Global Workspace Theory
The weakest noticeable sound is defined at 0 decibels. Imagine putting somebody into a scanner, and having them listen to two sounds:
- A trumpet playing at -5 dB
- A trumpet playing at 5 dB
The acoustic difference between the two waveforms are not very different. How similar are the patterns of brain activation?
Here we see that subliminal auditory stimuli only activate early perceptual areas. Consciousness brings with it a huge increase in neural activation! Why should this be?
Global Workspace Theory (GWT) posits that consciousness is involved in two mental operations:
- Binding: perceptual features, distributed across the brain, are bound together into discrete objects
- Broadcasting: these object networks are broadcast to the rest of cortex, allowing consumer systems to use & modify them.
- Consciousness is small: we can only retain a few (less than 7) objects in our head at one time.
- Consciousness is serial: we can’t read two books at the same time.
- Consciousness is flexible: unlike state of the art AI software, human reasoning can effortlessly enter new domains.
GWT explains these facts. Consciousness is…
- … small because it is hard to keep global object networks distinct from one another.
- … serial because it is a singleton: massively parallel modules engage the same centralized resource.
- … flexible because any consumer system can augment the processing of any perceptual object.
The Role of Attention
Attention is a gatekeeper. Our perceptual systems process myriad sensory events, these must bid for entry into the Global Workspace. The brain contains circuitry that implements this selective process, choosing which perceptual objects to bind & broadcast.
Let’s see if we can use this metaphor to make sense of the sprawling literature on attention.
There are three taxonomies of attention that you’ll find in the literature:
- Covert vs overt attention. As discussed in Attentional Spotlight, we can differentiate attending to objects in the periphery, versus saccading to attended targets.
- Bottom-up vs top-down attention. Distinguishes unplanned attention (e.g., to loud noises) vs goal-based attention (e.g., “count the number of times the soccer ball is passed”).
- Feature vs spatial attention. Distinguishes attending to a feature (“look for all red things”) vs an object (“look for a red triangle”)
In an influential paper, Peterson & Posner (1990) present three attentional networks: functionally independent brain systems which do attention. These are:
- Alerting. This network is tightly linked to wakefulness. Startling events induces strong alerting, lounging on a couch less so.
- Orienting. These two networks (one dorsal, the other located more ventral) orients the organism to process incoming stimuli.
- Executive. This network supports complex task execution, and goal-oriented attention.
Peterson & Posner’s framework allows us to simplify the conceptual landscape:
The Orienting network produces Bottom-Up (“externally-driven”) attention. Its dorsal arm contains mechanisms for covert and overt orienting.
The Executive network produces Top-Down (“internally-generated”) attention. Feature and Object attention are both a form of search template, and as such are constructed here.
An Attentional Organ
In my next post, I’m going to argue that the Dorsal Orienting network is the attentional gateway, full stop. It alone performs selection: a single gateway through which percepts pass into conscious awareness.
On this model, the arousal, ventral orienting, and executive networks play auxiliary roles, modulating our brain’s attentional gateway.
Until next time.
- Peterson & Posner (1990). The attention system of the human brain