Neurol Clin N Am

21 (2003) 501520

Visual agnosia
M. Jane Riddoch, PhD*, Glyn W. Humphreys, PhD
Brain and Behavioural Sciences Centre, School of Psychology,
University of Birmingham, Birmingham B15 2TT, UK

The father of agnosia: Lissauer

Lissauer was the rst scientist to provide a detailed account of a patient
with visual agnosia [1]. He presented his clinical observations at a meeting of
neurologists in 1888 in the context of a theoretic framework. He suggested
that, from early stages of visual processing, the processing of color, form,
and motion might be aected separately as a result of brain damage (as
indeed subsequent work has proved; see later discussion of Decits in
peripheral visual processing). At the level of form perception, he proposed
that visual recognition required processing through two distinct stages: the
rst (apperception) he described as the stage of conscious awareness of a
sensory impression; the second (or associative stage) was believed to result
from the simultaneous activation of many concepts related to the object
(ie, the activation of associated memories). Taking as an example a violin,
he stressed the wealth of sensory associations (image, sound, name, tactile experience from handling it, typical visual context, and so forth). The
specic nature of memories and associations for any object varies across
individuals; nonetheless, most people have a rich concept based on
such distributed associations. An associative agnosia results if perceptual
information for a violin, despite being derived normally, fails to activate the
stored associations. Lissauer was skeptical, however, that a purely associative agnosia could exist without some concomitant impairment in apperception. He also suggested that apperceptive decits could exist at dierent
levels of processing, although in all such cases, the decit should not be
attributable to impaired peripheral visual processing (which may result in

This work was supported by grants from the MRC, Wellcome Foundation, and the Stroke
Association.
* Corresponding author.
E-mail address: m.j.riddoch@bham.ac.uk (M.J. Riddoch).
0733-8619/03/$ - see front matter 2003, Elsevier Inc. All rights reserved.
PII: S 0 7 3 3 - 8 6 1 9 ( 0 2 ) 0 0 0 9 5 - 6

502

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

decreased acuity, reduction, or both in visual elds). Lissauers account of

agnosia continues to be used as a general framework, and we review evidence on agnosia in the light of Lissauers work. In particular, we evaluate
each of his proposals, dealing rst with whether or not agnosic decits are
attributable to low level decits in basic perception and whether or not disturbances to dierent properties of images can be established (eg, impaired
color or motion vision without impairments in form vision; see following
discussion). Subsequently, we discuss whether or not contrasting forms of
apperceptive agnosia exist (see later discussion of Apperceptive disorders)
and whether or not associative agnosia can occur without concomitant
problems in form processing (with apperceptive agnosia; see later discussion
of Associative disorders). These arguments are reviewed in light of cases
of visual agnosia reported since Lissauers inuential paper in 1888.

Decits in peripheral visual processing

Color and motion perception
Distinct areas of the brain are specic for the processing of color, form,
and motion, as indicated by independent neurologic decits. Achromatopsia
refers to the syndrome in which a patient loses the ability to see colors after
cortical damage. The loss may be partial or complete and it may or may not
be accompanied by other visual defects. Typically, the patient reports seeing
the world in black and white and shades of gray [2] but has no diculty in
the recognition of objects (provided recognition does not depend on color).
Relatively few cases have been reported with selective decits of motion perception. The most detailed study has been that of patient LM, who suered
a superior sagittal sinus thrombosis at age 43 [3]. An MRI showed severe
bilateral damage to the middle temporal gyrus and the adjacent part of the
occipital gyri with subcortical damage aecting lateral, occipital, and occipitoparietal white matter [4]. LM was profoundly disabled in everyday life and
found moving stimuli to be highly disturbing and unpleasant. She could not
perceive actual movement, so that when objects did move (eg, cars or people), they seemed to jump from one position to the next. She found that
understanding language was problematic (peoples lips seemed to hop up
and down), as was meal preparation (she had diculty in pouring and measuring liquids because they appeared frozen like a glacier) [5]. LM had no visual eld defect, her acuity was unimpaired, and she performed normally on
tests of visual identication and recognition of objects, faces, and places.
Form perception: a problem of low-level vision?
Efrons patient, Mr. S., was able to detect dierences in luminance, wavelength, and area and could respond to small movements of objects before him
[6]. Critically, however, he was unable to distinguish between two objects of

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

503

the same luminance, wavelength, and area when the only dierence between
them was shape. Over the years, there has been some controversy as to
whether or not decits in form perception, such as that shown by Mr. S., may
be classied as decits of low-level visual processing or whether or not they
form part of the agnosia continuum. In the 1950s, Bay argued that the sensitivity in visual testing was lacking in such patients. He used more detailed
examination techniques than were generally applied clinically, including tests
for local adaption and for sensation time, Bay dened sensation time as the
minimal time required for the appearing of a sensation (p. 516). Stimuli
were presented tachistoscopically in dierent parts of the visual eld. The
time taken to detect the stimuli was raised in areas of decreased function.
Adaption occurred if a constant stimulus was applied to a xed area of the visual eld. Over time, the perception of this stimulus slowly faded. The fading
time was dependent on the objects size and its positioning in the visual eld.
Fading occurred more rapidly in impaired areas of the visual eld and demonstrated disorders of visual functioning that had not been apparent on
standard perimetric testing in a variety of brain-damaged patients, including
a patient with visual agnosia [7]. Bay found that the retina in patients with
apparently normal visual elds showed abnormal fatigue. The abnormality
was greatest in the central retinal region but aected the periphery more (visual stimuli would tend to drop out of awareness). Bay argued that the visual
eld essentially was contracted and therefore would not allow the simultaneous perception of a whole visual stimulus in circumstances in which the visual
angle subtended by the stimulus was greater than the remaining functional
central area. As a result, it would not be possible to gain a holistic view of
the stimulus, but details of it may be seen in succession. These eects must
be contrasted with whose occurring in cases of simultanagnosia. Here also
identication is limited to one object at a time; however, now what denes
the object is critical. Luria [69] described a patient who was able to report
the star of David when shown the stimulus in a single color; however, the
patient only reported the presence of a triangle when the two triangles making up the star of David were shown in dierent colors. Identication here is
not determined by the spatial extent of the stimulus but by how the parts
group together (when in the same color) or segment apart when in dierent
colors. Bays [7] attribution of impaired visual recognition to disturbed
sensory processes resulted in more stringent assessments of many of the
subsequently reported cases of visual agnosia. For example, a detailed
investigation of visual abilities of a patient with visual agnosia, HJA [8],
showed no tunnel vision [9]; however, he was impaired in detecting local elements embedded in more global geometric forms. Thus, whereas he performed at normal levels in detecting a line target embedded in distractor
lines of a dierent orientation, he was unable to perform a similar task when
the target was a square made of four parallel lines (Fig. 1). Similarly, HJA
was abnormally slow in responding to local elements in compound global
letters [10], although he could respond to the global form itself (Fig. 2).

504

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

Fig. 1. Examples of the displays used by Alkhateeb et al [9]. In (A) the elements consist of single
lines; in (B) they consist of groups of lines. (Adapted from Alkhateeb W, Bromley JM,
Humphreys GW, Javadnia A, Riddoch MJ, Ruddock, H. Abnormal responses to multi-element
spatial stimuli in a subject with visual form agnosia. Clinical vision Science 1992; 7;16373; with
permission.)

Alkhateeb et al [9] argue that this pattern of performance in more complex search tasks mirrors HJAs ability to identify correctly local features of
visual scenes, while being unable to link them into a coherent percept (see
later discussion of Perceptual integration of form information). Others
also explored whether or not impaired low level vision (such as the thresholds for detecting and discriminating gratings, color, correlated movement,
reectance, shape from motion or luminance contrast, relative position, texture, and orientation) in separate studies in groups of patients with localized

Fig. 2. A Navon-type stimulus consisting of a global shape (the letter H) made up of smaller
letters (S).

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

505

brain damage can account for recognition disorders. The accumulated data
show that whereas there may be elevated thresholds, there is no correlation
between the pattern or the severity of impairments and the presence or
absence of visual agnosia (for a discussion of these issues, see Cowey [11]).
In the 1980s Campion and coworkers argued that the visual recognition
problem in patients with agnosia resulting from a particular etiology, carbon monoxide poisoning, is the result of multiple small scotoma (resulting
from the multiple lesions caused by the poisoning). They proposed that the
behavioral consequences of such lesions would be that patient viewed the
world through a peppery mask [12,13]. Given that we adapt over natural
holes because of the blind spot, however, and that any masking eect may
be reduced by movement, then the import of this argument is not clear. An
alternative account is that the problem results from impaired grouping processes [14,15] (see later discussion). Recently, Vecera and Gilds [16] attempted
to distinguish between these possibilities in a series of experiments with normal subjects. The basic paradigm required subjects to respond to a spatially
cued target. There were two conditions to the experiment: either subjects
viewed the displays (two rectangular bars) through a peppery mask (masking condition), or the displays were degraded in one of two ways (grouping
condition). In the grouping condition, either the midsections of the rectangles were removed (cues such as co-linearity and closure remained present,
allowing grouping of the parts into an object) or the corners were removed
(making it more dicult to group the parts into objects). The displays were
aligned either horizontally one above and one below the xation cross, or
they were aligned vertically, again one on each side of the xation cross.
A spatial cue appeared at one end of one of the bars. This was either valid
(75% of trials) or an invalid (25% of trials). After the cue, the two bars were
shown empty for 200 ms. On valid trials, the target then appeared in the
cued location. On invalid trials there were three possible locations for the
target: (1) in the same bar as the cue but at the opposite end; (2) in the other
bar in the same relative spatial location as the cue; or (3) in the other bar and
at the opposite end to the relative spatial location of the cue (for examples of
the stimuli, see Fig. 3). In the masking condition, RTs were slowed but the
pattern of results was similar to that obtained from unmasked displays (ie,
there was a valid cuing eect with responses to targets in the cued rectangle
faster than those to target in the in uncued rectangle). In the grouping conditions, there was no eect of degrading when only the corners of the rectangles were present, suggesting that grouping processes nullify this form of
degradation. There was an eect of degrading, however, when only the midsections of the rectangles were present (ie, in the circumstances when grouping could not occur). Vecera and Gilds [16] argue that their results support
the grouping-decit account of visual agnosia (see later discussion of Feature integration).
Warrington [17] also argued the poor recognition in cases such as those of
Efron [6] are conceptualized best as a partial or incomplete visual eld

506

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

Fig. 3. Stimuli similar to those used by Vecera and Gilds [16]: (A) illustrates the sequence of
events: a cue rst appears at the end of one of the rectangles, there is an interstimulus interval,
and the target then appears in one of three possible locations; (B) illustrates the masked
displays; and (C ) shows the stimuli with corners or mid sections only. (From Vecera SP, Gilds
KS. What processing is impaited in apperceptive agnosia? Evidence from normal subjects.
Journal of Cognitive Neuroscience 1998;19: 56880; with permission.)

defect. She describes them as pseudo-agnosic syndromes and argues

that adequate visual functioning could not be established. More recently,
Warrington and Rudge argued that adequate visual functioning reects the
ability to detect dierences between nonrepresentational visual patterns; in

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

507

particular, they propose that the term apperceptive agnosia be reserved for
disorders reecting damage to those sub-components of object recognition
that implicate the ability to construct, by recourse to stored representations,
an object specic structural description [18]. Tests used to assess whether or
not stored representations can be accessed include identifying overlapping,
familiar objects [19,20], incomplete outline drawings of objects [21], or
unconventional view object photographs [22]. In all these instances, patients
with posterior right hemisphere damage seem to be selectively impaired. One
diculty with this account is that, although right hemisphere patients show
impaired performance on a variety of perceptual tasks, they typically do not
have any diculty recognizing undegraded objects in conventional views;
such a diculty would seem to be central to the concept of visual agnosia.
Whilst acknowledging these arguments, the weight of opinion is that deficits of form processing can occur and that these are independent of early
visual processes that may be reected in poor visual acuity, reduced visual
elds or impairments in color, or movement or depth perception. In addition, some studies report measures of contrast sensitivity, visual evoked
responses, eye movements, and detection of targets dened by luminance,
which can be preserved even though problems in form perception are demonstrated [23].
Apperceptive disorders
Lissauer suggested two ways to assess the integrity of form perception: (1)
shape discrimination and (2) shape copying. He reported the results of these
tests with his patient, GL. With shape discrimination, GL initially was asked
to detect obvious dierences between simple shapes and then minimal dierences between complex shapes. He performed well with simple tests (eg, he
was asked to discriminate between shapes similar to the letter 3; one of the
shapes had an additional curl at the top), but performance slowed with more
complex tests (eg, he was asked to distinguish between two boxes, each of
which contained 12 equally distributed crosses; in one box, the center was
empty, whereas the other was lled with a cross). On the basis of these
observations, Lissauer argued that perceptual abilities might lie on a continuum, with the implication that a simple shape discrimination test is not sufcient to determine whether or not a perceptual decit is present. GLs
diering performance on the discrimination tests suggests that one may distinguish his processing of form (which the authors imply is intact based on
his performance in the easy discrimination test) from his ability to integrate
local and global aspects of form (which may be impaired, given his performance on the dicult discrimination test).
Lissauers patient was able to copy; however, again it is clear that Lissauer
did not regard copying ability as denitive proof of intact perception. He
indicates that the procedure used by the patient was very slow with many
pauses. He also noted that GL found it easier to copy line drawings than real

508

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

objects, because it was dicult to perceive the outline contour when shadows
were present. Recently, the dierences between drawing real objects and
copying line drawings have been explored empirically in a patient with visual
agnosia [24]. The patient, HJA, was asked to copy line drawings or to draw
objects, but in conditions of good lighting, without shadow. There were several conditions: either HJAs head was xed or he was allowed free movement. Also, objects were positioned within or outside reaching range in
order to vary the potential contribution of stereoscopic depth (stereo cues
being stronger within reachable space [25]) (see Fig. 4 for examples of HJAs
performance). This enabled judgment of the role of such cues in HJAs object
identication. All the drawings subsequently were presented to a group of
control subjects who were asked to name the depicted objects. The results
showed no reliable dierence in the identication of line drawings versus real
objects when the objects were close, even when head movements were prevented. When HJAs head was xed, however, and objects were viewed from
a distance, few of his drawings could be identied. In this condition, he typically failed to depict all the parts of the object successfully. The data indicate
that stereo depth cues may play an important part in such patients perceiving
objects, with perception abnormally impaired when stereo depth cues are
reduced, by preventing motion and stereo disparities (with a xed head position and more distant stimuli). In line with Lissauers original observations,
these data also suggest that the ability to copy items varies as a function of
test conditions and that apparently good copying cannot necessarily be taken
as evidence of intact perception. Given this proviso, we review the cases of
several patients in the literature, taking their ability to discriminate shapes
and to copy drawings as a starting point. We show that some patients are
unable to perform even these simple tasks (presumably because of impaired
form perception) and that, whereas others succeed, performance is far from
normal. In these instances, we propose that the decit is described best as a
perceptual decit specically aecting processes of visual integration.
Impaired form perception
Many patients have been described with a profound decit in form perception who are unable to either discriminate between simple shapes or to
copy drawings successfully [6,12,2628]. Interestingly, all these cases suered
carbon monoxide poisoning. Color, brightness, and movement discrimination typically were preserved, whereas the ability to perceive gural properties, such as size, orientation, and shape, was lost [6,28]. For instance,
Efrons patient, Mr. S., was unable to discriminate between a square and
a rectangle that were matched for total ux. Task diculty was manipulated
by varying the ratio of length to the width of the oblong. Mr. S. was barely
above chance, except on the easiest discriminations. In a more extensive
report of patient DF, Milner and colleagues [28] indicate that she performed
at chance on a similar test to that described by Efron [6]; she also was unable

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

509

Fig. 4. Examples of HJAs copies of (A) copies of line drawings; (B) real objects presented at a
distance with the head not xed; (C ) copies of real objects (distant condition, head xed); (From
Chainey H, Humphreys GW. The real-object advantage in agnosia: evidence for a role of
surface and depth information in object recognition. CognitiveNeuropsychology 2001;18:175
91; with permission.)

to segment gure from ground (ie, detecting an O, X, or shape against a

ground consisting of a noise mask); she was at chance at determining
whether or not abstract geometric shapes were symmetric or not, and so
forth. Although she was unable to discriminate between simple shapes, she
was able to use shape information in order to orient correctly her hand to
match the orientation of a slot and to appropriately scale her grasp to
objects of dierent sizes [28,29]. These data suggest dissociation between the
visual information used for recognition and perceptual judgments and the
visual information used for action.
A recent case who does not seem to have as extreme an impairment in
form perception as that shown in Mr. S. [6] and DF [28] is SMK. He suered

510

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

anoxia, which resulted in impairment in visual recognition that, although

relatively severe (he was reported as able to name only one of the rst six
of the Oldeld pictures and three of twelve silhouette drawings of common
objects), was not total (unlike DF, he could identify capital letters; also, DF
failed to recognize any line drawings) [30]. SMK also seemed to have
adequate elementary visual function: acuity was normal, visual elds were
full, he had normal stereopsis, and his ability to discriminate between surface appearances of lightness, color, and shading was normal [30]. DF, in
contrast, had severe visual decits: her visual elds could not reliably be
measured; brightness judgments were inaccurate; there was impaired detection of low spatial frequencies, although high spatial frequencies could be
detected normally; depth perception and movement perception were also
impaired. Although performance was poor, SMK was not at chance either
on the Efron test (assessing the ability to discriminate between squares and
rectangles), scoring 60% and 70% correctly for dicult and easy discriminations, respectively, or on a similar test using squares made from interrupted
lines, scoring 58% and 80% correctly for dicult and easy discriminations,
respectively. Although SMK found it dicult to make judgments on the
basis of shape, he performed well on tests of gureground segmentation
(for instance, detecting an X in the presence of background noise, detecting
the number of items when geometric shapes are arranged concentrically or
when they overlap, and so forth). From this, Davido and Warrington [30]
argued that the process of gureground perception is separate from processes involved in form perception, with only the latter impaired in SMK. It
is clearly the case that simple gureground discrimination can be dissociable from form perception, because gures can be distinguished from
ground on the basis of image properties other than form, including color
[31,32], depth [3337], and motion [3]. In these cases gure-ground segmentation is based on shape perception derived from cues other than luminancebased contours. These other cues (color, motion, depth) may result from
processing within separate modules to those involved in the processes in
deriving shape from luminance contours. A more critical issue for understanding the relations between form perception and gureground coding
is whether or not the two dissociate, even when form perception may be
important for gureground coding. Here the results are less clear. Peterson
and colleagues observed that the identication of occluded forms biases
gureground organization. They propose that objects are recognized rst on
the basis of image features and occluded contours are computed afterwards
(ie, a top-down model of gureground organization) [38]. Giersch et al [39],
alternatively, show that impaired gureground perception can be related to
processes involved in form coding. Their patient, HJA, was poor at discriminating occluded and occluding gures, but primarily when parts of the
occluded gure were close and so grouped, despite the presence of the
occluder. Under these circumstances, HJA sometimes ascribed the occluded
edge to the front shape, consistent with a problem in gureground assign-

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

511

ment. This problem in gureground assignment reects a more general

problem of linking parts to a whole in this patient, with the parts-to-whole
problem exacerbated under conditions of gureground ambiguity (Fig. 5).
Feature integration
Lissauers patient, GL, did not perform well when form perception was
tested stringently (eg, discriminating between complex shapes with minimal
perceptual dierences; see previous discussion); he also performed poorly on
a test of gureground segmentation. Lissauer placed two books in front
of him, which had identical ornamental borders on their covers. Lissauer
marked one aspect of the ornamental detail on one of the books and asked
GL to nd it on the other. This he was unable to do, presumably because
the assignment of the ornamental gure against the ground of the book was
dicult.
Many patients have been reported with intact shape perception (as
assessed using the Efron shape-matching test) [2,6,8,40,41] but who are

Fig. 5. Examples of the stimuli used by Giersch et al [39]: (A) non-overlapping, (B) overlapping
non-occluding, (C ) overlapping, occluding, and (D) overlapping silhouettes. (From Giersch A,
Humphreys GW, Boucart M, Koviaks I. The computation of occluded contours in visual
agnosia: evidence of early computation prior to shape binding and gure-ground coding.
Cognitive Neuropsychology 2000;17:73159; with permission.)

512

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

impaired when more stringent tests are used. One of these, HJA [8], was previously discussed, and his case is discussed in more detail here because it
reects a particular problem in coding complex shapes. Like Lissauers
patient, GL, HJA was able to produce accurate drawings of objects he could
not recognize, although, like GL, he was very slow. He also performed the
Efron [6] shape test well and could make orientation and size-matching
judgments at a normal level [42]. He showed at search functions (ie, no
eect of the number of distractors) on visual search tasks in which the target
diered from the distractors in terms of a single feature (eg, line orientation)
[14,43]. These data suggest that HJA is able to process simple form information in a spatially parallel manner. On other tests, however, he did not do as
well. For instance, he was at chance on an object decision task in which nonobjects were constructed using parts of real objects, although in an alternative version of the task in which silhouettes rather than line drawings were
used, he performed within normal limits. The internal detail in the line drawings seemed to disrupt rather than enhance his performance (nonbrain
damaged subjects nd line drawings easier to discriminate than silhouettes
[44]). Furthermore, as indicated in the previous section on Impaired form
perception, whereas HJA was able to identify the individual items in overlapping gures, he was slow relative to control performance [8]. In addition,
we also have shown that HJAs visual recognition is compromised if time
restraints are imposed. With unlimited exposure, he was able to name
80% of a set of line drawings, selected on the basis of previous successful
performance; however, at 100 ms, he was only able to name 15% of them.
These ndings suggest that HJAs perceptual abilities were impaired (1)
under speeded conditions and (2) when there were multiple segmentation
cues. Riddoch and Humphreys [10] characterized HJAs decit as a failure
to integrate local visual elements into perceptual wholes. This is not to say
that HJA is insensitive to global shape information. As noted, using Navontype stimuli in which large global letters are made up of smaller local
letters [45] (see Fig. 2), HJA responds faster to the global than to the local
letter (as with control subjects). Unlike normal subjects, however, the identity of the local letter had no eect on his responses to the global letter. His
responses to the local letter when presented in isolation were also within the
normal range, but he was slow when he had to identify the local letter in the
context of the global letter. This suggests that there is an abnormal eect of
global context that hinders his accurate perception of local detail. Consistent with this, HJA was impaired at grouping homogeneous distractors in
visual search tasks in order to detect targets eciently [15]. Normally, a
homogeneous context facilitates search by helping distractors form a separate group from the target. The detrimental eect of the distractor context,
for HJA, is indicative of interference from more global groups on the coding
of a target part.
To account for this pattern of performance, Riddoch and Humphreys
proposed that HJA had impaired perceptual integration of shape elements

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

513

into coherent wholes. It is not that HJA is insensitive to perceptual wholes,

but more that holistic contexts disrupt his coding and integration of parts.
This problem of part integration is particularly telling when there are local
edges within the wholes (eg, internal lines in line drawings), which prompt
parts to be segmented (in line drawings rather than silhouettes) (Fig. 6).
We have provided substantial detail about the case of HJA on the
grounds that the studies of his case provide perhaps the most thorough
investigation to date of perceptual organization in agnosia. Other patients
have been reported showing some similarities to HJA in the tasks they can
and cannot do. One example is Annalisa, a 21-year-old girl who suered a
trac accident resulting in damage to the right temporo-occipital region.
After this, Annalisa had a severe impairment in object recognition [41]. Like
HJA, she performed the Efron shape perception test well; she also performed within the control range in a (possibly) more stringent test of shape
perception devised by Talland [46] (here the stimuli consist of vertically oriented oblongs that end at their upper side with an arch of graded sharpness;
all possible combinations of pairings of the stimuli are presented and the
patient must decide which one of a pair has the sharper point). She scored
at ceiling on a samedierent match task with complex gures that were
distinguishable only by minor dierences in their external conguration

Fig. 6. Examples of overlapping letters as used by Riddoch and Humphreys [8].

514

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

(eg, hand postures) or internal structure (eg, gures from Rays Progressive Matrices). Against her good performance on tests of form perception,
Annalisa was impaired on tests of gureground segmentation: these included detecting a fragmented letter superimposed on a noisy background
[47] and a test based on that described by Ghent [48], in which displays of
three to ve overlapping drawings of either meaningful or meaningless items
(in separate conditions) were presented with 10 single drawings of items
from the appropriate condition. The task was to choose which of the single
items were present in the target array.
Similar eects have been reported recently in two further cases (SZ and
AP) [49]. SZ had bilateral lesions implicating the posterior watershed distribution with sparing of the primary visual cortex, and AP had decits aecting bilateral middle and inferior temporo-occipital cortices again sparing
the primary visual cortices. Both patients performed reasonably well when
asked to match a target item (a geometric shape) to one of three items
(80% and 100% correct for SZ and AP, respectively). Performance dropped,
however, when a similar task was performed with items consisting of partially overlapping geometric shapes (31% and 70% correct for SZ and AP,
respectively). Likewise, SB is reported as being able to match identical
shapes, but was at chance when asked to segment overlapping gures [50]
(also see Kartsounis and Warrington [41]). Butter and Trobe [51] reported
another patient, SM, who, like HJA, was not only impaired at overlapping
forms but was also worse with line drawings than with silhouettes. Indeed,
when presented with drawings of single objects, SM sometimes identied
them as two separate items, indicating segmentation of the parts and poor
coding of the whole. We suggest that normally an integrated representation
of the parts is linked to more holistic descriptions based on low spatial frequency components, so that we perceive wholes that are articulated and
identiable. This linkage process is impaired in integrative agnosia. Integrative agnosia represents a form of apperceptive agnosia in which the coding
of single shapes is relatively unimpaired, with the decits revealed under
conditions that stress visual segmentation and grouping.
Associative disorders
In contrast to the above cases, in which forms of perceptual decit are
apparent, many patients have been reported with impaired recognition of
visually presented objects but who perform well on tests of perception
[52,53]; thus, it seems that Lissauer was not correct in his assumption that
some form of perceptual decit underlies the recognition problem in all
patients with visual agnosia. We suggest that there exist relatively pure
forms of associative agnosia, without a concomitant apperceptive decit.
Riddoch and Humphreys [53,54] distinguished two forms of stored knowledge serving object recognition: stored structural descriptions and stored
semantic knowledge. Structural descriptions are stored representations of

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

515

the structural properties of objects, used to match on-line descriptions

encoded by the visual system. Semantic knowledge species information
about object function and interobject associations that forms the basis of
our full concept of an object. For objects to be recognized, with functional
and associative knowledge about them retrieved, stored structural and
stored semantic knowledge needs to be accessed. Accordingly, the association stage can be separated into two substages, each concerned with accessing a particular type of stored knowledge. Object decision tests have been
used to assess the ability to access stored knowledge of object shape. With
perceptually good nonobjects constructed from the parts of real objects,
object decisions are contingent on access to stored structural descriptions
and cannot rely on judgments from on-line perceptual representations alone.
In 1987, we [53] demonstrated that patients could perform dicult object
decisions (with nonobjects drawn from parts of two dierent objects from
the same category) when their ability to access semantic (associative and
functional) knowledge was impaired. The single case who showed this, JB,
was relatively poor at making judgments about which two of three objects
may be associated or used together, when they all came from the same category (eg, a hammer, a nail, and a screw). This was not because of a failure
to understand the test, because he performed perfectly when presented with
words rather than objects. The problem was in gaining access to this knowledge from vision. Given JBs good object decision performance, we proposed that there was a decit in visual access to stored semantic knowledge
but crucially this took place after intact access to stored structural knowledge
about objects. That is, there was an associative decit after intact perception
allowed contact to perceptual knowledge for particular stimuli. Since this
original case, this pattern of decit has been reported on several occasions
[52,55,56]. The pattern conrms the distinction between structural and
semantic knowledge and between apperceptive and associative agnosia. In
some of the other cases, the problem is not one of access to semantic knowledge per se, but more a disorder of semantic knowledge itself. For example,
the patient of Sheridan and Humphreys [55] was impaired at asking questions
about the functional properties of objects from their names and when the
objects were presented visually, so the problem generalized across modalities.
This may be expected from a disorder of semantic knowledge, rather than a
disorder of access to semantic knowledge from a specic modality (eg, spoken
words vs pictures).
One other interesting aspect of JBs decit is that his problem in object
identication was worse for some categories of objects than others. For
instance, he was worse at naming living things than at naming nonliving
things (tools, household items, and so forth). This apparent categoryspecic decit has been reported in substantial numbers of patients and was
rst investigated experimentally by Warrington and Shallice (see references
5760 for recent reviews). Warrington and Shallice [57] proposed that a deficit for living things could be linked with impairment to stored knowledge

516

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

about the perceptual properties of objects, because they assumed that perceptual knowledge was key to identifying living things. In contrast to this,
nonliving things may be distinguished more on the basis of their functional
properties (what they do or how they are used) and, thus, nonliving things
would not be as dependent on access to detailed perceptual knowledge.
Thus, damage to stored knowledge about the functional properties of
objects would disturb nonliving things more than living things (for evidence,
see Warrington and McCarthy [61,62]). Other investigators argued that such
category specic decits can reect the categoric nature of our semantic representations, which may be organized modularly to separate animals from
plant life and other things. At present, there is a good deal of study targeted at distinguishing these and other relevant accounts [60]. For our purposes, the important lesson is that disorders of semantic knowledge can be
described as associative agnosias, and they can arise in patients with no
apparent perceptual problems. A case such as JBs is particularly interesting,
however, because his problem is not in semantic knowledge itself, but in
accessing this knowledge from vision. How could a decit in visual access
also generate an apparent category-specic decit? Humphreys, Riddoch,
and Quinlan [63] demonstrated that category eects are found not only in
patients but also in normal participants (albeit on measures of reaction time
rather than error). They argued that the eects occurred because many living
things have a similar perceptual structure, and this created increased competition for recognition and naming. In a patient such as JB, competition at a
semantic level may be exacerbated if the lesion adds noise to activation
that is transmitted from the structural description system. Living things are
more dicult to identify as a consequence (for a simulation, see Humphreys
et al [64]).
The notion that one cause of category-specic decits for living things
arises in visual access to semantic knowledge also is supported by patients
with relatively milder decits than JB, who present with a naming rather
than a recognition impairment for living things. SRB and DM suered damage to the left medial and inferior occipito-temporal regions and they demonstrated a particular impairment in naming animate items [65,66]. Naming
inanimate items was relatively preserved (71% versus 95% correct for SRB,
and 46% versus 75% correct for DM for animate versus inanimate line drawings, respectively). In both cases there was no eect of name frequency, and
the patients provided specic semantic information about the items they
were unable to name (eg, for kangarooit lives in Australia, it carries its
young in a pouch, it hops). Both patients performed within the normal
range on standardized tests of semantics; however, on other tests, which
stress stored knowledge of the visual features of items, performance was not
as good. Thus, when drawing from memory, their drawings of animate items
received lower ratings from control subjects than their drawings of inanimate items (control subjects were presented with all the drawings and asked
to rate how good a representation of a target object each one was; good

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

517

representations received higher scores). Both patients were impaired somewhat, particularly when they were required to make within-category discriminations of real living objects from nonobjects created by combining
the parts of living things [67]. In addition, when the patients were required
to produce a name in response to denition stressing either the perceptual
properties of objects (eg, the name for an orange or cone-shaped vegetable)
or functional and associative properties (eg, the name of the root vegetable
that is said to help you see in the dark), the patients did well with the functional and associative denitions (92% and 89% correct for SRB and DM,
respectively) but more poorly on the perceptual denitions (51% and 42%
correct for SRB and DM, respectively (the control mean was 93.4% and
74% correct for inanimate and animate items, respectively). The data from
SRB and DM suggest that they have a mild disorder of stored perceptual
knowledge about objects, along with intact semantic knowledge of the functional characteristics of objects. Furthermore, this can produce an apparent
naming problem for living things, wherein patients sometimes seem to be
within a tip-of-the-tongue state. Humphreys, Riddoch, and Price [68] argue
that bottom-up activation of semantic knowledge from vision may be insufcient to invoke a name and that object naming requires recurrent activation of stored perceptual knowledge to dierentiate activation of a target
object from activation of other similar representations. Subtle impairments
of stored perceptual knowledge in cases such as SRB and DM produce naming decits because recurrent activation operates less well. This may be particularly important for the naming of living things, given the perceptual
similarity between their category members.

Summary
As was originally proposed by Lissauer, visual recognition may break
down either at an apperceptive or at an associative level. At an apperceptive
level, ner grain distinctions may be made; the authors distinguished here
between disorders of shape recognition and perceptual integration. It is not
the case, however, that all patients with visual recognition decits have
impaired perception: poor recognition and naming may also result from
damage to stored perceptual knowledge (eg, structural descriptions), from
problems in accessing semantic knowledge, from perceptual knowledge, or
from impairments to semantic knowledge itself. These represent dierent
types of associative decit. In some cases, mild damage to stored perceptual
knowledge also generates problems that are more severe on naming than on
recognition and more severe for some categories of objects than others.
References
[1] Lissauer H. Ein fall von seelenblindheit nebst einem beitrage zur theorie derselben. Arch
Psychiatr Nervenkr 1890;21:22270.

518

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

[2] Humphreys GW, Riddoch MJ. To see but not to see: a case of visual agnosia. London:
Lawrence Erlbaum; 1987.
[3] Zihl J, Von Cramon D, Mai N. Selective disturbance of movement vision after bilateral
brain damage. Brain 1983;106:31340.
[4] Zihl J, Von Cramon D, Mai N, Schmid C. Disturbance of movement vision after bilateral
posterior brain damage. Brain 1991;114:223552.
[5] Heywood CA, Zihl J. Motion blindness. In: Humphreys GW, editor. Case studies in the
neuropsychology of vision. Hove (UK): Psychology Press; 1999, p. 16.
[6] Efron R. What is perception? Boston Studies in Philosophy of Science 1968;4:13773.
[7] Bay E. Disturbances of visual perception and their examination. Brain 1952;76:51551.
[8] Riddoch MJ, Humphreys GW. A case of integrative agnosia. Brain 1987;110:143162.
[9] Alkhateeb W, Bromley JM, Humphreys GW, Javadnia A, Riddoch MJ, Ruddock KH.
Abnormal responses to multi-element spatial stimuli in a subject with visual form agnosia.
Clinical Vision Science 1992;7:16373.
[10] Humphreys GW, Riddoch MJ, Quinlan PT. Interactive processes in perceptual
organisation: evidence from visual agnosia. In: Posner MI, Marin OSN, editors. Attention
and performance XI. Hillsdale (NJ): Erlbaum; 1985.
[11] Cowey A. Cortical visual areas and the neurobiology of higher visual processes. In: Farah
MJ, Ratcli G, editors. The neuropsychology of high-level vision. Hillsdale (NJ): Lawrence
Erlbaum; 1994.
[12] Campion J, Latto R. Apperceptive agnosia due to carbon monoxide poisoning. An
interpretation based on critical band masking from disseminated lesions. Behav Brain
Res 1985;15:22740.
[13] Campion J. Apperceptive agnosia: the specication of constructs and their use. In:
Humphreys MJ, Riddoch, editors. Visual object processing: a cognitive neuropsychological
approach. Hove (UK): Lawrence Erlbaum Associates; 1987.
[14] Humphreys GW, Riddoch MJ, Quinlan PT, Donnelly N, Price CA. Parallel pattern
processing and visual agnosia. Can J Psychol 1992;46:377416.
[15] Humphreys GW, Riddoch MJ, Donnelly N, Freeman T, Boucart M, Muller H.
Intermediate visual processing and visual agnosia. In: Farah MJ, Ratcli G, editors. The
neurophysiology of high level Vision. Hillsdale (NJ): Lawrence Erlbaum Associates; 1992.
[16] Vecera SP, Gilds KS. What processing is impaired in apperceptive agnosia? Evidence from
normal subjects. Journal of Cognitive Neuroscience 1998;10:56880.
[17] Warrington EK. Neuropsychological studies of object recognition. In: Bruyn GW,
Klawans HL, editors. Handbook of clinical neurology. Amsterdam: Elsevier Science;
1985:33349.
[18] Warrington EK, Rudge P. A comment on apperceptive agnosia. Brain Cogn 1995;28:1737.
[19] De Renzi E, Spinnler H. Visual recognition in patients with unilateral cerebral disease.
J Nerv Ment Dis 1966;142:51325.
[20] De Renzi E, Scotti G, Spinnler H. Perceptual and associative decits of visual recognition.
Neurology 1969;19:63442.
[21] Warrington EK, James M. Disorders of visual perception in patients with localised cerebral
lesions. Neuropsychologia 1967;5:25366.
[22] Warrington EK, James M. Visual apperceptive agnosia: a clinico-anatomical study of three
cases. Cortex 1988;24:1332.
[23] Humphreys GW, Troscianko T, Riddoch MJ, Boucart M, Donnelly N, Harding GFA.
Covert processing in dierent visual recognition systems. In: Milner AD, Rugg MD,
editors. The neuropsychology of consciousness. London: Academic Press; 1992.
[24] Chainey H, Humphreys GW. The real-object advantage in agnosia: evidence for a role of
surface and depth information in object recognition. Cognitive Neuropsychology
2001;18:17591.
[25] Bruce V, Green PR, Georgeson MA. Visual perception: physiology, psychology and
ecology. Hove (UK): Psychology Press; 1997.

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

519

[26] Adler A. Disintegration and restoration of optic recognition in visual agnosia. Arch Neurol
Psychiatry 1944;51:24359.
[27] Benson DF, Greenberg JP. Visual form agnosia. A specic decit in visual discrimination.
Arch Neurol 1969;20:829.
[28] Milner AD, Perrett DI, Johnston RS, Benson PJ, Jordan TR, Heeley DW, et al. Perception
and action in visual form agnosia. Brain 1991;114:40528.
[29] Milner AD, Goodale MA. The visual brain in action. Oxford (UK): Oxford University
Press; 1995.
[30] Davido J, Warrington EK. A dissociation of shape discrimination and gure-ground
perception in a patient with normal acuity. Neuropsychologia 1993;31:8393.
[31] Heywood CA, Wilson B, Cowey A. A case study of cortical colour blindness with
relatively intact achromatopic discrimination. J Neurol Neurosurg Psychiatry 1987;50:
229.
[32] Troscianko T, Davido J, Humphreys GW, Landis T, Fahle M, Greelee M, et al. Human
colour discrimination based on a non-parvocellular pathway. Curr Biol 1996;6:20010.
[33] Gloning I, Gloning K, Ho H. Neuropsychological symptoms and syndromes in lesions of
the occipital lobes and adjacent areas. Paris: Gauthier-Villars; 1968.
[34] Holmes G, Horrax G. Disturbances of spatial orientation and visual attention with a loss
of stereoscopic vision. Arch Neurol Psychiatry 1919;1:385407.
[35] Riddoch G. Visual disorientation in homonymous half-elds. Brain 1935;58:37682.
[36] Rizzo M, Damasio H. Impairment of stereopsis with focal brain lesions. Ann Neurol
1985;18:147.
[37] Valkenberg CT. Zur Kenntis der gestoerten Tiefenwahrnehmung. Dtsch Z Nervenheilkd
1908;34:32237.
[38] Peterson MA, Gibson BS. Must gure-ground organisation precede object recognition?
Psychol Sci 1994;5:2539.
[39] Giersch A, Humphreys GW, Boucart M, Koviaks I. The computation of occluded contours
in visual agnosia: evidence of early computation prior to shape binding and gure-ground
coding. Cognitive Neuropsychology 2000;17:73159.
[40] De Renzi E, Lucchelli F. The fuzzy boundaries of apperceptive agnosia. Cortex
1993;29:187215.
[41] Kartsounis LD, Warrington EK. Failure of object recognition due to a breakdown
of gure-ground discrimination in a patient with normal acuity. Neuropsychologia
1991;2:96980.
[42] Humphreys GW, Riddoch MJ. Routes to object constancy: Implications from neurological
impairments of object constancy. Q J Exp Psychol 1984;36A:385415.
[43] Humphreys GW, Quinlan PT, Riddoch MJ. Grouping eects in visual search: eects with
single- and combined-feature targets. J Exp Psych [Gen] 1989;118:25879.
[44] Lawson R, Humphreys GW. The eects of view in depth on the identication of line
drawings and silhouettes of familiar objects: normality and pathology. Visual Cognition
1999;6:16595.
[45] Navon D. Forest before trees: the precedence of global features in visual perception.
Cognitive Psychology 1977;9:35383.
[46] Talland GA. Psychological studies of Korsakos psychosis: II Perceptual factors. J Nerv
Ment Dis 1958;127:197219.
[47] Warrington EK, Taylor A. The contribution of the right parietal lobe to object recognition.
Cortex 1973;9:15264.
[48] Ghent L. Perception of overlapping and embedded gures by children of dierent ages. Am
J Psychol 1956;69:57587.
[49] Grossman M, Galetta S, DEsposito M. Object recognition diculty in visual appercetive
agnosia. Brain Cogn 1997;33:30642.
[50] Le S, Cardebat D, Boulanouar K, Hena M.-A., Michel F, Milner D, et al. Seeing since
childhood, without ventral stream: a behavioural study. Brain 2002;125:5874.

520

M.J. Riddoch, G.W. Humphreys / Neurol Clin N Am 21 (2003) 501520

[51] Butter CM, Trobe JD. Integrative agnosia following progressive multifocal leukoencephalopath. Cortex 1994;30:14558.
[52] Hillis AE, Caramazza A. Cognitive and neural mechanisms underlying visual and semantic
processing: implications from Optic Aphasia. Journal of Cognitive Neuroscience 1995;7:
45778.
[53] Riddoch MJ, Humphreys GW. Visual object processing in optic aphasia: a case of semantic
access agnosia. Cognitive Neuropsychology 1987;4:13185.
[54] Riddoch MJ, Humphreys GW. Picture naming. In: Humphreys GW, Riddoch MJ, editors.
Visual object processing: a cognitive neuropsychological approach. London: Lawrence
Erlbaum Associates; 1987.
[55] Sheridan J, Humphreys GW. A verbal-semantic category-specic recognition impairment.
Cognitive Neuropsychology 1993;10:14384.
[56] Stewart F, Parkin AJ, Hunkin HN. Naming impairments following recovery from herpes
simplex encephalitis. Q J Exp Psychol 1992;44A:26184.
[57] Warrington EK, Shallice T. Category-specic semantic impairments. Brain 1984;107:82954.
[58] Caramazza A, Shelton JR. Domain-specic knowledge systems in the brain: The animate
inanimate distinction. Journal of Cognitive Neuroscience 1998;10:134.
[59] Forde E, Humphreys GW. Category-specic recognition impairments: a review of
important case studies and inuential theories. Aphasiology 1999;13:16993.
[60] Humphreys GW, Forde EME. Hierarchies, similarity, and interactivity in object recognition: category-specic neuropsychological decits. Behav Brain Sci 2001;24:
453509.
[61] Warrington EK, McCarthy RA. Category-specic access dysphasia. Brain 1983;106:
85978.
[62] Warrington EK, McCarthy RA. Categories of knowledge: further fractionations and an
attempted integration. Brain 1987;110:127396.
[63] Humphreys GW, Riddoch MJ, Quinlan PT. Cascade processes in picture identication.
Cognitive Neuropsychology 1988;5:67103.
[64] Humphreys GW, Lamote C, Lloyd-Jones T. An interactive activation approach to object
processing: eects of structural similarity, name frequency and task in normality and
pathology. Memory 1995;3:53586.
[65] Forde EME, Francis D, Riddoch MJ, Rumiati RI, Humphreys GW. On the links between
visual knowledge and naming: a single case study of a patient with a category-specic
impairment for living things. Cognitive Neuropsychology 1997;14:40358.
[66] Humphreys GW, Price CJ, Riddoch MJ. From objects to names: a cognitive neuroscience
approach. Psychol Res 1999;62:11830.
[67] Humphreys GW, Riddoch MJ. A case series analysis of category-specic decits of living
things: the HIT account. Cognitive Neuropsychology, in press.
[68] Humphreys GW, Riddoch MJ, Price CJ. Top-down processes in object identication:
Evidence from experimental psychology, neuropsychology and functional anatomy. Philos
Trans R Soc Lond 1997;B352:127582.
[69] Luria AR. Disorders of simultaneous perception in a case of bilateral occipito-parietal
brain injury. Brain 1959;82:43749.