Neuropsychologia, Vol. 36, No. I, pp.

109-114, 1998
~/ 1998 Elsevier Science Ltd. All rights reserved
~ Pergamon PII: S0028 3932(97)00017-1 Printed in Great Britain
0028 3932/98 $19.00+0.00

Impairments of mental rotation in Parkinson's

disease
A. C. LEE,*t J. P. HARRIS* and J. E. CALVERT~

*Department of Psychology, University of Reading, Whiteknights, Reading RG6 6AL, U.K.;

SDepartment of Psychology, University of Bristol, 8 Woodland Road, Bristol BS8 ITN, U.K.

(Received 17 June 1996; accepted 7 January 1997)

Abstract--It is controversial whether parkinsonian patients are impaired on visuospatial tasks. In the present study, patients and
normal control subjects judged whether pairs of wire-frame figures in different orientations were the same or different. The orientation
difference between the figures was either in the picture plane (around the z-axis, or two-dimensional) or in depth (around the y-axis,
or three-dimensional). Reaction times and error rates were measured. For the two-dimensional task, there were no significant
differences in errors between the two groups, though Parkinsonian subjects were significantly slower to respond than the control
group. In the three-dimensional task, patients had a different pattern of reaction times from the controls and made significantly
more errors, which were systematic at large angular differences. The results suggest a visuospatial deficit in Parkinson's disease,
which reflects problems in some aspect of the perception of extra-personal space. © 1998 Elsevier Science Ltd. All rights reserved

Key Words: visuospatial tasks; object-centred processing; image-centred processing; extra-personal space.

Introduction
One of the more controversial questions about possible
cognitive changes in Parkinson's disease (PD) has been
whether or not there is impairment of visuospatial pro-
cessing in the illness. Early studies have generally con-
cluded that there is evidence for such a deficit. For
example, patients are impaired on visuospatial subtests
of the WAIS [7] and in route-finding tasks [3]. However,
there are several problems with such studies [4, 5]. If the
dependent variable is time, patients will be at a dis-
advantage because of the accepted effects of the illness on
the motor system. Again, many tests require a sequence of
operations, and it may be in the organization or execution
of such a sequence that the patients' difficulties lie, rather
than in visuospatial processing per se.
In Brown and Marsden's study [4], a mental rotation
task [11] was employed. A Maltese cross was presented
on a computer screen, with an arrow in one arm and a
dot in one of the adjacent arms. The subjects' task was
to align themselves mentally with the arrow, and to press
one of two push-buttons, depending on whether the dot
t Address for correspondence: Department of Psychology,
University of Reading, Whiteknights, Reading RG6 6AL,
U.K.; e-mail: a.c.lee@reading.ac.uk; tel: + 1189 318522; fax:
+ 1189 316604.
was on their left or right. When the arrow pointed up, no
mental reorientation was required; when it was hori-
zontal, re-orientation through 90 ° was required; when it
pointed down, re-orientation through 180 ° was required.
PD subjects, although slower than controls in all
conditions, were not differentially worse in the conditions
that required greater amounts of re-orientation. Brown
and Marsden concluded that these results did not provide
evidence for an impairment of mental manipulation and
so for a general visuospatial deficit in the illness.
However, it can be argued that this task is not a strong
test of mental manipulation, for two reasons. First, it
requires the subject to change the relationship between
themselves and the stimulus, rather than the relationship
between two stimuli. The latter may put greater demands
on visuospatial mechanisms. Secondly, and perhaps more
importantly, even if the subjects were mentally man-
ipulating the stimulus as Brown and Marsden suggest,
this could be done as a two-dimensional rotation of
an image-based representation, rather than a three-
dimensional rotation of an object-based representation.
Other studies (e.g., [8]) have used other tasks to explore
the ability of the parkinsonian to mentally manipulate
objects in three dimensions. Ogden and co-workers used
a subsection of the Differential Aptitude Tests [1], in
which the subject was shown a drawing of a two-dimen-
sional shape and was asked to decide which of four pic-

109
110 A.C. Lee et al./Mental rotation in Parkinson's disease
tures of various three-dimensional shapes represented the
target shape if it were to be folded along specific lines. The
authors suggest that this task requires quite sophisticated
visuospatial abilities, as the subject has to both form and
manipulate mental images, without having to sequence
behaviour. Although the patient group were significantly
slower than the controls, both groups made a similar
amount of errors (between 33% and 35%). Ogden et al.
concluded that performance on this task demonstrated
that the patients did not have a visuospatial deficit. How-
ever, it could be argued that the task can be done in an
image-centred way, using local two-dimensional rather
than global three-dimensional information. For example,
if the local two-dimensional shape defined by the specified
line and the edges of the paper were a triangle, this would
immediately rule out a cube as a possible candidate three-
dimensional shape. Given the high error rates in the task,
in both patients and controls, such a strategy seems to
give a possible account of the data.
In this paper, we compare performance on two tasks
involving the kind of wire-frame figures introduced by
Shepard and Metzler [12]. Both tasks involve same/
different judgements of two such figures which vary in
orientation. Thus, subjects cannot perform the task by
mentally re-orientating themselves with respect to one or
both stimuli, but must re-orientate one stimulus with
respect to the other. In one experiment, the task involves
rotation in the picture plane (around the z-axis), as used
Fig. 1. Examples of stimuli shown at 20 ° separations.
by Brown and Marsden, whereas, in the other, rotation
in depth (around the y-axis) is required.
Method
Stimufi
Images based upon the Shepard and Metzler [12] figures
were used. Eight perspective line drawings of unfamiliar three-
dimensional figures were used (see Fig. 1). Each figure com-
prised 10 individual wire-frame cubes laid end-to-end.
After the third, the fifth and the eighth blocks, the next cube
was attached to one of the sides, rather than to the end, of the
last cube, thus creating a 90 ° angle in the figure. The shapes
were chosen for their ability to be rotated around the y-axis,
without one part of the figure occluding most of another part.
Thus, whatever the viewpoint, most of the figure was always in
view. The stimuli were presented in pairs, one on each side of a
computer monitor screen, in a range of orientations. Examples
of the stimuli are shown in Fig. 1.
Apparatus
The test stimuli were generated by a Pluto 11 graphics system,
(spatial resolution 768H x 288V pixels), and displayed on a
Digivision CD 14 RGB monitor. The stimuli were white against
a black background. The Pluto was programmed by an IBM
PC AT clone, to which two response buttons were connected
via an I/O card.
i
 A. C. Lee et al./Mental rotation in Parkinson's disease 111

Procedure that reaction times (RTs) increase with angular discrep-

E x p e r i m e n t I: t h r e e - d i m e n s i o n a l r o t a t i o n . The subjects were
seated 1 m from the screen so that the display subtended a
horizontal visual angle of 10°. They were shown a dem-
onstration of the principles of mental rotation, using models.
Then the subjects were shown examples of the stimuli on the
screen. In this experiment, one of the figures was rotated with
respect to the other (by a variable amount) around the y-axis.
Subjects were told that they had to decide if the figure on the
right was the same in shape (but globally rotated) as the figure
on the left. They were to press either of two buttons ('same' or
'different') to indicate their choice. The figures would remain
on screen until their decision was made, so there were no time
limits. They were to signal that they were ready for the next
trial by pressing one of the response buttons, when the next
stimulus would appear on screen.
Each subject was given a practice session of 15 trials, to
familiarize them with the task and the response box. In the
experimental sessions, there were 80 trials. On 50% of them,
the figures were the same, and on 50% they were different.
It was stressed to the subjects that they should try to be as
accurate, rather than as fast, as possible. Testing occurred
30 min after the PD subjects had taken a therapeutic dose of
L-DOPA.
E x p e r i m e n t 2." t w o - d i m e n s i o n a l r o t a t i o n . The procedure of
this experiment was identical to that of Experiment 1, except
that the demonstration and explanation of the principles of
mental rotation referred to two, rather than three, dimensions.
That is, rotation was around the z-axis (in the plane of the
screen).
Subjects.
Parkinsonian subjects. Experiment l: three-dimensional
rotation. Seven volunteer patients (six male, one female),
diagnosed by a consultant neurologist as having idiopathic PD,
were recruited from the regional Parkinson's Disease Society
[mean age 66.6 years (S.D. 5.4years), L-DOPA daily 878.3 mg
(S.D. 357.8 mg)]. These patients were selected because they all
had little or no akinesia in their dominant hand after medi-
cation, which might have affected their ability to use the
response box. Subjects scored between Stages II and III on the
Hoehn and Yahr scale.
Experiment 2: two-dimensional rotation. Two of the patients
who took part in Experiment 1 were unavailable for this experi-
ment. Testing of the remaining five subjects ensured that medi-
cation, duration of illness and symptoms were similar in the
two experiments.
Controlsubjects. Experiment 1: three-dimensional rotation.
Seven age- and sex-matched control subjects (six male, one
female) were used [mean age 63.6 years (S.D. 4.6 years)].
Experiment 2: two-dimensional rotation. The same five con-
ancy. Figure 2 shows that such a trend was generally
found in the present data when the figures were the same,
both for patients and controls, though the control data
are the m o r e variable, once the angular difference exceeds
140 r~. The graph of the RTs of parkinsonian subjects has
an apparently shallower slope for the 'same' stimuli than
that of the controls, though this difference was not sig-
nificant because o f the scatter in the data at the higher
angular discrepancies (t--0.03; P > 0 . 0 5 ) . The slopes of
the graphs of RTs for the 'different' stimuli are very
shallow, though slightly steeper for patients than
controls.
RTs to respond to the same images were significantly
slower [F(1,12) = 35.18; P = 0.0001] than those to different
images. This m a y reflect a difference in the way that the
two tasks can be solved. It can be obvious, on 'different'
trials, that the stimuli are different, from a c o m p a r i s o n
of individual limbs of the figures, which can occur early
in the presentation. However, if no incongruent limb is
immediately apparent, as on the ~same' trials, the subject
must continue the rotation of the images before being
sure that the figures are indeed the same. It appears that
the control group did indeed use this strategy, but the
PD group did not. Instead, they responded m o r e quickly
than the controls, but at the cost of systematic errors (see
below).
Overall, there was no significant difference in the reac-
tion times of the two subject groups [F(1,12)=0.07;
P > 0 . 7 ] , because of an interaction between group and
image type. Thus, for the 'same' stimuli, the P D group
were faster (for most angular differences) but for 'differ-
ent' stimuli they were slower than the control group.
Although this interaction just missed significance
[F(1,12) = 4.14; P = 0.06], analysis of the simple effects in
the interaction showed that they were all significant at
the 2 % level at least, except for the difference between
the P D and control group R T s for the 'same' images.
Errors. Figure 3 shows the mean percentage error
scores for P D subjects and controls. As the angular
SAME
24 [--
22 f
2018 _•PD PD DIFF
CO SAME
CO DIFF , , ,~t/"
/~
XX A

trols who had been matched to these PD subjects in Experiment

1 were used.

Results bl , ~ U].. . "53 412. . . . [5- " - - "C3- " - "~

41 . . . . [ 2 , . . _ . q D - - " ----E-'--"
2~ ~ o ~ & ~ o
Experiment l: three-dimensional rotation I I 1 [ I t I I I
20 40 60 80 100 120 140 160 180
ANGULAR DISCREPANCY
Reaction t i m e s . The three-way analysis of variance (in degrees)
( A N O V A ) showed a highly significant main effect of Fig. 2. Mean reaction times for judgements of the same and
angular discrepancy [F(9,108)=7.82; P < 0 . 0 0 0 1 ] , as different pairs of images for parkinsonian and control subjects
expected f r o m the finding of Shepard and C o o p e r [11] in Experiment 1.
112 A.C. Lee et al./Mental rotation in Parkinson's disease
PD SAME
100 - ,--c]-- PD DIFF
90 - ~ CO SAME
•-O-- CO DIFF
80
70-
~. 60
50
~ 40
30
20
0 20 40 60 80 100 120 140 160
ANGULAR DISCREPANCY
(in degrees)
Fig. 3. Mean percentage errors for judgements on the same and
different pairs of images for parkinsonian and control subjects
in Experiment 1.
difference between the stimuli was increased, PD patients
became relatively less accurate compared to controls. The
heavy black line indicates the level at which the subjects
would be performing if they were responding at chance.
As Fig. 3 shows, the PD subjects performed at worse than
chance on the 'same' stimuli once the angular discrepancy
exceeded about 140 ° .
Analysis of variance of the data shows a significant
main effect of angular discrepancy [F(9,108)=10.27;
P < 0.00001] and for whether the figures were the 'same'
or 'different' [F(1,12)= 10.44; P = 0.007]. There is also a
highly significant interaction between angular dis-
crepancy and image type [F(9,108) = 5.32; P < 0.00001 ].
Overall, there was a significant difference in the error
rates of the two subject groups [F(1,12) = 6.74; P < 0.02],
with the patients making more errors. There were also
significant interactions which can be best understood by
analysis of the underlying simple effects. The PD group
made significantly more errors than the controls for the
'same' images (at the 2% level), but not for the 'different'
images. In addition, the patients had significantly higher
error rates (at the 0.5% level) in recognizing the 'same'
images than the 'different' images, whereas the control
group had similar error rates for both types of stimulus.
For the last three angles of the sequence (140 °, 160 °,
180°), it can be seen that the parkinsonian subjects are
performing worse than would be expected if they were
responding randomly, i.e. at chance. This suggests that
the patients were making systematic errors, perhaps
through the use of an inappropriate strategy.
Experiment 2: two-dimensional rotation
Reaction times.. Figure 4 shows the mean RTs for the
PD and control groups in Experiment 2. The charac-
teristic increase in the time taken to respond with angular
discrepancy is still apparent for both subject groups,
14 "-0-" PD SAME
- C ~ PD DIFF i
12 .
CO SAME
_ Q _ ~
/\
A
10
s
4[
2~_
180
o I 1 I I r J I I I
20 40 60 80 100 120 140 t 60 180
ANGULAR DISCREPANCY
(in degrees)
Fig. 4. Mean reaction times for judgements on the same and
different pairs of images for parkinsonian and control subjects
in Experiment 2.
although the slopes of the graphs for the control group
are much flatter.
It can be seen that the pattern of results is quite differ-
ent from that found in Experiment 1, as both patients
and controls show similar reaction times to the 'same'
and 'different' stimuli. The PD group are always at least
2 sec slower than the control group, a difference which
was significant [from the ANOVA F(1,8)=10.86;
P < 0.02], but performed faster than they did in Experi-
ment 1. The group × angles interaction is also highly sig-
nificant [F(9,72)= 3.28; P = 0.02], reflecting the fact that
the difference between the two groups gets larger with
increasing angular discrepancy.
The differences in response times within both groups
of subjects for the 'same' and 'different' images are less
apparent in this experiment than in Experiment 1, and
neither reached significance at the 5% level.
Errors.. The mean percentage errors are plotted in Fig.
5. There is a large difference between the pattern of errors
found in Experiment 1 and that in this experiment. There
is a ceiling effect for smaller angular discrepancies, with
both the PD and control subjects making no or very few
errors. Even at large angular discrepancies, the mean
percentage of errors did not exceed 15 %. Non-parametric
analysis using the Friedman test (the data were not con-
sidered suitable for analysis with ANOVA because of a
lack of homogeneity) showed no significant differences
between the groups at any angular discrepancy for both
targets and distracters.
The graphs show that the number of errors made by
both groups increases with angular discrepancy, but not
as steeply as in Experiment 1. Above about 140 ° angular
discrepancy, the error rates for the PD group are slightly
higher than those for the control group. However, the
PD group shows no signs of the systematic errors at large
angular discrepancies found in the first experiment.
 A. C. Lee et al./Mental rotation in Parkinson's disease
PDDIFF [
CO SAME
CO DIFF
~- 15 1
~ 10
5 because the tasks were subtly different. Our task required
20 40 60 80 100 120 140 160 180
ANGULAR DISCREPANCY
(in degrees)
Fig. 5. Mean percentage errors for judgements on the same and
different pairs of images for parkinsonian and control subjects
in Experiment 2.
Discussion
The results of these experiments may be summarized
as follows. In two-dimensional mental rotation of one
wire-frame stimulus with respect to another, the patterns
of response times and errors are similar in parkinsonian
and in control subjects. The error rates are similar in both
groups, and, though slower, the PD response times show
similar increases with angular discrepancy to those of the
controls. In three-dimensional mental rotation, on the
other hand, though the response times show similar pat-
terns in both groups, there are large differences in error
rates. In particular, at large angular discrepancies, pat-
ients perform worse than chance, suggesting that their
errors are systematic. The results suggest that the patients
may be performing the two-dimensional rotation task in
a similar way to the controls, even though their responses
are slower, but that for three-dimensional rotation, at
large angular discrepancies, patients are using a different
strategy to that of the controls.
The findings raise a number of issues:
1. What light do the findings throw on the nature of
visuospatial processing in PD?
2. What is the strategy used by the patients in the three-
dimensional task?
3. How do the findings relate to broader PD symp-
tomatology?
On the first issue, the patients are clearly worse than
controls on two-dimensional mental rotation as indicated
by the significantly higher intercepts and steeper slope
RT functions, and slightly worse in terms of error rates.
However, it could be argued that this is not strong evi-
dence for a major visuospatial impairment in the illness,
but rather for a slowing of visuospatial processing which
is comparable to the slowing of other cognitive processes
in the illness. The PD error data from the large angular
113
discrepancies in three-dimensional rotation, on the other
hand, strongly suggest that there is a qualitative, not
simply a quantitative, difference between patients and
controls on this task. This finding suggests a positive
answer to the question (answered negatively by Brown
and Marsden [4]) of the existence of a visuospatial impair-
ment in PD. The discrepancy between Brown and
Marsden's finding that there was no differential slowing
with increased angular difference in a two-dimensional
rotation task and the present data could have arisen
subjects to rotate one stimulus with respect to another,
whereas that of Brown and Marsden required rotation
of the stimulus with respect to the subject. The latter task
may not allow any visuospatial impairment to become
apparent.
The nature of the patients' visuospatial problems is
less clear. One relevant piece of evidence comes from the
systematic errors made at large angular discrepancies in
the three-dimensional task. These errors imply that stim-
uli which were in fact the same but rotated close to 180°
with respect to each other were judged to be different.
One possible explanation for this effect is that patients
relied too heavily on local instead of global matching.
For example, they could select, say, the lowest vertical
segment of the figures (note that this was in the same
orientation in both figures), then compare the orien-
tations of the angles between those segments and the
adjoining segments. For rotations close to 180"~,the orien-
tations of the segments would be similar, but the upper
segments would be pointing in opposite directions in the
two figures. This would be taken (incorrectly) by the
patients as evidence for a difference, whereas the controls,
who (we assume) would perform the task with a more
global comparison, would make less errors of this kind.
This notion of local processing could also be an expla-
nation for the results of Ogden et al. [8] on a mental paper
folding task [1]. For example, a subject may make one
fold along a specified line, and on this basis may be able
to eliminate some of the four response options.
Visuospatial functions can be evaluated by tasks that
involve three components: the understanding of relative
positions of stimulus objects in space; the integration
of the stimuli into a simple spatial framework; and the
performance of mental tasks involving that spatial frame-
work, (e.g., [2]). Many parkinsonian patients report that
they experience a sense of constriction of visual space. In
Sacks' [10] evocative words: "what we observe, however,
is that all his space-time judgements are pushed out of
shape, that his entire co-ordinate system is subject to
expansions, contractions, torsions and warps...". Ano-
ther of Sacks' parkinsonian patients suffered with a deficit
in proprioception which made him unaware that he
walked with a 10Utilt [9]. This observation is in line with
the results of laboratory studies showing that patients
may mis-perceive visual vertical and/or horizontal [6]. In
a very different letter cancellation task, Villardita et
al. [13] demonstrated peripheral visual neglect in the
114 A.C. Lee et al./Mental rotation in Parkinson's disease
disease, another result consistent with a constriction
of peri-personal space. These results suggest that the
parkinsonian's perception of extra-personal space is
affected by the disease.
We suggest that it is this global distortion of their
spatial frame of reference which affects three-dimensional
mental rotation in PD. The local feature-matching strat-
egy, which, we suppose, is adopted to compensate for the
impairment, is adequate for two-dimensional, but not
three-dimensional, rotation tasks.
Acknowledgements--This work was supported by an MRC
studentship (A.C.L.) and grants from the Parkinson's Disease
Society (J.P.H.) and the Lord Dowding Fund (J.E.C.). We
thank Dr Gavin Brelstaff for help with programming.
References
1. Bennett, G. K., Seashore, H. G. and Wesman, Dif-
ferential Aptitude Tests Manual, 5th edn. Psycho-
logical Corporation, New York, 1972.
2. Boiler, F., Passafiume, D., Keefe, N. C., Rogers, K.
and Kim, Y. Visuospatial impairment in Parkinson's
disease. Archives of Neurology 41,485-490, 1984.
3. Bowen, F. P., Hoehn, M. M. and Yahr, M. D. Par-
kinsonism: alterations in spatial orientation as deter-
mined by a route walking test. Neuropsychologia 10,
355-361, 1972.
4. Brown, R. G. and Marsden, C. D. Visuospatial func-
tion in Parkinson's disease. Brain 109, 987-1002,
1986.
5. Brown, R. G. and Marsden, C. D. Cognitive function
in Parkinson's disease: from description to theory.
Trends in Neuroscience 131, 21-29, 1990.
6. Danta, G. and Hilton, R. C. Judgement of the visual
vertical and horizontal in patients with Park-
insonism. Neurology 25, 43-47, 1975.
7. Loranger, A. W., Goodell, H., McDowell, F. H.,
Lee, J. E. and Sweet, R. D. Intellectual impairment
in Parkinson's syndrome. Brain 95, 405-412, 1972.
8. Ogden, J. A., Growden, J. H. and Corkin, S. Deficits
on visuospatial tests involving forward planning in
high-functioning parkinsonians. Neuropsychiatry,
Neuropsycholoyy and Behavioural Neurology 3, 125-
139, 1990.
9. Sacks, O., The Man Who Mistook His Wife For A
Hat. Pan, London, 1985.
10. Sacks, 0., Awakenings, 2nd edn. Pan, London, 1990.
11. Shepard, R. N. and Cooper, L. A., Mental Images
and Their Transformations. MIT Press, Cambridge,
MA, 1982.
12. Shepard, R. N. and Metzler, J. Mental rotation of
three-dimensional objects. Science 171, 701-703,
1971,
13. Villardita, C., Smirni, P. and Zappala, G. Visual
neglect in Parkinson's disease. Archives of Neurology
40, 737-739, 1983.