DOI: 10.1093/brain/awg075

Brain (2003), 126, 908±919

Pure agnosia for mirror stimuli after right inferior
parietal lesion
Konstantinos Priftis,1,2 Elena Rusconi,1 Carlo UmiltaÁ1 and Marco Zorzi1
1University of Padova, Padova and 2University of Trieste,
Trieste, Italy

This study reports the experimental investigation of
G.R., a patient suffering from a highly speci®c disorder
in discriminating mirror stimuli following a right temporoparietal cerebrovascular accident. G.R. showed
intact perceptual, attentional, mnestic, linguistic and
executive abilities. Object recognition was accurate even
under unusual viewing conditions. He was highly accurate in de®ning the canonical orientation of common
objects and in discriminating misoriented objects
among identical distracters. However, he was severely

impaired in tasks requiring mirror-stimulus discrimination, a de®cit that persisted even when the object's
coordinates were systematically misaligned with respect
to his body. The disorder was also dependent upon the
frame of reference (allocentric versus egocentric) activated on the basis of task demands. These results demonstrate the existence of a highly speci®c disorder in
discriminating mirror stimuli de®ned in object-based
coordinates, suggesting a failure in processing the directionality of an object's intrinsic x-axis.

Keywords: mirror images; object recognition; frames of reference; agnosia
Abbreviations: BORB = Birmingham Object Recognition Battery; VOSP = Visual Object and Space Perception test

Introduction

An intriguing neuropsychological ®nding in the domain of
visual perception is that object recognition and the processing
of object orientation are supported by different brain mechanisms. Friedrich Best was the ®rst to describe striking
dif®culties in judging object orientation and preserved object
recognition, in a patient who suffered bilateral lesions to the
inferior parietal regions (Ferber et al., 2003).
The same pattern has been described recently in other
patients (Turnbull et al., 1997a; Turnbull, 1997; Karnath
et al., 2000; Cooper and Humphreys, 2000; Harris et al.,
2001). They show dif®culty in judging whether an object is
presented in its upright everyday-life orientation or misoriented. Some patients may copy line drawings accurately
yet systematically ignore the orientation of the original,
`normalizing' misoriented drawings to their upright orientation (Turnbull et al., 1995; Solms et al., 1998). At least one
case showed the converse dissociation: Patient D.M.
(Turnbull, 1997) correctly de®ned the orientation of the
majority of the objects he could not recognize. Finally, some
patients can recognize letters effortlessly despite lacking
information regarding their orientation (Solms et al., 1988;
Robertson et al., 1997).
ã Guarantors of Brain 2003

Other dissociations have been observed within the domain
of object orientation processing. Patients may show dif®culty
in detecting targets rotated 180° clockwise with respect to
upright distracters (Turnbull et al., 1997a; Harris et al., 2001).
At least in one case (Patient S.C.; Turnbull et al., 1997a),
discrimination of drawings rotated 180° clockwise was
impaired (29 correct out of 50), whereas mirror-stimulus
discrimination was spared (50 out of 50). The converse
pattern of impaired mirror-stimulus discrimination with
spared discrimination of other orientations has also been
reported. Patients R.J. (Turnbull and McCarthy, 1996) and
F.I.M. (Davidoff and Warrington, 2001) had severe dif®culty
discriminating mirror stimuli. However, their performance
was largely spared in discriminating other stimulus rotations
(e.g. 45°, 90° and 180° clockwise). Moreover, object
recognition, at least for objects observed from conventional
canonical viewpoints, was intact. These results suggest that
mirror-stimulus discrimination and processing of other object
rotations (e.g. 180° clockwise) might be subserved by
relatively independent brain systems.
Various theoretical accounts have been proposed to explain
the disorder for mirror stimuli. Turnbull and McCarthy

Downloaded from https://academic.oup.com/brain/article/126/4/908/331896 by guest on 11 April 2024

Summary

Correspondence to: Professor Marco Zorzi, Dipartimento
di Psicologia Generale, via Venezia 8, 35131, Padova,
Italy
E-mail: marco.zorzi@unipd.it

Agnosia for mirror stimuli

Patient R.J. (Turnbull and McCarthy, 1996) also had considerable dif®culty recognizing silhouettes and objects observed
from unusual viewpoints, despite intact object recognition
under normal viewing conditions. In summary, these studies
seem to indicate competition between viewer-dependent and
viewer-independent mechanisms. As a consequence, engagement of the viewer-independent mechanism would cause
mirror discrimination impairment, with object recognition
preserved except for the case of unusual viewing conditions.
The aim of the present study was to describe a new case
with a severe and highly speci®c impairment restricted to
tasks requiring discrimination of mirror stimuli. Experiments
1±5 aimed to verify the speci®city of the de®cit with respect
to other orientation manipulations, whereas the aim of
Experiments 6±9 was to test the different explanations
proposed for the mirror-stimulus discrimination de®cit.

Case presentation

G.R., a 72-year-old right-handed grocer with 8 years of
formal education, suffered from a right hemisphere cerebral
vascular accident in December 1993. Standard neurological
assessment demonstrated left hemiplegia and hemianaesthesia. Doppler echography showed a thrombotic obstruction of
the right internal carotid artery and stenosis of the left internal
carotid artery. A computed tomography (CT) scan evidenced
a hypodense region in the right hemisphere extending from
the insula to the inferoparietal region. Midline structures were
not changed. Older asymptomatic lesions were present in the
left inferior occipitotemporal cortex and in the cerebellum
bilaterally. Functional imaging by SPECT (single photon
emission computed tomography) (May, 1994) demonstrated
right temporoparietal hypoperfusion. R.G. underwent a right
STA-MCA (super®cial temporal artery-middle cerebral
artery) artery bypass operation, followed by a second
operation (May, 1995) to reconstruct the left internal carotid
artery.
A CT scan (January, 1996) did not reveal postoperative
complications and recon®rmed the hypodense area in the
right insula, extending to the corresponding subcortical white
matter, combined with dilatation of the lateral ventricular
structures. In May 1998 regional cerebral blood ¯ow (rCBF)
was assessed by xenon 133 clearance inhalation, demonstrating bilateral diffuse hypoperfusion, more evident in the right
parietal cortex (Fig. 1). A recent MRI brain scan (March
2002), recon®rmed all previous neuroradiological ®ndings
(Fig. 1). At the time of writing, G.R.'s left hemiplegia has
regressed to a moderate left hemiparesis, more evident in the
left inferior limb.

Neuropsychological assessment

Following his cerebrovascular accident, G.R. presented with
symptoms of left visuospatial neglect that recovered spontaneously and were completely absent during the formal
neuropsychological evaluation and experimental investiga-

Downloaded from https://academic.oup.com/brain/article/126/4/908/331896 by guest on 11 April 2024

(1996) suggested that it might be due to an inability to assign
suitable frames of reference to the observer's body (egocentric coordinates), to the inspected object (allocentric
coordinates) or to a translation from allocentric to egocentric
coordinates. R.J. had dif®culty judging stimulus orientation
with reference to a leftward or rightward arrow, suggesting
impairment of handedness encoding. Unfortunately, this
hypothesis could not be tested. Harris and colleagues have
recently advanced a similar hypothesis to explain orientation
agnosia as a de®cit in ®nding coordinate axes (Harris et al.,
2001).
Davidoff and Warrington (2001) have proposed an
explanation of mirror-stimulus discrimination de®cit in
terms of pathological engaging on viewer-independent representations supported by the posterior inferotemporal cortex
(i.e. the ventral visual pathway; Ungerleider and Mishkin,
1982; Milner and Goodale, 1995). At least for objects viewed
from angles in which critical features are available to the
observer, viewer-independent representations might be the
principal gateway for object recognition, identi®cation and
naming (Davidoff and Warrington, 1999; Warrington and
Davidoff, 2000). Microelectrode single-cell recordings from
area IT in monkeys have shown invariant responses for
original and mirror stimuli (Rollenhagen and Olson, 2000;
Baylis and Driver, 2001), suggesting a representational
format that is independent of object orientation. The spatial
attributes of objects, however, would be coded by viewerdependent mechanisms supported by parietal networks (i.e.
the dorsal pathway; Ungerleider and Mishkin, 1982; Milner
and Goodale, 1995). Viewer-dependent representations are
sensitive to orientation changes relative to the observer and
would subserve object recognition under non-optimal viewing conditions (e.g. when objects are fragmented, overlapping, shaded or foreshortened; for review see Turnbull et al.,
1997b). Thus, if, after parietal damage, object processing
takes place only in the ventral pathway, the engagement in
viewer-independent representations might prevent accurate
analysis of object orientation, leading to confusion of mirror
stimuli.
In support of this hypothesis, Warrington and Davidoff
(2000) reported the case of Patient J.B.A., who presented with
severe dif®culty in both object recognition and mirror
discrimination tasks. J.B.A.'s performance in mirror discrimination tasks was markedly better for objects that she was not
able to recognize. Warrington and Davidoff (2000) explained
the improvement in mirror-stimulus discrimination as a
consequence of disengagement from these viewer-independent representations. Furthermore, Patient F.I.M. (Davidoff
and Warrington, 2001), who showed a de®cit in discriminating mirror stimuli, was markedly impaired in recognizing
objects observed from unusual viewpoints, but was very
accurate with the same objects when viewed from conventional, canonical viewpoints, in which critical features can be
processed easily. The association between agnosia for
unusual views and impaired discrimination of mirror stimuli
was taken to support the engagement hypothesis. Note that

909

910

K. Priftis et al.

tions. Consent for our study was obtained according to the
Declaration of Helsinki. G.R. appeared alert, collaborative
and fully oriented for time, space and personal information.
He had no dif®culties of memory or communication in
everyday life and his social behaviour and emotional control
remained appropriate and well adapted. Standard neuropsychological testing revealed markedly preserved perceptual,
ideomotor, linguistic, memory and executive functions,
adjusted for his age and educational level (Table 1).
G.R.'s perceptual abilities were well preserved both in
tasks demanding feature integration (Street Completion Test)
and in tasks requiring coding of critical features [Silhouettes
and Birmingham Object Recognition Battery (BORB) subtest
7] or the de®nition of principal axes of elongation (BORB
subtest 8). Thus, in contrast with Patients R.J. (Turnbull and
McCarthy, 1996) and F.I.M. (Davidoff and Warrington,
2001), G.R. demonstrated well-preserved perceptual processing for both usual and unusual viewpoints. This ®nding
appears crucial in relation to Davidoff and Warrington's
(2001) viewer-independent engagement hypothesis (see
General discussion).
G.R.'s constructional praxis was preserved and he displayed normal local±global attentional shifting in copying
letters composed of smaller letters. Perception of symmetry
was intact. In a modi®ed version of the Benton Line
Orientation Test, G.R. was presented with 16 pairs of
vertically oriented lines and 16 pairs of horizontally oriented
lines. Four line pairs were parallel, whereas the remaining

line pairs formed angles between 5° and 15°. G.R.'s task was
to judge if the lines were parallel or formed an angle. His
performance was perfect for parallel (8 out of 8) and 15° (8
out of 8). He made two errors in discriminating differences
of 10° (6 out of 8) and was impaired in discriminating
differences of 5° (0 out of 8). Thus, G.R.'s ability to
discriminate angle variations was largely spared, at least for
lines that were parallel or formed angles of >5°. Furthermore,
G.R. was perfectly accurate in judging lines as horizontal or
vertical, suggesting intact assignment of fundamental spatial
axes of reference, at least with respect to his own body.
G.R. was accurate in detecting the odd one out among four
drawings in which the target could differ from the distracters
in size, colour or structure or was rotated 180° clockwise,
whereas he was severely impaired when it was a mirror
stimulus (Table 2). This prompted the experimental investigation.

Experiment 1: orientation judgement
Introduction

Some patients (Turnbull 1997; Turnbull et al., 1997a;
Karnath et al., 2000; Cooper and Humphreys, 2000; Harris
et al., 2001) have dif®culty deciding whether an object is
presented in its upright canonical position or rotated.
According to Turnbull and colleagues, they are orientation
agnosics (Turnbull et al., 1997a, b). The aim of the present

Downloaded from https://academic.oup.com/brain/article/126/4/908/331896 by guest on 11 April 2024

Fig. 1 (A) rCBF scan showing right parietal hypoperfusion. (B) MRI scans of G.R.'s brain. The right
hemisphere corresponds to the left half of each scan.

Agnosia for mirror stimuli

911

Table 1 Neuropsychological assessment

experiment was to rule out the presence of orientation
agnosia.

Stimuli, apparatus and procedure

Thirty-two object drawings (Snodgrass and Vanderwart,
1980) previously used by Turnbull and colleagues (Turnbull
et al., 1997a) were presented to G.R. printed on a 10 3 12 cm
¯ash card. All stimuli had a canonical, everyday-life orientation. Eight were presented at 0° (canonical orientation),
eight were rotated 90° clockwise, eight were rotated 90°
anticlockwise, and the remaining eight stimuli were rotated
180° clockwise. G.R. had to name the drawings and judge if
they were represented in their everyday-life orientation.

Results

G.R. named every drawing correctly and was accurate in
judging if the objects depicted were in upright canonical
orientation or were disoriented (32 out of 32). Thus, G.R.

Cut-off score

28/30

24/30

189 s
428 s
239 s
4.5 out of 5
7.6
6/6
3/12

>209 s
>558 s
>384 s
<3/5
<7.3
<1/6
0/12

35/36

<26/36

7.25/14
19/20
24/25
23/25
9/9
40/40
9/10
20/20
8/10
10/10

<2/14
±
19/25
19/25
<7/9
<32/40
<8/10
<18/20
<7/10
<6/10

13/14
6/6

<7.75
<5

4

<3.3

13/28
19/28
16/28
4
8.85
458/469

<6.3
<5.9
<6.25
<3.50
<5.50
<447

could accurately create a viewer-centred representation and
associate it with a more abstract viewer-independent representation, which speci®es canonical object orientation.

Experiment 2: disoriented drawings
Introduction

The aim was to compare G.R.'s performance with that of
other patients, who could copy object and non-object
drawings accurately but produced systematically disoriented
copies, suggesting preserved perception of object structure
with impaired processing of object orientation (Turnbull et al.,
1995; Solms et al., 1998).

Methods
Stimuli, apparatus and procedure

Twelve drawings depicting real objects (Snodgrass and
Vanderwart, 1980), previously used by Turnbull and
colleagues (Turnbull et al., 1997a), were presented to G.R.

Downloaded from https://academic.oup.com/brain/article/126/4/908/331896 by guest on 11 April 2024

Screening
Mini-Mental State Examination (Mondini et al., 2001)
Attention and executive functions
Trail-making test (Mondini et al., 2001)
Form A
Form B
Form BA
Cognitive estimation (Mondini et al., 2001)
Phonemic ¯uency (Mondini et al., 2001)
Verbal reasoning (Mondini et al., 2001)
Raven (short form) (Mondini et al., 2001)
Language
Token test (Spinnler and Tognoni, 1987)
Perceptual and spatial functions
Street Completion Test (Spinnler and Tognoni, 1987)
Silhouettes (Warrington and James, 1986)
BORB subtest 7 (Riddoch and Humphreys, 1993)
BORB subtest 8 (Riddoch and Humphreys, 1993)
Line bisection (Wilson et al., 1987)
Letter cancellation (Wilson et al., 1987)
VOSP, dot counting (Warrington and James, 1991)
VOSP, position discrimination (Warrington and James, 1991)
VOSP, number location (Warrington and James, 1991)
VOSP, cube analysis (Warrington and James, 1991)
Ideomotor and constructional functions
Constructional praxis (Spinnler and Tognoni, 1987)
Ideomotor praxis (Mondini et al., 2001)
Memory
Digit span (Mondini et al., 2001)
Prose memory (Mondini et al., 2001):
Immediate recall
Delayed recall
Total
Corsi span (Spinnler and Tognoni, 1987)
Corsi supraspan (Spinnler and Tognoni, 1987)
Semantic memory questionnaire (Laiacona et al., 1993)

G.R.'s score
(correct)

912

K. Priftis et al.

Table 2 Preliminary assessment of G.R. on odd one out of
four drawings
Condition

Correct

Colour
Size
Structure
180°
Mirror stimuli

8/8
7/8
6/8
6/8
1/8

Results

G.R.'s copies were accurate and well organized, respecting
both the global shape and the internal feature organization of
the stimuli (Fig. 2). More importantly, the overall orientation
of G.R.'s copies was well aligned with the corresponding
models. There was no evidence of copy rotations. These
results further con®rmed G.R.'s ability to create and reproduce correct viewer-centred representations.

Fig. 2 G.R.'s copies of disoriented drawings.

Experiment 3: discrimination of mirror
stimuli
Introduction

The aim was to investigate G.R.'s dif®culties with mirror
images using the same procedure (odd-one-out task) and
stimuli as those originally employed by Turnbull and
colleagues (Turnbull et al., 1997a).

Methods
Stimuli, apparatus and procedure

A set of ®ve black and white drawings (Snodgrass and
Vanderwart, 1980) was selected (living things: camel, bear;
non-living things: piano, shoe, motorcycle). G.R. was
presented with 90 trials (45 critical and 45 control trials).
Each critical trial included three drawings of the same object,
horizontally aligned one next to the other (Fig. 3A). Two of
the stimuli had the same orientation (distracters), whereas the
third stimulus was their mirror image (the odd one out). The
mirror image resulted from the rotation of the object around
its vertical axis (y-axis). Such manipulation de®nes directional changes along the horizontal axis (x-axis). Mirror
stimuli appeared with the same frequency in every position
and each critical trial was presented three times, giving a total
of 45 trials (5 objects 3 3 presentations 3 3 positions).
Control trials were constructed using the same set of stimuli,
but the three drawings had the same orientation.
The stimuli of each trial appeared in the centre of an 18 inch
PC monitor in pseudorandom order. G.R. sat ~40 cm from the

Fig. 3 (A) Mirror condition. (B) 180° (upside-down) condition.
(C) 45° condition.

monitor. His task was to name the stimuli and detect the
presence or absence of the odd one out. There was no time limit
and G.R. was free to move his eyes. The same task was also
administered to three healthy control subjects matched for age
and education (mean age 75.3 years, 8 years of education).

Results

G.R. named all stimuli without hesitation. His performance
was perfect on control trials (45 out of 45), whereas it was
impaired when discrimination of mirror images was required

Downloaded from https://academic.oup.com/brain/article/126/4/908/331896 by guest on 11 April 2024

printed on individual ¯ash cards. Four drawings were
presented at 0° (canonical orientation), four at 90°, four at
180° and four at 270° clockwise. G.R. was asked to copy the
stimuli, respecting their orientation.

Agnosia for mirror stimuli
(29 out of 45). The comparison between control and critical
trials was signi®cant [c2(1) = 17.10, P < 0.001]. Performance
of the control subjects was fast and errorless (control trials, 45
out of 45; critical trials, 45 out of 45). G.R. did not produce
random responses, nor did he produce false alarms on control
trials, showing that his errors resulted from a failure in
discriminating mirror stimuli.

This experiment investigated G.R.'s ability to discriminate
inverted stimuli. Turnbull and colleagues described patients
with dif®culties in discriminating stimuli rotated 180°
clockwise (Turnbull et al., 1997a). In the present experiment,
stimuli were rotated 180° upside-down instead of 180°
clockwise. An upside-down rotation of 180° is a more
stringent test than a clockwise rotation of 180° (Turnbull and
McCarthy, 1996; Davidoff and Warrington, 2001), because
the latter requires 180° rotation around the horizontal axis
followed by 180° rotation around the vertical axis. Moreover,
the same spatial transformation can be achieved by a
clockwise rotation around the object's perpendicular axis
(z-axis). The speci®c effect of rotation around the x-axis was
tested in the present experiment, whereas the effect of
rotation around the z-axis was tested in Experiment 5.

Methods
Stimuli, apparatus and procedure

These were the same as in Experiment 3, except that the oddone-out stimulus was rotated 180° upside-down with respect
to its horizontal axis (Fig. 3B).

Results

G.R. named correctly every stimulus (90 out of 90), produced
only one error on critical trials (44 out of 45) and was
errorless on control trials (45 out of 45). The difference in
accuracy between critical trials in Experiments 1 and 2 was
signi®cant [c2 (1) = 14.21, P < 0.001]. Control subjects made
no errors.
Thus, G.R. had spared abilities in discriminating 180°
upside-down inversions (i.e. object rotation around the x-axis
leading to directional changes along the y-axis), which is at
odds with his impaired performance in discriminating mirror
stimuli (i.e. object rotation around the y-axis leading to
directional changes along the x-axis).

Experiment 5: 45° and 225° clockwise rotation
Introduction

The aim was to clarify whether G.R.'s unimpaired performance in Experiment 4 for 180° upside-down stimuli would

extend to stimuli rotated around the z-axis, perpendicular to
the drawing plane (i.e. picture plane clockwise rotations).

Methods
Stimuli, apparatus and procedure

These were the same as in Experiment 3, except that the oddone-out stimulus appeared in half of the trials rotated 45°
clockwise and 225° clockwise in the other half (Fig. 3C),
giving a total of 60 critical trials.

Results

G.R.'s performance was accurate on both critical (59 out of
60) and control (60 out of 60) trials. The comparison between
critical trials in Experiments 3 and 5 was signi®cant [c2(1) =
19.33, P < 0.001], whereas the comparison between
Experiments 4 and 5 was not. Thus, G.R. had intact
discrimination for stimuli rotated 45° and 255° clockwise
(orientation changes along the z-axis).

Discussion of Experiments 1±5

The presence of a speci®c disorder regarding mirror-stimulus
discrimination was con®rmed by comparing G.R.'s performance in a variety of tasks. G.R.'s de®cit is clearly at odds with
his intact performance in both spatial (orientation judgement,
copying disoriented objects, 180° upside-down rotation, and
45° and 225° clockwise rotations) and perceptual (size, shape
and colour discrimination, feature integration, and recognition) tasks. Our ®ndings replicate the previous studies by
Turnbull and McCarthy (1996) and Davidoff and Warrington
(2001). The aim of the following set of experiments was to
establish the speci®c contribution of different coordinate
systems of reference.

Experiment 6: egocentric versus allocentric
coordinate frames of reference
Introduction

One possible explanation of the de®cit for mirror stimuli is
the impairment in processing coordinate frames of reference
(Turnbull and McCarthy, 1996; Harris et al., 2001). G.R.
might have speci®c dif®culties in de®ning the direction of a
stimulus with reference to its horizontal elongation axis (e.g.
distinguishing left from right). This, in turn, might render the
discrimination of mirror stimuli hard or altogether impossible, given that they differ with reference to their horizontal
(x) axis. For instance, the only difference between the two
mirror camels was that they were looking in opposite
directions along the x-axis.
An observer can use two types of coordinates to de®ne the
horizontal x-axis of an object: egocentric coordinates and
allocentric coordinates (Fig. 4). Egocentric coordinates de®ne
spatial parameters centred on the observer's body parts (e.g.

Downloaded from https://academic.oup.com/brain/article/126/4/908/331896 by guest on 11 April 2024

Experiment 4: 180° upside-down
discrimination
Introduction

913

914

K. Priftis et al.

gaze, head, trunk, limbs) and/or between the observer's body
and objects in the surrounding. Allocentric coordinates de®ne
relations within and between objects around the observer. If
mirror-stimulus discrimination depends on egocentric coordinates, performance in tasks having as their reference
G.R.'s body might be impaired. In contrast, if mirror-stimulus
discrimination is a function of allocentric coordinates, G.R.
might have dif®culty with tasks requiring the processing of
coordinates centred on objects or people in his environment.
This was tested in the present experiment.

Method

G.R. and the experimenter sat facing each other. Two
conditions were created to test verbal and non-verbal output.
In the naming condition, G.R. was asked to name body parts
indicated by the experimenter (24 trials); in the pointing
condition, G.R. was asked to point to body parts named by the
experimenter (24 trials). In each condition, for half of the
trials the target was G.R.'s body [e.g. `point to your left arm'
(egocentric coordinates)] and for half of the trials the target
was the experimenter's body [e.g. `point to my left arm'
(allocentric coordinates)]. Trials were presented in random
order. For pointing, G.R. used his right, non-paretic arm.

Results

G.R. was errorless on trials based on his body (24 out of 24),
but at chance level on trials based on the experimenter's body
[12 out of 24, c2(1) = 16, P < 0.001]. G.R. made the same
number of errors in the naming and pointing conditions,
demonstrating that his performance was not affected by
misuse of verbal labels (right±left). As a follow-up test, a
version of Piaget's `three mountains' task (Piaget and
Inhelder, 1956) was administered to G.R. He viewed an

Experiment 7: discrimination of vertically
aligned mirror stimuli
Introduction

Experiment 7 was designed to contrast egocentric and
allocentric coordinates in the context of mirror-stimulus
discrimination. To disalign the two coordinate systems, the
triplets of stimuli employed in Experiments 3 and 4 were
displayed rotated 90° clockwise. In this way, both axes (x and
y) were disaligned with respect to G.R.'s axes (note that the
y-axis was still aligned in Experiment 6).
This experiment also allowed us to address a possible
objection to the previous experiments. Since most stimuli
were derived from objects with a clearly de®ned canonical
upright orientation, canonical object representations could
facilitate discrimination of the odd-one-out stimulus in 45°,
225° and 180° upside-down conditions. This was not the case
for the mirror-stimulus condition, in which the orientation of
the odd-one-out stimulus remained canonical despite rotation
around the vertical axis. If G.R. ignores information regarding stimulus orientation and bases his judgements upon
violations of canonical representations, his performance
should be impaired in both mirror and upside-down conditions as a result of the rotation of the whole triplet of
drawings.

Methods
Stimuli, apparatus and procedure

The same stimuli as those used in Experiments 3 and 4 were
presented to G.R., but each triplet was rotated 90° clockwise.
Thus, stimuli now appeared one below the other in a vertical
arrangement. The mirror (Fig. 5A) and upside-down (Fig. 5B)
conditions were presented in separate sessions. Trial
sequence, apparatus and procedure were the same as in

Downloaded from https://academic.oup.com/brain/article/126/4/908/331896 by guest on 11 April 2024

Fig. 4 (A) Egocentric coordinates. (B and C) Allocentric
coordinates.

arrangement of three familiar objects and a set of four
drawings depicting the scene from four different viewpoints.
Remaining in one position, he had to select the drawing that
would re¯ect how the scene would look from other perspectives in the room. He was impaired for viewpoints different
from his own (50% errors).
Thus, G.R. demonstrated intact assignment of the egocentric coordinate frame (G.R.'s body or his own viewpoint)
but he was severely impaired when the same trials were
centred on an allocentric coordinate frame (experimenter's
body or viewpoints other than his own). Although these
results might not re¯ect a causal link with the mirror
discrimination de®cit, one should note that both impaired
tasks involved discrimination of directional changes along the
horizontal x-axis. However, one patient showing impaired
mirror-stimulus discrimination (R.K., Davidoff and
Warrington, 1999), in association with object agnosia and
impaired space perception, had no dif®culty in indicating left
and right with reference to both his and the examiner's body.

Agnosia for mirror stimuli

915

Experiment 8: mirror-letter discrimination
Introduction

Fig. 5 (A) Mirror condition, obtained by 90° clockwise rotation of
the triplets of Experiment 3. (B) 180° (upside-down) condition,
obtained by 90° clockwise rotation of the triplets of Experiment 4.

Experiment 3. The task was also administered to the three
healthy control subjects.

Results

In the mirror condition, G.R. made only one error on control
trials (44 out of 45) but was impaired on critical trials [31
correct out of 45, c2(1) = 13.5, P < 0.001). In the upside-down
condition, G.R.'s performance was perfect on control trials (45
out of 45) and very accurate on critical trials (41 out of 45). The
difference between critical trials in the two conditions was
highly signi®cant [31 out of 45 versus 41 out of 45, c2(1) =
13.5, P < 0.001]. Control subjects made no errors.

Some stimuli, such as letters, words, and some ¯ags, have
highly canonical orientations and their mirror versions are
virtually never experienced. Thus, a mirror letter might
strongly violate its canonical representation (Caramazza and
Hillis, 1990), whereas a mirror piano, for example, does not.
In the case of mirror letters the violation of canonical
representations might assist G.R. in mirror discrimination
tasks.

Methods
Stimuli, apparatus and procedure

They were the same as in Experiment 3, with the exception
that ®ve upper-case letters (B, E, F, G, R) were used to
construct the triplets. For each triplet, two of three identical
letters had the same orientation, whereas the third was their
mirror image. Mirror letters appeared twice in every position,
giving a total of 30 critical trials. Thirty control trials were
constructed using the same letters.

Results
Discussion

G.R.'s performance was unaffected by manipulations of the
overall orientation of the stimulus array. G.R. was severely
impaired in discriminating mirror stimuli even though they
were presented vertically. Clearly, G.R. had severe dif®culties in representing changes along an object's intrinsic x-axis
when they resulted from the rotation around the object's
intrinsic y-axis. This would happen regardless of the
relationship between the object's coordinates and G.R.'s
egocentric coordinates. Indeed, to achieve orientation discrimination in odd-one-out tasks, G.R. had to code differences between stimuli in each triplet (allocentric coordinates)
without any reference to his own body (egocentric coordinates). G.R.'s de®cit could not be attributed to defective use
of verbal spatial terms (left±right), given that mirror stimuli
differed along the up±down and not along the left±right axis.

G.R. was accurate on both critical (30 out of 30) and control
(30 out of 30) trials. His markedly improved performance in
the mirror-letters condition may indicate the possible use
of cues deriving from the violation of strongly canonical
letter representations, bypassing his de®cit for mirrorstimulus discrimination. These results con®rm previous
studies involving discrimination of mirror words (Turnbull
and McCarthy, 1996) and mirror letters (Davidoff and
Warrington, 2001).

Experiment 9: mirror-stimulus discrimination
versus virtual grasping
Introduction

It has been proposed that there are distinct brain systems for
object recognition (occipito-inferotemporal, ventral pathway)

Downloaded from https://academic.oup.com/brain/article/126/4/908/331896 by guest on 11 April 2024

The alternative hypothesisÐthat G.R. was using a strategy
based on violation of canonical representationsÐwas also
addressed in this experiment. If G.R. were using canonical
orientation information, one should expect that the 90°
clockwise rotation of the triplets would render access to
canonical representations equally dif®cult for all conditions.
However, G.R. was still impaired with mirror stimuli and
unimpaired with 180° upside-down stimuli, demonstrating
that the use of canonical orientation information cannot
account for his de®cit. Nevertheless, such a strategy might be
used when stimuli are canonical with reference to all axes
(e.g. letters). This hypothesis was tested in the next experiment.

916

K. Priftis et al.

Methods
Stimuli, apparatus and procedure

Triplets of stimuli, printed on the centre of A4 pages, were
constructed using coloured drawings of a knife and a cup
(Fig. 6). Twenty-four critical and 24 control trials were
presented to G.R. The procedure was identical to Experiment
3 for the mirror-stimulus discrimination condition. In the
virtual grasping condition, the same trials were presented and
on each trial G.R. was asked to indicate which of the depicted
objects could be grasped easily with his right hand. Although
not explicitly stated, these stimuli had the handle pointing to
G.R.'s right side.

Results

G.R.'s performance in discriminating mirror stimuli was
perfect on control trials (24 out of 24) but remained
signi®cantly impaired on critical trials [15 out of 24, c2(1)
= 11.07, P < 0.002]. In contrast, G.R. was highly accurate in
selecting the correct object to be grasped using his right hand
(critical trials, 22 out of 24; control trials, 24 out of 24). The
difference between critical trials in mirror-stimulus and
grasping conditions was signi®cant [15 out of 24 versus 22
out of 24, c2(1) = 5.77, P < 0.02].

Fig. 6 Triplets of knifes and cups used in Experiment 9.

Discussion

G.R.'s performance in Experiment 9 was signi®cantly
affected by the nature of the task. Although he had dif®culty
discriminating between mirror stimuli, he was accurate on
trials requiring virtual grasping. Note that G.R. was asked to
indicate an object that could potentially be grasped without
actually grasping it. This dissociation was not attributable to
speci®c processing demands (e.g. implicit versus explicit)
because in both conditions G.R. had to make an explicit,
conscious judgement, guided by object orientation.
Dif®culties in mirror-stimulus discrimination probably re¯ect
different underlying brain mechanisms, selectively supporting distinct coordinate systems.
Mirror-stimulus discrimination is largely based on
allocentric coordinates, requiring correct detection of
each stimulus and of the direction of its salient features
with respect to elongation axes. Furthermore, betweenstimuli comparison, an allocentric task, is required to
detect the odd one out. However, grasping is based on
the adaptation of the hand (egocentric coordinates) to
single object structure, size and orientation, without any
between-objects comparison. In order to grasp a knife, for
example, it is suf®cient to detect the orientation of the
critical feature (the handle). If the blade is pointing to the
viewer's left hand, the handle will point to the viewer's
right hand (egocentric coordinates). Consequently, the
right hand, being spatially compatible with the handle,
will be more suitable for grasping the knife. Furthermore,
during initial testing, G.R. was perfect in determining if
drawings of hands presented at various angles (clockwise
rotation by 0°, 45°, 90°, 135°, 180°, 225°, 270° and 315°)
corresponded to his right or his left hand. During followup, G.R. said that his strategy was to detect the position
of the thumb (critical feature) with respect to the rest of
the hand and then to adapt it to his own hand. This
seems to be a clear demonstration of G.R.'s intact ability
to detect an object's critical features and to align his
hand with them.
In short, the dissociation between mirror-stimulus processing and virtual grasping might re¯ect the differential
involvement of allocentric versus egocentric coordinates,
only the former being impaired.

General discussion

In this study we describe a patient, G.R., with severe dif®culty
discriminating mirror stimuli. His performance in the standard neuropsychological assessment was highly accurate.
Indeed, G.R. had largely spared perceptual, attentional,
mnestic, linguistic, praxic and executive functions. His
normal performance in discriminating changes in colour,
shape and size and on the Street Completion Test ruled out the
presence of de®cits at the stages of early or intermediate
perceptual processing. G.R. could effortlessly judge the
orientation of depicted objects (Experiment 1), make copies

Downloaded from https://academic.oup.com/brain/article/126/4/908/331896 by guest on 11 April 2024

and for object localization/action (occipito-parietal, dorsal
pathway) (Ungerleider and Mishkin, 1982; Milner and
Goodale, 1995). The ventral system would be viewerindependent and would permit object recognition regardless
of brightness, size, distance, location and orientation of the
stimuli. The dorsal system, however, has a reaching
component that is egocentric and a grasping component that
adapts hand (egocentric) coordinates to stimulus-centred
coordinates. Patients with lesions of inferior occipito-temporal regions have marked object recognition dif®culties but
intact object localization, whereas patients with damage to
the occipitoparietal regions have severe dif®culty localizing
and reaching for objects (optic ataxia) but demonstrate
preserved object recognition (for review see Ellis and Young,
1996).
The aim of the last experiment was to investigate if G.R.
was in¯uenced by task demands, searching for possible
dissociations between tasks requiring mirror-image discrimination and tasks requiring object-grasping judgement.

Agnosia for mirror stimuli

Taylor, 1973; Warrington and James, 1986), which requires
coding of accurate viewer-dependent representations. This
was exactly the case for Patients R.J. and F.I.M., who were
impaired in object recognition from unusual viewpoints, but
had no dif®culty recognizing an object from normal viewpoints. However, G.R. was intact in recognizing objects with
hidden critical features (BORB subtest 7) or a foreshortened
elongation axis (BORB subtest 8), and in copying drawings
based on accurate viewer-dependent coding of models
(Experiment 2). Thus, although G.R. could disengage from
viewer-independent representations and create accurate
viewer-dependent representations, he was still impaired in
discriminating mirror stimuli. These results indicate that
failure in discriminating mirror stimuli cannot be attributed to
inability to disengage from viewer-independent representations.
The coordinate frame of reference hypothesis seems more
plausible. According to Turnbull and colleagues, mirror
discrimination dif®culties might depend upon failure in
assigning a suitable frame of reference to observed objects
(Turnbull et al., 1997a). This frame is not necessary for object
recognition but is necessary for processing orientation,
including mirror-stimulus discrimination. Such a frame may
be centred on egocentric coordinates, having as its reference
the viewer's body, or allocentric coordinates having as their
reference the intrinsic structure of the object. The role of
coordinate systems was tested in Experiments 6, 7 and 9.
In Experiment 6, G.R. was perfectly accurate in naming
and localizing parts of his own body (egocentric coordinates)
but his performance was severely impaired when the same
task was performed with reference to the experimenter's body
(allocentric coordinates). In Experiment 7 the role of
allocentric coordinates was further investigated by rotating
90° clockwise each triplet from Experiments 3 and 4. In
Experiment 3, for example, G.R. was impaired in coding
directional changes along the x-axis as a result of rotation
around the y-axis, whereas he was accurate in coding
directional changes along the y-axis after object rotation
around the x-axis (Experiment 4). If G.R.'s performance was
affected to some degree by a subtle de®cit of egocentric
coordinates (or of translation from egocentric to allocentric
coordinates), one should expect a reversal of the de®cit after
90° clockwise rotation of each triplet from Experiments 3 and
4. In other words, G.R. should be accurate in discriminating
mirror stimuli and impaired in discriminating 180° upsidedown stimuli. Contrary to this, G.R. was still impaired in
discriminating vertically arranged mirror stimuli with respect
to vertically arranged 180° upside-down stimuli. Thus, G.R.'s
impairment remained clearly centred on allocentric coordinates, intrinsic to the object's structure and independent of
changes with respect to his egocentric coordinates. Finally, in
Experiment 9, G.R. was impaired when task execution
depended on allocentric coordinates (discrimination),
whereas he was accurate when task execution depended on
egocentric coordinates (virtual grasping).

Downloaded from https://academic.oup.com/brain/article/126/4/908/331896 by guest on 11 April 2024

respecting the orientation of the model (Experiment 2), and
perform odd-one-out discriminations in various orientations
(Experiments 4 and 5). Thus, he showed neither orientation
agnosia nor agnosia for unusual views. His spared abilities on
a variety of spatial tasks not involving mirror stimuli clearly
demonstrate the speci®city of his de®cit in discriminating
mirror stimuli.
The effect of object rotation around the three main
Cartesian axes, y, x and z, was tested in Experiments 3, 4
and 5. G.R. had dif®culties only in discriminating mirror
stimuli. These, in turn, were de®ned as the result of rotation
around the object's y-axis, which determines reversal along
the x-axis (Experiment 3). In contrast, G.R. was highly
accurate in discriminating directional changes with reference
to the x- and z-axes (Experiments 4 and 5). One might argue
that the stimuli of Experiment 4 were also mirror stimuli.
Davidoff and Warrington (2001), for example, have de®ned
such stimuli as vertical mirror images. We believe this
de®nition to be misleading. Indeed, true mirror stimuli
overlap perfectly when rotated around one and only one
axis of reference, whereas rotations around more than one
axis were required in Experiments 4 and 5 for the odd-one-out
and distracters to overlap. Thus, the odd-one-out stimuli of
Experiment 3 were true mirror stimuli, whereas the odd-oneout stimuli of Experiments 4 and 5 were not.
Davidoff and Warrington (2001) hypothesized that mirrorstimulus discrimination might re¯ect a processing bias
produced by intact, occipito-inferotemporal mechanisms
subserving viewer-independent object recognition. The intact
ventral pathway might prevent the damaged dorsal pathway
from creating viewer-dependent representations that carry
object orientation information. However, this explanation
seems to be inadequate for a number of reasons.
First, if patients are engaged in orientation-insensitive
representations, their performance should be impaired in
every task requiring orientation judgement. Contrary to this,
Patients R.J., F.I.M. and G.R. were selectively impaired only
in a mirror-stimulus discrimination task, whereas their
performance in tasks involving discrimination of other
orientations (e.g. 45° and 180°) was largely spared. Of
course, it may be argued that mirror stimuli are more dif®cult
to discriminate. For instance, discriminating mirror letters is
an extremely dif®cult task for 3-year-old children (Rudel and
Teuber, 1963). Although this dif®culty disappears during
development, it might reappear in adulthood because of brain
damage. An explanation in terms of task dif®culty predicts
that the dissociation between impaired discrimination for
mirror stimuli and intact discrimination of other orientations
should be observed, whereas the inverse dissociation should
not. However, patient N.L. (Turnbull et al., 1997a) showed
impaired discrimination for stimuli rotated 180° clockwise
(29 out of 50) with errorless mirror-stimulus discrimination
(50 out of 50).
Secondly, patients with a de®cit of mirror-stimulus
discrimination should show dif®culties in perceiving objects
under non-optimal viewing conditions (Warrington and

917

918

K. Priftis et al.

Within the domain of allocentric coordinates, the de®cit
was evident only when directional coding of the x-axis of the
object was required. In contrast, G.R. was accurate in de®ning
directional changes along other axes. We claim that the locus
of the de®cit is more in coding the direction of the x-axis than
just in ®nding the x-axis (Harris et al., 2001). If ®nding the
axis were the only problem, G.R. should have performed
nearly at random, at least in Experiment 7. In contrast, his
errors were not in¯uenced by overall orientation modi®cation
of the triplets. These results, together with the double
dissociation (Turnbull and McCarthy, 1996; Turnbull et al.,
1997a) between discrimination of mirror stimuli and
discrimination of other orientations, suggest further dissociations within the domain of allocentric coordinates.
The importance of the parietal region in discriminating
mirror stimuli has been suggested in previous studies
(Turnbull and McCarthy, 1996; Davidoff and Warrington
2001). For example, patients R.J. and F.I.M. had bilateral
occipitoparietal lesions. The case of G.R. suggests a speci®c
contribution of the right hemisphere for mirror-stimulus
discrimination: his lesion was mainly in the right hemisphere,
with evident hypoperfusion in the right inferior parietal
region. Brain imaging studies with healthy participants
should allow ®rmer conclusions regarding inter- and
intrahemispheric localization.
In summary, our study shows the existence of a highly
speci®c disorder in discriminating mirror stimuli in the
absence of other visuospatial disorders, such as orientation
agnosia or unusual-views agnosia. It is suggested that mirrorstimulus confusion depends on a failure in processing the
directionality of an object's intrinsic x-axis, i.e. within an
allocentric frame of reference.

Davidoff J, Warrington EK. A particular dif®culty in discriminating
between mirror images. Neuropsychologia 2001; 39: 1022±36.

Acknowledgements

Rudel RG, Teuber HL. Discrimination of direction of line in
children. J Comp Physiol Psychol 1963; 56: 892±8.

References
Baylis GC, Driver J. Shape-coding in IT cells generalizes over
contrast and mirror reversal, but not ®gure-ground reversal. Nat
Neurosci 2001; 4: 937±42.
Caramazza A, Hillis AE. Spatial representation of words in the
brain implied by studies of a unilateral neglect patient. Nature 1990;
346: 267±9.

Ferber S, Karnath HO. Friedrich Best's case Z. with
misidenti®cation of object orientation. In: Code C, Wallesch
CW, Joanette Y, Lecours AR, editors. Classic cases in
neuropsychology, Vol. II. 2nd ed. Hove (UK): Taylor and
Francis. In press 2003.
Harris IM, Harris JA, Caine D. Object orientation agnosia: a failure
to ®nd the axis? J Cogn Neurosci 2001; 13: 800±12.
Karnath HO, Ferber S, Bulthoff HH. Neuronal representation of
object orientation. Neuropsychologia 2000; 38: 1235±41.
Laiacona M, Barbarotto R, Trivelli C, Capitani E. Dissociazioni
semantiche intercategoriali: descrizioni di una batteria
standardizzata e dati normativi. Arch Psicol Neurol Psichiatr
1993; 54: 209±48.
Milner AD, Goodale MA. The visual brain in action. Oxford:
Oxford University Press; 1995.
Mondini S, Vestri A, Mapelli D, Bisiacchi
Neuropsicologico Breve. Padova: LIRIPAC; 2001.

P.

Esame

Piaget J, Inhelder B. A child's conception of space. London:
Routledge & Kegan Paul; 1956.
Riddoch JM, Humphreys GW. Birmingham Object Recognition
Battery. Hove (UK): Lawrence Erlbaum; 1993.
Robertson L, Treisman A, Friedman-Hill S, Grabowecky, M. The
interaction of spatial and object pathways: evidence from Balint's
syndrome. J Cogn Neurosci 1997; 9: 295±317.
Rollenhagen JE, Olson CR. Mirror-image confusion in single
neurons of the macaque inferotemporal cortex. Science 2000; 287:
1506±8.

Snodgrass JG, Vanderwart M. A standardized set of 260 pictures:
norms for name agreement, image agreement, familiarity, and
visual complexity. J Exp Psychol Hum Learn Mem 1980; 6:
174±215.
Solms M, Kaplan-Solms K, Saling M, Miller P. Inverted vision after
frontal lobe disease. Cortex 1988; 24: 499±509.
Solms M, Turnbull OH, Kaplan-Solms K, Miller P. Rotated
drawing: the range of performance and anatomical correlates in a
series of 16 patients. Brain Cogn 1998; 38: 358±68.
Spinnler H, Tognoni G. Standardizzazione e taratura Italiana di test
neuropsicologici. Ital J Neurol Sci 1987; 6 Suppl 8: 1±120.

Cooper ACG, Humphreys GW. Task-speci®c effects of orientation
information: neuropsychological evidence. Neuropsychologia 2000;
38: 1607±15.

Turnbull OH. A double dissociation between knowledge of object
identity and object orientation. Neuropsychologia 1997; 35:
567±70.

Davidoff J, Warrington EK. The bare bones of object recognition:
implications from a case of object recognition impairment.
Neuropsychologia 1999; 37: 279±92.

Turnbull OH, McCarthy RA. Failure to discriminate between
mirror-image objects: a case of viewpoint-independent object
recognition? Neurocase 1996; 2: 63±72.

Downloaded from https://academic.oup.com/brain/article/126/4/908/331896 by guest on 11 April 2024

This research was supported by grants from UniversitaÁ di
Padova, CNR and MURST to C.U. We are grateful to G.R.
for his willingness to collaborate and to Dr Francesco
Piccione for neurological advice.

Ellis AW, Young AW. Human cognitive neuropsychology. Hove
(UK): Psychology Press; 1996.

Agnosia for mirror stimuli
Turnbull OH, Laws KR, McCarthy RA. Object recognition without
knowledge of object orientation. Cortex 1995; 31: 387±95.
Turnbull OH, Beschin N, Della Sala S. Agnosia for object
orientation: implications for theories of object recognition.
Neuropsychologia 1997a; 35: 153±63.

919

Warrington EK, James M. Visual object recognition in patients with
right-hemisphere lesions: axes or features? Perception 1986; 15:
355±66.
Warrington EK, James M. Visual object and space perception test.
Bury St. Edmunds (UK): Thames Valley Test Company; 1991.
Warrington EK, Taylor AM. The contribution of the right parietal
lobe to object recognition. Cortex 1973; 9: 152±64.

Ungerleider LG, Mishkin M. Two cortical visual systems. In: Ingle
DJ, Goodale MA, Mans®eld RJW, editors. Analysis of visual
behavior. Cambridge (MA): MIT Press; 1982. p. 549±86.

Wilson B, Cockburn J, Halligan P. Behavioural Inattention Test.
Bury St. Edmunds (UK): Thames Valley Test Company, 1987.

Warrington EK, Davidoff J. Failure at object identi®cation
improves mirror image matching. Neuropsychologia 2000; 38:
1229±34.

Received May 27, 2002. Revised October 11, 2002.
Second revision November 13, 2002. Accepted November 14, 2002

Downloaded from https://academic.oup.com/brain/article/126/4/908/331896 by guest on 11 April 2024

Turnbull OH, Carey DP, McCarthy RA. The neuropsychology of
object constancy. J Int Neuropsychol Soc 1997b; 3: 288±98.