structure for 'Thought'    |     alphabetical list of themes    |     unexpand these ideas

18. Thought / D. Concepts / 4. Structure of Concepts / d. Concepts as prototypes

[concepts as built around typical examples of things]

22 ideas
Many concepts lack prototypes, and complex prototypes aren't built from simple ones [Fodor]
     Full Idea: Many concepts have no prototypes; and there are many complex concepts whose prototypes aren't related to the prototypes of their constituents in the way compositional explanation of productivity and systematicity requires.
     From: Jerry A. Fodor (Concepts:where cogn.science went wrong [1998], Ch.5)
     A reaction: His favourite example of the latter is 'pet fish', where the prototype of 'pet' is hardly ever a fish, and the prototype of 'fish' is usually much bigger than goldfish. Fodor is arguing that concepts are atomic.
Maybe stereotypes are a stage in concept acquisition (rather than a by-product) [Fodor]
     Full Idea: We needn't say that learning a stereotype is just a by-product of acquiring the concept; it could rather be a stage in concept acquisition.
     From: Jerry A. Fodor (LOT 2 [2008], Ch.5.4)
     A reaction: He rejects stereotypes because they don't give concepts the necessary compositionality in thought. But this idea would mean that children were incapable of compositionality until they had transcended the primitive stereotype stage.
One stereotype might be a paradigm for two difference concepts [Fodor]
     Full Idea: The same stereotype can give difference concepts; chickens are paradigmatic instances both of FOOD and of BARNYARD FOWL.
     From: Jerry A. Fodor (LOT 2 [2008], Ch.5.4)
     A reaction: And I'm guessing that lots of concepts could have two equally plausible stereotypes, even within a single mind. Stereotypes are interesting, but they don't seem to be the key to our understanding of concepts.
Categories are characterized by distance from a prototype [Gelman]
     Full Idea: On prototype views, categories are characterized by distance from a prototype.
     From: Susan A. Gelman (The Essential Child [2003], 05 'Causation')
     A reaction: Gelman observes that this view makes no reference to any causal features of things. This cuts them off from using underlying essences in the process of categorisation and concept-formation. How do you spot a prototype, with no category?
There is no 'ideal' bird or dog, and prototypes give no information about variability [Murphy]
     Full Idea: Is there really an 'ideal bird' that could represent all birds? ...Furthermore a single prototype would give no information about the variability of a category. ...Compare the incredible variety of dogs to the much smaller diversity of cats.
     From: Gregory L. Murphy (The Big Book of Concepts [2004], Ch. 3)
     A reaction: The point about variability is particularly noteworthy. You only grasp the concept of 'furniture' when you understand its range, as well as its typical examples. What structure is needed in a concept to achieve this?
Prototypes are unified representations of the entire category (rather than of members) [Murphy]
     Full Idea: In the prototype view the entire category is represented by a unified representation rather than separate representations for each member, or for different classes of members.
     From: Gregory L. Murphy (The Big Book of Concepts [2004], Ch. 3)
     A reaction: This is the improved prototype view, as opposed to the implausible idea that there is one ideal exemplar. The new theory still have the problem of how to represent diversity within the category, while somehow remaining 'unified'.
The prototype theory uses observed features, but can't include their construction [Murphy]
     Full Idea: Nothing in the prototype model says the shape of an animal is more important than its location in identifying its kind. The theory does not provide a way the features can be constructed, rather than just observed.
     From: Gregory L. Murphy (The Big Book of Concepts [2004], Ch. 6)
     A reaction: This makes some kind of mental modelling central to thought, and not just a bonus once you have empirically acquired the concepts. We bring our full range of experience to bear on even the most instantaneous observations.
The prototype theory handles hierarchical categories and combinations of concepts well [Murphy]
     Full Idea: The prototype view has no trouble with either hierarchical structure or explaining categories. ...Meaning and conceptual combination provide strong evidence for prototypes.
     From: Gregory L. Murphy (The Big Book of Concepts [2004], Ch.13)
     A reaction: Prototypes are not vague, making clearer classification possible. A 'mountain lion' is clear, because its components are clear.
Prototypes theory of concepts is best, as a full description with weighted typical features [Murphy]
     Full Idea: Our theory of concepts must be primarily prototype-based. That is, it must be a description of an entire concept, with its typical features (presumably weighted by their importance).
     From: Gregory L. Murphy (The Big Book of Concepts [2004], Ch.13)
     A reaction: This is to be distinguished from the discredited 'classical' view of concepts, that the concept consists of its definition. I take Aristotle's account of definition to be closer to a prototype description than to a dictionary definition.
Learning concepts is forming prototypes with a knowledge structure [Murphy]
     Full Idea: My proposal is that people attempt to form prototypes as part of a larger knowledge structure when they learn concepts.
     From: Gregory L. Murphy (The Big Book of Concepts [2004], Ch.13)
     A reaction: This combines theory theory (knowledge) with the prototype view, and sounds rather persuasive. The formation of prototypes fits with the explanatory account of essentialism I am defending. He later calls prototype formation 'abstraction' (494).
Prototype theory categorises by computing the number of shared constituents [Margolis/Laurence]
     Full Idea: On the prototype theory, categorization is to be understood as a similarity comparison process, where similarity is computed as a function of the number of constituents that two concepts hold in common.
     From: E Margolis/S Laurence (Concepts [2009], 2.2)
     A reaction: Again it strikes me that 'computing' similarity by mere 'number' of shared constituents won't do, as there is a prior judgement about which constituents really matter, or are essential. That may even be hard-wired.
The prototype theory is probabilistic, picking something out if it has sufficient of the properties [Margolis/Laurence]
     Full Idea: In the prototype theory of concepts, a lexical concept has probabilistic structure in that something falls under it if it satisfies a sufficient number of properties encoded by the constituents. It originates in Wittgenstein's 'family resemblance'.
     From: E Margolis/S Laurence (Concepts [2009], 2.2)
     A reaction: It would seem unlikely to be a matter of the 'number' of properties, and would have to involve some notion of what was essential to the prototype.
People don't just categorise by apparent similarities [Margolis/Laurence]
     Full Idea: When it comes to more reflexive judgements, people go beyond the outcome of a similarity comparison. Even children say that a dog surgically altered to look like a raccoon is still a dog.
     From: E Margolis/S Laurence (Concepts [2009], 2.2)
     A reaction: We can defend the theory by not underestimating people so much. Most categorisation is done on superficial grounds, but even children know there may be hidden similarities (behind the mask, under the bonnet) which are more important.
Complex concepts have emergent properties not in the ingredient prototypes [Margolis/Laurence]
     Full Idea: An objection to the prototype view concerns compositionality. A complex concept often has emergent properties, as when it seems that 'pet fish' encodes for brightly coloured, which has no basis in the prototypes for 'pet' or 'fish'.
     From: E Margolis/S Laurence (Concepts [2009], 2.2)
     A reaction: I would take 'pet fish' to work like a database query. 'Fish' has a very vague prototype, and then 'pet fish' narrows the search to fish which are appropriate to be pets. We might say that the prototype is refined, or the Mk 2 prototype appears.
Many complex concepts obviously have no prototype [Margolis/Laurence]
     Full Idea: Many patently complex concepts don't even have a prototype structure, such as 'Chairs that were purchased on a Wednesday'.
     From: E Margolis/S Laurence (Concepts [2009], 2.2)
     A reaction: [The example seems to be from Fodor] I disagree. If we accept the notion of 'refining' the prototype (see Idea 11135), then the compositionality of the expression will produce a genuine but very unusual prototype.
Knowing typical properties of things is especially useful in induction [Machery]
     Full Idea: Knowing which properties are typical of a class is particularly useful when you have to draw inductions about the members of a class.
     From: Edouard Machery (Doing Without Concepts [2009], 4.2.1)
The term 'prototype' is used for both typical category members, and the representation [Machery]
     Full Idea: The term 'prototype' is used ambiguously to designate the most typical members of a category, and the representation of a category. (I use the term in the second sense).
     From: Edouard Machery (Doing Without Concepts [2009], 4.2.1 n25)
Prototype theories are based on computation of similarities with the prototype [Machery]
     Full Idea: The most important property of prototype theories is that cognitive processes are assumed to involve the computation of the similarity between prototypes and other representations.
     From: Edouard Machery (Doing Without Concepts [2009], 4.2.3)
     A reaction: [He cites J.A.Hampton 1998, 2006] This presumably suits theories of the mind as largely computational (e.g. Fodor's account, based on the Turing machine).
Prototype theorists don't tell us how we select the appropriate prototype [Machery]
     Full Idea: We are typically not told how prototypes are selected, that is, what determines whether a specific prototype is retrieved from memory in order to be involved in the categorisation process.
     From: Edouard Machery (Doing Without Concepts [2009], 4.2.4)
     A reaction: One of the aims of this database is to make people aware of ideas that people have already thought of. This one was spotted 2,400 years ago. It's the Third Man problem. How do you even start to think about a particular thing?
Maybe concepts are not the typical properties, but the ideal properties [Machery]
     Full Idea: Barsalou (1983,1985) introduced the idea of ideals instead of prototypes. An ideal is a body of knowledge about the properties a thing should possess (rather than its typical actual properties). ... A 'bully' might be perfect, rather than typical.
     From: Edouard Machery (Doing Without Concepts [2009], 4.5.3)
     A reaction: [compressed] Machery offers this as an interesting minor variant, with little experimental support. I take idealisation to be one of the three key mental operations that enable us to think about the world (along with abstraction and generalisation).
It is more efficient to remember the prototype, than repeatedly create it from exemplars [Machery]
     Full Idea: Instead of regularly producing a prototype out of the exemplars stored in long-term memory, it seems more efficient to extract a prototype from category members during concept learning and to use this prototype when needed.
     From: Edouard Machery (Doing Without Concepts [2009], 6.3.2)
     A reaction: [This is a critique of Barsalou's on-the-fly proposal for prototypes] If the exemplar theory is right, then some sort of summary must occur when faced with a new instance. So this thought favours prototypes against exemplars.
The prototype view predicts that typical members are easier to categorise [Machery]
     Full Idea: The prototype paradigm of concepts makes the strong prediction that typical members should be easier to categorise than atypical members.
     From: Edouard Machery (Doing Without Concepts [2009], 6.4.1)
     A reaction: This is why philosophers should approach the topic of concepts with caution. Clearly empirical testing is going to settle this matter, not abstract theorising.