exeunt

The course is over, this blog is no longer active.

What we’ll discuss in the class on lexicalization

Next time we’ll try to discuss lexicalization – a process that relates concepts to words, and implicitly to natural language semantics. Here are the main issues we’ll seek to shed some light on:
1. Lexicalization: what it is and how it is manifested in language.
Remember what has been said last time in class about the representational approach to semantics. If we adopt this view then we’ll describe meaning as a combination between a conceptual element (a component of our conceptual structure, a component of a way (a model) in which we conceptualize the world we live in) and the denotation of the linguistic element we use to express that conceptual element (namely the concept). To put it simply, the meaning of a noun would be the combination between its denotation (i.e. the object) and the concept it expresses. Analogously, the meaning of a verb would be the action it denotes and the concept it expresses.
If a concept corresponds to a word in a language, this means that that concept is lexicalized. Therefore, lexicalization is a process that makes concepts expressible in language i.e. a process through which we assign (linguistic) form to concepts (mental entities). Not all concepts we use are lexicalized; for instance, some concepts are expressed by phrases (can you think of examples?).

2. What are the reasons why (some) concepts are lexicalized?
Here I would like you to think of some possible reasons. A helpful starting point would be the part 2.4.2 “Concepts” from Saeed’s chapter “Meaning, Thought, and Reality” that we referred to last time. But we encourage you to think of other reasons than those mentioned in the chapter.

3. Last, we’ll be concerned with lexical relations. This is the main preoccupation of what is called lexical semantics. The goal of lexical semantics is to study word meaning. Among the descriptive aims of lexical semantics there is that of showing how meanings of word in a language are interconnected and how speakers become aware of that.
In other words, we’ll try to see how the relational nature of concepts is reflected in language by means of characterizing the relations in which words corresponding to concepts (expressing concepts) enter.
The lexical relations we’ll discuss are:
• Homonymy
• Synonymy
• Opposites
• Polysemy
• Hyponymy
• Meronymy
Can you explain what each of these lexical relations means? Use your own words and try to give examples, this will help a lot.
Briefly, this is the plan for the next class. I hope it’ll help you organise both your ideas and questions about the topic of the class that you’ll raise on next Friday.

Reading: In order to prepare for this class you should read the paper I gave you last week called “Exploring the Relationship between Lexical Behaviour and Concept Formation in Design Conversation”. You will see that this is rather a technical article and it reports and discuss the results of a computational linguistic analysis. You don’t have to worry that you’ll not be able to understand all that technical language. You should concentrate on the subparts called “Lexicalized concepts” and “A psycholinguistic model” (pp. 2-3) and “Discourse analysis” (pp. 7-10).

summary of the 7th class

Three things happened in the last class. First, there were some introductory remarks about externalism. Second, we discussed Millikan’s project – mainly its goal and quite briefly [since again you didn’t read the text] the substantive part of the proposal. Third, we mentioned some of the motivations for such a move against descriptions and also possible practical consequences. Here I will focus on the first three sections of the paper.

1. Intro

Both the Classical Theory (CT) and the Prototype Theory (PT) consider concepts to be constituted by certain kinds of descriptions. In the first case, the descriptions are definitional, they consist of necessary and sufficient conditions for the application of concepts. To apply a concept C means to say of an object a that it is an instance of C. E.g. ‘Ana is a student.’ The predicate ‘student’ expresses the concept STUDENT. This concept is a description. It might be something like this description: x is human; x is enrolled in a higher education institution. In bi-conditional form: x is an instance of C if and only if x is human and x is enrolled in a higher education institution. The concept can be thus applied to all [sufficiency] objects that satisfy these conditions – and only [necessity] to them. The important point is this: what you have when you have a concept is a description like the one just mentioned. The idea is that when you apply/use a concept, you go over the list of conditions. Over all of them.
I assume that it is clear by now what changes when one moves to Prototype Theory. Concepts are still descriptions, but the conditions of application are relaxed. Roughly, we have features where we had conditions, and features can be weighted. We’re looking for relevant similarity, not for identity between objects that fall under the same concept. Note that in this case descriptions might not be of a linguistic nature – this was strongly suggested by the Classical Theory. Prototypes are structured sets of features – and features could be bits of sensory information – e.g. visual info.
There’s another thing [other than descriptionism but related to it] that CT and PT have in common. They are internalist (the common versions that we discussed at least). This means that the nature of a concept depends only on what goes on in the heads/minds of agents. The environment is irrelevant. Descriptions – definitions or prototypes – are something that we cook up in our heads, and then apply to the external world. One could imagine scenarios [google ‘brain in a vat’] in which there is little connection or correlation between the conceptual repertoire of a mind and its environment.
Millikan goes against both descriptions and internalism. It is therefore an alternative proposal in a more radical sense.

2. Millikan’s paper

Section #1
The project: Kripke and Putnam manage to put up a powerful criticism of descriptionist accounts [see ‘The problem of ignorance and error’!]; but we lack a detailed positive alternative built on their insights. So let’s try to have it done.
Note: The main insight of K&P was that that meaning of the word – the concept expressed, we would say – does depend on the external environment of the agent and not on what goes on in the agent’s mind. If the substance in the lakes and rivers around you is not H2O but XYZ, than your concept WATER does not consist of the description H2O, but of whatever reliably ‘points to’ the substance in the lakes and rivers around, that is XYZ; it doesn’t matter if you have the description H2O associated with WATER. Moreover, you can conceive of the external environment as social environment: so it’s not what you have in your head that matters, but what the community or the culture agrees on.
Scope of the project: concepts of substances. Substances = individuals, stuffs, real kinds [natural kinds, like hydrogen, Rosch’s basic level categories – natural objects, like mice, & artefacts, like houses].
Proposal: all substance concepts do something like pointing; they are ‘subject concepts’. We don’t classify objects when we use them – we don’t describe someone when we say ’Mama’, we don’t describe something when we say ‘milk’ – at least initially; we identify them (over multiple encounters).
What does this mean? Problem: ‘cash in the metaphor of pointing’ [p.526].
Note: the account will be causal (based on interactions with the environment) and will not depend directly on language > ‘Preverbal humans, indeed, any animal that collects practical knowledge over time of how to relate to specific stuffs, individuals, and real kinds, must have concepts of them.’ [p. 526 – my emphases].

Section #2
This section explains what is understood by ‘substances’. The central notion is that substances have ‘rich inductive potential’ [pp 527- ff]. You can learn about them quickly – maybe from a single encounter – and you can safely generalize [that is what induction means] what you learned to future encounters. Mama will continue to be nice next time you see her. Water will still be wet. ‘…knowledge will remain good ’ [p 528].
It makes sense to pay attention to substances then. They allow you to have at least some initial grip on the world. This is possible because having concepts of substances is not a matter of how your mind decides to cut up things, but depends on the real structure of the world: ‘It is not a matter of logic, of course, but rather of the makeup of the world’ [p. 528].
Note: by ‘the makeup of the world’ one should understand physical structure to which we are sensitive [what we can see, hear etc], but also socio-cultural structure. See e.g. the discussion of design and copying patterns on p. 529.
The project restated: ‘Throughout the history of philosophy and psychology, the tendency has been to project into the mind itself the structure of the object grasped by thought. I will argue the contrary, namely that substances are grasped not by understanding the structures or principles that hold them together but by knowing how to exploit these substances for information gathering purposes.’ [p. 530]

Section #3
‘The “concept” of a substance […] is the capacity to represent the substance in thought for the purpose of information gathering and storage, inference, and ultimately the guidance of action.’ [pp 530-531]
Here is how I think we should read this: the emphasis should be placed on ‘capacity’, not on ‘represent’. The concept is an ability to do certain things – e.g. guide action. This capacity depends on representing some of the stable properties of a substance. For example, we represent the wetness of water. But the capacity depends on representing some properties, not on having a certain representation. The capacity survives as representations change. So the concept is not the representation that supports re-identification.
Indeed, the representations do chance [again, see ‘The problem of ignorance and error’] as we learn new things about substances. We learn, for example, that water boils or freezes, that it can stop being wet. We still identify water as water throughout these changes. Remember that substance refers also to individuals. Maybe such an example is easier to follow: you re-identify your mother as your mother even though she – and what you know about her – changed a lot in 20+ years.
‘Information gathering’ says just this: we maintain some (minimal) label for the substance and we attach knowledge to it as that knowledge becomes available to us. We can ‘track’ a substance – this is primary. We can recognize water. The descriptions that we acquire in time are secondary. They do not belong to the nature of the concept of that substance. It doesn’t matter that we learn very late in our development that (pure) water is H2O.
‘Now think why a child, or animal, needs to carry knowledge of the properties of a substance from one encounter to another. If all of a substance’s properties were immediately manifest to the child upon every encounter there would be no need to learn and remember what these properties were. Carrying knowledge of substances about is useful only because most of a substance’s properties are not manifest but hidden from us most of the time. This is not, in general, because they are “deep” or “theoretical” properties, but because observing a property always requires a particular relation to it.’ [p. 532]
Carrying knowledge must be a robust process, since different properties of substances will be ‘visible’ on different encounters and in varying contexts. I could see the color of an object, but fail to taste it. I could see someone, but fail to hear her voice. But I must be able to ‘track’ the object and the person in situations where things change: say taste and not sight is available. As Millikan is saying, different representations are assigned ‘the same semantic value’ [p. 532]- they are representations of the same things [roughly, semantic value of a representation of object x = x].
A problem with this account that makes agents track/re-identify substances under various representations is that the connection between substances and agents might be too strong. That is, it might be hard to explain why agents are sometimes wrong in applying concepts of substances. Millikan’s answer is that our tracking capacities are generally reliable – because we use reliable indicators/cues – but not infallible – because cues can be misleading. We can take alcohol to be water if we fix on the substance just by using visual information/cues. But note that this is less likely to happen on multiple encounters. Next time we might taste the watery stuff – the mistake will surely be revealed.
Bottom line: don’t start by asking what a concept is, but what it is to have a certain concept. Millikan’s suggestion is that the answer to the second question is that to have a concept is to have a practical ability or skill – the ability to re-identify or track the substance throughout many encounters. If you find that convincing, it implies an answer for what a concept is: this very capacity of re-identification that makes use of a dynamic and evolving set of representations – see again the quote at the beginning of this section.
One problem to think about: if tracking depends on certain cues that are maximally – though not completely – reliable, what stops us from identifying the concept not with the ability, but with the description consisting of representations of those cues? Possible answer: the criticism of descriptionism put up by Krikpe and Putnam – a concept can survive a radical change in associated descriptions. But think of the example given by Millikan on p. 533 – when a child calls a tiger ‘kitty’ and then realizes that tigers are not cats, what is it that changes – what ‘cat’ means [i.e. the concept] or the cues used to recognize cats? Can we really maintain, as Millikan seems to, that the meaning of ‘cat’ is not affected?
I will not discuss section # 4-6, but I encourage you to read them.

3. Consequences

This is already too long, so I will just remind you an example we discussed briefly. What changes in re-branding and what are the expected – and the real consequences of such a procedure? Intuitively, what is meant to change is a description of, say, a product [it may be a symbolic object - whatever]. But if what guides action is not a specific representation, but an ability to see what is similar ‘through’ various representations, than we might need to think again. I will e-mail you a paper that seems to show that brand attributes [values etc – note these are all descriptive] don’t have a very strong impact in decision making @ consumers.

Summary of the 6th class by Ana

Below is the summary of the last class written by Ana. It’s an unedited draft for now, comments would be helpful. Please come back for the final version.

_____________

 
“Principles of Categorization” – Eleanor Rosch

On the 28th of March we discussed about the work of Eleanor Rosch called “Principles of Categorization”. In this seminar we first discussed how she begins the article with a non-conformist and quite strange example of categorization for the category “animal”
(extracted from an ancient Chinese encyclopaedia). This “type” of categorization proves that all the timed we categorize things, but nevertheless, not all systems of categorization are possible. A categorization means finding similarities and differences and introducing a certain object (or concept  ) in this or that “box”. But what happens if the box doesn’t exist? Can we categorize something in a category that doesn’t exist? Moreover, can we categorize something that doesn’t exist?
In my oppinion there is no such thing as a non-existent category. You create your own category, by the things you are familiar with. But the Chinese categorization was an counter-example. It seemed to me though that a stated category should exist, as it is stated/said/written. But this is another problem/matter that needs further and separate discussion.
Getting back to Rosch’s considerations and our seminar, we agreed that Rosch uses the Chinese categorization to demonstrate the fact that we do not have empty categories (empty “boxes”), as we must have a realistic view of (and over) the world. There are people and objects in the world that need to be organized, thus categorised. And because there is structure in the world, it helps us categorize (by making a correlation between principles and categories). In order for this correlation to happen, we need:
• the existence of the respective correlation;
• a knower;
• cultural categories.
The relationship between purposes defined by natural contextis based on the functional needs (with specific attributes).
Next, we stated the fact that the principles of categorization are based on the function of category systems and the structure of information. In such way, we therefore have the cognitive economy (with a minimum of effort and a maximum of information) and the perceived world (in which we categorize objects according to our perception – we admit the fact that there is a structure in the world).

The category system has 2 dimensions:
• a vertical dimension (associated to inclusion and hierarchies – one class includes another one based on a certain level of inclusiveness),
• a horizontal one (based on segmentation).
An example for this separation would be the scheme below:

But where do we stop? What level is basic? All this must be based on the perceived attributes (what we perceive in the world). But this is the same problem that we had with “features” (a problem of onthology – the onthology of prototypes). It is very important that we are culturally conditioned to see something. We first have the cathegory, then we have the list of features.
  Here we encountered the problem of typicality. Catgories must respect the maximality disctinction: the problem of cue validity (cues as features). The best cues are thise that work all the time (they have a maximum validity) – such as the concept “guils” has for the concept “fish”. Accordingly, “guils” has a cue validity of 1 for “fish” and a null cue validity for other objects/concepts.
 The same thing happens to the concept “animal”. The most typical concept (if we can put it this way) for “animal”, choosing between “dog” and “fish”, is of course “dog”. Why is that though? Do we use it more often? Do we hear it more often? Does a dog look more like an animal than a fish? Or do we associate a dog more with an animal than a fish?

Thus we reached the point of discussing basic level objects, which best mirror the correlational structure of the environment and is the most inclusive level. They have common attributes to most elements in the category they belong to. The basic level object comes to mind when people are asked what is the most representative object for a certain category. And it also is the first level in the lexicon of a language, a certain “label”. For example, kids first learn “dog” and then “animal”, because it is more intuitive this way.
 What is most interesting is that when people are asked to name the attributes of a category, they list the attributes of the basic level object (eg. Attributes of the category “vehicle”: four wheels, a stirring wheel, a gear box, color, dynamicity and so on, which are all attributes of the basic level object – “car”). This is explained by the fact that we identify these attributes quicker within the respective category.

Regarding the matter of context, basic level objects have a neutral context (such as a laboratory), perfect for experiments. For example, if we say that: “There is a dog on the porch”, we do not need to know the species or the name of the dog. But if we say that “There is a collie on the porch” or “a mammal on the porch” (here we “see” much less) we don’t have a neutral context, but a particular one.
 Here we encountered the problem of wrong categories, thinking of the phrase: “There is a whale on the porch” (  ). The concept “whale” is commonly thought as a fish, even though it is a mammal.
 Then we talked about prototypes and the role of objects. Prototypes are best examples for a category (they state the category membership) and it also is the point where most attributes meet. But of course, a prototype depends on the judgement – on the degree of typicality. But it is very important to state that prototypes do not exist: they are a “label”. They are the internal structure of a category and reflect in the mental representation (according to the last period from a list of 3 in Rosch’s work).
 There exists a conventional category and a non-conventional one (we can add to it; it can be a fictional or an existent category).
 Categories are not always labeled with a single word (there can always exist the category of “others”). Categories can also be exhaustive (but not always; although, why do we have the category of “others” if not for making a category exhaustive?).
 Returning to the basic level, we stated that we need it to go to the superordinate level (just like little children do) in order to understand and to perceive it better.
 In conclusion we said that Rosch made a reconsideration of the classical theory (based on necessary and sufficient conditions and on definitions – which put things and objects on equal footing; but there are more or less typical objects that surround us, such as central and periferal members  ). We also stated that the properties are in the members, but they are asscociated with the knower (that is an independet being – be it human, dog, parrot, monkey and so on). Thus, a certain categorization depends on various factors, that all need to be analyzed and of course, not skipped away.

 

exam questions

Questions for the exam on April 18th [there will be two more questions after the next meeting]. You will get two of these questions, and you will have one hour to answer them.

1. Discuss two arguments for the classical/definitional theory of concepts.
2. Discuss two arguments against [i.e. problems for] the classical/definitional theory of concepts.
3. Explain in a few words Frege’s distinction between sense and reference. Do you think the distinction is relevant for a theory of concepts? Why – or why not?
4. What is Frege trying to show when he uses the analogy involving the moon, its image on a telescope lens, and the retinal image of that? Does the analogy succeed in supporting the intended claim?
5. Could we apply or understand a concept with imprecise/vague borders? Think of Wittgenstein’s discussion of the concept GAME.
6. Explain in your own words Wittgenstein’s notion of family resemblance. What is the relevance of this notion for a theory of concepts? Please use examples.
7. What was the main motivation[s] for the introduction of the prototype theory?
8. Some of the arguments against the classical theory are also arguments against the prototype theory. Discuss one such example.
9. Please state and discuss briefly Rosch’s two principles.
10. What are prototypes according to Rosch?

list*2

Two discussion lists that might help you; I’m pretty sure most of you know the second, but here they are anyway:

Philos L

Romstudyabroad [RO]

next

No class this week, next one: April 11th.

Reading: Ruth Millikan’s ‘A Common Structure…‘. [It's 1st of April, but this ain't no joke folks.]

Summary of the 5th class

The Prototype Theory of Concepts (1)

Last Friday we discussed a few introductory matters concerning the Prototype Theory of Concepts.  This collection of theories called the Prototype Theory (PT) emerged in the ‘70s and was developed to address some of the difficulties in Classical Theory. The most significant problem for the definitional structure of concepts that PT tries to account for is the problem of psychological reality. The main tenets of PT are extrapolated from data gathered from psychological experiments. These data proved to be unaffected when studying concepts as definitions.
Here are the main ideas of PT that we mentioned in class:
- Concepts have a structure that encodes a statistical analysis of the properties their members tend to have. Hence, we can identify central or typical members of a category on the one hand, and less typical or peripheral members on the other hand. Concepts with prototypical structure –> their extensions as being graded: birdier, fruitier.
- The features of a concept are not necessary, as there may be items in the extension of a concept that fail to instantiate a certain property(ies). See the zebra without stripes or the 3-legged albino tiger examples.
- Classical Theory –> concept application must satisfy all of the concept’s features, PT –> must satisfy a sufficient number of features, some of which weigh more than others.
- Family resemblance: formal criteria are neither logical nor psychological necessity; open-ended set of properties; falling under the boundaries of a similarity space.
We didn’t say much about the advantages or the criticisms of PT (we’ll do this during the next meeting), but instead we tried to do an exercise.  We chose SCIENCE from a list of four other concepts, and we suggested a list of characteristic features for this concept’s prototype. Furthermore, we discussed some members of the category and tried to grade them for typicality. This task proved to be quite challenging, as you remember, and we had a hard time trying to grade the members we found. Now, let me suggest a rethinking of the exercise as follows:

	SCIENCE	PHYSICS	MATHEMATICS	MUSIC	LAW
Works with demonstrated data	Yes	yes	Yes	No	No
Is built on theories	Yes	Yes	Yes	Yes	No
Has a rigid structure  (is formalized)	Yes	Yes	Yes	No	Yes
Guides our perception of the world	Yes	Yes	?	?	No
Is innovative	yes	Yes	No	Yes	No
Makes predictions about facts and behaviours	Yes	Yes	No	No	No

re: Rosch’s “Principles of Categorization”

This is a classic text. Here are some things that might help you deal with it:

- first, the paper itself [you have it also in the reader]

- Prof. Rosch’s website

- in case you’re curious about the Borges quote, this will help

- btw, that list made quite an impression; please read page xvi of this book

visual note on Wittgenstein & more

A few pics taken yesterday in Vienna. To understand more, please take a look at this paper.

The house built by Wittgenstein for his sister, Margarete:

Cafe Central [gathering place in the early days of the Vienna Circle]

*** 

Alex had us read a poem by Rilke recently. Here is the German original again from the Rilke portal. And an English translation I found on Wikipedia:

I am so afraid of people’s words.
They describe so distinctly everything:
And this they call dog and that they call house,
here the start and there the end.
I worry about their mockery with words,
they know everything, what will be, what was;
no mountain is still miraculous;
and their house and yard lead right up to God.

I want to warn and object: Let the things be!
I enjoy listening to the sound they are making.
But you always touch: and they hush and stand still.
That’s how you kill.

Translated by Annemarie S. Kidder