I read this paper in late 2020 as part of my ongoing study of our neuroscientific understanding of consciousness. These are the notes I took.

The full title of the paper is Homing in on consciousness in the nervous system: An action-based synthesis, by Morsella et al, and it can be accessed for free here.

  • Thesis: passive frame theory of consciousness
    • Passive frame theory proposes that the primary function of consciousness is well circumscribed, serving the somatic nervous system. For this system, it serves as a frame that constrains and directs skeletal muscle output, thereby yielding adaptive behavior. The mechanism by which consciousness achieves this is more counterintuitive, passive, and “low level” than the kinds of functions that theorists have attributed to consciousness. (2)

  • 1. Terms and assumptions

    • focusing on the most basic form of consciousness
      • We believe that, to advance the study of consciousness, one should focus not on high forms of consciousness (e.g., “self-consciousness”), but on the most basic forms of consciousness (e.g., the experience of a smell, visual afterimages, tooth pain, or urges to scratch an itch). This form of consciousness has fallen under the rubrics of “sentience” (Pinker 1997), “primary consciousness” (Edelman 1989), “phenomenal consciousness” (Block 1995b), “qualia” (J. A. Gray 2004), “phenomenal states” (Tye 1999), and “subjective experience.” In our framework, we refer to a thing of which one is conscious (e.g., an afterimage) as a conscious content (Merker 2007; Seth 2007). All of the contents of which one is conscious at one time can be construed as composing the conscious field (Freeman 2004; Köhler 1947; Searle 2000). The contents of the conscious field change over time.

    • descriptive rather than normative (describing how consciousness evolved rather than how it should function)
      • from a descriptive standpoint, the distinction between conscious and unconscious processes is apparent in many fields:
        • Thus, in every field of inquiry, there is the de facto distinction between the two kinds of processes, though often without mention of the taboo term “consciousness.” For example, in perception research, there exists the distinction between supra- versus subliminal. In memory research, there is the distinction between “declarative” (explicit) processes and “procedural” (implicit) processes (Schacter 1996; Squire 1987). In motor and language research, the conscious aspects of voluntary action or of speech production are contrasted with the unconscious aspects of, say, motor programming (Levelt 1989; Rosenbaum 2002; J. A. Taylor & Ivry 2013). Various fields also contrast “controlled” processing (which tends to be conscious) and “automatic” processing (which is often unconscious; Lieberman 2007). In summary, from a descriptive approach, the contrast between conscious and unconscious processes in the brain is somewhat inevitable (Morsella & Poehlman 2013).

    • minimalistic, focusing on simple cases
      • hypothetical organism: human-like mammal that is occupied with basic operations like breathing, locomoting, avoiding tissue damage
        • the organism can experience approach-avoidance conflicts but is incapable of indirect cognitive control (e.g. stimulating hunger or fear by deliberately imagining the things that would induce those states)
        • organism is inside a cave, sees an opening, is warm and comfortable in the cave (Stage 1), but notices a disturbing smell (Stage 2).
      • this focus on low-level functions of consciousness contrasts with many theorists' views that the function of consciousness pertains to high-level activities such as social interaction, language, formation of the self, “theory of mind”, simulations of behavior.
      • caveat: some have argued that consciousness does not contribute to action. (Hommel 2013; Masicampo & Baumeister 2013).
    • action-based theory
      • Thus, from our untraditional, action-based approach, we subscribe to an uncommon theoretical position – that the nature of consciousness is best understood by examining the requirements of adaptive (efferent) action control rather than the needs of perceptual analysis.

    • focusing on olfactory consciousness rather than vision or auditory. vision is deeply interlinked with higher-level processing like conceptualization and semantics. also, with vision and hearing, the individual can self-generate experiences, but with smell this is limited
    • EASE framework: elemental, action-based, simple, evolutionary-based
  • 2. The role of consciousness in the nervous system

    • **Tenet 1: ** the subset consensus that consciousness is associated with only a subset of all processes and regions of the nervous system, and this subset is qualitatively distinct from the unconscious regions of the brain (in terms of function & physical organization)
      • also, even some complex nervous functions are unconscious
        • Evidence for the complexity of unconscious processing is found in cases in which the entire stimulus-response arc is mediated unconsciously, as in the case of unconsciously mediated actions. (4)

          • there are even conditions in which actions are decoupled from consciousness, e.g. alien hand syndrome, anarchic hand syndrome, utilization behavior syndrome (16)
        • this raises the question, what does consciousness contribute?
    • Tenet 2: the integration consensus that the conscious field integrates various forms of information processing that would otherwise be independent
      • conscious information is available “globally”
      • there is some evidence that “the conscious processing of a percept involves a wider and more diverse network of regions than does the subliminal (unconscious) processing of the same percept” (4)
    • understanding the physiology of consciousness requires understanding its neuroanatomy
      • cortical-subcortical controversy – is consciousness associated with the cerebral cortex or subcortical structures? evidence from lobotomies and the psychophysiology of dreams suggests cortical lobes may not be necessary for consciousness
      • authors propose that progress can be made in this question by examining the role of consciousness in nervous function
    • Tenet 3: the function of consciousness is as a specific kind of integration involving musculoskeletal output
      • not all integration requires consciousness
        • afference binding is integration across sensory modalities (e.g. McGurk effect) and is generally unconscious
        • efference binding is integration between stimulus and muscle activity; it can also be unconscious, e.g. pressing a button in response to a subliminal stimulus
      • however some integrations are very much conscious, e.g. holding breath while underwater, or experiencing an approach-avoidance conflict. these often involve action selection rather than perceptual processing
        • action selection should also be distinguished from motor control/motor programming, which are largely unconscious
      • In short, conflicts at the stage of processing of action selection are experienced consciously, whereas conflicts at perceptual stages of processing are unconscious. It has been proposed that, unlike unconscious integrations, these integrations involve competition for control of the skeletal muscle (“skeletomotor,” for short) output system (Morsella 2005). The skeletomotor output system contains the unconscious motor plans that are necessary to enact one skeletomotor act versus another (Bizzi & Mussa- Ivaldi 2004; Rizzolatti et al. 2004; Rosenbaum 2002). It stores, for example, the unconscious articulatory plans that are necessary for speech production (Buchsbaum 2013) and the plans for blinking (Graziano 2008). When these plans are stimulated sufficiently, overt actions arise. (5)

      • the conscious field is necessary to integrate what appear to be multiple inclinations toward the skeletomotor output system, as captured by the principle of Parallel Responses into Skeletal Muscle (PRISM). (6)

    • Tenet 4: the conscious field is for adaptive voluntary action
      • Just as a prism combines different colors to yield a single hue, the conscious field permits for multiple response tendencies to yield a single, integrated action. Absent consciousness, skeletomotor behavior can be influenced by only one of the efference streams, leading to unintegrated actions. (6)

      • only information in the field of consciousness can in turn affect voluntary action
        • in voluntary action, when the appropriate contents are absent, there is no independent system or repository of knowledge that can step in to fill their role. Thus, the conscious field wholly and exclusively determines what in everyday life is called voluntary behavior. Conversely, for every voluntary action, the organism can report a conscious content responsible for that action, regardless of the veracity of the introspection. (6)

      • this relates to why skeletal muscle is thought of as “voluntary” muscle
        • Since at least the nineteenth century, it has been known that, though often functioning unconsciously (as in the frequent actions of breathing and blinking), skeletal muscle is the only bodily effector that can be consciously controlled, but why this is so has never been addressed theoretically. PRISM introduces a systematic reinterpretation of this age-old fact (Morsella 2005): Skeletomotor actions are at times “consciously mediated” because they are directed by multiple systems that require consciousness to influence action collectively – what we refer to as collective influence. (6)

      • all systems that seek to influence skeletomotor output must “go through it”. each system has its own operations and evolutionary origins. a given muscle can only perform one action at a time (e.g. uttering one word), so the inclinations from various systems need to be “understood” and processed collectively. for actions to be adaptive, information from other sources (e.g. the physical environment) must also be integrated.
        • To a degree greater than that of any other effector system (e.g., smooth muscle), distinct regions/systems of the brain are trying to control the skeletomotor output system in different and often opposing ways. All inclinations toward it, from primitive plans about basic needs to complex plans associated with language, must engage this system. Thus, the skeletomotor output system is the “final common path” for processes capable of influencing skeletomotor function. (6)

  • 3. Conscious contents

    • the role of consciousness as guiding adaptive voluntary action suggests that the contents of consciousness are those things that render voluntary action adaptive.
      • they adopt the assumption that consciousness cannot be “content-free”, but this contradicts the idea that a content can enter the conscious field (for if content can “enter” the field, there must be a separate “container” in addition to the content). to avoid contradictory terminology, they state that contents do not enter consciousness but rather become conscious. there remains the question of whether contents “constitute the field or modulate it” (footnote n.10, p 16-17). see Searle, Consciousness, 2000
    • Tenet 1: conscious contents are “perceptual-like” in nature (sensorium hypothesis)
      • will use the term percept for conscious contents, acknowledging it can be misleading
        • we acknowledge that conscious contents are neither purely sensorial nor purely motor related; instead, they represent well-crafted representations occurring at a stage of processing between sensory analysis and motor programming (6)

      • proposal that contents are “perceptual-like” based on several areas of study
      • sensorium hypothesis: conscious contents are primarily influenced by perceptual (rather than motor) events and processes; motor processes are largely unconscious
        • During action, for example, one is unconscious of the efference to the muscles that dictates which fibers should be activated at which time (Rosenbaum 2002). Although one is unconscious of these complex programs (Johnson & Haggard 2005), one is often aware of their proprioceptive and perceptual consequences (e.g. perceiving the hand grasping). (7)

        • language is an example of how “mechanisms of action production can be complex but unconscious”, as in syntax
      • there’s an isomorphism in conscious content between acting (saying “hello”), dreaming (saying “hello” in a dream), and observing the act of another (hearing “hello”) → in every case, it is the perceptual dimension of the experience that is available to consciousness
      • 3.1.1. it has also been proposed that perceptual-like contents are the only kind of content that are understandable by multiple brain systems, i.e. are the most “broadcastable”. they are the “lingua franca” of action systems.
        • reasons for this are out of scope but they reference a few papers/books. related to “isotropic information”
    • Tenet 2: conscious contents can directly activate actions in the skeletal muscle output system
      • ideomotor theory: “the mental image of the (perceptual-like) action effects (in the body or in the world) of an instrumental action leads to the execution of that action, with the motor programming involved being unconscious” (8)
        • From this standpoint, the urge to move the arm leftward is isomorphic to the perceptual consequences of what would be observed if the act were performed. (8)

        • the Stroop test is evidence of the link between action-related perceptual processes and unconscious action systems (unsure why)
    • Tenet 3: action selection is the result of action-related representations competing with each other
      • consciousness allows for conflicts between action representations, but it’s not the mechanism by which the conflicts themselves are resolved. the resolution depends on the relative strengths and interactions of the representations themselves
    • Neural evidence for sensorium hypothesis
      • regions of the brain involved in perceptual processing seem to be responsible for consciousness
        • directly stimulating parietal areas gives rise to the urge to perform a certain action, and further stimulation gives subjects the impression that they actually performed the action, even though they didn’t (Desmurget 2009). conversely, stimulation of premotor areas gives rise to the action itself, but the subject does not believe they performed it
      • note that circuits associated with perception are located throughout the brain, so the sensorium hypothesis is consistent with cortical and subcortical accounts of consciousness
    • Tenet 4: the olfactory system provides clues to the neural correlates of conscious content
      • the olfactory system is relatively simple: it primarily consists of paleocortex (which has half the number of layers as neocortex) and is entirely in the frontal cortex, as opposed to vision or audition which involve large-scale interactions between frontal and parietal cortices
        • unlike most senses, olfaction is directly processed in the cortex ipsilaterally (on the same side) instead of being relayed by the thalamus first.
        • after processing, the olfactory cortical regions do send signals to a thalamic nucleus, but this nucleus does not seem to be required for conscious smell
      • olfactory input is processed in the orbitofrontal cortex (OFC) after being collected by the olfactory bulb, and lesions in this area often cause loss of smell, seeming to indicate the necessity of the cortex for consciousness
      • lesions in thalamic areas cause problems with olfactory discrimination and memory, but never cause anosmia (loss of smell)
      • altogether, the current best hypothesis to falsify is that olfactory consciousness requires cortical processing
      • other interesting properties of olfaction
        • unlike other senses, olfaction generates no subjective experience whatsoever when under-stimulated (e.g. in the absence of odors, or during habituation to existing odors)
          • in comparison, with eyes closed, we still have the conscious content of a dark field. but with smell, there is “experiential nothingness”; one would not know they had a sense of smell if not for the memory of it
          • this allows us to consider smell as a “true addition” to the conscious field and analyze its properties from this angle
  • 4. The generation of conscious contents

    • 4.1 Tenet: Content generation is encapsulated
      • conscious contents generally enter awareness with no effort. they are generated reflexively. Helmholtz called these “unconscious inferences”
      • similarly, action-related urges can also enter awareness reflexively. e.g. the urge to breathe while holding one’s breath underwater cannot be voluntarily hidden from awareness; it is insulated or “encapsulated” from voluntary control
        • specifically, the behavior of breathing can be suppressed, but the conscious content of the urge cannot be
      • it is believed to be adaptive for conscious contents to be encapsulated from each other in this way; it would be maladaptive if one conscious content had direct influence on another
        • e.g. if one’s motivations could voluntarily “turn off” the content of pain or hunger, those contents would lose their adaptive value
        • motivation and belief have an impact on high-level processes like memory, but much less on perceptual contents
      • caveat: “this is not to mean that the configuration of afference engendering one content cannot influence the generation of other contents – a form of context sensitivity in afference processing that occurs unconsciously” (12)
        • e.g. when watching a video of a snake, there is much less of a fear response than seeing a snake in real life. but this is not an example of the conscious contents themselves influencing each other, but rather the afferent signals that generate one content influencing the generation of another
      • this encapsulation is the reason why “illusions persist despite one’s knowledge regarding the actual nature of the stimuli” (12)
      • Because of the principle of encapsulation, conscious contents cannot influence each other either at the same time nor across time, which counters the everyday notion that one conscious thought can lead to another conscious thought. (12)

      • not only can conscious contents not influence each other, but it is postulated that conscious content generators cannot communicate content with each other.
        • For example, the generator charged with generating the color orange cannot communicate “orange” to any other content generator, because only this generator (a perceptual module) can, in a sense, understand and instantiate “orange.” (12)

      • instead, generators can influence each other not by transmitting content but by transmitting activation:
        • This activation, in turn, influences whether the receiver module will generate, not the kind of content generated by the module from which it received activation, but rather its own kind of content (e.g., a sound). (12)

    • 4.2 Tenet: Field contents must meet multiple-constraint satisfaction, be unambiguous, and appear as if apprehended from a first-person perspective
      • at a given time, conscious contents are complete and unambiguous
        • because of the very nature of the constitution of the systems giving rise to conscious sensory representations, these systems are incapable of representing stimulus ambiguity (e.g., as in the Necker cube), at least at one moment in time (13)

        • (although, what do they mean by “ambiguous”? is seeing a somewhat grainy picture at night not an example of ambiguity?)
      • the well-constructed contents could be a result of feedback mechanisms, e.g.
        • “multiple drafts” of content being made and revised (Dennett, Consciousness Explained)
        • “reentrant processing” in which “a module, in order to give rise to a conscious representation, must receive feedback activation from other modules about that representation” (13)
      • contents are generally sensed to be different from the “observing agent”
        • this helps the conscious system to discern the location of different objects in space, as distinct from the organism itself
    • 4.3 Tenet: The conscious field serves as a frame that represents encapsulated contents for collective influence over, not itself, but the skeletal muscle output system
      • Importantly, the collective influence of the combination of contents in the conscious field is not toward the conscious field itself; instead, according to PRISM, the conscious field is apprehended by the (unconscious) mechanisms composing the skeletomotor output system. Thus, the conscious contents of blue, red, a smell, or the urge to blink are the tokens of a mysterious language understood, not by consciousness itself (nor by the physical world), but by the unconscious action mechanisms of the skeletomotor output system. Why do things appear the way they do in the field? Because, in order to benefit action selection, they must differentiate themselves from all other tokens in the field – across various modalities/ systems but within the same decision space. (13)

  • 5. Passive frame theory: an action-based synthesis

    • the primary function of the conscious field is collective influence of otherwise encapsulated contents on the skeletomotor output system (14)

    • defines a “frame check” in which conscious contents are compared with each other
      • For the selection of any skeletomotor plan to be adaptive, selection must transpire ^^in the frame of the other conscious contents^^ composing the conscious field at that instant. We refer to this as a “frame check.” (14)

        • not sure what they mean by “transpire in the frame of other conscious contents” though
      • Consciousness is associated only with frame checks and not with the more active aspects of the conscious-unconscious cycle (e.g., content generation, conflict resolution, motor programming). (14)

    • because consciousness is a “continuous feed system”, conscious contents exist even in the absence of any conflict or any explicit need for skeletomotor actions/decisions
    • remember in this framework, all conscious contents are perceptual. so when a smell triggers an avoidant urge, the content of that urge is about “the perceptual aspects of the to-be-produced action” (14)
    • with regard to the creature in the cave – a certain set of things are all conscious because they all need to be integrated together
      • Again, as with the case of (a) anticipated action effects, (b) actual action effects, and (c) information about the immediate environment, adaptive action selection requires that the conscious contents associated with both Stage 1 (e.g., the percept of the opening and the warmth) and Stage 2 (e.g., the smell and the inclination to stay in the cave) be, in terms of their functional consequences for action selection, ^^the same kind of thing^^ – comparable tokens existing in the same decision space. (14)

      • to me this is giving a great answer as to why a certain set of things are conscious and not others, but it still doesn’t quite explain consciousness itself (but perhaps this really is a metaphysical/unfalsifiable problem 🤷‍♂️)
  • 6. Implications and concluding remarks

    • they describe consciousness as more passive and structure-based than processing based – since there is just an integrated field conscious field that is generated anew in each moment.
      • Figuratively speaking, at one moment in time, there are few “moving parts” in the conscious field. (The field itself has no memory and performs no symbol manipulation; for these high-level mechanisms, it only presents, to action systems, the outputs of dedicated memory systems and of executive processes, respectively.) (15)

    • in the context of this theory, it makes more sense to study consciousness across sensory modalities (“systems-level approach”) than to study just a single sensory modality
      • I wonder where “attention” is relevant in all this? the ability to “focus more” on one sensory modality versus another?