Brain potentials associated with conscious aftereffects induced by unseen stimuli in a blindsight subject

Abstract

The study is of brain activity in a blindsight subject (D.B.), who reports conscious visual afterimages of stimuli of which he is unaware when they are presented. This contrast offered a unique opportunity to study event-related potential recordings of conscious versus unconscious visual phenomena generated by the very same stimulus in the identical locus of the visual field. The behavioral results confirmed the reliability of the difference in the subject's report for inducing stimuli versus their aftereffects. The rationale of the event-related potential analysis was to subtract “on” signals from “off” signals, the latter associated with the onset of conscious events and the former for events that remained unconscious. Because there are inherent differences in on and off potentials, the subtractive resultants for the blind hemifield were compared with the same subtractions for the good hemifield when the subject was aware both of the stimuli and their afterimages. A differential pattern in subtractive resultants emerged with a strong anterior left frontal focus for the blind field and a posterior focus for the intact field. The results are compared with other studies suggesting an anterior focus for conscious visual events.

In the absence of primary visual cortex (V1), some visual discriminations are still possible by humans and monkeys, necessarily mediated by optic pathways that bypass the geniculostriate system. In human subjects the discriminations are accompanied by altered or absent visual awareness (”blind-sight”) and typically must be uncovered by methods such as forced-choice two-alternative guessing. There are homologous effects in monkeys (1, 2). Historically, the first blindsight subject to be tested intensively over a period of >10 years was D.B. (3, 4). Recently he was reexamined after a gap of 17 years, confirming his original discriminative capacities, but a new phenomenon has emerged: He now reports conscious visual negative afterimages (including complementary colored afterimages) induced by a range of visual stimuli of which he is unaware while they are present (5). The complementary colors and contrasts of the negative afterimages match the shapes and contents of the inducing stimuli and also change their size directly with the projected distance (Emmert's law). It is not clear when this feature of his visual processing first emerged.

This conjunction, unconscious stimuli inducing conscious aftereffects, offers a unique opportunity to compare and contrast the sequential brain events generated by precisely the same stimulus in the same visual-field locus. Moreover, because D.B.'s visual-field defect is restricted to his left hemifield, the other hemifield can serve as a control in which the processing and aftereffects are both conscious.

D.B.'s right primary visual cortex was removed by surgical treatment for a nonmalignant venous tumor lodged in the right calcarine cortex (V1) at the age of 33 (see ref. 3). The tumor had probably been in situ for several years, perhaps even prenatally. At the time of testing reported here he was 62 years old. His surgical treatment involved the attachment of several aneurysm clips and other metal clips, which preclude MRI. Computed tomography scans of the occipital region are also poorly resolved because of the clips. But event-related potentials (ERPs) can still be recorded. Although their precise localization may also be subject to some distortion by metal clips, the more global patterns of anterior versus posterior activations are still informative.

The main point of interest is the comparison between unconscious activity induced by the onset and continued presence of the visual stimulus on the one hand versus the effects of the offset of the very same stimulus giving rise to a conscious afterimage on the other. We focused on the subtraction of the “onset” ERPs from the “offset” ERPs for the same stimulus in an attempt to isolate the effects uniquely associated with his visual awareness. If “on” and “off” ERP effects in the good field, both associated with conscious visual percepts (the seen stimulus and its conscious afterimage), were exactly equivalent in terms of cortical events, then subtraction of off from on should yield zero-size resultants. And, if so, for the blind hemifield the difference could be assumed directly to reflect differences between consciously perceived and unconsciously processed events. But, of course, on and off are not equivalent. Therefore the subtraction for the blind field has to be judged in relation to the same subtraction for the good field. The question is whether the greater inequality in the blind-field comparison reveals a differential pattern relative to the good field. An age-matched control subject was also studied for comparison with D.B.'s normal field and also for possible hemisphere differences related to the left versus right hemifields.

Methods

Stimuli. Black/white or red/green vertical sine-wave gratings, 13.5° wide by 16° high, were displayed on an E120 Flexscan T660 monitor such that they fell into the lower quadrant of the blind or the intact hemifield, with the nearest edges 14° from the vertical meridian and 3.6° below the horizontal meridian. The stimuli were generated by the Cambridge Research Systems VSG II/3, running on a Pentium III personal computer. The sinusoidal 3 cycles per degree achromatic grating had a mean luminance of 9 cd/m² and was equiluminant with the background. Luminance contrast was 0.5. The 3 cycles per degree red/green grating was constructed by sinusoidally varying the red and green guns with the same depth of luminance modulation but in antiphase. Chromatic contrast was therefore 0.5, but luminance did not vary across the grating. The fixation distance was 57 cm. A white opaque mask covered the edges of the monitor screen to eliminate edge artifacts. The contrast and other parameters were determined in preliminary sessions to ensure that D.B. reliably reported no experience of the stimuli but nevertheless reliably reported afterimages. The control subject always reported experience of the stimuli and a brief afterimage on approximately half of the trials. [D.B.'s blind-field afterimages typically last much longer than those in his seeing field or those of normal control subjects (5).]

Stimulus-Response Sequence. All testing took place in a sound-attenuated, electrically shielded, windowless cubicle. On each trial an experimenter presented the visual stimulus by pressing the computer's mouse; its onset was signaled to the subject by an auditory blip. The subject was provided with a response box with two response keys, with which he was instructed to respond “aware” or “unaware.” He was instructed to respond “unaware” only when he had no awareness whatever, no matter how faint or its “feeling” quality.

The subject's first response (aware or unaware of onset) turned the stimulus off after a short variable temporal gap averaging 500 ms (to separate the manual response from the off signal in the recordings). His second response indicated whether he then had an afterimage, but he was asked to withhold this response until after any afterimage disappeared completely to eliminate confusion between it and the signaled onset of the next trial.

Stimuli were delivered in blocks of 216 trials. Four blocks were conducted for red/green gratings, followed by four for achromatic gratings. In each case, three of the four blocks were for stimuli in the blind field, and one was for stimuli in the good field. Each block of trials contained ≈10% of “blank” stimuli randomly inserted. ERP recordings were made throughout.

The same procedure was followed for the control subject (A.C., aged 66) with both types of gratings, with the same stimulus parameters, and the same response alternatives, with one block for red/green and one for black/white gratings.

ERP Recording. The electroencephalogram activity was recorded continuously from 54 scalp sites by using nonpolarizable tin electrodes mounted on an elastic cap (Electro-Cap, Eaton, OH) and positioned according to the 10-20 international system (6). The montage included 8 midline sites (FPZ, FZ, FCZ, CZ, CPZ, PZ, POZ, and OZ) and 23 lateral sites over each hemisphere (FP1/FP2, AF3/AF4, AF7/AF8, F3/F4, F5/F6, F7/F8, FC1/FC2, FC3/FC4, FC5/FC6, FT7/FT8, C3/C4, C5/C6, T7/T8, CP1/CP2, CP3/CP4, CP5/CP6, TP7/TP8, P3/P4, P5/P6, P7/P8, PO3/PO4, PO7/PO8, and O1/O2). In addition a ground electrode was positioned between FPZ and FZ at the midline, and the right mastoid served as the active reference for all electrodes. Recordings obtained from the left mastoid electrode were used off line to rereference the scalp recordings to the average of the left and right mastoids. Electrode impedance was <5 kΩ. The signal was amplified 20,000 times and digitized at a sampling rate of 250 Hz. Data were recorded with a band-pass filter of 0.03-100 Hz.

Eye position, movements, and blinks were detected by recording the electrooculogram as well as by video-based remote infrared tracking. In the electrooculogram, the voltage difference between two electrodes located lateral to the left and right external canthi recorded horizontal eye movements. The voltage difference between electrodes located above and beneath the right eye recorded vertical eye movements and blinks. The infrared tracker (Senso Motoric Instruments, Teltow, Germany) was based on the principle of using the position of the corneal reflection as well as the center of the pupil to define the pupil position and had a resolution >0.1°. The standard sampling rate was 60 Hz. Eye-tracker signals were calibrated by using a nine-point calibration procedure for each subject. During the calibration procedure the subject has to fixate all calibration points one after another in a predetermined order. Trials with incorrect gaze direction, eye movements, or blinks were discarded.

Electroencephalogram data were allocated to epochs off line. Epochs started 100 ms before and ended 600 ms after stimulus presentation for onset and had a similar arrangement for offset. There was a baseline correction based on activity between -50 and 0 ms. Epochs of each kind were averaged separately, as were epochs where awareness or unawareness were signaled by the subject. Eight classes of waveform were therefore obtained: to onset, left and right hemifields and aware and unaware; to offset, left and right hemifields and aware and unaware. In practice certain classes were extremely rare. Separate averages were made for each class when the number of trials permitted. Some ERP trials were rejected because of eye blinks, eye movements, or other recording artifacts. For D.B.'s blind-field records, 640 trials were used for each average waveform for onset and 507 for offset for red/green gratings and 499 and 486, respectively, for achromatic gratings. For the good field, the respective numbers of trials were 168 and 180 for red/green and 220 and 141 for achromatic gratings. For the control subject (A.C.) there were 181 trials for onset waveform and 187 for offset for red/green in the left hemifield and 173 and 178 for the right hemifield. For the waveforms for achromatic gratings for A.C., there were 188 for onset and 191 for offset in the left hemifield and 186 and 191, respectively, for the right hemifield.

Results

Behavioral. In his blind field, in all but a small number of trials (2 of 1,154) D.B. reported no awareness of the inducing grating stimuli. In contrast, he reported seeing afterimages in the vast majority of trials, both for achromatic and chromatic gratings (1,140 of 1,154; see Table 1). He never reported awareness of blanks in his blind field but gave a small number of “false-positive” reports of afterimages to blanks (8 of 142), possibly due to persistence from the previous trial. In his good field he reported awareness of gratings in all but one trial (of 384) and also reported afterimages in all but two trials. He never reported awareness of blanks (0/48) or any afterimages after blanks. These strong contrasting patterns of results make statistical analysis redundant.

Table 1. Grand totals for red/green and achromatic gratings.

	Aware/unaware	Afterimages
Red/green*
Gratings	2/576	573/576
Blanks	0/72	3/72
Achromatic†
Gratings	0/578	567/578
Blanks	0/70	5/70

^*Spatial frequency = 3 cycles per degree; chromatic contrast = 0.5 (blocks IV + VI + VII).

^†Spatial frequency = 4 cycles per degree; luminance contrast = 0.5 (blocks XII, XIV, and XV).

ERP Results. Off and on separately. Video sequences of the temporal events from 80 to 600 ms, with a 20-ms frame width, were examined for off and on ERPs separately for each hemifield for both chromatic and achromatic gratings and also for off minus on. It was clearly evident that there was greater positive potential activity for off signals in the blind field than in the good field for both types of gratings. In the blind field, strong positivity occurred to off for both types of gratings starting initially in the right (”cortically blind”) hemisphere at ≈100 ms for achromatic gratings and 160 ms for red/green gratings (Fig. 1). This was followed by negative-going activity distributed centrally. In contrast, posterior responses in the good field to off were more variable and weaker. For both types of gratings, late positive frontal activity was seen for off in the good field. On signals (data not shown) in both the blind and good fields were also associated with posterior negative-going activity extending from ≈200 to 400 ms. The size of the potentials to visual stimuli tended to be weak, almost certainly because the stimuli used in this study were themselves relatively weak (contrast of 0.5, relatively desaturated colors, and an extramacular location) to satisfy the behavioral conditions of the experiment for the occurrence of blindsight with conscious afterimages.

Fig. 1.

Topographies of activity after the offset of the two types of grating stimuli (when D.B. reported seeing afterimages) showing early posterior positive activity starting around 100 ms for offsets of presentations of gratings in the blind hemifield. No clear pattern is seen for offsets in the good hemifield. (Upper) Red/green gratings: offset. (Lower) Black/white gratings: offset. B, blind field; G, good field.

Off-minus-on subtractions. The same pattern of greater activity to off in the blind field was also preserved in the off-minus-on subtractions. In the blind field strong positivity occurred ≈140 ms after off for achromatic gratings and ≈180 ms for red/green gratings in the right (cortically blind) hemisphere, later switching to the left hemisphere. In contrast, the same subtraction in the good hemifield revealed a slightly later and smaller positive posterior difference (200 ms) for red/green and somewhat later negative activation for black/white (300 ms). However, the good field, especially for red/green gratings, showed a strong early left frontal resultant positive activation not evident in the blind field. These effects that were evident in the video sequences were also confirmed in the topographic brain maps. In the control subject, the off responses for the two hemifields were less consistent regarding the locus of the early positive responses, and there were inconsistencies in the signs of the early potentials to on.

The spatial distribution of off-minus-on subtractions was examined by comparing their relative sizes for blind- vs. good-field stimuli at each electrode location, irrespective of latency, within a range of 100-400 ms. The mean amplitudes of the resultant subtractions between 100 and 400 ms were computed for both types of gratings (Table 2). The relative sizes of these values for the blind and good hemifields were compared for each electrode position to measure whether the off-minus-on resultant for the blind field was greater or smaller than that of the good field. That is, a double subtraction was carried out: (off minus on, blind field) minus (off minus on, good field).

Table 2. Mean amplitudes (100-400 ms) of subtractions: Offset minus onset.

Red/green gratings	Blind field	Good field	Black/white gratings	Blind field	Good field
FPZ	−5.05	2.42	FPZ	−3.64	−1.33
FZ	−4.68	0.02	FZ	−3.85	−2.76
FCZ	−4.17	−1.18	FCZ	−3.89	−3.16
CZ	−5.20	−2.91	CZ	−4.25	−4.47
CPZ	−3.45	−2.47	CPZ	−3.27	−4.30
PZ	−1.68	−1.45	PZ	−1.55	−2.70
POZ	0.56	−0.45	POZ	−0.14	−1.52
OZ	2.50	1.70	OZ	2.44	0.23
FP1	−3.85	1.01	FP1	−3.34	−1.00
FP2	−3.01	1.77	FP2	−3.11	−0.97
AF3	−3.62	2.05	AF3	−3.24	−0.70
AF4	−1.67	0.13	AF4	−3.14	−1.60
AF7	−3.70	0.97	AF7	−2.79	−0.37
AF8	−2.71	0.69	AF8	−2.94	−1.66
F1	−3.86	0.37	F1	−3.25	−1.38
F2	−3.48	−0.20	F2	−3.27	−2.10
F3	−3.74	0.85	F3	−2.82	−0.76
F4	−3.06	−0.28	F4	−2.77	−2.34
F5	−3.59	1.04	F5	−2.78	−0.65
F6	−3.15	−0.18	F6	−2.71	−2.22
F7	−3.59	1.02	F7	−2.58	−0.51
F8	−1.84	0.50	F8	−2.47	−1.96
FC1	−4.00	−0.76	FC1	−3.39	−2.54
FC2	−3.60	−1.40	FC2	−3.36	−3.70
FC3	−3.82	0.88	FC3	−2.86	−1.32
FC4	−3.31	−1.67	FC4	−3.35	−4.57
FC5	−3.64	−0.53	FC5	−2.68	−1.17
FC6	−2.30	−0.83	FC6	−2.23	−3.03
FT7	−5.06	1.18	FT7	−3.27	−0.78
FT8	0.49	0.12	FT8	−2.41	−1.58
C1	−3.62	−1.48	C1	−2.94	−2.72
C2	−2.86	−1.57	C2	−3.03	−3.50
C3	−3.43	−0.88	C3	−2.74	−2.33
C4	−2.41	−2.36	C4	−2.50	−3.80
C5	−2.72	0.31	C5	−1.92	−1.01
C6	−1.00	−0.90	C6	−1.51	−3.26
T7	−1.08	0.70	T7	−1.22	−0.04
T8	−0.56	0.07	T8	−0.58	−1.90
CP1	−2.72	−1.55	CP1	−2.04	−2.58
CP2	−1.86	−2.03	CP2	−2.13	−3.28
CP3	−2.15	−0.99	CP3	−1.33	−1.79
CP4	−1.25	−1.54	CP4	−1.50	−3.67
CP5	−1.90	−0.23	CP5	−1.33	−1.20
CP6	−0.60	−0.56	CP6	−0.30	−3.06
TP7	−1.18	0.55	TP7	−0.79	−0.31
TP8	−0.55	−0.46	TP8	−0.17	−2.72
P1	−1.36	−0.81	P1	−1.05	−1.63
P2	−0.67	−1.03	P2	−1.08	−1.91
P3	−1.37	−0.60	P3	−0.50	−0.95
P4	−0.07	−1.01	P4	−0.87	−2.71
P5	−1.07	0.09	P5	−0.42	−0.91
P6	−0.02	−1.24	P6	0.07	−3.09
P7	−0.77	0.59	P7	−0.41	−0.92
P8	0.11	−1.62	P8	0.00	−3.07
PO3	−0.88	−0.09	PO3	−0.14	−0.86
PO4	0.06	−0.71	PO4	−0.06	−1.68
PO7	−0.15	0.92	PO7	−0.15	−0.85
PO8	0.00	−0.63	PO8	0.29	−1.91
O1	0.91	1.10	O1	1.21	−0.21
O2	0.84	0.16	O2	1.26	−0.42

A general pattern emerged for both chromatic and achromatic gratings, namely that blind-hemifield off-minus-on resultants (over the 100- to 400-ms range) were greater than good-hemifield resultants for electrode placements in frontal locations, and in contrast, the resultants were greater for good than blind hemifields in posterior locations. The strongest frontal focus was in the anterior left hemisphere. The posterior foci tended to be mainly in the right hemisphere. Examples of representative subtypes are shown for selected anterior and posterior electrode positions (Figs. 2 and 3) for D.B. as well as the control subject.

Fig. 2.

Representative samples of resultants of subtractions of off-minus-on potentials for anterior and posterior loci for the patient D.B. and the control subject. For D.B., absolute values of blind-field resultants are larger than good-field resultants in the blind half-field for frontal loci and the opposite for posterior loci. For the control subject, right-field resultants tend to be more independent of locus. Shown are red/green gratings. (Inset) Electrode positions are marked in solid black.

Fig. 3.

Data are as described for Fig. 2 but for black/white gratings.

Considering scores for all 25 electrode placements anterior to the central row (C1, 2, etc.) for both gratings combined, 46 of the 50 possibilities (25 × 2 gratings) had blind-field resultants greater than good-field resultants (yielding double-subtraction resultants with negative signs). Conversely, only 8 of 52 (26 × 2) placements posterior to the central row had greater blind-field than good-field resultants, yielding outcomes with positive signs. Such a distribution is extremely unlikely by chance (P < 0.001 by χ² analysis). When the size of the resultants is taken into account, there is a clear bias for greater blind-field resultants to be located most anteriorly (mean of seven most anterior loci = -3.58 for red/green gratings and -1.57 for black/white gratings) and a bias for greater good-field resultants to be located most posteriorly (mean = +0.71 for red/green gratings and +2.19 for black/white gratings). There is no overlap between the sets of individual values of the most anterior and most posterior loci.

A similar outcome is obtained when the peak values instead of mean values were considered (range of 100-400 ms). For both achromatic and chromatic gratings, the largest negative values are for anterior electrode positions, and the largest positive values are for posterior positions. (The 10 largest negative values for red/green double subtractions are FPZ, AF3, FP2, FT7, FC3, AF7, F3, FZ, F5, and F7, mean value = -5.49; the 10 largest positive values for red/green double subtractions are P8, POZ, P6, P4, PO4, PO8, PZ, O2, and CP2, mean = +1.72; the 10 largest negative values for black/white gratings are AF3, FP1, FPZ, F3, F5, AF7, F1, FP2, F7, and FT7, mean = -0.85; and the 10 largest positive values for black/white gratings are P6, P8, CP6, TP8, OZ, PO8, C6, CP4, POZ, and T8, mean = +3.57.)

The topographic distribution shown in Fig. 4 illustrates the contrasting pattern of opposite signs of the posterior and anterior loci arising between 300 and 400 ms for both types of gratings. The topography also helps to identify the temporal properties of the differential distribution. For both types of gratings, a clear pattern of the two principal foci emerges in the left anterior frontal and the right posterior regions at ≈260 ms for red/green gratings and 300 ms for black/white gratings. The signs of the foci are opposite, as expected, given that the direction of the subtractive differences are reversed.

Fig. 4.

Topographies of double-subtraction resultants. For both types of gratings a right posterior focus and a left anterior focus emerge after 200-300 ms of opposite sign. The charts at the far right of each row are summaries of activity between 100 and 400 ms. (Upper) Red/green gratings: blind field (offset-onset) - good field (offset-onset). (Lower) Black/white gratings: blind field (offset-onset) - good field (offset-onset).

We also examined the video sequences and brain maps for the pattern of results for good field (on) minus blind field (on) and good field (off) minus blind field (off), separately for black/white and red/green gratings. The results of this subtraction showed strong left frontal activation for off for both types of gratings, which is consistent with the frontal foci seen in the off-minus-on subtraction for the blind field versus the good field, but the latency was a bit later for red/green gratings (≈200 ms) than for black/white gratings (140 ms). The expected early increased posterior activation for off was not seen and must have been masked by the subtraction. The pattern for on was variable for black/white, but central posterior activation was seen for red/green emerging at ≈200 ms.

Control Comparison. It is logically possible that the results had nothing to do with a blind versus an intact half-field but stemmed from a cerebral hemispheric difference, given that each half-field projects uniquely to its contralateral hemisphere, coupled with the possibility of cerebral dominance. Perhaps the frontal vs. posterior clusters of relative off-minus-on differences could emerge from hemispheric asymmetry. The control subject showed consistently stronger potentials for stimuli to his left hemifield than to his right. However, and crucially, this difference appeared for electrodes independently of their anterior-posterior locus both for achromatic and chromatic gratings. There was not a single electrode placement in which this pattern was reversed.

Discussion

Given that the good-field subtraction reflects conscious-minus-conscious and the blind-field subtraction conscious-minus-unconscious visual perception, the question is whether there is a differential pattern in the blind field relative to the good field. When that decisive comparison is made between the subtractions in each field, a clear outcome emerges: At frontal and other anterior loci, the resultant subtractions for the blind field, irrespective of latency, are greater than those for good field, and at posterior loci good-field subtraction resultants are greater than those for the blind field. The differential pattern is not attributable to a hemisphere difference per se, because no such pattern emerged in a control subject.

D.B.'s metal surgical clips in the occipital lobe make precise and detailed resolution of posterior activation uncertain, but they do not mask the striking difference in findings between frontal and posterior loci for differential stimulation of the blind and good-field results. Right frontal activation has been implicated in a functional MRI study of another blindsight subject (G.Y.) with left hemisphere damage to V1 in comparing stimuli of which the subject was aware versus those for which he was unaware in his blind hemifield (7). D.B.'s pathology was in his right hemisphere, which might account for his predominantly left frontal focus. Frontal activation has also been implicated in similar comparisons of conscious and unconscious events in unilateral neglect and in change blindness (8). The results support the view that visual awareness entails and requires a further stage of processing beyond the posterior input stage (8-10). The subtractive bias toward posterior loci for good-field presentation might be expected if the on signals are greater than the off signals for the same stimulus, and indeed this was strongly confirmed in the results of the age-matched normal control subject for both types of gratings over the entire distribution of electrodes. It also could account for the predominance of negative-going subtractive resultants for the majority of off-minus-on scores.

The present investigation only considers size of subtractive resultants independent of size, latency, cross-correlations, or rhythmic components. The stress is on the differential distribution of subtractive resultants. The evidence from off signals alone, immersed in a complex matrix, suggests that the blind field shows stronger posterior activity and that the good field shows stronger frontal activity for off. However, although the various subtractions are consistent with this, just what the chain of events is that leads to the resultant frontal pattern, of course, remains to be discovered.

The striking overall difference in the distribution of subtractive off-minus-on resultants occurs notwithstanding the limitation on resolution stemming from spatial distortion of field potentials from metallic clips. To date we have not seen any other blindsight subjects who report conscious afterimages of the unseen inducing stimuli, and this may reflect a history of migraine in D.B. There are indications that brain potentials are enhanced and are slower to habituate in migraineurs (11-13), and interestingly, migraineurs show increased visual adaptation effects of uniquely cortical phenomena such as motion and tilt aftereffects (14). If other subjects are identified for whom functional MRI is suitable, they would offer a unique opportunity for finer detailed analysis of the contrast between consciously aware and unaware percepts to be related to the present ERP patterns.