BIG, BIGGER, BEST: ON ABSOLUTE VS. RELATIVE JUDGMENT
DB: "I would like a criterion of polyadicity. What jumps to mind is that polyadic properties require more than one object for their instantiation."
Okay, I'll bite, but only for epistemic aspects of properties/relations, not ontic ones -- i.e., not about what properties/relations are, but about the conditions under which people judge things to have properties or relations. For my example, I will use "big" and "bigger."
The inclination is to say that being "big" is a monadic, absolute property and being "bigger" is a dyadic, relative property, requiring at least two things to be compared.
The problem is with "thing." For I can certainly look at a circle alone and judge that it is big. And I can look at a second circle alone, and judge that it is big too. And I can also compare the circles, and judge that the second circle is bigger than the first circle.
So far, everything is in conformity with the idea that properties are monadic and absolute, whereas relations are polyadic and relative.
But I can also consider the two circles as one composite object, and I can judge that composite object to be "plus" if the bigger circle is on the right and "minus" if the bigger circle is on the left. So being plus or minus is a monadic, absolute property, yet it is based largely on the same judgment that we just called relative.
There is a tug toward saying that the plus/minus judgment has a relative component because it is based on a sub-thing judgment that is dyadic and relative. But there the cognitive phenomena of "subitizing" and "overlearning" suggest that this too may be an oversimplification:
"Subitizing" occurs when a property immediately "pops out" holistically, rather than as a result of a slower sequential analysis. A good example is counting: If we see a few discrete objects, we can count them, and the more there are, the longer it takes. But when there are only 1-4 objects, we can detect their "numerosity" (cardinality) instantly, without counting, from their (absolute) "shape."
The capacity to subitize cardinality 1-4 without counting is probably inborn, but subitizing in general can also be learned. Often we will be able to subitize after we have practised, through repetition, to detect a property through slow, sequential analysis, and then we continue learning -- "overlearning" -- until the detection becomes as fast and unconscious as a reflex. The plus/minus judgment described above could almost certainly be overlearned to the point where it is subitized.
The reply may be that although one is not conscious of the comparison process implicit in the plus/minus judgment, it is nevertheless going on, unconsciously. One might also argue that there is unconscious counting going on in the case of subitizing 1-4. But it is much more likely that what has happened is the kind of recoding that, since George Miller's celebrated paper on the "
magical number 7 +/- 1" has come to be called "chunking" and "rechunking": The brain simply recodes what it takes to be a "thing": In the case of 1-4 judgments, this chunking is innate; in the case of perceptual overlearning, it is learned, but the resultant property "detector" is much the same: it detects what would otherwise seem to be a series of simpler properties as a single complex property.
Miller did make an important distinction between relative and absolute judgment, however. An absolute judgment is made on the basis of a single thing (usually presented as a sensory stimulus or "input"), whereas a relative judgment is made on the basis of two or more things, presented simultaneously or successively. "Big" would then be an absolute judgment, whereas "bigger" would be a relative judgment. Miller's interest, in that paper, was in
information, which is the reduction of uncertainty, and in particular he was interested in the limits of our capacity to reduce uncertainty in the case of relative and absolute judgments.
The limit on relative judgments (in each sensory modality) is the "just noticeable difference" or JND. Miller pointed out that the size of the smallest difference we can discriminate relatively, and hence the number of JNDs, is largely immutable and depends on the sensory modality and the property dimension in question, as constrained by the inborn sensitivity of our sense-organs and brain.
In contrast, Miller pointed out, the size of the "chunks" we can identify absolutely seems to be 7 +/- 2 (since revised downward by Maxwell Cowan to something closer to
4), but that limit can be increased substantially by recoding more information into a single chunk through overlearning. His famous example was digit span, in which we can usually only repeat by rote a series of about 7 random 0/1 digits we have heard or seen. But if we learn (and overlearn) how to recode the 0's and 1's into composite binary code, using our overlearned decimal names for the cardinal numbers, till we can reliably subitize the decimal names for triplets of 0/1 digits (much the way we can recode dots and dashes in morse code), then we thereby increase the number of 0's and 1's that we can remember considerably. (In the case of morse code, we can extend it to the number of random words we can remember by rote, again about 7 +/- 2 -- and presumably even more if the words make sense.)
There is more: Even "big" is an implicit relative judgment. (That is why I chose it.) It presupposes some sort of a comparative scale as a context. A circle the size of a quarter is big relative to the size of atoms and small relative to the size of planets. We often have an implicit ("subitized") default context for absolute judgments of degree, but that still leaves what first looked like an absolute judgment now looking like an implicit relative judgment instead.
But aren't all absolute judgments like that? There's an old Maine joke that speaks volumes about relative vs. absolute judgment: "How's yir wife?" Reply: "Compayured to whot?"
Even shape detection -- say, circle vs. non-circle -- has an implicit background context, which will result in different judgments when the context of alternatives amongst which uncertainty needs to be minimized is circles vs. elipses and when the context is circles vs. squares. One context or the other (or both, if cued) can likewise be overlearned to the point where it is subitized.
The upshot seems to be either that (1) whether a judgment is absolute or relative depends on what chunks the observer has overlearned to subitize or (2) all judgments are implicitly context-dependent, hence always relative, never absolute.
I have not provided the requested criterion of polyadicity (but perhaps I have given some reason to think we may not need one).
(And I remind you that this was all based on the epistemic, not the ontic properties of objects and properties, as detected by sensorimotor systems like ourselves. On the ontological status of properties and "relations,",
nolo contendere.)
-- SHHarnad, S. (1987)
Category Induction and Representation, Chapter 18 of: Harnad, S. (ed.) (1987)
Categorical Perception: The Groundwork of Cognition. New York: Cambridge University Press.
Harnad, S. (2005)
To Cognize is to Categorize: Cognition is Categorization. In Lefebvre, C. and Cohen, H., Eds.
Handbook of Categorization. Elsevier.