Prototype theory of concept formation

Ted,

You have not read Merlin's text, but if you're more keen on shooting down PT and committing your error, than you are on understanding it, so be it.

"Novel" existent is just another way of stealing the concept. How do you identify an existent as novel?

See my post 8, where I wrote: "From what I gather, an initial core existent is an existent that isn't easily sorted into preexisting categories because it's so different from all other existents. It's a novel stimuli; it's interference pattern in the brain is unique. So it gets mentally filed away in a place of its own." PT and CT posit the same reason for why people form concepts -- it mentally organizes reality into bite-sized chunks.

Yesterday is not a counterexample. Nothing has been offered for it to counter, unless someone unbeknownst to me has explained whether or how PT applies to higher order concepts. More importantly, as I explained, because yesterday is a concept peculiar in either theory, it is a bad example both in CT and PT for explaining abstract concepts.

When two people disagree as to whether a chicken or a sparrow is the actual core example, how do they decide?

This question doesn't make sense. First, people don't go about deciding which existents are core examples. Core examples mentally emerge over time based on correlations among existents. People identify core examples. Second, there isn't just one definitive core example. It works on a sliding scale. Some existents are more typical of a category than others.

As for typical examples from the words you listed under your point 5, I'll answer so as not to come off as dodgy. A car accident with one guy speeding into another's bumper immediately comes to mind for "fault," and a misspelled word immediately comes to mind for "error." But this is neither here nor there. As I've said, I've found criticisms but not explainations of how PT applies to higher order abstract concepts.

The following criticism, which I mentioned in my last post, applies to Rand's theory: "PT complains that CT lumps all members into a category as equals, ignoring the PT notion that some members are more important for the category than others." As I see it, acknowleding an internal structure (a) might help explain how a case can be borderline, and (b) might provide guidance on what members of a category should stay or go when deciding whether to expand or shrink the category.

Just a reminder: I haven't adopted PT, nor am I defending it. I'm trying to understand/explain it. Please keep this in mind.

Jordan

Post 41

Sunday, September 27, 2009 - 11:07am

"Yesterday is not a counterexample. Nothing has been offered for it to counter, unless someone unbeknownst to me has explained whether or how PT applies to higher order concepts. More importantly, as I explained, because yesterday is a concept peculiar in either theory, it is a bad example both in CT and PT for explaining abstract concepts."

Jordan, "yesterday" is a valid disproof of PT by counterexample because PT can provide no core example for that concept. (If you want something more tangible, "critter" is another counterexample.) If PT does not address higher order concepts then it is simply overbilled, a typical academic Anne Elkism, it is not a theory of concepts. I don't know what "CT" is, other than a self-servingly weak description of the definition by essence theory of concepts, but what you have described of it is not an accurate description of Rand's theory, which is the comparandum with which you began this thread. I can provide several valid definitions for "yesterday" off the top of my head. (The waking cycle prior to the given speech act, the 24 hour period that proceed the contextually current one, the most recent day passed, yatta yatta yatta. . . .) Hence "yesterday" is not a counterexample at all for conventional theory.

I see that all my other questions, like the request for a core example for your very own words (error, fault, misdescribe, concept, identifying, typify, except, perhaps, because, category, later) are met with silence.

Post 42

Sunday, September 27, 2009 - 1:51pm

Ted,

Yet again, I cannot find an example of PT applying to higher order concepts. I'm not sure it ever claimed to, nor how it would if it did. So yesterday as an example counters nothing but my guesses and your hasty conclusions. As I explained, under CT, the problem of yesterday is its unique membership, not its definition. If you'd like to debate the pecularity of yesterday under CT, then have it. For this thread, it's a distraction, as I explained.

And I'm not going to explain CT, nor how Rand's theory is akin to CT with few exceptions. There's plenty of literature out there, including Merlin's book, on the subject.

As for being silent to your questions -- check again. You gave a list and asked whether I could provide a "typical image of any one of them" (verbatim). I provided an example for two of them: fault and error.

Not much more we can discuss when you ignore my answers, refuse to familiarize yourself with PT, and yet persist blindly in attacking it. This is unpleasant.

Jordan

Post 43

Sunday, September 27, 2009 - 2:37pm

You did not provide a core image of fault or error, you asserted that one could have a "mental snapshot" of the same, which is simply begging the question.

You keep repeating that I have not read Merlin's writing. I have read about PT in several upper level courses in psychology and epistemology, as well as in places more suited for such notions, like Discover and Scientific American. If Merlin Jetton produces something of value you think I have missed, you can certainly quote him and enlighten all the readers here, not just me.

I think it's fair to summarize that PT provides no relevant criticisms of or counterexamples to the Randian theory. PT does not apply to abstractions from abstractions. PT cannot show when it is necessary to form a new concept, except by smuggling in that "novel" experiences do not fit the definitions of existing concepts, a criterion which it leaves implied. PT offers not one single core example of of anything other than a perceptual level concrete, and even then only a limited number. PT cannot provide a basis for deciding the proper scope of core examples, for example does the image of a robin stand for the concept "robin" or the concept "songbird" or the concept "bird" in general. (And if the same image is the core of multiple concepts, then what, other than the definition, differentiates those concepts from each other if the core example does not?) Finally, if PT were a new medicine it would fail clinical trials on two grounds. It has serious flaws, and it does nothing better than the tools we already have.

If I understand you, you are retreating from your original statement:

"Rand's theory of concept-formation is basically an iteration of the classical theory, a process of separating like objects apart from unlike objects based by reference to some definitive feature(s). In the 70s, Eleanor Rosch introduced an alternative theory dubbed the prototype theory, a process of comparing how similar objects are to to some definitive "core" object(s). Both seem plausible to me. Both hold different epistemic implications. I'm not sure what to make of this. Thoughts?"

and you now hold that PT does not compare to Rand's theory, but applies only in a limited way to a limited subset of concrete-level concepts. As I have said, and as Rand grants, concepts do admit of associations other than their essence. They do have connotations and the like. I think most people will happily admit that if pressed, when asked to imagine an exemplar of a perceptual level concept that the image will be a more typical one. But what is esssential to concepts, also called universals, what makes the infinitely open-ended, is that their definitions apply to any units that fit the criteria, past present and future. As perceptual concretes, prototypes can serve no such universal purpose.

Post 44

Sunday, September 27, 2009 - 2:54pm

While I obviously reject its wider implications, prototype theory does reflect a real psychological phenomenon. When I asked my father to think of a bird and then tell me what he saw, he agreed with Ryan, he did not see any particular bird. When I asked him to visually imagine any specific bird he closed his eyes, and said okay. When asked he said he had imagined a robin, hence my examples above.

I think the phenomenon of synesthesia is also relevant. Many people do associate things such as colors with tomes, or numbers, letters, and days of the week. I associate letters and numbers with certain colors. This tells us something about human psychology. Presumably I associate letters with colors because of words learned early on. The letter A is red for me, presumably because apple begins with the letter A. But these associations are subjective. Tone-color synestheticians do not at all agree with what tones go with what colors. And I fear that prototype images are similarly subjective.

Post 45

Sunday, September 27, 2009 - 5:53pm

Hi Ted,

You did not provide a core image of fault or error. . .

I really did. In post 40, I wrote:

As for typical examples from the words you listed under your point 5, I'll answer so as not to come off as dodgy. A car accident with one guy speeding into another's bumper immediately comes to mind for "fault," and a misspelled word immediately comes to mind for "error." But this is neither here nor there. As I've said, I've found criticisms but not explainations of how PT applies to higher order abstract concepts.

*

Yes, as I explained to Ryan in post 25, I misrepresented PT re "definitive 'core' objects." But it'd be premature to conclude that PT doesn't compare to Rand's theory, or that it applies only to lower level concepts. I need more info on it.

*

I'm not going to reproduce Merlin's section on PT because it's worth reading in its entirety (about 5 minutes or so). It's too much to reproduce here, but the link to it suffices perfectly.

Jordan

Post 46

Monday, September 28, 2009 - 8:07am

Thanks, Jordan. The example you provided of the bumper accident is what I had referred to as an editorial cartoon. The problem is that the core example should be immediately and unambiguously perceivable (intuitable if you like) as a primary example of that concept and only that concept.

The bumper accident example is like an editorial cartoon. Its meaning is not perceived directly, but is inferred conceptually. You can guess at the meaning, like in a game of charades. But the meaning is not perceived, it is interpreted.

It is also ambiguous. It could just as easily be the core example of accident or impact or an infinite number of other interpretations as for fault. How do we know which? This is not a fault of your example, so much as a fault of the theory. By substituting a token of the concept for the definition, PT loses the specifying power of the genus/differentia definition. Consider a Venn diagram. The conventional theory says a concept is defined by the area where the genus circle overlaps with the differentia circle. Conventional theory defines an unambiguous space. All items and only such items as fall within the intersection are species of that concept. PT picks one species, one X, and says that any item that is sufficiently similar falls under the same concept. PT picks a point and presumably other points within some undefined distance of that point. But X's within Venn diagrams do not define sets. An X is equally the member of any and every circle which encloses it.

(Edited by Ted Keer on 9/28, 8:15am)

Post 47

Monday, September 28, 2009 - 9:44am

The prototype theory makes sense but is far from complete as a theory of concepts. In my opinion it best explains how children hold concepts of commonly perceived entities and qualities. Similarity provides the "glue" for the prototype theory, as it does for Rand's theory.

Babies don't exit the womb and start forming concepts the same way adults do. They lack the experience and mental development. They see, for example, a small number of balls or dogs, and the similarities constitute a rudimentary concept of 'ball' or 'dog', the label typically learned from older people. Children's category judgments tend to be based on examples, on easily perceived traits like shape, and holistic (arguably based on strong neural responses), not a complete feature-by-feature comparison of a candidate to the defining elements of a concept. The most familiar example functions as a prototype (there may be multiple prototypes for some categories) to decide if a newly encountered candidate is similar enough to fall under the concept. As a wider variety of tokens are experienced, the concept expands. Less reliance is based on easily observed similarities, while other sorts of "deeper" similarities are recognized.

Ted wrote:

By substituting a token of the concept for the definition, PT loses the specifying power of the genus/differentia definition.

The prototype theory's main concern is recognition of instances of a concept. It is not concerned with genus/differentia definitions, nor supplanting them. Ask a few toddlers who know milk when they see it to define "milk". They are incapable of a genus/differentia definition. All they can do is point to an instance of it.

(Edited by Merlin Jetton on 9/28, 10:14am)

Post 48

Monday, September 28, 2009 - 10:41am

Thanks, Merlin. Again, I happily concede that concepts have connotations, associations other than just their definitions. The concept bird is undoubtedly connected in the mind with "early x gets the worm" and "hitchcock" and "fighter jet's target" and a whole host of phrases, images, sounds and other associations. That is how we can do crossword puzzles, make jokes, etc. Of course our mental notions for some perceptual level concepts are linked to "typical" images such as the schoolbook image of a robin catching a worm.

But this does not capture how a child like my nephew forms the concept "more" which it holds before it begins to speak! It does not capture how concepts are formed at all, since a core image is a point in mental space, while a concept is a multidimensional space bounded by the intersection of the surfaces of the Venn volumes of the genus and the differentia.

Jeff Hawkins' On Intelligence has a much more fruitful theory of the mental structure of concepts than this. His theory is entirely consonant with conventional definition-by-essence theories, and it applies to concepts at all levels. He actually provides a posited brain structure that shows a hierarchical network consonant with conventional theory.

Yes, a small child cannot express a definition, but he holds definitions implicitly. Many adults hold their definitions only implicitly. This is not a problem for conventional theory. The Socratic method shows that those definitions are there to be elicited when prompted by one's own attention or an instructor's coaching.

There are still concepts like critter or toy where the idea of a "typical" core is absurd. A ball and a block and jacks are toys but the first resembles a fruit, the second a brick and the third caltrops far more than each does the other. There is no plausible "typical" core image for the concept toy. If we want to say, quite cautiously, that people often form core images for perceptual level concepts that's fine. That then is a theory of core images, not a theory of concepts or of concept formation. Having read the literature I know that this theory is not presented in that limited reasonable way.

Post 49

Monday, September 28, 2009 - 2:05pm

Hi Ted,

The fender-bender is akin to an editorial cartoon. I see what you're saying. Its meaning is indeed less apparent than that of images evoked in connection with lower level concepts. The similarities between the fender-bender and other instances of "fault" are not readily apparent from mere imagery. Perhaps such a mental-snapshot just serves as a marker for those super-perceptual aspects that are not mentally "snap-shotable," but that underpin the similarities giving rise to the concept "fault"? Or maybe PT dispenses altogether with mental snapshots when it comes to those higher level concepts? To belabor the point, I'm still really in the dark when it comes to whether/how PT treats higher level concepts.

*

Notably, I should mention that I've seen Objectivists (on a now defunct list-serv) veer away from a strictly definition-oriented approach to categorization, specifically when it came to exceptional individuals being included under the concept Man defined as a rational animal. To parallel that example, take the concept bird defined as a feathered animal -- or if you'd prefer, defined as a feathered, bipedal, egg-laying animal. Strictly, a chicken without any feathers wouldn't be a bird under this definition. But the chicken would still be lumped into the category because it is in so many other ways like other chickens that are definitively birds. Again, this is how I've seen some Objectivists reason, and it might well be couched as a non-Objectivist approach since it falls outside the strictures of the Objectivist theory of concept formation. I still thought it worth mentioning.

Jordan

Post 50

Monday, September 28, 2009 - 2:52pm

Rand does argue, and I agree with her, that you need a perceptual tag to effectively manipulate concepts, and that is the purpose that a word serves. She goes so far as to say that a phrase like weapon of mass destruction is not a real concept until you come up with an acronym for it, like dubyaemdee.

You could certainly argue that prevocal children and feral children or the neglected deaf-dumb might operate on the level of visual tags. Temple Grandin of Animals in Translation and An Anthropologist on Mars seems to think this way. She would be the one to ask if she has a core image of "fault." She speaks of having things filed away as mental video tapes. I don't see how one could really test it, since learning to talk most likely rewires and overwrites earlier tags. As a matter of cognitive economy, visual images cost vastly more than do phonemic tags. Just compare the compression of an ascii coded text with a gif of the same text. Grandin speaks of having to run thru her tapes to think things out and I think that in so far as that is true she is very much thinking in a method different from and alien to word-tagged concepts with definitions by essentials. That may be the source of her relative strengths and weaknesses.

I assume it's obvious I would oppose any attempt to use PT as a justification for teaching look-say over phonetics or any other way of dumbing down the teaching of definition-by-essential thinking. I am highly skeptical of any approach to this that treats it as anything but a technical and academic side issue to definition essentialism.

(Edited by Ted Keer on 9/28, 10:11pm)

Post 51

Monday, September 28, 2009 - 10:09pm

Jordan, can you be more specific as to what you mean by this: "Notably, I should mention that I've seen Objectivists (on a now defunct list-serv) veer away from a strictly definition-oriented approach to categorization, specifically when it came to exceptional individuals being included under the concept Man defined as a rational animal." I would like to comment but I want to here more from you first.

Post 52

Tuesday, September 29, 2009 - 2:00pm

Hi Ted,

It was the wetheliving list-serv. Paraphrasing the question, it was asked (not by me) whether severely mentally retarded, comatose, and senile individuals belong under the concept human in light of their failure to meet the species criterion, "rational," in the genus-species definition, "rational animal."

Some discussers insisted that those individuals in question really did have whatever "rational" referred to. Others simply excluded those individuals from the category. Yet others suggested that the individuals belonged, despite their being exceptional, because of their affiliation with those who clearly did belong, pointing out that the exceptional individuals still have human DNA, anatomy, parents, and various other "human" features.

I posed the parallel example of whether a featherless chicken belongs under the concept birds, so as to present those discusser's thinking in a plainer and less contentious context.

Jordan

Post 53

Tuesday, September 29, 2009 - 5:26pm

The following appears in this thread with the encouragement of Ted Keer, after I'd sent it as a private message to Jordan and Ted]

[message subject:]
the humanity of those not reasoning (e.g., senile, comatose, or severely retarded people)

Ordinarily, I'd post the following publicly on RoR. However, recent messages there make me wary of sending the forum any question which any member may, conceivably, misidentify as meaning something other than the literal meaning of its words in the sequence in which I arranged them. (Since I cannot know in advance what someone else will misidentify, that may well mean "any question whatsoever.")

Jordan -- thanks for your recent message about some past Oist discussions re whether we can regard senile, comatose, or severely retarded people as human.

/a/
May I observe that a "no" answer would make it philosophically unobjectionable to (e.g.) use senile, comatose, and/or severely retarded (but otherwise physically healthy) people as a source of dietary protein? Contemplate, then, the difficulty of ensuring that a slaughterhouse or butcher shop or restaurant, free to procure and sell such meat, indeed sells only the meat of guaranteed non-rational humans -- or, rather, guaranteed non-rational "whatever-we-ought-to-call-them-if-we-shouldn't-call-them-humans-because-they-aren't-rational."

(If we classify such people as "not human," what term would we assign to the concept expressed by the above cumbersome phrase?)

Since one cannot easily measure the (former) IQ of a corpse, it would just get too easy for murderers to dispose of the body by sneaking it into a slaughterhouse production line, a meat wagon, etc.. The diner who has ordered a platter of "nonrational rib roast -- IQ during life guaranteed below 5" would not know whether the restaurant had served him what he ordered -- or had served him instead the restaurateur's quite rational, if inconvenient, mother-in-law.

/b/
One might also point out that societies which classify certain infants, children, teens, or adults as "not human" have, historically, had poor records as defenders of human freedoms.

Post 54

Tuesday, September 29, 2009 - 5:46pm

Oh sure, once more, blame me!

:)

Post 55

Tuesday, September 29, 2009 - 6:45pm

Hi Kate,

The topic resurfaces periodically in Objectivist forums. Consider making a new thread if you'd like to explore further. I should've also mentioned a fourth route individuals on that past list-serv took, namely, modifying the definition of human to accomodate those otherwise exceptional cases.

Jordan

(Edited by Jordan on 9/29, 7:20pm)

Post 56

Tuesday, September 29, 2009 - 8:01pm

Thanks, but I don't see a need to explore further: it seems unlikely that Objectivists look forward to dining on roast imbecile under glass.

Post 57

Tuesday, September 29, 2009 - 8:12pm

Here's a thought: Suppose that there were a genetic treatment that raised the average person's IQ to that of Einstein x 2 - or about that estimated for William James Sidis, often cited as possibly the most intelligent human ever...

But, it only worked on 75% of the population. For the rest of humans, their IQ would be unchanged.

Relatively, however, that 25% would be retards, on a relative par with a typical Down Syndrome person today.

Most communications with the unfortunate 25% would require a painful effort of reducing everything to their cognitive level. For someone in a close, long term relationship or between family members, this would not generally be a devastating problem, because healthy long-term value relationships would generally last, I'm guessing. And, the 25% would still be just as productive, on average, as they had been before the upgrade of their fellows. There are many people with Down Syndrome today who live productive, happy lives.

However, just as we - referring to those in the majority who were successfully upgraded - instantly spot someone who is seriously retarded today, we would naturally spot those people who were operating at about 1% of our capacity, and a natural reaction would be to avoid them, just to save time and the difficulty of spelling out everything that the new "normals" would grasp at a hint.

This sounds like pure science fiction. However, there is nothing in the laws of biology that forbids it, and examples such as Einstein and Sidis show that such mental prowess is in fact possible.

Post 58

Tuesday, September 29, 2009 - 9:15pm

The problem, in part, is the failure to recognize that a concept may have different definitions under different contexts, and that two separate concepts may share the same or largely the same referents yet be distinguished because they are differentiated from different things.

The biological concept human, for instance, and the ethico-metaphysical concept man are two different things. Person, man, adult, child, human, the concepts are many and each has many senses. The philosophical concept man is normative. It asks what type of life is proper for man. The biological concept human is descriptive. It treats all members of the species Homo sapiens as humans merely by the fact that they are are the offspring of an interbreeding population of apes that does not interbreed with other population. Biological species do not have essences per se, they merely have type specimens.

Plucking a chicken does not make it cease to be a member of the species Gallus gallus domesticus. While we argue what is proper for the law based on a concept of man, legally, all humans are presumed persons until shown otherwise. There are many other issues such as the fact that we are all only potentially rational most of the time. One does not lose the right to life because unconsciousness removes one's irrationality. Such complexities fill libraries.

Were it only fair to do so we might turn the tables on sophomoric rationalists who substitute a definition of a concept for the concrete reality which that concept is a tool to describe and require them to prove their rationality before we recognize their rights. As it stands, we take it as a given that all born humans have rights unless by their own actions they determine otherwise.

(Edited by Ted Keer on 9/29, 9:17pm)

Post 59

Wednesday, September 30, 2009 - 9:38am