Wednesday, 11 November 2026

The Theory of Information Pickup and Its Consequences (Gibson, 1979, Chapter 14)

So far, all the traditional accounts of perception have been rejected. Perception is not based on sensations detected by the retina, enriched by memories or processed into experience. Meaning is not simply associations from the past, nor innately given. In this chapter, Gibson brings together things we have already reviewed to lay out the theory of specifying information, and then discusses the (very radical and far ranging) implications of this theory. This is the key to the ecological approach.

Gibson notes there are five key ways this theory "differs radically" from other theories, and the theory of information needs to be able to support these changes:

  1. It redefines perception. Perception is not "an appearance in the theatre of his consciousness" (pg 228); rather, it is an "achievement of an individual", a "keeping-in-touch- with the world" and it entails awareness-of both self and the environment. Perception is a continuous process, not a series of discrete moments. 
  2. What is perceived is different. Based on the discussion of the environment (Chapters 1-3) and how that changes (Chapter 6), we perceive places, objects, substances, event, and affordances (not time, space, colours, etc). 
  3. It requires a new conception of information. Information is not 'knowledge transmitted to a receiver'; it is not transmitted at all. Ecological information is the ambient patterns that surround us, and information is there at the potential point or path of observation, not transmitted from the environment to us. It is inexhaustible; no matter where you look or how closely, more things can still be revealed (pictures are weird because they do not support this). This is not Shannon-Weaver information, but information remains the best term (Gibson notes the potential for confusion and that he 'would use another term if I could'. Information must also allow misperceptions possible, because they can occur; but misperception is not a single problem, but a name for a few different issue, and the errors are based in information (lack of information due to the situation or a failure of adequate exploration, annoying optical circumstances like things looking bent as they enter water, etc). 
  4. Perception as information pickup entails perceptual systems, not channels. Gibson laid this out the the 1966 book but tries again given the misunderstandings that followed. Part of the problem is considering things like looking around as behaviours producing feedback, the outcome of a chain of events, rather than a constitutive part of the process of vision. He lists 5 differences between senses and perceptual systems (pg 234-5)
    1. Senses are defined in term of receptors; perceptual systems are made of organs and their adjustments across various scales, all of which serve the pick up of information in some way
    2. Senses are more passive, systems are more active, in the way they obtain information
    3. Senses take on board a "repertory of innate sensations"; new ways to combine them can be learned, but you can't learn to detect a new sensation. Systems, in contrast, can develop and mature and change what they detect (ecological perceptual learning, c.f. Jacobs & Michaels, 2007)
    4. Sensations only have the quality of the receptor they stimulate; the idea of specific nerve energies. There is nothing inherently visual about the activation of a retinal cell and these things must be inferred. Systems, in contrast, pickup information that does preserve (via specification) the qualities of the environment (e.g. affordances). 
    5. For senses, attention is a central process on top of the detection of sensations; for systems, attention is embodied in the activity of the entire perception-action loop. For systems, attention is also not necessarily a narrowing of focus, but can easily be an expansion (as we investigate gradients, events, etc).
  5. Perception as information pickup entails detecting both persistence and change, all the time, at the same time, as the flip sides of the same coin. This is in contrast to things like space, time, shape, etc. Sense based theories need memory: persistence is when the now and the past are the same, change is when they are different; but there has to be a comparison. This also poses a problem: you need to know which two things to compare before you can even ask this question (this is the basic loan of intelligence problem with association as a mechanism). In contrast, for information, what is invariant and what flows (in, say, the optic array) can simply be detected. We perceive that things have persisted by detecting invariants; we perceive that different things are similar by the more abstract detection of invariants across objects (not as a mental act, though). 

In the second half of the chapter, Gibson takes the theory of perception as information pickup and views a variety of issues through this lens. He makes a series of quite bold claims!

First, Gibson reviews some of the traditional issues in sensory psychology, listing some of the kinds of mental operations required to get them to work. He notes that computational theories are just the latest way of implementing these, and they produce theories with the same kinds of problems and loans of intelligence. In particular, he notes how they all rely on being able to use memory to enrich sensation into perception. This is circular (it explains our knowledge of the world via knowledge of the world) and it rests on a separation of the present and the past, which Gibson claims is a false dichotomy. The nature of information is such that it blurs or erases this distinction; it occurs over space and time, information pickup is continuous, and how things persist and change are specified, not remembered. 

Next he talks about nonperceptual awareness; so called 'higher order cognition' such as thinking and knowing. He suggests that this need to be explained with reference to information, and that one path might be that any perceptual system that has learned to differentiate information variables can also operate without the continued presence of the flow of information. (He doesn't propose any more details, and many more seem needed; work on things such as calibration might help here). 

He then talks about the distinction between imagining and perceiving. In traditional theories, these might be easily confused, because they are both, say, the activity of some part of the brain. Ecologically, the distinctions are clear; things can be imagined, but imagined things simply do not support the very many ways in which real things are experienced by perceptual systems actively exploring the environment. Sometimes things 'seem' real (dreams, hallucinations) but those are all cases where the active exploring that constitutes our typical ongoing 'reality checking' aren't invoked. 

Next up is the notion of knowing, and what that might be in the context of the pickup of information. Traditionally, perceiving is more like imagining and knowing is something quite different; the latter entails storing things so they outlast the sensory present. Information pickup by perceptual systems instead make perception and imagining very distinct activities, and perception and knowing much more alike. Gibson considers knowing as an extension of perceiving, the same kind of thing as perceiving, supported by information and the observer's increasing sophistication in interacting with the perceptual arrays. 

There are some kinds of knowledge that are a little different. Knowledge mediated by instruments and tools can come about by the tool bringing something otherwise out of view into the range of our perception (e.g. a microscope or telescope) and perception can be mostly normal. It can also be less direct (e.g. if it is about some nonecological but real physical quantity like 'length' or 'mass). We learn to perceive and read the tool, so it's harder than just looking through a microscope, but the difference is not one of kind. 

There are other mediations for knowledge, specifically words and depictions. You can put some information into words, but interacting with the word is not the same as interacting with the information, and words do not support the typical exploratory activity of a perceptual system. So words can do some things but not others. Pictures and other depictions are closer to perception; you don't need to learn to perceive a picture on top of learning to perceive the environment, for example; but they are again only able to contain some information, and they again do not support the kinds of exploring we do of the actual environment. So while these are quite powerful ways we have to communicate some things, they remain mediations with limits we can distinguish. Depictions will be discussed in detail in the next and final section of the book.

No comments:

Post a Comment