Dissociation of human caudate nucleus activity in spatial and nonspatial working memory: an event-related fMRI study.
ABSTRACT We employed a novel event-related fMRI design and analysis technique to explore caudate nucleus contributions to spatial and nonspatial working memory. The spatial condition of a delayed-response task revealed greater mnemonic activation in four of six subjects when the delay period preceded immediately a probe stimulus requiring an overt motor response, as contrasted with a probe requiring no response. This effect was not seen in frontal or parietal cortical areas, and was seen in the caudate nucleus in a formally identical object condition in just one of six subjects. We hypothesized that this pattern of activity represented spatially dependent motor preparation. A second experiment confirmed this hypothesis: delay-period activity of the caudate nucleus showed greater time dependence in a task that featured spatial and motoric memory demands than in a comparable nonspatial task that featured the same response contingencies. These results suggest an important subcortical locus of the dissociation between spatial and nonspatial working memory, and a role for the human caudate nucleus in the integration of spatially coded mnemonic information with motor preparation to guide behavior.
- SourceAvailable from: Edward de Haan[Show abstract] [Hide abstract]
ABSTRACT: Many studies have identified the prefrontal cortex as the brain area that is critical for spatial memory, both in humans and in other primates. Other studies, however, have failed to establish this relation. Therefore, the aim of this paper was to review the literature regarding the role of the human prefrontal lobe in spatial memory. This was done by examining the evidence obtained from neuropsychological patients and from studies using brain-imaging techniques (PET and fMRI). Evidence supporting the notion that the prefrontal cortex is extensively involved in spatial working memory was found. The majority of these studies, however, suggests that frontal-lobe involvement is not related to the type of material that is being processed (e.g., spatial vs. nonspatial), but to process-specific functions, such as encoding and retrieval. Theoretically, these functions could be linked to the central executive within Baddeley's working-memory model, or to recent theories that emphasize the various processes that play a role in working memory. Also, methodological issues were discussed. Further research is needed to enhance our understanding of the precise interaction of domain-specific and general processes.Neuropsychology Review 01/2000; 10(2). · 6.42 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Cognitive deficits are recognized in Parkinson's disease. Understanding cognitive functions mediated by the striatum can clarify some of these impairments and inform treatment strategies. The dorsal striatum, a region impaired in Parkinson's disease, has been implicated in stimulus-response learning. However, most investigations combine acquisition of associations between stimuli, responses, or outcomes (i.e., learning) and expression of learning through response selection and decision enactment, confounding these separate processes. Using neuroimaging, we provide evidence that dorsal striatum does not mediate stimulus-response learning from feedback but rather underlies decision making once associations between stimuli and responses are learned. In the experiment, 11 males and 5 females (mean age 22) learned to associate abstract images to specific button-press responses through feedback in Session 1. In Session 2, they were asked to provide responses learned in Session 1. Feedback was omitted, precluding further feedback-based learning in this session. Using functional magnetic resonance imaging, dorsal striatum activation in healthy young participants was observed at the time of response selection and not during feedback, when greatest learning presumably occurs. Moreover, dorsal striatum activity increased across the duration of Session 1, peaking after most associations were well learned and was significant during Session 2 where no feedback was provided, and therefore no feedback-based learning occurred. Preferential ventral striatum activity occurred during feedback and was maximal early in Session 1. Taken together, the results suggest that the ventral striatum underlies learning associations between stimuli and responses via feedback whereas the dorsal striatum mediates enacting decisions.NeuroImage 07/2014; · 6.25 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Motivation is important for learning and cognition. While dopaminergic (D2) transmission in the ventral striatum (VS) is associated with motivation, learning and cognition are more strongly associated with function of the dorsal striatum, including activation in the caudate nucleus. A recent study found an interaction between intrinsic motivation and the DRD2/ANKK1 polymorphism (rs1800497), suggesting that A-carriers of rs1800497 are significantly more sensitive to motivation in order to improve during working memory (WM) training. Using data from the two large-scale imaging genetic datasets - IMAGEN (n=1080, age 13-15 years) and BrainChild (n~300, age 6-27) - we investigated whether rs1800497 is associated with WM. In the IMAGEN-dataset, we tested whether VS/caudate activation during reward anticipation was associated with WM performance and whether rs1800497 and VS/caudate activation interact to affect WM performance. We found that rs1800497 was associated with WM performance in IMAGEN and BrainChild. Higher VS and caudate activation during reward processing were significantly associated with higher WM performance (p<0.0001). An interaction was found between the DRD2/ANKK1 polymorphism rs1800497 and VS activation during reward anticipation on WM (p<0.01), such that carriers of the minor allele (A) showed a significant correlation between VS activation and WM, while the GG homozygotes did not, suggesting that the effect of VS BOLD on WM is modified by inter-individual genetic differences related to D2 dopaminergic transmission.Neuropsychopharmacology accepted article peview online, 09 April 2014; doi:10.1038/npp.2014.83.Neuropsychopharmacology: official publication of the American College of Neuropsychopharmacology 04/2014; · 8.68 Impact Factor
Cognitive Brain Research 8 1999 107–115
Dissociation of human caudate nucleus activity in spatial and nonspatial
working memory: an event-related fMRI study
Bradley R. Postle), Mark D’Esposito
Department of Neurology, UniÕersity of PennsylÕania Medical Center, 3 West Gates, Area 9 3400 Spruce St. Philadelphia, PA 19104 USA
Accepted 23 March 1999
We employed a novel event-related fMRI design and analysis technique to explore caudate nucleus contributions to spatial and
nonspatial working memory. The spatial condition of a delayed-response task revealed greater mnemonic activation in four of six subjects
when the delay period preceded immediately a probe stimulus requiring an overt motor response, as contrasted with a probe requiring no
response. This effect was not seen in frontal or parietal cortical areas, and was seen in the caudate nucleus in a formally identical object
condition in just one of six subjects. We hypothesized that this pattern of activity represented spatially dependent motor preparation. A
second experiment confirmed this hypothesis: delay-period activity of the caudate nucleus showed greater time dependence in a task that
featured spatial and motoric memory demands than in a comparable nonspatial task that featured the same response contingencies. These
results suggest an important subcortical locus of the dissociation between spatial and nonspatial working memory, and a role for the
human caudate nucleus in the integration of spatially coded mnemonic information with motor preparation to guide behavior. q1999
Elsevier Science B.V. All rights reserved.
Keywords: Basal ganglia; Motor preparation; Planning
A role for the caudate nucleus in spatial working mem-
ory has been established through studies of the behavioral
effects of lesions to 6,17,18,22,36 and electrical stimula-
tion of 10,28,39 this subcortical structure in monkeys.
Despite this evidence, the preponderance of neuroimaging
studies of human working memory to date have focused on
cortical correlates of working memory function. These
studies have established that spatial and nonspatial work-
ing memory are differentially reliant on dorsal stream and
ventral stream regions, respectively, in posterior cortex
7,12,31,38 , whereas the extent of spatialrnonspatial seg-
regation of working memory function in frontal cortex
remains a topic of debate 14,31,37,44 . Evidence of a
deficit in spatial working memory in patients with Parkin-
son’s disease PD19,41 , a disease characterized by
reduced supply of dopamine to the striatum, has led to
proposals that the caudate nucleus might be a site where
spatial and nonspatial mnemonic processing are carried out
differentially. Evidence consistent with this model has
emerged from demonstrations of a selective impairment of
spatial working memory in early PD, despite spared work-
ing memory for objects 29,32,33 , and of differential
patterns of activation of the caudate nucleus in monkeys
performing spatial and object working memory tasks 26 .
The studies reported here were intended to elucidate the
contribution of the caudate nucleus to spatial and nonspa-
tial working memory in humans.
2. Expt 1: ‘What’-then-‘Where’ delayed response
Most experiments designed to study visual working
memory do not permit resolution of the theoretically disso-
ciable contributions of sensory memory and motor prepara-
tion to delay period activity. Evidence of selectively com-
promised spatial working memory in PD 29,32,33 , for
example, may arise from degraded spatial sensory mem-
ory representations or from disordered motor planning and
execution when behavior must be organized with exclu-
0926-6410r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.
PII: S0926-6410 99 00010-5
B.R. Postle, M. D’EspositorCognitiÕe Brain Research 8 1999 107–115
sively spatial information. Our ‘what-then-where’ delayed-
response task based on 35
periods during each trial, enabling us to analyze delay-
period activity by stimulus material spatial, object and by
featured two discrete delay
position in the trial first, second . Importantly, only the
second delay period was followed by a motor response
Fig. 1a . This 2=2 design permitted us to unconfound
mnemonic and motor contributions to delay-period activity
B.R. Postle, M. D’EspositorCognitiÕe Brain Research 8 1999 107–115
in the caudate nucleus. We tested two hypotheses about
caudate nucleus delay-period activity in a delayed-re-
sponse task: i we would observe a main effect of stimu-
lus condition, such that spatial activity would be greater
than nonspatial activity consistent with the sensory mem-
. Ž .
ory hypothesis ; or ii we would observe an interaction of
stimulus condition with position of the delay, such that
activity would be greater during spatial memory periods
followed immediately by a motor response consistent with
the motor planning hypothesis . Because segregation of
striatal efferents focuses the majority of caudate nucleus
output via subsequent stations of the basal ganglia and the
thalamus on prefrontal and limbic cortical regions associ-
ated with higher cognitive function, and the majority of
putaminal output on motor areas of frontal cortex 4 , we
limited our analyses to the caudate nucleus.
We studied 6 right handed subjects 5 males; mean
ages23 years . All subjects were recruited from the
undergraduate and medical campuses of the University of
Pennsylvania, and all gave informed consent.
2.2.2. BehaÕioral procedure
Each trial began with an instructional cue 500 ms ,
followed by an initial target stimulus presentation 1 s
followed by a delay period ‘delay 1’; 6.5 s followed by
the presentation of the initial target stimulus ‘match’ and
a foil stimulus ‘intermediate stimulus presentation’; 1.5
s , followed by a second delay period ‘delay 2’; 6.5 s ,
followed by a probe stimulus ‘final probe’; 1 s; Fig. 1a .
A fixation cross appeared with the onset of the initial
target, and remained on the screen until the offset of the
final probe. An intertrial interval ITI of 15 s separated
each trial; the time from trial onset to trial onset was 32 s.
The instructional cue read ‘shape first’ or ‘location first’ in
a pseudorandomly determined order. In ‘shape first,’ or
what-then-where, trials, subjects encoded the featural de-
tails of the initial target, ignoring its location on the screen,
and retained this featural information during delay 1. The
two intermediate stimuli both appeared in a location differ-
ent from that occupied by the initial target, and their onset
prompted a discrimination as to which of the two was an
identical featural match with the initial target. Immediately
upon making this discrimination, subjects encoded the
location of the match stimulus and retained this location
information during delay 2. In this way, the match probe
for the ‘what’ portion of the trial became the target for the
‘where’ portion of the trial. Finally, subjects indicated
whether or not the final probe occupied the same location
as the location target stimulus i.e., as the match stimulus
from the intermediate stimulus presentation , and indicated
their decision with a ‘yes’ or ‘no’ button press. In ‘loca-
tion first,’ or where-then-what, trials, subjects were trained
to perform spatial delayed response during the first half of
the trial, and to encode featural information about the
location match stimulus from the intermediate stimulus
presentation in order to perform object delayed response
during the second half of the trial. Each block of trials,
corresponding to one fMRI scan, contained six what-then-
where and six where-then-what trials presented in a pseu-
dorandomized order, each featuring an equal number of
‘yes’ and ‘no’ final probes. Five of the six experiments
consisted of eight scanned blocks of testing, or 96 trials
Fig. 1. a Schematic diagram and timeline of a what-then-where trial of the delayed-response task featured in Experiment 1. Each box represents a
stimulus display event, and the dotted lines connecting each box to the timeline represent the sequence and duration of each of these events. The numbers
along the timeline represent seconds, and the two arrows indicate the positioning of the two delay-sensitive independent variables in our statistical model.
A where-then-what trial would be formally identical, but with the instructions reading ‘location first,’ one of the two intermediate stimuli occupying the
same location on the screen as the initial target, and neither of the two intermediate stimuli matching the featural identity of the initial target. See text for a
description of the task. b Results of single subject analyses in the four subjects exhibiting significantly greater caudate nucleus activity during spatial
delay 2 than during spatial delay 1. Analogous behavior was not observed in the object condition for subjects AK, MA, and EW. Subject EP also exhibited
greater delay 2 than delay 1 activity in the object condition; note that these suprathreshold object delay voxels were located on a more inferior slice than
the voxels exhibiting suprathreshold spatial delay activity for this subject. c Superimposed trial averaged BOLD fMRI data from what-then-where and
where-then-what trials for the suprathreshold voxels identified in the spatial delay 2yspatial delay 1 contrast for subject EP. The gray bars represent the
two delay periods of each trial; the time on the horizontal axis corresponds to the diagram of trial events presented in panel a. Changes in BOLD data lag
behind putatively causal behavioral events by 4 to 6 s. Note that spatial delay-period activity is higher in what-then-where than in where-then-what trials,
whereas object delay activity is at a comparable level during the two conditions.
Fig. 2. a Schematic diagram and timeline for the motor-set task featured in Experiment 2; the conditional visuo-motor task is illustrated above the
delayed-matching task. All graphical conventions are the same as Fig. 1. Note that the two delay-specific independent variables are positioned within the
same delay period. b Results of single subject analyses in the four subjects exhibiting significantly greater caudate nucleus activity during conditional
visuo-motor delay 1 than during conditional visuo-motor delay 2. Analogous behavior was not observed in the delayed-matching task for subjects JHo and
JL. Subjects JHa and WK also exhibited greater delay 1 than delay 2 activity in the delayed-matching task; note that suprathreshold conditional
visuo-motor and delayed-matching voxels are overlapping for these two subjects. c Superimposed trial averaged BOLD fMRI data from conditional
visuo-motor and delayed-matching trials from the suprathreshold voxels identified in the conditional visuo-motor delay 1yconditional visuo-motor delay
x 2 contrast for subject JHo. The gray bar represents the delay period of each trial; the time on the horizontal axis corresponds to the diagram of trial events
presented in panel a. Note that conditional visuo-motor delay-period activity is higher in the first portion of the delay period, whereas delayed-matching
delay-period activity is relatively stable throughout the delay period.
B.R. Postle, M. D’EspositorCognitiÕe Brain Research 8 1999 107–115
Žand, therefore, 96 spatial delay periods and 96 object
delay periods ; one subject was scanned for five blocks.
2.2.3. fMRI procedure
Imaging was carried out on a 1.5 T SIGNA scanner
GE Medical Systemsequipped with a prototype fast
gradient system for echo-planar imaging. A standard ra-
diofrequency RF head coil was used with foam padding
to restrict comfortably head motion. High resolution axial
T1-weighted images 21 axial slices
every subject. A gradient echo, echoplanar sequence TR
s2000 ms, TEs50 ms was used to acquire data sensi-
tive to the blood oxygen level dependent BOLD signal.
Scans of the behavioral task were preceded by a scan in
which we derived the impulse response function IRF for
each subject. The IRF, which characterizes the fMRI re-
sponse resulting from a brief pulse of neural activity 9 ,
was used to smooth independent variables in the general
linear model GLM that we used to analyze the results of
the scans of our behavioral task see fMRI data process-
ing, below . The derive-IRF scan lasted 5 min 20 s 160
imagesrslice . Our method for deriving the IRF is de-
scribed in detail elsewhere 16 . Each fMRI scan of our
behavioral task lasted 6 min 24 s
FMRI data collection during all scans was preceded by 20
s of dummy gradient and RF pulses to achieve a steady
state of tissue magnetization.
were obtained in
180 imagesrslice .
2.2.4. fMRI data processing
The principle of the fMRI time series analysis was to
model the fMRI signal changes occurring during particular
periods of the behavioral trials with covariates comprised
of shifted, BOLD IRFs 49 . The fMRI time series were
tested with covariates that modeled the expected BOLD
signal response in the event of an increase in neural
activity relative to the ITI occurring during each behav-
iorally significant portion of each trial i.e., stimulus pre-
sentation, delay, and response periods . Importantly, we
used this method to obtain measures of delay-period activ-
ity in voxels in the caudate nucleus that were not contami-
nated by variance in the fMRI time series attributable to
stimulus presentation or response activity. We could ac-
complish this because smoothness of the fMRI response to
neural activity allows fMRI responses to be resolved on
the order of 4 s 49 . False positive rates were controlled at
as0.05 by Bonferroni correction for the number of vox-
els per region of interest ROI
All our analyses were performed with ROIs. Caudate
nucleus ROIs were drawn for each subject on that subject’s
T1 anatomical images, and incorporated the head of the
caudate nucleus, beginning rostrally and ventrally at ap-
proximately the level of the anterior commissure, and the
body of the caudate nucleus, extending caudally along the
lateral wall of the lateral ventricle and ending at the
ventral-most level at which the body of the lateral ventricle
appeared intact in one slice i.e., one slice dorsal to the
slice in which the atrium became visible . We also per-
formed analyses on three cortical areas that are linked
anatomically and functionally with the basal ganglia: areas
9 and 46 of dorsolateral prefrontal cortex DLPFC , pri-
mary motor cortex M1 , and area 7 of posterior parietal
cortex PPC . DLPFC and PPC both exhibit delay-related
activity during working memory tasks, both are important
sources of afferent input to the caudate nucleus, and
DLPFC is an important recipient via subsequent stations
of the basal ganglia and the thalamus of caudate nucleus
output. We created ROIs for these two cortical regions by
drawing them onto the ‘canonical’ representation of a
brain in Talairach space that is provided in SPM96b, using
the atlas of Talairach and Tournoux 40 to confirm our
identification of anatomical landmarks. Next, we trans-
formed these ROIs from Talairach space into the native
space in which each subject’s data had been acquired by
applying the 12 parameter affine transformation
with nonlinear deformations
effectively, a ‘reverse normalization’ . Movement-related
activity in M1 is closely correlated temporally with move-
ment-related activity in the striatum 2,3,13 . M1 thereby
represents an effective control region for dissociating mo-
toric from mnemonic activity in the caudate nucleus. We
defined the M1 ROI in each subject as the cortex immedi-
ately anterior to the central sulcus.
5 , routine in SPM96b
2.2.5. Single subject analyses
Our analyses were performed in two steps: single sub-
ject analyses and group analyses. Single subject analyses
permitted us to maintain the high spatial resolution af-
forded by fMRI, and to detect intersubject variability. Such
information is lost in analysis approaches that combine
data from all subjects at an early stage of analysis, and are
thus restricted to testing for activation patterns that are
consistent enough across subjects in a standard space to be
detected after group-averaging. Our single subject analy-
ses, in contrast, treat each subject as a case study, and
permit us to assess replication of as well as variation in
effects across individual cases. In essence, data from 6
subjects performing the same task represent a single result
with 5 replications. Single subject analyses with methods
comparable to those used in the present study and, impor-
tantly, with a large number of observations per subject, as
in the present study have been demonstrated to feature
ample sensitivity to detect signal intensity changes of
interest 16,50 . Contrasts performed with single subject
data in the present study had in excess of 1200 effective
degrees of freedom.
We effected tests of our first hypothesis, a main effect
of stimulus material, by generating a two-tailed t-map of
the contrastspatial delay 1qspatial delay 2 y object
delay 1qobject delay 2and detecting suprathreshold
voxels. We tested our second hypothesis, interactions of
stimulus material and position in delay-period activity, by
generating two-tailed t-maps of the contrast spatial delay
B.R. Postle, M. D’EspositorCognitiÕe Brain Research 8 1999 107–115
1yspatial delay 2
1yobject delay 2 . The identification of suprathreshold
t-values in this analysis would indicate that delay-period
activity within a particular working memory condition
spatial or object was sensitive to position within the trial.
In the event of significant interactions, tests for main
effects of delay position spatial delay 1qobject delay
1 y spatial delay 2qobject delay 2
formed to discount simple order effects. Analysis of activ-
ity during the response phase of the task
probeyobject final probe , which would reflect motor
activity, served as a control for the memory-related analy-
ses described above.
and of the contrastobject delay
would be per-
2.2.6. Group analyses
Our group analyses were performed as random effects
models, an approach that permits generalization of results
obtained from a sample to the population represented by
that sample. This inferential step cannot be made with the
fixed effects group analyses that have been employed by
the majority of fMRI experimentalists to date 24,47 .
Importantly, random effects analyses are invulnerable to
spurious results that can arise if a disproportionately large
effect size in a single subject ‘drives’ the mean effect size
for the group, as can happen with fixed effects analyses.
All our group analyses used t-values as dependent mea-
sures, because they represent an index of the signal-to-noise
ratio for a given contrast. T-values are proportional to the
magnitude of the hypothesized effect, and they are normal-
ized measures because they are scaled by the error in each
subject. Group tests for a main effect of stimulus material
were performed by first identifying for each subject the
voxels within the caudate nucleus ROI showing a main
effect of delay period activity spatial delay 1qspatial
delay 2qobject delay 1qobject delay 2 and, from these
voxels, extracting a spatially averaged time course and
calculating the orthogonal contrast of
spatial delay 2 y object delay 1qobject delay 2 . The
resultant t-value represented, for each subject, the extent to
which the sensitivity of delay-period activity was greater
for spatial or for object stimuli. A positive t-value would
indicate that spatial delay-period activity was greater than
object delay-period activity, a negative t-value the con-
verse. A paired t-test on these t-values, one from each
subject, assessed the significance of any trends in the data
To conduct group analyses of our second hypothesized
effect, a greater influence of trial position on delay-period
activity with one stimulus type than with another, we
generated an index of the sensitivity of caudate nucleus
activity to trial position by identifying critical voxels
showing a main effect of delay period activity within the
caudate nucleus ROI, extracting a spatially averaged time
course from these critical voxels, and calculating the or-
thogonal contrast ofspatial delay 1yspatial delay 2 y
object delay 1yobject delay 2 . The two-tailed t-value
spatial delay 1q
arising from this contrast represented, for each subject, a
normalized measure of the interaction of stimulus material
and delay period position. A paired t-test on these t-val-
ues, one from each subject, assessed the significance at the
group level of this interaction. Comparable analyses were
performed within the DLPFC, M1, and PPC ROIs.
2.3.1. Single subject analyses
The test of our first hypothesis yielded a null result:
direct contrast of spatial vs. object delay-period activity
collapsed across trial position revealed no suprathreshold
voxels in the caudate nucleus in any subject. We did,
however, observe an interaction of stimulus material with
delay position in several subjects: Trial-position differ-
encesdelay 2)delay 1 achieved significance in the
caudate nucleus for four subjects in the spatial condition as
contrasted with only one subject in the nonspatial condi-
Ž . Ž
tion Fig. 1b . The contrast of delay 1 vs. delay 2 col-
lapsed across stimulus material
differences in any subject.
suprathreshold voxels were located in the right head of the
caudate nucleus in two subjects, in the left head of the
caudate nucleus in one subject, and bilaterally in the head
of the caudate nucleus in one subject subject EP ; object
delay 2)delay 1 voxels were also found in the head of
the caudate nucleus, bilaterally, in subject EP, although in
a different, nonoverlapping set of voxels. Probe differences
spatial final probe)object final probe were significant
in only one subject. No significant trial-position effects
were observed in DLPFC, M1, or PPC.
yielded no significant
Spatial delay 2)delay 1
2.3.2. Group analyses
All group analyses failed to achieve significance in each
of the four ROIs caudate nucleus, DLPFC, M1, PPC .
Experiment 1 revealed the absence of main effects of
stimulus material or of delay position, but an effect of
delay position on spatial delay-period activity in four of
six subjects, indicating that spatial delay-period activity in
the caudate nucleus was more sensitive to trial position
than was object delay-period activity in the majority of our
subjects. Thus, the caudate nucleus may make functionally
different contributions to working memory for ‘where’
than to working memory for ‘what.’ Voxels with delay-
period activity that exhibited significant sensitivity to trial
position were all located in the head of the caudate nu-
cleus, with no clear trend in lateralization. These results
suggest that the caudate nucleus, unlike its important
sources of cortical afferents or the cortical targets of its
efferents, may be preferentially activated in the delay
period of working memory tasks by the coincidence of
spatial memory and the need to formulate a response based
B.R. Postle, M. D’EspositorCognitiÕe Brain Research 8 1999 107–115
on that information. This pattern of activation could be
characterized as spatially mediated motor preparation. It is
dissociable from final probe-related activity in the caudate
nucleus, which was greater in the spatial condition in only
one subject. Although this effect was significant in four of
six subjects, the nonsignificant result of our group analysis
prompted us to test the hypothesis that arose from Experi-
ment 1 with a second experiment.
3. Expt 2: ‘Motor set’
An important feature of the hypothesis generated in
Experiment 1 is that caudate nucleus delay period activity
is more sensitive to motor contingencies when memoranda
are coded spatially than when they are coded nonspatially.
We tested this hypothesis with a ‘motor set’ experiment
Fig. 2a , based on Ref. 21 , that featured a spatial condi-
tional visuo-motor task and a nonspatial delayed-matching
task. In this experiment the delay period was lengthened to
12 s to permit assessment of caudate nucleus activity
during two discrete portions of the same delay period see
Section 3.2.2 . Our design permitted us to test hypotheses
analogous to those tested in Experiment 1:1 a main effect
of task type; and 2 an interaction of task type and delay
position. A finding of a greater interaction in the condi-
tional visuo-motor task would be consistent with a prefer-
ential role for the caudate nucleus during working memory
task performance in spatially mediated motor preparation.
The six subjects participating in our study 3 males,
mean ages22.7 were recruited from the undergraduate
and medical campuses of the University of Pennsylvania,
and all gave informed consent.
3.2.2. BehaÕioral procedure
The conditional visuo-motor task and the delayed-
matching task both featured color-behavior associations
that subjects learned prior to scanning. In the conditional
visuo-motor task, the initial presentation of either of two
colored stimuli ‘cue’; 600 ms in the top position of a
three-circle array indicated whether the position to choose
at the end of the trial was on the left blue cue or the right
yellow cue . After the 11.4 s delay the three circle array
was re-presented, but with all three circles colored white
‘probe’; 1 s . Subjects chose the circle on the left or the
right with a button press Fig. 2a . Cue presentation in this
task informed subjects explicitly about the correct response
for that trial, and thus subjects could guide performance by
retaining a representation of either the spatial location
associated with the cue or the motor response associated
with the cue during the delay period a prospective mem-
ory code 34,46 .
The initial cue in the delayed-matching task was red or
green. Subjects remembered this color until the probe
presented the two colors, in pseudorandomly determined
order, in the two bottom positions, whereupon they chose
the color that matched the cue. Timing, sequence, and
layout of this task were identical to the conditional visuo-
motor task Fig. 2a . In contrast to the conditional visuo-
motor task, however, the delayed-matching task required
memory for the color of the initial cue, and task-relevant
spatial and motor information was unavailable until a
response was prompted by the re-presentation of the cued
color at the end of the trial Fig. 2a . Previous behavioral
testing with this task has indicated that subjects respond
more quickly on conditional visuo-motor than on delayed-
matching trials, consistent with the assumption that condi-
tional visuo-motor trials permit motor preparation during
the delay period 15 .
3.2.3. fMRI scanning and data processing procedures
These were the same as those used in Experiment 1,
with the exception that the two delay-period independent
variables of interest were not separated by intervening task
events. The ‘delay 1’ independent variable was positioned
4 s into the trial in order to model variance in the fMRI
time series that was attributable to the first portion of the
delay period, and the ‘delay 2’ independent variable was
positioned 8 s into the trial in order to model activity that
was attributable to the second portion of the delay period.
The 4 s spacing between delay 1 and delay 2 covariates
ensured that they would not model a significant amount a
shared variance 49 , and, therefore, that the loading on
each would index delay-period activity uncontaminated by
delay-period activity occurring later or earlier in the trial,
3.2.4. Single subject analyses
We tested our first hypothesis, a main effect of task, by
generating a two-tailed t-map of the contrast
visuo-motor delay 1qConditional visuo-motor delay 2 y
Delayed-matching delay 1qDelayed-matching delay 2
and detecting suprathreshold voxels. We tested our second
hypothesis, interactions of task type and position in delay-
period activity, by generating two-tailed t-maps of the
contrast Conditional visuo-motor delay 1yConditional
visuo-motor delay 2 and of the contrast Delayed-match-
ing delay 1yDelayed-matching delay 2 . The identifica-
tion of suprathreshold t-values in this analysis would
indicate that delay-period activity within a particular task
varied systematically over time. In the event of significant
interactions, tests for main effects of delay position
ditional visuo-motor delay 1qDelayed-matching delay 1
y Conditional visuo-motor delay 2qDelayed-matching
delay 2would be performed to discount simple position
B.R. Postle, M. D’EspositorCognitiÕe Brain Research 8 1999 107–115
effects. We also tested for differential motor activity with a
contrast of probe-related activity in the two tasks.
3.2.5. Group analyses
Random effects group analyses were performed on our
first hypothesis by generating a one-tailed t-map of the
main effect of delay-period activity for each subject Con-
ditional visuo-motor delay 1qConditional visuo-motor
delayqDelayed-matching delay 1qDelayed-matching
delay 2 , identifying suprathreshold voxels and extracting
from them a spatially averaged time series, applying to this
time series the orthogonal contrast of
motor delay 1qConditional visuo-motor delay 2 y
Delayed-matching delay 1qDelayed-matching delay 2 ,
and performing a paired t-test on the resultant t-values
contributed by each subject. We tested our second hypoth-
esis by extracting a spatially averaged time course from
the voxels in the caudate nucleus ROI showing a main
effect of delay-period activity, calculating the orthogonal
Conditional visuo-motor delay 2 y Delayed-matching de-
lay 1yDelayed-matching delay 2 , and performing a
paired t-test on the resultant t-values contributed by each
subject. As in Experiment 1, this analysis would be re-
peated in M1 to confirm the mnemonic nature of this
3.3.1. Single subject analyses
Analysis of delay-period activity in caudate nucleus
ROIs revealed no overall effects of task
visuo-motor, delayed matching or of position in the delay
delay 1, delay 2 —a failure to reject the null hypotheses
that there was no main effects of task. There was, how-
ever, a significant position effect delay 1)delay 2 in the
conditional visuo-motor task in four of the six subjects, as
contrasted with a significant position effect in two of the
six subjects in the delayed-matching task
Suprathreshold voxels identified in the conditional visuo-
motor task were located in the right head of the caudate
nucleus in two subjects JHa and JL , in the left head and
left body of the caudate nucleus in one subject WK , and
bilaterally in the head of the caudate nucleus in one subject
JHo . There were no differences in probe-related activity
in any subjects. Suprathreshold voxels identified in the
delayed-matching task were overlapping with, or adjacent
to, the voxels identified in the conditional visuo-motor
task: right head and left body in one subject JHa , and left
head and left body in one subject WK .
Fig. 2b .
3.3.2. Group analyses
These revealed no significant effects of task or of
position in the delay. The analysis of the task by delay-
position interaction in caudate nucleus, however, revealed
greater sensitivity to delay position in the conditional
visuo-motor task than in the delayed-matching task in each
of the five subjects for whom we performed this analysis
Ž Ž .. Ž
t 4 s6.9; p-.005 . Data from one subject were ex-
cluded from the analyses because a significant difference
between tasks in fMRI signal associated with cue presenta-
tion rendered analysis of delay-period activity equivocal
49 . This effect was not observed in frontal areas 9 and
46 or M1, or in posterior parietal area 7. Finally, there was
no significant difference, at the group level, in probe-re-
Delay-period activity of the caudate nucleus in the
Experiment 2 showed greater time dependence in a task
that featured spatial and motoric memory demands than in
a comparable task requiring nonspatial memory. The motor
effector function of the caudate nucleus, in contrast, did
not differ in the two tasks. These results are consistent
with the model, generated by the results of Experiment 1,
of caudate nucleus working memory activity reflecting
spatially mediated motor preparation. The spatial informa-
tion needed for preparation of a motor plan was available
to subjects with the presentation of the cue only on condi-
tional visuo-motor trials. The reliable pattern of delay 1
activity exceeding delay 2 activity
group analysis as well as in the majority of the single
subject analyses suggests that this motor preparation takes
place in the initial portion of the delay period.
The fact that caudate nucleus activity was greater dur-
ing the first portion of the delay period of motor-set
conditional visuo-motor trials suggests that the caudate
nucleus modulation that we observed in the spatial condi-
tion of the what-then-where task in Experiment 1 did not
reflect a ‘ramping up’ of caudate nucleus activity as the
response drew nearer. Indeed, the delay-period behavior of
caudate nucleus voxels in the conditional visuo-motor task
of the motor-set experiment stands in contrast with such
‘ramping up’ behavior that Fuster observed in many pre-
frontal cortical neurons in monkeys performing this task,
behavior that he characterized as ‘motor set’ activity 21 .
Such motor-set activity has been observed in individual
PFC voxels of the subjects who participated in this experi-
ment, and is described elsewhere 15 .
significant in the
4. General discussion
The present study confirms that the caudate nucleus is
an important neural locus of the dissociation between
spatial and nonspatial visual working memory, as has been
suggested by previous studies of PD patients 29,32,33 .
Although the ‘what’r‘where’ dissociation in human work-
23,32,38,42 and electrophysiological 27 studies, and
candidate cortical substrates have been identified in neu-
roimaging studies 12,14,31,38 , the present report offers
B.R. Postle, M. D’EspositorCognitiÕe Brain Research 8 1999 107–115
the first evidence in humans that this dissociation may
reflect more than simply an extension of the parallel,
segregated spatial and nonspatial visual information pro-
cessing pathways 43,45 . Our results suggest that spatial
delay-period activity task performance may feature greater
interaction with the motor system than does nonspatial
delay-period activity. Spatial working memory must play
an important role in planning and executing motor action
in the service of behavioral objectives, and the data pre-
sented here suggest that the caudate nucleus is an impor-
tant mediator of this function. One implication of the
results of these two experiments is that the deficits ob-
served in PD patients on tests of spatial memory
19,29,32,33,41 may reflect disrupted formulation of a
motor response based on a spatially coded mnemonic
representation, rather than a purely sensory deficit, as may
have been predicted from previous models 8,25 . The
present results are also not easily reconciled with a pro-
posed role for the neostriatum in the governance of amodal
strategic and general attentional processes 30 , because
they did not demonstrate reliable evidence for differential
position- or time-dependent activity during delay periods
of nonspatial working memory tasks.
Our results are consistent with the view that the caudate
nucleus, with its central position in the motor system, is
preferentially involved in spatial mnemonic processing
because of the integral role of spatial information in the
formulation motor behaviors 1,11 . An alternative model
positing a role for the caudate nucleus in nonspatial work-
ing memory task performance, however, derives support
from an autoradiographic study in monkeys that revealed
topographical differences in caudate nucleus activity asso-
ciated with performance of spatial and object delayed
alternation tasks, with the former associated with activa-
tion of more anterior regions, and the latter with more
posterior regions 26 . Our single subject analyses did not
find reliable evidence for nonspatial delay-period activity
in the caudate nucleus that was dependent on impending
motor contingencies, nor did they reveal evidence of topo-
graphically dissociable patterns of spatial and nonspatial
mnemonic activity. Our group analysis technique for con-
trasting spatial and object delay-related activity, however,
did not permit an investigation of the possibility of topo-
graphically differential representation of spatial and non-
spatial working memory. Rather, by collapsing across all
voxels demonstrating delay-period activity, it represented a
‘winner-take-all’ approach. We opted for this approach to
maximize our sensitivity to experimentally induced changes
in activation. Methods affording greater sensitivity and
higher spatial resolution than those employed in this study
e.g., fMRI at higher field strength will be necessary to
determine conclusively whether the human caudate nu-
cleus participates preferentially in spatial working memory
function, or whether topographical differences may also
characterize differential representation of spatial and non-
spatial working memory in the human caudate nucleus.
Supported by the Charles A. Dana Foundation, Ameri-
can Federation for Aging Research, and NIH grants
NS01762 and AG13483. We thank Dustin Ballard, Jessica
Lease, Rajiv Singh, and Elizabeth Wheeler for assistance
with programming and data collection, and Geoffrey
Aguirre and Eric Zarahn for helpful discussions of this
w x 1 L. Abraham, M. Potegal, S. Miller, Evidence for caudate nucleus
involvement in an egocentric spatial task: return from passive trans-
port, Physiol. Psychol. 11 1983 11–17.
w x 2 G.E. Alexander, M.D. Crutcher, Neural representations of the target
goal of visually guided arm movements in three motor areas of the
monkey, J. Neurophysiol. 64 1990 164–178.
w x 3 G.E. Alexander, M.D. Crutcher, Preparation for movement: neural
representations of intended direction in three motor areas of the
monkey, J. Neurophysiol. 64 1990 133–150.
w x 4 G.E. Alexander, M.R. DeLong, P.L. Strick, Parallel organization of
functionally segregated circuits linking basal ganglia and cortex,
Annu. Rev. Neurosci. 9 1986 357–381.
w x 5 J. Ashburner, K. Friston, Fully three-dimensional nonlinear spatial
normalization: a new approach, NeuroImage 3 1996 S111.
w x 6 K. Battig, H.E. Rosvold, M. Mishkin, Comparison of the effects of
frontal and caudate lesions on delayed response and alternation in
monkeys, J. Comp. Physiol. Psychol. 53 1960 400–404.
w x 7 A. Belger, A. Puce, J.H. Krystal, J.C. Gore, P. Goldman-Rakic, G.
McCarthy, Dissociation of mnemonic and perceptual processes dur-
ing spatial and nonspatial working memory using fMRI, Human
Brain Mapping 6 1998 14–32.
w x 8 F. Boller, D. Passafiume, N.C. Keefe, K. Rogers, L. Morrow, Y.
Kim, Visuospatial impairment in Parkinson’s disease: role of percep-
tual and motor factors, Arch. Neurol. 41 1984 485–490.
w x 9 G.M. Boynton, S.A. Engel, G.H. Glover, D.J. Heeger, Linear sys-
tems analysis of functional magnetic resonance imaging in human
V1, J. Neurosci. 16 1996 4207–4221.
10 S.M. Cohen, Electrical stimulation of cortical-caudate pairs during
delayed successive visual discrimination in monkeys, Acta Neurobi-
ologiae Experimentalis Warsaw 32 1971 211–233.
11 D. Cook, R.P. Kesner, Caudate nucleus and memory for egocentric
localization, Behav. Neural Biol. 49 1988 332–343.
12 S.M. Courtney, L.G. Ungerleider, K. Keil, J. Haxby, Object and
spatial visual working memory activate separate neural systems in
human cortex, Cerebral Cortex 6 1996 39–49.
13 M.D. Crutcher, G.E. Alexander, Movement-related neuronal activity
selectively coding either direction or muscle pattern in three motor
areas of the monkey, J. Neurophysio. 64 1990 151–163.
14 M. D’Esposito, G.K. Aguirre, E. Zarahn, D. Ballard, Functional
MRI studies of spatial and non-spatial working memory, Cogn.
Brain Res. 7 1998 1–13.
15 M. D’Esposito, D. Ballard, E. Zarahn, G.K. Aguirre, The role of
prefrontal cortex in sensory memory and motor preparation: an
event-related fMRI study. submitted for publication .
16 M. D’Esposito, B.R. Postle, D. Ballard, J. Lease, Maintenance
versus manipulation of information held in working memory: an
event-related fMRI study. Brain and Cognition in press .
17 W.H. Dean, G.D. Davis, Behavior changes following caudate lesions
in rhesus monkey, J. Neurophysiol. 22 1959 525–537.
18 I. Divac, H.E. Rosvold, M.K. Scharcbart, Behavioral effects of
B.R. Postle, M. D’EspositorCognitiÕe Brain Research 8 1999 107–115
selective ablation of the caudate nucleus, J. Comp. Physiol. Psychol.
63 1967 184–190.
19 M. Freedman, M. Oscar-Berman, Selective delayed response deficits
in Parkinson’s and Alzheimer’s disease, Arch. Neurol. 43 1986
20 K.J. Friston, J. Ashburner, C.D. Frith, J.-B. Poline, J.D. Heather,
R.S.J. Frackowiak, Spatial registration and normalization of images,
Human Brain Mapping 2 1995 165–189.
21 J.M. Fuster, Memory in the Cerebral Cortex, MIT Press, Cambridge,
22 P.S. Goldman, H.E. Rosvold, The effects of selective caudate lesions
in infant and juvenile rhesus monkeys, Brain Res. 43 1972 53.
23 R. Hecker, B. Mapperson, Dissociation of visual and spatial process-
ing in working memory, Neuropsychologia 35 1997 599–603.
24 A.P. Holmes, K.J. Friston, Generalisability, random effects and
population inference, NeuroImage 7 1998 S754.
25 A. Hovestadt, G.J. DeJong, J.D. Meerwaldt, Spatial disorientation as
an early symptom of Parkinson’s disease, Neurology 17
26 R. Levy, H.R. Friedman, L. Davachi, P.S. Goldman-Rakic, Differen-
tial activation of the caudate nucleus in primates performing spatial
and nonspatial working memory tasks, J. Neurosci. 17 1997 3870–
27 A. Mecklinger, N. Muller, Dissociations in the processing of ‘what’
and ‘where’ information in working memory: an event-related po-
tential analysis, J. Cogn. Neurosci. 8 1996 453–473.
28 E.F. Mordinov, Effect of successive electrical stimulation of differ-
ent parts of the brain on delayed spatial choice in monkeys, Neu-
rosci. Behav. Physiol. 11 1981 164–169.
29 A.M. Owen, J.L. Iddon, J.R. Hodges, B.A. Summers, T.W. Robbins,
Spatial and non-spatial working memory at different stages of
Parkinson’s disease, Neuropsychologia 35 1997 519–532.
30 B. Pillon, B. DeWeer, M. Vidailhet, A.-M. Bonnet, V. Hahn-Barma,
B. DuBois, Is impaired memory for spatial location in Parkinson’s
disease domain specific or dependent on ’strategic’ processes?,
Neuropsychologia 36 1998 1–9.
31 B.R. Postle, M. D’Esposito, ‘What’-then-‘Where’ in Visual Work-
ing Memory: an Event-Related fMRI Study. J. Cogn. Neurosci. in
32 B.R. Postle, J. Jonides, E. Smith, S. Corkin, J.H. Growdon, Spatial,
but not object, delayed response is impaired in early Parkinson’s
disease, Neuropsychology 11 1997 1–9.
33 B.R. Postle, J.J. Locascio, S. Corkin, J.H. Growdon, The time
course of spatial and object visual learning in early Parkinson’s
disease, Neuropsychologia 35 1997 1413–1422.
34 G. Rainer, S.C. Rao, E.K. Miller, Prospective coding for objects in
the primate prefrontal cortex, J. Neurosci., in press.
35 S.C. Rao, G. Rainer, E.K. Miller, Integration of what and where in
the primate prefrontal cortex, Science 276 1997 821–824.
36 H.E. Rosvold, J.M.R. Delgado, The effect on delayed-attention test
performance of stimulating or destroying electrically structures within
the frontal lobes of the monkey’s brain, J. Comp. Physiol. Psychol.
49 1956 365–372.
37 Rushworth, A.M. Owen, The functional organization of the lateral
frontal cortex: conjecture or conjuncture in the electrophysiology
literature?, Trends Cogn. Sci. 2 1998 46–53.
38 E.E. Smith, J. Jonides, R.A. Koeppe, E. Awh, E.H. Schumacher, S.
Minoshima, Spatial vs. object working memory: PET investigations,
J. Cogn. Neurosci. 7 1995 337–356.
39 J.S. Stamm, Electrical stimulation of monkeys’ prefrontal cortex
during delayed-response performance, J. Comp. Physiol. Psychol. 67
40 J. Talairach, P. Tournoux, Co-Planer Stereotaxic Atlas of the Human
Brain, Thieme Medical Publishers, New York, 1988.
41 A.E. Taylor, J.A. Saint-Cyr, A.E. Lang, Frontal lobe dysfunction in
Parkinson’s disease. The cortical focus of neostriatal outflow, Brain
109 1986 845–883.
42 M.C. Tresch, H.M. Sinnamon, J.G. Seamon, Double dissociation of
spatial and object visual memory: evidence from selective interfer-
ence in intact human subjects, Neuropsychologia 31 1993 211–219.
43 L. Ungerleider, J. Haxby, ‘What’ and ‘where’ in the human brain,
Curr. Opin. Neurobiol. 4 1994 157–165.
44 L.G. Ungerleider, S.M. Courtney, J.V. Haxby, A neural system for
visual working memory, Proc. Natl. Acad. Sci. U.S.A. 95 1998
45 L.G. Ungerleider, M. Mishkin, Two cortical visual systems, in: D.J.
Ingle, M.A. Goodale, R.J.W. Mansfield Eds. , Analysis of Visual
Behavior, MIT Press, Cambridge, MA, 1982, pp. 549–586.
46 M. Watanabe, Reward expectancy in primate prefrontal neurons,
Nature 382 1996 629–632.
47 R.P. Woods, Modeling for intergroup comparisons of imaging data,
NeuroImage 4 1996 S84–S94.
48 E. Zarahn, G.K. Aguirre, M. D’Esposito, Empirical analyses of
BOLD fMRI statistics: I. Spatially unsmoothed data collected under
null-hypothesis conditions, Neuroimage 5 1997 179–197.
49 E. Zarahn, G.K. Aguirre, M. D’Esposito, A trial-based experimental
design for fMRI, Neuroimage 6 1997 122–138.
50 E. Zarahn, G.K. Aguirre, M. D’Esposito, Temporal isolation of the
neural correlates of spatial mnemonic processing with functional
MRI, Cogn. Brain Res. 7 1999 255–268.