Discussion 2

Neuropsychology

Visual agnosia and focal brain injury

Olivier Martinaud a,b,* a Department of neurology, Rouen university hospital, CHU Charles-

Nicolle, 1, rue de Germont, 76031 Rouen cedex, France b Inserm U1077, poˆ le des formations

et de recherche en sante´, universite´ de Caen–Normandie, UMR-S1077, 2, rue des

Rochambelles, 14032 Caen cedex, France

1. Introduction The cognitive consequences of focal brain injury create a window into greater

understanding of the visual recognition network. Despite the strengths of functional

neuroimaging, including the study of normal brains to reveal distinguishable activations that

correlate with selective cognitive process, detailed patient-based research is the only method for

revealing causal relationships among brain systems. Visual agnosia encompasses all disorders of

visual object recognition confined to a selective perceptual (visual) modality not due to an

impairment of elementary visual processing or some other cognitive deficit (such as language or

memory). Visual agnosia usually refers to visual object agnosia. However, this report considers

the entire spectrum of visual agnosia disorders, including visual spatial agnosia. There is no

unique taxonomy for visual object agnosia [1]. Based on a sequential dichotomy between

perceptual and memory systems, the most accepted proposition distinguishes ‘apperceptive

agnosia’ and ‘associative agnosia’ [2]. Patients with apperceptive agnosia fail to recognize a

visual stimulus because of an impairment in perceptual processing, excluding elementary visual

deficits (such as a r e v u e n e u r o l o g i q u e 1 7 3 ( 2 0 1 7 ) 4 5 1 – 4 6 0 i n f o a r t i c l e

Article history: Received 27 February 2017 Received in revised form 11 July 2017 Accepted 17

July 2017 Available online 24 August 2017 Keywords: Visual agnosia Prosopagnosia Alexia

Topographagnosia Orientation agnosia a b s t r a c t Visual agnosia encompasses all disorders of

visual recognition within a selective visual modality not due to an impairment of elementary

visual processing or other cognitive deficit. Based on a sequential dichotomy between the

perceptual and memory systems, two different categories of visual object agnosia are usually

considered: ‘apperceptive agnosia’ and ‘asso- ciative agnosia’. Impaired visual recognition

within a single category of stimuli is also reported in: (i) visual object agnosia of the ventral

pathway, such as prosopagnosia (for faces), pure alexia (for words), or topographagnosia (for

landmarks); (ii) visual spatial agnosia of the dorsal pathway, such as cerebral akinetopsia (for

movement), or orientation agnosia (for the place- ment of objects in space). Focal brain injuries

provide a unique opportunity to better unders- tand regional brain function, particularly with the

use of effective statistical approaches such as voxel-based lesion–symptom mapping (VLSM).

The aim of the present work was twofold: (i) to review the various agnosia categories according

to the traditional visual dual-pathway model; and (ii) to better assess the anatomical network

underlying visual recognition through lesion-mapping studies correlating neuroanatomical and

clinical outcomes. # 2017 Elsevier Masson SAS. All rights reserved. * Service de neurologie,

CHU Charles-Nicolle, 1, rue de Germont, 76031 Rouen cedex, France. E-mail address:

[email protected]. Available online at ScienceDirect www.sciencedirect.com

http://dx.doi.org/10.1016/j.neurol.2017.07.009 0035-3787/# 2017 Elsevier Masson SAS. All

rights reserved. visual field deficit), whereas those with associative agnosia fail to associate the

correct result of their visual analysis with their memory stores of the functional and semantic

properties of the stimulus. However, this distinction between the two broad categories has been

refined and extended ever since to reflect more elaborate perceptual models, such as the

computational approach to vision [3] and the hierarchical model of object recognition [4]. Within

the so-called apperceptive domain, three kinds of object agnosia are described: visual form

agnosia; integrative agnosia; and transformational agnosia.

Within the so-called associative domain, two sorts are described—multimodal associative

agnosia (not confined to the visual modality), and semantic agnosia (a deficit of conceptual

knowledge)— although both extend beyond the scope of visual agnosia stricto sensu. Patients

with visual form agnosia are unable to recognize, match, copy or discriminate simple visual

stimuli, such as a square from a circle. They perceive the traits and lines of objects, but are

impaired at discerning distances, lengths and orientations. Only a few cases have been reported,

with most having suffered from carbon monoxide poisoning [5–9], one from mercury poisoning

[10], another from oligodendroglioma [11] and a further one from stroke [12]. Patients who have

integrative agnosia are able to proceed with the individual elements of each form, but are unable

to put those elements together into a perceptual whole. Object recognition is sometimes

preserved through a feature-by- feature identification strategy. Two such patients have been

extensively tested, one after a stroke [13], the other following meningoencephalitis [14]. The

apperceptive nature of inte-grative agnosia has been described, and suggests that this kind of

agnosia could be a specific type [15]. Finally, patients with transformational agnosia are unable

to create a viewpoint- independent representation of an object, leading to difficulties in

identifying objects in unusual perspectives, but not when seen in canonical views [16]. This

disorder is also known as ‘perceptual categorization deficit’ [1].

Most patients with visual—whether object and spatial— agnosia have extensive lesions

involving both hemispheres. However, studies of deficits resulting from more focal brain injuries

have allowed descriptions of selective agnosia categories, such as prosopagnosia, pure alexia and

topogra- phagnosia (see Martinaud [17] for a review). In addition, the lesion-mapping approach

offers a better understanding of regional brain function, beginning with a description of the

traditional visual dual-pathway model [9]. Methodological improvements using statistics, such as

voxel-based lesion– symptom mapping (VLSM) [18], have further enhanced our capacity to

identify critical anatomical sites [19]. Thus, the aim of the present work was to review the

various agnosia categories and to better assess the anatomical network underlying visual

recognition, based on lesion-mapping studies correlating neuroanatomical and clinical outcomes.

2. Visual dual-pathway model

Two distinct visual pathways, the ‘what’ and ‘where’, were first reported in macaque monkey

brains [20], then adjusted with some modifications in humans to include a dissociation between

perception and action [21]: (i) the ventral pathway, involving the occipitotemporal cortex, allows

identification of visual stimuli and their semantic attributes; and (ii) the dorsal pathway,

involving the occipitoparietal cortex, allows visual control of actions, spatial localization of

visual stimuli, and identification of their spatial attributes, such as orientation, depth and

movement.

This model of two visual systems was inspired by the observation of double dissociations in

patients with visual form agnosia due to a lesion in the ventral pathway and patients with optic

ataxia due to a lesion in the dorsal pathway [22]. Two patients, R.V. and D.F., were reported as

classic illustrations of this dissociation [23]. R.V., a 55-year-old woman, suffered from optic

ataxia after several strokes involving bilateral lesions of the occipitoparietal region [23]. Patients

with optic ataxia are unable to reach out or grasp objects, although they have no difficulty in

recognizing or describing those same objects. D.F., a 34-year-old woman, developed visual form

agnosia due to lesions in the ven-trolateral occipital region following carbon monoxide poison-

ing [9]. In a well-known experiment using a slot that can be placed in many different orientations

and a hand-held card, D.F. showed great difficulty in indicating the orientation of the slot, yet

performed as well as normal subjects in reaching out and inserting the card into the slot [24].

More experiments with patients suffering from optic ataxia (for example, A.T.) [25,26] and those

with visual form agnosia (such as J.S.) [12] have since been reported, thereby supporting this

double dissociation, as has the recent study of a severe developmen- tal impairment affecting the

perception of visual objects while sparing motion processing [27].

To provide a more accurate account of reported patients with visual agnosia, different subclasses

have been proposed according to the damaged pathway. A lesion in the ventral pathway could

lead to a deficit of visual object identification, depending on the category of visual stimuli (face

or word), whereas a lesion in the dorsal pathway could lead to a deficit linked to visuospatial

attributes (movement or orientation) of the visual object. What follows here is a brief description

of the main agnosia categories and their neural correlates according to lesion-mapping studies.

3. Visual object agnosia of the ventral pathway 3.1.

Cerebral achromatopsia In this syndrome, the patient is unable to perceive colors, which is

different from color agnosia, an impaired knowledge of colors but with no difficulties in

perceptual tests [28], and color anomia, an impaired ability to name colors despite no difficulties

in perception or knowledge of colors [29]. The first review of 14 patients with perturbed

perception of colors emphasized the bilateral involvement of the fusiform gyrus, the lingual

gyrus, or both, close to the striate cortex [30]. The rarity of cerebral achromatopsia may be due to

the fact that large lesions destroy the striate cortex, leading to homony-mous hemianopia or

cortical blindness. The most recent review of 92 cases of cerebral achromatopsia underlined the r

e v u e n e u r o l o g i q u e 1 7 3 ( 2 0 1 7 ) 4 5 1 – 4 6 0452 high frequency (72% of cases) of

associated prosopagnosia [31]. In that study, the maximum overlap in the 11 cases with

achromatopsia, but without prosopagnosia, was found in the right hemispheric region located

next to V4, which plays a crucial role in color processing, as confirmed by the historical first

case report of hemiachromatopsia in Madame R. [32,33].

3.2. Prosopagnosia

This kind of agnosia is defined as the inability to recognize faces. Despite the relative rarity of

this disorder, many cases of prosopagnosia have been reported since the first case in 1844 (for

reviews, see Farah [1], Barton [34], and Mayer and Rossion [35]). Most were associated with

other visual disorders, such as achromatopsia [31], topographagnosia and object recognition

impairment [35]. Isolated prosopagnosia seems more like an exception, and is still debated

because of methodological issues, including the complexity of the tasks [36] (but see Farah [1])

and reaction times [36] (but see Rossion et al. [37]). At least three cases of patients with visual

agnosia for objects, but without prosopagnosia, have been reported [38,39], suggesting the

classic double dissociation between faces and objects, and favoring the idea of two distinct

systems that are at least partly independent. Consistent with the taxonomy mentioned above,

‘apperceptive prosopagnosia’, characterized by difficulty in copying faces, and ‘associative

prosopagnosia’, characterized by difficulty in matching faces, have also been reported [1].

Lesions causing prosopagnosia usually involve the bilateral occipitotemporal cortex, especially

the fusiform and lingual gyri, as demonstrated by autopsy studies [40]. However, many cases

have been reported of prosopagnosia patients with unilateral right hemispheric lesions (see

Martinaud [17] for a review, and also Schmidt [41], Jansari et al. [42], and Barton and Corrow

[43] for more recent cases). On the other hand, only four cases of prosopagnosia following

unilateral left-sided lesions have been described [44–47], three of which were left-handed,

suggesting anomalous lateralization of perceptual functions [44,45,47]. The fourth case suffered

from non-convulsive status epilepticus, with ‘‘focal right temporal abnormality and generalized

irregular spike and wave discharges’’ on electroencephalography [46]. There is also growing

evidence of right hemispheric dominance in facial recognition, with contributions from the left

hemisphere. However, to our best knowledge, only two lesion-mapping studies have attempted

to determine the lesions causing prosopagnosia [19,31]. The first, focused on cerebral achro-

matopsia, was a meta-analysis involving 52 patients with prosopagnosia [31]. Several non-

contiguous regions of the right hemisphere were found in the eight cases with prosopagnosia, but

no achromatopsia, suggesting that facial processing is distributed across the occipitotemporal

cortex, with maximum overlap with the inferior occipital gyrus. The second study was designed

as a systematic analysis of visual disorders following posterior cerebral artery strokes, using a

VLSM approach [18], in a consecutive series of 31 patients [19]. There was maximum overlap of

lesions in the right fusiform and parahippocampal gyri. These findings are consistent with the

results of an extensive analysis of individual prosopagnosia patients such as P.S. [48], and with

functional imaging studies in healthy subjects of the ‘cortical face network’ [49–51].

3.3. Pure alexia

Acquisition of an efficient reading system lies in the develop- ment of specialized mechanisms

linking vision and language. Pure alexia, also known as ‘alexia without agraphia’, ‘word

blindness’ and ‘agnosic alexia’, is characterized by various degrees of impaired word reading

with no deficits in the production or comprehension of oral language, and a preserved ability to

write spontaneously or on dictation [52]. In global alexia, patients are severely impaired and

cannot identify single letters [53]. Other patients are able to identify single letters and develop

letter-by-letter reading strategies, leading to a reading latency depending on word length, the

hallmark of pure alexia [1,54]. The usual description of this syndrome is a breakdown of ‘visual

word form representation’ [55] as the final result of the computationofthe abstract

identityofvisually perceived strings of letters. Such representation does not depend on visual

features such as letter size or font [53,56]. Clearly, acquired dyslexia without aphasia could arise

from syndromes other than pure alexia, especially hemianopic dyslexia (reading difficulties

confined to a visual hemi field as a result of hemianopia), and neglect dyslexia and simultagnosic

dyslexia, depending on the dorsal visual pathway injury. In addition, pure alexia is usually

associated with right hemianopia, although cases without visual field defects have been

documented [57]. Lesions causing pure alexia usually involve the left occipi- totemporal cortex,

especially the left fusiform and lingual gyri, as demonstrated in the first historically detailed

neuropatho- logical case [53]. The first lesion study using cerebral computed tomography (CT)

was conducted in 17 left posterior cerebral artery stroke patients [58], and compared the lesions

of five patients with pure alexia, five patients with global alexia and seven patients with no

reading impairment, and found a critical region in the left middle fusiform gyrus. Similar results

have been found by other, more recent, work using brain magnetic resonance imaging (MRI) and

VLSM [19, 52,59,60]. Lesions of the left middle fusiform gyrus appear to be good predictors of

pure alexia: in the most recent anatomical studies [19, 60], all patients with pure alexia (three

patients in each study) revealed lesions involving this region, whereas none of those without pure

alexia had such lesions, and only five pure-alexia right-handed patients were reported to have

unilateral right hemispheric lesions [61–65]. Thus, pure alexia results from a direct lesion of the

left middle fusiform gyrus, but also from disconnection, with disrupted projections to or from

that region [66], although only one case of completed e afforestation of an intact fusiform cortex

from both the left and right hemispheric lower-level visual- processing cortices has been

extensively studied [67]. Another case study highlights the role of degeneration of the inferior

longitudinal fasciculus following surgery for epilepsy [68]. Moreover, cases have been reported

of alexia restricted to the left half of the visual field due to lesions of the posterior callosal

pathways [52,69], as well as alexia restricted to the right visual field resulting from left

hemispheric lesions between an intact lower-level visual cortex and the fusiform cortex [70].

3.4. Topographagnosia

Also called ‘landmark agnosia’ or ‘topographic agnosia’, this is an inability to identify buildings,

landscapes or scenes (see r e v u e n e u r o l o g i q u e 1 7 3 ( 2 0 1 7 ) 4 5 1 – 4 6 0 453 Sewards

[71] for a review). Patients with this condition usually demonstrate topographical disorientation

or impaired navi-gation, although these deficits could be related to other difficulties, such as

memory impairment or general spatial cognitive disability (see Barrash [72], and Claessens and

van der Ham [73] for reviews). Note that even when both syndromes are associated,

topographagnosia differs from topographical disorientation (see below). According to the most

recent review of navigational impairment involving 67 patients (including 43 stroke patients),

whatever the cognitive explanation of their deficit [73], clear cases of topographa- gnosia due to

stroke lesions are more scarce, with heteroge-neous difficulties in recognizing famous and

familiar landmarks [74,75] and new landmarks [75–78], and even perceptual deficits with

complex scenes [79–81]. Double dissociation of these two deficits was described in two patients:

G.N. was unable to discriminate scenes without enough salient clues, yet recognized objects

perfectly [80]; and D.F. had object agnosia, yet could accurately discriminate between scenes

[82]. The right occipitotemporal cortex is consistently damaged in the majority of

topographagnosia patients [73]. One study of four stroke patients with topographagnosia (two

with pro-sopagnosia) and two other stroke patients with prosopagno- sia, but no landmark

agnosia, based on MRI scans but without statistical analysis, found that the right posterior

parahippo-campal gyrus was involved in the acquisition of novel information about scenes as

well as the identification of familiar landscapes, associated with the anterior lingual gyrus and

adjacent fusiform gyrus [75]. It is now accepted that lesions in the parahippocampal cortex affect

scene and landmark recognition [71], and a recent lesion-mapping study has demonstrated that

this cognitive ability probably depends on bilateral hemispheric lesions in this area [19]. 3.5.

Body form agnosia Representations of the human body—its general shape or body parts,

excluding the face—require a different kind of analysis. Three distinct types of representations

have been suggested [83]: (i) ‘body schema’, the coding of body postures; (ii) ‘body structural

description’, a topological map of the body; and (iii) ‘body image’, semantic information of the

names and functions of body parts. Based on this classification system, deficits of body

perception were investigated in a large study of 64 unilateral stroke patients, and found

associations between lesions of the left temporal lobe and body structural descriptions as well as

body image [83]. The body schema implicates a parie to frontal network, which means dorsal

rather than ventral involvement of the visual pathway. Unfortunately, anatomical analysis did not

enable more precise determination of the anatomo-clinical correlation. A subsequent lesion-

mapping study of 28 stroke patients distinguished two different deficits: body form and body

action deficits [84]. VLSM analysis found that lesions of the bilateral inferior and middle

occipitotemporal cortex and left superior temporal sulcus were associated with impaired

discrimination of body parts. Interestingly, selective impair-ment of body-part discrimination

(but not faces or parts of objects) was associated only with lesions of the bilateral middle

occipitotemporal cortex. These findings provide evidence of selective visual body form agnosia

and of partially distinct neural substrates of face and body processing. It must be stated that

functional imaging studies are making significant contributions in this area by defining more

precise neural networks, involving focal brain areas such as the extrastriate body area (EBA) and

fusiform body area (FBA) [85], which are also involved in the anatomical clusters found in the

reported lesion-mapping studies. However, this is beyond the scope of the present work.

4. Visual spatial agnosia of the dorsal pathway

Spatial neglect could be considered spatial agnosia linked to a focal brain lesion in the dorsal

visual pathway (but see Thiebaut de Schotten et al. [86] for the probable contribution of

subcortical white-matter pathways). This syndrome is des-cribed elsewhere in this volume.

4.1. Cerebral akinetopsia

In this condition, the patient is unable to perceive visual motion [87]. Cerebral akinetopsia, also

called ‘motion blind-ness’, has been described in only a few cases of focal brain injury [88–91].

Patient L.M. has been extensively reported in several studies [87,88,92–96] because of the

remarkable selectivity of her deficit, with no other visual or cognitive dysfunction except for

mild anomic aphasia [96]. The anatomical lesion in L.M. was localized to the posterior part of

the middle temporal gyri and adjacent portion of the occipital gyri in both hemispheres [92],

involving the V5/ medial temporal (MT) cortex known as the ‘motion center’. Similar subtypes

of cerebral akinetopsia have been proposed, such as the Zeitraffer phenomenon, where percep-

tion of the speed of an object in motion is impaired [97], or motion blindness confined in one

visual hemifield following a unilateral brain lesion [98–100]. The most recent group study of 21

unilateral acute stroke patients investigated lesion location in motion direction discrimination,

and mapped all lesions to the left hemisphere [89]. On superimposing lesions of the 10 motion-

blind patients on those of the 11 with no such deficits, this analysis revealed two areas of

overlap, one in the posterior parietal cortex close to the occipitoparietal sulcus, the other in the

occipitotemporal junction close to V5/MT. Complex motion-processing deficits, such as form

perception when derived from motion, are also found, but with no clear lesion correlations.

4.2. Optic ataxia

This is a deficit of visually guided movements to reach objects. Unlike cerebellar ataxia, patients

with optic ataxia are able to perform reaching tasks while receiving proprioceptive or auditory

information [101]. If the deficit results from a unilateral lesion, symptoms are then contralateral

to the arm and hemifield of vision, and have to be sought by asking patients to reach for targets

in the peripheral visual field. Optic ataxia is usually observed in the context of Ba´ lint’s

syndrome, r e v u e n e u r o l o g i q u e 1 7 3 ( 2 0 1 7 ) 4 5 1 – 4 6 0454which associates dorsal

simultanagnosia (see below), ocular apraxia (inability to direct voluntary eye movements to

visual targets) and optic ataxia. Based on a number of case reports with either unilateral or

bilateral lesions (R.V. [23], A.T. [25,26], I.G. [102,103], G.T. [104], U.S. and G.H. [105], O.K.

and C.F. [106], C.A.N. [107], M.H. [108] and K.E. [109]), it has been assumed that the

occipitoparietal junction is the neural correlate of optic ataxia. However, since the first mapping

study using CT data and the superimposition method in a series of 10 patients with unilateral

stroke was done [110], more recent group studies have pointed to the role of the junction

between the inferior parietal lobule and superior occipital cortex, as well as the junction between

the occipital cortex and superior parietal lobule [111–113].

4.3. Dorsal simultanagnosia

Patients with this deficit are able to recognize most objects, but cannot process more than one at

a time [1], which means that descriptions of a complex scene can be used to assess the

corresponding visual abilities leading to slow and fragmentary reports. Navon stimuli, which

consist of letters arranged to form a different letter [114], as well as Arcimboldo pictures, which

are portraits created by combining various diverse objects [115], can also be used to analyze the

restricted visual attention characterizing dorsal simultanagnosia. As already mentioned, this

deficit often appears in the context of Ba´ lint’s syndrome. Note that the term usually refers to

dorsal simultanagnosia, although ventral simultanagnosia can hap-pen. In this deficit, patients

have a reduced capacity to rapidly recognize multiple visual objects [1]. Lesions causing dorsal

simultanagnosia usually involve bilateral medial occipitoparietal junctions, yet analyses of the

neuroanatomical network primarily rest on just a few case reports [116–121]. To the best of our

knowledge, there is only one lesion-mapping study of a focal brain injury population comparing

lesion patterns in seven patients with simultana-gnosia (five with a stroke and two with

corticobasal degenera- tion) and 52 patients without simultanagnosia (49 with stroke) [122] (but

see also Neitzel et al. [123] for a study of 12 patients with posterior cortical atrophy). The results

of that voxel-based morphometry study indicated that simultanagnosia is associated with gray-

matter lesions of the bilateral occipito-parietal cortices and right intraparietal sulcus. It should

also be noted that extensive white-matter damage was also found and that diffusion tensor

imaging (DTI) analyses have provided emerging evidence of disruption of the integrity of three

long association pathways (superior longitudinal fasci-culus, inferior fronto-occipital fasciculus

and inferior longi-tudinal fasciculus) [124].

4.4. Topographical disorientation

This is defined as loss of the ability to navigate within large-scale environments [125]. A more

general condition than topographagnosia, this may be the consequence of many different

cognitive deficits (see above). In addition to agnosic or amnestic deficits, a disorder within the

egocentric spatial reference frame could lead to topographical disorientation. For instance,

patient G.W., who had no prosopagnosia or object agnosia, was unable to provide an accurate

route description [126]. Patients with ‘egocentric disorientation’ usually perform poorly on a

wide range of visuospatial tasks, suggesting that the deficit is not specific to the topographical

domain [125]. All patients reported so far [125] had right or bilateral posterior parietal lesions

that often involved the superior parietal lobe. Three patients with topographical disorientation,

but no difficulty recognizing landmarks or no egocentric disorientation, seemed to have trouble

coding allocentric spatial relationships between objects [127]. Their lesions were related to the

right retrosplenial region extending to the medial parietal lobe. Nevertheless, this taxonomy is

questionable, and more recent classifications have proposed three main categories of

navigational impairment [72]. The first is ‘landmark-based’ and essentially similar to topogra-

phagnosia; the two others are ‘location-based’ and ‘path-based’, and correspond to location and

path knowledge, respectively. The heterogeneous nature of these concepts does not favor a

comprehensive clinical anatomical correla-tion, which probably involves both visual pathways.

4.5. Autotopagnosia and heterotopagnosia

As already mentioned in the description of body form agnosia, body perception implicates both

visual pathways. A lesion in the dorsal visual pathway could be responsible for either

autotopagnosia (patients D.L.S. [128], D.A. and V.M. [129], G.L. [130], J.D. [131] and E.C.

[132]) or heterotopagnosia [133] (patients B.E.G., C.O.G. and R.O.M. [134]), involving an

inability to point to human body parts, either one’s own or those of someone else. Double

dissociation of these two deficits has been described in two patients, J.R. and A.P., who had

neurodegenerative disorders [135]. Both disorders were asso-ciated with posterior parietal

lesions of the dominant hemisphere [82,129] but, to our best knowledge, no lesion-mapping

study is available. Finger agnosia could be considered a particular form of autotopagnosia

limited to a selective deficit of identifying fingers, but it may also be absent in autotopagnosia

patients [136]. This deficit is usually observed in the context of Gerstmann’s syndrome [137].

4.6. Orientation agnosia and agnosia for mirror stimuli

An inability to visually identify the orientation of an object in space may differ according to the

plane used to mentally rotate the object. Patients with orientation agnosia are unable to

discriminate between objects rotated in the picture plane (x–y plane), whereas dissociation has

been observed in patients with agnosia for mirror stimuli, who are unable to discriminate

between objects rotated through the picture plane (mirrored on the x–z plane). Only 15 such

brain-injured patients have been reported so far (see Martinaud et al. [138] for a review): four

had both deficits; four had agnosia for mirrored stimuli with no orientation agnosia; one had a

reversed pattern; and six had orientation agnosia, but no evaluation of their performance for

mirrored stimuli. These patients all presented with large and mostly bilateral lesions, except for

one who had only a small right occipitoparietal hematoma [139]. r e v u e n e u r o l o g i q u e 1

7 3 ( 2 0 1 7 ) 4 5 1 – 4 6 0 455 The recent unique lesion-mapping study of 34 stroke patients

with unilateral lesions involving the parietal lobe to investigate the ability to discriminate

between different rotations has provided evidence of a double dissociation between (i) selective

agnosia for mirrored stimuli associated with a focal area involving the right inferior parietal

lobule, and extending to the intraparietal sulcus and posterior part of the temporal gyrus, and (ii)

orientation agnosia associated with a nearby, but distinct, anatomical region [138].

5. Study limitations

5.1. Ventral and dorsal visual pathway dichotomy

Ever since the visual dual-pathway model was described [140], extensive anatomical and

functional revisions have been proposed [141–143]. In addition, binary segregation of the ventral

and dorsal routes has been thrown into question by neuropsychological and neuroimaging

evidence [144]. As presented above, certain concepts, such as topographical disorientation and

body perception, require cognitive function distributions across both visual pathways. Moreover,

the ventral pathway might decode spatial properties of a visual stimulus, whereas the dorsal

pathway might contain object representations [143]. Furthermore, anatomical connectivity has

been reported between the occipitotemporal and occipi-toparietal networks [145]. The caudal

part of the inferior parietal lobule projects directly to hippocampal and para-hippocampal areas

[146], while the posterior arcuate fascicu-lus and vertical occipital fasciculus have been reported

as additional connections [147]. Nevertheless, the distinction between the occipitotemporal and

occipitodorsal streams remains useful for clinical purposes, such as describing the different

forms of visual agnosia. However, growing evidence challenges the functional independence of

the two visual pathways and instead suggests coordinated processes for integrating perceptual

information [148].

5.2. Localizationist and associationist models

Lesion-mapping studies are confronted by two limitations: (i) different lesion localizations could

lead to similar neuropsy- chological deficits; and (ii) symptoms may result from lesions in the

connections between two intact regions [149]. Locali-zationist models consider that one specific

function lies within one independent region, and lesions of that region will lead to complete loss

of function. However, it is well known that interregional connections can be found between

distant brain areas and that large-scale networks are dedicated to cognitive functions, as

demonstrated by functional imaging (see Catani et al. [150] for a discussion on localizationist

and associationist models). Associationist models take into account disconnec- tion syndromes

such as pure alexia, which could result from a direct lesion of the left fusiform gyrus or from

deafferentation of this intact region, as described above. Thus, interpretations of lesion-mapping

studies should always consider the entire neural network, including the disconnection hypothesis.

Diffusion MRI tractography and the atlas-based segmentation approach may also be very useful

together with the VLSM approach and functional imaging techniques.

6. Conclusion

The study of visual agnosic patients with focal brain injuries has provided relevant information

on the neural networks of visual recognition. Case reports and lesion-mapping studies, and

particularly the VLSM approach, are still of major importance for better identification of the

critical centers of cognitive function. Anatomical and functional connectivity should also be

considered.