Está en la página 1de 14

Downloaded from http://jnnp.bmj.com/ on March 13, 2015 - Published by group.bmj.

com

Journal of Neurology, Neurosurgery, and Psychiatry, 1974, 37, 489-501

The anatomical basis of prosopagnosia


J. C. MEADOWS'
From the Aphasia Unit, Boston V.A. Hospital, Boston, U.S.A.,
and National Hospital, Queen Square, London

SYNOPSIS Evidence is presented that patients with prosopagnosia have right anterior inferior
occipital lesions in the region of the occipital temporal junction. Many if not all cases have an
additional lesion in the left hemisphere; this is often but apparently not always symmetrical with
the right hemisphere lesion. This evidence is discussed in relation to the anatomical connections of
these regions and the results of experiments in animals.

Prosopagnosia is the specific inability to recog- recently drawn attention to the occurrence of
nize familiar faces. Its clinical features have been bilateral lesions in published cases that have
well reviewed by previous authors-for example, come to necropsy, although they did not con-
Hecaen and Angelergues (1962). It will be sider in detail the possible significance of the
pointed out here only that the severely prosopa- location of these lesions.
gnosic patient typically has no difficulty recogniz- In the present author's view, there is already
ing everyday objects, although he may be quite good evidence concerning the localization of
unable to recognize even members of his family lesions in prosopagnosia. This conclusion is
unless they speak, when he immediately recog- based upon an analysis of clinical case reports
nizes their voices. Some severely prosopagnosic and a re-evaluation of pathological findings in
patients may recognize certain people by utilizing reported cases. It is the purpose of this paper to
some specific attribute, such as a mole on the marshal this evidence and discuss possible
face, unusual height, spectacles of particular underlying mechanisms.
design, individual clothes, etc. However these
features can be described in verbal terms, and the
NEUROPSYCHOLOGICAL ASPECTS
visual percept of faces without such features does
not lead to recognition. The possible significance of the anatomical
Patients with this rare symptom have poster- lesions to be described will be better understood
iorly placed cerebral lesions with physical signs if consideration is first given to the normal pro-
suggesting bilateral involvement in some cases cess of facial recognition and then to the neuro-
but in others there is clinical evidence only of psychological features of prosopagnosia.
right hemisphere disease. Visual field defects are
nearly always present, but these are not sufficient FACES AS A CATEGORY It should be realized that
to account for the prosopagnosia, for other recognition of faces is a most complex and
visual material is usually perceived normally. sophisticated visual achievement which is par-
Unfortunately, pathological studies have been ticularly gestalt-like in nature, for most faces are
few and have not led previous authors to any resistant to verbal interpretation. We can each
firm conclusions on the anatomical lesions in- probably recognize more than 1,000 faces, the
volved. Rondot and Tzavaras (1969), who have majority of which differ in fine detail, but we do
reviewed the subject extensively, mostly from this in a manner which it is quite impossible to
the neuropsychological point of view, have convey verbally. It depends on learning what are
I Present address: The National Hospital, Queen Square, London
essentially visual pattern discriminations of great
WCIN 3BG. complexity.
489
Downloaded from http://jnnp.bmj.com/ on March 13, 2015 - Published by group.bmj.com

490 J. C. Meadows

There is no obvious counterpart to facial and bird watcher mentioned in the last para-
recognition in other visual experience. Thus, graph who had acquired skills with other
there is no other visual category within which we categories, and subsequently lost them when
are required to distinguish and remember such they developed prosopagnosia, would tend to
minor differences in detail by predominantly support this 'acquired' view, as does the well-
non-verbalizable means. It can be argued that known observation that 'all Chinamen look
certain individuals have special experience with alike' to the untravelled Westerner, but not to
particular categories (the forester with trees, the the Westerner who has spent years in the
horticulturalist with flowers, etc.) and in this Orient.
respect it is interesting that Bornstein (1963) The cautious conclusion is thus reached that
described a birdwatcher with prosopagnosia there are special characteristics about the brain's
who could no longer differentiate similar birds ability to discriminate and remember faces but
and a prosopagnosic farmer (Bornstein et al., these may be acquired rather than inborn. The
1969) who could no longer identify his individual same skills may not be acquired with other
cows. These however are special cases. Neverthe- categories simply because there is no cultural
less, the prosopagnosic patient of DeRenzi et al. pressure to acquire them. Accordingly, facial
(1968) had difficulty in distinguishing fruit (an recognition becomes by far the most complex
apple from a peach) and there have been some and frequently encountered example of relatively
cases with inability to distinguish particular pure visual discrimination learning that occurs in
chairs (Faust, 1955) or buildings (Pallis, 1955; everyday life. Considered in this way it becomes
Beyn and Knyazeva, 1962; Cole and Perez Cruet, less surprising that it may be disturbed in rela-
1964). These are less striking than the prosopa- tive isolation.
gnosia possibly because they are less difficult
visual tasks and also because they are less purely PROSOPAGNOSIA AND THE AMNESIC SYNDROME
visual (since it is easier to describe the differences
in verbal terms than with faces). Perhaps the most important recent discovery in
Cases such as those just mentioned are relevant relation to prosopagnosia is that patients with
to the question of whether there are special this condition, who are unable to recognize
characteristics (apart from the complexity of the familiar faces, perhaps even those of close
task) about the brain's ability to distinguish faces. friends and family, may be able to discriminate
Certain experimental observations suggest that and match faces normally. In other words the
the brain processes information about faces in a disturbance appears to be one of memory for
different way from other visual material (see for faces. To understand this better, it is helpful to
example, Yin, 1970). This is borne out by the consider some of the ways in which facial
everyday observation that we are strikingly recognition has been studied, together with the
better at recognizing a face as familiar after a results that have been obtained.
single encounter than we are other material-for A considerable literature has built up in
example, an individual horse or a tree. However, recent years showing that the right cerebral
this does not mean that skill in perception of hemisphere plays a special role in visual percep-
faces is a fundamental and primitive capacity of tion and visual memory, and indeed patients
the brain, as has been argued by some authors- with right hemisphere lesions who are not
for example, Bodamer (1947). It may be that we suffering from clinical prosopagnosia perform
learn to distinguish faces to a degree not seen less well than those with left hemisphere lesions
with other categories because facial recognition on scored tests involving facial discrimination
is from the very earliest age and throughout life (DeRenzi and Spinnler, 1966; Warrington and
such an essential and determining aspect of James, 1967; Benton and Van Allen, 1968;
daily living. Thus, it can be argued that we might Tzavaras et al., 1970). Such tests have been of
have acquired the same perceptual skill in rela- three broad types:
tion to the configuration of one tree relative to 1. Those testing visual memory for faces-such
the next, if tree configuration were as major a as recognition of photographs of well-known
determinant of behaviour as is faces. The farmer public figures (Warrington and James, 1967), or
Downloaded from http://jnnp.bmj.com/ on March 13, 2015 - Published by group.bmj.com

The anatomical basis ofprosopagnosia 491

of previously unknown faces after an appropriate recall was not, and from subsequent reports that
delay (Milner, 1968). This type of test is the most patients with true prosopagnosia may perform
closely related to the clinical symptom of pro- normally on matching tasks. This has been so in
sopagnosia and was shown statistically by these three patients so far reported (Tzavaras et al.,
authors in unselected series of patients to be most 1970; Benton and Van Allen, 1972). However a
impaired in right temporal lesions. (In none fourth patient reported by DeRenzi et al. (1968)
of these patients was clinical prosopagnosia performed badly on matching tasks and on both
present.) immediate and delayed recall, indicating that a
2. Those testing immediate recall of pre- restricted form of amnesia cannot account for
viously unfamiliar faces (DeRenzi and Spinnler, all cases.
1966; Warrington and James, 1967; Milner, There are objections to considering proso-
1968). The subject might be shown a facial pagnosia as a restricted form of the amnesic
photograph which is then removed, and he is syndrome. If this were so, one would expect
then immediately told to pick out the face from a amnesic patients-for example, patients with
display of several faces. alcoholic Korsakoff's psychosis-also to be
3. Various matching tasks, in which a facial prosopagnosic. In fact, it is very unusual for the
photograph is matched with other photographs difficulty in recognizing faces that is experienced
of the same face (Benton and Van Allen, 1968; by amnesic patients to compare in severity with
Tzavaras et al., 1970). In some tests, matching true prosopagnosia. The severely amnesic patient
has been with photographs taken from different is certainly unable to learn new faces but the
angles or in different ambient lighting, or modi- author is aware of no convincing case where this
fied by grimaces, false beards, spectacles, etc. difficulty extended retrogradely to faces learnt in
While performance in all these tests is im- the remote past, as in prosopagnosia. The
paired statistically in unselected groups of severest degree of deficit is exemplified by the
patients with right hemisphere lesions, they are much studied subject H.M., who was amnesic
not necessarily impaired to the same extent. after bimedial temporal lobectomy. This subject
Thus, Warrington and James (1967) found no has for many years since operation had severe
correlation in their series of patients (none of difficulty in facial recognition, being unable even
whom was clinically prosopagnosic) between im- to recognize friends and neighbours (first met
paired recognition of well-known faces and im- after operation) who have been visiting him
paired immediate recall of unfamiliar faces. They regularly for years (Milner, 1966). He has even
speculated that prosopagnosia might be a been known to invite total strangers into the
localized form of the amnesic syndrome (a point house to await his mother's return, thinking that
previously hinted at by Hecaen et al., 1957). The they must be friends he has not recognized. By
amnesic syndrome is characterized by impair- contrast, the reports indicate that he recognizes
ment of memory (both verbal and visuospatial) his parents normally. Such preservation of re-
with preservation of adequate intellectual func- mote memory is a regular feature of the amnesic
tions, and results from bilateral damage to syndrome which seems to be absent in prosopa-
central core limbic structures (see, for example, gnosia. Thus, the disturbance of memory for faces
Brierley, 1966). Amnesic patients certainly per- in prosopagnosia appears to be more severe and
form poorly in recognizing well-known faces (as different in character from that produced by bi-
they do in other tests of memory) and they are lateral lesions of central core limbic structures.
typically unaffected in matching tasks and in It is interesting to note, however, that patients
immediate recall of faces. with prosopagnosia very frequently have disturb-
This notion would hold, therefore, that ance of another memory function-topo-
prosopagnosia is a disturbance of one aspect of graphical memory (Hecaen et al., 1957;
visuospatial memory. Some support for this Bornstein, 1963; Rondot and Tzavaras, 1969).
view comes from Milner's (1968) finding that This is of some relevance, since patients with the
delayed recognition of faces was impaired amnesic syndrome itself have obvious clinical
(though not to the extent seen in prosopagnosia) difficulty in two major visual functions-topo-
after right temporal lobectomy, while immediate graphical orientation and facial recognition. In
Downloaded from http://jnnp.bmj.com/ on March 13, 2015 - Published by group.bmj.com

492 J. C. Meadows

fact, topographical orientation (memory) is less known to the author which have been published
purely visual than is facial recognition, since in during the last 40 years, together with five cases
general it involves more verbalizable material. from the records of the National Hospital,
Nevertheless, this relationship provides addi- Queen Square and Maida Vale. The cases briefly
tional support for those trying to relate proso- mentioned in Bornstein's review (1963) are not in-
pagnosia and amnesia, in spite of the objections cluded because of inadequate neurological details.
that have already been voiced.
There is some evidence that topographical The present cases (1-5) will not be reported in detail
orientation and facial discrimination may be dis- in this paper. Case 1 had a large posteriorly placed
sociable at cortical level: Newcombe and right hemisphere glioma with radiological evidence
Russell's (1969) study of missile-injured patients of invasion of the corpus callosum. Cases 2, 3, and 5
showed that a maze-learning task was more im- had each had two or more strokes with historical
paired with right parietal lesions, while defects of evidence of involvement of both hemispheres. Case
facial discrimination (Mooney's closure test) 4, a previously healthy 33 year old man, developed
were most severe in patients with right temporal prosopagnosia after evacuation of a right posterior
lesions. We can therefore speculate on whether temporal haematoma, arising from a minute angio-
matous malformation. In this case, the prosopa-
prosopagnosia might depend upon some lesion gnosia was not as severe as in the other cases; he
intermediate between a posterior area required could, for example, recognize his wife without
for facial discrimination and more anterior tem- difficulty, but failed to recognize people that he met
poral structures (with possible limbic connect- only once or twice a year (such as his cousins and
ions) required for learning. Such a lesion might certain close friends) even though he had known
also interrupt similar connections concerned them for many years. In all these cases there was
with topographical memory. The anatomical evidence of impaired topographical orientation.
evidence to be presented is compatible with
this hypothesis. Several points emerge from examination of Fig.
1. Visual field defects are nearly always present,
THE ANATOMICAL EVIDENCE and bilateral defects occur in about half the
In the majority of published cases of prosopa- cases. However, in the unilateral cases, the
gnosia exact localization of the lesions is not defect is nearly always on the left. Most notable
described. Investigations have been scanty and of all is the frequency of left upper quadrant
analysis of the reported clinical details is fre- involvement, suggesting a low, posteriorly
quently unrewarding from the viewpoint of pre- placed lesion of the right cerebral hemisphere.
cise localization. It is clear that in many cases Thus left upper quadrant involvement was noted
there is evidence of bilateral disease (shown in all the 19 cases with unilateral left-sided
usually by bilateral field defects), although in defects and in all but one of the 15 cases with
some there is evidence only of right hemisphere bilateral defects. There were however four cases
involvement. Whether the overall clinical picture with unilateral right-sided defects and four cases
suggests unilateral or bilateral disease, however, without visual field defects. It is worth noting at
there is one clinical feature which is of good this point that two of the four cases with uni-
localizing value. A left upper quadrantic visual lateral right-sided defects subsequently died and
field defect is a remarkably consistent finding in at necropsy had discrete right occipitotemporal
reported cases. In a few cases, radiological or pathology in addition to lesions in the left hemi-
operative findings have been described and in a sphere (Arseni and Botez, 1958; Benson et al.,
small number there have been necropsies. How- 1974).
ever, to the author's knowledge, there are only Of the cases without field defects, two resulted
seven detailed reports of necropsies in the from carbon monoxide poisoning (Hecaen et al.,
literature (but see Addendum). 1957; Benton and Van Allen, 1972), while a
third (Macrae and Trolle, 1956) followed closed
VISUAL FIELD DEFECTS The visual field defects in head trauma with coma for three weeks-con-
a large series of patients with prosopagnosia are ditions which usually cause bilateral brain
shown in Fig. 1. These comprise all reports damage. Bilateral damage in the case of Macrae
Downloaded from http://jnnp.bmj.com/ on March 13, 2015 - Published by group.bmj.com

The anatomical basis of prosopagnosia 493

Milian (1932) 4 Meadows (1973)-case 1 Cole and Perez Cruet

He'caen et at. (1952)-


case2
Medw I93-ae2Gloning
Meadows(1973)-case2 case 2
et at. (1966)

C Klein and Stack (1953) Meadows (1973)-case 3 Rondot et al. (1967)

C Alajouanine et al. (1953) Meadows (1973)-case 4 Gloning et al. (1970)

C) Faust (1955)-case I Meadows (1973)-case S

5 Faust (1955)-case 2 Holf and Potzl (1937)

t) de Busscher et al. (1956) Donini (1939) Potzl (1953)

j Hecaen et al. (1956)- Bodamer (1947)-case 1 Arseni and Botez (1958)


Hcase eta.(15)

E7j case22 Bodamer (1947)-case 2 Faust (1955)-case 4

Hecaen et at. (1957) Faust (1955)-case 5 h Rubens and Benson


case 2 KF (1971)
Bornstein and Kidron Fut(95-cs
5D (1959) Faust (1955)-case 6

5D case 1et at. (1966)- Faust (1955)-case 7 Faust (1955)-case 3

4>) De Renzi et al. (1968) H =caen et al. (1952)-


case 1
Macrae and Trolle
(1956)

jt') Bornstein et al. (1969) Pallis (1955) Wm= et at. (1957)


case 1

Assal (1969) Beyn and Knyazeva Benton and Van Allen


(1962) (1972)
FIG. 1. Visualfield defects in cases ofprosopagnosia published since 1930, including those offive cases from the
records of the National Hospital, Queen Square. The defects were homonymous but for convenience are here
represented by a single field. The cases have been arranged into four groups depending upon whether there were
left-sided defects alone, bilateral defects, right-sided defects alone, or no defects.
Downloaded from http://jnnp.bmj.com/ on March 13, 2015 - Published by group.bmj.com

494 J. C. Meadows

and Trolle was supported by the finding of small area of softening in the left hemisphere.
tachistoscopic tunnel vision for depth and Heidenhain's case (1927) showed infarction
colour. Carbon monoxide poisoning in particu- inferiorly, in the region of both occipitotemporal
lar may damage visual association cortex bi- junctions, involving fusiform and lingual gyri.
laterally and cause cortical visual impairment The case of Hecaen et al. (1957) had a right-
(Garland and Pearce, 1967) or visual agnosia sided glioblastoma which at operation was
(Benson and Greenberg, 1969). The author invading the occipital lobe, the posterior parts of
knows of an additional case of carbon monoxide the parietal and temporal lobes, and also the
induced prosopagnosia without visual field posterior part of the corpus callosum. Eventually
defect in whom less troublesome visual agnosia the patient died and necropsy confirmed the
was also present (M. Espir, personal com- presence of a glioma which by now was infiltra-
munication). ting extensively throughout the right temporal
The fourth published case without visual fieldand occipital regions. Involvement was however
defect (Faust, 1955) had clear evidence of right maximal in the medial and inferior aspects of the
occipitotemporal damage, even though no field occipital and posterior temporal regions, with
defect was recorded. In this case, which resultedextension into the corpus callosum as far as the
from a missile injury, there was a right occipito-
opposite ventricular wall.
basal bone defect associated with a piece of metal The case of Arseni and Botez (1958) was of a
3 cm from the midline close to the petrous bipolar spongioblastoma which at necropsy was
temporal bone, with dilatation of the trigone andinfiltrating extensively the left occipitotemporal
posterior horn of the right lateral ventricle on region. There was an additional deposit, about
contrast radiography. 1-5 cm in diameter, described as involving the
region of the tapetum on the right.
NECROPSY LOCALIZATION Pathological studies The case of Pevsner et al. (1962), reported in
confirm the presence of right occipitotemporal life by Bornstein and Kidron (1959) showed the
pathology in all cases so far published and in following findings at necropsy:
most cases there was an approximately sym-
'In the middle of the left inferior parietal lobule, in
metrical lesion in the left hemisphere. In all the angular gyrus and in the upper part of the
cases there was evidence of bilateral disease superior temporal sulcus, a yellow-brown, slightly
(Table). In Wilbrand's case (1892) there was depressed lesion, 3 cm in diameter was found.The
right occipital softening extending inferiorly anterior border of this lesion was formed by the
from the calcarine fissure, with a corresponding angular gyrus and it extended on to the parieto-
occipital sulcus which also formed its posterior
border. These macroscopic findings at autopsy
extended to a depth of 1 mm. On microscopy signs
TABLE of mild atrophy, gliosis and rarefaction of cells was
LOCATION OF LESIONS IN CASES OF PROSOPAGNOSIA seen.... On section of the brain, a narrow cystic
THAT CAME TO NECROPSY cavity, 1 cm in length ... was found in the region of
the inferior border of the medial surface of the right
Right hemisphere Left hemisphere hemisphere, 3 cm anterior to the occipital pole,
Wilbrand
under the calcarine fissure which also formed the
(1892) Inferior occipital Inferior occipitotemporal roof of the lesion. It was 4-6 mm below the cortex.'
Heidenhain
(1927)
Hecaen et al.
Inferior occipitotemporal Inferior occipitotemporal The patient of Gloning et al. (1970) sustained a
(1957) Occipito-parietotemporal Invasion through series of cerebrovascular accidents in the few
splenium to
ventricular wall
months before death and at necropsy there were
Arseni and areas of softening in various parts of the brain.
Botez (1958) Occipito-parietotemporal Occipitotemporal
Pevsner et al. Most striking in relation to his long-standing
(1962)
Gloning et al.
Inferior occipitotemporal Angular gyrus region symptom of prosopagnosia, however, were old,
(1970) Inferior occipitotemporal Inferior occipitotemporal bilateral ischaemic infarcts, situated almost
Benson et al.
(1974) Inferior occipitotemporal Inferior occipitotemporal
symmetrically in the region of the occipito-
temporal junction, destroying the central parts
Downloaded from http://jnnp.bmj.com/ on March 13, 2015 - Published by group.bmj.com

The anatomical basis of prosopagnosia 495

of both fusiform gyri. They extended deeply and RADIOLOGICAL OR OPERATIVE LOCALIZATION
involved the basal parts of the optic radiations There are a few cases in which approximate
but spared the striate cortex. localization can be deduced from radiological
Finally, the patient with associative visual and operative findings. In these cases the findings
agnosia reported in life by Rubens and Benson are in keeping with those already described.
(1971) has recently died. This patient had severe Hecaen and Angelergues (1962) mention three
prosopagnosia which persisted after ability to cases with difficulty in facial recognition after
name common objects on visual presentation had surgery to the posterior part of the right cerebral
become almost flawless. Necropsy in this patient hemisphere: removal of right parieto-occipito-
(Benson et al., 1974) showed infarction in left temporal region and right occipitotemporal
posterior cerebral artery distribution with region in two cases with intractable epilepsy, and
splenial destruction and a second, discrete lesion removal of a large right parieto-occipital
involving the right fusiform gyrus. The latter meningioma in the third. The case of Pallis (1955)
involved predominantly the white matter, extend- showed defective filling of the right posterior
ing from a point about 2 cm in front of the occi- cerebral artery on angiography but there were
pital lobe tip posteriorly to a point about 15 cm bilateral visual field defects indicating bilateral
anterior to the junction of parieto-occipital involvement. Assal (1969) described a patient
sulcus with calcarine sulcus. with a proven right parietotemporal intracerebral
These are the only well-described necropsy haemorrhage, and haemorrhage in the right
reports but there are two other cases that came posterior temporal region was responsible in the
to necropsy mentioned briefly in the literature, present case 4. Finally, the present case 1 had a
both reported by Bornstein (1963, 1965). His large deeply placed glioma in the posterior part
first case is confusing; in the case report the of the right hemisphere. Its localization is most
lesion is described as a right occipital lobe easily seen from the brain scan (Fig. 2). Carotid
meningioma causing left-sided hemianopia, yet and vertebral angiograms indicated a large
later in the paper he refers to the 'meningioma in malignant tumour lying deep in the right, pos-
the left parieto-occipital lobe'. Necropsy is not terior temporoparietal region and involving the
mentioned in this paper. In his second publica- splenium of the corpus callosum. At operation, a
tion he refers again to this case and also mentions large tumour was encountered at a depth of 2-5
an additional case. Without giving further details cm from the surface convexity of the brain. Fur-
he states simply that postmortem examination ther exploration showed that it had come through
'revealed a left parieto-occipital meningioma in the medial surface of the hemisphere and was
one case and a glioblastoma in the same region extending under the free edge of the falx in the
in the second'. It is difficult to comment on region of the vein of Galen. It also occupied the
either of these cases without further details. body of the right lateral ventricle, bulging into it

77-7

FIG. 2. Technetium brain


scan of the present case 2,
(a) anteroposterior and (b)
77 right lateral views.

(a f hn)
(b)
Downloaded from http://jnnp.bmj.com/ on March 13, 2015 - Published by group.bmj.com

496 J. C. Meadows

from its medial side. An incomplete removal lesion is not accompanied by a correspondingly
was performed. The patient was not followed sited lesion in the left hemisphere.
up. What is the mechanism of the disturbance
produced by these occipitotemporal lesions?
DISCUSSION
Their gross location has been discussed, but the
possible pathways affected have not yet been
LOCATION OF LESIONS The evidence that has considered. For the present, we will leave aside
been presented points clearly to the importance the question of whether the lesion at this site is
of occipitotemporal lesions in prosopagnosia. unilateral or bilateral and just consider the
The evidence for right-sided involvement is very possible structures involved.
strong: lesions in this region were found in all The lesions that have been described affect
the seven cases that came to necropsy so far mainly inferior occipitotemporal cortex (lingual
reported, and this correlates with the extremely and fusiform gyri) and do not usually extend for-
high incidence of left upper quadrantic visual ward far into the temporal lobe. They also in-
field defects in clinical case reports. On the other volve the white matter deep to these gyri, com-
hand, lesions at this site in the left hemisphere prising incoming and outgoing fibres to this
were found in only five of the seven necropsies, cortex (including tapetal fibres descending from
and right upper quadrantic field defects are not the posterior part of corpus callosum), inferior
particularly common clinically. longitudinal fasciculus (ILF) and lower fibres of
The lesions in the right hemispheres involved optic radiation.
fusiform or lingual gyri in six of the seven cases. In fact, the lower part of the optic radiation
Immediately deep to these gyri lies the optic fuses and is intermingled with ILF, which is the
radiation, and this structure was obviously main pathway from the classical visual associa-
involved in most if not all these cases (a clear tion cortex (areas 18 and 19) into the temporal
statement is not made in every case). In the lobe. These two structures are together known as
seventh case (Arseni and Botez, 1958) the lesion the external sagittal stratum. A lesion of the
was more dorsally placed, and was described as lower part of the right optic radiation in the
involving the tapetum (callosal fibres sweeping occipitotemporal region (accounting for the left
down into the occipito-temporal region). upper quadrantic field defect) will thus almost
Left hemisphere lesions were present in all certainly involve ILF and is likely in addition to
seven necropsies but only in five were these interrupt callosal (tapetal) connections between
occipitotemporal as in the right hemisphere inferiorly placed occipitotemporal cortex on the
(Table). In the case of Pevzner et al. (1962) that two sides. Even the more dorsally placed lesion
came to necropsy the only left-sided lesion in the case of Arseni and Botez (1958) probably
described was mild gliosis in the angular gyrus involved the external sagittal stratum, for this
region. In the case of Hecaen et al. (1957) there structure is closely applied to the tapetum, upon
was invasion by tumour from the right hemi- which the tumour was centred, just lateral to the
sphere through the posterior part of the corpus ventricle.
callosum, but only as far as the ventricular wall. These basal occipitotemporal lesions may thus
As already stated, the purely clinical data also interrupt the main visual input to the temporal
fail to support the consistent presence of sym- lobe in addition to any effects arising from
metrical lesions, for right upper quadrantic field damage to the occipitotemporal cortex itself. In
defects are not a regular finding in published case fact, some fibres of ILF probably interconnect
reports; in about half the cases there is no right various parts of inferior occipital and temporal
field defect at all and in some of the cases cortex, and ILF may even form a chain of
where a right-sided defect is present it spares the connections concerned with visual function, ex-
right upper quadrant (Fig. 1). We may conclude tending from the occipital into the temporal
rather cautiously at this point that bilateral lobe (see below). Damage to ILF in this
occipitotemporal lesions may underlie many region may thus have some effects similar to
cases of prosopagnosia but that in some (and lesions of occipitotemporal cortex itself, by in-
possibly many) cases, the right occipitotemporal volving the latter's incoming and outgoing con-
Downloaded from http://jnnp.bmj.com/ on March 13, 2015 - Published by group.bmj.com

The anatomical basis of prosopagnosia 497

nections. Damage to the long fibres in ILF is lesions in the left hemispheres have not all been
probably not important, for prosopagnosia does occipitotemporal in location. Similarly, the evi-
not appear to be associated with lesions more dence from clinical studies, in particular the
anteriorly in the temporal lobe, further along the visual field data, leads one to question the role of
course of these same fibres. the left hemisphere (at the very least, it must be
different from the role of the right hemisphere).
COMPARISON WITH ANIMAL EXPERIMENTS Some This, of course, accords with the statistical
insight into the possible neuropsychological studies in unselected brain damaged patients
mechanism of lesions at this site may be gained which show that the right hemisphere is of much
by considering the results of experiments in greater importance in facial discrimination and
animals. Prosopagnosia associated with bilateral recall. It is natural to question whether the left
occipitotemporal lesions might be compared hemisphere lesion is necessary at all. The evi-
with the defect of visual discrimination learning dence, however, is overwhelming that bilateral
(with sparing of both primary visuosensory func- lesions are important in the overall aetiology of
tion and learning in other modalities) which fol- prosopagnosia. The high incidence of bilateral
lows bilateral ablations of inferotemporal cortex disease clinically and the invariable presence of
in the macaque (see review by Gross, 1974). This bilateral lesions in the few necropsies so far per-
analogy may be relevant even to cases with in- formed can scarcely have occurred by chance.
volvement of only the right occipitotemporal Moreover, bilateral lesions conveniently explain
region, for this is the side of the brain which, in the rarity of the condition. However, this still
man, is most concerned with visuospatial func- does not explain the inconsistent location of the
tion and, more specifically to the present con- left hemisphere lesion. It is not unreasonable to
text, with facial discrimination and recall (see the suggest, for example, that the left occipito-
statistical studies on unselected series of brain temporal lesions are important (being located
damaged patients already discussed). roughly symmetrically to those in the right
A simple explanation of prosopagnosia can hemisphere) but that in the two cases that came
thus be suggested based on the occipitotemporal to necropsy where the left hemisphere lesions
location of the lesions and the fact that similarly were not occipitotemporal (Hecaen et al., 1957;
placed lesions in monkeys impair visual dis- Pevzner et al., 1962) they might be incidental and
crimination learning. In man, facial recognition irrelevant.
is so much more complicated and more fre- Some alternative possibilities might be con-
quently undertaken than any other task involv- sidered to account for some of the anomalies
ing complex visual discrimination learning that that have been discussed-in particular the
part of the inferior occipitotemporal cortex clinical evidence implicating the right hemi-
appears to become functionally modified for this sphere more than the left. Some of those patients
task. It is suggested that destruction of this with clinical evidence of only unilateral disease
cortex or its connections underlies prosopa- might indeed have only unilateral lesions and be
gnosia. Connections from this region to more rare individuals in whom visual function, or at
anterior temporal cortex and thence to medial least the function of facial recognition, is unusu-
limbic structures are apparently still necessary ally lateralized in the right hemisphere, whereas
for learning new faces, as has already been dis- in most individuals bilateral lesions are required
cussed in relation to the amnesic syndrome. to cause prosopagnosia. Thus, some cases might
have unilateral (right) occipitotemporal lesions,
THE PROBLEM OF THE LEFT HEMISPHERE LESION So as in the case of Pevzner et al. (1962), while
far we have avoided discussing the difficult prob- other cases have bilateral occipitotemporal
lem of the variable location of the left hemisphere lesions.
lesions and the question that follows from it- Another possibility is that the combination of
whether bilateral lesions are essential or not for lesions in bilateral cases has the same effect, so
the development of prosopagnosia. It has already far as physiological mechanism is concerned, as a
been noted that bilateral lesions have been present single strategically placed right-sided lesion. For
in all cases that came to necropsy but that the example, recognition of familiar faces might
Downloaded from http://jnnp.bmj.com/ on March 13, 2015 - Published by group.bmj.com

498 J. C. Meadows

depend on connections from both occipital cor- dence that a unilateral lesion can cause proso-
tices to the right occipitotemporal cortex. A pagnosia. It is to be hoped that further necropsy
lesion causing destruction of the right primary studies will explain some of the paradoxes con-
visual cortex in combination with destruction at cerning lesion location, at least at the gross
some point along the pathway from the left anatomical level.
striate cortex to the right occipitotemporal
region (probably via the left visual association VARIETIES OF PROSOPAGNOSIA It has already been
cortex, including the left occipitotemporal region, pointed out that discrimination of faces may
and callosal fibres passing in the splenium) be normal in prosopagnosia and that, at least
might then have the same effect as a single right in some cases, the disorder can be considered
occipitotemporal lesion. Damage to the callosal to be one of memory for faces. However, it
fibres might be in the splenium itself, or more is unlikely that a disturbance of visual memory
laterally placed in either right or left hemisphere. underlies all cases of prosopagnosia. Thus the
It will be recalled that there was gliomatous patient of DeRenzi et al. (1968) was unable
infiltration of the corpus callosum from the right even to discriminate faces, judging from his
hemisphere in both the case of Hecaen et al. performance on matching tasks and tests of
(1957) that came to necropsy and case 1 of the immediate recall. This is not too surprising, for
present author. However, this explanation fails facial recognition is probably not a single-step
to account for the rarity of prosopagnosia, process but is likely to involve a series of stages.
which would, for example, be expected to occur In fact, different types of abnormality can be pre-
regularly with right posterior cerebral artery dicted. Discrimination of faces might be im-
occlusion. paired and the subject be unable to say whether
Could the last explanation apply if an isolated two identical facial photographs are the same or
right hemisphere lesion were particularly dis- not. At a more complex level, he might succeed
crete in location ? Could, for example, the in this task but be unable to say whether two
neuronal networks in the chain of connections faces are the same when photographed from
beyond a particularly discrete right occipito- different angles or in different ambient lighting.
temporal lesion still have a decisive effect on the In turn, these processes of perceptual integration
brain's response? In other words, might these might be intact but the patient be unable to draw
networks actively 'inform the brain' that a on memory function required for recognition.
familiar face is unfamiliar simply because they do The last-mentioned seems to be the commonest
not receive the information necessary to identify type of defect.
it? It is known that the left hemisphere can The present study suggests that the anatomical
identify faces when disconnected from the right substrate for these functions includes the region
hemisphere (Levy et al., 1972) and this is pre- of the occipitotemporal junction, particularly on
sumably how facial recognition is achieved, for the right. Gross et al. (1972) have provided some
example, when right occipital and basal temporal evidence in the monkey that the sort of hierarchi-
lobe is destroyed by right posterior cerebral cal perceptual processes demonstrated by Hubel
artery occlusion. But, in a particularly discrete and Wiesel (1962, 1965, 1968) in the geniculo-
right occipitotemporal lesion of the type en- striate system of the cat and monkey may be
visaged, is it possible that the 'negative' informa- extended into the inferotemporal cortex for
tion from the right hemisphere in some manner more complex material, and it is reasonable to
suppresses the 'positive' information from the consider whether a similar mechanism might
left hemisphere and leads to an actual denial of occur with complex perceptual processes, like
recognition ? With a larger right hemisphere facial recognition, in man. The known lateralizat-
lesion, the right hemisphere mechanism would ion of visuospatial function in man is presumably
be more deranged and identification by the left relevant to the apparent greater importance of
hemisphere would take place. the right hemisphere but as already discussed
This is an interesting speculation but not one certain questions remain unanswered in relation
which can at the present time be supported; as to this.
already discussed, there is so far no firm evi- Such a hierarchical arrangement could under-
Downloaded from http://jnnp.bmj.com/ on March 13, 2015 - Published by group.bmj.com

The anatomical basis of prosopagnosia 499

lie clinical differences between reported cases. hemisphere, two came to necropsy and in both
Against this background, it is easier to under- cases bilateral occipitotemporal lesions were
stand cases such as that of DeRenzi et al. (1968) found (see previous section); rather surprisingly
with impaired facial discrimination, and cases the two right hemisphere lesions did not cause
associated with metamorphopsia (altered or dis- detectable left visual field defects. Thus there is
torted vision), as discussed by Hecaen and no evidence at the present time that prosopagno-
Angelergues (1962), where there is a more pri- sia can result from an isolated left hemisphere
mary disturbance of perception. The cases of lesion.
Bodamer (1947) fall into the latter category and In conclusion, however, it is worth considering
his first patient, for example, complained that the possible combined effects of partial interrup-
faces appeared 'strangely flat; white with dark tion of visual input to speech area together with a
eyes, as if in one plane, like white oval plates ... right occipitotemporal lesion, for such an ex-
all the same' (see Critchley, 1953). Metamor- planation could explain the problematical case of
phopsia may also be very pronounced without Pevzner et al. (1962), that came to necropsy
actual prosopagnosia, as in Bodamer's third where the only reported lesion in the left hemi-
patient who really had a specific metamorphop- sphere was mild gliosis in the angular gyrus
sia involving faces, rather than prosopagnosia, region.
since the patient could still recognize faces. SUMMARY
Finally, it must be remembered that difficulty
in facial recognition may occur as part of less Patients with prosopagnosia nearly always have
specific neurological disturbances. The question a left upper quadrantic visual field defect,
of facial recognition in amnesic subjects has correlated clinically and pathologically with a
already been considered and it has been pointed right occipitotemporal lesion. The few necropsies
out that the defect in recognition here does not all show bilateral lesions but the right hemisphere
appear to extend retrogradely to faces learned in lesions have all involved the region of the
the remote past, but only as far retrogradely as occipitotemporal junction.
the amnesia itself extends; the difficulty is in The lower part of the optic radiation fuses
learning new faces during the period of retro- with the inferior longitudinal fasciculus (ILF)
grade and anterograde amnesia. and these two structures lie immediately dorsal
Facial recognition may also be impaired in to inferior occipitotemporal cortex. ILF is the
patients with visual agnosias (for example, main direct pathway by which visual information
Benson and Greenberg, 1969; Taylor and reaches temporal lobe (and thence limbic struc-
Warrington, 1971). In these conditions objects tures), but along its path its fibres also inter-
cannot be identified when presented visually connect the different areas of inferior occipito-
and it is therefore not too surprising that faces temporal and posterior temporal cortex. As
may not be identified either. In 'apperceptive' lesions far forward in ILF are apparently not
visual agnosia there is bilateral disturbance of associated with prosopagnosia, involvement of
visual association cortex. On the other hand inferior occipitotemporal cortex or its incoming
' associative' visual agnosia has been inter- and outgoing connections is probably the im-
preted as indicating disconnexion between the portant factor.
visual areas and the posterior speech area All cases of prosopagnosia that have come to
(Geschwind, 1965) and can result from lesions necropsy have also shown lesions in the left
confined to the left hemisphere. It is theo- hemisphere, mostly occipitotemporal, so that
retically possible that a lesion confined to the approximately symmetrical lesions were present.
left hemisphere might similarly cause isolated But in two out of seven necropsies no left occipito-
prosopagnosia. This would depend upon whether temporal lesion was described. Right-sided visual
fibres transmitting information about faces are field defects have been present in only about half
segregated sufficiently from other visual input to the recorded cases. The problem of the inconsis-
the speech area. Of the four cases of prosopa- tent location of the left hemisphere lesion has
gnosia shown in Fig. 1, where the clinical data not been resolved but various possibilities have
suggested confinement of pathology to the left been considered.
Downloaded from http://jnnp.bmj.com/ on March 13, 2015 - Published by group.bmj.com

500 J. C. Meadows

Bilateral lesions of inferior temporal cortex in Bornstein, B. (1965). Prosopagnosia. In 8th International
monkeys impair visual discrimination learning. Congress of Neurology, Vienna, 1965. Proceedings, Vol. 3,
pp. 157-160. Wiener Medizinischen Akademie: Vienna.
Facial recognition can in many respects be Bornstein, B., and Kidron, D. P. (1959). Prosopagnosia.
regarded as an extremely complex learning task Journal of Nelurology, Neutrosurgery, and Psychiatry, 22,
124-131.
in visual discrimination. Bornstein, B., Sroka, H., and Munitz, H. (1969). Proso-
pagnosia with animal face agnosia. Cortex, 5, 164-169.
Brierley, J. B. (1966). The neuropathology of amnesic states.
ADDENDUM In Amnesia, pp. 150-180. Edited by C. W. M. Whitty and
0. L. Zangwill. Butterworths: London.
Since this paper was written, three further cases have Cole, M., and Perez-Cruet, J. (1964). Prosopagnosia. Neulro-
been drawn to the author's attention, in one of psychologia, 2, 237-246.
Critchley, M. (1953). The Parietal Lobes. Arnold: London.
which necropsy findings were published. A case is Critchley, M. (1965). Acquired anomalies of colour percep-
reported in Critchley's (1965) paper on anomalies of tion of central origin. Brain, 88, 711-724.
colour vision of cerebral origin; this patient had a De Busscher, J., Hoffman, G., and Kluyskens, J. (1956).
superior altitudinal hemianopia with prosopagnosia, Agnosia visuelle temporaire pour les personnes et optico-
spatiale pour les objets a la suite d'un ictus unique. Acta
loss of topographical sense, and impaired colour Neurologica Belgica, 56, 162-176.
vision with xanthopsia. Lhermitte et al. (1972) report De Renzi, E., and Spinnler, H. (1966). Facial recognition in
a case that came to necropsy who, in addition to brain-damaged patients. An experimental approach.
difficulty with faces was also unable to recognize Neurology (Minneap.), 16, 145-152.
De Renzi, E., Faglioni, P., and Spinnler, H. (1968). The
familiar cars, a feature shown to some extent also by performance of patients with unilateral brain damage on
the present case 5 and by the case of Macrae and face recognition tasks. Cortex, 4, 17-34.
Trolle (1956). At necropsy, Lhermitte's case showed Donini, F. (1939). Su di un caso di aprasia costruttiva con
bilateral lesions of fusiform gyrus with, on the right, grave disorientamento esospaziale e perdita della facolta
del viconoscimento della fisonomia della personne. Note
involvement also of the adjacent lingual gyrus. Psichiatre, 68, 469-485.
Finally, Tzavaras, Merienne, and Masure (personal Faust, C. (1955). Die zerebralen Herdstorungen bei Hinter-
communication) describe a left-handed patient with a hauptsverletzungen und ihre Beurteilung. Thieme: Stuttgart.
left temporal lesion who presented with proso- Garland, H., and Pearce, J. (1967). Neurological complica-
tions of carbon monoxide poisoning. Quarterly Journal of
pagnosia, amnesia, and language difficulty. Medicine, 36, 445-455.
Geschwind, N. (1965). Disconnexion syndromes in animals
My thanks are due to Dr. D. Frank Benson for his and man. Part 2. Brain, 88, 585-644.
Gloning, I., Gloning, K., Hoff, H., and Tschabitscher, H.
comments on this manuscript. (1966). Zur prosopagnosie. Neuropsychologia, 4, 113-131.
Gloning, I., Gloning, K., Jellinger, K., and Quatember, R.
(1970). A case of 'prosopagnosia' with necropsy findings.
REFERENCES Neuropsychologia, 8, 199-204.
Alajouanine, T., Lhermitte, F., and Sabouraud, 0. (1953) Gross, C. G. (1974). In press, quoted by Gross et al., 1972.
Gross, C. G., Rocha-Miranda, C. E., and Bender, D. B.
Agnosie visuelle sans alexie. Revue Neurologique, 89, 158. (1972). Visual properties of neurons in inferotemporal
Arseni, C., and Botez, M. I. (1958). Consideraciones sobre un cortex of the macaque. Journal of Neurophysiology, 35, 96-
caso de agnosia de las fisonomias. Revista Neuro- 111.
psiquiatria, 21, 583-593. Hecaen, H., and Angelergues, R. (1962). Agnosia for faces
Assal, G. (1969). Regression des troubles de la reconnaissance (prosopagnosia). Archives of Neurology (Chlic.), 7, 92-100.
des physionomies et de la memoire topographique chez un Hecaen, H., Ajuriaguerra, J. de, Magis, C., and Angelergues,
malade opere d'un hematome intracerebral parieto- R. (1952). Le probl&me de l'agnosie des physionomies.
temporal droit. Revue Neurologique, 121, 184-185. Ence~phale, 41, 322-355.
Benson, D. F., and Greenberg, J. P. (1969). Visual form Hecaen, H., Angelergues, R., Bernhardt, C., and Chiarelli, J.
agnosia. Archives of Neurology (Chic.), 20, 82-89. (1957). Essai de distinction des modalites cliniques de
Benson, D. F., Segarra, J. M., and Albert, M. L. (1974). l'agnosie des physionomies. Reviue Neurologique, 96, 125-
Archives of Neurology. (In press.) 144.
Benton, A. L., and Van Allen, M. W. (1968). Impairment in Hecaen, H., Penfield, W., Bertrand, W., and Malmo, R.
facial recognition in patients with cerebral disease. Cortex, (1956). The syndrome of apractognosia due to lesions of
4, 344-358. the minor cerebral hemisphere. Archives of Neurology
Benton, A. L., and Van Allen, M. W. (1972). Prosopagnosia (Chic.), 75, 400-434.
and facial discrimination. Journal of Neurological Sciences, Heidenhain, A. (1927). Beitrag zur Kenntnis der Seelen-
15, 167-172. blindheit. Monatsschriftfur Psychiatrie und Neuirologie, 66,
Beyn, E. S., and Knyazeva, G. R. (1962). The problem of 61-116.
prosopagnosia. Journal of Neurology, Neurosurgery, and Hoff, H., and Potzl, 0. (1937). Ueber eine optisch-agnostische
Psychiatry, 25, 154-158. Storung des 'Physiognomie-Gedachtnisses' (Bezeihungen
Bodamer, J. (1947). Die Prosop-Agnosie. Archiv fur Psy- zur Ruckbildung eine Wortblindheit). Zeitschrift fur die
chiatrie und Nervenkrankheiten, 179, 6-53. gesamte Neurologie und Psychiatrie, 159, 367-395.
Bornstein, B. (1963). Prosopagnosia. In Problems of Dynamic Hubel, D. H., and Wiesel, T. N. (1962). Receptive fields,
Neurology, pp. 283-318. Edited by L. Halpern. Hadassah binocular interaction and functional architecture in the
Medical Organization: Jerusalem. cat's visual cortex. Journal of Physiology, 160, 106-154.
Downloaded from http://jnnp.bmj.com/ on March 13, 2015 - Published by group.bmj.com

The anatomical basis of prosopagnosia 501

Hubel, D. H., and Wiesel, T. N. (1965). Receptive fields and 1'allis, C. A. (1955). Impaired identification of faces and
functional architecture in two nonstriate visual areas (18 places with agnosia for colours. Journal of Neurology,
and 19) of the cat. Journal of Neurophysiology, 28, 229-289. Neurosurgery, and Psychiatry, 18, 218-224.
Hubel, D. H., and Wiesel, T. N. (1968). Receptive fields and I'evzner, S., Bornstein, B., and Loewenthal, M. (1962).
functional architecture of monkey striate cortex. Journal of Prosopagnosia. Journal of Neurology, Neurosurgery, and
Physiology, 195, 215-243. Psychiatry, 25, 336-338.
Klein, R., and Stack, J. J. (1953). Visual agnosia and Potzl, 0. (1953). Zur Agnosie des Physiognomiegedachtnisses.
alternating dominance; analysis of a case. Journal of Mental JWiener Zeitschrift fir Nervenheilkunde, 6, 335-354.
Science, 99, 749-762. Rondot, P. Tzavaras, A., and Garcia, R. (1967). Sur un cas
Levy, J., Trevarthen, C., and Sperry, R. W. (1972). Percep- de prosopagnosie persistant depuis quinze ans. Reviue
tion of bilateral chimeric figures following hemispheric Neurologique, 117, 424-428.
deconnexion. Brain, 95, 61-78. Rondot, P., and Tzavaras, A. (1969). La prosopagnosie apres
Lhermitte, F., Chain, F., Escourolle, R., Ducarne, B., and vingt annees d'etudes cliniques et neuropsychologiques.
Pillon, B. (1972). Etude anatomoclinique d'un cas de Journal de Psychologie Normale et Pathologiqie, 66, 133-
prosopagnosie. Revue Neurologique, 126, 329-346. 166.
Rubens, A. B., and Benson, D. F. (1971). Associative visual
Macrae, D., and Trolle, E. (1956). The defect of function in agnosia. Archives of Neurology (Chic.), 24, 305-316.
visual agnosia. Brain, 79, 94-110. Taylor, A., and Warrington, E. K. (1971). Visual agnosia: a
Milian, G. (1932). Cecite morphologique. Bulletin de single case report. Cortex, 7, 152-161.
l'Academie de Medecine, 107, 664-666. Tzavaras, A., Hecaen, H., and Le Bras, H. (1970). Le prob-
Milner, B. (1966). Amnesia following operation on the leme de la specificite du deficit de la reconnaissance du
temporal lobes. In Amnesia, pp. 109-133. Edited by visage humain lors des lesions hemispheriques unilaterales.
C. W. M. Whitty and 0. L. Zangwill. Butterworths: Neuiropsychologia, 8, 403-416.
London. Warrington, E. K., and James, M. (1967). An experimental
Milner, B. (1968). Visual recognition and recall after right investigation of facial recognition in patients with uni-
temporal-lobe excision in man. Neuropsychologia, 6, 191- lateral cerebral lesions. Cortex, 3, 317-326.
209. Wilbrand, H. (1892). Ein Fall von Seelenblindheit und
Newcombe, F., and Russell, W. R. (1969). Dissociated visual Hemianopsie mit Sectionsbefund. Deutsche Zeitschrift fiir
perceptual and spatial deficits in focal lesions of the right Nervenheilkunde, 2, 361-387.
hemisphere. Journal of Neurology, Neurosurgery, and Yin, R. K. (1970). Face recognition in brain injured patients
Psychiatry, 32, 73-81. a dissociable ability? Neuropsychologia, 8, 395-402.
Downloaded from http://jnnp.bmj.com/ on March 13, 2015 - Published by group.bmj.com

The anatomical basis of


prosopagnosia
J. C. Meadows

J Neurol Neurosurg Psychiatry 1974 37: 489-501


doi: 10.1136/jnnp.37.5.489

Updated information and services can be found at:


http://jnnp.bmj.com/content/37/5/489

These include:

Email alerting Receive free email alerts when new articles cite
service this article. Sign up in the box at the top right
corner of the online article.

Notes

To request permissions go to:


http://group.bmj.com/group/rights-licensing/permissions

To order reprints go to:


http://journals.bmj.com/cgi/reprintform

To subscribe to BMJ go to:


http://group.bmj.com/subscribe/