Exp Brain Res (2010) 200:3–24

DOI 10.1007/s00221-009-1914-2

REVIEW

The blindsight saga

Alan Cowey

Received: 24 May 2009 / Accepted: 16 June 2009 / Published online: 1 July 2009© Springer-Verlag 2009

Abstract Blindsight is the ability of patients with clini-cally blind Weld defects, caused by damage to the primaryvisual cortex V, to detect, localise and even discriminatevisual stimuli that they deny seeing. Blindsight tells usmuch about the nature of visual processing in the absenceof the primary visual cortex and is a paradigmatic exampleof implicit knowledge. It has attracted widespread interestand debate amongst philosophers, cognitive neuropsychol-ogists and visual neuroscientists. Its downside is that possi-ble artefacts abound, much more so than with examples ofimplicit memory or deaf hearing and numb touch. Unfortu-nately the artefacts are still frequently ignored, or dismissedas captious, with the result that many of the genuine quali-ties of blindsight remain uncertain. Now that blindsight inmonkeys has been established the substantial literature onthe eVects of removing parts or all of V1 in monkeys on theresidual physiological cerebral responses to visual stimuliin their Weld defects is at last directly relevant to humanblindsight. Whether blindsight is, or could be, useful ineveryday life is the next unsolved problem.

Keywords Blindsight · Consciousness · Awareness · Guessing · Visual areas

Introduction

The engaging oxymoron ‘blindsight’ has survived for the35 years since it was coined (Sanders et al. 1974; Weiskrantzet al. 1974), although Shakespeare might claim precedencewith his triple oxymoronic “Blind sight, dead life, poormortal living ghost, …” (Richard III, Act IV, scene IV).Over the years philosophers (e.g. Hyman 1991; Bennett andHacker 2003) criticised the term on the grounds that it pro-moted a nonsensical distinction between sensing somethingand monitoring those sensations in order to perceive them.Some scientists argued that if a subject can correctly andvoluntarily respond to a visual stimulus the subject cannotbe blind. And others dismissed it as an artefact of slipshodexperimentation. It is unfortunate that this continuing andvexatious wrangle, at times amounting to a diatribe, drewattention away from a fascinating neuroscientiWc problem,namely how the brain continues to process visual informa-tion in the absence of both acknowledged visual perceptionand the primary visual cortex that retinotopically corre-sponds to the relevant part of the visual Weld, whileenabling the subject voluntarily and correctly to respond tothe unseen stimuli. If blindsight had instead been calledresidual visual processing after destruction of primaryvisual cortex—a topic that had existed already for decadesin research on animals—it is likely that much of its contro-versial nature would not have arisen. But we might havemissed the essence and the fascination of the phenomenon:how can we voluntarily respond to visual events that wedeny seeing?

So what is blindsight? It is the ability of patients withabsolute, clinically established, visual Weld defects causedby occipital cortical damage to detect, localize, and dis-criminate visual stimuli despite being phenomenally visu-ally unaware of them (Pöppel et al. 1973; Richards 1973;

A. Cowey (&)Department of Experimental Psychology, University of Oxford, South Parks Road, Oxford OX1 3UD, UKe-mail: alan.cowey@psy.ox.ac.uk

123

4 Exp Brain Res (2010) 200:3–24

Sanders et al. 1974; Weiskrantz et al. 1974). The ‘blind’ inblindsight depends on the patients’ verbally reporting ‘no’when they are asked if they saw a visual stimulus but the‘sight’ is demonstrated by forced-choice procedureswhereby the patients have to guess where or what the stim-ulus is even though they deny seeing it. Typically thechoice is binary, e.g. was the stimulus red or green, station-ary or moving, horizontal or vertical, high or low in thevisual Weld? Performance can be impressive, especiallywith stimuli that move or have high luminance contrastwith their background, and scores of nearly 100% correcthave been reported, as shown in Fig. 1, which epitomisesblindsight. But much lower scores are commoner withstimuli that are coloured or sparsely contoured or blurred ordiVer only in shape, or texture. This survey explores howblindsight might be achieved and what it tells us aboutvisual processing, in both patients and monkeys.

The properties of blindsight

For now let it be assumed that blindsight is whatever visuo-motor responses survive the total destruction or deaVerenta-tion of all or some part of the striate cortex, area V1. Thisallows reference to evidence from early neurological stud-ies before the introduction of forced-choice procedures, andfrom monkeys before the Wrst attempts to measure phenom-enal visual awareness in monkeys.

Around the turn of the nineteenth century visual Welddefects caused by occipital damage were widely investi-gated and several studies stressed the total absence ofacknowledged vision in Weld defects caused by stroke orpenetrating head wounds that probably destroyed (asopposed to just damaged) part of striate cortex, now knownas V1. For example, Wilbrand and Sänger (1900) noted thelack of “all and any sensory quality” (“alle und jeglicheEmpWndungsqualität”) and Holmes (1918) stressed thatexcept at the edge of a Weld defect visual stimuli are notseen at all. There were, and still are dissenters, notablyRiddoch (1917) who was the Wrst to argue that a patient canbe blind to visual stimuli except those that move or Xicker,and Zeki (see Barbur et al. 1993) who reached a similarconclusion in a study of patient GY, who certainly has nosurviving V1 corresponding to his almost total hemianopia.Although Riddoch’s patients responded in terms that indi-cate they experienced genuine phenomenal visual motionwe shall never know whether their V1 damage was com-plete or whether during recovery from incomplete occipitaldamage, motion perception recovered before other aspectsof vision. There is no evidence that Riddochs’s moving andstationary stimuli were matched for detectability in the see-ing Weld and this is a general problem that continues to dogthe blindsight literature nearly a century later. Nor is clearthat GY experienced real visual percepts as opposed to ‘afeeling that something happened’, a problem that is alsounresolved.

Visual reXexes

Whatever the resolution of the above problems, it hasalways been clear that subjects who report no visual per-cepts commonly retain reXexive responses to light, whichmight, in theory, underlie their correct forced-choice per-formance. For instance, the pupil still responds to changinglight levels (Magoun and Ranson 1935; Bender and Krieger1951; Brindley et al. 1969) and even to spatial patterns likeisoluminant gratings that yield no overall change in lightlevels (Weiskrantz et al. 1999). In hemianopic monkeystoo, the pupils respond to a stimulus in the blind hemiWeldand, importantly, its variation with spatial frequency closelyresembles that seen in human hemianopia (Weiskrantz et al.1999). As well, the blink reXex to bright light persists(Bender and Krieger 1951; Hackley and Johnson 1996),and the eyes continue to track a unidirectional movingvisual scene, i.e. optokinetic nystagmus or OKN Pizzamiglioet al. 1984; (van Hof-van Duin and Mohn 1983; Pizzamiglioet al. 1984; Heide et al. 1990). However, residual OKN iscontroversial because it persists after unilateral brain dam-age (Heide et al. 1990; van Hof-van Duin and Mohn 1983;Braddick et al. 1992) but could not be evoked in total cortical

Fig. 1 GY’s performance and associated awareness in a discrimina-tion between vertical or horizontal 20° displacement of a target at var-ious speeds. Note that although awareness varies from zero to 100%as speed increases, performance is high at all speeds. Reproduced fromWeiskrantz et al. (1995). Copyright 1995 National Academy ofSciences, USA

123

Exp Brain Res (2010) 200:3–24 5

blindness (Brindley et al. 1969; Perenin et al. 1980; Perenin1991), with the single exception of the patient studied byBraak et al. (1971).

Several studies reported that cortically blind monkeysretained a visual blink reXex and appropriate pupillarychange in response to a bright light but the reports wereunconvincing until the pioneering studies of Klüver (1941)demonstrated them in monkeys in which all or nearly all V1was subsequently shown to have been removed. He waseven able to show that stimulating the central retina wasmore eVective than eccentric stimulation, even though thelatter is more likely to correspond to spared regions of V1in the surgically inaccessible rostral calcarine Wssure. Klü-ver also noted an absence of the blink response to a threat-ening visual gesture, but in presentations of a large andcontrasty looming stimulus in the hemianopic Weld of mon-keys, King and Cowey (1992) readily elicited an avoidanceresponse. The photic blink reXex was subsequently studiedby Pasik and Pasik (1964, 1982) who demonstrated itsindependence of the pupil by paralysing the latter. Theyalso showed a clear relationship between stimulus intensityand the occurrence of blinking (probability of blinking >0.9with the most intense stimuli) and that blinking was not anartefact of heat from bright lights.

OKN was Wrst thoroughly investigated by Pasik andPasik (1964, 1982), with a greater range of stimulus condi-tions than those used by Braak and van Vliet (1963), whodemonstrated that 6 months after histologically veriWedtotal removal of V1, OKN was ‘not essentially diVerentfrom before the operation’. According to the Pasiks, OKNwas present as little as 1 week after bilateral removal ofstriate cortex, with substantial recovery after 1 month,especially for stimulus velocities of 22–45°/s. But the peakfrequency of response was roughly halved, and OKN wasabolished at velocities of 80–90°/s. Like ter Braak and vanVliet, they noted after-nystagmus. Flicker-induced nystag-mus in which monocular stimulation provokes nystagmusin the direction of the stimulated eye, was essentially normal(Pasik et al. 1970).

Forced-choice responding

The visual reXexes just described are easy to induce but theydo not necessarily explain how blindsight might arise. Evenwhen they correlate well with the properties of forced-choiceguessing the correlation could be coincidental. This is whyvoluntary responses are ultimately more informative and,potentially, useful to the patient. They were extensively stud-ied in monkeys long before they were examined in humansubjects with blindsight but their existence was assumed—now evidently wrongly—to indicate residual real sight. Thishistorical sequence is followed here.

Residual visual sensitivity in monkeys

Before the introduction of methods that involved trainingmonkeys to maintain Wxation on a particular part of a dis-play so that visual stimuli could be presented within a uni-lateral Weld defect, several experimenters removed all of V1bilaterally, or attempted to do so. In patients this—fortu-nately rare—event renders the patient totally corticallyblind and presumably the same is true of monkeys. How-ever, monkeys with no V1 can learn to respond to, and dis-criminate between, diVerent visual stimuli, usually byhaving to touch them or avoid them (Klüver 1941; Pasikand Pasik 1982; Weiskrantz 1963; Weiskrantz et al. 1977;Keating 1975). They could discriminate between stimulidiVering in luminous Xux (Klüver 1941; Schilder et al.1971), brightness (Schilder et al. 1971) orientation (Keating1975) and the extent of their achromatic contrast borders(Weiskrantz 1963).

Wavelength discrimination has always been dubious.Successful discrimination was reported by Schilder et al.(1972), Pasik and Pasik (1982) and Keating (1979), whoshowed that Wve monkeys relearned a red versus green dis-crimination despite substantial and randomized variation inthe intensity of the broad-band stimuli following totalremoval of V1 (subsequently veriWed histologically). Evenafter additional removal of extra-striate cortex that includedareas V2, V3, V4, TEO, and even caudal IT, four of themonkeys reached at least 80% correct, and those who wereswitched to blue versus green achieved 70–85% correct. Sowhy the doubts? If the spectral sensitivity of monkeyslacking V1 is shifted towards scotopic levels, even a pho-tometer would have registered the longer-wave stimulus(red) as dimmer than the green, and the green as dimmerthan the blue at all except the physically dimmest longer-wave stimulus. In accordance with this interpretation,Humphrey (1974) found no evidence that what to a normalobserver would be red and green were discriminable, irre-spective of intensity, by his famous cortically blind monkeyHelen. By varying the ‘brightness’ of the green stimulusHumphrey found an equivalence point where her discrimi-nation collapsed. Results from Malmo (1966) and Leporéet al. (1975) support this interpretation, for after total bilat-eral removal of V1, peak spectral sensitivity under light-adaptation was almost the same as under dark-adaptation,i.e. about 500 nm. This apparent extreme diminution ofcone function contrasts with the clear indication of a Pur-kinje-shift in the hemianopic monkeys studied by Coweyand Stoerig (1999). The diVerence between partial and totalremoval of V1 could underlie the latter diVerence as themonkeys of Malmo (1966) and Leporé et al. (1975) hadextensive bilateral lesions that strayed beyond V1, whereasthose of Cowey and Stoerig were hemianopic and withoutextensive damage to extrastriate cortex.

123

6 Exp Brain Res (2010) 200:3–24

The results of shape or pattern discrimination are just ascontroversial. Klüver (1941) showed that a destriated mon-key could discriminate, at 90% correct, between a luminoussquare and 76 small circles of the same total area and Xux.His conclusion that the pattern of the stimulus might deter-mine the response was conWrmed by Weiskrantz (1963)who showed that total length of contour in a stimulus is amajor discriminable feature. Nonetheless, shape discrimi-nation is diYcult in cortical blindness, as demonstrated inan extensive set of experiments by the Pasiks and their col-laborators (Pasik et al. 1969; see Pasik and Pasik (1982) forreview). Even after years of training on a variety of visualtasks macaque monkeys without any V1 required severalthousand trials to relearn a discrimination between a circleand a triangle at 90% correct, and performance wasimpaired by changing the size of the triangle or by invertingit, indicating that they were not responding to triangularityper se. The impoverished shape perception was also dem-onstrated by Dineen and Keating (1981), for only three ofWve monkeys could reach 80% correct when the amount ofcontour, shape of sub-elements and number of corners weresystematically varied. Finally, Humphrey (1974) showedthat the destriated monkey Helen could discriminatebetween several targets that diVered only in shape. How-ever, when the discriminability between each of severaldiVerent shapes and a standard shape was measured andequated, he found that when diVerent shapes of similar dis-criminability from the standard were pitted against eachother the monkey could not tell them apart. For example, acircle was indiscriminable from a triangle. This led Hum-phrey to propose the idea of stimulus salience in blindsight,whereby diVerent stimuli might ‘catch the eye’ to diVerentextents. But if two stimuli have similar salience, even ifgrossly diVerent in normal vision, they will be inseparablein blindsight. This idea, perhaps because it was conceivedbefore its time, is unjustly neglected. Whether Humphrey’sWnding is evidence against any genuine shape discrimina-tion in total cortical blindness in monkeys (interestingly, ithas never been studied in hemianopic monkeys) remainsuncertain. But it is incontrovertible that such monkeys, orhemianopic monkeys, display no evidence of visually rec-ognizing complex objects (such as desirable food or fright-ening toy snakes or a monkey doll) in their cortically blindvisual Welds even after many years of experience (Coweyand Weiskrantz 1963; Humphrey 1974). The latter observa-tions are relevant to current studies of aVective blindnessdiscussed later.

Residual visual discrimination in patients with blindsight

Forced choice non-verbal paradigms, already widely usedwith normal observers by visual psychophysicists, wereintroduced by Richards (1973), who demonstrated a

residual crude stereoscopic mechanism operating in corti-cally blind Welds. With the same patients, Pöppel et al. 1973reported that they could direct their eyes towards theapproximate position of a brieXy presented stimulus theydenied seeing. Sanders et al. (1974) and Weiskrantz et al.(1974) tested several residual visual functions that includedtarget localization and shape discrimination in the famouspatient DB, whose surgical removal of right V1 has unfor-tunately never been veriWed by MRI because of the intra-cranial wound clips. They too used forced-choice methods,and coined the term ‘blindsight’ to describe the paradoxicalsurviving visual properties. The cardinal feature of blind-sight was that the stimuli were not consciously seen; eventhough much later it became clear that patients can beaware that a stimulus has occurred without yielding a visualpercept, i.e. the type 2 blindsight described by Weiskrantzet al. (1995). Blindsight rapidly became a striking exampleof ‘implicit knowledge’.

Within a few years many properties of human blindsighthad been explored by forced-choice guessing. Localising anunseen stimulus was conWrmed and extended (Perenin andJeannerod 1975; Blythe et al. 1987). Blindsight patientscould also detect stationary or moving stimuli interleavedrandomly with blank trials, and discriminate Xicker, stimu-lus orientation, target displacement, and wavelength (seeStoerig and Cowey (1997); Stoerig (2006), for review).Only when shapes had to be discriminated and this couldnot be achieved on the basis of orientation of their contours,was discrimination impossible in patient DB (Weiskrantz1987). However, although another blindsighted patient’sdiscrimination between diVerent rectangular (Efron-type)shapes was also at chance level, the patient’s reaching andgrasping movements to the same unseen stimuli correlatedwell with their shape and orientation (Perenin and Rossetti1996). Subsequently DB’s form discrimination recovered(Trevethan et al. 2007) to an extent that even allowed iden-tiWcation of complex digital images. In what must be one ofthe most remarkable, deeply puzzling, provocative, andstimulating examples of blindsighted performance, DB sub-sequently identiWed 25 of 28 low-contrast outlined Snod-grass Wgures of animals presented in his blind Weld despitenot knowing the category of the stimulus and, even moreextraordinarily, performed better than when the sameimages were presented to his seeing hemiWeld or to normalsubjects, despite saying that he was almost always unawareof anything, i.e. pure blindsight. A similar result was foundwith even more complex coloured images. It is unfortunatethat it is currently impossible to determine whether DBachieves this without surviving V1 and that it is unknownwhether other blindsight patients are in any way similar.

Almost all examples of forced-choice responding requireone of two allowable responses. In a very diVerentprocedure Brown et al. (2007) asked two subjects, one

123

Exp Brain Res (2010) 200:3–24 7

hemianopic the other with quadrantanopia, to adjust theirright-hand Wnger grip to make it appropriate as if reachingfor an unseen object presented in the blind left Weld. Whenthe left hand rested on the lap there was no evidence thatthe right hand’s grip, or estimated size, of the unseen objectwas appropriate. But both were correctly scaled by the righthand when the left hand was placed next to the unseenobject and additional tests showed that this was not just aproduct of selective attention to spatial location. Instead theauthors attribute the result to activation of bimodal visuo-tactile cortical neurons whose receptive Welds track theposition of the hand, which could explain why reaching andpointing are better able to reveal blindsight than are verbalresponses.

The artefact problem and objections to blindsight

In a statement amounting to a hostage of fortune, the exis-tence of blindsight was challenged by Campion et al.(1983) and their trenchant arguments and conclusions areoften recited. Their concerns were fourfold. First, it wasargued that the phenomenally unconscious aspect of blind-sight was “essentially trivial” because normally sightedobservers behave in this way at and around the psychophys-ically determined threshold for detection; second, that itcould depend on islands of spared striate cortex; third, thatlight from the ‘unseen’ stimulus might be detected by itsscatter on to normal parts of the retina; fourth that blind-sight merely reXects the fact that a subject uses diVerent cri-teria for responding to stimuli in the seeing and the blindWelds. All four criticisms continue to be invoked but also tobe ignored in many studies of blindsight, which is unfortu-nate because their criticisms remain important. Given thelatter they are reconsidered here.

Is the unconscious nature of blindsight essentially trivial?

This possibility stems from the view that in blindsight “thestimuli are not unconscious…..all subjects in all studies are,by conventional criteria, aware of stimuli to some extent”(Campion et al. 1983, p 480). The latter part of the state-ment might be correct but it ignores the fact that visualstimuli in the blind Weld can sometimes be discriminatedalmost faultlessly (e.g. Weiskrantz et al. 1995) and aretherefore by deWnition well above conventional thresholdlevel where reported awareness can be unreliable, whereasin the seeing Weld stimuli even just above threshold they areinvariably accompanied by conscious visual awareness. Itis the huge gulf between performance and reported aware-ness that characterizes blindsight. If anything is trivial it isthe fact that the subjects are aware (know) that stimuli arebeing presented—because they are told so. It is the absence

of visual qualia yet discriminating as if they are present thatmakes blindsight so interesting. However, critics of thealleged unconscious nature of blindsight do have somethingto bolster their views, namely more recent observations thatblindsighted subjects may become aware that a visual stim-ulus which they deny seeing can provoke a non-modalimpression that “something happened”, although callingthis a sixth sense might be unwise. This led Weiskrantz(1998) to distinguish between type 1 blindsight (no aware-ness of any kind) and type 2 blindsight, where the subjecthas a non-visual experience that something occurred andeven where in the blind Weld it took place. However, whenawareness was reported after every stimulus trial in a studyof contrast sensitivity in blindsight (Stoerig et al. 2002) per-cent correct performance declined hand-in-hand withawareness. The striking dissociation between awarenessand performance reported by Weiskrantz et al. (1995)might therefore be restricted to moving stimuli, which ifcorrect itself requires an explanation. A more mundaneexplanation might be that diVerent subjects, or the samesubject at diVerent times, change their criterion foracknowledging awareness. This is no mere quibble. Forexample, a subject with total phenomenal blindness in aWeld defect and who is required to guess about the presenceand nature of visual stimuli he cannot see and who is keento “do well”, might begin to notice events that he initiallyregards as irrelevant. The subject might learn that if hiseyes ‘want’ to move in a particular direction, the feelingcorrelates with performance and he or she begins to reportbeing aware of this rather than being aware of a visual per-cept. As a consequence the criterion for awareness couldchange. This general point was raised long ago (Cowey andWeiskrantz 1963) but is rarely considered.

Does blindsight depend on islands of surviving V1?

Critics were right to raise this question, especially aspatient DB, whose performance led to the birth of the termblindsight, cannot be structurally scanned to determinewhether his striate cortex is really missing in the righthemisphere. As well, there is still no explanation for thetemporary recovery of real sight in one quadrant of hisimpaired hemiWeld (see Weiskrantz 2009) and for the reve-lation nearly 30 years after his operation, that he experi-ences visual after-images when a visual stimulus is turnedoV (Weiskrantz et al. 2003). How ironic if the discovery ofblindsight proves to be based on a patient who does notpossess it! Still, even if shreds of striate cortex underlieblindsight, this would not explain the absence of visualawareness. Fortunately it can at last be discounted in cru-cial instances. Many patients with blindsight have nowbeen structurally imaged by MRI and in some of them—most notably the much studied GY—there is no evidence

123

8 Exp Brain Res (2010) 200:3–24

that any striate cortex is spared in the region correspondingto the huge Weld defect (e.g. Barbur et al. 1993; Bridgeet al. 2008). There is also evidence from functional neuro-imaging (fMRI) that even when there is spared striate cor-tex it is not functionally activated by visual stimuli thatelicit blindsight in other parts of the visual Weld (Goebelet al. 2001) But this apparently convincing demonstration issubject to the reservation that surviving but damaged cor-tex, unlike normal cortex, might be capable of processingvisual information even while showing no detectablechanges in local cerebral blood Xow. Lastly, although it hasbeen suggested that investigations on monkeys, where thecomplete removal of striate cortex has been repeatedlyhistologically veriWed, are inadmissible because we do not,and cannot, know whether monkeys have blindsight(Campion et al. 1983, pp 470, 479), the much later demon-stration that monkeys do possess blindsight (see below)rebutted this criticism (Cowey and Stoerig 1995).

Is blindsight just the detection of stray light in the seeing Weld?

This possibility has been raised more than any other in crit-icisms of blindsight, although even if true one wonders whythe stray light is not consciously perceived despite sustain-ing excellent performance. The latter would amount toblindsight in normal vision! Perhaps it is perceived, but asan amorphous lightening within the seeing Weld that iscategorised by the patient as “a feeling that somethinghappened” (which might sustain detection but can hardlyunderpin discrimination of locus or orientation etc.), inwhich case it could be type 2 blindsight. Because the latteris plausible it remains important to evaluate any contribu-tion of stray light. For instance Campion et al. (1983) diddemonstrate that some of their patients were responding toa bright target in the blind temporal Weld by detecting lightscattered from the subject’s nose into the seeing part ofthe retina. This is an example of extra-ocular light scatter,which can potentially arise from any reXective surface infront of the subject. A simple control for this potential arte-fact is to place a half-patch over the viewing eye so that theblind retina is obscured, leaving any extra-ocular reXectionsvisible to the uncovered normal retina. By these meansKing et al. (1996) showed that the blindsight of all four oftheir hemispherectomized subjects was entirely based onlight scattered into the seeing hemiWeld whereas in thehemianope GY, whose lesion is solely occipital, it was not.Despite this alarming demonstration the control is still notused routinely, even with hemispherectomized subjects.

As its name indicates, intra-ocular scatter arises withinthe eye itself. Transmissive scatter depends on wavelengthand on opacities in the lens and vitreous, which increasewith age. The shorter the wavelength the greater the scatter,

which suggests that blue light should be more detectable byscatter than red. With rare exceptions (see below) there isno evidence that this is a quality of blindsight. Intraocularscatter is both more diYcult to assess and to eliminate.Light from a visual stimulus is always smeared on the ret-ina because of optical imperfections. If the point-spreadfunction is extensive, the spread might be wide enough forthe normal retina to detect it. The closer the stimulus is tothe edge of the blind region the more likely the smearedimage is detectable by the normal visual Weld. The spreadcontributes to the fuzziness of the perimetrically measuredoutlines of Weld defects. This is probably unimportant whenthe stimulus is far from the border of the Weld defect butseveral published studies of blindsight have used targetsclose enough to normal retina for a smeared image to stim-ulate normal retina. As a rule of thumb, anything within 3°or 4° is suspect, depending on the background luminanceand its contrast with the stimulus.

The other form of intraocular scattering is reXective orLambertian, which is roughly equal in all directions and ismuch more widespread across the retina as a result of mul-tiple aberrations in the path of light between the cornea andthe retina. Lest this seem fanciful it is salutary that whenLambertian scatter was modelled for the human eye(Faubert et al. 1999; Faubert and Diaconu 2001) the blind-sight reported previously by the same experimenters inhemispherectomy patients was entirely accounted for byLambertian scatter. In other words the patients were reallyblind and their blindsight was an artefact. In conWrmationthey showed by in vivo spectroreXectometry that the rela-tive insensitivity to greenish wavelengths in the blind hemi-Welds of the patients was explicable by selective spectralintraocular absorption of those wavelengths (Faubert et al.1999). These lessons continue to be ignored in several stud-ies of blindsight. However, they support those studies ofspectral sensitivity in blindsight in patients and monkeyswith occipital damage where there was no selective depres-sion of sensitivity in the blind Welds to greenish wave-lengths (Stoerig and Cowey 1992; Cowey and Stoerig1999).

The best control for all forms of light scatter is to showthat a stimulus that indicates blindsight no longer does sowhen presented monocularly on the natural blind spot,which measures about 6° £ 4°. This was shown with hemi-anopic monkeys (Cowey and Weiskrantz 1963; Cowey2004) and patients (Stoerig and Cowey 1991). But targetsoften used to study blindsight are often larger than the blindspot. Nevertheless, when circumstances allow it is the bestcontrol for assessing the contribution of light scatter. Forexample, Cowey and Stoerig (1995, 1997, 1999) showedthat monkeys with complete unilateral removal of striatecortex can localise small targets in the ‘blind’ hemiWeld andthat such targets are classiWed as ‘not a light’ i.e. blindsight.

123

Exp Brain Res (2010) 200:3–24 9

Two hemianopic monkeys, with head movements severelyrestricted and eye position monitored, were trained monoc-ularly to detect a brief small target subtending 0.5°, muchsmaller than the optic disc, while the monkey Wxated asmall light at the centre of the screen. After establishingthat in the blind Weld the monkeys could score about 90%correct to small targets, the target was moved a positionwhere the blind spot would be expected on anatomicalgrounds. At an eccentricity of about 13°, just below the hor-izontal meridian, performance deteriorated sharply, fromabout 90% correct to 1 and 27.0% correct. When the blindWeld target was moved away from the blind spot both mon-keys again scored better than 90% correct (Cowey 2004).This control showed that the monkeys were not respondingon the basis of either extraocular or intraocular light scatter,and were therefore like the patients studied by Stoerig andCowey (1992). Even earlier (Cowey 1967) it was shownthat monkeys with a retinal lesion larger than the naturalblind-spot and made by photocoagulation could not detectvisual targets directed at the induced blind region eventhough monkeys with a cortically induced Weld defect ofthe same size and position could do so. But stray lightwould be similar in both conditions.

Much of the early work on blindsight was performedwith targets projected from behind the subject, or usingMaxwellian view (Blythe et al. 1987) or even hand-heldtargets during clinical screening. But a new form of extra-ocular cue arose from the increasing use of raster displays.When a stimulus is presented on one side of a raster dis-play, it can produce a faint but visible ghost image thatspreads horizontally across the entire width of the screen(Cowey 2004). The artefact arises when the VDU electronbeam is incapable of faithfully restoring the speciWed lumi-nance values to their background values in the next rasterline of the stimulus. The ‘ghost’ is especially evident whenhigh luminance contrasts and horizontal gratings are used.Its detectability can readily be demonstrated by coveringthe part of the screen in front of the stimulus. When testinga hemianope the raster artefact can be rendered ineVectiveby turning the VDU on its side so that the ghost imagebecomes vertical and can be conWned to the blind hemiWeld.The signiWcance of this potential artefact was shown byCowey and Azzopardi (2001), who found that the directionof vertical motion of a horizontal grating or random dotkinematogram to the right side of a central Wxation point onthe VDU could be identiWed by a normal subject, wearing ahalf-patch over the right half of the viewing eye, by notic-ing the raster artefact at the left of the screen. It wasreported that blindsight subject GY can discriminate thedirection of such global motion in his blind right hemiWeld(Benson et al. 1998) but when the display was arrangedsuch that no raster artefact was visible to normal vision,three blindsighted subjects (including GY) failed to dis-

criminate direction of motion although they could still dis-criminate between stationary and moving displays (Coweyand Azzopardi 2001). Although the raster artefact is com-mon, especially in older and technically less advancedVDUs, its existence is rarely acknowledged, making it diY-cult to evaluate several reports of blindsight. That it is ahardware rather than a software problem is shown by thefact that some but not other VDUs exhibit it via the samesoftware programme.

Signal detection and the problem of response bias

The Wrst alarm bells about the possible eVect of thepatient’s diVerent criteria when asked if he saw a visualstimulus (yes/no) or had to guess (forced-choice respond-ing) whether a stimulus occurred in one or other of twointervals or in one of two ways, like stationary or moving,were sounded by Campion et al. (1983). Yes/no methodsare characteristic of clinical tests of vision whereas forced-choice is the hallmark of psychophysical measures ofvision. But the two procedures measure diVerent things. Asubject’s judgement in a yes/no task depends not only onsensitivity to the target, but also on any tendency to selectone or other response independently of his sensitivity, i.e.response bias. If the subject’s criterion for saying ‘yes’ is acautious one, ‘no’ responses will predominate. But forcedchoice methods are less prone to response bias (althoughnot necessarily devoid of them) and characteristically leadto lower detection thresholds. Substantial variation in biasin the cortically blind Weld was also demonstrated by Stoeriget al. (1985) by simply varying the ratio of target-to-blanktrials in a target detection task.

Whether or not blindsight diVers from normal near-threshold vision depends in large part on response bias,which varies according to the measure of performancethat is used. In signal detection theory (Green and Swets1966; Macmillan and Creelman 1991) sensitivity, or d�, iscalculated so as to be independent of bias. But routineclinical tests and many laboratory tests are conWned topercent correct, which accurately measures sensitivityonly when there is no response bias. There are many illus-trations of how percent correct varies with bias, for vari-ous values of d�, calculated for a yes-no detection task inwhich the proportions of targets and blanks are equal. Per-formance can range from 50 to 95% correct, depending onbias. The problem of assessing awareness (or its absence)in blindsight is that responding in forced-choice teststends to be free of bias with respect to the stimulus,whereas responding in yes–no tasks does not. Thus a dis-sociation between percent correct scores of awareness(assessed by yes–no detection tasks) and forced-choiceperformance could simply be due to the extreme responsebias in the yes–no task.

123

10 Exp Brain Res (2010) 200:3–24

A further problem is that bias is not necessarily stable.For example it can be controlled by asking a subject toguess whether or not an unseen stimulus was present oneach trial presented in a detection task (Azzopardi andCowey 1997). They informed blindsighted subject GY thata high contrast stimulus would occur on half the trials atrandom and asked him to distribute his guesses accord-ingly. Bias was readily minimized in this way (Azzopardiand Cowey 1998). Moreover, the shift in bias could accountfor a statistically signiWcant dissociation between chanceand above-chance percent-correct performance given anunderlying sensitivity of d� > = 1.0. Clearly forced-choiceproduces such a shift in bias. With an extremely conserva-tive criterion a subject rarely if ever says ‘yes’, whereasimposed and appropriate instructions to guess increases theproportion of yes responses and reduces the bias. There-fore, simple instructions and the way a subject interpretsthem, can produce a prominent dissociation betweenperformance and awareness.

Response bias can also be altered by providing or with-holding information about when a stimulus is likely to bepresented. Simply varying the time interval between aprompt and a target changes a normal subject’s responsecriterion, even when the cue and target are in diVerentmodalities (Treisman 1964). Although response criterionhas not been measured quantitatively in the presence orabsence of prompts in cortically blind patients, there is evi-dence that it can signiWcantly aVect performance. Forexample, GY never detected a high-contrast target in hisblindsighted Weld when its appearance was not signalledwith an auditory cue, even though he easily detected it withthe prompt (King et al. 1996). A lesser but still severe dete-rioration of performance occurred in hemianopic monkeysin their blind hemiWeld when the cue that normally sig-nalled the instantaneous presentation of a visual stimulus inany one of the four visual quadrants no longer did so(Cowey and Stoerig 2003). In the same study GY alsofailed to respond at all on many trials, echoing the demon-stration that he more often reports (non-visual) awarenessof targets when they are signalled by a cue (Kentridge et al.1999a, b).

Despite the above considerations, does response biasexplain all the dissociation in performance between yes/noand forced choice guessing? This was speciWcally investi-gated by Azzopardi and Cowey (1997). They Wrst mea-sured GY’s ability to detect large (20° £ 20°), highcontrast (0.95), square-wave gratings (0.25 cycles/deg)presented for 200 ms in his hemianopic Weld. The proce-dures were yes–no or temporal 2-alternative forced-choice,in which GY also had to rate his conWdence on a 4-pointscale immediately after each decision. The results wereanalysed to generate receiver operating characteristic(ROC) curves from which estimates of d� were derived

which incorporated the appropriate mathematical transfor-mation for equating sensitivity in yes–no and 2afc tasks.GY’s results were compared with those from three controlsubjects tested in the same way. GY was approximatelyhalf as sensitive when detecting the stimuli in the yes/nocondition as in the forced-choice condition, and the diVer-ence was signiWcant at every stimulus contrast. By contrast,control subjects were equally sensitive to the targets in thetwo kinds of task. Therefore, any diVerence between GY’spercent correct scores to the targets presented in his blindWeld measured by yes-no and forced-choice tests, cannot beentirely explained by shifts in response criterion from onetask to another and that visual processing in the blind Weldis not just like normal vision near detection threshold.However, when he was tested with moving targets therewas no longer any diVerence in sensitivity for yes-no andforced-choice procedures. As mentioned earlier there issomething special about moving stimuli, which harks backto Riddoch (1917) and to Zeki and Vytche (1998). Variationin bias in the cortically blind Weld was also demonstratedand measured by simply varying the ratio of target-to-blanktrials in a target detection task (Stoerig et al. 1985; Stoerigand Pöppel 1986; Stoerig and Cowey 1989, 1992). Muchremains to be discovered about the roles of criterion shifts,criterion instability, and diVerent stimulus qualities inblindsight.

Is there any way of eliminating response bias by anchor-ing it in some way? Post-decision wagering oVers a promis-ing possibility. For example, in the Iowa gambling task andthe learning of an artiWcial grammar task subjects arerequired to wager on their decisions, incorrect wagers beingsubtracted from their winnings. It is intuited that they willonly wager high when they are conWdent about theirchoices. But it is already established that their performanceis signiWcantly better than chance when they are still bettinglow and expressing no conWdence in their responses (e.g.Dienes and Scott 2005) and this has become an acceptedindication of implicit knowledge. This was conWrmed byPersaud et al. (2007) in several normal subjects and withstimuli in GY’s seeing hemiWeld. GY was then tested withgrating stimuli presented in his hemianopic Weld or blanktrials. When he was aware of the stimulus, despite sayingthat he did not see anything, he wagered high and per-formed at better than 90% correct. But when he denied anyform of awareness and accordingly wagered low he stillscored signiWcantly better than chance, i.e. he demonstratedblindsight. However, CliVord et al. (2008) have argued thatpost-decision wagering fails to distinguish between perfor-mance without awareness and a reluctance to gamble whenthe sensory stimulus is weak, i.e. the response bias problemagain. Fortunately there is yet another method of assessingwhether a subject is truly unaware of the stimulus in blind-sight (Persaud and Cowey 2008). Using an exclusion task,

123

Exp Brain Res (2010) 200:3–24 11

in which subjects had to report the opposite location of theactual location of a stimulus in the upper or lower quadrant,normal subjects and GY could perform this perfectly taskwhen the stimulus was in the sighted visual Weld. But whenit was in GY’s blind hemiWeld and he denied any awarenessof it he responded signiWcantly more often to its real posi-tion, demonstrating that blindsight is unlike real consciousvision.

A diVerent attempt to eliminate bias was made byKunimoto et al. (2001), which involves asking subjects torate their conWdence in their response judgements, thenderiving a bias free measure (a-prime or á) of awareness bya calculation derived from signal detection theory. The ideaseemed to work as long as response bias did not change. Ifit does change then á will Xuctuate, as demonstrated byEvans and Azzopardi (2007) in normal subjects but, impor-tantly, in GY too. The problem of totally eliminating orstabilising bias seems to be intractable.

Pathways for blindsight

Blindsight exists but how is it mediated; what is its anatom-ical basis?’. There are ten pathways from the eye into thebrain, shown on the right in Fig. 2. The two most prominentpathways (thick lines) are to the superior colliculus (about1.5 £ 104 Wbres in the macaque monkey) and to the dorsallateral geniculate nucleus (dLGN) of the thalamus (at least106 in macaques). When V1, which is indubitably the corti-cal target of almost all of the projection neurons in theparvocellular (P) and magnocellular (M) layers of the dLGN,is destroyed there is rapid and nearly total retrograde degen-eration of these P and M projection neurons (Mihailovicet al. 1971). It was accordingly widely believed that path-ways other than those to and from the LGN must be mediat-ing blindsight. The explanation may be more complex, asfollows:

1. Destruction of striate cortex does not cause the inter-laminar layers of the dLGN to degenerate conspicu-ously, although the extent has never been quantiWed.The interlaminar projection neurons are part of thekoniocellular (K) projection (for review see Hendryand Reid 2000). It originates from bistratiWed retinalganglion cells, which provide information about wave-length (Dacey and Lee 1994). The fact that they sur-vive destruction of V1 presumably reXects the fact theyalso project to extra-striate visual areas. Their numbershave never been assessed but they may be as commonas M cells, i.e. about 105 in each LGN. This means thatthey are as abundant as all ganglion cells in the labora-tory rat, which has excellent vision!

2. A few projection neurons in the parvo and magnocellu-lar layers also survive removal of V1, presumably forthe same reason, i.e. they too project to extra-striatecortex, from which they can be retrogradely labelled(Cowey and Stoerig 1989). This means that a popula-tion of K, M, and P cells in the dLGN continues to pro-ject to several extra-striate visual areas in the absenceof striate cortex. The fact that blindsight varies amongpatients and is sometimes absent could reXect the con-tribution of these non-striate projections, especially asextra-striate cortex is often involved in occipital lesionsin human subjects. This is supported by the Wndingin monkeys that, other things being equal, the extentof transneuronal retrograde degeneration correlates

Fig. 2 Schematic diagram of the ten established pathways from theeye to their retinorecipient targets in the brain. The right side of thediagram shows the normal arrangement, the left side the eVect ofremoving striate cortex V1. Thicker lines and arrows indicate heavierprojections. Dashed lines show numerically weak projections. Forsimplicity the many onward cortical pathways from V2, V3, V4 etc.are not shown, except those destined for the inferior temporal cortexand, presumably, the amygdala. Back-projections are ignored. Notethat on the left side, lacking V1, the visual input to the ventral process-ing stream is severely impoverished but that the pregeniculate nucleushas expanded. Labelling from bottom upwards: SCN suprachiasmaticnucleus; MTN, LTN and DTN medial, lateral and dorsal terminal acces-sory optic nuclei; NOT nucleus of the optic tract; ON olivary nucleus;PGN pregeniculate nucleus; SC superior colliculus; PI inferior pulvi-nar; dLGN dorsal lateral geniculate nucleus. ModiWed from Stoerig andCowey (1997)

123

12 Exp Brain Res (2010) 200:3–24

signiWcantly with the extent of extra-striate involve-ment in the cortical lesion.

3. Following total hemispherectomy in monkeys, even theipsilateral superior colliculus shrinks, by about 30% involume, and about a third of its neurons degenerate(Theoret et al. 2001). It is hardly surprising thereforethat blindsight revealed by forced-choice guessing, asopposed to residual visual processing revealed by othermeans (Tomaiuolo et al. 1997), is absent in hemispher-ectomized patients.

4. After total unilateral removal of striate cortex in infantmonkeys the pre-geniculate nucleus increases involume and its neurons are also larger (Dineen et al.1982). Although the morphological types of ganglioncells that project to the PGN in primates have neverbeen thoroughly studied, an identiWed input remainsfrom wavelength-opponent retinal P cells (Cowey et al.2001). The physiological properties of retinal P cellsequip them to register chromatic borders in the absenceof luminance borders. This could explain why certainisoluminant coloured targets can be detected inblindsight whereas discriminating between two diVerentcolours is diYcult or impossible.

5. There is no known change in the retinal projection tothe dorsal, lateral and medial terminal nuclei of theoptic tract, which mediate the response to Xow Welds.Hardly surprising therefore that these reXexive responses,like that of the pupil, survive in blindsight.

6. After destruction of V1, the projection neurons of the Pand M layers of the LGN die within 12 weeks inmonkeys (Mihailovic et al. 1971). This is followed byextensive transneuronal retrograde degeneration of ret-inal ganglion cells (TRD) over the following few years.First described, but never quantiWed, in the humanretina by van Buren (1963) it was subsequently shownthat the famous blindsight monkey Helen had lostabout 80–90% of the ganglion cells in her central retina(Cowey 1974). Later morphological studies pinpointedthe loss to P ganglion cells (Cowey et al. 1989) andshowed, as mentioned above, that its variation amongmonkeys with complete unilateral removal of striatecortex depended not only on age at operation andlength of survival (Dineen and Hendrickson 1981;Weller and Kaas 1989) but also on the extent of extra-striate and white matter damage (Cowey et al. 1999).Given that this variation occurs in monkeys, where thelesion can be made precisely, it is bound to be moreprevalent in patients, where the extent of accidentallesions varies extensively. But what is the evidence thatTRD occurs at all in patients? The two patients studiedby van Buren (1963) had brain damage that could haveinvolved the optic tracts, including damage to the retinaitself. The problem was therefore more recently inves-

tigated in the blindsight subject GY by using opticalcoherence tomography (OCT) to determine whetherthe ‘blind’ hemiretina shows evidence of death of reti-nal ganglion cells. As illustrated in Cowey (2004), GYhas conspicuous thinning of the ganglion cell layer ofthe hemiretina of both eyes corresponding to his Welddefect. Very recently, the same result was reported in alarge group of patients with long standing Weld defectscaused by occipital damage, although they had notbeen assessed for blindsight (Jindahra et al. 2009).

It was reported by Trevethan et al. (2007) that blindsightcan even be superior to real sight in the unaVected part ofthe retina. Subject DB was able to detect a low contrast 2°Gabor patch that was invisible at the same eccentricity inhis sighted hemiWeld. This counterintuitive result may wellbe a consequence of the massive selective loss of P gan-glion cells in the blind Weld, as described above, whichcould change the sensitivity of the surviving M cells byabolishing any competition between the two processingchannels. As yet there is no direct evidence for thisalthough the neglected paper by Wilson (1967) shows thatspatial summation, and therefore sensitivity, increaseswithin the Weld defects caused by lesions in the visual radi-ations and that although sensitivity within the Weld defectwas lower than outside it at the same eccentricity, it wasstill improving steeply at the largest stimulus used, i.e.80 arcmin, unlike the now shallower curve in the normalWeld. As Trevethan et al. used a larger, 2°, stimulus at aneccentricity of 11.3°, spatial summation may have renderedthe retinal sensitivity higher within the Weld defect. Unfor-tunately, the phenomenon has not been reported in hemi-anopes who, unlike DB’s experience of afterimages, lackany phenomenal vision in the blind Weld.

Do monkeys have blindsight?

This question is important given that the vast majority ofneuronatomical and single cell physiological studies rele-vant to blindsight have been performed on monkeys. But ifmonkeys do not possess blindsight such studies could beirrelevant, except for investigations of the eVects of remov-ing striate cortex on visual performance without respect tovisual conscious awareness of the stimuli, which is whatinvestigators believed they were studying for most of thetwentieth century. It is also possible that they have blind-sight but that “no one has ever shown, nor can ever show,blindsight in monkeys” (Campion et al. 1983, original ital-ics). It was the latter stark message that led Cowey andStoerig (1995, 1997) to attempt to assess blindsight in hem-ianopic monkeys, as follows. On each trial a brief lumi-nance target appeared in one of the four quadrants of a

123

Exp Brain Res (2010) 200:3–24 13

VDU when the monkey Wxated and touched a start-lightnear the centre of the display. Three hemianopic monkeysand one normal monkey (who was subsequently made hem-ianopic) were studied. Stable detection thresholds for thenormal and the hemianopic Welds were determined overseveral thousand trials. When the targets were about 0.5 logunits above threshold the monkeys scored better than 90%correct even in the hemianopic Weld, like patients withblindsight. The monkeys then learned a diVerent task. Onhalf of the trials a target was brieXy presented in the normalhemiWeld. But on the remaining trials no target was pre-sented (called blanks) and the monkey was rewarded if itpressed a prominent rectangular outline permanently pres-ent on the screen. In the next and crucial stage a visual tar-get, known to be well above the previously determineddetection threshold, was presented in the hemianopic Weldon 10% of trials. Would the monkeys respond to theseprobes by touching them—as they had previously doneover thousands of trials in the detection task—or wouldthey respond by touching the rectangle reserved for blanktrials? The answer was unambiguous. All three hemianopicmonkeys almost always categorized the probe in the hemia-nopic Weld as a blank, i.e. as not a visual stimulus. Theyoccasionally responded by touching the probe, i.e. catego-rizing it as a light, which is consistent with conscious visualperception. However, they also occasionally responded to atarget in the normal hemiWeld as a blank, suggesting that onsuch trials they were unsure. The simplest explanation forthis result is that detectable lights in the hemianopic Weldare not experienced as lights, as in patients with blindsight.

Rightly, the demonstration of blindsight in monkeysneeded an evaluation of several controls before it could beaccepted: (1) Perhaps the monkeys did have a phenomenalvisual percept in the hemianopic Weld but because it alwayslooked dimmer than in the normal hemiWeld they perceivedit to be more like a blank than a target in the normal Weld, asargued by Macphail (1998). But even when the monkeyswere trained with a target in the good Weld that was justabove detection threshold, and therefore perceptually faint,and the probe in the hemianopic Weld was well abovethreshold and should therefore appear bright, the sameresult was obtained. (2) The probes were necessarily lessfrequent than targets in the good Weld in order not to makethe monkey make many unrewarded mistakes and therebyrisk a general breakdown in performance. But when thesame low incidence of (dim) targets was used in the normalhemiWeld the monkeys did not treat them as blanks. Fur-thermore, when control sessions were made with probesand targets equally probable the probes were still treated asblanks. (3) When the targets in the normal hemiWeld andthe probe in the blind Weld on the right were made muchlarger (14° £ 14°) and consisted of a 0.3 cyc/deg horizontalgrating in an attempt to increase their salience, the same

result was obtained. (4) Even when the grating was movedvertically at 18 deg/s in order to determine whether a mov-ing stimulus might be categorized diVerently, the sameresult was obtained. (5) The reinforcement schedule in anydiscrimination experiment is important. Perhaps a particu-lar reward schedule encourages the monkeys to categorizetargets in the blind Weld as blanks. One of the hemianopicmonkeys was accordingly tested on the blindsight task withthe usual schedule (reward on a probe trial only by reachingfor it, a second was tested in extinction on probe trials (notrewarded whether it pressed the probe or the blank rectan-gle) and the third was rewarded whichever it pressed. Thismade no diVerence to the categorization of probes asblanks. (6) When the normal monkey was subsequentlymade hemianopic by removing her left striate cortex andher performance compared with that of several blindsightpatients on a localisation task, the monkey and the patientsperformed similarly well in the blind hemiWeld. But whenthe same targets were presented in a categorization task themonkey always categorized the target in the blind Weld as ablank whereas the patients (who in addition had to saywhether they were aware of the stimulus) responded justlike the monkey when they were unaware but not whenthey were aware, implying that the monkey was neveraware or had not yet learned the signiWcance of being aware(Stoerig et al. 2002).

A sceptical contrary view was expressed by Mole andKelly (2006), who argued that the working memory load isnot the same in the localization task and the signal detectiontask and that this might be suYcient to explain the diVerentpatterns of responding without having to invoke blindsightin the monkeys. For example an increased working memoryload “may lead to a diVerence in a monkey’s focusing ofattention [which] could in turn lead to a diVerence in whatthe monkey notices”. Their point is valid and promptedCampion (2007, p 207) to assert that “Cowey and Stoerig’sclaim [has] been clearly refuted”. But there are three rea-sons why Mole and Kelly are unlikely to be correct. First,even if the target frequency in the localisation task is madelow in the blind Weld and high in the seeing Weld in order todirect expectations and attention preferentially to the latter,the monkeys continue to perform well with the now rareblind Weld targets. Second, Mole and Kelly directed theircriticism at the very Wrst demonstration of blindsight inmonkeys (Cowey and Stoerig 1995) where there weremany possible target positions in the seeing Weld and all tar-gets were small. When the targets were subsequently mademuch larger and more contoured and there was only a sin-gle target in either Weld and the no-target blank responsewas placed centrally instead of entirely in the seeing Weld,all this in order to make the task easier and reduce memoryload, the same result was obtained. Third, reaction timeswere collected for every trial in the latter experiment and

123

14 Exp Brain Res (2010) 200:3–24

the results are shown, for the Wrst time, in Fig. 3. The rele-vant point is that when correctly responding to a blank trial,reaction time is signiWcantly faster than when making thesame motor response to indicate, wrongly, that a target inthe blind Weld is a blank. In other words, the blind-Weld tar-gets are still being processed and it cannot be that in the sig-nal detection task they are categorised as blanks for thereason proposed by Mole and Kelly. These controls showthat hemianopic monkeys have blindsight like humanblindsight, as also concluded by Moore et al. (1995, 1998),in experiments involving saccadic eye movements to tar-gets in the blind and normal Welds. This conclusion is ofmajor importance with respect to the behavioural evidencealready mentioned and with respect to investigations of the

anatomical and physiological basis of blindsight, almost allof which are based on monkeys.

Although monkeys have blindsight in the sense of indi-cating that visual targets are not perceived as lights, arethey consciously detected in a non-visual fashion? This wasinvestigated by Stoerig and Cowey (2009) who attemptedto train a blindsighted monkey to respond correctly to aprominent high contrast target in the blind Weld by provid-ing extensive practice on the signal detection task alreadydescribed. After initially always responding by categorisingthe target in the blind Weld as a blank she gradually begancorrectly to touch its position. However, she never scoredbetter than 50% correct even after extensive practice andher reaction times were lengthy, suggesting that she

Fig. 3 Reaction times for one normal monkey (Rosie) and threehemanopic monkeys averaged over several thousand trials for eachmonkeys on the blindsight task. On each trial the monkey Wxated thecentre of a VDU and a high contrast square wave grating isoluminantwith the surround was brieXy presented in the left or the right hemiWeldor there was no stimulus (blank trial). The monkeys had previouslybeen trained to touch the remembered position of a target no matterwhere it was presented in either hemiWeld in order to obtain reward andeven in the bind hemiWeld the hemianopic monkeys scored around90% correct on this localisation task. They were then trained to toucha permanently outlined rectangle above the centre of the VDU when-ever no target appeared. In the Wnal arrangement targets were again

presented in either hemiWeld on half the trials and the remainder wereblank trials. DiVerent proportions of the three types of trial were usedin diVerent sessions but the Wgure shows the overall result. When thetarget was in the hemianopic Weld the three hemianopic monkeysalmost invariable categorized it as a blank and responded (incorrectly)by touching the central rectangle. But they took signiWcantly longer todo so than when making the identical motor response (correctly) ongenuine blank trials. The result shows that the target in the blind hemi-Weld is being processed despite being mistakenly categorized as ablank. From Cowey, Stoerig and Alexander (unpublished). For moredetails of the procedure see Cowey and Stoerig (1997)

123

Exp Brain Res (2010) 200:3–24 15

remained uncertain about the nature of the stimulus in theblind hemifeld, or that she was sometimes totally unawareof it (blindsight) and sometimes aware of something (Type2 blindsight). The monkey was continuously watched onCCTV and her slight hesitation when responding to the tar-get in the blind Weld was obvious.

Notwithstanding the arguments presented above, selec-tive spatial and temporal attention does have an eVect onblindsighted performance in both monkeys and patients.Performance declines steeply when the moment at which astimulus occurs in the blind Weld becomes uncertain(Cowey and Stoerig 2003) and, using a diVerent procedurewhere a visual cue at the beginning of each trial usually, butoccasionally invalidly, cued GY to the position of the targetKentridge et al. (1999a, b) showed that performance wassigniWcantly superior when the cue was valid, presumablybecause attention could be directed to the position of thetarget. Incidentally, this result also shows that if a humansubject can attend to a particular place within a hemianopicWeld it is hardly surprising that monkeys can also attend tothe blind Weld.

Although studies of temporal uncertainty and selectivespatial attention in blindsight are rare they have a practicalmessage. In everyday life, there is huge uncertainty aboutwhen and where visual events will occur. It may thereforebe more diYcult than envisaged to train subjects to usetheir blindsight.

Indirect methods of studying blindsight: implicit processing

Direct methods of assessing blindsight can be discomfort-ing for the subject, who must make judgements aboutthings he denies seeing. To demonstrate implicit processingof an unseen stimulus in a Weld defect its eVect on theresponse to a seen stimulus in the normal visual Weld isevaluated. If it is altered by the unseen stimulus, the lattermust have been implicitly processed. There are now manysuch demonstrations. For example, simultaneous or priorpresentation of an unseen stimulus can substantially alterthe reaction time to the seen stimulus in human subjects(Marzi et al. 1986; Corbetta et al. 1990; Rafal et al. 1990).The perceived hue of after-images induced by Wxationinside a coloured surround can change when the colour ofthe surround is changed, even when the change is restrictedto the blind Weld (Pöppel 1986). In a phi motion task theperformance of patient FS improved when a third and invis-ible stimulus was presented between the two visible onesand the patient even reported motion between the stimulirather than just detecting a time diVerence between them(Stoerig and Fahle 1995). Even more remarkably, highlevel cognitive processing in blindsight was suggested by

the inXuence of a word presented in the blind Weld, e.g.‘river’ or ‘money’, on the semantic interpretation of a poly-semous heard word like bank (Marcel 1983, 1998). Unfor-tunately eye movements were not fully controlled ormeasured and the demonstration remains arresting ratherthan convincing. It is hardly surprising that such methodshave rarely been investigated with monkeys. But whendone so, a blind Weld target slowed reaction time to a seentarget in the normal hemiWeld (Cowey et al. 1998), as inblindsighted patients.

Total cortical blindness

Although the Wrst investigations of the eVects of removingall striate cortex bilaterally in monkeys have already beendiscussed, total human cortical blindness, which is distress-ing but fortunately rare, provides an opportunity to assessblindsight in the absence of at least some of the artefactsalready described. Such a patient is also more likely toattempt to use blindsight because no normal vision is avail-able. In one of the Wrst such studies (Kiper et al. 2002) ayoung girl MS was studied over a period of years followingrecovery from bacterial meningitis at age 5 weeks. When3 years old, structural MRI revealed bilateral occipitallesions with no occipital cortex on one side and only a thinribbon on the other. Despite this, several ‘early’ visualfunctions (acuity, colour, form, motion) were almost nor-mal when assessed by 2afc. Only higher-order functionslike Wgure ground segregation based on texture, wereseverely impaired although they too showed subsequentimprovement. Although the report does not make it certain,but suggests, that MS was displaying real sight rather thanblindsight, it does show that performance can be unusuallygood with little or no V1. How does she achieve this? Aprevious article on MS by the same investigators (Innocentiet al. 1999), using fMRI and EEG-coherence, indicated thather cortico-cortical and callosal connections had re-organ-ised in a novel fashion. Although MS is probably not agood guide to blindsight following occipital lesions muchlater in life, she does provide fascinating evidence ofconscious visual perception without V1. And a slightlyworrying note is that visual perimetry shows that that MSdoes have a superior monocular crescent where vision isespecially spared. Several years ago the present writerattempted to determine the possible contribution of theslight monocular visual sparing in monkey Helen to hernavigational skills. Wearing a pair of skiing goggles inwhich one eye was completely covered in black tape andthe other eye was similarly covered apart from a narrowcrescent corresponding in size to Helen’s spared peripheralvision, and a tiny phosphorescent glowing spot was stuck tothe inside of one goggle as a blurred Wxation ‘point’ in the

123

16 Exp Brain Res (2010) 200:3–24

centre, the task was to navigate by means of the small mon-ocular peripheral visual window. After spinning round on arevolving stool so that any recollection of the surroundingscould not guide navigation, it proved possible to moveabout, slowly but eVectively, by using only head move-ments to scan the environment. Even 30 min practice pro-duced considerable improvement. One wonders whetherseveral years of experience might have made me as eYcientas monkey Helen!

The recent study by De Gelder et al. (2009) is particu-larly interesting because it focuses on the navigationalskills of a totally blind man, TN, whose successive strokescompletely destroyed striate cortex bilaterally, as veriWedby both structural and functional magnetic imaging. TNcould successfully navigate a corridor containing largeobstacles without colliding with them. His astonishing suc-cess was plausibly attributed to surviving extra-striate path-ways concerned with the registration of Xow Welds when hemoved. Despite this striking demonstration, serious doubtsremain. The authors recognized the possibility of echoloca-tion but somewhat egregiously concluded that it “can besafely discounted”. The extensive literature on echolocationin blind subjects or blindfolded normal subjects, spanningover 60 years, suggests otherwise (e.g. Supa et al. 1944;Rosenblum et al. 2000; Schwitzgebel and Gordon 2000).The latter authors also draw attention to the frequent inabil-ity of their blind subjects to explain how they are navigat-ing and that even when they oVer an explanation it can beerroneous. TN might have been displaying not blindsightbut the related phenomenon of unconscious knowledge ofauditory signals. As TN was also closely followed alongthe corridor by one of the experimenters in order to alerthim in case of impending collisions, it is possible that thelatter unwittingly provided clues now legendary in the taleof the horse Clever Hans. Whatever the resolution of thisuncertainty it could be avoided by requiring TN to wear earmuVs to block echoes, or opaque spectacles to eliminateXow Welds or to be alerted to possible collisions via a loud-speaker. Doubtless these essential controls will eventuallybe carried out because they are absolutely essential.Whether TN also has aVective blindsight is discussed in thenext section.

AVective blindsight

AVective blindsight is the ability to physiologically registerand/or discriminate between diVerent ‘emotional’ stimulisuch as happy or scowling faces. For example De Gelderet al. (1999) found that subject GY could discriminate—byguessing—between diVerent facial expressions in his hemi-anopic Weld. Subsequently, Morris et al. (2001), usingfMRI, reported that GY showed diVerent patterns of activa-

tion in the amygdala to fearful and happy facial expressionswhen the unseen stimuli were in his blind Weld. Unsurpris-ingly their conclusions have been questioned because thestimuli were also shown to GY in his normal hemiWeld andit is possible that GY learned to pick out observed salientfeatures of fearful expression, like wide-open eyes and,knowing that the same stimuli were being used in his blindWeld, selected a salient stimulus as the fearful one. It is stillnot known whether GY or any other hemianope could iden-tify facial expressions if the set of stimuli was restricted tothe blind Weld and never shown beforehand to the seeingWeld. However, these reservations cannot apply to a patient,like TN, with total cortical blindness and veriWed destruc-tion of V1 bilaterally. When TN was shown facial stimulihe exhibited increased activation in the right amygdala inresponse to fearful faces, as shown by fMRI (Pegna et al.2005). Moreover, by using electric source imaging of theEEG, Gonzalez-Andino et al. (2009) showed that all facialexpressions elicited responses of short latency (70–120 ms)in the temporal polysensory areas of TN but that emotionspeciWc responses occurred later (>120 ms) in more rostraltemporal cortex and latest of all (>200 ms) in the rightamygdala. The authors conclude that the pathway is fromretina to superior colliculus and thence to pulvinar, extra-striate cortex, temporal cortex, amygdala and not a pathwayfrom colliculus to pulvinar to amygdala as previously sug-gested. As the diVerent facial expressions were presented inrandom order and at a time when TN could not even tellwhether the room was light or dark and, apparently, he didnot even know the nature of the stimuli, it is diYcult toattribute the results to anything else but pure blindsight viapathways independent of V1 and of any sensory awareness.

Physiological correlates of blindsight and awareness

There are several investigations in monkeys of single unitresponses evoked by targets in the ‘blind’ Weld, althoughtheir existence does not necessarily prove that they underlieblindsight. In anaesthesised monkeys, parts of whose striatecortex were ablated or reversibly inactivated, the extrastri-ate visual areas in the dorsal stream retain many of theirvisual receptive Weld properties (Rodman et al. 1989, 1990;see Bullier et al. 1994 for review). But even neurons in themotion area MT/V5 lose their directional sensitivity despitecontinuing to respond to motion per se (Azzopardi et al.2003). In contrast, visually sensitive neurons of the ventralstream were rare, notwithstanding the anatomically demon-strated sparse direct projections of surviving dLGN cells toareas V2, V4, TEO, and IT (Bullier et al. 1994). The ventralstream thus seems to be more dependent on striate cortex.However, the immediate outcome of removing or inactivat-ing striate cortex might be a poor guide to the role of an

123

Exp Brain Res (2010) 200:3–24 17

extrastriate visual pathway in an alert subject long afterdamage to V1. This was recently investigated by Schmidet al. (2009) who, using fMRI in monkeys, found unambig-uous evidence of activation in extrastriate areas V2 and V3in response to visual stimuli presented in the ‘blind’ regioncorresponding to a lesion in V1 made up to 2 years earlier,although the signal strength was far less than in regionsretinotopically corresponding to the intact visual Weld. Itcould be argued that increased blood Xow might not indi-cate spiking activity but the authors also showed that visualstimuli conWned to the Weld defect still elicited multi-unitactivity in the retinotopically corresponding parts of V2/V3. This is currently the clearest evidence for the survivalof visual responsiveness in extrastriate cortex of the ventralprocessing stream following a lesion of V1.

Single cell recordings like those just described for mon-keys are understandably absent for humans, but extra-cra-nial or functional neuroimaging studies are increasinglyused. One of the Wrst such experiments on human bindsight(Barbur et al. 1993) showed that when GY was presentedwith moving stimuli in his hemianopic Weld that produced a“feeling” that something had occurred in his blind Weld,there was increased blood Xow in area MT/V5 of the‘blind’ hemisphere, indicating that that this area has avisual input that bypasses V1 and can mediate motiondetection. Similar results were reported when MEG wasused with GY (Holliday et al. 1997). There are many subse-quent demonstrations along the same lines but none is freeof the following problems:

1. Functional activations, whether revealed by PET orfMRI, do not necessarily show that an activated regionis necessary for performing the task and is actually reg-istering the stimuli in the blind Weld, nor whether itunderlies discrimination as opposed to detection. Forexample, if a subject is informed about the kind ofstimulus that will be presented, top-down processingcould impose anticipatory changes of blood Xow in theregions that are normally involved in processing suchstimuli. For example Azzopardi and Cowey (2002)demonstrated that blood Xow increases in striate cortexeven when the subject expects a visual stimulus whichis then not delivered, i.e. on a catch trial. Similarly(Cowey unpublished) there can be a small pupillaryresponse when an expected visual stimulus is not deliv-ered, as long as the time of delivery is predictable, as itoften is. This could, at least partly, explain why thepupil still responds to a stimulus presented in the blindWeld and of which the patient says he is completelyunaware (Weiskrantz et al. 1999).

2. Functional activation studies of blindsight continue tolack a condition where the targets are neither shownnor described beforehand so that the subject cannot

know what to expect and accordingly cannot use somesimple feature of the stimulus that relates it to the samefeature if the stimulus is presented in the seeing Weld.As already mentioned, in studies of ‘recognition’ ofemotional expression in faces presented in the blindWeld (e.g. De Gelder et al. 2001), if the subject isshown the faces in the seeing Weld and knows that thesame pictures will be presented in the blind Weld, thephysical salience of two wide-open eyes could cue thedetection of a fearful expression, which by top-downprocessing might instigate the observed activations inthe amygdala. When a hemianopic monkey was unex-pectedly and for the Wrst time shown a monkey doll ora banana in his blind Weld there was no overt emotionalresponse but when they were then presented in the nor-mal Weld the appropriate response—fright or desire—was immediate (Cowey and Weiskrantz, 1963). Thesame conclusion was reached by Humphrey (1974)who noted that the bilaterally cortically blind monkeyHelen, who “sees everything but recognises nothing”never displayed emotional responses to visual events,which is puzzling if she retained a visual projectionto the amygdala concerned with the registration ofemotional visual stimuli.

3. If a cortical region is active in blindsight, e.g. area MT/V5, and the activity is indispensable for blindsight, per-formance should be perturbed if the activity is experi-mentally disrupted. This was studied by using rTMS todisrupt activity in area MT/V5 while two subjects,including GY, performed global motion discrimina-tions. The rTMS severely impaired a simple 2afcdiscrimination between motion and non-motion in theblind Weld of both subjects but not in the control sub-jects, whereas the same stimulation above area V3,which is also functionally activated by global motion,did not (Alexander and Cowey 2009). However, eventhis seemingly clear interpretation is clouded by thepossibility that TMS to MT/V5 activates its eVerentprojections to the koniocellular layers of the LGN, dis-rupting the latter without aVecting the still unknownsource of the visual input to V3 in the absence of striatecortex.

Transcranial magnetic stimulation (TMS) oVers a means ofinvestigating some of these problems. rTMS delivered overarea MT/V5 in normal subjects, and to the normal hemi-sphere of GY, while they are Wxating the centre of a blankWeld induces reproducible illusory motion (moving phosph-enes). But when the TMS was applied to area MT in GY’s‘blind’ hemisphere he never experienced phosphenes,stationary or moving (Cowey and Walsh 2000), suggestingeither that area MT is the wrong target or that in the absenceof V1 area MT/V5 on the blind side is not physiologically

123

18 Exp Brain Res (2010) 200:3–24

receptive to TMS. The latter possibility was supported in asubsequent study where TMS was applied over MTV5 ofboth hemispheres in GY at a variety of stimulus onset asyn-chronies (Silvanto et al. 2007). GY now perceived phosph-enes, “for the very Wrst time” he declared, in both hisnormal and his ‘blind’ Welds and the most important factorwas the strength of the TMS delivered to the normal hemi-sphere, indicating that it is the projection from the latter, viathe corpus callosum, that ‘primes’ area MT/V5 of the blindhemisphere. The bilateral phosphenes can also be colouredif GY is Wrst adapted to coloured backgrounds but it is thebackground colour in the seeing hemiWeld that determinesthe colour of the phosphene in the blind hemiWeld (Silvantoet al. 2008), further indicating the ‘priming’ arises in thenormal hemisphere. GY is known, from recent diVusionimaging (Bridge et al. 2008) to possess a hypernormal con-nection between area MT/V5 of the normal and damagedsides of the brain. This connection could explain the abovephenomena and the Wnding (Silvanto et al. 2009) that whenTMS is applied above MT/V5 of the blind hemisphere itinXuences the appearance of phosphenes evoked by TMSapplied above V1 of the normal hemisphere in a mannernever observed with normal subjects.

Awareness without seeing: Type2 blindsight

There are now several reports that patients with blindsightcan be conWdently aware that a stimulus was presented butwithout their experiencing a visual percept. This idea isneither as new nor as counterintuitive as might seem, forCowey and Weiskrantz (1963) speculated that monkeysmight detect the visual stimuli delivered in their Welddefects by ‘noticing’, and learning the relevance of, a ten-dency to make a saccadic eye movement provoked by avisual target. Despite this, awareness unaccompanied bysome form of visual quale has not been widely accepted.The most recent evidence from electrophysiological workwith monkeys was not designed to address this issue and isrelevant but unfortunately inconclusive. By studying theability of normal monkeys to indicate in which of two briefintervals microstimulation was delivered to the cortex viaindwelling electrodes that could not stimulate any struc-tures outside the cortex, Murphey and Maunsell (2007)showed that monkeys could discriminate the stimulation inseveral extrastriate visual areas at stimulation levels littlediVerent from those in striate cortex, indicating that neuro-nal signals evoked in any part of visual cortex can generatepercepts. The areas included frontal eye Welds, where TMSin human subjects does not evoke visual qualia. In asubsequent similar investigation on human subjects withindwelling electrodes Murphey et al. (2009) found thatmicrostimulation of early extrastriate visual areas was not

only detectable but evoked visual percepts (as does TMS).But it remains unclear whether identical unilateral microsti-mulation in the absence of V1, could sustain detectionwithout any accompanying visual percept, as does TMS inGY. We are tantalisingly close to solving this problem.

When asked about their awareness without seeing any-thing, blindsight subjects provide many descriptions oftheir ‘feeling’, one of which is that it is just a ‘gut feeling’.This might be more than metaphorical. By giving multiplebolus infusions of the short lasting beta adrenergic agonistisoproterenol to normal subjects and then asking them todetect and rate any changes in their interoceptive state,notably cardiac and respiratory sensations, Khalsa et al.(2009) showed in a double-blind and placebo-controlledstudy that all 15 subjects achieved a high dose-dependentcorrelation between their experienced internal feelings andthe drug. If unseen visual stimuli such as Xicker, motion,and changes in intensity in blindsight can inXuence auto-nomic systems they might be detectable by these means,although this could hardly explain abilities like orientationdiscrimination.

Endogenous phenomenal visual percepts in blindsight?

Although it has proved to be little short of impossible toinduce phenomenal visual percepts in the blind Weld, mightthey arise spontaneously, or is the extensive swathe ofextrastriate visual areas incapable of generating them?Given the evidence cited above about microstimulation ofthese areas in normal subjects it seems extraordinary thatthe blind Weld should be forever blind. Several lines ofinvestigation bear on this issue. Farah et al. (1992) studieda patient before and after unilateral occipital lobectomy andfound that her ability to imagine common objects that Wlledthe mental visual Weld changed after lobectomy in line withher hemianopia, failing to include the hemianopic Weld.Using a diVerent approach, several investigators reportedthat if part of a visual stimulus occupies the blind Weld, thestimulus might be phenomenally completed (for review seeStoerig and Cowey 1997). However, this might beexpected, given that completion of Wgures occurs whenparts of them cross the natural blind spot in normal vision.But what of dreams and hallucinations? Unsurprisingly, theformer have proved too diYcult to evaluate. However,there are many reports that patients experience visual hallu-cinations in their blind Weld, especially in the early stagesof recovery from stroke, e.g. Kölmel (1985), which are eas-ier to interpret because the patient is conscious and canlocalise the hallucination with respect to the line of sight.They presumably arise from pathological activation ofextrastriate areas, although this leaves mysterious why theycannot be reproduced by TMS unless their source is deep in

123

Exp Brain Res (2010) 200:3–24 19

the temporal lobe and therefore beyond the reach of TMS.But their existence indicates that ipsilesional extrastriatecortex is not doomed to permanent blindness.

Are brain rhythms important for awareness and attention in blindsight?

Following the discovery of the EEG it became clear that inall animal brains there is an underlying rhythm to the col-lective neural discharges, which varies with the behaviouralstate. But it was not until the early 1990s that particularrhythms in localized regions of the cerebral cortex wererelated to speciWc behavioural conditions like selectivelyattending to a particular stimulus (Singer 1993). Theirimportance has been repeatedly contested, in large partbecause of the absence of any explanation of why a rhythmmight add anything useful to neuronal processing. But thecomputational advantage of rhythms, especially withrespect to the ‘binding problem’ concerning how informa-tion about diVerent features of a sensory stimulus that areregistered by diVerent sets of neurons, often in diVerent cor-tical areas, can lead to a uniWed perception of an object orselective attention to that object, has been splendidlyexplained in the monograph by Buzsaki (2006). The impor-tance of rhythms to cognitive mechanisms has often beenquestioned on the grounds that rhythms occur throughoutthe brain, even in the eye! True, but it is the power of rhyth-mic activity in diVerent frequency wavebands in diVerentregions that seems to matter, not their existence. A contraryview is presented by Gold (1999), who argues that “… untilwe can separate the problems of visual binding and visualawareness experimentally, the role, if any, of 40-Hz oscilla-tion in consciousness will remain an open question”. Theyhave certainly not yet been separated!

Given the above, it is easy to appreciate why measuresof the power in diVerent frequency bands and their correla-tion with diVerent cognitive states, like sensory awareness,are increasingly popular and blindsight is no exception.Two examples must suYce. Schurger et al. (2006) usedMEG to test the hypothesis that gamma oscillations signalthe entry of a neural representation into awareness in GY,while controlling for other measures of neural informationprocessing such as discrimination accuracy and reactiontime. GY was tested on an orientation-discrimination taskusing stationary stimuli at a near-threshold level of con-trast, to which he sometimes responded “aware” and some-times “unaware”, despite never having a visual percept.GY’s above-chance accuracy was no diVerent whether ornot he reported awareness of the stimulus. But activity inthe gamma band (44–66 Hz) above the occipito-parietalregion of the blind hemisphere correlated signiWcantly withawareness (but not accuracy), whereas activity in the alpha

band (8–12 Hz) did not. In a further study Schurger et al.(2008) used similar stimuli and methods to study the eVectof cueing attention to speciWc locations in GY’s blind Weld.They found increased gamma band activity—again in dor-sal parietal cortex—in relation to selective attention thatwas independent of reported eVects of awareness, as well aseVects of awareness that were independent of attentionalstate. Increased gamma band activity therefore indicatesboth awareness and directed attention even in a blind hemi-Weld. However, like many correlational studies, it does notof itself physiologically explain how the presumed top-down modulation of attention drives regionally selectivegamma band activity, nor how such activity instigatesawareness (pace Buszáki), which may prove to be anotherexample of the so-called hard problem of consciousness(Chalmers 1996).

The EEG has good temporal but not spatial resolution.Nevertheless, like MEG, it should be able to indicate pre-cisely when and roughly where events accompanying andperhaps underlying blindsight are taking place. There areseveral studies of cortical potentials evoked by visual stim-uli presented in Weld defects (for review see Rao et al.2001). Still unaccountably, only the study by Vytche et al.(1995) reported an early evoked potential in area MT/V5.Rao et al. did record a potential in the region of MT/V5 insubject GY when fast moving stimuli were used but it waseven later than on the normal side, which itself was laterthan in V1. Especially late are potentials that correlate withawareness of the stimuli (Weiskrantz et al. 2003). They areprominent in ventrolateral prefrontal cortex (area 46) wherethey also occur in fMRI studies but with unknown latency(Sahraie et al. 1997). But as with fMRI and MEG, theirexistence does not show where or how the awareness ofType II blindsight is initiated.

Rehabilitation and “learning blindsight”

Rehabilitation in patients with brain damage is a majormedical concern. A unilateral stroke that produces hemiple-gia is severely disabling and research into its alleviationabounds. But although hemianopia can slow the patient’sability to read and in many countries prevents them fromdriving legally, the symptoms are rarely enduringly dis-tressing and hemianopic patients and monkeys appear to benormal to all but experienced observers. This is because theblind hemiWeld is not experienced as a plight, unlike aparalysed limb. Hemianopic subjects learn instead to usetheir intact hemiWeld to view and scan the visual scene.How else can subject GY perform downhill skiing?! Andrecent research shows how training hemianopic subjects tomake exploratory eye movements into the blind Weldimproves their performance, for example reading (e.g.

123

20 Exp Brain Res (2010) 200:3–24

Schuett et al. 2008). Nonetheless, there are increasingattempts to see whether residual visual sensitivity in theblind Weld itself improves with practice and—especiallyrecently—to correlate any improvements with concomitantneuroimaging. The commonest design of such investiga-tions is to give monkeys or patients repeated practice atdetecting targets in their blind Weld (e.g. Cowey 1967;Cowey and Stoerig 1995, Wg 16; Zihl and von Cramon1985; Kasten and Sabel 1995; Kasten et al. 1998) but thisencounters the following problems: 1. Improvements mightreXect nothing more than the improvement that occurs innormal vision with repeated practice at stimuli close tothreshold. 2. Subjects might learn to notice the kinds ofcues described earlier under artefacts. 3. They are learningthe relevance of being aware of something even though it isnot phenomenally visual. 4. They are changing theirresponse bias. It is such factors that have led to the occa-sionally astringent controversies about the nature ofimproved performance (e.g. Reinhard et al. 2005; Horton2005). Aware of these problems, several recent investiga-tions provide good evidence for genuine training-basedrecovery of sensitivity in Weld defects. For example, Huxlinand Pasternak (2004) showed prominent and stable recov-ery of motion discrimination in the Weld defects of cats withneurochemical extrastriate lesions, but only in the trainedand not in the untrained location. Using comparable meth-ods Huxlin (2008) obtained similar results in two patientswith initially absolute Weld defects, who also began to usetheir recovered sensitivity to track moving objects in theblind Weld. In their most recent investigation, in which thepossible artefacts mentioned above were eliminated, Huxlinet al. (2009) report that in Wve subjects with initially abso-lute Weld defects produced by damage to V1, extensivetraining over months led to the restoration of almost normallevels of discriminating global motion, including its direc-tion, and that the restoration was conWned to the trainedarea within the scotoma, and that there was even accompa-nying recovery of conscious visual perception. An indica-tion of the likely neural basis is provided by Henrikssonet al. (2006) who, using MEG and fMRI, found that thetraining- induced improvement of Xicker detection in ahemianopic subject correlated with increased activation ofarea MT/V5 and areas V1, V2, V3 and V3A of the undam-aged hemisphere. Further examples and a comprehensivereview of this topic are provided by Stoerig (2007).

Conclusions

Much of the controversy that has bedevilled blindsightarises from the use of terms that are not satisfactorily deW-ned. For example, subjects are too seldom asked to describeexactly what they mean when they say they are ‘aware’, the

fact that awareness is itself liable to bias is ignored, andexperimenters too often give the impression that if twostimuli can be told apart the discrimination is based on aparticular quality, like colour or motion, rather than onaccompanying properties like salience or change in posi-tion. This general problem was identiWed well over a cen-tury ago by Dodds (1885, p 22), who bleakly wrote, in adiscussion of brain damage and localization of function,…”. Much confusion has arisen from the introduction ofthe terms of a subjective psychology into the study ofobjective brain physiology. Writers have puzzled theirreaders, and often we fear themselves, in their attempts todeWne consciousness, and distinguish between sensation,perception, apperception, cognition and the like”. But evenDodds would have to agree that we are making progress.

A further problem concerns how to integrate the disparateapproaches to investigating blindsight. For example, canTMS indicate which of the many areas activated in FMRIare essential for awareness?; can the putative importance ofrhythmic activity in particular frequency bands be assessedby imposing or disrupting such activity?; can single unitrecordings from cells in diVerent visual areas of monkeyswith damage to V1 be correlated with the animal’s concom-itant behavioural performance?; might deep brain stimula-tion in blindsight patients produce phenomenal percepts inthe blind Weld?; could modern DTI tractography in patientswith damage to V1 uncover connexions that might explainthe basis of any plasticity, whether it is age related, and thequalities of spared or recovered visual sensitivity. An endur-ing problem concerns why V1, and only V1, is indispens-able for phenomenal vision, as much evidence suggests,given theoretical network accounts of consciousness and therole of top-down processing, which ‘should’ be intact. Astart has already been made on these problems, as men-tioned in this review, but there is some way to go.

Acknowledgments This research was supported by a UK MedicalResearch Council project grant. I am specially grateful to Dr. IonaAlexander for her help with Wgures and editing. It is a pleasure toacknowledge Dr. Robert Doty and the University of Rochester, NewYork, for jointly inviting me to present the Wrst Elizabeth Doty Memo-rial Lecture in 2008, which prompted me to prepare this review. I alsothank Charles Heywood for alerting me to the 19th century writings ofW.J. Dodds.

References

Alexander I, Cowey A (2009) The cortical basis of global motiondetection in blindsight. Exp Brain Res 192:407–412

Azzopardi P, Cowey A (1997) Is blindsight like normal, near-thresholdvision? Proc Natl Acad Sci USA 94:14190–14194

Azzopardi P, Cowey A (1998) Blindsight and visual awareness.Conscious Cogn 7:292–311

Azzopardi P, Cowey A (2002) Cerebral activity related to guessing andattention during a visual detection task. Cortex 38:833–836

123

Exp Brain Res (2010) 200:3–24 21

Azzopardi P, Fallah M, Gross CG, Rodman HR (2003) Responselatencies of neurons in visual areas MT and MST of monkeys withstriate cortex lesions. Neuropsychologia 41:1738–1756

Barbur JL, Watson JD, Frachowiak RSJ, Zeki S (1993) Consciousvisual perception without V1. Brain 116:1293–1302

Bender MB, Krieger HP (1951) Visual function in perimetrically blindWelds. Arch Neurol Psychiat 65:72–79

Bennett MR, Hacker PMS (2003) Philosophical foundations of neuro-science. Blackwell, Oxford, p 461

Benson PJ, Guo L, Blakemore C (1998) Direction discrimination ofmoving gratings and plaids and coherence of dot displays withoutprimary visual cortex (V1). Eur J NeuroSci 10:3767–3772

Blythe IM, Kennard C, Ruddock KH (1987) Residual vision in patientswith retrogeniculate lesions of the visual pathways. Brain110:887–905

Braddick OJ, Atkinson J, Hood B, Harkness W, Jackson G, Vargha-Khadem F (1992) Possible blindsight in infants lacking one cere-bral hemisphere. Nature 360:461–463

Bridge H, Thomas O, Jbabdi S, Cowey A (2008) Changes in connec-tivity after visual cortical brain damage underlie altered visualfunction. Brain 131:1433–1444

Brindley GS, Gautier-Smith PC, Lewin W (1969) Cortical blindnessand the functions of the non-geniculate Wbres of the optic tracts.J Neurol Neurosurg Psychiat 32:259–264

Brown LE, Kroliczac G, Demonet J-F, Goodale MA (2007) A hand inblindsight: hand placement near target improves size perceptionin the blind visual Weld. Neuropsychologia 46:786–802

Bullier J, Girard P, Salin P-A (1994) The role of area 17 in the transferof information to extrastriate visual cortex. In: Peters A, RocklandKS (eds) Cer cortex, vol 10. Plenum Press, New York, pp 301–330

Buzsaki G (2006) Rythms of the brain. Oxford University Press,Oxford

Buzsaki G, Draguhn A (2004) Neuronal oscillations in corticalnetworks. Science 304:1926–1929

Campion J (2007) Blindsight–Fact or Myth? Psychologist 20:207–208Campion J, Latto R, Smith YM (1983) Is blindsight an eVect of

scattered light, spared cortex, and near-threshold vision? BehavBrain Sci 3:423–486

Chalmers D (1996) The conscious mind: in search of a fundamentaltheory. Oxford University Press, Oxford

CliVord CWG, Arabzadeh E, Harris JA (2008) Getting technical aboutawareness. Trends Cog Sci 12:54–58

Corbetta M, Marzi CA, Tassinari G, Aglioti S (1990) EVectiveness ofdiVerent task paradigms in revealing blindsight. Brain 113:603–616

Cowey A (1967) Perimetric study of visual Weld defects in monkeysafter cortical and retinal ablations. Quart J Exp Psychol 19:232–245

Cowey A (1974) Atrophy of retinal ganglion cells after removal ofstriate cortex in a rhesus monkey. Perception 3:257–260

Cowey A (2004) Fact, artefact and myth about blindsight. Quart J ExpPsychol 57A:577–609

Cowey A, Azzopardi P (2001) Is blindsight motion blind? In: deGelder B, de Haan E, Heywood CA (eds) Out of mind. OxfordUniversity Press, Oxford, pp 87–103

Cowey A, Stoerig P (1989) Projection patterns of surviving neurons inthe dorsal lateral genicualte nucleus following discrete lesions ofstriate cortex: implications for residual vision. Exp Brain Res75:631–638

Cowey A, Stoerig P (1995) Blindsight in monkeys. Nature 373:247–249

Cowey A, Stoerig P (1997) Visual detection in monkeys with blind-sight. Neuropsychologia 35:929–939

Cowey A, Stoerig P (1999) Spectral sensitivity in hemianopic macaquemonkeys. Eur J NeuroSci 11:2114–2120

Cowey A, Stoerig P (2003) Stimulus cueing in blindsight. Prog BrainRes 144:261–277

Cowey A, Walsh V (2000) Magnetically induced phosphenes in sight-ed, blind and blindsighted observers. NeuroReport 11:3269–3273

Cowey A, Weiskrantz L (1963) A perimetric study of visual Welddefects in monkeys. Quart J Exp Psychol 15:91–115

Cowey A, Stoerig P, Perry VH (1989) Transneuronal retrogradedegeneration of retinal ganglion cells after damage to striate cor-tex in macaque monkeys: selective loss of Pß cells. Neuroscience29:65–80

Cowey A, Stoerig P, Bannister M (1994) Retinal ganglion cellslabelled from the pulvinar nucleus in macaque monkeys. Neuro-science 61:691–705

Cowey A, Stoerig P, Le Mare C (1998) EVects of unseen stimuli onreaction times to seen stimuli in monkeys with blindsight.Conscious Cogn 7:312–323

Cowey A, Stoerig P, Williams C (1999) Variance in transneuronalretrograde ganglion cell degeneration in monkeys after removalof striate cortex: eVects of size of the cortical lesion. Vision Res39:3642–3652

Cowey A, Johnson H, Stoerig P (2001) The retinal projection to thepregeniculate nucleus in normal and destriate monkeys. EurJ NeuroSci 13:279–290

Dacey DM, Lee BB (1994) The blue-ON opponent pathway in primateretina originates from a distinct bistratiWed ganglion cell type.Nature 367:731–735

De Gelder B, Vroomen J, Pourtois G, Weiskrantz L (1999) Non-conscious recognition of aVect in the absence of striate cortex.NeuroReport 10:3759–3763

De Gelder B, Vroomen J, Pourtois G (2001) Covert aVective cognitionand aVective blindsight. In: de Gelder B, de Haan E, HeywoodCA (eds) Out of mind. Oxford University Press, Oxford, England,pp 205–221

De Gelder B, Tamietto M, van Boxtel G, Goebel R, Sahraie A, van denStock J, Steinen BMC, Weiskrantz L, Pegna A (2009) Intactnavigation skills after bilateral loss of striate cortex. Curr Biol18:R1128–R1129

Dienes Z, Scott R (2005) Measuring unconscious knowledge:distinguishing structural knowledge and judgement knowledge.Psychol Res 69:338–351

Dineen J, Hendrickson A (1981) Age-correlated diVerences in theamount of retinal degeneration after striate cortex lesions inmonkeys. Invest Ophthalmol 21:749–752

Dineen J, Keating EG (1981) The primate visual system after bilateralremoval of striate cortex. Survival of complex pattern vision. ExpBrain Res 41:338–345

Dineen J, Hendrickson A, Keating EG (1982) Alterations of retinalinputs following striate cortex removal in adult monkey. ExpBrain Res 47:446–456

Dodds WJ (1885) On some central aVections of vision. Part I. Brain8:21–39

Evans S, Azzopardi P (2007) Evaluation of a ‘bias-free’ measure ofawareness. Spat Vis 20:61–77

Farah MJ, Soso MJ, DashieV RM (1992) Visual angle of the mind’seye before and after unilateral occipital lobectomy. J Exp PsycholHum Percep Perform 18:241–246

Faubert J, Diaconu V (2001) From visual consciousness to spectralabsorption in the human retina. Prog Brain Res 134:300–409

Faubert J, Diaconu V, Ptito M, Ptito A (1999) Residual vision in theblind Weld of hemidecorticated humans predicted by a diVusionscatter model and selective spectral absorption of the human eye.Vision Res 39:149–157

Vytche DH, Guy CN, Zeki S (1995) The parallel visual motion inputsinto areas V1 and V5 of human cerebral cortex. Brain 118:1375–1394

Goebel R, Muckli L, Zanella FE, Singer W, Stoerig P (2001) Sustainedextrastriate cortical activation without visual awareness revealed byfMRI studies of hemianopic patients. Vision Res 41:1459–1474

123

22 Exp Brain Res (2010) 200:3–24

Gold I (1999) Does the 40-Hz oscillation play a role in visualconsciousness? Conscious Cogn 8:186–195

Gonzalez-Andino SL, Grave de Peralta Menendez R, Khateb A,Landis T, Pegna AJ (2009) Electrophysiologicl correlates ofaVective blindsight. NeuroImage 44:581–589

Green DM, Swets JA (1966) Signal detection theory and psychophys-ics. Wiley, New York

Hackley SA, Johnson SL (1996) Distinct early and late components ofthe photic blink relex: I. Response characteristics in patients withretrogeniculate lesions. Pathophysiol 33:239–251

Heide W, Koenig E, Dichgans J (1990) Optokinetic nystagmus, selfmotion sensation and their after eVects in patients with occipito-parietal lesions. Clin Vis Sci 5:145–156

Hendry SHC, Reid RC (2000) The koniocellular pathway in primatevision. Ann Rev Neurosci 23:127–153

Henriksson L, Raninen A, Näsänen R, Hyvärinen L, Vanni S (2006)Training induced cortical representation of a hemianopic Weld.J Neurol Neurosurg Psychiat 78:74–81

Holliday IE, Anderson SJ, Harding GFA (1997) Magnetoencephalo-graphic evidence for non-geniculostriate visual input to humancortical area V5. Neuropsychologia 35:1139–1146

Holmes G (1918) Disturbances of vision by cerebral lesions. BritJ Ophthal 2:353–384

Horton JC (2005) Disappointing results from NovaVision’s visualrestoration therapy. Brit J Ophthalmol 89:1–2

Humphrey NK (1974) Vision in a monkey without striate cortex: a casestudy. Perception 3:241–255

Huxlin KR (2008) Perceptual plasticity in damaged adult visualsystems. Vision Res 48:2154–2166

Huxlin KR, Pasternak T (2004) Training induced recovery of visualmotion perception after extrastriate cortical damage in the adultcat. Cer Cortex 14:81–90

Huxlin KR, Martin T, Kelly K, Riley M, Friedman DI, Burgin WS,Hayhoe M (2009) Perceptual relearning of complex visual motionafter V1 damage in humans. J Neurosci 29:3981–3991

Hyman J (1991) Visual experience and blindsight. In: Hyman J (ed)Investigating psychology: sciences of the mind after Wittgen-stein. Routledge, London, pp 166–200

Innocenti GM, Kiper DC, Knyazeva MG, Deonna TW (1999) Natureand limits of cortical developmental plasticity in a child. J RestNeurol Neurosci 15:219–227

Jindahra P, Petrie A, Plant GT (2009) Retrograde trans-synaptic retinalganglion cell loss identiWed by optical coherence tomography.Brain. doi:10.1093/brain/awp001)

Kasten E, Sabel BA (1995) Visual Weld enlargement after computer-training in brain-damaged patients with homonymous deWcits—an open pilot trial. Rest Neurol Neurosci 8:113–127

Kasten E, Wüst S, Behrens-Baumann W, Sabel BA (1998) Computer-based training for the treatment of partial blindness. Nature Med4:1083–1087

Keating EG (1975) EVects of striate and prestriate lesions on the mon-key’s ability to locate and discriminate visual forms. Exp Neurol47:16–25

Keating EG (1979) Rudimentary color vision in the monkey afterremoval of striate and preoccipital cortex. Brain Res 179:379–384

Kentridge RW, Heywood CA, Weiskrantz L (1999a) EVects of tempo-ral cueing on residual visual discrimination in blindsight. Neuro-psychologia 37:479–483

Kentridge RW, Heywood CA, Weiskrantz L (1999b) Attention withoutawareness in blindsight. Proc Roy Soc Lond B 266:1805–1911

Khalsa SS, Rudrauf D, Sandesara C, Olshansky B, Tranel D (2009)Bolus isoproterenol infusions provide a reliable method forassessing interoceptive awareness. Int J Psychophysiol 72:34–45

King SM, Cowey A (1992) Defensive responses to looming visualstimuli in monkeys with unilateral striate cortex ablation. Neuro-psychologia 30:1017–1024

King SM, Azzopardi P, Cowey A, Oxbury J, Oxbury S (1996) The roleof light scatter in the residual visual sensitivity of patients withcomplete cerebral hemispherectomy. Visual Neurosci 13:1–13

Kiper DC, Zesiger P, Maeder P, Deonna T, Innocenti GM (2002)Vision after early–onset lesions of the occipital cortex: 1. Neuro-psychological and psychophysical studies. Neural Plasticity 9:1–25

Klüver H (1941) Visual functions after removal of the occipital lobes.J Psychol 11:23–45

Kölmel HW (1985) Compex visual hallucinations in the hemianopicWeld. J Neurol Neurosurg Psychiat 48:293–298

Kunimoto C, Miller J, Pashler H (2001) ConWdence and accuracy ofnear-threshold discrimination responses. Conscious Cogn10:294–340

Leporé F, Cardu B, Rasmussen T, Malmo RB (1975) Rod and conesensitivity in destriate monkeys. Brain Res 93:203–221

Macmillan NA, Creelman CD (1991) Detection theory: a user’s guide.Cambridge University Press, Cambridge

Macphail EM (1998) The evolution of consciousness. Oxford Univer-sity Press, Oxford

Magoun HW, Ranson SW (1935) The central path of the light reXex.Arch Ophthalmol 13:791–811

Malmo RB (1966) EVects of striate cortex ablation on intensitydiscrimination and spectral intensity distribution in the rhesusmonkey. Neuropsychologia 4:9–16

Marcel AJ (1983) Conscious and unconscious perception: an approachto the relations between phenomenal experience and perceptualprocesses. Cog Psychol 15:238–300

Marcel AJ (1998) Blindsight and shape perception: deWcit of visualconsciousness or of visual function? Brain 121:1565–1588

Marzi CA, Tassinari G, Aglioti S, Lutzemberger L (1986) Spatial sum-mation across the vertical meridian in hemianopics: a test ofblindsight. Neuropsychologia 30:783–795

Mihailovic LT, Cupic D, Dekleva N (1971) Changes in the number ofneurons and glial cells in the lateral geniculate nucleus of themonkey during retrograde cell degeneration. J Comp Neurol142:223–230

Mole C, Kelly S (2006) On the demonstration of blindsight in mon-keys. Mind and Language 21:475–483

Moore T, Rodman HR, Repp AB, Gross CG (1995) Localization ofvisual stimuli after striate cortex damage in monkeys: parallelswith human blindsight. Proc Natl Acad Sci USA 92:8215–8218

Moore T, Rodman HR, Gross CG (1998) Man, monkey and blindsight.Neuroscientist 4:227–230

Morris JS, DeGelder B, Weiskrantz L, Dolan RJ (2001) DiVerentialextrageniculate and amygdala responses to presentation ofemotional faces in a cortically blind Weld. Brain 124:1241–1252

Murphey DK, Maunsell JHR (2007) Behavioral detection of electricalmicrostimulation of cortical visual areas. Curr Biol 17:862–867

Murphey DK, Maunsell JHR, Beauchamp MS, Yoshor D (2009)Perceiving electrical stimulation of identiWed human visual areas.Proc Natl Acad Sci USA 106:5389–5393

Pasik P, Pasik T (1964) Oculomotor function in monkeys with lesionsof the cerebrum and the superior colliculi. In: Bender MB (ed)The oculomotor system. Hoeber, New York, pp 40–80

Pasik P, Pasik T (1982) Visual functions in monkeys after total remov-al of visual cerebral cortex. Contrib Sensory Physiol 7:147–200

Pasik P, Pasik T, Schilder P (1969) Extrageniculostriate vision in themonkey: discrimination of luminous Xux-equated Wgures. ExpNeurol 24:421–437

Pasik P, Pasik T, Valciukas JA (1970) Nystagmus induced by station-ary repetitive light Xashes in monkeys. Brain Res 19:313–317

Pegna AJ, Khateb A, Lazeyras F, Seghier ML (2005) Discriminatingemotional faces without primary visual cortices involves the rightamygdala. Nat Neurosci 8:24–25

Perenin MT (1991) Discrimination of motion direction in perimetri-cally blind Welds. NeuroReport 2:397–400

123

Exp Brain Res (2010) 200:3–24 23

Perenin MT, Jeannerod M (1975) Residual visual functions in corti-cally blind hemiWelds. Neuropsychologia 13:1–7

Perenin MT, Rossetti Y (1996) Grasping without form discriminationin a hemianopic Weld. NeuroReport 7:793–797

Perenin MT, Ruel J, Hecaen H (1980) Residual visual capacities in acase of cortical blindness. Cortex 16:605–612

Persaud N, Cowey A (2008) Blindsight is unlike normal conscious vision:evidence from an exclusion task. Conscious Cogn 17:1050–1055

Persaud N, McLeod P, Cowey A (2007) Post-decision wagering objec-tively measures awareness. Nature Neurosci 10:257–261

Pizzamiglio L, Antonucci G, Francia A (1984) Response of the corti-cally blind hemiWelds to a moving visual scene. Cortex 20:89–99

Pöppel E (1986) Long-range colour-generating interactions across theretina. Nature 320:523–525

Pöppel E, Frost D, Held R (1973) Residual visual function after brainwounds involving the central visual pathways in man. Nature,London 243:295–296

Rafal R, Smith W, Krantz J, Cohen A, Brennan C (1990) Extragenicu-late vision in hemianopic humans: saccade inhibition by signals inthe blind Weld. Science 250:118–121

Rao A, Nobre AC, Cowey A (2001) Disruption of evoked potentialsfollowing a V1 lesion: implications for blindsight. In: De GelderB, de Haan E, Heywood CA (eds) Out of mind. Oxford UniversityPress, Oxford, pp 69–86

Reinhard J, Schreiber A, Schiefer U, Kasten E, Sabel BA, Kenkel S,Vonthein R, Trauzettel-Klosinski S (2005) Does visual restitutiontraining change absolute homonomous visual Weld defects? Afundus controlled study. Br J Ophthalmol 89:30–35

Richards W (1973) Visual processing in scotomata. Exp Brain Res17:333–347

Riddoch G (1917) Dissociation of visual perceptions due to occipitalinjuries, with especial reference to appreciation of movement.Brain 40:15–57

Rodman HR, Gross CG, Albright TD (1989) AVerent basis of visualresponse properties in area MT of the macaque: I. eVects of striatecortex removal. J Neurosci 9:2033–2050

Rodman HR, Gross CG, Albright TD (1990) AVerent basis of visualresponse properties in area MT of the macaque. II. EVects ofsuperior colliculus removal. J Neurosci 10:1154–1164

Rosenblum LD, Gordon M, Jarquin L (2000) Echolocating distance bymoving and stationary listeners. Ecol Psychol 12:181–186

Sahraie A, Weiskrantz L, Barbur JL, Simmons A, Williams SCR,Brammer MJ (1997) Pattern of neuronal activity associated withconscious and unconscious processing of visual signals. Proc NatlAcad Sci USA 94:9406–9411

Sanders MD, Warrington EK, Marshall J, Weiskrantz L (1974) ‘Blind-sight’: vision in a Weld defect. Lancet 1:707–708

Schilder P, Pasik P, Pasik T (1971) Extrageniculostriate vision in themonkey. II. Demonstration of brightness discrimination. BrainRes 32:383–389

Schilder P, Pasik P, Pasik T (1972) Extrageniculostriate vision in themonkey. III. Circle vs triangle and ‘red vs green’ discrimination.Exp Brain Res 14:436–448

Schmid MC, Panagiotaropoulos T, Augath MA, Logothetis NK,Smirnakis SM (2009) Visually driven activation in macaque areasV2 and V3 without input from the primary visual cortex. PLoSONE 4(e5527):1–14

Schuett S, Heywood CA, Kentridge RW, Zihl J (2008) Rehabilitationof hemianopic dyslexia: are words necessary for re-learningoculomotor control? Brain 131:3156–3168

Schurger A, Cowey A, Tallon-Baudry C (2006) Induced gamma-bandoscillations correlate with awareness in hemianopic patient GY.Neuropsychologia 44:1796–1803

Schurger A, Cowey A, Cohen J, Treisman A, Tallon-Baudry C (2008)Distinct and independent correlates of attention and awareness ina hemianopic patient. Neuropsychologia 46:2189–2197

Schwitzgebel E, Gordon MS (2000) How well do we know our ownconscious experience? The case of human echolocation. PhilosTopics 28:235–246

Silvanto J, Cowey A, Lavie N, Walsh V (2007) Making the blindsightedsee. Neuropsychologia 45:3346–3350

Silvanto J, Cowey A, Walsh V (2008) Inducing conscious perceptionof colour in blindsight. Curr Biol 18:950–951

Silvanto J, Walsh V, Cowey A (2009) Abnormal functional connectiv-ity between ipsilesional V5/MT + and contralesional striatecortex (V1) in blindsight. Exp Brain Res 193:645–650

Singer W (1993) Synchronisation of cortical activity and its putativerole in information processing and learning. Ann Rev Physiol55:349–374

Stoerig P (2006) Blindsight, conscious vision, and the role of theprimary visual cortex. Prog Brain Res 155:217–234

Stoerig P (2007) Functional rehabilitation of partial cortical blindness.Restorative Neurol Neurosci 25:1–14

Stoerig P, Cowey A (1989) Residual target detection as a function ofstimulus size. Brain 112:1123–1139

Stoerig P, Cowey A (1991) Increment-threshold spectral sensitivity inblindsight. Brain 114:1487–1512

Stoerig P, Cowey A (1992) Wavelength discrimination in blindsight.Brain 115:425–444

Stoerig P, Cowey A (1997) Blindsight in man and monkey. Brain120:535–559

Stoerig P, Cowey A (2009) Blindsight. In: Baynes T, Cleermans A,Wilken P (eds) Oxford companion to consciousness. OxfordUniversity Press, Oxford, pp 112–116

Stoerig P, Fahle M (1995) Apparent motion across a scotoma: animplicit test of blindsight. Eur J Neurosci Suppl 8:76

Stoerig P, Pöppel E (1986) Eccentricity-dependent residual targetdetection in visual Weld defects. Exp Brain Res 64:469–475

Stoerig P, Hubner M, Pöppel E (1985) Signal detection analysis ofresidual vision in a Weld defect due to a post-geniculate lesion.Neuropsychologia 23:589–599

Stoerig P, Zontanou A, Cowey A (2002) Aware or unaware: assess-ment of cortical blindness in four men and a monkey. Cer Cortex12:565–574

Supa M, Cotzin M, Dallenbach KM (1944) Facial vision: the percep-tion of obstacles by the blind. Am J Psychol 57:133–183

ter Braak JG, van Vliet AGM (1963) Sub-cortical optokinetic nystag-mus in the monkey. Psychiat Neurol Neurochirurg 66:277–283

ter Braak JWG, Schenk VWD, van Vliet AGM (1971) Visual reactionsin a case of long-standing cortical blindness. J Neurol NeurosurgPsychiat 34:140–147

Theoret H, Boire D, Herbin M, Ptito M (2001) Anatomical sparing inthe superior colliculus of hemispherectomized monkeys. BrainRes 16:274–280

Tomaiuolo F, Ptito M, Marzi CA, Paus T, Ptito A (1997) Blindsight inhemsipherectomized patients as revealed by spatial summationacross the vertical meridian. Brain 120:795–803

Treisman M (1964) The eVect of one stimulus on the threshold ofanother: an application of signal detectability theory. BritJ Statistical Psychol 17:15–35

Trevethan CT, Sahraie A, Weiskrantz L (2007) Can blindsight besuperior to ‘sighted sight’? Cognition 103:491–501

Van Buren JM (1963) Trans-synaptic retrograde degeneration in thevisual system of primates. J Neurol Neurosurg Psychiat 34:140–147

Van Hof-van Duin J, Mohn G (1983) Optokinetic and spontaneousnystagmus in children with neurological disorders. Behav BrainRes 10:163–175

Weiskrantz L (1963) Contour discrimination in a young monkey withstriate cortex ablation. Neuropsychologia 1:145–164

Weiskrantz L (1987) Residual vision in a scotoma: follow-up study ofform discrimination. Brain 110:77–92

123

24 Exp Brain Res (2010) 200:3–24

Weiskrantz L (1998) Consciousness and commentaries. In: Towards ascience of consciousness II—the second Tucson discussions anddebates, M.I.T. Press, Cambridge, pp 371–377

Weiskrantz L (2009) Blindsight: a case study spanning 35 years and newdevelopments, 3rd edn. Oxford University Press, Oxford, p 255

Weiskrantz L, Warrington EK, Sanders MD, Marshall J (1974) Visualcapacity in the hemianopic Weld following a restricted corticalablation. Brain 97:709–728

Weiskrantz L, Cowey A, Passingham C (1977) Spatial responses tobrief stimuli by monkeys with striate cortex ablations. Brain100:655–670

Weiskrantz L, Barbur JL, Sahraie A (1995) Parameters aVecting con-scious versus unconscious visual discrimination in a patient withdamage to the visual cortex (V1). Proc Natl Acad Sci USA92:6122–6126

Weiskrantz L, Cowey A, Barbur JL (1999) DiVerential pupillary con-striction and awareness in the absence of striate cortex. Brain122:1533–1538

Weiskrantz L, Rao A, Hodinott-Hill I, Nobre AC, Cowey A (2003)Brain potentials associated with conscious after eVects induced byunseen stimuli in a blindsight subject. Proc Nat Acad Sci USA100:10500–10505

Weller RE, Kaas JH (1989) Parameters aVecting the loss of ganglioncells of the retina following ablations of striate cortex in primates.Vis Neurosci 3:327–342

Wilbrand H, Sänger A (1900) Die Neurologie des Auges, vol III.JF Bergmann, Wiesbaden

Wilson ME (1967) Spatial and temporal summation in impairedregions of the visual Weld. J Physiol 189:189–208

Zeki S, Vytche DH (1998) The Riddoch syndrome: insights into theneurobiology of conscious vision. Brain 121:25–45

Zihl J, von Cramon D (1985) Visual Weld recovery from scotomota inpatients with postgeniculate damage. Brain 108:335–365

123