Separate, Causal Roles of the Caudate
in Saccadic Choice and Execution
in a Perceptual Decision Task
Long Ding1,* and Joshua I. Gold1
1Department of Neuroscience, University of Pennsylvania, Philadelphia, PA 19104, USA
In contrast to the well-established roles of the stria-
tum in movement generation and value-based deci-
sions, its contributions to perceptual decisions lack
direct experimental support. Here, we show that
electrical microstimulation in the monkey caudate
nucleus influences both choice and saccade re-
sponse time on a visual motion discrimination task.
Within a drift-diffusion framework, these effects
consist of two components. The perceptual compo-
nent biases choices toward ipsilateral targets, away
from the neurons’ predominantly contralateral re-
sponse fields. The choice bias is consistent with
a nonzero starting value of the diffusion process,
which increases and decreases decision times for
contralateral and ipsilateral choices, respectively.
The nonperceptual component decreases and in-
creases nondecision times toward contralateral and
ipsilateral targets, respectively, consistent with the
caudate’s role in saccade generation. The results
sions used to select saccades that may be distinct
from its role in executing those saccades.
The basal ganglia have been known for more than a century to
play important roles in movement control (Ferrier, 1873; Wilson,
functions, including various forms of decision making, have also
become better appreciated (Brown et al., 1997; Divac et al.,
1967; Middleton and Strick, 2000). For example, the basal
ganglia havebeen causallylinked toreward-modulated behavior
and represent a key component in value-based decision making
(Barto, 1995; Cai et al., 2011; Hikosaka et al., 2006; Hollerman
et al., 2000; Kable and Glimcher, 2009; Samejima and Doya,
2007). It is unclear if and how the basal ganglia also contribute
to perceptual decisions that link sensory input to oculomotor
Support for the basal ganglia’s role in perceptual decision
making comes from several sources. The basal ganglia receive
diverse anatomical inputs from almost all parts of sensory and
sensory-motor cortical areas (Figure 1A). These areas include
the middle temporal (MT) and medial superior temporal (MST)
areas of extrastriate cortex, lateral intraparietal cortex (LIP),
and parts of prefrontal cortex including the frontal eye field
(FEF) (Maunsell and van Essen, 1983; Saint-Cyr et al., 1990; Se-
lemon and Goldman-Rakic, 1985, 1988; Yeterian and Pandya,
1995), all with well-characterized activity related to a task linking
a decision about visual motion to saccadic eye movements (Brit-
ten et al., 1992, 1996; Ding and Gold, 2012; Ditterich et al., 2003;
Hanks et al., 2006; Kim and Shadlen, 1999; Newsome et al.,
1989;Roitman andShadlen, 2002;Salzmanetal.,1992;Shadlen
and Newsome, 1996). Theoretical studies have ascribed several
decision-related computations to specific components of the
basal ganglia (Berns and Sejnowski, 1995; Bogacz and Gurney,
2007; Lo and Wang, 2006; Rao, 2010). Single-unit activity in the
encode a number of decision-related signals in monkeys per-
forming the visual motion saccade task (Ding and Gold, 2010).
fMRI studies revealed striatal activation in human subjects per-
forming visual motion discrimination tasks (Forstmann et al.,
2008; van Veen et al., 2008). In contrast, the frequency of clini-
cally observed perceptual impairments is much lower than that
of motor deficits for diseases associated with basal ganglia
dysfunction (e.g., Parkinson’s disease). This observation seems
to argue against a major role of the basal ganglia in perceptual
decision-making, although non-motor symptoms are often
under-reported or unrecognized by clinicians (Chaudhuri et al.,
In this study, we used electrical microstimulation in the
caudate nucleus in monkeys performing a visual motion discrim-
ination task (Figure 1) to address three questions: (1) is there
a causal link between caudate activity and perceptual decision
behavior? (2) What are the specific decision-related computa-
tions that are influenced by caudate activity? (3) How do the
basal ganglia’s roles in perceptual decisions relate to their roles
in movement control? The results indicate that the basal ganglia
can bias perceptual decisions toward particular alternatives.
These effects are distinct from their role in movement execution.
Thus, the basal ganglia appear to make multiple causal contri-
butions to simple decisions that link sensory input to motor
Neuron 75, 865–874, September 6, 2012 ª2012 Elsevier Inc. 865
Microstimulation Did Not Alter the Monkeys’ Task
As described in previous reports (Ding and Gold, 2010, 2012),
the performance of the two monkeys on the RT dots task de-
pended critically on the strength (coherence) of the motion stim-
ulus. Bothmonkeysachieved near-perfect accuracyandhadthe
shortest RTs for coherences >20%, with steadily decreasing
accuracy and increasing RT at lower coherences (Figure 2).
We fit choice and RT performance simultaneously with a drift-
diffusion model (DDM; see Experimental Procedures and curves
in Figure 2), and we fit choice data alone using logistic functions
(see Figure S2 available online). We quantified performance
using two measures estimated from the fits: choice bias, corre-
sponding to the horizontal position of the psychometric curve
(Figure 2, top panels), and discrimination threshold, correspond-
ing to the steepness of the psychometric curve.
We examined the effects of electrical microstimulation on
performance in 43 sessions (n = 29 and 14, for monkey C and
F, respectively). The microstimulation sites were within the
general regions sampled in our previous recording study (Fig-
ure S1; Ding and Gold, 2010). The motion directions used were
similar to our previous recoding studies for the caudate nucleus
and FEF (Table S1; Ding and Gold, 2010, 2012). The inclusion of
randomly interleaved microstimulation trials did not appear to
affect the monkeys’ overall strategy for solving the task. For
example, when comparing performance on trials without micro-
stimulation from this study to performance of the same monkeys
on the same task in a recent study in which no microstimulation
was used (Ding and Gold, 2012), choice bias and discrimination
threshold were not significantly different for both monkeys and
for all motion axes tested (Wilcoxon rank-sum test, p > 0.05).
Moreover, the DDM fit separately to trials with and without mi-
crostimulation in this study had comparable goodness of fits
(Wilcoxon signed-rank test for H0: equal log-likelihood, p = 0.14).
Microstimulation Influenced Choice and RT
The effects of caudate microstimulation on performance are
shown for two representative sessions in Figure 2. In both cases,
microstimulation caused the monkeys to favor the T1 choice
(ipsilateral to the microstimulation sites), reflected in a leftward
shift of the psychometric function (top panels). The T1 choices
reflected in a downward shift in the chronometric function for
positive coherence values (bottom panels). Using the DDM fit
simultaneously to psychometric and chronometric data, the
change in bias when comparing trials with and without micro-
stimulation (Dbias; positive/negative values imply more T2/T1
choices on microstimulation trials) was ?4.2% and ?5.0%
coherence for monkeys C and F, respectively, for these sessions
(bootstrap methods, p < 0.05 for both). In contrast, Dthreshold
(positive/negative values imply higher/lower threshold on micro-
stimulation trials) was ?1.1% and 2.2% coherence, respectively
(bootstrap methods, p > 0.05 for both).
Across sessions, electrical microstimulation had a consistent
effect on choice biases, inconsistent effects on thresholds, and
mixed effects on RTs (Figure 3). A significant Dbias was
observed in 18 out of 29 and 7 out of 14 sessions for monkeys
Cand F,respectively(we defined asignificant effectasa session
in which the value measured on trials with microstimulation fell
using bootstrapping from trials without microstimulation). More-
over, Dbias tended to be negative, representing an increased
preference for ipsilateral or upward choices (Figure 3A; mean
Dbias = ?2.7% coherence, t test for H0: mean = 0, p < 0.0001).
Incontrast, asignificant Dthreshold wasobserved inonly8and1
sessions for monkeys C and F, respectively, with a mean value
across all sessions that did not differ significantly from zero (Fig-
ure 3B; mean = 0.2% coherence; p = 0.47). For sessions with
significant nonzero Dbias, the mean RT for correct, microstimu-
lation-favored choices was shorter on microstimulation trials
(Wilcoxon signed-rank test, p = 0.0026), an effect that was larger
for lower coherences (Figure 3C, circles). The mean RT for
correct, other choices was not different between microstimula-
tion conditions (p = 0.37; Figure 3C, triangles). For other ses-
sions, RT was not significantly affected for either choice (p =
0.41 and 0.15 for T1 and T2 choices, respectively).
The observed Dbias was robust and independent of our
choice of fitting with a DDM. Using logistic-only psychometric
Figure 1. Simplified Oculomotor System
Diagram and the Behavioral Task
(A) A schematic drawing illustrating the major
connections of the oculomotor basal ganglia for
temporal visual area; LIP, lateral intraparietal
cortex; FEF, frontal eye field; GPe, external
segment of the globus pallidus; STN, subthalamic
nucleus; SNr, substantia nigra pars reticulata; SC,
superior colliculus; DA, dopamine.
(B) Behavioral task. The monkey decides the
direction of random-dot motion and then re-
sponds, at a self-determined time, by making
a saccade to foveate one of two choice targets.
Saccades to the target in the direction of coherent
motion (assigned randomly for 0% coherence)
are followed by juice reward. Electrical micro-
stimulation is delivered during motion viewing and
terminated when the monkey makes a saccade.
Caudate Stimulation Biases Perceptual Decisions
866 Neuron 75, 865–874, September 6, 2012 ª2012 Elsevier Inc.
fits, microstimulation-induced Dbias was observed in 17 out of
29 and 9 out of 14 sessions for monkeys C and F, respectively
(Figure S2). The mean Dbias was ?2.5% coherence (p <
0.0001). The mean Dthreshold was ?0.2% coherence, which
did not differ significantly from zero (p = 0.57). The microstimula-
tion effects also did not persist beyond the trial on which it was
applied: there were no consistent effects on bias or threshold
for the next trial (considering only such trials without microstimu-
lation) in sessions with a significant current-trial effect (paired
t test, p = 0.42 and 0.34 for bias and threshold, respectively).
Microstimulation-Induced Choice Bias Correlated with
Neuronal Spatial Tuning
Some of the session-by-session variability in Dbias reflected
differences in the spatial tuning properties of nearby neurons
(Figure 4). We quantified the spatial selectivity of isolated units
by computing an ROC index constructed from average firing
rates recorded during motion viewing, separated by T1 and T2
choices (Figure S3). ROC index values <0.5 represent higher
firing rates for T2 choices, whereas values >0.5 represent higher
firing rates for T1 choices. Sites where more than one spatially
tuned neurons were recorded on the same electrode were
excluded (n = 3).
For sites with statistically significant Dbias, the microstimula-
tion-induced bias tended to be in the opposite direction as the
spatial selectivity of nearby neurons (Figure 4). As previously
reported, caudate neurons have predominantly contralateral
spatial preferences for this task (Ding and Gold, 2010), reflected
as more data points to the left of the vertical dashed line in the
figure (ROC indices <0.5). In contrast, the microstimulation-
induced choice bias predominantly favored ipsilateral choices,
reflected as more data points below the horizontal dashed line
in the figure (Dbias < 0).
In addition, the magnitude of Dbias varied systematically with
the value of the ROC index (linear regression in Figure 4). Micro-
stimulation at sites with the strongest contralateral-preferring
responses tended to have the strongest ipsilateral-biasing
effects. In contrast, microstimulation at sites with the strongest
eral-biasing effects. Forthe ninesiteswith statistically significant
Dthreshold, wedidnotobserve any relationshipbetween thresh-
old change and neuronal spatial selectivity (data not shown;
linear regression, p = 0.88).
Microstimulation-Induced Choice Bias Reflected
a Nonzero Starting Value in DDM
We further used the DDM to test two (not necessarily mutually
exclusive) hypotheses about the source of the microstimula-
tion-induced choice biases (Hanks et al., 2006). One possibility
is that microstimulation corresponds to an asymmetry in the
amount of evidence needed for each of the two choices. When
the decision bounds are fixed, the starting value (SV) of the
accumulating decision variable in the DDM controls the relative
diffusion distance for the two choices. In this case, a positive
SV indicates a decrease in the distance for T1 choices and an
increase in the distance for T2 choices, thus creating a bias
toward T1 choices. The second possibility is that microstimula-
tion adds momentary evidence (ME) to the accumulating deci-
sion variable, favoring the more frequent choice. In this case,
ME modulates the rate of accumulation, with a positive value
indicating that extra evidence for the T1 choice is added at every
time step during evidence accumulation, thus creating a bias
toward T1 choices.
Based on parameter fits using revised DDMs, the microstimu-
lation-induced bias was better characterized using nonzero
values of SV than ME (Figure 5). Using a model containing both
SV and ME terms, best-fitting values of SV, but not ME, tended
to be different from zero and thus account for the choice biases
(Figure 5A; sign test for zero median: p = 0.004 and 0.286,
respectively). Using two reduced models with either SV or ME
terms,but notboth, thefits yielded positive values forboth terms
and thus could, in principle, account for a negative Dbias (Fig-
ure 5B; median: 12% of bound distance and 2.7% coherence;
p = 0.0002 and 0.0041, respectively). However, the SV-only
model accounted for the observed Dbias better than the ME-
only model (Figure 5C), resulting in a larger log-likelihood (equiv-
alent to smaller Bayesian information criteria, or BIC, given the
same number of parameters for the two models; Wilcoxon
signed-rank test, p = 0.012). Similar results hold if only sessions
with negative Dbias were included in the analyses. Thus, within
the DDM framework, microstimulation-induced choice bias
was better characterized as a change in the relative amount of
evidence needed for each choice than a change in the actual
Microstimulation-Induced Changes in RT Reflected
Both a Nonzero Starting Value and Changes in
Nondecision Times in DDM
However, the SV term alone did not fully explain the microstimu-
starting value alone is expected to decrease and increase
Figure 2. Example Performance
as thefractionof trials inwhichthe monkey chose the T1 target as afunction of
signed coherence, where positive/negative coherence indicates motion
toward T1/T2. Chronometric functions are plotted as the mean RT, measured
as the time between motion stimulus onset and saccade onset, on correct
trials as a function of signed coherence. Solid curves are simultaneous fits of
both functions to a drift-diffusion model (DDM) with asymmetric bounds,
separately for trials with (red) and without (black) electrical stimulation.
Caudate Stimulation Biases Perceptual Decisions
Neuron 75, 865–874, September 6, 2012 ª2012 Elsevier Inc. 867
decision time toward T1 and T2 choices, respectively, with
similar magnitudes (for example, see Figure S4 and the shaded
areas in Figure 6H). In contrast, caudate microstimulation re-
sulted in increases in RT toward T2 that were much smaller in
absolute magnitude than the decreases in RT toward T1 (Figures
3C and 6H, blue and red curves, respectively).
These RT effects did not result from our microstimulation
protocol evoking inappropriate eye movements. For example,
microstimulation did not evoke saccades or cause small eye
movements: the standard deviation of eye position before
saccade onset did not differ between trials with and without mi-
crostimulation (0.17?± 0.06?versus 0.16?± 0.04?; paired t test,
individual sessions). Microstimulation also did not change the
percentage of trials when the monkey broke fixation or made
a saccade to neither choice target (paired t test, p = 0.59 and
more subtle, asymmetric changes in saccade initiation. In the
DDM framework, the time for decision formation is controlled
by the two decision bounds (A and B), drift rate (scaled by k),
tion, including motor preparation and saccade initiation, is
aggregated as nondecision times. RT is the sum of the decision
and nondecision times. We thus hypothesized that microstimu-
lation might also influence nondecision times. We compared
six variants of the DDM and three variants of race models to
the full DDM (i.e., the model used in Figure 2) to identify the
model that can best account for the microstimulation effects
on psychometric and chronometric functions and also capture
the microstimulation effect on cumulative RT distributions. The
fitting parameters for the ten models are listed in the second
column of Table S2. Briefly, the DDM variants use combinations
of parameters to capture the microstimulation effects: SV; ME;
choice-dependent changes in nondecision times (DT01 and
DT02); and changes in A, B, and k. The race model variants use
mulator alone, and changes in rectification threshold (q). For
these comparisons, we focused on sessions with negative Dbias
(n = 22) and used as the baseline the fitting results from the full
DDM that fits trials with and without microstimulation separately
In general, the DDM variants performed much better than the
race models in fitting psychometric and chronometric functions
(Figures 6A and 6B). We assessed goodness of fit using two
methods: log likelihood (Figure 6A; Table S2), which does not
take into account different numbers of fitting parameters, and
BIC (Figure 6B), which does. Both methods had smaller values
(i.e., better fits) for all of the DDM variants than for all of the
race models. BIC could not distinguish between the different
variants of the DDM.
A more detailed analysis of the RT distributions indicated that
our results were best matched by the two DDM variants that
included an SV term and independent changes in nondecision
times(model 2: SV +ME +2DT0; and model 3:SV +2DT0). These
two DDM variants (along with the full model) and a race-model
variant (model 8) produced the smallest sum-of-squares error
between observed and simulated cumulative RT distributions
(Figure 6C). The better fits provided by these models can be
panels depict the change incumulative RT distributions between
all trials with microstimulation versus all trials without microsti-
mulation, computed separately for T1 (red) and T2 (blue) choices
(see Figure S4 for details). Models 2, 3, and 8 most effectively
captured the asymmetry in microstimulation-induced changes
in RT (larger effects for T1 than T2 choices) that we observed
(Figures 6D, 6E, and 6J, respectively; data for model 1 not
shown). Models 2 and 3 also most effectively captured the
dynamics of the microstimulation-induced changes in T1 RTs,
including little change in short RTs but rapidly increasing effects
for RTs >?500 ms (arrows). The goodness of fits of models 2, 3,
and8forthe changesincumulative RTdistribution, asmeasured
by sum of squared error or R2, do not differ significantly (t test,
p > 0.05). However, model 8 is worse than models 2 and 3 for
fitting both psychometric and chronometric functions (Figures
6A and 6B; Wilcoxon signed-rank test, p < 0.0001), indicating
that the DDM models provided better overall fits.
Using model 3, which had the fewest parameters of models
1–3, best-fitting values of SV had a mean value of 12.2% of
bound distance (sign test for nonzero median, p < 0.0001),
the nondecision time for choice T1 was prolonged by a median
value of 41 ms (p = 0.004), and the nondecision time for T2
was shortened by a median value of 62 ms (p = 0.0008). Thus,
caudate microstimulation seemed to have two effects: (1)
a motion stimulus-dependent effect that promoted choices to
T1, and (2) a motion stimulus-independent effect that delayed
the execution of saccades to T1 and facilitated the execution
of saccades to T2. These results were consistent with the
influence of caudate microstimulation on separable decision
and saccade processes, as opposed to two independent deci-
sion processes corresponding to the two alternatives in a race
Figure 3. Summary of Microstimulation
Effects (n = 43 Sessions)
(A and B) Scatterplots of bias (A) and threshold (B)
estimates from DDM fits for trials with (ordinate)
and without (abscissa) caudate microstimulation.
Sessions with significant stimulation effects (see
Experimental Procedures) are indicated by filled
(C) Median difference in mean RT between trials
with and without microstimulation. Error bars
indicate 25thand 75thpercentiles.
Caudate Stimulation Biases Perceptual Decisions
868 Neuron 75, 865–874, September 6, 2012 ª2012 Elsevier Inc.
The caudate nucleus has been shown previously to contribute
causally to saccade generation, the evaluation of expected
outcomes, and mediation of reinforcement-based and associa-
tive learning (Kitama et al., 1991; Nakamura and Hikosaka,
2006a, 2006b;Watanabeand Munoz,2010;Williams and Eskan-
dar, 2006). In this study, we used electrical microstimulation to
demonstrate for the first time that the caudate also causally
contributes to perceptual decision making. Applying microsti-
mulation in the caudate of monkeys performing a direction-
discrimination task affected both choice and RT. The effect on
choice was consistent with an offset in the starting or ending
value of an evidence-dependent accumulation process defined
by a commonly used model of decision making, the DDM. The
effect on RT was consistent with the combined effects of the
offset and concomitant facilitation and suppression of saccades
toward contralateral and ipsilateral targets, respectively.
A main goal of this study was to help to position the basal
ganglia pathway computationally in the overall decision process
for this task. Anatomically, the caudate receives input from
numerous cortical structures that contribute to the decision (Fig-
ure 1A). These structures include area MT in extrastriate visual
cortex, which contains direction-selective neurons that provide
sensory evidence and areas LIP in parietal cortex and FEF in
prefrontal cortex, both of which contain neurons that encode
the process of accumulating the sensory evidence over time
intoadecision variablethatgovernsthesaccadic response(Brit-
ten et al., 1992, 1996; Ding and Gold, 2012; Kim and Shadlen,
1999; Roitman and Shadlen, 2002; Shadlen and Newsome,
1996). Outputs of the oculomotor basal ganglia pathway
target the superior colliculus, which also receives direct input
from LIP and FEF and contains neurons that similarly encode
the evidence-accumulation process (Horwitz and Newsome,
1999). We recently showed that certain task-driven neuronal
like in LIP, FEF, and the superior colliculus but not in MT (Ding
and Gold, 2010). Our present results are consistent with these
findings, indicating that caudate plays a similar, causal role in
decision making as that found previously for LIP but not MT
using a comparable microstimulation protocol (Ditterich et al.,
2003; Hanks et al., 2006). Together, these findings suggest
that evidence accumulation used to instruct saccadic choices
is implemented in a set of interconnected brain regions including
LIP, FEF, the superior colliculus, and the basal ganglia pathway
that indirectly links these cortical and subcortical structures.
Despite the similarities between our results and those for area
LIP, we note two striking differences. The first is in the sign of
choice bias, which for caudate is toward the target ipsilateral
to the site of microstimulation but for LIP is toward the target
contralateral to the site of microstimulation. The opposite signs
are unlikely simply due to a difference in microstimulation pulse
Figure 4. Microstimulation-Induced Choice Bias Correlates with the
Spatial Tuning of Neurons Recorded at the Site of Microstimulation
Spatial tuning was quantified using an ROC index (see also Figure S3); values
greater/less than 0.5 indicate stronger responses for T1/T2 choices. Choice
bias is quantified as Dbias, the difference in estimated bias for trials with and
without microstimulation; positive/negative values indicate increased prefer-
ences for T2/T1 choices. Data clustering in the top right and bottom left
quadrants indicate that the microstimulation biases choice away from
induced Dbias and single recorded neurons are included (n = 22). Line is
a linear fit (F value = 7.41, p = 0.013).
Figure 5. Microstimulation-Induced Dbias Is Captured by a Nonzero
Starting Value in the DDM
(A) Histograms of best-fitting parameters SV (starting value, left) and ME
(momentary evidence, right) in a DDM model with both terms (model 2; see
also Table S2 and Experimental Procedures). Arrows indicate median values.
SV is expressed as the percentage of bound distance (A + B in DDM).
(B) Histograms of best-fitting parameters SV (left)and ME (right) in the reduced
models with each term alone (models 3 and 4, respectively).
(C) Scatterplots of Dbias measured from experimental data (ordinate), using
independent DDM fits for trials with and without microstimulation, and Dbias
predicted (abscissa) using the SV-alone model (left, model 3) or the ME-alone
model (right, model 4). Solid lines are linear fits (slope: 0.99 and 0.45; F value:
251.8 and 9.9; p < 0.0001 and 0.0045, respectively). For these analyses, sites
with significant Dbias were included (n = 25).
Caudate Stimulation Biases Perceptual Decisions
Neuron 75, 865–874, September 6, 2012 ª2012 Elsevier Inc. 869
frequency, given that caudate microstimulation tends to have
consistent effects on saccade behavior over a large frequency
range (5–333 Hz; Watanabe and Munoz, 2010). The ipsilateral
choice bias with caudate microstimulation is also unlikely anarti-
ship with the nearby neurons’ tuning properties (Figure 4). It is
conceivable that caudate microstimulation antidromically acti-
vates a distal, upstream region that has an opposite role to
LIP’s in perceptual decision making, although such a region
has not yet been identified.
We thus consider an alternative explanation based on the
intrinsic organization of the basal ganglia. The basal ganglia
are organized into direct and indirect pathways (Figure 1A),
which are first segregated in the striatal population of projection
neurons (DeLong, 1990; Graybiel and Ragsdale, 1979; Hikosaka
et al., 1993; Hikosaka and Wurtz, 1983, 1985; Niijima and Yosh-
ida, 1982). Activation of striatal projection neurons in the two
pathways is assumed to have opposite effects on the basal
ganglia output, resulting in net excitation or inhibition of the
Capturing Both Choice and RT Effects
(A) Difference in mean negative log likelihood
between the given model and model 1 (two inde-
pendent DDMs fit to trials with and without
microstimulation). Note that the first bar is at zero
by definition. The value for model 2 is close to
zero. Smaller values indicate better fits. See Table
S2 and (D)–(L) for model identities. Filled bars
indicate a significant nonzero difference in means
(t test, p < 0.05, with corrections for multiple
(B) Difference in Bayesian information criterion
from model 1. Same conventions as (A). Smaller
values indicate better fits.
(C) Sum of squared error (SSE; mean ± SD)
between the observed and simulated micro-
stimulation-induced change in cumulative RT
distributions. Filled bars indicate significantly
smaller SSE in models 2, 3, and 8 than the others
(p < 0.05, with corrections for multiple compar-
(D–L) Difference in the cumulative distribution of
RT between trials with and without micro-
stimulation (see Figure S4) for models 2–10, as
indicated. Experimental data are shown as solid
curves. Simulation results are shown as shaded
areas (mean ± SD). Red: T1 choice; blue: T2
choice. Arrows point to the rising portion of the T1
curve for the experimental data.
6. Comparisonof Modelsfor
superior colliculus for the direct or indi-
rect pathway, respectively (Figure 1A).
these pathways has been shown to
and reinforcement learning (Kravitz et al.,
2010, 2012; Nakamura and Hikosaka,
2006b) and thus, in principle, could
have opposite effects on perceptual
decisions. These two subpopulations of
striatal projection neurons, although physically intermingled
and indistinguishable with extracellular recordings, differ in their
somatodendritic and synaptic properties (Ade et al., 2008; Ce-
peda et al., 2008; Day et al., 2008; Flores-Barrera et al., 2010;
Gerfen et al., 1990; Gertler et al., 2008; Shen et al., 2007). We
speculate that our electrical microstimulation preferentially acti-
downstream oculomotor structures including the superior colli-
culus. It will be interesting to test this hypothesis by specifically
targeting the direct and indirect pathways with pharmacological
manipulations or by comparing activity patterns in caudate and
FEF/LIP withthose foundincomponents of theindirect pathway,
such as the subthalamic nucleus or the external segment of
The second difference between caudate and LIP microstimu-
lation is their effects on RT. LIP microstimulation shortens RT for
the favored choice and increases RT for the other choice with
a similar magnitude (Hanks et al., 2006). In contrast, the effect
of caudate microstimulation on RT is not symmetric for the two
Caudate Stimulation Biases Perceptual Decisions
870 Neuron 75, 865–874, September 6, 2012 ª2012 Elsevier Inc.
choices. Based on our modeling efforts, the best explanation for
these caudate microstimulation results is a combined effect on
a perceptual process favoring ipsilateral choices and a nonper-
saccades. The two effects may result from the influence of mi-
crostimulation on different neural assemblies in the caudate
nucleus. This idea is consistent with the functional anatomy of
the basal ganglia pathway, which is known to contain multiple
parallel loops both in overall function (e.g., limbic, motor, asso-
ciative) and in more microscopic domains (e.g., topographic
projections throughout the pathway for body regions; Alexander
and Crutcher, 1990; Alexander et al., 1986; Parent and Hazrati,
1995). A highly speculative scenario may be that activation of
the direct pathway of the ‘‘motor’’ loop decreases and increases
nondecision times for contra- and ipsilateral saccades, respec-
tively, whereas activation of the indirect pathway of the ‘‘percep-
tion’’ loop biases choice toward ipsilateral targets.
The idea that caudate encodes two distinct, task-related
processes—one involved in forming the perceptual decision,
the other in oculomotor control—may also help to bridge seem-
ingly conflicting results from previous studies of caudate micro-
stimulation. Specifically, caudate microstimulation can evoke
contraversive saccades and, when delivered before saccade
onset and at sites with neural activity modulated on a simple
visually guided saccade task, reduces RT for contraversive sac-
cades (Kitama et al., 1991; Nakamura and Hikosaka, 2006a).
These results suggest a facilitatory effect of microstimulation
on contraversive saccades. In contrast, when delivered before
saccade onset at ‘‘blindly’’ sampled sites, caudate microstimu-
lation increases RT for contraversive saccades and, to a lesser
extent, decreases RT for ipsiversive saccades on a pro-/antisac-
cade task (Watanabe and Munoz, 2010, 2011). These results
suggest a suppressive effect of microstimulation on contraver-
sive saccades. In light of our observations, these previous
reports may have resulted from differential activation of distinct
that preferentially activates neurons participating in saccade
generation facilitates generation of contraversive saccades. In
contrast, microstimulation that preferentially activates neurons
participating in perceptual-decision formation or other cogni-
tively demanding forms of saccade selection facilitates selection
of ipsilateral saccade targets. The former effect dominates for
evoked saccades and for simple saccade tasks with targeted
microstimulation sites. Both effects are in place for pro-/antisac-
cade tasks with blindly sampled microstimulation sites and for
the dots task. The dots task enables the dissociation of percep-
tual decision-making and saccade effects, with manipulations of
stimulus strength (Petrov et al., 2011).
In contrast to the microstimulation effects on choice bias, we
did not observe a consistent effect on discrimination threshold.
This result is consistent with our interpretation of caudate
response properties in the context of the DDM (Ding and Gold,
2010). According to that framework, discrimination threshold is
determined by the decision bounds and a constant of propor-
tionality used to convert the evidence to a log likelihood ratio-
related quantity (Gold and Shadlen, 2002; Ratcliff, 1978). The
decision bounds govern the speed-accuracy tradeoff and in
our previous study were not encoded in caudate: unlike in LIP
and FEF, evidence-accumulation activity in caudate did not
converge at a DDM-like bound just prior to saccade onset on
theRT dotstask(Dingand Gold, 2010, 2012;Roitman andShad-
len, 2002). The constant of proportionality may already be incor-
porated in the inputs from MT and thus not influenced by
caudate microstimulation. However, despite this consistency
with our previous recording study, the lack of an effect on
discrimination threshold is not consistent with previous compu-
tational modeling and fMRI studies that posit a role for the basal
ganglia pathway in mediating the appropriate speed-accuracy
tradeoff (Bogacz et al., 2010; Brown et al., 2004; Forstmann
et al., 2008; Frank, 2006; Gurney et al., 2004; Lo and Wang,
2006; Rao, 2010; van Veen et al., 2008). This discrepancy might
kind of task-modulated neural activity we described previously.
More work is needed to identify other possible neural correlates
of the decision boundand determine their causal role in the deci-
In summary, caudate microstimulation influences both choice
and RT in monkeys performing a demanding perceptual deci-
sion task. These effects support causal roles of the caudate
nucleus—and by extension the basal ganglia—in mediating
perceptual decision formation and saccade generation. In con-
junction with their reported roles in valuation of different options,
the basal ganglia are well positioned to play important roles in
real-life, complex decisions that must take into account of
multiple sources of external inputs and internal preferences.
We used two adult male rhesus monkeys (Macaca mulatta) that were previ-
ously trained on the direction-discrimination (dots) task used in this study
(Ding and Gold, 2010, 2012). Each monkey was implanted with a head holder
for identifying and recording from caudate neurons are described previously
(Ding and Gold, 2010). Prior to the microstimulation experiment, monkey C
was trained on various versions of the dots task for >5 years and used for
data collection in three previous studies (Ding and Gold, 2010, 2012; Law
and Gold, 2008); monkey F was trained for two years and used for data collec-
tion in two previous studies (Ding and Gold, 2010, 2012). Both monkeys
showed clear sensitivity to motion strength and stimulus duration on a fixed-
duration version of the task (monkey C, Law and Gold, 2008; monkey F, Fig-
ure S5) and speed-accuracy tradeoff on the RT task (Ding and Gold, 2010).
All training, surgery, and experimental procedures were in accordance with
the National Institutes of Health Guide for the Care and Use of Laboratory
Animals and were approved by the University of Pennsylvania Institutional
Animal Care and Use Committee.
The dots task requires the subject to decide the direction of random-dot
motion and respond as soon as the decision is formed with a saccadic eye
movement (Figure 1A; Ding and Gold, 2010). Briefly, after the monkey main-
tained central fixation for an exponentially distributed duration, a random-
dot motion stimulus was presented in a 5?aperture centered on the fixation
point, with a fixed velocity of 6?/s in one of two opposite motion directions.
Motion direction and strength (the percent of dots moving coherently in one
used were 0%, 3.2%, 6.4%, 12.8%, 25.6%, and 51.2%. To increase the
number of trials per condition for microstimulation experiments, 51.2% coher-
ence trials were omitted in 14 sessions for monkey C, who consistently
performs at 100% correct for 51.2% and nearly 100% correct at 25.6% coher-
ence without microstimulation. After stimulus onset, the monkey was free to
indicate its decision about the motion direction at any time by making
a saccade to the corresponding visual choice target. The choice targets
Caudate Stimulation Biases Perceptual Decisions
Neuron 75, 865–874, September 6, 2012 ª2012 Elsevier Inc. 871
tion point and at an eccentricity of 10?(Figure 1B). The stimulus was turned off
once a saccade was detected. The monkey was rewarded with juice for
choosing the correct choice target (congruent with the motion direction at
nonzero coherence levels; randomly picked for 0%-coherence trials). Eye
position was monitored using a video-based system (ASL) sampled at
240 Hz. Reaction time (RT) was measured as the time from stimulus onset to
saccade onset, the latter identified offline with respect to velocity (>40?/s)
and acceleration (>8,000?/s2).
At the beginning of a session, we identified a caudate site with single- or
multiunit activity modulated on the dots task. Neural activity was recorded
using glass-coated tungsten electrodes (Alpha-Omega) or polyamide-coated
tungsten electrodes (FHC, Inc.). The motion direction that elicited the largest
responses was determined by online visual inspection and then used to define
the axis of motion for the dots task used in the remainder of the experimental
session (Table S1). Unlike cortical regions such as MT and LIP, the caudate is
not topographically organized, and nearby neurons do not necessarily share
the same response profiles (Ding and Gold, 2010; Hikosaka et al., 1989). We
without considering nearby neural activity. Electrical microstimulation was
delivered at the same site during motion stimulus presentation (negative-
parameters were chosen to maximize potential effect sizes while avoiding
evoked saccades (Nakamura and Hikosaka, 2006a; Watanabe and Munoz,
2010, 2011). Because higher currents are needed to activate the thinner,
sparsely myelinated projection axons in the caudate nucleus compared to
the thicker, more myelin-dense projection axons of the cortex, the current
intensity used is expected to have similar effective current spread to that of
comparable microstimulation studies in cortex (Adinolfi and Pappas, 1968;
Blatt et al., 1990; Felleman and Van Essen, 1991; Spatz and Tigges, 1972; Te-
hovnik, 1996; Tomasi et al., 2012). Trials with and without microstimulation
were equally divided and randomly interleaved in a session. The neural
responses were sorted offline (Plexon, Inc.). Each neuron’s spatial selectivity
was quantified as a receiver operating characteristic (ROC) index, which is
the area under the ROC curve constructed using average spike rate during
motion viewing (from 200 ms after stimulus onset to 100 ms before saccade
onset, all coherence levels were included; also see Figure S3).
Performance was quantified with psychometric and chronometric functions
(Figure 2), which describe the relationship of motion strength (signed coher-
ence, Coh, positive for toward T1, negative for toward T2) with choice and
RT, respectively. Per laboratory convention, T1 choice targets were placed
in the right hemifield (ipsilateral to the microstimulation sites) or directly above
the central fixation point (for T1 = 90?). Performance on trials with and without
microstimulation was analyzed using five methods, as follows (all model fits
were accomplished using maximum-likelihood methods):
First, we fit psychometric and chronometric functions simultaneously to
a drift-diffusion model (DDM), which has been used successfully to describe
performance of both monkey and human subjects on the RT dots task
(Hanks et al., 2006; Palmer et al., 2005). Here, we used separate fits for
trials with and without microstimulation, using a model with five free parame-
ters (model 1): A, B, k, T01, and T02. According to this model, momentary
motion evidence is assumed to follow a Gaussian distribution N(m, 1), the
mean of which, m, scales with coherence m = k 3 Coh, where k governs the
coherence-dependent drift. A decision variable is computed as the temporal
accumulation of this momentary motion evidence. A decision (T1 or T2) is
reached when the value of the decision variable reaches a decision bound
(+A or –B, respectively). Decision time is defined as the interval between
time and non-decision time (T01for a T1 choice and T02for a T2 choice).
Within this framework, the probability of choosing T1 (i.e., the probability
e2 mB? 1
e2 mB? e?2 mA.Theaverage
mcothðmBÞ for T1 decisions and
decision time is
from the choice function as one-half the difference in coherence correspond-
mcothðmAÞ for T2 decisions. Threshold was estimated
ing to 25% and 75% T1 choices (Klein, 2001). Bias was defined as the signed
percent coherence corresponding to 50% T1 choices. Distributions of esti-
mated threshold and bias for trials without microstimulation were estimated
by repeating fits with resampled trials. A statistically significant microstimula-
tion effect was identified if the value from microstimulation trials fell outside
the mean ± 2 SD of the values from resampled no-microstimulation trials. An
alternative bootstrapping method, which estimates the probability of obtaining
the experimentally observed Dbias/Dthreshold from all trials with shuffled
microstimulation conditions, gave similar results (data not shown).
Second, to ensure that our results were not overly conditioned on the
assumptions of the DDM, we also fit psychometric data alone using a logistic
1+e?bðCoh?aÞ, separately for trials with and without microstimula-
tion. Discrimination threshold was defined as one-half the difference in
coherence corresponding to 25% and 75% T1 choices from the fitted function
(Klein, 2001). Bias was defined as the value of a. Statistically significant micro-
stimulation effects were detected using the bootstrap methods described
Third, we adopted a modified DDM to fit psychometric and chronometric
functions simultaneously for trials with and without microstimulation (model
2). This model uses the same basic parameters as in the above drift-diffusion
model (A, B, k, T01, and T02). In addition, we introduced two terms similar to
a previous study to account for the microstimulation-induced choice biases
(Hanks et al., 2006): starting value (SV) and momentary evidence (ME). SV
was implemented as a change in decision bounds: +A/-B for no microstimula-
tion trials and +A-SV/-B-SV for microstimulation trials. ME was implemented
as a change in momentary motion evidence: m = k 3 Coh for no
microstimulation trials and m = k 3 (Coh + ME) for microstimulation trials. Posi-
tive SV or ME corresponds to an increased bias toward T1. To account for
possible microstimulation effects on nondecision processes, we introduced
two additional nondecision times (T010and T020) for trials with microstimulation.
Fourth, to further investigate effects of microstimulation on both choice and
RT, we compared goodness of fits of six versions of the DDM (models 2–7). All
of these models use the five basic parameters as in the above drift-diffusion
model: A, B, k, T01, and T02. In addition, they use combinations of additional
parameters to capture the microstimulation effects (see Table S2 for more
details): SV; ME; choice-dependent changes in nondecision times (two sets
of T01and T02for trials with and without microstimulation); and changes in
A, B, and k (two sets of A, B, and k for trials with and without microstimulation).
We also implemented race models of independent accumulators with rectified
inputs (models 8–10; Smith and Vickers, 1988) to test for the possibility that
caudate’s role in the decision process is inconsistent with a basic assumption
of DDM, that a single decision variable governs the decision process. Accord-
ing to the basic race model, momentary motion evidence is assumed to follow
a Gaussian distribution N(m, 1), the mean of which, m, scales with coherence:
m = k 3 Coh, where k governs the coherence-dependent drift. The motion
evidence is compared to a threshold q. One accumulator integrates the differ-
encebetweenthemotion evidence and qonlyifthe difference ispositive,while
the other accumulator integrates the difference only if the difference is nega-
tive. If the first accumulator reaches bound +A before the other reaching
bound -B, a choice toward T1 is made; if the second accumulator reaches
bound -B first, a choice toward T2 is made. The steps of accumulation is
converted to actual decision time by a scaling factor, a. Similar to the DDM,
RT is the sum of decision and nondecision times (T01and T02). To capture
the microstimulation effects, we considered three variations of the basic
ME value added at each step of accumulation for the first accumulator, and (3)
a change in q. Goodness of fit was measured as the log-likelihood for each
model and compared across models using BIC to take into account different
numbers of fitting parameters.
Fifth, we examined microstimulation-induced effects on RT distributions.
For each session, we collapsed trials (correct and error) across coherence
levels and computed the cumulative RT distributions, separately for the two
choices and microstimulation conditions (Figure S4). For each choice, we
computed the difference in cumulative RT distributions between trials with
and without microstimulation. The microstimulation effect on the RT distribu-
tion was measured as the average difference across sessions, separately for
Caudate Stimulation Biases Perceptual Decisions
872 Neuron 75, 865–874, September 6, 2012 ª2012 Elsevier Inc.
the two choices. For model predictions, choice and RT data were simulated
with session-specific fitting parameters and with trial numbers for the different
coherence 3 direction conditions matched to the experimental data. Simu-
lated data were analyzed in the same way as the experimental data. Mean
and standard deviation of the simulated difference in cumulative RT distribu-
tion were estimated using bootstrap methods.
Supplemental Information includes five figures and two tables and can
be found with this article online at http://dx.doi.org/10.1016/j.neuron.2012.
Wethank Takahiro Doi, Matt Nassar, and Yin Lifor helpful commentsand Jean
Zweigle for animal care. This work was supported by NIH K99–EY018042 and
ARRA supplement (L.D.) and R01–EY015260 (J.I.G.) from the National Eye
Accepted: July 19, 2012
Published: September 5, 2012
Ade, K.K., Janssen, M.J., Ortinski, P.I., and Vicini, S. (2008). Differential tonic
GABA conductances in striatal medium spiny neurons. J. Neurosci. 28,
Adinolfi, A.M., and Pappas, G.D. (1968). The fine structure of the caudate
nucleus of the cat. J. Comp. Neurol. 133, 167–184.
Alexander, G.E., and Crutcher, M.D. (1990). Functional architecture of basal
ganglia circuits: neural substrates of parallel processing. Trends Neurosci.
Alexander, G.E., DeLong, M.R., and Strick, P.L. (1986). Parallel organization of
functionally segregated circuits linking basal ganglia and cortex. Annu. Rev.
Neurosci. 9, 357–381.
Barto, A.G. (1995). Adaptive critics and the basal ganglia. In Models of
Information Processing in the Basal Ganglia, J.C. Houk, J.L. Davis, and D.G.
Beiser, eds. (Cambridge, MA: MIT Press), pp. 215–232.
Berns, G.S., and Sejnowski, T.J. (1995). How the basal ganglia make deci-
sions. In Neurobiology of Decision-Making, A.R. Damasio, H. Damasio, and
Y. Christen, eds. (Berlin: Springer-Verlag), pp. 101–113.
nization and cortico-cortical connections of the lateral intraparietal area (area
LIP) in the macaque. J. Comp. Neurol. 299, 421–445.
Bogacz, R., and Gurney, K. (2007). The basal ganglia and cortex implement
optimal decision making between alternative actions. Neural Comput. 19,
The neural basis of the speed-accuracy tradeoff. Trends Neurosci. 33, 10–16.
Britten, K.H., Shadlen, M.N., Newsome, W.T., and Movshon, J.A. (1992). The
analysis of visual motion: a comparison of neuronal and psychophysical
performance. J. Neurosci. 12, 4745–4765.
Britten, K.H., Newsome, W.T., Shadlen, M.N., Celebrini, S., and Movshon, J.A.
(1996). A relationship between behavioral choice and the visual responses of
neurons in macaque MT. Vis. Neurosci. 13, 87–100.
Brown, L.L., Schneider, J.S., and Lidsky, T.I. (1997). Sensory and cognitive
functions of the basal ganglia. Curr. Opin. Neurobiol. 7, 157–163.
Brown, J.W., Bullock, D., and Grossberg, S. (2004). How laminar frontal cortex
and basal ganglia circuits interact to control planned and reactive saccades.
Neural Netw. 17, 471–510.
Cai, X., Kim, S., and Lee, D. (2011). Heterogeneous coding of temporally dis-
counted values in the dorsal and ventral striatum during intertemporal choice.
Neuron 69, 170–182.
Cepeda, C., Andre ´, V.M., Yamazaki, I., Wu, N., Kleiman-Weiner, M., and
Levine, M.S. (2008). Differential electrophysiological properties of dopamine
D1 and D2 receptor-containing striatal medium-sized spiny neurons. Eur. J.
Neurosci. 27, 671–682.
Chaudhuri, K.R., Healy, D.G., and Schapira, A.H.; National Institute for Clinical
Excellence. (2006). Non-motor symptoms of Parkinson’s disease: diagnosis
and management. Lancet Neurol. 5, 235–245.
Day, M., Wokosin, D., Plotkin, J.L., Tian, X., and Surmeier, D.J. (2008).
Differential excitability and modulation of striatal medium spiny neuron
dendrites. J. Neurosci. 28, 11603–11614.
DeLong, M.R. (1990). Primate models of movement disorders of basal ganglia
origin. Trends Neurosci. 13, 281–285.
Ding, L., and Gold, J.I. (2010). Caudate encodes multiple computations for
perceptual decisions. J. Neurosci. 30, 15747–15759.
Ding, L., and Gold, J.I. (2012). Neural correlates of perceptual decision making
before, during, and after decision commitment in monkey frontal eye field.
Cereb. Cortex 22, 1052–1067.
Ditterich, J., Mazurek, M.E., and Shadlen, M.N. (2003). Microstimulation of
visual cortex affects the speed of perceptual decisions. Nat. Neurosci. 6,
Divac, I., Rosvold, H.E., and Szwarcbart, M.K. (1967). Behavioral effects of
selective ablation of the caudate nucleus. J. Comp. Physiol. Psychol. 63,
Felleman, D.J., and Van Essen, D.C. (1991). Distributed hierarchical process-
ing in the primate cerebral cortex. Cereb. Cortex 1, 1–47.
Ferrier, D. (1873). Experimental researches in cerebral physiology and
pathology. J. Anat. Physiol. 8, 152–155.
E. (2010). Different corticostriatal integration in spiny projection neurons from
direct and indirect pathways. Front. Syst. Neurosci. 4, 15.
Forstmann, B.U., Dutilh, G., Brown, S., Neumann, J., von Cramon, D.Y.,
Ridderinkhof, K.R., and Wagenmakers, E.J. (2008). Striatum and pre-SMA
facilitate decision-making under time pressure. Proc. Natl. Acad. Sci. USA
Frank, M.J. (2006). Hold your horses: a dynamic computational role for the
subthalamic nucleus in decision making. Neural Netw. 19, 1120–1136.
Gerfen, C.R.,Engber, T.M.,Mahan, L.C., Susel,Z.,Chase, T.N., Monsma,F.J.,
Jr., and Sibley, D.R. (1990). D1 and D2 dopamine receptor-regulated gene
expression of striatonigral and striatopallidal neurons. Science 250, 1429–
Gertler, T.S., Chan, C.S., and Surmeier, D.J. (2008). Dichotomous anatomical
properties of adult striatal medium spiny neurons. J. Neurosci. 28, 10814–
Gold, J.I., and Shadlen, M.N. (2002). Banburismus and the brain: decoding the
relationship between sensory stimuli, decisions, and reward. Neuron 36,
Graybiel, A.M., and Ragsdale, C.W., Jr. (1979). Fiber connections of the basal
ganglia. Prog. Brain Res. 51, 237–283.
Gurney, K., Prescott, T.J., Wickens, J.R., and Redgrave, P. (2004).
Computational models of the basal ganglia: from robots to membranes.
Trends Neurosci. 27, 453–459.
Hanks, T.D., Ditterich, J., and Shadlen, M.N. (2006). Microstimulation of
macaque area LIP affects decision-making in a motion discrimination task.
Nat. Neurosci. 9, 682–689.
Hikosaka, O., and Wurtz, R.H. (1983). Visual and oculomotor functions of
monkey substantia nigra pars reticulata. IV. Relation of substantia nigra to
superior colliculus. J. Neurophysiol. 49, 1285–1301.
Hikosaka, O., and Wurtz, R.H. (1985). Modification of saccadic eye move-
ments by GABA-related substances. II. Effects of muscimol in monkey sub-
stantia nigra pars reticulata. J. Neurophysiol. 53, 292–308.
Caudate Stimulation Biases Perceptual Decisions
Neuron 75, 865–874, September 6, 2012 ª2012 Elsevier Inc. 873
Hikosaka, O., Sakamoto, M., and Usui, S. (1989). Functional properties of
monkey caudate neurons. I. Activities related to saccadic eye movements.
J. Neurophysiol. 61, 780–798.
Hikosaka, O., Sakamoto, M., and Miyashita, N. (1993). Effects of caudate
nucleus stimulation on substantia nigra cell activity in monkey. Exp. Brain
Res. 95, 457–472.
Hikosaka, O., Nakamura, K., and Nakahara, H. (2006). Basal ganglia orient
eyes to reward. J. Neurophysiol. 95, 567–584.
Hollerman, J.R., Tremblay, L., and Schultz, W. (2000). Involvement of basal
ganglia and orbitofrontal cortex in goal-directed behavior. Prog. Brain Res.
Horwitz, G.D., and Newsome, W.T. (1999). Separate signals for target selec-
tion and movement specification in the superior colliculus. Science 284,
Kable, J.W., and Glimcher, P.W. (2009). The neurobiology of decision:
consensus and controversy. Neuron 63, 733–745.
Kim, J.N., and Shadlen, M.N. (1999). Neural correlates of a decision in the
dorsolateral prefrontal cortex of the macaque. Nat. Neurosci. 2, 176–185.
Kitama, T., Ohno, T., Tanaka, M., Tsubokawa, H., and Yoshida, K. (1991).
Stimulation of the caudate nucleus induces contraversive saccadic eye move-
ments as well as head turning in the cat. Neurosci. Res. 12, 287–292.
function: a commentary. Percept. Psychophys. 63, 1421–1455.
Kravitz, A.V., Freeze, B.S., Parker, P.R., Kay, K., Thwin, M.T., Deisseroth, K.,
and Kreitzer, A.C. (2010). Regulation of parkinsonian motor behaviours by op-
togenetic control of basal ganglia circuitry. Nature 466, 622–626.
Kravitz, A.V., Tye, L.D., and Kreitzer, A.C. (2012). Distinct roles for direct and
indirect pathway striatal neurons in reinforcement. Nat. Neurosci. Published
online: April 29, 2012. http://dx.doi.org/10.1038/nn.3100.
Law, C.T., and Gold, J.I. (2008). Neural correlates of perceptual learning in
a sensory-motor, but not a sensory, cortical area. Nat. Neurosci. 11, 505–513.
Lo, C.C., and Wang, X.J. (2006). Cortico-basal ganglia circuit mechanism for
a decision threshold in reaction time tasks. Nat. Neurosci. 9, 956–963.
Maunsell, J.H., and van Essen, D.C. (1983). The connections of the middle
temporal visual area (MT) and their relationship to a cortical hierarchy in the
macaque monkey. J. Neurosci. 3, 2563–2586.
Middleton, F.A., and Strick, P.L. (2000). Basal ganglia output and cognition:
evidence from anatomical, behavioral, and clinical studies. Brain Cogn. 42,
Nakamura, K., and Hikosaka, O. (2006a). Facilitation of saccadic eye move-
ments by postsaccadic electrical stimulation in the primate caudate.
J. Neurosci. 26, 12885–12895.
Nakamura, K., and Hikosaka, O. (2006b). Role of dopamine in the primate
caudate nucleus in reward modulation of saccades. J. Neurosci. 26, 5360–
Newsome, W.T., Britten, K.H., and Movshon, J.A. (1989). Neuronal correlates
of a perceptual decision. Nature 341, 52–54.
Niijima, K., and Yoshida, M. (1982). Electrophysiological evidence for branch-
ing nigral projections to pontine reticular formation, superior colliculus and
thalamus. Brain Res. 239, 279–282.
on the speed and accuracy of a perceptual decision. J. Vis. 5, 376–404.
Parent, A., and Hazrati, L.N. (1995). Functional anatomy of the basal ganglia. I.
The cortico-basal ganglia-thalamo-cortical loop. Brain Res. Brain Res. Rev.
Petrov, A.A., Van Horn, N.M., and Ratcliff, R. (2011). Dissociable perceptual-
learning mechanisms revealed by diffusion-model analysis. Psychon. Bull.
Rev. 18, 490–497.
Rao, R.P. (2010). Decision making under uncertainty: a neural model
based on partially observable markov decision processes. Front. Comput.
Neurosci. 4, 146.
Ratcliff, R. (1978). Theory of memory retrieval. Psychol. Rev. 85, 59–108.
Roitman, J.D., and Shadlen, M.N. (2002). Response of neurons in the lateral
intraparietal area during a combined visual discrimination reaction time task.
J. Neurosci. 22, 9475–9489.
Saint-Cyr, J.A., Ungerleider, L.G., and Desimone, R. (1990). Organization of
visual cortical inputs to the striatum and subsequent outputs to the pallido-
nigral complex in the monkey. J. Comp. Neurol. 298, 129–156.
Salzman, C.D., Murasugi, C.M., Britten, K.H., and Newsome, W.T. (1992).
Microstimulation in visual area MT: effects on direction discrimination perfor-
mance. J. Neurosci. 12, 2331–2355.
Samejima, K., and Doya, K. (2007). Multiple representations of belief states
and action values in corticobasal ganglia loops. Ann. N Y Acad. Sci. 1104,
Selemon, L.D., and Goldman-Rakic, P.S. (1985). Longitudinal topography and
interdigitation of corticostriatal projections in the rhesus monkey. J. Neurosci.
Selemon, L.D., and Goldman-Rakic, P.S. (1988). Common cortical and
subcortical targets of the dorsolateral prefrontal and posterior parietal cortices
in the rhesus monkey: evidence for a distributed neural network subserving
spatially guided behavior. J. Neurosci. 8, 4049–4068.
Shadlen, M.N., and Newsome, W.T. (1996). Motion perception: seeing and
deciding. Proc. Natl. Acad. Sci. USA 93, 628–633.
Shen, W., Tian, X., Day, M., Ulrich, S., Tkatch, T., Nathanson, N.M., and
Surmeier, D.J. (2007). Cholinergic modulation of Kir2 channels selectively
elevates dendritic excitability in striatopallidal neurons. Nat. Neurosci. 10,
Smith, P.L., and Vickers, D. (1988). The accumulator model of 2-choice
discrimination. J. Math. Psychol. 32, 135–168.
Spatz, W.B., and Tigges, J. (1972). Experimental-anatomical studies on the
‘‘middle temporal visual area (MT)’’ in primates. I. Efferent cortico-cortical
connections in the marmoset Callithrix jacchus. J. Comp. Neurol. 146,
responses. J. Neurosci. Methods 65, 1–17.
Tomasi, S., Caminiti, R., and Innocenti, G.M. (2012). Areal differences in diam-
eter and length of corticofugal projections. Cereb. Cortex 22, 1463–1472.
van Veen, V., Krug, M.K., and Carter, C.S. (2008). The neural and computa-
tional basis of controlled speed-accuracy tradeoff during task performance.
J. Cogn. Neurosci. 20, 1952–1965.
Watanabe, M., and Munoz, D.P. (2010). Saccade suppression by electrical
microstimulation in monkey caudate nucleus. J. Neurosci. 30, 2700–2709.
Watanabe,M., and Munoz, D.P. (2011). Saccadereaction times are influenced
bycaudatemicrostimulation following andpriortovisualstimulusappearance.
J. Cogn. Neurosci. 23, 1794–1807.
Williams, Z.M., and Eskandar, E.N. (2006). Selective enhancement of associa-
tive learning by microstimulation of the anterior caudate. Nat. Neurosci. 9,
Wilson, S.A.K. (1914). An experimental research into the anatomy and physi-
ology of the corpus striatum. Brain 36, 427–492.
Yeterian, E.H., and Pandya, D.N. (1995). Corticostriatal connections of extras-
triate visual areas in rhesus monkeys. J. Comp. Neurol. 352, 436–457.
Caudate Stimulation Biases Perceptual Decisions
874 Neuron 75, 865–874, September 6, 2012 ª2012 Elsevier Inc.