topics: |  philosophy | linguistics | linguistics

Ten essays that have appeared in various journals and books. Much of the work deals with the notion of convention in language, but more generally about the prescriptive normative rules, the partial regularities underlying language.

The first chapter analyzes the notion of convention, and establishes some conditions for "convention" to hold - such as reproducibility - However, I found the chapter on Why (most) kinds are not classes, where she develops the notion of historical continuity, or eternal essence that defines a category, as opposed to any notion of present-day similarity etc - this turns out to be rather similar to the biological norms used to define a species (which turn out to be historical as well). Some more details of the argument.

Chapter 1: Language conventions made simple

Much of this essay is an analysis of the process of convention, and draws heavily on the work of Lewis; particularly "Convention: A Philosophical Study" (1969), based on his phd, at the start of which he says: "It is a platitude that language is ruled by convention", and proceeds to to give us 'an analysis of convention in its full genearlity, including tacit convention not created by agreement. In this work, Lewis uses concepts of game theory to analyze the nature of social conventions, and claims that [wiki:]

   social conventions, such as the convention that one is
   driving on the right (not on the left), the convention
   that the original caller will re-call if a phone
   conversation is interrupted, etc., are solutions to
   so-called "'co-ordination problems'".  Co-ordination
   problems were at the time of Lewis's book a much
   under-discussed kind of game-theoretical problem; most of
   the game-theoretical discussion had circulated around
   problems where the participants are in conflict, such as
   the prisoner's dilemma.

[Of course, these solutions also exhibit Nash equilibrium properties - AM]

Requirements for being conventional

This lays out an approach for tackling these based on game-theoretic notions.
Lewis' analysis is very complex, calling for a number of factors, e.g. in a
driving context, both agents must be aware of the keep-to-the-left rule,
and both must know that the other knows that she knows that the other knows
etc.

Millikan tries to simplify these situations where conventions arise in terms
of two requirements - that it should be possible to imitate it (reproduce) in
a hand-me-down manner; and that this is how it "has been done" (weight of
precedent).  In addition she outlines several types of coordination situation
- e.g. the leader-follower where one person follows the cues of the other,

A. Reproducibility:

[ RM uses the word "reproduce" or "handed-down", but imitative maybe clearer,
  or does this alter RM's point altogether? ]

considers conventions like "shirt-buttoning" - if it is efficient to
button it from top downwards, such behaviour may have been independently
discovered by a population, but would not be "handed-down" hence it would not
be conventional.  Thus to be conventional, a behaviour must be imitated.
Several ways:

Conventions may reproduce by being told.  An example of a 2-step conventional
   pattern reproducing by telling arises when Johnny's mother tells him that
   he is to put his letter in the mailbox and put up the flag.  The result is
   that he reproduces the first part of a conventional pattern of activity,
   the second part of which will be reproduced by the mailman, who, on seeing
   the flag, will pick up the contents of the mailbox, and then put the flag
   down.
Counterpart reproduction: Arises when a bolt and a nut are to have the same
   gauge. e.g. positions assumed by men and women in traditional ballroom
   dancing.  "Copies of copies may drift away, but the need to fit
   counterparts retards drift." 5
Unconscious reproduction: e.g. interpersonal distance - if too small, one
   moves back, but it may be reinforced by the need not to circumambulate
   around the room.

B. Weight of precedent

Patterns are conventional only if there are other patterns that could have
been substituted, given other historical accidents.  Quotes bird songs as
being non-conventional but arbitrary.
  [but this is perhaps belied by other evidence, see, e.g. Meredith West's
  study of cowbird song across different ecotopes in the continental USA: NC
  males can learn TX songs, IN males can learn SD, can become bilingual - NC
  males could learn TX song from TX females who can't themselves sing it...
  so the males can adapt to a contingency.  But is it the females here then
  who become "standard-bearers"?  Is this yet another instance of
  philosophy being undermined and rendered pointless by empirical data? ]

weight of precedent makes a convention socio-economically stable - e.g. forks in
the west and chopsticks in the far east.

Coordination conventions:

Here the action requires a number of participants to coordinate, but each
person's action is not clear a priori.  E.g. moving a sofa, or re-dialling
after a phone call gets cut, or the doorway do-ci-do.

Conventions are not regularities

Conventions are not mandatory - not everyone hands out cigars after a son is
born.

Blind coordination: both people do something w.o. waiting to see what the
other is doing - e.g. driving on the left.

leader-follower conventions: e.g. while moving a sofa, one becomes a leader
and the other coordinates to follow.
[or does the follower consider all the plans L may have, considers what L's
current action fits into, and then adopts the one that may be best suited to
the overall goal? RM seems to think that this would be possible only if the
overall process was also "traditional". But isn't this also possible for a
novel algorithm, say? ]

Lewis: The conventions of language are truthfulness [say only what is true] and
trust [believe what you hear].  RM feels these rules elliptically describe
conventions, but may not describe "regularities"
  [but I think Lewis's argument is more strong.  The majority of speech acts
   are underlied by such assumptions; and these are conventions, because we
   too get to follow them after being exposed to it... ]

Combining conventions: RM's uses the term crisscrossing conventions - Lg
  is a "tangled jungle of overlapping, crisscrossing traditional patterns,
  reproducing themselves whole or in part for a variety of reasons. - and am
  I wrong to think that she is really discussing the act of combining
  conventions - if so, how is it different from, say the combining of
  constructions, or "conceptual blending", say?

The notion of combining conventions seems rather similar to the idea of
combining concepts... as in Langacker, say, or in Fauconnier's Conceptual
Blending.

[NOTE: convention as probability - high-probability events are coded more
compactly and patterns for it are more easily accessed - e.g.
"conventional moves in chess" - expectations are already known and bayesian
rules building on these are already available.

A convention in this probabilistic view, arises not by maximizing expectation
alone (which would maximize benefits, and would indicate the absence of
choices) but because it has already been analyzed, and the computations
required to compute expectations are known.  Thus the benefits are not in the
search space per se, but in the meta-space of computational cost.

bio: David Kellogg Lewis

 ... a metaphysician and a philosopher of mind, language and
 logic at Princeton University, died on Sunday at his home in Princeton,
 N.J. He was 60.  ...  Mr. Lewis was once dubbed "a mad-dog modal realist"
 for his idea that any logically possible world you can think of actually
 exists. He believed, for instance, that there was a world with talking
 donkeys.  ...  Born in Oberlin, Ohio, on Sept. 28, 1941, Mr. Lewis was
 educated at Swarthmore College, Oxford University and Harvard, where he
 earned his doctorate, under the supervision of W. V. Quine, in 1967. He
 taught philosophy at U.C.L.A. from 1966 to 1970, and then at Princeton
 University, where he became the Class of 1943 University Professor of
 Philosophy. He was a fellow of the American Academy of Arts and Sciences and
 received honorary degrees from the University of York and Cambridge
 University, in England.		[obituary NYT: October 20, 2001]

Chapter 6: Why (most) kinds are not classes

[More complete argument: On Clear and Confused Ideas (Millikan, 2000) = OCCI]

hundreds of experiments by psychologists have tried to discover how people
`classify' or `categorize' items under kind (category) words such as dog,
chair, and fruit.

such words do not designate classes but rather units of an entirely different
kind. A very few, uncompounded nouns that designate classes do exist, but
words like dog, chair, and fruit are not among them.

Biological Species

What is a biological species?

Biologist MT Ghiselin
(1974, 1981) and philosopher David Hull (1978):
   to be members of the same species, individual animals must belong to
   historical lineages that have a common origin. They do not have to be
   similar to one another in any specified way.

All dogs do not share any feature - not a gene - every dog gene has
alternative forms (alleles).  Similarly no properties or resemblance.

Two groups of highly similar animals, if they have different historical
origins, form multiple species.
    species are not classes that share a similarity, but rather are big,
    scattered, historical individuals enduring through time. [GH]

However, usually the historical relationship ("blood ties") are assoc with
considerable similarity in properties.

Homeostasis:
Each gene must help produce a viable individual frequently enough so that it
can get itself reproduced sufficiently frequently so as not be eliminated
from the gene pool.  Diff individuals in a species resemble one another in a
great variety of ways, but do not all resemble one another in any particular
ways.

What pulls them together as a group is not just that they have common or
overlapping properties, but that they tend to have common and overlapping
properties for a good reason [causality: they are caused by genetic
overlap]. ...

On the other hand, classes (categories) are defined by the members having
certain common properties. Fuzzy classes may be defined by the members having
overlapping properties, or by their having many properties in common with a
paradigm, or paradigms.
But the members of a class don't have to be causally like one another; they
may be so by accident.

Species are not categories

If there is a reason why one dog is likely to be similar to the next dog in a
good number of respects, then there is a reason why studying one dog is
likely to yield a considerable amount of probable knowledge about the next
dog. ... all this knowledge is merely probable knowledge. Whatever one learns
about the properties of dogs, it will not be analytic or necessary that every
individual dog has each of those properties.

  [WN: analytic: of a proposition that is necessarily true independent of
  fact or experience; "`all spinsters are unmarried' is an analytic
  proposition")

E.g. consider the class "red triangles".  If one member is sweet, it doesn't
say much about any other member, either necessarily or probabilistically.
This is because they are not causally related.

Possibility of Inductive knowledge:
The value of being in a class comes from the ability to predict that one
indiv will be like the next, so we can obtain knowledge without examining
every member separately.

[The knowledge generalizn argument may not extend the way it is proposed
here.

Probabilistically, if K members out of K tasted in class "red triangle" is
sweet, we can construct an estimated prob 1 that all red tris are sweet.  The
reliability of this estimate would depend on the sample size N etc.  It is of
course possible that just these K were sweet, and if N>>K, then N-K/N is
~ 1, so that red-tri's are generally not sweet.

This latter would also be possible for some class like dogs.  Say out of 1000
dogs, 10 are blue.  Then if these are the ones I see first, I may have a
wrong notion of dog colour.  Thus inductive knowledge about dog colour and
red-tri sweetness both may be impossible.

So, just as in any class, only some attrib's are similar, so also in a
species. ]

If empirical knowledge about a species is inductive, it would be useful to
refer to it as a group, and it could acquire a name.

Any group where one individual within the unit, or one part of the unit, is
liable to be like another, can be called "substances," (from Aristotle).
Non-philosophers may need to read this as a new technical term, but
philosophers may recognize it as fairly traditional.

3. Three kinds of (Aristotelian) "substances"

Aristotelian "substances" fall roughly into at least three basic sorts, which
I call "historical kinds," "eternal kinds," and "individuals."

3.1. Historical kinds

A "historical kind" (like dogs, for example) is a collection of individuals
scattered over a definite spatiotemporal area.

Furthermore, the individuals are causally
related to one another in such a manner that each is likely to be similar to
the next in a variety of aspects.
[because
1. something akin to reproduction or copying has been going on;
2. members have been produced by, in, or in response to, the  the same
   historical environment;
3. an ubiquitous causal factor often supporting the first factor- some
   "function" is served by members of the kind, where "function" is
   understood roughly in the biological sense as an effect raising the
   probability that its cause will be reproduced.
]

Artifacts are often good examples of this - e.g. chairs.

Designed for ergonomic and aesthetic preferences of humans, who are
themselves much alike (for the same reasons dogs are).
Moreover, the design of a chair is invariably
influenced by the design of prev chairs (partly because the prev designs fit
the culture) - hence chairs form a rough "historical kind."

There are historical reasons, which have nothing to do with any arbitrary
points of definition, why one knows roughly what to expect when someone
offers to bring a chair.

Musical renditions of a folk tune ... form historical kinds. The
renditions are copied from other people, or from scores that had been
transcribed from earlier renditions or had been copied from earlier
scores.
McDonald's restaurants form a historical kind.  There are historical,
cultural causes for their being so much alike.

Professors, doctors, and businessmen form historical kinds that are
especially well integrated when the groups are studied in particular
historical and cultural contexts.  Individual professors, doctors, and
businessmen are likely to act in similar ways and to have attitudes in common
as a result of: similar training handed down from person to person
(reproduction or copying), and/or custom (more copying), and/or natural human
dispositions (compared to dog dispositions), and/or social pressures to
conform to role models (copying again), and/or legal practices handed down
from univocal sources.

There is a reason why it may be productive to investigate, for example, "the
attitudes of American doctors toward acupuncture".  These attitudes are
contagious. They spread.

3.2. Eternal kinds

... are alike because of a common inner nature of some sort, such as an inner
molecular structure, from which the observable properties are determined.
e.g. chemical elements and compounds, or H20 which is ice,
liquid water, or steam.

Stars, planets, comets, asteroids, and geodes are eternal kinds, not because
their properties flow always from exactly the same inner nature, but because
they were formed by the same natural forces, in the same sort of
circumstances, out of materials that are similar in relevant ways.

  [Yet one atom is quite dissimilar from another, in velocity, position,
  energy, mass, etc. Water, ice and steam - how are they really similar?
  Could this also not be historical?  Why
  would stars not be historical, once we understand their history?
  w: The SUBSTANCE is that part of water that is unchanged when it turns from
  water to ice; it is the "bare particular", without which it would not be
  water. It is "bare" because it is considered without its properties and
  "particular" because it is not abstract. The properties that the substance
  has are said to inhere in the substance.  Arguments for substance theory
  include "Arguments from Grammar" (Snow is white), and Arguments from
  conception.  ]

3.3. Individuals

The last kind of (Aristotelian) substances are individuals.  Ghiselin and
Hull claimed that species are actually individuals, because they are held
together not by a traditional essence, but rather through historical causal
connections.

Like species, properties of individuals are stable.  ...
A species is a "homeostatic system .... amazingly well-buffered to resist
change and maintain stability in the face of disturbing influences" [Eldredge
and Gould (1972, p.114].

If a woman is tall, brownhaired, knowledgeable about electronics, and a good
piano player today, it is likely, though not certain, that she will also
possess these traits tomorrow.
Individual objects are things that inductive knowledge can
be collected about over time; similarly, a historical kind and, more broadly,
an eternal kind are things that knowledge can be collected about over time.

[I am not sure this third kind is sufficiently different from a historical
kind.  A living person (or an artifact) is changing from day to day, some
atoms are sloughing off, some others are building up.  But something
continues to be similar to what it was before; this is perhaps a
causal historical continuity not sufficiently diff from "historical
kinds".  The woman-of-yesterday and the woman-of-today are two different
entities, but historically (and therefore causally) connected).    ]

Concepts

Thus: historical kinds, eternal kinds and individuals - three
basic kinds of (Aristotelian) substances - are similar with respect to why it
is possible to gain inductive knowledge about one part of such a
cemented-together unity from other parts.

... there is no central set of properties, all or some of which one must be
able to think of, recognize or discriminate in order to think of the
(Aristotelian) substance dog, in order to learn about dogs, to understand
things said about dogs, and so forth.

Twin-Earth Scenario: Hilary Putnam

As an aside, this discussion reminded me of Hilary Putnam's famous twin earth
scenario - if our mental categories can't be sure about identity, how can we
relate to "kinds"?

    We begin by supposing that elsewhere in the universe there is a planet
    exactly like earth in virtually all respects, which we refer to as ‘Twin
    Earth’. (We should also suppose that the relevant surroundings of Twin
    Earth are identical to those of earth; it revolves around a star that
    appears to be exactly like our sun, and so on.) On Twin Earth there is a
    Twin equivalent of every person and thing here on Earth. The one
    difference between the two planets is that there is no water on Twin
    Earth. In its place there is a liquid that is superficially identical,
    but is chemically different, being composed not of H2O, but rather of
    some more complicated formula which we abbreviate as ‘XYZ’. The Twin
    Earthlings who refer to their language as ‘English’ call XYZ
    ‘water’. Finally, we set the date of our thought experiment to be several
    centuries ago, when the residents of Earth and Twin Earth would have no
    means of knowing that the liquids they called ‘water’ were H2O and XYZ
    respectively. The experience of people on Earth with water, and that of
    those on Twin Earth with XYZ would be identical.

    Now the question arises: when an earthling, say Oscar, and his twin on
    Twin Earth say 'water' do they mean the same thing? (The twin is also
    called 'Oscar' on his own planet, of course. Indeed, the inhabitants of
    that planet call their own planet 'Earth'. For convenience, we refer to
    this putative planet as 'Twin Earth', and extend this naming convention
    to the objects and people that inhabit it, in this case referring to
    Oscar's twin as Twin-Oscar, and twin-earth water as twater.) Ex
    hypothesi, their brains are molecule-for-molecule identical. Yet, at
    least according to Putnam, when Oscar says water, the term refers to H2O,
    whereas when Twin Oscar says 'water' it refers to XYZ. The result of this
    is that the contents of a person's brain are not sufficient to determine
    the reference of terms they use, as one must also examine the causal
    history that led to this individual acquiring the term. (Oscar, for
    instance, learned the word 'water' in a world filled with H2O, whereas
    Twin Oscar learned 'water' in a world filled with XYZ.)