One Kind of Consciousness or Two?
Access and phenomenology in visual perception
The neuroscience literature concerning the putative distinction between conscious phenomenology and conscious access, with respect to visual perception, is surveyed. Several sets of empirical findings are considered (blindsight, iconic memory and change blindness, as well as signal detection theory experiments on monkeys). Two theoretical positions are described. One holds that conscious states are intrinsically accessible. A contrasting view is that there are two distinct categories, access consciousness and phenomenal consciousness, for mental representations of visual percepts. On the latter view, there are phenomenal states unavailable to conscious access, and access and phenomenology have distinct neural correlates. On balance, the experimental evidence is inconclusive. Some suggestions are given for further empirical work which could resolve certain remaining questions.
Phenomenal and Access Consciousnesses
Psychologists and neuroscientists studying visual consciousness distinguish between several types of consciousness; notably, they appeal to a distinction between access consciousness and phenomenal consciousness (Pinker, 1997). Phenomenal consciousness (PC) refers to the qualitative experience of content associated with sensory perception. Access consciousness (AC), in contrast, is the property of content which is made widely available to neural systems, including those involved in memory, perceptual categorization, decision-making, and others (Block, 2005). Access conscious content is generally subject to some form of report, verbal or otherwise. (The distinction between these two types of visual consciousness is closely related to the distinction between attention to and awareness of visual inputs; see Lamme, 2003.) There is disagreement as to whether PC and AC should be modeled as distinct or coextensive categories for mentally represented visual information (Wiens, 2007), as well as concerning whether one or two neural correlates are responsible for these types of consciousness (Block, 2005; 2005b; Baars & Laureys, 2005).
Two Points of View on Phenomenology and Access
In what follows two scientific viewpoints are considered. The first is part of what might be called “orthodox” global workspace theory, which hypothesizes that there is a network of neurons that integrates the activity of a diverse multitude of specialized networks into a unified global workspace (Baars, 2002). The view under consideration holds that an item of information (such as a visual percept) represented by a particular neural population is subjectively experienced as conscious when that population is mobilized via top-down attentional amplification into a state of synchronized activity involving neurons throughout the brain. This synchronized activity “broadcasts” the information in question to the global workspace, making it available for a variety of processing operations (Dehaene & Naccache, 2001). On this view, visual information becomes both conscious and access conscious (in the sense of the previous subsection) when the corresponding patterns of activity in the posterior cortex (Crick & Koch, 1995) undergo top-down attentional amplification. Consequently, this interpretation of global workspace theory holds that there cannot be PC without AC.
An alternative point of view, most strongly associated with the philosopher Ned Block (1995, 2005, 2007) but also advocated by several neuroscientists (Lamme, 2003; Koch & Tsuchiya, 2006) hypothesizes that conscious awareness of a visual percept hinges upon the occurrence of recurrent processing of the visual information in question. This recurrent processing involves networks of neurons linking activated regions of the visual cortex to higher level nuclei in the frontal cortex, as well as to other areas of the brain (Lamme & Roelfsema, 2000). Such recurrent processing, on this view, comprises a neural correlate (or causal mechanism) of PC, and can - but need not - be activated via the top-down attentional amplification process posited by the global workspace model. Instead, a “winner-take-all competition” for broadcast into the global workspace takes place among neural representations of visual information involved in recurrent loops of varying strength. Only once they are attended and broadcast to the global workspace do these representations become access conscious (Block, 2005). As a result, it is possible that strong, but not winning, representations are phenomenally conscious without having been attended or become accessible (Block, 2005; Lamme, 2003).
Evidence for the Orthodox Global Workspace View
As a self-consistent theory integrating much of what is known empirically concerning consciousness (in a broad sense) and working memory, global workspace theory has proved its conceptual usefulness many times over, and its predictions have been corroborated by a vast body of supporting evidence from neuroimaging studies (Baars, 2002). It would seem that this lends the orthodox global workspace view of access and phenomenology some credence, as it fits within a successful theoretical framework for understanding the mind and the brain. However, what is at stake in the debate concerning PC and AC is not the correctness of the global workspace theory in the large (this is widely accepted by the scientific community), but rather its “chauvinism” with regard to which neural events underlie conscious experiences (Block, 2007). The success of the global workspace hypothesis as a broad theoretical framework therefore does not entail the correctness of the orthodox view that only attended, and hence accessible, mental representations are conscious. Indeed, alternate accounts have been proposed (Block, 2005; Lamme, 2003) which incorporate almost the entirety of global workspace theory while nevertheless admitting into (phenomenally) conscious awareness certain mental representations that have not received top-down attentional amplification. Since these alternatives sacrifice no explanatory power with respect to describing access consciousness and other phenomena scientifically, the coherence and success of global workspace theory offers no theoretical evidence for rejecting a conceptual and neural distinction between PC and AC. We next turn to empirical, rather than theoretical, justification for the unitary, orthodox global workspace theory account of phenomenology and access.
Blindspots and hemi-neglect. Dehaene & Naccache (2001) review one class of empirical research often invoked in support of the orthodox global workspace account of access consciousness. This research centers on contrasting phenomenology among (a) normal subjects with an ordinary retinal blindspot, (b) subjects who have developed a retinal scotoma producing an abnormal blindspot, and (c) subjects with parietal brain lesions suffering from hemi-neglect. There is robust empirical evidence, surveyed by Dehaene & Naccache (2001), that in case (a) the subject is unconscious of visual information presented before the blindspot, unconscious of this visual deficit (we do not “see” a hole in our visual field), and unable to process such visual information at an unconscious level. In case (b) the subject is unconscious of visual information presented before the blindspot and unable to process such information, yet conscious of his or her visual deficit. Finally, the most surprising results concern inattentional “blindsight” in hemi-neglect patients (c) who are unconscious of visual information before their neglected field and unconscious of their deficit, but are nevertheless able to process such information (as indicated, for example, by behavioral measurements of priming; see McGinchley-Berroth et al., 1993).
Dehaene & Naccache (2001) interpret these findings to support the theory that consciousness requires top-down attentional activation. In case (a) subjects' normal retinal deficit entails that there is no neural representation of visual information corresponding to the topographic location of the blindspot; hence a fortiori no attentional amplification can occur and the subject is unconscious of this part of the visual field. However, neither are there visual representations in long-term memory corresponding to the blind part of the visual field, so when remembered visual scenes are attended this part of the visual field remains unconscious. The lack of discrepancy between ongoing and remembered visual consciousness explains subjects' lack of awareness of their visual deficit. In case (b) the only difference is that neural representations corresponding to the scotomic portion of the retinal field persist in long-term memory; when these representations are attended subjects become conscious of the remembered visual information that was then before what is now their blindspot. Such conscious visual experience contrasts with their phenomenal experience of their current, damaged visual field, and this contrast makes them aware of their deficit. (Note that this explanation is cogent because aside from their dysfunctional retinas, scotomic patients have normal neural activity, so in particular the neural basis for attentional amplification and broadcast to the global workspace is intact.) Finally, hemi-neglect patients have intact retinas and primary visual cortices, but are lesioned in parietal areas which may be involved in attentional processing. This hypothesis combined with the orthodox global workspace theory is consistent with the phenomenology of case (c), in which subjects report themselves unconscious of visual information concerning the neglected field and unconscious of this deficit, but the visual information is nonetheless available to other cognitive processes. In particular, although information concerning the neglected visual field presumably exists as intact neural representations in the brains of such subjects, their lesions prevent these representations from undergoing attentional amplification and becoming conscious; nonetheless, unconscious processing of these representations remains possible.
In case (c), the orthodox global workspace interpretation therefore rests upon two claims: (1) neglect patients maintain largely intact neural representations of visual information concerning their neglected field; and (2) parietal lesions prevent such representations from becoming conscious by interfering with the process of attentional amplification. We now turn to specific empirical findings which speak to the validity of these claims. As mentioned above, some evidence for claim (1) is afforded by behavioral studies involving priming. When hemi-neglect patients are shown pairs of images in their left and right visual fields, they report themselves unaware of the images in their neglected fields, which is consistent with the phenomenology (c) described above. But when given a word in the center of their visual fields and asked to say whether or not it accorded with the previously displayed images, neglect patients performed comparably with normal subjects. The implication is that the unconscious visual information triggers the same semantic priming in neglect patients as in normal subjects, indicating that the information undergoes ordinary semantic processing despite remaining unconscious (McGinchley-Berroth et al., 1993).
In addition to such behavioral evidence, neuroimaging studies have also borne out predictions of neural activity consistent with claim (1). Rees et al. (2000) used functional magnetic resonance imaging to measure activity in the visual cortex of the hemi-neglect patient G.K., suffering from parietal lesions of the sort considered above. In this study, G.K. was shown objects in either of his visual fields, or both, and in the bilateral trials reported unawareness of objects in the neglected field. Importantly, G.K. and other hemi-neglect patients are not unconscious of stimuli in their neglected field in the absence of stimulus in the other field. This provides a basis for comparing activation of the visual cortex in the hemisphere processing information from the neglected field (the ipsilesional hemisphere) both when that information is conscious (unilateral trials) and when it is unconscious (bilateral trials). The fMRI results showed conclusively that activation in the visual cortex of the ipsilesional hemisphere can occur without awareness of objects in the contralesional visual field. Moreover, the pattern of activation was found to be “strikingly similar” in the unilateral and bilateral trials (Rees et al., 2000). However, the fMRI study could only show the stastical absence of visual cortical loci which are reliably activated by visual stimuli of which G.K. was aware rather than unaware, or vice versa. Qualitative differences between the unilateral and bilateral trials were observed (namely, the size of the activated region), and possible differences in temporal (rather than spatial) activation patterns could not be ruled out by this study. Hence, while these neuroimaging results constitute forceful empirical evidence for claim (1) and thus bear out the orthodox global workspace interpretation of visual consciousness in hemi-neglect patients (Dehaene & Naccache, 2001), such evidence is not incontrovertible.
There is also a body of empirical work supporting the claim (2) that parietal lesions prevent visual information in the neglected field from becoming conscious by interfering with attentional amplification (see Driver & Mattingley 1998 for a survey of findings on neglect). For example, Posner et al. (1984) studied the effects of various brain lesions on covert orienting of visual attention, meaning orienting visual attention without overt reorientation of the eyes or the head. The authors of this study instructed patients to detect a target at one of two possible positions in the visual field (contralesional or ipsilesional), but first used arrow cues to direct subjects' attention either to the correct target location or to another location. Reaction time to target was measured as a function of the location of the cue and the time between cue and target; these reaction times were then correlated with the extent and anatomical location of the patients' lesions. Parietal patients exhibited a unique reaction time pattern, with an especially marked increase in reaction time to trials in which an invalid directional cue was given for a target in the contralesional field. Such trials with invalid cues differ from trials with valid cues in that after an invalid cue has been given attention must be disengaged from an incorrect location before it can be oriented to the target. Posner et al. conclude that parietal lesions interfere particularly with the cognitive operation of disengaging attention from its current focus. Findings such as this do not demonstrate conclusively that parietal lesions prevent visual representations from becoming conscious in hemi-neglect patients by obstructing attentional amplification. Nonetheless, such findings do indicate a prominent role for the parietal cortex in the cognitive processes involved in reorienting attention, which makes the orthodox global workspace interpretation of hemi-neglect patients' phenomenology at least somewhat plausible.
Visual masking. Another set of neuroimaging studies invoked by proponents of the orthodox global workspace view (Baars & Laureys, 2005) is the work of Dehaene et al. (2001). In this study functional magnetic resonance imaging and event related potentials were used to measure activation in parts of the cortex known to be involved in the conscious processing of language, while subjects attempted to read a word flashed for approximately 10 milliseconds with and without masking. Masking is the phenomenon whereby subjects usually find themselves able to read the word in this situation, but fail to do so when distracting visual stimuli are presented in close proximity to the target. It was found that under masking conditions, the neural markers of conscious language processing, notably significant activation in large brain-scale neural networks thought to be involved in comprising the global workspace, did not occur (Dehaene et al., 2001). Baars & Laureys (2005) argue that since activity in the visual cortex persists, this study indicates that visual consciousness and activity in the visual cortex are uncorrelated. This is consistent with the orthodox global workspace theory (since consciousness requires attentional amplification that masking may block), but not with the theory that PC has a neural basis consisting of activity in the visual cortex which is not broadcast to the global workspace.
A Theoretical Critique
Block (2005b) argues that none of the aforementioned empirical findings are inconsistent with a neural distinction between AC and PC. The neat account of various blindness phenomena offered by global workspace theory fails to disprove the existence of phenomenology without access because it conflates the two. In particular, the criterion for consciousness it presumes is the presence or absence of a reportable, and hence accessible, phenomenology. Whether or not hemi-neglect patients, for example, are phenomenally conscious of information in their neglected visual field, the fact that lesions in brain areas involved in attentional amplification of representations in the neglected field prevent access consciousness of those representations does not speak to any scientific issue concerning phenomenal consciousness of those representations. The same objection applies to the corroborating neuroimaging results of Dehaene et al. (2001), which moreover do not detect the presence or absence of recurrent neural activity in the visual cortex, a crucial feature of the theory that PC and AC are distinct. These arguments indicate that the adduced empirical evidence does not uniquely support the orthodox global workspace theory.
Empirical Evidence for Distinct Access and Phenomenal Consciousnesses
The theory that AC and PC are distinct is subject to a number of subtle philosophical objections on epistemic and metaphysical grounds. These issues will not be considered here, but it has been argued (Block, 2007; see also the included commentaries by philosophers and neuroscientists and Block's replies) that it is an empirical question whether the neural basis for PC contains the neural basis for AC (and hence whether AC is necessary for PC). We will presume there is no logical obstruction to a distinction between AC and PC for the remainder of our discussion.
Since global workspace theory is capable of explaining a wide variety of consciousness phenomena along the orthodox lines mentioned previously, a proponent of the theory that PC and AC are distinct and correspond to distinct neural machinery must adduce empirical evidence that the latter theory is better suited to explaining. We now turn to the evidence that has been offered for this purpose by Block (2005; 2007) and Lamme (2003).
Signal Detection Theory
Supér et al. (2001) showed monkeys textured patches that either did or did not contain a target of altered texture in one corner, training the monkeys to saccade to the target when it appeared and to fixate on the center of the patch otherwise. They then used implanted “figure” and “ground” microelectrodes to measure activity in receptive areas of the primary visual cortex (V1) corresponding to the location of the target and another location on the patch. Finally, they recorded both the monkey's behavioral response (presence or absence of saccade towards the target upon its appearance) and neural “modulation” (change in activity measured by the electrodes upon the appearance of the target) under varying conditions of saliency (the distinction in texture between the target and the patch was varied along a continuous gradient). The authors found that the monkey's neural modulation was manipulable by varying the saliency and the number of “catch” trials in which no target was presented.
Moreover, with a very high saliency or a very low proportion of catch trials, correlation between neural modulation and saccade behavior was high; with a very low saliency or a very high proportion of catch trials, this correlation dropped almost to zero.The authors also note that in similar studies on monkeys anesthetized with isoflurane (but with their eyes open) the observed patterns of neural modulation measured by the figure and ground electrodes disappeared (Lamme et al., 1998). Supér et al. demonstrate using a form of analysis called signal detection theory that these findings are consistent with a model (corroborated by further experimentation) in which neural modulation is associated with a neural representation intermediate between visual perception and conscious access to the corresponding visual information (see Block, 2005).
This remarkable experiment in neural signal detection theory is presented by Block (2005) as evidence for a distinction between the neural correlates of PC and AC. Drawing upon other findings linking neural modulation in V1 to recurrent activity in the visual cortex, as well as the finding that neural modulation in the monkeys of the Supér et al. study disappeared under anesthesia, he suggests that such modulation is plausibly involved in (if not partially constitutive of!) phenomenal consciousness. Lamme (an author of the Supér et al. study) shares this interpretation of the data (Lamme, 2003). Assuming this, the low correlation measured between modulation and saccade behavior under low saliency or high catch-trial frequency conditions indicates that this particular PC correlate can occur independently of conscious access, providing evidence that PC and AC are distinct on a neural level.
Of course, with respect to challenging the orthodox global workspace interpretation of consciousness, the results of this study are inconclusive. For by the orthodox global workspace theory the measured neural modulation is a neural correlate not of PC but of some preconscious visual representation, which nevertheless likely has a link to recurrent activity in the visual cortex. So as evidence for a PC-AC distinction, this experiment relies on independent evidence for a correlation between recurrent activity in the occipital cortex and PC.
Iconic Memory and Change Blindness Phenomena
Block (2007) argues that the results of several classical experiments lead to interpretations comprising a psychological-neuroscience “mesh” of plausible explanation, whereby the informational content of phenomenal consciousness can overflow that of access consciousness in a way explained by the hypothesis that recurrent loops comprise a neural basis for PC. The inference is that, combined with evidence such as the neural signal detection theory result just described, plus other psychological signal detection theory results (see Block (2005) for a review), a distinction between PC and AC and between their neural correlates is the best explanation cognitive science currently possesses for the observed data.
This subsection will outline the iconic memory and change blindness experiments which suggest a cognitive overflow of PC beyond the capacity of AC. Subjects in the experiment of Sperling (1960) were briefly shown an array of symbols and reported a persistent (on the order of hundreds of ms) phenomenal perception of the entire array; however, when asked to verbally report the precise symbols in the array, subjects are only able to recall about four of them. Interestingly, which four are recalled depends on which four subjects are asked to recall. When the pitch of a tone played immediately after the visual stimulus is removed dictates whether, e.g., the highest, middle, or lowest row of symbols is to be reported verbally, subjects accurately report four symbols from the correct row. This experiment was then adapted by Landman et al. (2003) to a change blindness paradigm (Block, 2007; Gray, 2007) as follows. Subjects are briefly shown an array of rectangles for 0.5 sec, and the array is then replaced by a blank screen for a variable period, followed by another array of rectangles in which one indicated object may or may not have changed orientation. Subjects are asked to say whether or not an orientation change took place. As in ordinary change blindness experiments, the blank screen prevents subjects from detecting changes in objects whose original orientations are not held in working memory. After correcting for guessing, Landman et al. found that subjects display a limited capacity to keep track of the orientations of only four rectangles in the array. But as in Sperling (1960), subjects report phenomenal awareness of perceiving all or most of the rectangles.
Block (2007) suggests that both the Sperling and Landman et al. experimental paradigms indicate that the informational content of phenomenal consciousness overflows the cognitive capacity of conscious access. This takes for granted the fact that subjects' self-reported phenomenal experiences are accurate, and that they hold an entire array of information in phenomenal consciousness even though only a small portion is available to AC. This “meshes” with the neuroscience hypothesis that recurrent loops provided a neural basis for PC, which is supported by evidence such as the results of Supér et al. (2001) and other studies surveyed in Block (2007), as follows. On this hypothesis various coalitions of neurons compete for attentional amplification and concomitant broadcast to the global workspace, placing them in access consciousness. This competition entails a loss of information during the amplification stage. Consequently, this hypothesis can explain the psychological observation of “information overflow.” On the other hand, the orthodox global workspace theory of unitary, accessible consciousness provides no such ready explanation. Therefore, the distinction between PC and AC should be accepted as an inference to the best explanation of the available data.
Certainly a weak point of this argument is that it depends on subjects in these experiments really being phenomenally consciousness of the larger array of objects. Block (2007) argues that there is no reason to reject their reports of such phenomenal consciousness as faulty, but Dehaene et al. (2006) suggest that these reports might be the result of a type of illusion (which Block (2007) calls a hyperillusion). To wit, content which is merely potentially accessible - and hence (by the orthodox global workspace interpretation) merely potentially phenomenally conscious - might seem phenomenally conscious because when asked to report whether they saw the entire array, subjects attentionally amplify the entire array in a “dilute” way that loses information regarding the precise content of the characters in the array (Dehaene et al., 2006). We will not consider this objection further here, as it remains to be empirically tested.
Since the crux of the empirical argument in support of a PC-AC distinction rests upon an inference to the best explanation from available psychological and neuroscientific data, it is worth examining the scientific reliability of such forms of inference. In any such situation the danger with using inference to the best explanation to support a single given hypothesis is that one overfits the available data. A more reliable approach is to generate a large variety of possible explanations and compare their fit to the given data points. Viewed from this standpoint, the “mesh” argument of Block (2007; namely, that PC overflow of AC information capacity meshes with known neural facts about recurrent loops in way that empirically supports both phenomena) is inadequate (Hulme & Whiteley, 2007).
Conclusions and Directions for Further Research
As the preceding analysis illustrates, a variety of subtle methodological questions, as well as a collection of perplexing philosophical quandaries, complicates the dispute over the putative distinction between phenomenal and access consciousness. As a number of vehemently opposed and a number of vehemently supportive commentaries on recent work of Block (2007; see the same paper for the commentaries) illustrates, there is a decided lack of scientific and philosophical consensus on the relevant questions. Indeed, the most relevant empirical evidence, such as the neuroimaging studies done to date and the neural signal detection theory study of Supér et al. (2001), is inconclusive in supporting or refuting the existence of a distinction between the neural correlates of PC and AC. Behavioral and neuroimaging studies largely bear out explanations of phenomena involving access consciousness according to the global workspace theory, but do not speak to the correctness of the orthodox version of this theory in ruling out phenomenal consciousness without access. On the psychological side, the question remains whether iconic memory and change blindness paradigms suggest an overflow of the information content of PC beyond that of AC, or merely an illusion of phenomenal perception in certain circumstances which can be explained in the context of an orthodox interpretation of global workspace theory. The issues surrounding this question seem too philosophically fraught to be resolved through simple psychological experimentation. (For example, what distinguishes an “illusion of perception” from a perception?)
Certain further empirical work would be valuable, however. Ideally, sufficiently sensitive non-invasive electrical probes will be developed to perform an experiment along the lines of Supér et al. (2001) with humans engaged in a task along the lines of Sperling's (1960). Alternately, perhaps existing neuroimaging techniques (such as fMRI) can be refined to measure the recurrent neural activity, potentially free from top-down attentional amplification, in the visual cortices of such subjects. If such activity can be measured in humans independently of conscious access, it would also be interesting to see whether it persists under various eyes-open states of sleep and anesthesia. While such experiments will not convince a skeptic that the neural mechanisms in question correlate to phenomenal consciousness rather than a preconscious intermediate representation between perception and access consciousness, they would at least indicate whether this intermediate representation possesses the most salient features of what we intuitively refer to as PC (for example, seeming like PC, even if such an appearance is illusory).
Baars, B. (2002). The conscious access hypothesis: origins and recent evidence. TRENDS in Cognitive Sciences 6(1), 47-52.
Baars, B., & Laureys, S. (2005). One, not two, neural correlates of consciousness. Letter, TRENDS in Cognitive Sciences 9(6), 269.
Block, N. (1995). How many concepts of consciousness? Behavioral and Brain Sciences 18(2), 272–84.
Block, N. (2005). Two neural correlates of consciousness. TRENDS in Cognitive Sciences, 9(2), 46-52.
Block, N. (2005b). The merely verbal problem of consciousness. TRENDS in Cognitive Sciences, 9(6), 270.
Block, N. (2007). Consciousness, accessibility, and the mesh between psychology and neuroscience. (With commentaries and reply.) Behavioral and Brain Sciences 30, 481–548.
Crick, F., & Koch, C. (1995). Are we aware of neural activity in primary visual cortex? Nature 375, 121-123.
Dehaene, S. et al. (2001). Cerebral mechanisms of word masking and unconscious repetition priming. Nature Neuroscience 4(7), 752-758.
Dehaene, S., & Naccache, L. (2001). Towards a cognitive neuroscience of consciousness: basic evidence and a workspace framework. Cognition 79, 1-37.
Dehaene, S. et al. (2006). Conscious, preconscious, and subliminal processing: a testable taxonomy. TRENDS in Cognitive Sciences 10(5), 204-211.
Driver, J., & Mattingley, J. (1998). Parietal neglect and visual awareness. Nature Neuroscience 1(1), 17-22.
Gopnik, A. (2007). Why babies are more conscious than we are. (Commentary on Block 2007.) Behavioral and Brain Sciences 30, 503-504.
Gray, P. (2007). Psychology. New York: Worth Publishers.
Hulme, O., & Whiteley, L. (2007). The “mesh” as evidence – model comparison and alternative interpretations of feedback. (Commentary on Block 2007.) Behavioral and Brain Sciences 30, 505-506.
Koch, C., & Tsuchiya, N. (2006). Attention and consciousness: two distinct brain processes. TRENDS in Cognitive Sciences 11(1), 16-22.
Lamme, V., et al. (1998). Figure–ground activity in primary visual cortex is suppressed by anaesthesia. Proceedings of the National Academy of Sciences, U.S.A. 95, 3263–3268.
Lamme, V., & Roelfsema, P. (2000). The distinct modes of vision offered by feedforward and recurrent processing. Trends in Neuroscience 23, 571-579.
Lamme, V. (2003). Why visual attention and awareness are different. TRENDS in Cognitive Sciences 7(1), 12-18.
Landman, R., Spekreijse, H. & Lamme, V. A. F. (2003). Large capacity storage of integrated objects before change blindness. Vision Research 43(2), 149–64.
McGlinchey-Berroth, R., et al. (1993). Semantic priming in the neglected field: evidence from lexical decision task. Cognitive Neuropsychology 10, 79-108.
Pinker, S. (1997). How the mind works. New York: W.W. Norton & Co.
Posner, M., et al. (1984). Effects of parietal injury on covert orienting of attention. Journal of Neuroscience 4, 1863–1874.
Rees, G., et al. (2000). Unconscious activation of visual cortex in the damaged right hemisphere of a parietal patient with extinction. Brain, 123 (Pt. 8), 1624-1633.
Sperling, G. (1960). The information available in brief visual presentations. Psychological Monographs: General and Applied 74(11, Whole No. 498), 1–29.
Supèr, H., Spekreijse, H., & Lamme, V. (2001). Two distinct modes of sensory processing observed in monkey primary visual cortex (V1). Nature Neuroscience 4(3), 304-310.
Wiens, S. (2007). Concepts of visual consciousness and their measurement. Advances in Cognitive Psychology 3(1-2), 349-359.