantly check for response
by the requirement to inhibit the incipient response (Coxon, conflict—as when multiple possible responses are generated
Stinear, & Byblow, 2006). This suggests that stopping is achieved by the current stimulus—and the conflict mechanism could then
via active inhibition of (potential) response tendencies rather trigger the controlling processes accordingly (e.g. Botvinick,
than via activation of the antagonist muscle. Cohen, & Carter, 2004; see also Norman & Shallice, 1986).
These data provide a mechanistic account of how the frontal Thus, one possibility is that the preSMA may monitor for
cortex can interact with the motor system, and how an executive conflict between an intended response and a countervailing
function such as response inhibition is implemented via con- signal (for an alternative response, or for no response). Then,
necting circuitry. They suggest that the behavioral requirement when such conflict is detected, the ‘‘brakes’’ could be put on via the
to inhibit a response has its counterpart in a neurocognitive connection between the right IFC and the STN region. This may
inhibitory mechanism that actively suppresses the go command. explain how responses are completely stopped, and also how
This bears on the important question of whether ‘‘inhibition’’ is they are slowed, pending a decision. When we presented irrel-
a valid concept in psychological research (see Aron, 2007; evant stop signals to subjects, we found that their responses on go
MacLeod, Dodd, Sheard, Wilson, & Bibi, 2003). trials were slowed, without complete cancellation. Moreover, the
An interesting possibility is that this ‘‘control circuit,’’ or one degree of slowing corresponded to the degree of activation in the
that works along similar lines, could be recruited during per- right preSMA, the IFC, and the STN region (Aron et al., 2007).
formance of a wide variety of tasks. Thus, observed activation Thus, braking (slowing) a response and stopping (canceling) a
of such brain regions as the right IFC and STN region can then response may occur via the same system. This is important
provide evidence for whether an inhibitory mechanism is oper- because many requirements for control, e.g., in Stroop and other
ational during planning and switching and other control situa- interference tasks, are not so much about stopping outright as
tions, even when rapid stopping is not required. (It is important about, ostensibly, braking an inappropriate response tendency
to bear in mind, however, that inferring whether a cognitive until a discrepancy can be resolved (Frank, 2006). Future ex-
process is active on the basis of brain activation is limited by the periments could address whether the above-mentioned control
fact that particular brain regions are not activated by one type of circuitry is recruited by, for example, Stroop interference, which
cognitive process only; Poldrack, 2006). would thereby help to clarify its cognitive constituents.
Overall, these results demonstrate how different putative
functions can be localized to different brain regions within an
THE WIDER FUNCTIONAL NEUROANATOMIC overall structurally connected network. The results also provide
NETWORK testable predictions about the functional neuroanatomy under-
lying other kinds of tasks, such as Stroop interference, planning,
As we saw, one of the problems that has confounded executive- and switching.
function research is the poor correspondence between functional
terms such as ‘‘inhibition,’’ ‘‘maintaining,’’ or ‘‘monitoring’’ and
underlying brain regions. By identifying critical nodes such as SWITCHING BETWEEN TASKS OR RESPONSES
the right IFC and the STN region with well-defined hypothetical
functions (such as inhibition), it is possible to ask what other Consistent with the possibility that the revealed circuitry be-
regions they are connected to, and thus build up a network of tween the right preSMA, IFC, and STN regions has wider ap-
putative interacting functional regions. We found that both the plicability, it has been shown that the ability to switch between
right IFC and the STN are connected to the preSMA (part of different tasks or responses also engages key nodes in this net-
the superior frontal lobe, adjacent and dorsal to the anterior work. ‘‘Switching’’ refers to the ability to intentionally change
cingulate) in the right hemisphere (Aron et al., 2007; Fig. 1d). from a task such as answering a phone, to visually searching for
This is striking because preSMA damage also affects stopping something, to doing something else. It has been carefully studied
(Floden & Stuss, 2006; Nachev et al., 2007). with sophisticated cognitive-psychology paradigms in which
While the proper functional role of the preSMA continues to subjects perform a series of trials of task A and then switch to
draw considerable research interest, its role could, as a first performing a series of trials of task B (reviewed by Monsell,
approximation, be described as ‘‘monitoring and resolving re- 2003). For each subject, engagement of executive functions is
sponse conflict’’ (e.g., Nachev et al., 2007). This putative func- measured in terms of the switch cost, which is computed by
tion fits nicely with its being a third node in the response-control subtracting the average reaction time of nonswitch trials from the
system, because it helps to solve the riddle of how control is average reaction time of switch trials. While it is still unclear
generated. It has often seemed that, to explain how control exactly which cognitive processes underpin the switch cost, it

Volume 17—Number 2 127

 Executive Functions

seems that one constituent process could relate to the control of functions interact is still incomplete, the assignment of such
irrelevant response tendencies, as patients with damage to the putative functions as inhibition and conflict monitoring/resolu-
right IFC (who had impaired SSRT in the stop-signal paradigm) tion to discrete but connected nodes within the frontal cortex,
also had the longest switch costs (reviewed in Aron, 2007; also and the specification of functional connections of these nodes to
see Hodgson et al., 2007). the subcortex, represents an advance in psychological under-
But where does the signal to switch come from? While ques- standing. As we have seen, these results bear on important
tions about the origin of control are mysterious, it is intriguing controversies such as whether inhibition is a valid construct in
to note that disruption of the preSMA by transcranial magnetic psychology, whether responses are controlled by suppressing an
stimulation affected switch trials but not nonswitch trials agonist or activating an antagonist muscle, what the mechanism
(Rushworth, Hadland, Paus, & Sipila, 2002). Further, a recent underlying Stroop interference is, what the cognitive constitu-
study recorded directly from neurons in the preSMA of monkeys ents of the switch cost are, and how exactly the control homun-
while they performed an oculomotor switch from a well-prac- culus can be banished. Future progress will come from further
ticed (automatic) task to a less practiced one (Isoda & Hikosaka, interaction between cognitive psychology and neuroscience.
2007b). Switch trials were associated with increased activity of The development of a new generation of cognitive tasks with
preSMA cells. Moreover, their activity fell within a sufficiently well-operationalized functional components could do much to
short duration to influence behavior, and also at a time scale that complement neuroscience investigations with increasing spatial
preceded increased switch-related activity within STN neurons and temporal resolution.
in the same paradigm (Isoda & Hikosaka, 2007a). The authors
also showed that injecting electric current into preSMA neurons
led to a greater proportion of successful switch trials. In a study Recommended Reading
of a single patient with selective preSMA damage, an impair- Goldberg, E. (2001). The executive brain: Frontal lobes and the civilized
mind. Oxford, UK: Oxford University Press. An introduction to the
ment was found in stopping one response and performing another
topic for lay readers, including colorful accounts of individual
one (Nachev et al., 2007; also see Floden & Stuss, 2006). To- patients with frontal lobe damage and implications for law and
gether, these studies strongly suggest that the preSMA is im- society.
portant for switching. Combined with the finding that the right Knight, R.K., & Stuss, D.T. (Eds.). (2002). Principles of frontal lobe
IFC is also important for switching and with the revealed three- function. Oxford, UK: Oxford University Press. A collection of
way white-matter network between the preSMA, the IFC, and short papers by experts in the field.
the STN, it could be postulated that the functional mecha- Miller, E.K., & Cohen, J.D. (2001). (See References). An important
monograph expounding a theoretical account of executive function
nism that brakes and stops responses could also underlie in terms of the way in which the prefrontal cortex influences other
switching between tasks. Therefore, this three-way functional- brain regions during task performance.
anatomic system for response control does apparently have
generality for different effectors (eye, speech, and hand) as well
as for different tasks (stopping and switching at least, and maybe
others). Acknowledgments—Thanks to David Badre and Vic Ferreira
for helpful comments and to Bill Lanouette for editing assis-
CONCLUSIONS tance.

Innovations in cognitive psychology, embodied in stop-signal REFERENCES

and task-switching paradigms, have been keenly adopted by
neuroscientists. The combined approach has enabled a finer Aron, A.R. (2007). The neural basis of inhibition in cognitive control.
dissection of executive functions than was possible with clas- Neuroscientist, 13, 214–228.
sical paradigms such as the WCST. The emerging picture is that Aron, A.R., Behrens, T.E., Smith, S., Frank, M.J., & Poldrack, R.A.
(2007). Triangulating a cognitive control network using diffusion-
performance of control tasks depends on different anatomically
weighted magnetic resonance imaging (MRI) and functional MRI.
connected brain regions, making up a network. As a first ap- Journal of Neuroscience, 27, 3743–3752.
proximation, different nodes in the network can be assigned Aron, A.R., Durston, S., Eagle, D.M., Logan, G.D., Stinear, C.M., &
different executive functions. These include ‘‘inhibition’’ for the Stuphorn, V. (2007). Converging evidence for a fronto-basal-gan-
right IFC and the right STN and ‘‘conflict monitoring and reso- glia network for inhibitory control of action and cognition. Journal
lution of competition’’ for the preSMA. It is likely that other of Neuroscience, 27, 11860–11864.
regions such as the dorsolateral PFC are important for main- Aron, A.R., & Poldrack, R.A. (2006). Cortical and subcortical contri-
butions to Stop signal response inhibition: Role of the subthalamic
taining goals and rules (reviewed by Bunge, 2004) and that other nucleus. Journal of Neuroscience, 26, 2424–2433.
regions are important for maintaining attention, computing Botvinick, M.M., Cohen, J.D., & Carter, C.S. (2004). Conflict monitoring
the value of outcomes, and adjusting behavior accordingly. and anterior cingulate cortex: An update. Trends in Cognitive
Although the picture of how, when, and for which tasks such Sciences, 8, 539–546.

128 Volume 17—Number 2

 Adam R. Aron

Bunge, S.A. (2004). How we use rules to select actions: A review of MacLeod, C., Dodd, M., Sheard, E., Wilson, D., & Bibi, U. (2003). In
evidence from cognitive neuroscience. Cognitive, Affective, & Opposition to Inhibition. In B.H. Ross (Ed.), The psychology of
Behavioral Neuroscience, 4, 564–579. learning and motivation (Vol. 43, pp. 163–214). San Diego:
Coxon, J.P., Stinear, C.M., & Byblow, W.D. (2006). Intracortical inhi- Elsevier Science.
bition during volitional inhibition of prepared action. Journal of Miller, E.K., & Cohen, J.D. (2001). An integrative theory of prefrontal
Neurophysiology, 95, 3371–3383. cortex function. Annual Review of Neuroscience, 24, 167–202.
Floden, D., & Stuss, D.T. (2006). Inhibitory control is slowed in patients Monsell, S. (2003). Task switching. Trends in Cognitive Sciences, 7, 134–
with right superior medial frontal damage. Journal of Cognitive 140.
Neuroscience, 18, 1843–1849. Nachev, P., Wydell, H., O’Neill, K., Husain, M., & Kennard, C. (2007).
Frank, M. (2006). Hold your horses: A dynamic computational role for The role of the pre-supplementary motor area in the control of
the subthalamic nucleus in decision making. Neural Networks, 19, action. Neuroimage, 36(Suppl. 2), T155–T163.
1120–1136. Norman, D., & Shallice, T. (1986). Attention of action: Willed and
Hodgson, T., Chamberlain, M., Parris, B., James, M., Gutowski, N., automatic control of behavior. In R.J. Davidson, G.E. Schwartz, &
Husain, M., et al. (2007). The role of the ventrolateral D. Shapiro (Eds.), Consciousness and self-regulation: Advances in
frontal cortex in inhibitory oculomotor control. Brain, 130, 1525– research and theory (Vol. 4, pp. 1–18). New York: Plenum.
1537. Poldrack, R.A. (2006). Can cognitive processes be inferred from ne-
Isoda, M., & Hikosaka, O. (2007a, November). Role for the subthalamic uroimaging data? Trends in Cognitive Sciences, 10, 59–63.
nucleus in switching from automatic to controlled behavior. Paper Rabbitt, P.M. (Ed.). (1997). Methodology of frontal and executive
presented at the annual meeting of the Society for Neuroscience, function. Hove, UK: Psychology Press.
San Diego, CA. Robbins, T. (1996). Dissociating executive functions of the prefrontal
Isoda, M., & Hikosaka, O. (2007b). Switching from automatic to con- cortex. Philosophical Transactions of the Royal Society of London–
trolled action by monkey medial frontal cortex. Nature Neu- B: Biological Sciences, 351, 1463–1471.
roscience, 10, 240–248. Rushworth, M.F.S., Hadland, K.A., Paus, T., & Sipila, P.K. (2002).
Logan, G.D., & Cowan, W.B. (1984). On the ability to inhibit thought Role of the human medial frontal cortex in task switching: A
and action: A theory of an act of control. Psychological Review, 91, combined fMRI and TMS study. Journal of Neurophysiology, 87,
295–327. 2577–2592.

Volume 17—Number 2 129