CHAPTER 14
Two Visual Streams:
Neuropsychological Evidence
DAVID P. CAREY
he science of goal-directed movement has always been multidisciplinary.
Recently, scientists from motor control and kinesiology have developed substantial
interest in the neurobiology of reaching in nonhuman primates and in patients with
brain damage and deficits in looking and reaching. The two-visual systems model of
posterior neocortex has enjoyed considerable popularity outside of clinical
neuropsychology and certainly has generated substantial praise (in parallel with
criticism) within it. After more than 15 years, a reappraisal of one variant of the model
is well overdue. This particular variant is especially relevant for scientists interested in
sensorimotor control and its interface with cognition. In this chapter I argue that early
concerns about the overreliance of the model on one patient with visual agnosia have
been addressed through new work on optic ataxia. Some of this latter research on
dorsal stream function and dysfunction, inspired by the two-visual systems theory, is
now extending beyond the limits of the model.
Two Visual Pathways in the Cerebral Cortex
Work in the 1960s by Trevarthen, Ingle, Schneider, and others anticipated the heavily
cited two visual systems hypothesis of Ungerleider and Mishkin (1982) and its
subsequent revision by Milner and Goodale (1995; Goodale & Milner, 1992, Milner &
Goodale, 2008).1 In fact, several works from clinical
Jeannerod (1994) also suggested a variant of the model that differs from that of Milner and Goodale mainly in how much
action representation is represented in the dorsal stream (see Jeannerod, 1994). In addition, Glover and Dixon have a variant
that subdivides motor functions into planning and control (see Glover, 2004; Glover & Dixon, 2001).
T
neurology (e.g., Kleist, 1934, cited in Grusser & Landis, 1991) and neuroanatomy also
suggested a distinction between spatial and object vision (e.g., Glickstein, May III, &
Mercier, 1985). Nevertheless, the Ungerleider and Mishkin and in particular the Milner
and Goodale models have become the most influential. The purpose of the present
chapter is to review these alternatives in terms of the two decades of research on
patients with brain damage that have passed since the initial formulations.2
The two visual streams (see figure 14.1) have been discussed so extensively in visual
neuropsychology that only a brief precis is provided here. The ventral chain of areas in
the occipitotemporal cortex plays crucial roles in the processing of perceptual
information about form, faces, color, and other visual attributes crucial for recognition
and perception of visual stimuli. The distinction between the two variants of the twovisual
systems hypothesis is in what the dorsal stream of the occipitoparietal cortical
regions is specialized for. In Ungerlieder and Mishkin, spatial vision, including
judgments of spatial attributes (such as which of two targets is closer to a landmark)
and the use of spatial information to guide movement, is crucial, where as in the
Milner and Goodale account, the use of any visual attribute (not just spatial position
such as distance or orientation) for the guidance of sensorimotor acts is crucial.
Early Arguments Against the Milner and Goodale Account
Critics of any two-visual systems account frequently argue that cross talk between
visual cortical areas in both streams invalidates these simple dichotomies. However,
these critics often ignore much of the excellent anatomy performed by Ungerleider and
colleagues (e.g., Boussaoud, Desimone, & Ungerleider, 1992; Ungerleider & Mishkin,
1982), which has stood the test of time. The two pathways are relatively independent of
one another—they are interconnected much more heavily within stream than across
stream (e.g., Catani, Jones, Donato, & ffytche, 2003; Felleman, Xiao, & McClendon,
1997; Galletti et al., 2001; Shipp, Blanton, & Zeki, 1998) and mainly communicate with
one another via the superior temporal polysensory area, the frontal eye fields, and a
few other subdivisions of the premotor and prefrontal cortex (Boussaoud, Ungerleider,
& Desimone, 1990; Young, 1992). The two streams also have relatively distinct targets
in the frontal lobes (Grol et al., 2007; Petrides & Pandya, 2007; Tomassini et al., 2007).
2
Many neuroimaging and behavioral studies that speak to these issues will not be addressed in this chapter (see Culham,
Gallivan, Cavina-Pratesi, & Quinlan, 2008; Culham, Cavina-Pratesi, & Singhal, 2006; Konen & Kastner, 2008), but in my eyes
the key issues are related to imaging studies that show superior parietal activation in the absence of covert or overt motor
responses. Behavioral studies on the dissociations between illusion and action as well as the effects of delay on sensorimotor
responses are also important but cannot be addressed here.
A second critique of two-visual systems accounts is that any idea involving only two
visual streams must be grossly oversimplified and is at best a crude heuristic. It is
difficult to argue with the former statement, but the latter (the crudeness of the
heuristic) can be addressed by empirical evidence. In fact, in spite of all of the caveats
associated with generalizing from damaged to neurologically intact brains, the
evidence suggests that the heuristic is anything but crude. Furthermore, the staunchest
critics of two- visual systems models must acknowledge that these models have
inspired several interesting, challenging, and productive lines of experimentation,
whatever the precise veracity of the original ideas that drove them.
► Figure 14.1 An anatomically detailed variant of the primate two-visual systems schematic. The back of the brain is to the left of the figure.
Lighter gray boxes represent distinct cortical areas of the dorsal stream (labeled by this group as where); darker gray boxes represent areas of the
ventral stream (what). Areas of integration in the superior temporal sulcus (STS) and prefrontal cortex are also depicted.
"Pathways for motion analysis: Cortical connections of the medial superior temporal and fundus of the superior temporal visual areas in the
macaque," D. Boussaoud, L.G. Ungerleider, and R. Desimone, Journal of Comparative Neurology Vol. 296,1990, pgs. 462-495. Copyright 1990.
Adapted by permission of John Wiley & Sons, Inc.
Parietal-"where"
Much of the most compelling
formulation comes from patients with optic ataxia and from one well
D.F., who exhibited visual form agnosia. In effect, numerous experiments with D.F.
suggested that her perceptual system
monoxide poisoning, such that she could no longer read (alexia), recognize familiar
faces (prosopagnosia), or identify common objects (visual agnosia). Her variant of this
latter disorder extended to making discriminat
areas but different shapes, the so
In 1991, we (Goodale, Milner, Jakobson, & Carey) described how, despite serious
difficulties with size judgments and with judgm
be rotated in the picture plane, D.F. demonstrated move
completely sensitive to these visual attributes. When working with the rectangles that
were equal in area but different in aspect ratio,
perceptually, her maximum grip aperture was perfectly scaled to the size of the object
while still on the move to grasp it. Similarly, when we required D.F. to use a handheld
card to show us the orientation of a slot
trial, she performed extremely poorly. However, if instead we asked her to post her
card in the slot as she would a letter in a postbox,
orientation well before the card m
examples show how D.F could use visual
Figure 14.2 The perception-action dissociation in D.F., a visual agnosic patient, (a) The stimulus, which was a
handheld card, and a sample visuomotor posting trial. For perceptual orientation matching, participants rotated
the handheld card to match the slot's orien
tasks, with correct performance normalized to the vertical (indicated by the arrow).
Adapted, by permission, from A.D. Milner and M.A. Goodale, 1998, "Visual brain in action,"
4-12-milner.html.
3
Mailbox for the North American reader
Much of the most compelling evidence cited in support of the Milner and Goodale
formulation comes from patients with optic ataxia and from one well-studied patient,
D.F., who exhibited visual form agnosia. In effect, numerous experiments with D.F.
suggested that her perceptual systems were severely compromised by carbon
monoxide poisoning, such that she could no longer read (alexia), recognize familiar
faces (prosopagnosia), or identify common objects (visual agnosia). Her variant of this
latter disorder extended to making discriminations among simple rectangles of equal
ferent shapes, the so-called visual form agnosia (Benson & Greenberg, 1969).
In 1991, we (Goodale, Milner, Jakobson, & Carey) described how, despite serious
difficulties with size judgments and with judgments of orientation of a slot that could
be rotated in the picture plane, D.F. demonstrated movements that were still
completely sensitive to these visual attributes. When working with the rectangles that
were equal in area but different in aspect ratio, although D.F. could not tell them apart
perceptually, her maximum grip aperture was perfectly scaled to the size of the object
while still on the move to grasp it. Similarly, when we required D.F. to use a handheld
card to show us the orientation of a slot in a picture plane that we rotated from trial to
trial, she performed extremely poorly. However, if instead we asked her to post her
card in the slot as she would a letter in a postbox,1 she rotated the card to match the
orientation well before the card made contact with the slot (see figure 14.2). These
examples show how D.F could use visual
action dissociation in D.F., a visual agnosic patient, (a) The stimulus, which was a
handheld card, and a sample visuomotor posting trial. For perceptual orientation matching, participants rotated
the handheld card to match the slot's orientation. (b) D.F.'s performance versus a control performance on the two
tasks, with correct performance normalized to the vertical (indicated by the arrow).
by permission, from A.D. Milner and M.A. Goodale, 1998, "Visual brain in action," Psyche 4(12). Available: http://psyche.cs.monash.edu.au/v4/psyche
Mailbox for the North American reader
evidence cited in support of the Milner and Goodale
studied patient,
D.F., who exhibited visual form agnosia. In effect, numerous experiments with D.F.
s were severely compromised by carbon
monoxide poisoning, such that she could no longer read (alexia), recognize familiar
faces (prosopagnosia), or identify common objects (visual agnosia). Her variant of this
ions among simple rectangles of equal
(Benson & Greenberg, 1969).
In 1991, we (Goodale, Milner, Jakobson, & Carey) described how, despite serious
ents of orientation of a slot that could
ments that were still
completely sensitive to these visual attributes. When working with the rectangles that
although D.F. could not tell them apart
perceptually, her maximum grip aperture was perfectly scaled to the size of the object
while still on the move to grasp it. Similarly, when we required D.F. to use a handheld
in a picture plane that we rotated from trial to
trial, she performed extremely poorly. However, if instead we asked her to post her
she rotated the card to match the
ade contact with the slot (see figure 14.2). These
action dissociation in D.F., a visual agnosic patient, (a) The stimulus, which was a
handheld card, and a sample visuomotor posting trial. For perceptual orientation matching, participants rotated
tion. (b) D.F.'s performance versus a control performance on the two
http://psyche.cs.monash.edu.au/v4/psyche-
attributes such as orientation and size to control her hand actions and yet, when asked to
show us what she perceived, could not function normally.
Some critics of the two-visual systems account of Milner and Goodale have suggested that
the model is overly reliant on data from the single case of D.F. (e.g., Ungerleider & Haxby,
1994). Of course, if D.F. is a valid instance of a patient with a compromised ventral stream,2
then evidence for the double dissociation with deficits and spared abilities of patients with
compromised dorsal streams is logically possible. In a nutshell, double dissociations show
that task A isn't failed in a single case patient because it is more difficult than task B (which
patient A can do), because a different patient passes task A and fails task B (see Dunn &
Kirsner, 2003).
Double Dissociations in Perception and Action
The first claim for a double dissociation between impairments after lesions to the dorsal and
ventral streams appeared in 1994. Goodale and colleagues (Goodale, et al., 1994) contrasted
the ability of D.F. (compromised ventral stream, intact dorsal stream) to reach and grasp
irregularly shaped pebbles (see figure 14.3 for examples) with that of patient R.V. (originally
described in Jakobson, Archibald, Carey & Goodale, 1991) experienced optic ataxia as part of
Balint-Holmes syndrome. Optic ataxia is a disorder of reaching to visual targets that cannot
be explained by low-level visual deficits. It is thought to be a consequence of damage in the
superior parietal lobe (De Renzi, 1982; Husain & Stein, 1988; Ratcliff & Davies Jones, 1972;
Rondot, De Recondo, & Ribadeau Dumas, 1977; although see Karnath & Perenin, 2005). The
parietal lobe regions damaged in patients with optic ataxia include several important dorsal
stream areas implicated in the control of eye and hand movements. R.V. had a compromised
dorsal stream but a largely intact ventral stream.
Goodale and colleagues (Goodale et al., 1994) found that D.F. was extremely poor at making
same-different judgments (under conditions of free viewing) about irregularly-shaped
"pebble" stimuli (see figure 14.3), while R.V. was virtually flawless at this task. In other
words, R.V.'s visuoperceptual performance was relatively intact, while D.F.'s was not. The
opposite pattern was obtained when the participants were required to grasp the pebble
stimuli between their index fingers and thumbs. Figure 14.3 is a schematic representation of
the performance of R.V. and D.F. in terms of selected grasp points (as depicted by grasp lines
that show the orientation of the index finger and thumb before contact). As can be seen, D.F.'s
grasp
1
The James, Culham, Humphrey, Milner, and Goodale (2003) fMRI study of D.F. suggests that some portionsf her ventral
stream are intact, more than a decade after her accident. The most extreme structural damage is to the lateral occipital
cortex (area LO), which recent functional image studies (e.g., Kourtzi, Erb, Grodd,1 Biilthoff, 2003) have implicated as
being crucial for perceptual encoding of shape.
points tended to intersect the pebble's center of mass,1 a finding that suggests that her dorsal
stream was perfectly capable of generating effective grasps using visual information that
could not be used for perceptual judgements (Goodale et al., 1994).
The second type of double dissociation between optic ataxia and visual agnosia that supports
the two-visual streams model depends on a theoretiFigure
14.3 Grasp lines preceding object contact by D.F., who demonstrates visual agnosia; R.V., who demonstrates optic ataxia; and
a matched control. Grasp lines of the control and D.F. tend to intersect near the center of mass of each object. While R.V. had no
difficulty making same-different judgments about these objects, she clearly did not pick them up normally. On the contrary, D.F.'s
grasps were indistinguishable from the control's, in spite of her difficulty in telling any two pebbles apart.
Reprinted from Current Biology, Vol. 4, M.A. Goodale et al., "Separate neural pathways for the visual analysis of object shape in perception and prehension,"
pgs. 604-610. Copyright 1994, with permission of Elsevier.
1
Monika Harvey, David Milner, and I have argued elsewhere that using the principal axis to guide grasping (i.e., grasping the
perpendicular axis) would produce roughly the same results, at least with these types of stimuli (Carey, Harvey, & Milner, 1996).
cal property of the dorsal stream: Its default modus operandi is to work in real time with
visual information for the guidance of eye, head, and hand movements. In D.F., for example,
sensitivity to both target size (her grip scaling was eliminated; Goodale, Jakobson, & Keillor,
1994) and location (her pointing accuracy dropped appreciably; Milner, Dijkerman, & Carey,
1998) was dramatically reduced when we required her to move after a delay. Paradoxically,
these same delays may improve sensorimotor responses for patients with optic ataxia.
Milner and colleagues (Milner, Paulignan, Dijkerman, Michel, & Jean- nerod, 1999)
investigated the effects of delay on sensorimotor responses of patients who misreach due to
optic ataxia. In one study, patient A.T., who experienced bilateral parietal lobe damage 12
years previously (Jeannerod, Decety, & Michel, 1994) was required to point to peripheral
targets immediately or 5 s after they were extinguished. Five control subjects performed
somewhat worse in the delayed condition; A.T. performed substantially better. In a second
study, patient I.G. (who also demonstrated optic ataxia) was tested on her ability to scale the
size of her grip to different targets in immediate conditions versus a delay of 5 s, which
required a pantomimed grasp (as used by Goodale and colleagues with D.F.) or a delayed
real grasp. Like A.T., I.G. demonstrated an improvement after delay. The authors also found
that I.G. was better when she pantomimed grasps than when she had to wait 5 s and then
grasp (delayed real grasping). During delayed real grasping, presumably some dorsal stream
impairments attenuated the good skills that could be driven by ventral stream mechanisms.
Additional evidence for this interpretation is that I.G. tended to scale her grip to the
originally presented target when, in delayed real grasping trials, the target was replaced with
an object of a different size. Control participants never did this (Milner, Dijkerman, Mcintosh,
Rossetti, & Pisella, 2003). The authors concluded that A.T. and I.G. could rely on a
representation of target position or size mediated by the ventral stream to guide their actions
after the compromised dorsal system response was not utilized. D.F., the previously
described patient with visual agnosia, had no such representation to fall back on to guide her
motor responses. Therefore, her motor skills were substantially compromised by even
relatively short delays (Goodale, Jakobson, & Keillor, 1994). This line of research was
reviewed by Milner and colleagues in 2003 (Milner, Dijkerman, Mcintosh, Pisella, & Rossetti,
2003).
Later Controversies: Diagnosing Optic Ataxia
The use of double dissociation between perception and action in optic ataxia and visual
agnosia to support two-visual systems models is not without controversy. First, diagnosis in
neuropsychology is not based exclusively on a particular neurochemistry, neuropathology,
genetic marker, or other
incontrovertible piece of hard medical evidence. The diagnosis of visual agnosia, and more
specifically visual form agnosia, can depend on interpretation of the magnitude of
recognition dysfunction juxtaposed with the degree of associated low-level visual
disturbance. How poor does a patient have to be at recognizing Snodgrass and Vanderwart
line drawings (a standard set of drawings whose psychometric properties are well described;
Snodgrass & Vanderwart, 1980) before the diagnosis is made? If patients fail low-level visual
subtests of standardized instruments such as the Visual Object and Space Perception Battery
(VOSPB; Warrington & James, 1991) or the Birmingham Object Recognition Battery (BORB;
Riddoch & Humphreys, 1993), could we reasonably expect them to be capable of normal
object recognition? If not, visual agnosia should not be diagnosed.
For optic ataxia, how exaggerated does misreaching have to be (particularly in peripheral
vision) before we can be confident in a diagnosis? Many of the studies examining this
disorder, with only a few recent exceptions, include no control subjects matched in age or
visual status. When control subjects are included, the control sample is small (and there are
some serious statistical issues as a consequence; see Crawford & Garthwaite, 2005) and the
selection mechanism is unspecified—often the controls are highly functioning individuals
from a university community who might be not be comparable with patients with stroke,
even when matched for age and sex. In some cases hand- or field-specific deficits (which of
course must be consistent with lesion laterality) may allow patients to serve as control
participants for themselves. However, these studies usually only describe the relatively intact
performance in the unaffected arm or field—they don't typically measure it.
A second caveat for experimental work on patients with optic ataxia is that the most florid
form, in which misreaching is observed even for fove- ated targets, occurs as part of BalintHolmes
syndrome, a complex disorder with at least a triad of symptoms. These other
symptoms, including atten- tional and gaze difficulties, could be indirectly responsible for
the misreaching, or the ataxia could be an independent symptom that depends on a different
component of the large bilateral occipitoparietal lesions. Accurate eye tracking in the elderly
is difficult, and when it has been done, details on calibration (which requires accurate eye
movement and good control of fixation), recalibration frequency, system accuracy, and drift
correction are suspiciously absent from the method sections of the reports.
The third caveat regarding optic ataxia is a critical but typically omitted demonstration: The
misreaching deficits in patients who have been diagnosed with optic ataxia by definition are
restricted to deficits in moving to visual targets. As I have argued elsewhere (Carey, 2004),
auditory and proprioceptive guidance of movement are rarely assessed in these patients in
any formal way For example, magnetic misreaching (a very rare disorder in which patients
slavishly reach to the fixation point rather than an
extrafoveal target; Carey, Coleman, & Delia Sala, 1997; see also Buxbaum & Coslett,
1997,1998; Jackson, Newport, Mort, & Husain, 2005; van Donkelaar & Adams, 2005) has been
labelled as a variant of optic ataxia by Buxbaum and Coslett as well as Jackson and
colleagues, but we have evidence that, in our patient at least, poor guidance of limb
movement is not restricted to the visual modality. Blanghero andcolleagues (2007) have made
similar observations in two patients previously described in the literature as demonstrating
optic ataxia.
The fourth caveat relates to the distinction between foveal and nonfoveal variants of optic
ataxia, which is becoming blurred or even ignored (perhaps because the former are so rare, it
is becoming convenient to discuss the extrafoveal variant as the prima facie case).
Misreaching in the full-blown foveal variant is so fascinating because the errors happen in
spite of accurate gaze, which also in a sense serves as a crude control of visual status: The
patient sees the target well enough to fixate on it. The same cannot always be said when
targets are presented well into the visual periphery, which is one of the reasons why I worry
about the visual status of control participants. Of course, conducting clinical experiments
with the more common extrafoveal cases means ensuring fixation compliance, with all of the
problems related to monitoring eye position briefly outlined earlier.
How do these concerns constrain the usefulness of any particular patient for evaluating
cortical models of perception-action dissociation? I have little doubt that, in spite of her
diffuse brain damage, D.F.'s recognition problems are restricted to the visual domain (for
example, I tested her haptic recognition abilities in the 1990s and they were effectively
normal). In spite of several attempts, other patients with a similar agnosia and perceptionaction
profile have been difficult to find. A patient who had experienced meningoencephalitis
in childhood and subsequently demonstrated relatively intact visuomotor skills
in conjunction with poor visuoperceptual abilities has been described (Le et al., 2002).
Nevertheless, he differs from D.F. in several important respects, including visuoperceptual
matching skills and a lifetime of development that was not afforded to D.F., who acquired
her disabilities in adulthood. Similarly, a few patients with agnosia have shown, similarly to
D.F., dependence on binocular vision for grasping, although these people do not share many
features of D.F.'s unusual profile. Recently Yang, Wu, and Shen (2006) described an anoxic
patient who may have demonstrated a similar perception-action profile, but that report
focussed on implicit processes related to shape and not sensorimotor skills per se. Goodale,
Wolf, and others (Goodale et al., 2008; Wolf et al., 2008) have recently described a patient
with visuoperceptual problems even more severe than those of D.F., who has intact grasping.
Very recently, Karnath, Riiter, Mandler, and Himmelbach (2009) have found a patient with a
ventral stream lesion who shows a very similar perception-action profile to D.F. using Efron
square matching and grasping, as well as orientation posting and matching.
The difficulty in finding other patients with visual agnosia and such a pure dissociation has
not hindered research inspired by the Milner and Goodale variant of the two-visual systems
theory Much of the recent work in support of their model has come from studies of
neurologically intact samples. Work on dissociations in perception and action has proven to
be fruitful as well as contentious (see Bruno, 2001; Carey, 2001; chapter 13 of this volume for
reviews of dissociations in perceptual and action systems in response to visual illusions;
although see Biegstraaten, de Grave, Smeets, & Brenner, 2007, for a contrary view). In
parallel, claims about the rather short-term memory of sensorimotor systems (Goodale, et al.,
1994) mediated by dorsal stream mechanisms have remained rather unchallenged.
One patient study (Himmelbach & Karnath, 2005) attempted to evaluate whether the effects
of delay in patients represent a step function (the strongest support for a switch in processing
mode from a sensorimotor representation to a more enduring, less metric perceptual
representation) or a continuous drop (or improvement) in performance. These authors found
some evidence for improvement with delay in two patients with optic ataxia, although this
effect was relatively continuous and therefore did not suggest a step change indicative of a
switch from impaired dorsal representation to intact perceptual representation. However, the
two cases described in this study are not without controversy. First, no details beyond mean
age were provided on the elderly control participants. Given the acuteness of the patients, a
nonataxic control group with similar visual status or neglect might have been appropriate.
Second, one of the patients (U.S.) had profound right-sided neglect, but we were not told of
her error rates in right and left sides of space. In addition, monitoring fixation in these
patients must have been extremely difficult (how was the infrared eye tracker calibrated in
patients with potential difficulties in eye movement and hemispatial neglect?). Third, these
cases were tested in a very acute phase of the patients' illness (6 and 8 d poststroke). In
neuropsychological syndromes such as hemispatial neglect, patients with left-brain damage
can show large right neglect (e.g., De Renzi, 1982) that often resolves very quickly and is
made of very little by researchers. Perhaps optic ataxia, like neglect after left-brain lesions, is
easier to diagnose in acute patients (although Himmelbach & Karnath do not say so
explicitly). If this is the case, a more specific control group is probably more appropriate.
Optic ataxia and delay have also been linked in experiments by Schndler and coworkers
(2004) and Rice and colleagues (2008). In the first study, Schindler and coworkers found that
two patients with optic ataxia were impaired in adjusting their reach trajectories when two
obstacles were moved in the work space from trial to trial. In the follow-up study, Rice and
colleagues (2008) showed that introducing a delay between stimulus presentation and
response abolished this failure. In other words, the patients showed trajectories that were
appropriate for the obstacle locations in the
same manner as control participants did in immediate and delayed conditions. These effects
represent a third example of how optic ataxic patients behave in ways consistent with the
Milner and Goodale account.
Recently colleagues of David Milner have changed their tune somewhat by criticizing the
use of patients with optic ataxia as the appropriate counter to patient D.F. (Pisella, Binkofski,
Lasek, Toni, & Rossetti, 2006). One of their claims is that the deficits studied in peripheral
vision in the optic ataxia cases are not strictly comparable to D.F., who performs with full
foveal vision in conditions where perception and action responses are contrasted with one
another. The critique is not strictly true of course: D.F. shows similar perception-action
dissociations when forced to work with stimuli in peripheral vision. We have shown that D.F.
reaches to target positions in peripheral vision with accuracy similar to that of controls
(although interposing short delays before the response seriously impairs this skill; Milner,
Dijkerman, & Carey 1998). Figure 14.4 (from Milner & Goodale, 2008) shows data contrasting
the performance of D.F. with that of A.T. (data from Milner et al., 1999). And there are
examples of patients with optic ataxia operating in central vision who show the perceptionaction
distinction, such as V.K., who we described in 1991 (Jakobson, Archibald, Carey, &
Goodale, 1991).
A final study worthy of note questions the assumptions that the superior parietal lobule is the
crucial site of the lesion that produces optic ataxia in humans (Karnath & Perenin, 2005). The
authors of this study used a lesion overlap method (Rorden & Brett, 2001) with a large sample
of patients screened for optic ataxia and found that the crucial regions of overlap did
Figure 14.4 Reaching into the periphery in (a) D.F., a patient with agnosia (from Milner, Dijkerman, & Carey, 1998), and (b)
A.T., a patient with optic ataxia (from Milner et al., 1999). Note that worse performance is higher on the y-axis (absolute error
= unsigned error of end point from actual target position). The scales have been modified to show the patterns from the two
separate studies.
Adapted from Neuropsycholgia, Vol. 46, A.D. Milner and M.A. Goodale, 2008, "Two visual systems re-viewed," pgs. 774-785. Copyright 2008, with
permission of Elsevier.
not center on the superior parietal lobule, as suggested by early studies (e.g., Ratcliff &
Davies-Jones, 1972) and the Milner and Goodale account. Instead, they found that lesion
overlap centered on a region around the border of the inferior parietal lobe with
occipital cortex. The precise anatomy of optic ataxia is not crucial for the Milner and
Goodale model in a functional heuristic sense, but there are suggestions in the literature
from neuroimaging and patient work that this region may be utilized in nonperceptual
operations. This controversy warrants additional research.
Summary and Future Directions
In spite of these unresolved issues, the identification of several patients with
extrafoveal optic ataxia, such as R.V., A.T., and I.G., and the first round of experiments
inspired by the two-visual systems model of Milner and Goodale have reintroduced this
category of dysfunction to contemporary neuroscience. For example,
neurophysiological models of eye-hand transformations, inspired by data from singleunit
studies in nonhuman primates (e.g., Andersen, Snyder, Batista, Bueno, & Cohen,
1998; Chang, Dickinson, & Snyder, 2008), have inspired recent experiments on the
nature of eye- hand coordination deficits in neurologically intact participants as well as
in patients with optic ataxia. These latter studies may never have occurred if the twovisual
systems debate on dorsal and ventral streams had not reintroduced this
interesting type of patient to the mainstream literature.
For example, studies by Crawford, Khan, and colleagues (e.g., Khan et al., 2007) have
supported the idea that hand movements may be coded in eye-centered coordinate
schemes. Some of these models may be limited because they do not necessarily
acknowledge a specific role of foveation in eye-hand coordination (and retinal
coordinates will do in many of these hierarchical accounts; Carey, Ietswaart, & Delia
Sala, 2002). Nevertheless, the eye-centered accounts have inspired additional studies of
patients and neurologically intact participants from a sensorimotor control perspective
(c.f., Dijkerman et al., 2006; Scherberger, Goodale, & Andersen, 2003;Verha- gen,
Dijkerman, Grol, & Toni, 2008). These studies will inevitably suggest greater
specifications for simple two-stream accounts. Indeed, the two- visual systems models
may fade from the neuropsychological literature as the specifics become elaborated. Of
course, the indebtedness of those studies of the future to two-visual systems theory will
remain. Ideas about a monolithic multi-feature representation of the visual world (see
Goodale, 1983) have gradually been laid to rest, thanks in no small part to the literature
referred to in this chapter.
In conclusion, the patient work related to the two-visual systems account of Milner
and Goodale has grown dramatically in the past 15 years or so since the model first
came to the attention of neuroscientists. Although
contentious elements remain and some of the details have required minor rethinks (c.f.,
Milner & Goodale, 2008), for the most part the model has stood the test of time. What is
of little debate is the amount of directed research that it has driven or at least inspired.
The growing recognition of sensorimotor processes and their rightful place in cognitive
neuroscience has been a long time coming.