Are you conscious? 17 Criteria for Consciousness

A lack of clear definitions for terms like "intelligence" and "consciousness" plagues any serious discussion of those concepts. A recent article by Seth, Baars & Edelman argues for a core set of 17 properties that are characteristic of consciousness, and could be used in the "diagnosis" of consciousness in humans and other animals.

Property 1: "Irregular" patterns of brain activity

Electrical oscillations occuring between 20 and 70 times per second are common in awake humans, but epilepsy, sleep, anesthesia and some forms of brain damage are accompanied by the dominance of highly regular oscillations slower than 4 Hz. Others have argued that the conscious EEG is characterized by a particular noise signature known as "pink" or "1/f" noise. Although the purpose of these peculiar electrical oscillations is poorly understood, they appear to be a consistent feature of primate consciousness.

Test: electroencephalography (EEG)

Property 2: Connectivity "loops" between thalamus and cortex

While an entire hemisphere can be removed without apparent loss of consciousness in human subjects, bilateral damage to specific regions of sensory cortex seems to remove those senses from consciousness (such as in blindsight, where subjects lose the ability to consciously report visual stimulation, but can nonetheless act reflexively or non-conscoiusly in response to such stimulation).

However, damage to the thalamus (and associated brainstem regions) results in a complete loss of consciousness. Seth et al. argue that the thalamus maintains the state of consciousness through its interactions with neocortex, and the neocortex is responsible for providing the contents of consciousness.

Test: white matter tractography or dtMRI

Property 3: Distributed Representations

While automatic or reflexive tasks are associated with fairly retricted distributions of neural activity, more novel and complex tasks (presumably those involving consciousness) are associated with more widespread and distributed representations throughout the brain (particularly involving a distributed "task network" of frontal regions).

Test: PET, fMRI or MEG

Property 4: Embodiment and Multiple Sensory Modalities

Seth et al's fourth property suggests that consciousness requires diversity of content that may be specifically provided by embodiment, perhaps because bodies provide diverse sensory (and proprioceptive) inputs for processing. The basic idea is that a critical threshold of informational complexity is required before consciousness can emerge.

Test: Not currently testable

Property 5: Emergent Information Maximization

Seth et al. argue that when information is redundant or uninformative, it tends to fade from conscious awareness. Furthermore, according to their "dynamic core" theory of neuronal oscillations, brain activity self-partitions into statistically-independent "clusters" of co-active and collaborative activity patterns. The degree of complexity in each cluster determines its probability of entering conscious awareness (as complexity is defined by, say, Kolmogorov). In other words, the dominant or largest cluster of active neurons is the manifestation of the "dynamic core" of consciousness. Seth et al. do not specify exactly why redundant activity patterns do not self-reinforce one another (and thus become dominant and enter consciousness).

Test: information theoretic analyses of EEG, fMRI, MEG, PET, single-cell recording, etc. (an example of this kind of analysis)

Property 6: Present-centeredness

Seth et al. argue that consciousness is centered on the present, probably for evolutionary reasons, and that this present-centeredness may arise from the fact that the dynamic core of neural activity is in constant flux.

Test: tests of "rapid adaptivity" or "fleetingness," for example the Sperling partial report

Property 7: Internal Consistency

Seth et al. argue that because functional clusters compete for entry into consciousness, consciousness at any given time is itself dominated by an internally-consistent dynamic core of activity. Although this dynamic core may rapidly shift (displaying "rich intermittency" in its attractor dynamics), at any given time conscoiusness is internally consistent.

Test: Bistable images such as the Necker Cube, EEG "flicker tagging" with the Figure/Ground Illusion

Property 8: Capacity Limitations and Seriality

Seth et al argue that while consciousness is not equivalent to working memory or attention it is similarly limited in capacity, due to the fact that only a single cluster can dominate the dynamic core of activity patterns that make up consciousness at any given piont.

Test: Illusory Motion Reversal, Binocular Rivalry

Property 9: Unity and Coherence of Action and Perception

Following from property 8, Seth et al. argue that the "binding together" of widely distributed representations (Property 4) is a central feature of consciousness. This may serve to accomplish both present-centeredness (Property 6) and internal consistency (Property 7). Although Seth et al. endorse the hypothesis that binding is accomplished through synchronous neural oscillations, they admit that focused attention may also contribute (as hypothesized by Anne Treisman, among others).

Tests: "Recombination span" (aka binding) tests, measures of neuronal synchrony from EEG, MEG, or local field potential recordings

Property 10: Self-awareness

Seth et al. point to evidence that many spatial representations in the brain are egocentric, meaning that the locations of objects and actions are "self-centered". They suggest that consciousness requires at least this form of self-awareness, and that "higher-order consciousness" (e.g., being aware that you are conscious) may have been "evolutionarily bootstrapped" by this kind of elemental and spatial representation of self.

Test: Mirror Test

Property 11: Accurate Reportability

In perhaps their most controversial property, Seth et al argue that consciousness is accompanied by the ability to explicitly indicate the nature of subjective experience.

Test: Behavioral training with output devices (keyboards, sign language, etc)

Property 12: Privacy of consciousness

Seth et al. argue that consciousness is by definition bounded within a self (Property 10) and that subjectivity directly arises from this property as it interacts with emergent information maximization (Property 5).

Test: Not currently testable

Property 13: Focus-fringe structure

Seth et al suggest that consciousness is composed of a central focus and a sense of surrounding vagueness or uncertainty, perhaps representing the presence of nondominant funcitonal clusters outside the dynamic core. Examples may include tip-of-the-tongue and deja-vu.

Test: Not currently testable

Property 14: Advantages for Learning

Consciousness has the property that it improves learning, relative to learning mediated by unconscious processes, perhaps because of the increased computational power provided by thalamocortical architectures (Property 2).

Test: lesions or rTMS of brainstem and examination of effects on learning

Property 15: Independence from sensory input

Seth et al. argue that consciousness is also stable from interference from sensory input, such that certain characteristics of consciousness may last the organism's entire lifetime yet differ between organisms (such as personality characteristics).

Test: Delay-match-to-sample tasks

Property 16: Allocentricity

In addition to egocentric representations, Seth et al. argue that consciousness is accompanied by an alternative representational scheme of "other-centered"-ness. For example, we can name not only the spatial relationship of a nearby object to ourselves, but also its relationship to other objects in the room. (Seth et al. do not specify why allocentricity cannot be built out of iterated egocentric calculations: e.g., "the cup is to my left, the spoon is farther to my left, therefore the spoon is to the left of the cup").

Tests: Mental rotation

Property 17: Intentionality

If a certain goal cannot be accomplished with reflexive behaviors, consciousness can overcome those reflexes to accomplish the goal.

Tests: decision making and planning tests such as Tower of Hanoi, or Wason selection, where reflexes are unlikely to suffice. Also post-decision wagering (Thanks Neil!)

Conclusions

Seth et al have provided an interesting set of criteria for evaluating consciousness - the implication is that later criteria build on the earlier ones. It has applications to even the most controversial cases (for example, Terri Schiavo) and could be useful as a benchmark in the evaluation of animal intelligence and animal consciousness.

That said, there are lurking problems with this account. For one, there could be alternative forms of consciousness which satisfy the later criteria without satisfying the earlier ones - we just don't know. The entire approach reeks of human-centeredness, though of course that's a necessary evil: we don't really know that any other species is conscious (and furthermore, we're merely giving our fellow humans the benefit of the doubt).