Languages Magazine

How Information Gets Its Meaning (Specification II)

By Andrew D Wilson @PsychScientists
Gibson proposed that specification was required in order for perceptual information to have meaning that was tied to the world in a manner an organism could use. The concept of specification has been placed back under the microscope by recent theoretical and empirical work. Here I want to briefly summarise the theoretical argument put forth by Withagen & van der Kamp (2010), who worry that specification places too strong a constraint on what a perceiving-acting organism might find informative. They suggest that (visual) perception can still be direct with non-specifying patterns, if you stop thinking information is in the relation between the environment and the optic array but rather, in the relation between the optic array and the organism. They propose this because recent empirical work suggests that organisms can happily get around using non-specifying variables; they want to keep directness, however and they don't think Chemero's solution to the problem does the trick. I'll review the studies they cite over the next few posts; first, let's lay out the solutions they propose.
Again, I want to emphasize that this is very much a work in progress for me. I'm using these posts to come to grips with the arguments, and I don't yet endorse any of these various critiques. My goal is simply to have a clear understanding of what everyone says, so that we can evaluate those claims later on when I review some data.
How information gets its meaning I: Gibson (well, Turvey, Shaw & Mace, anyway)For Gibson and those who followed shortly after, most prominently Turvey, Shaw, Reed & Mace, (1981), specification was critical for a theory of direct perception to succeed. The idea is this: there are properties of the environment to which organisms need to be sensitive. These properties interact with energy such as light and created patterns in structured arrays of energy which can be detected by an organism. The pattern then provides information about the properties if and only if the property caused the pattern by a lawful process that guarantees the two are uniquely related to each other. Specification guarantees that pattern is informative, because of the Shaw's symmetry principle:
We can represent the symmetry principle as follows. Let E = ‘‘The environment is the way it is,’’ I = ‘‘The information is the way it is,’’ and P = ‘‘Perception is the way it is.’’ Also, let ‘‘>’’ stand for the logical relation of adjunction, a nontransitive conjunction that we can read as ‘‘specifies.’’ Then, the symmetry principle is
[(E > I) & (I > P)] & [(P > I) & (I > E)].
In English, this says: ‘‘That the environment is the way it is specifies that information is the way it is and that information is the way it is specifies that perception is the way it is, and that perception is the way it is specifies that the information is the way it is and that information is the way it is specifies that the environment is the way it is.’’ We can simplify this to say that the environment specifies the information, which specifies perception, and perception specifies the information, which specifies the environment. This principle is symmetrical in that the environment, information, and perception determine one another. This, on the Turvey-Shaw-Mace view, is what it is for perception to be direct.
Chemero, 2009, p. 111
The environmental property is projected as a pattern into light according to the laws of ecological optics. By virtue of this lawful basis, picking up the pattern simply is perceiving the environment because the symmetry principle underwrites a legitimate path back to the environment. This principle is critical, Turvey et al argue, because only it can establish a direct path along which information about things in the world can flow to the organism. The whole system depends on specification, however, so losing specification means losing this path.
How information gets its meaning II: Chemero & situation semanticsI've reviewed Chemero's suggestion here; he takes the situation semantics of Barwise & Perry and uses that to underpin the relationship between the world, the pattern and the organism, instead of the lawful basis of the symmetry principle. This removes the need for specification and thus the symmetry principle, and I suggested in my post that this really is a problem. Withagen & van der Kamp (2010) think so to and summarise it very effectively:
Arguing that variables that correlate with an environmental property can also carry information about it, Chemero cannot explain the object of perception solely in terms of the variable that is detected....After all, a pattern in the array can correlate with and thus carry information about many environmental properties. For example, because of constraints (i.e., the laws of mechanics) the above-mentioned variable M correlates with the length, the weight, and perhaps even with the color of rods. So what determines that a participant perceives the length of the rod and not its color when detecting this variable?
Withagen & van der Kemp, 2010, pg, 7, emphasis added
This is the cost Chemero incurs by abandoning symmetry; the mapping between pattern and environment is no longer 1:1. He needs another process to constrain the meaning of the pattern to one rather than another property of the world, and if this is internal then we're back with representational, indirect theories of perception. Withagen & van der Kamp still want directness but don't want specification. They attempt to solve this problem by appeal to Oyama's relational theory of information from developmental systems biology (Oyama, 1985, 2000). 
How information gets its meaning III: Withagen, van der Kamp and developmental systems theory Developmental systems theory is the idea (Oyama, 1985, 2000) that the form of an organism is the result of a developmental process in which multiple factors participate, rather than being specified by, for example, the genome (another classic in the literature is D'Arcy Wentworth Thompson's 1917 book On Growth and Form). In this context, Oyama defines (genetic) information relationally. That is, patterns in a genome (genes) cannot be described as containing information until they take part in a developmental process. Only then can what that a gene means (it's informational content) be revealed, and that meaning will vary with the process in which it takes part. 
The analogy for Gibsonian information is to stop calling the pattern in the optic array information, and instead simply think of it as just a structure. It should only be thought of as containing information when it takes part in a specific process and is placed in a particular relation (i.e. when it is perceived by a particular organism engaged in a particular task). The informational content, i.e what the pattern means, is then defined by the specifics of the relation. The details of the act of perception constrains the meaning of the pattern, making it information.
On this view of information, a given pattern in an array can mean different things to different organisms, or to the same organism at different times. The empirical work I'll review next claims to show that this sort of thing can occur. What an organism perceives at any given time, on this account, is a function of the relation between the pattern detected and that current exploratory behaviour: this, Withagen & van der Kamp claim, is enough to reduce the mapping between pattern and environment back to 1:1. 
ComparisonEveryone into direct perception agrees that properties of the world interact with energy arrays to create patterns in, say, ambient light (making it an optic array). Everyone also agrees that organisms know about the world by detecting this pattern, after a little learning. The main difference is the mechanism that underpins how this last bit works without mental representations providing additional content (see Figure 1; click to embiggen).

How Information Gets Its Meaning (Specification II)

Figure 1. How structure in light can come to directly inform an organism about the world, according to Gibson vs Withagen & van der Kamp (click for huge)

For Gibson and Turvey et al, the critical work is in the projection of the world into a pattern in light. This occurs according to the laws of ecological optics and the resulting pattern is therefore specific (maps 1:1) to the interesting property in the world. All the organism has to do is pick up that pattern and it has perceived the property of the world. The pattern is therefore information, and it is defined with respect to a potential point of observation.
For Withagen & van der Kamp, the critical work is in the perception of that pattern by the organism. Perception is a relational act: it involves an organism placing itself in a relation to some pattern, and the details of the relation depend on what the organism is up to in order to access the pattern. That pattern then has meaning as a function of the relation created by the organism's activity.
This has a side effect. Placing the hard work in the relation between pattern and information means you can relax the rules about the relation between the world and the pattern. In particular, that relation no longer needs to be a specification relation; the pattern just has to be reliable enough so that when the organism perceives it, the meaning that emerges is correct often enough so that the organism doesn't get itself killed. Evolution, they argue, only favours 'slightly better than the other guy' and does not demand 'using the one correct answer'. The overall process, however, is still direct: no mental gymnastics required.
ConclusionsAt this point I will reserve judgment about whether I think this move is a good idea. The first thing to do is to review the empirical evidence that motivated this particular move; this paper gives me the Discussion section summaries, but I need to spend some time on the Methods and Results.
I will say that I'm still worried about the loss of specification; making any old thing that gets projected into the optic array a potential source of information means it's complicated to get from that projection back to some particular property of the world. In addition, the issue of 'works well enough' is a real problem; trying to control, say, locomotion using information that actually means something relevent to locomotion 70% of the time is a recipe for disaster. I'm inclined to think evolution might eat such an organism for breakfast (see? I can appeal to evolution too! :) The account here sounds plausible, but lacks detail. That said, this analysis might come in handy when we start talking about language,where specification seems unlikely. So for now, count me as intrigued but as yet unconvinced.
ResearchBlogging.orgChemero, A. (2009). Radical Embodied Cognitive Science. Cambridge, MA: MIT Press.,, MIT Press e-book
Turvey, M. T., Shaw, R. E., Reed, E. S., Mace W. M. (1981). Ecological laws of perceiving and acting: In reply to Fodor and Pylyshyn (1981) Cognition, 9 (3), 237-304 DOI: 10.1016/0010-0277(81)90002-0 Download 
Withagen, R., & van der Kamp, J. (2010). Towards a new ecological conception of perceptual information: Lessons from a developmental systems perspective Human Movement Science, 29 (1), 149-163 DOI: 10.1016/j.humov.2009.09.003

Back to Featured Articles on Logo Paperblog