Abstract
The minimally conscious state (MCS) is usually ascribed when a patient with brain damage exhibits observable volitional behaviors that predict recovery of cognitive functions. Nevertheless, a patient with brain damage who lacks motor capacity might nonetheless be in MCS. For this reason, some clinicians use neural signals as a communicative means for MCS ascription. For instance, a vegetative state patient is diagnosed with MCS if activity in the motor area is observed when the instruction to imagine wiggling toes is given. The validity of using neural signals in ascribing MCS requires a special sort of inference. That is, no-report paradigmatic assessments must have inductively strong ways of inferring a purported informational content from the observed neural signal that grounds the fact that the patient has top-down cognitive control (or residual volition). Shannon’s mathematical theory of communication and Bayes’ theorem reveals the formal structure of neural communication. On the basis of relevant data from the neuroscience literature, I conclude that the formal structure combined with the data shows that neural signals can be used as a communicative means for operational diagnostic criteria for MCS ascription.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Explore related subjects
Discover the latest articles, news and stories from top researchers in related subjects.Notes
By ‘behavioral’, I mean not only motor but also non-motor actions like speaking.
In contrast to behavioral assessments, assessments of the no-report paradigm do not depend on observable volitional behaviors such as verbal or bodily reports or eye movement. In this paper, the no-report paradigm refers to cases where “non-behavioral” brain activity is taken as the evidence for MCS ascription. Particularly, I discuss cases where brain activity is taken as a neural signal that evidences residual volition of the subject.
CRS-R measures the degree of MCS on the bases of the following observable behavioral factors: auditory, visual, motor, oromotor or verbal, and communication (Giacino et al. op. cit.; Schnakers et al. op. cit.). CRS-R quantifies the responsiveness of a patient with respect to all these factors (but not only to a specific one) in ascribing MCS.
Suppose that a patient is able to track her face in a mirror for several seconds with considerable stability. The patient has a better chance of recovery given rehabilitation, which typically presupposes long-term (e.g., a few months) intensive treatment (Ansell 1993: 91–92). For this reason, Ansell (1993, 1995) writes that visual tracking performance predicts rehabilitation-ready status rather than consciousness. Following Ansell, I take MCS as a state that predicts recovery of consciousness given rehabilitation.
For instance, there can be a locked-in syndrome patient who is fully conscious but unable to perform any observable behavior.
Some clinicians seem to interpret VT in this sense. For instance, Owen et al. (2006: 1402) write that the observed brain activity clearly demonstrates that the vegetative state patient in their experiment is consciously aware of herself and her surroundings. Owen (2013: 123) writes that the “philosophical” possibility of zombies (i.e., human-like machines without any phenomenal experience) is practically irrelevant because there is no neurophysiological data that supports the zombie hypothesis. Nevertheless, Owen’s claim begs the question because we have no compelling neurological account for neural correlates of consciousness (NCC).
For more detailed discussion about this topic, see Shea and Bayne (op. cit.) and Drayson (op. cit.). Hohwy (2009) is a nice guide to different types of consciousness and their relations to agency.
For instance, Laureys et al. (2000) attempt to identify thalamocortical connectivity as a condition for consciousness through results showing that such connectivity is observed in controls but not in vegetative state patients.
Nevertheless, although both motor action and mental action are functionally connected to the motor area, percentage increases in signal intensity (of the BOLD signal) during mental action are on average 30% as great as increases during motor action (Porro et al. op. cit.). This means that there is a difference between the functional connectivity of motor action and that of mental action.
This is the so-called ‘semantic priming effect’. The semantic priming effect includes cases like the following: automated activities in the somatotopy of the motor and premotor cortex are observed for a few milliseconds when a subject hears a word related to foot or leg such as ‘kick’ (Pilvemüller 2005).
The relevant localization with respect to the Toe-Imagery task could involve a network of areas including the left paracentral lobule, the posterior part of the supplementary motor area, and the right inferior frontal gyrus, which are the areas known to be related to the imagery of toe movement (Ehrsson et al. 2003). For the sake of simplicity, however, in this paper I define the Toe-Motor-signal as pertaining to the medial premotor cortex only rather to the aforementioned network of areas.
In this paper, any use of ‘log’ implies the logarithm to base 2 (e.g., log20.5). The amount of information that a particular state of a system, Si, carries is –logP(Si) (Harms 2006).
The information of S in the state i is \(I\left( {{\text{S}}_{i} } \right) = - { \log }_{2} P\left( {{\text{S}}_{i} } \right)\).
\(I\left( {{\text{S}}_{i} } \right) = {\text{log}}_{2} {\mathbf{n}}\) (n is the number of possible states of S) when the systems is unbiased (or the possible states are equiprobable).
Intuitively, as P(Si) gets lower, Si carries more information, as it gets larger, Si carries lesser information, and if P(Si) = 1, then Si carries no information. P(Si) = 0 implies the amount of information is infinite, and –logP(Si) ≥ 0. (Isaac forthcoming).
The purported informational content is ‘the subject is mentally wiggling toes’ rather than ‘the subject is volitionally mentally wiggling toes’. This is because volition is ascribed on the basis of the informational content. That is, the content implies that the subject has cognitive capacities like language comprehension and working memory that involve top-down control.
The semantic priming effect by itself does not preclude the possibility that a vegetative state patient has residual volition. Nevertheless, we cannot infer that the patient has residual volition from the informational content about the semantic priming effect because such effect does not seem to require top-down cognitive control.
A full description of the mental event is ‘E1 given the Toe-Imagery task instruction (T)’. Thus a full description of P(E1) is P(E1|T). Ei should be taken as a possible mental event (or state) that a subject could undergo when the Toe-Imagery task is performed by the subject.
It is worth noting that activity in the medial premotor cortex tokens a signal in that such activity is observed under the task circumstances (or when the instruction is given). In other words, we are trying to measure the changes that the tokening of M as a response to the given instruction brings about with respect to E1 comparing to P(E1) unconditioned to the tokening of M. So, as the full description of P(E1) is P(E1|T), a full description of P(E1|M) is P(E1|M) given T (or P(E1|M and T)).
Schnakers et al. (2009) report that of the 44 patients diagnosed with the vegetative state based on the clinical consensus of the medical team, 18 (41%) were found to be in MCS (i.e., the rehabilitation-ready state) following standardized assessments of CRS-R. This suggests that assessments of CRS-R can discriminate MCS patients from comatose patients more reliably compared to non-standardized behavioral assessments, thus 0.15 seems to be a reasonable number. One might claim that I am biased in evaluating the reliability of behavioral assessments and CRS-R should be taken as more reliable. As a matter of fact, however, setting P(E1) lower than 0.15 would increase the inductive reliability of VT (refer to footnote 28 for details).
Poldrack uses Bayes’ theorem to explain a reverse inference, which is a practice in brain image studies by which the engagement of a particular cognitive process in a new study (that prima facie does not involve that process) is inferred from the activation of a particular brain region that has been associated in previous studies with that cognitive process (Poldrack 2006).
About 80% of brain activity (measured by EEG) during continuous hand movements by clenching softly a ball is observed in the hand area (i.e., the lateral premotor cortex) with fourteen healthy subjects (Neuper et al. 2005), which suggests that voluntary motor action functionally requires activation in some specific areas of the brain.
Neuper et al. (2005) categorize motor-mental action into two types, namely, the kinesthetic motor mental action and the visual motor mental action. The former kind refers to cases where a subject is imagining doing something from the first-person perspective. The latter kind refers to cases where a subject is imagining motor imagery from the third-person perspective. Neuper et al. claim that the primary reason motor mental action sometimes fails to involve activity in the relevant motor areas is that subjects often perform the visual motor mental action instead of the kinesthetic mental action. For example, misunderstanding of a given instruction to imagine clenching a ball in the right hand, according to Neuper et al., might result in activity in occipital areas instead of the hand area. Given this possibility, it is reasonable to assume that P(M|E1) pertaining to the vegetative state group (i.e., 0.5) is lower than P(M|E1) pertaining to the control group (i.e., 0.75) since following an instruction correctly would be harder for a vegetative state patient.
This estimation roughly corresponds to the actual outcome of the Toe-Imagery task, where 3 vegetative state patients showed activity in the toe area (Cruse et al. op. cit.).
Similar to the case of E1 and E2, one might claim that P[(E1|M) and (E2|M)] > 0 and P(E1|M) + P(E2|M) < 1. In the text, I have explained why the co-occurrence scenario is insignificant. Nevertheless, there could be E3 given M, the case where a subject has residual volition but simply fails to follow the instruction. E3, if it is possible, implies that the activation of the medial premotor cortex (i.e., E3 given M) is either purely a random noise, which is statistically insignificant, or a replicable phenomenon. If it is the latter, then we can treat P(E3|M) as included in P(E1|M) because P(E3|M) is a reliable informational indicator of residual volition (although the observed neural activity, or M, might not be physiologically or causally related to residual volition). In other words, given that ‘E3 given M’ is replicable, the obtaining of E3 given the tokening of M reliably predicts that the subject has residual volition even if it does not have the content ‘wiggling toes’ (yet, P(E3|M) must be very low since the relationship between the tokening of M and the obtaining of E3 under the task circumstances seems hard to explain). So, I also assume that P[(E1|M) and (E2|M)] = 0 and P(E1|M) + P(E2|M) = 1.
Technically, the fact that the semantic priming effect is context dependent decreases P(E2|M and T). For the sake of argument, however, I will ignore the technical connection between P(M|E2) and P(E2|M and T).
The supplementary motor area was active in Owen et al.’s patient for about 30s (Owen et al. 2006). This sort of evidence decreases P(M1|E2), where Mj pertains to variation in signal intensity. Nevertheless, for the sake of simplicity, I will ignore possible states of the signal.
E2 in P(M|E2) includes all the competing hypotheses, namely, \(\mathop \sum \limits_{{{\text{k}} = 1}}^{\text{n}} {\text{A}}_{k}\) (where Ak indicates different types of automated mental event given M). In this sense, P(M|E2) is the sum of probabilities of all competing hypotheses. We have some relevant evidence (cited in the text) that makes the possibility of the semantic priming effect unlikely, but no good evidence that it is likely. Nor have we other sorts of good competing hypotheses (that do not involve volition or top-down cognitive control) that explain tokening of M under the task circumstances (Shea and Bayne, op. cit. 466–469). So, assigning 0.05 to P(M|E2) is to some extent arbitrary, but is reasonable.
It is worth emphasizing that v(M) is basically about the changes in the probabilistic space. That is, v(M) is about the changes that the conditional probabilities make with respect to the unconditional probabilities. So, as P(E1) gets lower and P(E1|M) increases the more probable it is that the signal conveys the purported informational content.
In the Toe-Imagery task, there were 16 vegetative state patients and 12 controls. Suppose that 75% of controls follow the instruction (i.e., 9 out of 12), and 15% of the vegetative state group has been misdiagnosed and 50% of them follow the instruction (i.e., 2 out of 3). If so, P(E1) is roughly 0.4.
One way to make the signal significant is to increase P(E1|M) larger than P(E1). Naci et al. (2015, 2014) modeled neural correlates for movie-viewing by observing changes in the brain activity relevant to some specific events in a movie. Such a ‘narrative’ model captures the brain activity required for the executive demand of movie-viewing: e.g., when a subject is watching a certain plot, it is expected that certain areas are activated. Given that “synchronized activity fluctuations in the frontal and parietal regions tracked the common cognitive experience of different [healthy] individuals while watching the same movie,” the authors expected that the same activity will be observed if a vegetative state patient in the same task is in fact conscious (Naci et al. 2014). This narrative model captures conscious brain activity in the sense that the (diachronic) activation in some particular regions indicates that the subject is indeed watching the movie. This Movie-Viewing test is a method of ascribing a full-fledged conscious state by diachronically individuating a neural signal as pertaining to some diachronic activity in a network of areas. In other words, Movie-Viewing test increases P(E1|M) almost to 1 and decreases P(E2|M) almost to 0. But this has little to do with MCS or VT.
In the Tennis-Imagery task, the subject is asked to imagine playing a game of tennis and activity in the supplementary motor area is expected. In the Spatial-Navigation-Imagery task, the subject is asked to imagine walking around a familiar place like home and activity in areas relevant to memory function (e.g., the parahippocampal gyrus) is expected (Owen et al. 2006).
LogP(\({\text{E}}_{1}^{*}\)|M*)/P(\({\text{E}}_{1}^{*}\)) cannot express the fact that the accuracy rate increases the probability that the patient is mentally acting. A posterior distribution by the lights of one Bayesian update (e.g., a correct answer to Q1) may serve as prior to another update. Performing the experiment described in the text allows one to progressively update one’s probability that the subject performs the requested mental action. If the probability is updated 0.99, then the formal structure of v(M*) = 〈log P(\({\text{E}}_{1}^{*}\)|M*)/P(\({\text{E}}_{1}^{*}\)), log P(\({\text{E}}_{2}^{*}\)|M*)/P(\({\text{E}}_{2}^{*}\))〉 is of, roughly, v(M*) = 〈0, −3〉. Nevertheless, this doesn’t undermine the validity of the Q&A task because the prior with high probability by itself confirms that the subject is mentally acting. Unlike the case of log P(\({\text{E}}_{1}^{*}\)|M*)/P(\({\text{E}}_{1}^{*}\)), the post-signal probability, P(\({\text{E}}_{1}^{*}\)|M*), can sensibly reflect P(\({\text{E}}_{1}^{*}\)) with high probability. Thanks to the anonymous reviewer for Minds and Machines for clarifying how Bayesian update works.
References
Andrews, K., et al. (1996). Misdiagnosis of the vegetative state: Retrospective study in a rehabilitation unit. The British Medical Journal, 313, 13–16.
Ansell, B. J. (1993). Slow-to-recover patients: Improvement to rehabilitation readiness. The Journal of Head Trauma Rehabilitation, 8(3), 88–98.
Ansell, B. J. (1995). Visual tracking behavior in low functioning head-injured adults. Archives of Physical Medicine and Rehabilitation, 76(8), 726–731.
Armstrong, D. M. (1980). The nature of mind and other essays. Ithaca, NY: Cornell University Press.
Bargh, J. A., & Chartland, T. L. (1999). The unbearable automaticity of being. American Psychologist, 54(7), 467–479.
Cruse, D., et al. (2011). Bedside detection of awareness in the vegetative state: a cohort study. The Lancet, 378, 2088–2094.
Drayson, Z. (2014). Intentional action and the post-coma patient. Topoi, 33, 23–31.
Ehrsson, H. H., Geyer, S., & Naito, E. (2003). Imagery of voluntary movement of fingers, toes, and tongue activates corresponding body-part–specific motor representations. Journal of Neurophysiology, 90, 3304–3316.
Frith, C. D., Blakemore, D., & Wolpert, D. M. (2000). Abnormalities in the awareness and control of action. Philosophical Transactions of the Royal Society of London. Series B, 355, 1771–1788.
Giacino, J. T., & Kalmar, K. K. (1997). The vegetative and minimally conscious states: A comparison of clinical features and functional outcome. The Journal of Head Trauma Rehabilitation, 12(4), 36–51.
Giacino, J. T., Kalmar, K., & Whyte, J. (2004). The JFK coma recovery scale-revised: Measurement characteristics and diagnostic utility. Archives of Physical Medicine and Rehabilitation, 85(12), 2020–2029.
Giacino, J. T., et al. (2002). The minimally conscious state: Definition and diagnostic criteria. American Academy of Neurology, 58, 349–353.
Godfrey-Smith, P. (2012). Review of brian Skyrms’s signals. Mind, 120, 1288–1297.
Grafton, S. T., et al. (1991). Somatotopic mapping of the primary motor cortex in humans: Activation studies with cerebral blood flow and positron emission tomography. Journal of Neurophysiology, 66(3), 735–743.
Haibo, D., et al. (2008). Neuroimaging activation studies in the vegetative state: predictors of recovery? Clinical Medicine, 8(5), 502–507.
Harms, W. F. (2006). What is information? Three concepts. Biological Theory, 1(3), 230–242.
Hohwy, J. (2009). The neural correlates of consciousness: New experimental approaches needed? Consciousness and Cognition, 18, 428–438.
Isaac, A. M. C. (2010). The Informational Content of Perceptual Experience. Unpublished dissertation.
Isaac, A. M. C. (forthcoming). The Semantic Latent in Shannon Information. The British Journal for the Philosophy of Science.
Laureys, S., et al. (2000). Restoration of thalamocortical connectivity after recovery from persistent vegetative state. The Lancet, 335, 1790–1791.
Monti, M. M., Laureys, S., & Owen, A. M. (2010a). The vegetative state. The British Medical Journal, 341, 292–296.
Monti, M. M., et al. (2010b). Willful modulation of brain activity in disorders of consciousness. The New England Journal of Medicine, 362(7), 579–589.
Naci, L., Sinai, L., & Owen, A. M. (2015). Detecting and interpreting conscious experience in behaviorally non-responsive patients. NeuroImage. http://dx.doi.org/10/1016/j.neuroimage.2015.11.059.
Naci, L., et al. (2014). A common neural code for similar conscious experiences in different individuals. PNAS, 111(39), 14277–14282.
Neuper, C., et al. (2005). Imagery of motor actions: Differential effects of kinesthetic and visual–motor mode of imagery in single-trial EEG. Cognitive Brain Research, 25, 668–677.
Overgaard, M. (2015). Behavioral Methods in Consciousness Research. Oxford: Oxford University Press.
Owen, A. M. (2013). Detecting consciousness: A unique role for neuroimaging. Annual Review of Psychology, 64, 109–133.
Owen, A. M., et al. (2006). Detecting awareness in the vegetative state. Science, 313, 1402.
Owen, A. M., et al. (2007). Response to comments on “detecting awareness in the vegetative state”. Science, 315(5816), 1221.
Pattamadilok, C., et al. (2017). Automaticity of phonological and semantic processing during visual word recognition. NeuroImage, 149, 244–255.
Pfurtscheller, G., & Neuper, C. (1997). Motor imagery activates primary sensorimotor area in humans. Neuroscience Letters, 239, 65–68.
Poldrack, R. A. (2006). Can cognitive processes be inferred from neuroimaging data? Trends in Cognitive Sciences, 10(2), 59–63.
Porro, C. A., et al. (1996). Primary motor and sensory cortex activation during motor performance and motor imagery: A functional magnetic resonance imaging study. The Journal of Neuroscience, 16(23), 7688–7698.
Pulvermüller, F. (2005). Brain mechanisms linking language and action. Nature Reviews Neuroscience, 6, 576–582.
Raposo, A., et al. (2009). Modulation of motor and premotor cortices by actions, action words and action sentences. Neuropsycologia, 47, 388–396.
Schnakers, C., et al. (2009). Diagnostic accuracy of the vegetative and minimally conscious state: Clinical consensus versus standardized neurobehavioral assessment. BMC Neurology, 9, 35.
Shannon, C. (1948). A mathematical theory of communication. The Bell System Technical Journal, 27(379–423), 623–656.
Shea, N., & Bayne, T. (2010). The vegetative state and the science of consciousness. The British Journal for the Philosophy of Science, 61, 459–484.
Shea, N., Godfrey-Smith, P., & Cao, R. (2017). Content in Simple Signalling Systems. The British Journal for the Philosophy of Science, in press.
Shewmon, D. A. (2004). THE ABC OF PVS: Problems of definition. In C. Machado & D. A. Shewmon (Eds.), Brain death and disorders of consciousness. New York: Kluwer Academic/Plenum Publishers.
Skyrms, B. (2010). Signals: Evolution, learning, and information. Oxford: Oxford University Press.
Stender, J., et al. (2016). The minimal energetic requirement of sustained awareness after brain injury. Current Biology, 26, 1494–1499.
Stins, J. F. (2009). Establishing consciousness in non-communicative patients: A modern-day version of the Turing test. Consciousness and Cognition, 18, 187–192.
Tsuchiya, N., et al. (2015). No-report paradigms: Extracting the true neural correlates of consciousness. Trends in Cognitive Sciences, 19(12), 757–770.
Wegner, D. M., & Wheatley, T. (1999). Apparent mental causation: Sources of the experience of will. American Psychologist, 54(7), 480–492.
Acknowledgements
I am indebted to Carrie Figdor for many fruitful discussions and her invaluable comments made on earlier drafts. I am also grateful to Gregory Landini, Richard Fumerton, as well as two anonymous reviewers of this journal for their well-placed suggestions on previous versions. Many thanks to Seung Wook Kim, David Redmond, and members of the University of Iowa Graduate Philosophical Society for helpful discussions.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Noh, H. No-report Paradigmatic Ascription of the Minimally Conscious State: Neural Signals as a Communicative Means for Operational Diagnostic Criteria. Minds & Machines 28, 173–189 (2018). https://doi.org/10.1007/s11023-017-9433-6
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11023-017-9433-6