Dear And and Mark,
Thanks for this interesting exchange. My view of Dick is that there is
something permanent running through my various appearances on the world
stage (though I have no idea how to push this idea back to me when I was
much younger). I imagine this is a fundamental part of human (and maybe
animal) cognition - think of all the evidence of children expecting
objects to reappear when hidden, etc. and recognising individuals. On
the other hand, as Mark says, my properties vary from time to time, and
the only way to reconcile these time-bound properties with this
permanence is to distinguish tokens of Dick from the type Dick. So I
agree with And that we think of it as the same Dick who comes into the
room and who leaves it, I also agree with Mark that I'm like an event
which is bound to a particular time; and even if a word is like a
person, we can still distinguish tokens (with time-bound properties)
from the (timeless) type.
Dick
And Rosta wrote:
> Jasp to me:
>
>> > This leaves me wondering why WG distinguishes between word types
>> and > word
>> > tokens. If the scene described above involves one and the same
>> Dick, > seen
>> > involved 3 different situations, why do we see "three boys met
>> three > girls"
>> > as involving 3 tokens of THREE? Why not say that one and the same >
>> concept,
>> > THREE, is involved in two different situations, first as the parent of
>> > "boys" and then as the parent of "girls"?
>>
>> One reason is that the two (I assume you meant two) tokesn can have
>> different properties: "three boys met threw [sic] girls".
>
>
> Right. But when Dick engages in a series of activities through the
> day, he
> has different properties in the different situations (hungry, then full;
> energetic, then tired; and so forth). But in neither ordinary English
> nor ordinary WG representations of knowledge do we represent this as
> different tokens of Dick. We allow that an individual can have variation
> in what properties they have. It strikes me as inconsistent not to treat
> words like individuals.
>
> Mark:
> [thanks for this very useful reply...]
>
>> And Rosta wrote:
>>
>>>
>>> Consider a scene:
>>> Dick enters the room.
>>> Dick picks up a sandwich.
>>> Dick takes a bite of the sandwich.
>>>
>>> We think of each segment of this scene as involving one and the same
>>> Dick, not different tokens of Dick. (Of course we can think of
>>> Dick-tokens, e.g. "Dick-in-the-1970s", "the Dick who entered the
>>> room was
>>> a much less hungry Dick than the Dick who had eaten the sandwich". But
>>> this is not our usual mode of thought when thinking about
>>> individuals) In
>>> other words we think of Dick as a Kind
>>
>>
>> It is because of this difficulty (Dick seems to be operating like a
>> kind,
>> but we really would like to think of Dick as an individual) that my
>> ontology treats (at an ultimate level) referents like Dick as events.
>> Events can have parts. Dick entering the room, his picking up a sandwich
>> and his taking a bite of the sandwich are all parts of Dick. Some very
>> complex events, such as Dick, are perceived as a trajectory of change
>> happening to a stable, identity-laden "object" -- the figure and
>> ground of
>> a gestalt.
>>
>> Under this ontology, Dick is not a class (or a kind): Dick is an
>> instance
>> (or an individual). Each sub-event that is part of Dick is likewise an
>> instance (any part of an instance must itself be an instance).
>
>
> OK. The only thing I'm unclear about is whether "my [your] ontology"
> means (a) your ontological model of the universe or (b) your model of
> the (anglophone?) human mind's ontological model of the universe.
>
>>> (A kind is the entire membership of a category abstracted into a single
>>> exemplar.)
>>
>>
>> That seems rather unintensional.
>
>
> It was intended to be explanatory rather than definitional. In the
> vocabulary introduced in your reply, I mean, by Kind, a category
> conceived of not as a set but as an individual with the properties
> that would belong to members of the set.
>
>>> This leaves me wondering why WG distinguishes between word types and
>>> word
>>> tokens.
>>
>>
>> It's one of the things I like best about WG. I think it distinguishes
>> between types and tokens for the same reason formal ontologies and
>> programming languages distinguish between classes and instances: in
>> order
>> to capture generalizations in one's descriptions. Otherwise, all you
>> really have (empirically) is a pile of tokens.
>>
>>> If the scene described above involves one and the same Dick, seen
>>> involved 3 different situations, why do we see "three boys met three
>>> girls" as involving 3 tokens of THREE?
>>
>>
>> I only see 2 word tokens of THREE. What have I missed?
>>
>>> Why not say that one and the same concept, THREE, is involved in two
>>> different situations, first as the parent of "boys" and then as the
>>> parent of "girls"?
>>
>>
>> I'm not sure I'm in the same headspace as you, because the question
>> doesn't make any sense to me.
>>
>> "three boys met three girls" has two tokens of THREE because the
>> utterance
>> contains two instances of [Tri:].
>
>
> My understanding of WG is that it treats classes as individuals. That
> is, WG's ontology has only individuals. These are related by Isa, which
> in Langackerian terms means something like "schematic elaboration" (I
> can't face opening my Langacker to check this; am crossing my fingers
> & hoping I'm right).
>
> So my point is that the single individual Dick can participate in two
> different situations, such that in each situation he has properties
> he doesn't have in the other, why can't words be seen in exactly the
> same way? (E.g. THREE in the example I gave.)
>
> It seems inconsistent to treat Dick and THREE differently. If there
> are two tokens of THREE in the sentence I gave, then there should be
> multiple tokens of Dick. Or in terms of your ontology, if there are
> multiple event-segments of Dick, then there should be 2 event-segments
> of THREE in the sentence.
>
> Furthermore, the evidence of English is that although we can
> conceptualize individuals in terms of their instances/tokens/segments,
> we prefer not to (i.e. we tend to say "Dick" rather than "a Dick",
> which is very much the marked option). So one might extrapolate a
> preference for not distinguishing word types from word tokens either.
>
> Mark in another message:
>
>>> As you might have inferred, the Clark/Superman case strikes me as
>>> analogous to cases where there are multiple tokens of the same type.
>>> And
>>> hence the type/token distinction is not the same as Isa.
>>
>>
>> Of course it's not the same -- it doesn't even relate the same kinds of
>> things.
>>
>> Isa relates a class to another class (one of its superclasses).
>> Type/token
>> relates an instance to a class (to which it belongs).
>>
>> I remember having an earlier discussion about this with Dick and Jasper,
>> in conjunction with the Babbage simulator. We decided that, in practice,
>> it's okay for our 'isa' link to ambiguously represent both "real" isa
>> (class to superclass) as well as the token/type relationship
>> (instance to
>> class) because we explicitly represent token nodes as distinct from type
>> nodes. If we're ever called upon to do any processing that is
>> sensitive to
>> the difference (which I anticipate), we'll have the information we need
>> that way.
>>
>> Most formal ontologies actually do both (instances and classes are
>> represented distinctly, as are the kindOf and instanceOf relationships),
>> but that's an unnecessary redundancy in terms of our network processing.
>
>
> Certainly in the last piece of heavy-duty strictly-WG work I did (my PhD
> a decade ago), I distinguished SubclassOf from InstanceOf. But as I've
> said here and there in recent messages, in those days I wasn't taking
> seriously the cogntivist/mentalist dimension of WG. Now I am taking that
> dimension seriously; & I understand one of WG's substantive claims to be
> that knowledge can be modelled using only Individuals and what is called
> 'Isa' (= schematic elaboration). I.e. no classes, no SubclassOf. 'Isa'
> is claimed not to be ambiguous, from a cognitive perspective. It's clear
> that this move throws up lots of serious logical difficulties, but at the
> same time it has a considerable degree of prima facie cognitive
> plausibility (prototype theory, schematicity instead of Aristotelean
> classes; behaviour of mass nouns & generics).
>
> So why I'm worrying away at this issue is that I don't want to ditch
> WG's model of the formatives of conceptual structure unless I'm convinced
> that it cannot be an accurate model of cognition.
>
> Since you have a background in Strat Grammar & Cognitive Lx more broadly,
> I'm especially eager to discover your views on these issues.
>
> --And.
>
>
--
Richard Hudson, FBA,
Emeritus Professor of Linguistics,
University College London
www.phon.ucl.ac.uk/home/dick/home.htm
|