Human and Animal Minds: The Consciousness Questions Laid to Rest

The global workspace theory, the phenomenal concept strategy, and the distribution of consciousness.

The Free Energy Principle and Active Inference Framework (FEP-AI) begins with the understanding that persisting systems must regulate environmental exchanges and prevent entropic accumulation. In FEP-AI, minds and brains are predictive controllers for autonomous systems, where action-driven perception is realized as probabilistic inference. Integrated Information Theory (IIT) begins with considering the preconditions for a system to intrinsically exist, as well as axioms regarding the nature of consciousness. IIT has produced controversy because of its surprising entailments: quasi-panpsychism; subjectivity without referents or dynamics; and the possibility of fully-intelligent-yet-unconscious brain simulations. Here, I describe how these controversies might be resolved by integrating IIT with FEP-AI, where integrated information only entails consciousness for systems with perspectival reference frames capable of generating models with spatial, temporal, and causal coherence for self and world. Without that connection with external reality, systems could have arbitrarily high amounts of integrated information, but nonetheless would not entail subjective experience. I further describe how an integration of these frameworks may contribute to their evolution as unified systems theories and models of emergent causation. Then, inspired by both Global Neuronal Workspace Theory (GNWT) and the Harmonic Brain Modes framework, I describe how streams of consciousness may emerge as an evolving generation of sensorimotor predictions, with the precise composition of experiences depending on the integration abilities of synchronous complexes as self-organizing harmonic modes (SOHMs). These integrating dynamics may be particularly likely to occur via richly connected subnetworks affording body-centric sources of phenomenal binding and executive control. Along these connectivity backbones, SOHMs are proposed to implement turbo coding via loopy message-passing over predictive (autoencoding) networks, thus generating maximum a posteriori estimates as coherent vectors governing neural evolution, with alpha frequencies generating basic awareness, and cross-frequency phase-coupling within theta frequencies for access consciousness and volitional control. These dynamic cores of integrated information also function as global workspaces, centered on posterior cortices, but capable of being entrained with frontal cortices and interoceptive hierarchies, thus affording agentic causation. Integrated World Modeling Theory (IWMT) represents a synthetic approach to understanding minds that reveals compatibility between leading theories of consciousness, thus enabling inferential synergy.

How do airborne plumes of molecules docking on olfactory receptors emerge as conscious odors in the brain? How are they interpreted in space, time, and biological meaning? And how do they lead to fast and adaptive decisions and actions? In general, conscious (reportable) perception supports neural adaptation to novelty, judgments of self-relevance, and voluntary decision-making. Conscious processes have a number of established properties that are markedly different from unconscious ones (Seth et al., 2005). A growing experimental literature has explored conscious perception with a wide array of recording techniques. Similarly, conscious olfaction “as such” can be studied by comparing novel vs. habituated odors, attended vs. unattended ones, and rivaling olfactory percepts, comparable to visual rivalry (Stevenson and Mahmut, 2013). State comparisons of odor processing during waking vs. sleep, general anesthesia, and impaired consciousness are also important. Neural activity underlying conscious percepts should follow the known psychophysical features of the stimulus. A specific conscious odor should correspond to a specific trajectory in the olfactory perceptual space derived from psychophysical stimulus matching and discrimination (e.g., Berglund et al., 1973; Koulakov et al., 2011). Global workspace theory (GWT) has been used as a framework for experimental studies of conscious brain processes for more than two decades, leading to a family of related models and experimental predictions. GWT begins by analyzing a reliable set of properties of conscious events (Baars, 1988, 2002). For example, while conscious perception shows limited momentary capacity, it also supports access to non-conscious functions, like memory, executive control, and automatic skills (Baars and Franklin, 2003). By comparison, unconscious stimuli do not afford such very widespread access to unconscious brain capacities. In general terms, a global workspace (GW) is a functional hub of binding and propagation in a population of loosely coupled signaling elements, such as neurons (Izhikevich, 1999). A GW is commonly compared to the stage of a theater, or a playing field in a large football arena, allowing many specialized knowledge sources to compete and cooperate to resolve focal problems. GW architectures are useful to resolve ambiguous and novel stimuli, such as words in natural language. Conscious percepts often result from a process of ambiguity reduction, and GW architectures have therefore been proposed as models of conscious perception. They are also consistent with highly interactive information flow in the cortico-thalamic (C-T) system (Baars et al., 2013). Edelman et al. (2011) suggest that GW theory is consistent with Neural Darwinism and its many ramifications. GWT predicted widespread “broadcasting” of conscious events, a prediction that is now widely accepted. In a recent study of visual rivalry in the macaque, content-specific “global broadcasting” from temporal to lateral prefrontal cortex was observed for both oscillatory population signaling and multi-unit recordings (Panagiotaropoulos et al., 2012). Similarly, long-distance cortical phase-linking is associated with the waking state but not slow-wave sleep (see Baars et al., 2013 for a review). In general, conscious sensory input has been repeatedly found to evoke more widespread, high amplitude, and phase-linked oscillations in cortex. Baars et al. (2013) have proposed that neuronal source coalitions may emerge anywhere in cortex, becoming subjectively conscious and reportable when a convergent winner-take-all source coalition comes to a momentary equilibrium, able to drive many other regions. During the waking state the visual cortex shows reentrant signaling among more than 40 visuotopical maps (Steriade, 2006). In vision the occipito-temporal cortex identifies the perceptual features that emerge in consciousness, from high-resolution visual details to lower-resolution object and event representation (IT and MTL). For the sight of a visual coffee cup or a flower garden, input convergence is believed to occur at high levels of the visual hierarchy, including object perception in area IT and event perception in MTL. However, a simple stimulus, like the sight of a single star on a dark night, might equilibrate early in the visual hierarchy, since areas V1 and LGN have the highest spatial resolution. This highly flexible version of GWT in the C-T system has been called Dynamic GWT (dGWT). Thus, visual cortex may integrate visual gestalts and broadcast them to frontoparietal, anterior temporal, and subcortical regions. In contrast, non-sensory “feelings of knowing” (FOKs), including expectations and intentions, may arise and propagate from frontal and anterior-temporal regions to caudal sites (Cole et al., 2010; Baars et al., 2013). Direct brain recordings in human patients show widespread neocortical signaling by way of cross-frequency phase-linking among cortical arrays, especially theta-gamma and alpha-gamma signaling. Single neurons have been shown to phase-adapt to dominant theta oscillations, suggesting a mechanism by which individual neurons may be recruited by population oscillations. Such spatiotemporal coding allows for an extremely rich signaling vocabulary, but specific coding schemes are just beginning to be understood (see Baars et al., 2013 for a review).

An extended theory of global workspace of consciousness

The study of consciousness is gaining importance in both neuroscience and the development of Artificial Intelligence (AI). We show here that an advanced White Matter (WM) tractography method, termed gridography, can explore the potential integration of two prominent theories of consciousness: Global Workspace Theory (GWT) and Integrated Information Theory (IIT). Using gridography on high-resolution diffusion MRI data from the Human Connectome Project, we demonstrate that gridography obtains WM connections between the anterior brain regions associated with GWT and posterior regions linked to IIT in a form which agrees with the Epiontic Consciousness Theory (ECT), which is an intermediary theory between GWT and IIT. We evaluate how experimental gridography data aligns with the physiological structures implicated in consciousness by analyzing: (i) Information characteristics of consciousness theories; (ii) Improvement of diffusion MRI tractography by use of gridography; (iii) Expected gridography results based on consciousness theory. Our findings suggest that these connections, particularly those of the Superior Longitudinal Fasciculus (SLF), support a ECT unified model of consciousness integrating aspects of both the primarily epistemic GWT and the primarily ontic IIT. This study proposes a novel framework that could reconcile existing theoretical divisions between GWT and IIT through the use of the ECT approach.

In his paper ‘Are we ever aware of concepts? A critical question for the Global Neuronal Workspace, Integrated Information, and Attended Intermediate-Level Representation theories of consciousness’ (2015, this journal), Kemmerer defends a conservative account of consciousness, according to which concepts and thoughts do not characterize the contents of consciousness, and then uses that account to argue against both the Global Neuronal Workspace theory of consciousness and Integrated Information Theory of Consciousness, and as a point in favour of Prinz’s Attended Intermediate-level Representations theory. We argue that there are a number of respects in which the contrast between conservative and liberal conceptions of the admissible contents of consciousness is more complex than Kemmerer’s discussion suggests. We then consider Kemmerer’s case for conservatism, arguing that it lumbers liberals with commitments that they need not – and in our view should not – endorse. We also argue that Kemmerer’s attempt to use his case for conservatism against the Global Neuronal Workspace and Integrated Information theories of consciousness on the one hand, and as a point in favour of Prinz’s Attended Intermediate Representations theory on the other hand, is problematic. Finally, we consider Kemmerer’s overall strategy of using an account of the admissible contents of consciousness to evaluate theories of consciousness, and suggest that here too there are complications that Kemmerer’s discussion overlooks.

Different theories explain how subjective experience arises from brain activity1,2. These theories have independently accrued evidence, but have not been directly compared3. Here we present an open science adversarial collaboration directly juxtaposing integrated information theory (IIT)4,5 and global neuronal workspace theory (GNWT)6–10 via a theory-neutral consortium11–13. The theory proponents and the consortium developed and preregistered the experimental design, divergent predictions, expected outcomes and interpretation thereof12. Human participants (n = 256) viewed suprathreshold stimuli for variable durations while neural activity was measured with functional magnetic resonance imaging, magnetoencephalography and intracranial electroencephalography. We found information about conscious content in visual, ventrotemporal and inferior frontal cortex, with sustained responses in occipital and lateral temporal cortex reflecting stimulus duration, and content-specific synchronization between frontal and early visual areas. These results align with some predictions of IIT and GNWT, while substantially challenging key tenets of both theories. For IIT, a lack of sustained synchronization within the posterior cortex contradicts the claim that network connectivity specifies consciousness. GNWT is challenged by the general lack of ignition at stimulus offset and limited representation of certain conscious dimensions in the prefrontal cortex. These challenges extend to other theories of consciousness that share some of the predictions tested here14–17. Beyond challenging the theories, we present an alternative approach to advance cognitive neuroscience through principled, theory-driven, collaborative research and highlight the need for a quantitative framework for systematic theory testing and building.

Tools and tests for measuring the presence and type of consciousness are becoming available, but there is no established theoretical approach for what these tools are measuring. This paper looks at various categories of tests for measuring the presence and type of consciousness and suggests ways in which different theories of consciousness may be empirically distinguished. We label the various testable correlates of consciousness as the "measurable correlates of consciousness" (MCC). There are three sub-categories of MCC: 1) Neural correlates of consciousness (NCC); 2) Behavioral correlates of consciousness (BCC); 3) Creative correlates of consciousness (CCC). We also look specifically at ways in which the General Resonance Theory of consciousness may be tested and compared to other theories like the Integrated Information Theory of consciousness and Global Workspace Theory. We suggest additional simplified approaches under the hypothesis that electrical and magnetic fields are the seat of consciousness. Last, we reflect on how broader philosophical views about the nature of consciousness, such as materialism and panpsychism, may also become scientifically tractable.IntroductionHow can we know if any person, animal or any thing is actually conscious and not just simulating various aspects of consciousness? The nature of consciousness makes it by necessity a wholly private affair (Libet 2005; Koch 2019). The only consciousness I can know with certainty is my own. Everything else is inference.How do we create a reliable “consciousness-ometer” (what I’ll call apsychometer in the rest of this paper)? This inquiry has been relegated to philosophical musings until the last few years, but we are at a juncture where tools for measuring consciousness are starting to mature. This paper looks at the various kinds of tools and tests available, how they can be used to test for the presence and type of consciousness, and makes some suggestions for how a reliable psychometer could be created and refined over time.Theories of consciousness are abundant, but often untested or even untestable (Michel et al. 2019). A major coordinated testing program has yet to be conducted, but the Templeton World Charity Foundation embarked in 2019 [Fn 1] on a multi-year effort to examine a number of the more prominent theories of consciousness in a series of one-on-one adversarial experimental tests, with the express intent of distinguishing the various theories. The first head-to-head contest will feature Global Neuronal Workspace theory (Dehaene 2014) and the Integrated Information Theory of consciousness (Oizumi, et al. 2013).Footnote 1. Limited details are available at Templeton’s website here: https://www.templetonworldcharity.org/arc. Additional details on the program and approach are available here: https://www.quantamagazine.org/neuroscience-readies-for-a-showdown-over-consciousness-ideas-20190306/. Additional details were released at an October 2019 announcement: https://sci-hub.tw/https://science.sciencemag.org/content/366/6463/293.full. In thinking about ways to test theories of consciousness, it is important to keep in mind at all times that we can’t know if any person, any animal, or anything else at all is actually conscious, rather than a sophisticated simulation of consciousness. We can and frequently do in practice, nevertheless, make reasonable inferences about the presence of other consciousnesses. Libet 2005 agrees: “[S]ubjective experience cannot be directly measured by external objective devices or by external observations. Conscious subjective experience is accessible only to the individual having the experience.”Attempts to assess the presence or nature of consciousness in any particular circumstance, and related attempts to assess different theories of consciousness and their predictions, will face the problem of reasonable inference (abduction) because of this fundamental limitation on our individual and collective knowledge. But this problem is surmounted frequently in practice in that we, each of us, reasonably infer that other people are conscious, based on their behavior and appearance. The same holds true for pets and many other animals. Testing for the presence of consciousness throughout the physical world relies on making similar reasonable inferences.Koch 2019 (p. 155) makes a similar argument: “Because you are so similar to me, I abduce that you too have subjective, phenomenal states. The same logic applies to other people. Apart from the occasional solitary solipsist this is uncontroversial.” Koch proceeds through the course of his book to offer various ways that scientists may, now and in the future, test for the presence and character of consciousness in humans, animals and even non-biological entities – all based on abduction (reasonable inference).We propose in the present paper a general quantification framework that rests on various “measurable correlates of consciousness” (MCC). This rubric includes the “neural correlates of consciousness” and the related but broader notion of “behavioral correlates of consciousness.” It also includes a newly-coined “creative correlates of consciousness” (CCC) category that is explained below. MCC refers to any means identified for measuring aspects of consciousness.This paper identifies various ways in which MCC can be identified and tested. We also suggest ways for testing and contrasting specific theories of consciousness, including the General Resonance Theory (GRT) of consciousness that has been developed by Hunt and Schooler over the last decade. We also argue that the various metaphysical positions with respect to the nature of consciousness may, contrary to widespread opinion on this subject, be tested.These questions are more than philosophical. With the coming age of intelligent digital assistants, self-driving cars, and other robots serving us and increasingly running our lives, does it matter if these AIs are actually conscious or just simulating consciousness?More relevant for today’s needs, how can we know that coma victims, or patients in vegetative or minimally conscious states, are conscious or not? Or if they are likely to recover? How can a family know whether to take a patient off life support or not, if they don’t know with any certainty what kind of consciousness is or is not present, or is likely to re-enter over time?The Measurable Correlates of ConsciousnessThere is a small but growing field looking at how to assess the presence and even quantity of consciousness in various entities. I’ve divided possible tests into three broad categories that comprise collectively what I call the “measurable correlates of consciousness” (MCC) (Fig. 1). The MCC represent all possible scientific measures for inferring the presence of consciousness. They are “correlates” because we can’t know with certainty, as discussed above, whether consciousness is actually present. We can only infer, based on our measurements and best judgments. But they, nevertheless, “measurable,” and this term is meant to capture both of these key features.

Different theories of consciousness describe different mechanisms of consciousness. Mechanisms can be used to identify scientific kinds, which offers another way of evaluate what the targets of consciousness science really are. Global neuronal workspace theory and local recurrent processing are used as case studies of mechanisms of consciousness. By referring to criteria of mechanistic demarcation, it is argued that the mechanism described in global workspace theories is inadequately demarcated, and that ‘reportability’ is not in fact a single phenomenon as is often assumed. Second, the mechanism of local recurrent processing would generate far more conscious content than currently acknowledged, so seems fundamentally implausible. Instead of describing mechanisms of consciousness, it is suggested that both accounts are best described as identifying selected parts of sensory and cognitive processes that are well-documented elsewhere. In this case, ‘consciousness’ does not seem to refer to a scientific kind, and so is unlikely to be a viable target for scientific investigation.

A global workspace model for phenomenal and access consciousness

We present a functional magnetic resonance (fMRI) dataset collected as part of an adversarial collaboration aimed at arbitrating between the Global Neuronal Workspace theory (GNWT) and the Integrated Information Theory (IIT) of consciousness. Participants (N = 118) were presented with suprathreshold visual stimuli belonging to four different categories (faces, objects, letters, false fonts) with three orientations (front, left, right view), and three durations (0.5, 1.0, 1.5 seconds). Participants were asked to identify infrequent targets that changed in each block, thereby rendering two categories task-relevant and two task-irrelevant. The simplicity of the experimental design and of the task given to the participants ensures that these data are broadly reusable. Besides testing predictions from other theories of consciousness, these data can be used to examine various aspects of visual processing. The anonymized data were converted to Brain Imaging Data Structure (BIDS), and can be easily accessed through a web platform or an API. The dataset contains quality reports, demographics, behavioral performance, and eye-tracking data. We also provide code for preprocessing and analyzing the data.

The study of consciousness is gaining importance in both neuroscience and the development of Artificial Intelligence (AI). We show here that an advanced White Matter (WM) tractography method, termed gridography, can explore the potential integration of two prominent theories of consciousness: Global Workspace Theory (GWT) and Integrated Information Theory (IIT). Using gridography on high-resolution diffusion MRI data from the Human Connectome Project, we demonstrate that gridography can accurately map WM connections between the anterior brain regions associated with GWT and posterior regions linked to IIT. We evaluate how experimental gridography data aligns with the physiological structures implicated in consciousness by analyzing: i. the current status of consciousness-related brain locations; ii. structures necessary for transmitting consciousness-related information; iii. characteristics necessary for tractography to identify consciousness-related WM connections. Our findings suggest that these connections, particularly those of the Superior Longitudinal Fasciculus (SLF), support a unified model of consciousness integrating aspects of both the primarily epistemic GWT and the primarily ontic IIT. This study proposes a novel framework that could reconcile existing theoretical divisions between GWT and IIT.

Recent years have seen a blossoming of theories about the biological and physical basis of consciousness. Good theories guide empirical research, allowing us to interpret data, develop new experimental techniques and expand our capacity to manipulate the phenomenon of interest. Indeed, it is only when couched in terms of a theory that empirical discoveries can ultimately deliver a satisfying understanding of a phenomenon. However, in the case of consciousness, it is unclear how current theories relate to each other, or whether they can be empirically distinguished. To clarify this complicated landscape, we review four prominent theoretical approaches to consciousness: higher-order theories, global workspace theories, re-entry and predictive processing theories and integrated information theory. We describe the key characteristics of each approach by identifying which aspects of consciousness they propose to explain, what their neurobiological commitments are and what empirical data are adduced in their support. We consider how some prominent empirical debates might distinguish among these theories, and we outline three ways in which theories need to be developed to deliver a mature regimen of theory-testing in the neuroscience of consciousness. There are good reasons to think that the iterative development, testing and comparison of theories of consciousness will lead to a deeper understanding of this most profound of mysteries.

To locate consciousness in the flow of synaptic activity in the brain, we must first locate it in the flow of information processing in the mind. Two different positions have been debated for centuries. The liberal view maintains that the contents of experience include not only sensory, motor, and affective states, but also concepts and the thoughts they enter into. In contrast, the conservative view maintains that concepts have no intrinsic qualia of their own, and that the contents of experience are therefore restricted to sensory, motor, and affective states. Here I discuss how this long-standing controversy is relevant to several contemporary neuroscientific theories of consciousness. I do so, however, in a manner that is admittedly biased toward the conservative view, since I am among those who believe that it is more consistent than the liberal view with a number of key findings. I focus first on two of the most prominent neuroscientific theories of consciousness—namely, Stanislas Dehaene's Global Neuronal Workspace Theory and Giulio Tononi's Integrated Information Theory. I argue that because both of these approaches assume the liberal view, they are challenged in significant ways by data favoring the competing conservative view. I then turn to a third framework—namely, Jesse Prinz's Attended Intermediate-Level Representation Theory. I contend that because it explicitly endorses the conservative view, it has a unique advantage over the other two approaches. I also point out, however, that it has independent shortcomings that prevent it from achieving adequate explanatory coherence. I conclude by emphasizing that, if the conservative view is in fact correct, a central goal of future research should be to distinguish, at both psychological and neurobiological levels of analysis, between the following two kinds of information processing that often occur simultaneously: first, activation of the modality-specific sensory, motor, and affective representations that constitute the sole ingredients of conscious experiences; and second, activation of the conceptual representations that give those experiences meaning and that may even influence them in a top-down manner, but that never themselves reach awareness.

Identifying the neural correlates of consciousness (NCC) is widely regarded as imperative to our understanding of consciousness. Although several scientific theories of consciousness make predictions about the NCC, they often differ in their level of detail, making direct comparison through experimental findings difficult. Here, we report an adversarial collaboration protocol in which unbiased experimentalists will test predictions formed by theory proponents of Global Neuronal Workspace Theory (GNWT) and Integrated Information Theory (IIT), concerning the location, timing and functional connectivity of the NCC. To enable cross-species comparisons, non-human primates (NHPs) and mice will view/listen to supra-threshold visual/auditory stimuli for variable durations in a go-nogo task that controls for report confounds, while we record with Neuropixels electrodes neuronal responses from visual, auditory, posterior parietal and/or prefrontal cortical areas. To causally test predictions about the timing and location, we will manipulate activity in prefrontal cortical regions using electrical stimulation in NHPs or optogenetic silencing in mice. This Study Protocol details the experimental design, analyses, divergent predictions, and the anticipated outcomes, along with their interpretation.