What does the blindsight phenomenon imply with respect to the relationship between conscious awareness and perceptual processing?

Cited by (0)

Decision Criterion

Blindsight can be demonstrated in the form of a dissociation between visual performance in two different paradigms/tasks, namely clinical perimetry and FC tasks. In humans, the apparent discrepancy between an area of clinically blind visual field and the ability to make some form of visual discrimination was only revealed by the implementation of ‘animal type’ FC methodologies. The basis of clinical perimetry is a ‘yes-no’ (yn) task in which one of two possible stimuli (target or blank) is presented on each trial and the participant’s task is to judge which one was presented. This task allows the participant the freedom to say ‘blank’ or ‘no stimulus’ on every trial presentation if they wish to do so. Consistently replying with ‘no’ is what one would expect if they are subjectively unaware of visual stimuli throughout the task. In an m Alternative FC task (mAFC), each of m different stimuli is presented at every trial and the participant has to judge which of m intervals contained a specified stimulus, either in a temporal or spatial (i.e., localization) interval. This paradigm ensures that the participant is effectively forced to make a judgment, for example, between two temporal intervals, as the option to judge ‘no stimulus’ simply does not exist. This facilitates the revelation of above-chance detection or discrimination performance where it exists, in the absence of subjective awareness (generally required for yn task). Blindsight is characterized by the dissociation in response between the two paradigms, for example, performance levels of c. 0% detection in yn paradigm compared to >90% correct discrimination in the FC task (despite the denial of subjective awareness throughout testing). One of the criticisms of blindsight theory is that the apparent dissociations between yn responding in perimetry and FC performance could result from the use of different decision/response criteria in the two tasks. The implication of this assertion is that the distinction between blindsight and normal, near-threshold vision would not be clear, which implies that the study of blindsight would not add anything to the understanding of mechanisms of visual awareness that could not be drawn from the investigation of normal participants operating at the lower limits of their vision.

According to signal detection theory (SDT) the judgment of a participant in a detection or discrimination task depends not only on his/her sensitivity (d′) but also on his/her response criterion/bias (the tendency to select one or other of the stimuli, irrespective of sensitivity). In SDT, sensitivity is calculated independently of response bias and vice versa. However, in the majority of blindsight research percent correct is used to denote performance which, it can be argued, represents performance accurately only in the absence of response bias. In contrast, 2AFC tasks are criterion free as any bias reflects a bias to one or other interval rather than to one or other stimulus. This raises the potential issue of whether the dissociation reported in blindsight between perimetry and FC tasks is due to a difference in response criteria between the two tasks. This has been tested directly in GY during yn and FC detection of static and moving stimuli. GY’s response criterion differed significantly between yn and FC responding, and the difference was sufficient to result in a blindsight-type dissociation with bias-sensitive measures of performance. When measured independently of bias, GY’s sensitivity to static targets was greater in the FC compared to the yn task (in contrast to normal control participants), but GY’s sensitivity to moving targets did not differ. These results suggested that differences in response criterion could account for dissociations between yn and FC detection of motion stimuli, but not for static target presentations. This may explain the trend for blindsight cases to report increased awareness in response to motion stimuli. Importantly, these results also suggest that blindsight is not qualitatively the same as normal, near-threshold vision and that the neural mechanisms for pattern and motion-detection in blindsight may differ.

The question of decision criterion/response bias has also been addressed by altering the proportion of stimuli to blank trials in perimetry. Despite variations in response criterion, the patient’s ability to detect stimuli remained essentially impervious to such variations, again suggesting that the discrepancy between subjective awareness and detection ability cannot be attributed to a difference in response criterion.

Implications

The anatomic implications of blindsight seem fairly obvious to us now; there are multiple pathways for visual information to reach cortical regions and not all need to be relayed through primary visual cortex to influence behavior. The fact that such pathways exist and that they can support some residual visual function in the absence of conscious awareness belies another question—what function do these residual pathways serve? One possibility is that secondary pathways provide low-resolution representations of visual information that are suitable for rapid responding. In the context of affective blindsight, such a rapid function may be adaptive in generating fight-or-flight behaviors in response to emotionally arousing stimuli. This speculation is perhaps supported by recent evidence that affective blindsight relies on the processing of low spatial frequencies (Burra et al., 2019). In our own work on affective blindsight (Striemer et al., 2019b), the patient's performance fell to chance when she was asked to provide a subjective confidence rating of her discrimination choices after each trial. Perhaps, at least for this patient, the demonstration of affective blindsight relies on patients making fast “gut” reactions, reactions that were hindered by the requirement to indicate confidence.

The idea that blindsight pathways may serve rapid responses to visual input, separate from the geniculostriate pathway that supports conscious vision, also accords with the frequently demonstrated blindsight response to moving stimuli. There are many circumstances in which a rapid response to movement in the periphery is critical, from obstacle avoidance to intercepting moving objects (e.g., imagine dodging an errant baseball thrown in your direction). These pathways may have had a more prominent role in controlling behavior earlier in our evolutionary history. In a similar vein, these pathways may be more active early in neural development when visual-cognitive capacities are immature and the environment is hazardous (Bridge et al., 2016). Prior to developing the sophisticated visual recognition capacities supported by the geniculostriate and ventral visual pathways, we may rely more heavily on signals that convey looming objects, obstacles, or movement, in these more rudimentary pathways (Bridge et al., 2016). Such a hypothesis is difficult if not impossible to test in humans. However, recent work in marmosets suggests that V1 lesions acquired in early life are accompanied by preservation of connections from the retina via the pulvinar to motion sensitive area MT (Warner et al., 2015). In contrast, retinal input to the lateral geniculate nucleus deteriorates following these early V1 lesions. Previous work by Payne et al. (1996) and Payne and Lomber (1998) in cats and Moore et al. (1995, 1996) in monkeys have also shown that residual visual abilities are more robust in animals lesioned early in development. Although speculative, this may explain why blindsight seems to be more prominent in humans whose visual cortical damage was sustained early in life. For example, GY, who has demonstrated an extraordinary range of residual visual abilities, suffered occipital damage after a car accident when he was 7 years old (Barbur et al., 1993).

From Bard’s, Pick’s, and Riddoch's early findings in the early part of the 20th century, to modern tractography and functional neuroimaging studies, the residual visual capacities in the fascinating phenomenon of blindsight demonstrate the sheer complexity of visual processing in the primate brain. Visual processing is highly distributed not only within visual cortex itself, but along distinct subcortico-cortical pathways. We may yet be at the threshold of new discoveries as we start to ask more challenging questions about the functions that are served by these so-called “secondary” visual pathways. Furthermore, the fact that we can still process visual information outside of awareness forces us to contemplate difficult philosophical questions regarding the nature and function of consciousness (e.g., Block, 1995; Brogaard, 2011, 2015).

Blindsight and Subjective Threshold Models

Blindsight is a striking neuropsychological syndrome in which rare individuals suffering from certain forms of brain damage perform quite well on simple direct discrimination tasks (e.g., position discrimination) despite vigorous denials of visual awareness, thus yielding a particularly dramatic subjective threshold effect. Further, blindsight effects differ qualitatively from what simply very weak conscious perception would produce. On the other hand, confidence ratings in blindsighted observers predict performance, in line with weak conscious accounts. Overall, the status of blindsight is unclear. Although many researchers seem convinced that blindsight is genuinely unconscious, it may instead reflect qualitatively different but nonetheless weakly conscious processes. After all, given the brain damage intrinsic to blindsight, one would expect any residual visual capacities to qualitatively differ—whether conscious or unconscious.

Abstract

Blindsight grew out of efforts to compare the functional and anatomical properties of the visual system in human and non-human primates. It refers to the uncanny dissociation between the lack of visual awareness following damage to the primary visual cortex, and the persistence of non-conscious abilities that enable patients to discriminate, respond or act toward “unseen” stimuli in their blind field. Here we review the historical origins of this discovery, the variety of visual properties that can still be processed non-consciously, the methods used to evaluate them, and the neuroanatomical and neurophysiological substrates that mediate blindsight.

Control experiments

Blindsight is a controversial issue. It has been suggested by its detractors that results such as those of our pointing experiments could be artefactual because subjects could use light scattering from the targets into unimpaired parts of their field, as a localized cue [18]. There is no doubt that scattered light may be a problem in such experiments, particularly when bright targets and a high target-to-background contrast are used. However, the amount of blindsight has been found to vary across groups of patients examined in the same experimental situation [41, 84]. This fact clearly indicates that differences between groups cannot be related to experimental variables, but rather pertains to specific characteristics of the groups, such as age, or type or extent of lesion.

This problem of poorer blindsight performance in certain patients, particularly in cases with lesions occurring late in life, is quite apparent in the study of Campion et al. [18]. These authors have reported on pointing performance of three hemianopic patients with lesions of the occipital cortex acquired in adulthood. Stimuli used as targets were as bright as those used in our experiments, and were presented for a longer duration (1 sec). Accuracy of pointing at targets presented within the scotoma was very poor in all three subjects, correlation between target position and pointing position being clearly significant in one case only, non-significant in one case and marginally significant in one case.

The same criticism of scattered light cannot apply to our own control experiment in two adult patients with a bitemporal hemianopia resulting from chiasmatic lesions [85, 86]. In such cases no blindsight is expected to occur, since all retinal projections subserving the temporal fields are blocked by the lesion. Indeed in these two patients, positions of pointings directed at targets appearing within the temporal scotomata were found to be unrelated to target positions (correlations coefficients ranging between 0.09 and 0.37, non-significant). This result indicates that scattered light could not be used by these patients for localizing the targets. Scattered light is an unspecific factor common to all hemianopic subjects examined for blindsight. Therefore, there is no reason why it should not be used by all of them for directing the movements at target location. By contrast, blindsight is a delicate effect revealing the residual activity of an impaired system, and there is no reason why it should be present in all hemianopic patients with cortical lesions.

Introduction

The word blindsight is an oxymoron and the phenomenon is a paradox. I treat this subject from the perspective of a clinical neuro-ophthalmologist who has background in experimental psychology and visual neuroscience.

In the least ambiguous sense, blindsight is the discernment of a visual stimulus in the absence of conscious awareness (blindsight type 1 of Weiskrantz). However, it will soon be evident that much of the literature deals with cases in which the subject acknowledges some perception in the blind region of visual field without actually recognizing the stimulus (blindsight type 2 of Weiskrantz). Further subcategorization of blindsight has been done and will be discussed later. All cases of blindsight require some damage to the primary visual cortex (V1) with sparing of parts of visual association cortex and/or associated neural circuitry. Blindsight, in some ways, can be considered a condition in which there is unawareness of awareness, although this term could be applied to certain forms of agnosia, especially anosoagnosia, or the unawareness of a body part. Furthermore, unawareness and denial of cortical visual loss is present in Anton’s syndrome. Most types of blindsight, however, are not agnosia but something more primitive, involving damage to, and sparing of, specific neural circuitry dealing with the basics of visual perception.

Induction of blindsight

Blindsight is the residual processing of visual stimuli in areas of the field where the subject denies being aware of those stimuli. The neurobiological processes underlying blindsight are covered elsewhere in this series (Chapters 7 and 16). Here, we ask if it can be usefully rehabilitated, with the short answer being that it can in some individuals, though no-one has yet demonstrated that this leads to any practical gains in visual function.

The first attempts to retrain blindsight occurred in Aberdeen (Sahraie et al., 2006). Twelve patients with chronic, poststroke homonymous hemifield defects were studied. The therapeutic stimuli were circular patches of gratings (1 cycle per degree, 6 degree in diameter) flickering at 10 Hz. Three regions were stimulated relatively deep within the blind field (10 − 300) on a daily basis over 3 months, with a total of about 4500 trials per region. The participants had to report in which of two time periods the stimulus had appeared, as well as whether they were aware of the stimulus. Generally, participants improved gradually in their ability to detect stimuli (Fig. 19.1, top), but their awareness did not. More detailed analysis of a subset showed that the training effects were limited to the regions and the spatial frequency trained—that is, benefits did not generalize to other areas of the blind field or to gratings of a different spatial frequency in the trained areas. Subjective improvements in participants’ fields were reported, but no functional outcomes were collected (see Fig. 19.1, bottom.).

A group in Rochester took this work a step further using a different form of stimulation deep in the blind hemifields of five patients with unilateral cortical lesions (Huxlin et al., 2009). They trained subjects to discriminate the left–right direction of the global motion of random-dot stimuli shown for 500 ms. Subjects received trial-by-trial feedback and performed 50–200 sessions, each with 300 trials, to give 15,000–60,000 trials in total. Training improved global direction discrimination (i.e., the same stimulus as used for training) but also generalized to a grating motion detection task and a more difficult motion signal task. As with the Saharie results, the training effects were limited to the same retinotopic location (Fig. 19.2). There was no report on the effect in daily life. In a follow-up paper, the same group demonstrated that retinotopic training effects could be induced by static stimuli, but these did not generalize to motion detection tasks (Das et al., 2014). A further study showed small improvements on clinical automated static perimetry (24–2 Humphrey), but the overall effect was small (~ 1 dB in mean deviation) and tended to be at the edges of the homonymous defects (Cavanaugh and Huxlin, 2017). While the authors argued that improvements of this magnitude are considered clinically meaningful in glaucoma, the analogy may be spurious given the dense visual loss in hemianopia (Huxlin and Cavanaugh, 2018; Wall and Schiefer, 2018). Again, no measures of visual function in daily life were reported. In their latest paper, currently available in nonpeer review form (Saionz et al., 2020), they report two interesting findings using the same treatment method. First, patients treated in the acute phase of their hemianopia seem to learn quicker and with more generalization. However, the second finding was that, compared to a control group with acute hemianopia and no therapy, there was no significant expansion of the visual field on automated Humphrey perimetry in the treatment group.

In summary, in contrast with the disappointing results for restoration of conscious vision, it does appear that blindsight can be induced through training, with generalization across motion detection tasks, although the effects are constrained to the areas of the field that are trained. The ability to induce visual awareness is variable and functional improvements in the tasks of daily life have yet to be demonstrated.

Training blindsight

Whether blindsight can be trained is controversial. Though some deny that training helps (Balliet et al., 1985; Blythe et al., 1987), others claim that practice can improve saccadic or manual localization (Zihl, 1980; Bridgeman and Staggs, 1982; Zihl and Werth, 1984a, b; Magnussen and Mathiesen, 1989). A study that trained 9 hemianopic subjects with a variety of forced-choice localization and form discrimination tasks found improvement after 5–6 months (Chokron et al., 2008). Training of motion perception may improve performance in the retinotopic region trained (Huxlin et al., 2009), and repeated training over 3 months with detection of gratings can improve contrast sensitivity in hemianopic fields (Sahraie et al., 2006). There are even claims that such training may expand visual fields (Zihl and von Cramon, 1979; Zihl, 1981; Chokron et al., 2008).

The Role of Attention in Unconscious Visual Processing

It was noted above that blindsight patients can direct attention to the location of a stimulus presented in the visual field defect. It is also becoming clear that attention can modulate the effects produced by unconsciously processed stimuli. Such attentional modulation has been found in the unconscious processing of stimuli whose visibility is suppressed during backward masking and during visual crowding. Nonetheless some visual stimuli, such as natural scenes or faces and their emotional expressions, whose visibility is suppressed by an aftercoming mask seem to require little if any attention in order to be processed unconsciously. Natural scenes and faces occur very often in our normal everyday commerce with the visual world. In addition faces conveying emotional expressions are particularly salient social stimuli. For any of these reasons such stimuli might be processed automatically with little or no need of attentional modulation. On the other hand, the unconscious processing of abstract and nonecological stimuli typically used in laboratory experiments may be subject to attentional modulation since they are both less familiar and less salient.

Publisher Summary

This chapter discusses the phenomenon of blindsight. This phenomenon has attracted the attention of philosophers, psychologists, and neuroscientists because it highlights the nature of covert vision, indicates that striate cortex is indispensible for visual awareness, and provides a means of studying the visual information carried by pathways other than the major route through striate cortex. Denial of visual experience may be present even when the subject's detection or discrimination approaches 100% correct, when localization of an unseen target is excellent, or when the threshold for detection is reduced by less than a log unit. It has been seen that the presentation of a line stimulus in the blind field influences judgments about the relative distance of targets presented in the normal field and that the color of targets in the seeing field is affected by unseen colors in the scotoma. Some patients report occasional sensations, which may even be of a visual nature. The chapter presents evidence that suggests that cortex may be involved in processing visual information in blindsight. A recent study has demonstrated that in visual cortex, frequency locking, especially between different visual areas, is difficult without feedback between areas that are connected. Removing striate cortex, with its massive reciprocal connections with several secondary cortical visual areas that are in turn further connected with each other and with additional visual areas, is likely to disrupt or even destroy this delicate phase locking.