The Continuum of Detection and Awareness of Visual Stimuli Within the Blindfield: From Blindsight to the Sighted-Sight

Abstract

Purpose.: We investigated systematically the effect of repeated exposure on detection and reported awareness of visual stimuli presented deep within the field defect of 5 hemianopic patients.

Methods.: An objective measure of sensitivity (detection in a temporal two-alternative forced-choice paradigm) and subjective reports of awareness were recorded on trial by trial bases. Visual stimulus to be detected was a temporally modulated (10 Hz) circular patch (6° diameter) of vertical grating (1 c/°). Hemianopic patients took part in the study 8 to 15 months after injury, so that the findings could not be attributed to spontaneous recovery.

Results.: Initially, high contrast (90%) target stimuli were detected at or near chance level with little reported awareness. In 4 of 5 cases, repeated stimulation led to improved sensitivity, indicated by increased detection scores and higher incidence of awareness. In a fifth case, there was no change in sensitivity despite extensive exposure (>22,000 trials).

Conclusions.: At retinal locations deep within the field defect, repeated stimulation can lead to blindsight performance (type I detection without awareness), followed by detection with reported awareness (type II blindsight), and eventual reported visual experiences. The findings indicate that conscious awareness of stimuli lies on a continuous spectrum and repeated systematic training can lead to improved visual sensitivity.

Introduction

A century ago, a handful of neurologists were investigating the visual effects of traumatic brain injury, mapping out the visual deficits and trying to get into grips with relationships between the site of the lesion after a gunshot wound, extent of the resultant blindness, and effects it had on veterans of wars. The work of Tatsuji Inouye on veterans of Russo-Japan war, ¹ and those of Holmes, ² Poppelreuter, ³ and Riddoch ⁴ during the First World War documented the topographic relationship between the extent of occipital brain injury, resultant visual deficits, and residual visual capacities maintained within so-called “blind” areas. Although Poppelreuter suggested some rudimentary visual function remained, ³ and Riddoch reported the persistence of motion sensitivity within the blind field in a few cases, ⁴ the general consensus was that visual deficits in the chronic phase were absolute and stable over time. ⁵ The subsequent 100 years witnessed a wide range of research on nonhuman animals with lesions or ablations of visual pathways, for example. ⁶ The ensuing extensive body of literature by and large suggested that the deficits in visual capacities were transient and the animals tended to recover postinjury. ⁷ These observations were not in line with experiences in human subjects. Of note are the studies on monkey Helen, with bilateral ablation of the striate cortex, in whom Humphrey and Weiskrantz reported only some minor localization impairment. ⁸  

Bridging the gap between the nonhuman and human observations was the use of forced-choice guessing paradigms in patients with blindness as a result of occipital lesions. ⁹ The technique replaced verbal reports with a paradigm more akin to those in animal research, and the findings revealed a wide range of residual visual capacities even when patients reported themselves to be purely guessing, with no acknowledged awareness of visual events. ¹⁰ This capacity is termed blindsight type I, to be distinguished from type II blindsight when some rudimentary awareness of visual events may be reported, even though the patient denies seeing per se. ^11,12  

In the last four decades, numerous studies have documented the range of visual capacities remaining in blindsight patients, including above-chance discrimination of spatial patterns, form, motion, and color; evidence of implicit processing of emotional faces and bodies; and the ability to reach, grasp, and make eye movements to targets, to name just a few (see the report of Danckert and Rosetti ¹³ for a review). Therefore, there is no doubt that information presented within the field defect is extensively processed in many cortical and subcortical structures. ^14,15  

These findings raise a number of fundamental questions as regards the possibility of rehabilitation of deficits; namely, how is processed information made available to subjective commentary? What are the prerequisites for being consciously aware of the outcome of processing of visual information presented within the field defect? ¹⁶ More fundamentally, can neural processing be modified such that stimuli falling within a chronically blind area of the visual field lead to correct discrimination and even awareness of visual events? ¹⁷  

Common approaches to rehabilitation of visual deficit after brain injury include using optical aids ¹⁸ or training patients to improve their visual search efficiency. ¹⁹ Although a significant number of patients find the eye movement training useful for detecting/avoiding obstacles, neither of the approaches could be claimed to change the sensitivity within the blind area. ²⁰  

Reading difficulties following occipital brain injury (termed hemianopic dyslexia) also are a common problem in stroke survivors. ^21,22 Following brain injury, reading problems manifest in slow reading speed, omission of words, and disorganized oculomotor reading patterns. ^23,24 Rehabilitation training approaches specific to reading also have been developed and have shown effectiveness in improving reading ability following training. ^23–25 In a recent clinical investigation, it has been demonstrated that methods aimed at rehabilitation of visual exploration and those addressing reading are highly specific interventions, each affecting the intended function only and, therefore, there is no between intervention transfer effect. ²⁶  

Restitution techniques aim to change blindfield sensitivity through repeated stimulation over extended periods of time. Four different stimuli are used in the most common approaches. These include detection of small light targets, ²⁷ flickering letters, ²⁸ coherent movement in random dot patterns, ²⁹ and temporally modulated spatial gratings. ³⁰ It is likely, and indeed stipulated, that there are multiple ways in which neuronal plasticity can lead to remediation, although direct evidence for any one of these mechanisms is sparse. For example, placement of small light targets at blind/sighted borders, as used in Vision Restoration Therapy (VRT), may lead to changes in local neuronal interactions, ³¹ whereas stimulation of areas deep within the blind field may lead to processing in multiple cortical and subcortical regions, ³² or even increased activity in areas within the contralesioned hemisphere. ³³  

Investigation of detection ability within the blind region often has involved patients participating in forced-choice paradigms in the laboratory environment, with performance at levels above statistically-derived chance levels, indicating the presence of residual capacities. An interesting question is whether sensitivity may change over time with repeated stimulation, such that patients performing at chance level in a detection task would perform significantly above chance after prolonged, repeated exposure. We report four cases of hemianopia after ischemic occipital brain injury and show how the detection ability changes with extensive stimulation. In addition, the subjective reports of awareness of visual events appear to mirror those of detection performance. Inclusion of negative findings in a fifth case despite extensive training indicates that caution is needed in generalizing the findings to all hemianopic cases. 

Methods

Patients

Five patients participated in this study. A brief summary of their case history is given below. 

AM, a 56-year-old female patient, started the study 15 months after stroke. The radiologic report based on her magnetic resonance imaging (MRI) scans taken 8 months after stroke showed an old infarct in the left occipital lobe involving the occipital pole, and the cortex medially and laterally. There also was an incidental finding of an extra-axial soft tissue mass in the left side of the posterior fossa, abutting the posterior aspect of the left petrous temporal bone, inferior to the level of the internal auditory meatus. This was roughly spherical in shape, measuring a maximum of 2.1 cm in diameter, and was indenting the anterior aspect of the left cerebellar hemisphere. The appearances were typical of an incidental meningioma. In addition, there was cortical atrophy, which was most obvious in the right parietal lobe. 

SR, a 55-year-old female patient, was tested 9 months after stroke. The radiologic report based on her MRI scans showed the existence of gliosis extending from the occipital horn of the right lateral ventricle to the primary visual cortex, typical of old infarction. There were no other intracranial abnormalities and the thalamus was not involved. 

EB is a 54-year-old female patient, tested 11 months after stroke. The radiologic report based on her MRI scans showed evidence of an old infarct in the left occipital lobe involving the occipital pole and underlying white matter, but with some sparing of the primary visual cortex. There was no involvement of the lateral geniculate nucleus (LGN). However, there was evidence of more widespread cerebrovascular disease with small infarcts in the right internal capsule, and corona radiata, right thalamus, and pons. 

LD (female) suffered a stroke at the age of 36. An MRI scan revealed evidence of high signal following the sulci and gyri of the left occipital lobe extending to the medial aspect of the left temporal lobe. There was no evidence of significant hemorrhaging. As a result of stroke, LD had an upper right quadrantopia. LD first started the training sessions 8 months after stroke, and the training sessions were spread over a 21-month period. 

DE is a 40-year-old male with an upper left quadrantopia who started training 8 months after stroke. The damage visible on MRI scans extended anteriorly from the medial portion of the right occipital lobe and the right medial temporal lobe to the thalamus. DE also experienced some loss of sensation and weakness of the left arm and hand in the acute phase, which recovered in subsequent months. 

All testing was done in accordance with the ethical guidelines of the School of Psychology, Aberdeen University and Grampian Research Ethics Committee, in accordance with the 1964 Declaration of Helsinki. 

Visual Field Testing

The extent of visual field deficits were determined using program 30-2 of the Humphrey Visual Field Analyser (HVFA; Carl Zeiss, Cambridge, UK) twice before the study and twice after the completion of training sessions. The average sensitivity values are plotted in Figure 1 together with an MRI showing the extent of the lesion. A schematic binocular visual field also is plotted, with the corresponding detection and reported awareness for each target location plotted in Figure 2. All five cases had excellent fixation stability (all fixation losses ranged from 0/22 to 3/20, false-positive errors ranged from 0/19 to 3%, false-negative error ranged from 0/16 to 1/11). They all were able to concentrate for the duration of the 30-minute daily training. 

Figure 1. 

 

MRI scans and visual fields are shown for all patients. The monocular visual fields were obtained using program 30-2 HVFA, before and after the training. A schematic representation of binocular visual field also is shown for each patient, and the light disc within the dark area denotes the location of the training targets.

Figure 2

 

The detection score (solid lines), reported awareness (grey lines), and the stimulus contrasts (dashed line) are plotted for each training session for each patient. The chance level performance for detection in a two-alternative, forced-choice paradigm is 50%. With increasing performance, the target contrasts were lowered. There is no improved detection in DE, even after 448 training sessions.

Apparatus

The training task was conducted at patients' homes on an IBM compatible personal computer. Gamma corrections were conducted on all monitors using an LMT luminance meter (LS100; Konica Minolta, Inc., Tokyo, Japan) at 256 equi-stepped logical colors. The task consisted of detection of a temporally modulated spatial grating patch using a temporal two-alternative forced-choice (2AFC) task. That is, the task was to report whether a target stimulus was presented during the first or second of two intervals. The stimuli were circular patches of vertically-oriented sinewave gratings (1 c/°, 10 Hz, 6° diameter) and were presented at 3 predetermined retinal eccentricities in a randomly interleaved order. The exact location was tailored to each individual patient's deficit (see Fig. 1). In this report, we have only included the detection and reported awareness for target stimuli placed deepest within the field defect, because there would be no effect of occasional small eye movements on the visibility of stimuli placed well away from the border of the scotoma. The stimulus interval was reported by choosing one of the two buttons on a response box. In addition, after each trial, the patient was required to report whether s/he was aware of any stimulus presentation by pressing one of two additional buttons on a response box. The patient was required to report “no awareness” when there was no awareness or feeling of the stimulus being presented, or to report “aware” otherwise. The locations of other stimuli were chosen so that at least one location straddled the sighted field and a second target appeared close to sighted/blind field boundary. There were 50 trials at each of the three locations (total 150 trials) in each training session. To manipulate task difficulty, the target contrast was modulated algorithmically such that if the detection performance at any given location was at or above 84% for 3 consecutive sessions, the stimulus contrast was reduced by 10%. Reduction in performance to 64% and below resulted in an increase of 5% in contrast in the next training session at that location. Therefore, the algorithm for determining the subsequent target contrasts was biased to provide a conservative estimate of contrast detection. That is, in the absence of improvements in detection performance, and when the detection ability is noisy, it is more likely for contrast levels to remain high than for them to be reduced. No feedback was provided to AM, SR, EB, and LD throughout the training sessions, and the patients were debriefed on their performance only after the training had finished. Therefore, they were totally unaware of their performance scores for the duration of the experiment. DE was provided positive feedback after each correct trial after an earlier study had indicated that the provision of feedback can accelerate the recovery rate. ³² Presentation of trials for all patients was self-paced. In addition, a minimum of 2 minutes' rest was imposed after 75 trials to avoid fatigue. 

AM, SR, and EB conducted 62, 69, and 96 training sessions over a 3-month period, whereas LD and DE performed 203 and 448 sessions over a 21-month and 3-year period, respectively. 

Results

Figure 2 depicts the measured parameters during each training session, including detection performance (solid dark lines), reported awareness (solid grey lines), and the target contrast (broken dark lines) for each patient. 

Overall, it appears that at the early stages, the detection performance is at or near chance level (50%) despite the presentation of high contrast targets (initial contrast 90%). The detection performance and reported awareness changed over multiple training sessions. There was a total of 3100, 3450, 4800, 10,150, and 22,400 trials for AM, SR, EB, LD, and DE, respectively. To quantify the progression over this period, the average of detection scores, and reported awareness responses over the first and the last 5 sessions were calculated. 

Patient AM

The correct detection for 90% contrast target, averaged over the initial 5 sessions, was 66.4% (SD = 8.3), which was significantly above chance (t = 4.42, df = 4, P = 0.012). This improved to 82.4% (SD = 3.85) at average contrast of 38%, which again was significantly above chance (t = 18.83, df = 4, P < 0.001). The percentage of trials when AM reported being aware of a target presentation also changed from 5.6% (SD = 3.29) to 35.6% (SD = 11.9), which was significantly higher (t = 4.72, df = 4, P < 0.01) despite lowering of target contrast. 

Patient SR

The correct detection for 90% contrast target was initially 68.4% (SD = 6.69), which was significantly above chance (t = 6.147, df = 4, P = 0.004). This improved to 98.8% (SD = 1.10) at average contrast of 5% that was significantly above chance (t = 99.613, df = 4, P < 0.001). The percentage of trials when SR reported being aware of a target presentation were initially 13.6% (SD = 7.92) and reached 19.6% (SD = 5.55), which were not significantly different from each other (t = 2.176, df = 4, P = 0.095) despite an 85% drop in stimulus contrast. 

Patient EB

The correct detection for a 90% contrast target initially was at chance level 50% (SD = 0), and improved to 84.8% (SD = 4.60) at average contrast of 80% that was significantly above chance (t = 16.9, df = 4, P < 0.001). EB did not report being aware of any of trials during initial and final training sessions. 

Patient LD

The correct detection for 90% contrast target initially was at chance 49.6% (SD = 4.98, t = −0.18, df = 4, t = 0.866), and improved to 80.4% (SD = 5.55) at average contrast of 20% that was significantly above chance (t = 12.249, df = 4, P < 0.001). The percentage of trials when LD reported being aware of a target presentation also changed from 0% (SD = 0) to 98% (SD = 2.45, t = 89.461, df = 4, P < 0.001) despite lowering of target contrast by 70%. 

Patient DE

The correct detection for 90% contrast target initially was 52% (SD = 8.71), which was not significantly above chance (t = 0.513, df = 4, P = 0.635). At the final 5 training sessions he performed at an average of 48.8% (SD = 9.23) correct, which again was at chance level (t = −0.291, df = 4, P = 0.786). The target contrast also remained at 90% throughout. There were no incidences of reported awareness for any of the trials, throughout the training sessions. 

Figure 1 shows visual field sensitivity as measured using the HVFA (Carl Zeiss). This measure does not show any change in sensitivity except for the case SR. Nevertheless, there may be changes in adjacent locations, as some also were stimulated by additional targets during training. To compare the overall visual field sensitivity before and after the training, the total sum of sensitivity (in decibels) from each eye was calculated and then averaged for both eyes for the affected hemifield. The results for before and after training showed no significant change in overall visual sensitivity as measured using HVFA (Carl Zeiss) (before M = 308.15, SD = 187.17; after M = 322.35, SD = 216.27; t = −0.725, df = 4, P = −0.508). Although HVFA (Carl Zeiss) data did not show a significant change, the visual sensitivity also may be measured in terms of the target contrasts at the start and end of the training. As a group the average stimulus contrast lowered from the initial 90% to 46.6% (SD = 37), which was a significant reduction (t = 2.614, df = 4, one-tail P < 0.03). 

To investigate whether the improved psychophysically determined detection performance is specific to the training area, or if it is a more global improvement, the detection of a 1 c/°, 10° diameter Gabor patch at a nontrained (control) location was measured in three of the patients (AM, SR, and EB) before and after training. The control location coordinates were AM (15,9), SR (−12,8), and EB (15,9). Training did not affect performance at this control location (AM, before 72%, after 76%; EB, before 62%, after 42%; SR, before 72%, after 76%; confidence interval 13.9% based on binomial distribution, chance level 50%). 

We also have measured the subjective reports of awareness as a function of stimulus contrast before and after daily training, using Gabor stimuli (10° diameter, SD of spatial Gaussian 2.5°, duration 2 seconds, at SD of temporal Gaussian 250 ms) in four cases who have shown improvements following training. In these measurements, fixation stability was ensured in every trial using an infrared pupillometer (ASL5000; Applied Science Laboratories). The apparatus allowed for shifts in fixation larger than 1.5° to be identified reliably. Data plotted in Figure 3 show higher incidence of reported awareness in 3 cases in the absence of any detectable eye movements, indicating that improvements were not likely to be due to increased fixation instability. SR, in contrast, showed equally high sensitivity before and after training. As the stimulus size used for laboratory testing was larger (10°) than those used in daily training (6°), for the case of SR and only for this patient, it may have been possible that the stimuli not always were confined to the blind areas. Small eye movements may have shifted the targets into areas adjacent to the scotoma that had reduced sensitivity relative to the intact field, but were not completely blind to the 10° target even before training. This would explain the relatively good contrast responses plotted in Figure 2 for SR. 

Figure 3

 

Percentage of trials when patients reported awareness of the presentation of a Gabor patch is plotted as a function of Gabor contrast before (dark squares) and after (grey squares) completion of the daily training program. For all measurements, fixation stability was monitored and ensured using an infrared pupillometer (ASL5000; Applied Science Laboratories, Bedford, MA).

Discussion

Detection performance in a temporal two-alternative forced choice paradigm improved in 4 of 5 cases reported (AM, SR, EB, and LD). At the early stages, performance either was low or at chance level when high contrast targets were presented. With repeated stimulation, detection improved until it was consistently close to ceiling. Subsequently, the target contrast was reduced automatically by the training program. This continued for a large number of trials, until patients were performing consistently at well above chance with low-contrast targets. Overall, the incidence of reported awareness also increased with repeated stimulation and above-chance detection, except for SR who was minimally aware of targets at very low contrast (5%). The pattern of findings indicates improved detection ability and awareness of stimuli with repeated and systematic training and exposure. We would like to expand on the implications of the findings in relation to subjective and objective measures of sensitivity, putative mechanisms for recovery, and prospects for recovery of function after occipital brain injury. 

Subjective reports of visual experiences usually are related to the physical properties of stimuli. At the limits of perception for healthy observers and clinical populations with damaged visual pathways, reports of the subjective experience are affected by the subjective bias. Some observers (or even the same observer, but in different testing sessions), may not be consistent in their reports of visual events. ³⁴ This often is referred to as a shifting subjective criterion. For this reason, criterion-free measures provide an important objective measure of sensitivity. We have collected subjective and objective measures of sensitivity by asking our patients to report their subjective awareness of visual events as well as obtaining their detection scores in a temporal two-alternative forced-choice paradigm. Above-chance performance in the absence of any reported subjective experience, such as the case in patient EB, is referred to as blindsight type I. In 3 of 5 cases reported here, there also is a clear link between detection and reported awareness (see Fig. 2) for stimuli presented within the perimetrically blind field. In these cases, patients reported some awareness of stimuli presented within the field defect, even initially, when they did not report seeing the patterns (blindsight type II). A divergence of detection performance and reported experience also is evident in the perimetric data, which showed little change following training. This dissociation is worthy of further exploration. 

Perimetric evaluation of visual sensitivity typically is conducted using subjective methods. Patients are asked to indicate the onset of a briefly presented small light target by pressing on a response box. Targets are presented in various field locations, commonly within the central 30°, while the patient fixates on a small target. The target size used in an HVFA (Carl Zeiss), often referred to as gold standard in clinical practice, is a small disc (Goldmann size III, 0.3° diameter). Perimetry is a conservative estimate of visual processing, since it cannot tap into residual visual capacities and it relies on subjective yes/no technique. In addition, the minimum target size needed for detection is variable across patients after brain injury, and depends on the extent and location of the lesion. Some patients may have little problem in detecting targets of few degrees wide, while they are at chance for detecting smaller targets. ³⁵ Some patients also report being able to detect moving targets better than the stationary ones. This is not surprising as the preservation of and improvements in motion detection within the blind field, ³⁶ together with its neuronal substrates, ³⁷ have been well documented. Subjectively, except for DE, all patients participating in our study reported improved ability to detect moving objects within their field defect following training. These changes, although useful in aiding navigational skills, cannot be measured using standard perimetry. Questionnaires designed to tap into evaluation of activities of daily living have been used to assess such changes, ³⁸ but again they have the limitation of being solely based on subjective responses. It is likely, but has not been established clearly to our knowledge, that objective improvements in detection are associated with improvements in daily life, even in the absence of any change in subjective experience. Therefore, there is a need for the development of criterion-free, objective assessments of functional vision. Such instruments then could be used to assess the efficacy of interventions, as well as to quantify abilities that currently are inaccessible via subjective reports. 

The extent of visual field loss in acute stage of brain injury may not be stable. Indeed, some hemianopic patients may recover fully within a few days after injury, with the improvements often attributed to a reduction in brain inflammation. The probability of this recovery, termed “spontaneous recovery,” diminishes rapidly with time and, although it remains a possibility, becomes rare after a few months postinjury. ³⁹ Spontaneous recovery can be identified through improvements in visual field plots between two time intervals. Patients who participated in our study were tested between 8 and 15 months after injury, and the stability of fields plotted in Figure 1 shows that there was no evidence of spontaneous recovery. 

The case of DE, who shows no improvement despite extensive training, is of great interest. Recently, Schmid et al. reported that blindsight performance in macaque monkeys relies on intact LGN, as temporary inactivation of LGN in monkeys with blindsight after chronic V1 lesion led to chance level performance on visual performance measures. ⁴⁰ Structural MRI data on DE also shows that his lesion extends anteriorly to the thalamus and probably affects the Pulvinar, and possibly LGN. If LGN is, indeed, damaged, DE would be the first human case to our knowledge that could corroborate the findings in nonhuman animal research. Detailed analysis of fiber tracks in DE will be needed to establish the lesion boundaries. 

In conclusion, the findings showed that with repeated systematic stimulation, patients initially performing at chance level in a detection task may progress to performance levels significantly above chance without any awareness (blindsight type I), followed by some limited awareness of visual events (blindsight type II), and finally to report seeing visual objects (conscious vision). The findings supported the view of a continuum of conscious awareness and that systematic exposure can lead to improved detection within the blind field. 

Acknowledgments

The authors thank James Urquhart for technical support. 

Supported by grants from the Chief Scientists Office, Scottish Government (CZB/4/30), a BBSRC award Grant BB/H01280X/1 (ARH), and a McDonnell Foundation award (LW). 

Disclosure: A. Sahraie, NovaVision, Inc. (C), Sight Science, Ltd. (C), P; C.T. Trevethan, None; M.-J. MacLeod, Sight Science, Ltd. (C); L. Weiskrantz, None; A.R. Hunt, None