topics: |  cognitive-psychology | language | categories | semantics

Cognitive and Linguistic models of "category"

The categories and types that we isolate from the world of phenomena we do not find there because they stare every observer in the face; on the contrary, the world is presented as a kaleidoscopic flux of impressions which has to be organized by our minds -- and this means largely by the linguistic systems in our minds. We cut up nature, organize it into concepts, and ascribe significances as we do, largely because we are parties to an agreement to organize it in this way -- an agreement that holds throughout our speech community and is codified in the patterns of our language - Benjamin Whorf [1940/1956, 213-4]

Because my mother tongue is English, it seems self evident that bushes and trees are different kinds of things. I would not think this unless I had been taught that it was the case. - Edmund Leach (anthropologist), 1964

That all categories are merely learnt cultural artifacts - Intuitively rejectable. Colours: reality as a continuum - is self-evident.

human visual system: discriminates 7.5 mn colours

[Bloomfield 1933]: Colour-names of different languages do not embrace the same gradations...

Structuralist view of word meaning

- Language as an autonomous, self-contained system (structure): "La Langue et
  us syst\`eme dont tous les termes sont solidaires et o\`u la valeur de l'un
  ne r\'esulte que de la pr\'esence simultan\'ee des autres" - [Saussure
  (1916/1964, 159]

- concepts are purely differential: "non pas positivement par leur contenu,
  mais negativement par leurs rapports avec les autres termes du systeme"
  [p.162; not positively, in terms of their content, but negatively by
  contrast with other terms in the same system (tr. Harris 83)]

COLOUR : structuralist view:

- all colour terms have equal status.  Frequencies can vary, but since they
  are all defined w.r.t. the entire system
- all referents of the term have euqal status.  Two colours categorized as
  red - no sense in saying one is redder than the other.
- only object of linguistic study is the system, not indiv terms.

Colour cognition: Berlin and Kay 1969

Ninety-Eight languages surveyed, based on speakers

BASIC COLOUR TERMS:
- are not subsumed by other terms - e.g. crimson and scarlet are varieties of
	red
- are morphologically simple.  bluish, golden etc excluded
- not collocationally restricted, e.g. "blond" - used only hair
- are of frequent use (exclude "puce" or "xanthic")

Claims:

1. FOCAL COLOURS:
   when people across cultures are asked to pick good exemplars of the basic
   colour terms, cross-linguistic variation largely disappears.  Although
   range of colours differ for what is "red", a "good red" is largely the
   same.

   [By and large substantiated, though Russian appears to have twelve
	focal colours, etc. ]

2. IMPLICATIONAL HIERARCHY: (more controversial)
   All the 98 languages appear to select basic colour terms from an inventory
   of 11 colours.  If a lang has only two colour terms (no lang has fewer
   than two), then these will be focal black and focal white.  If there is a
   third term, it will always be red.  The fourth term will be yellow or
   green. etc. :

         black            yellow                      grey, orange,
         white  < red <   green   < blue < brown  <   purple, pink

   i.e. if a term exists for a colour to the right, then it exists for all
   colours to its left.

Only terms to the left undergo productive derivation (blacken, whiten,
redden, but *bluen, yellowen); similarly -ness is less likely to the right -
blueness, but *purpleness, *orangeness.
Frequency: black, white, red, brown, blue, green, grey, yellow, pink, orange,
	   purple [Kucera and Francis 67 - Amer Engl computational survey]

Controversy: For 20 of the languages, the speakers were in the San F area,
and all bilingual.  Remaining 78 are even more suspect - mostly from
dictionaries, some of them anthropologists' reports dating back to the 19th c.

ElEANOR ROSCH / (after-marriage HEIDER)

[Heider 72] : Four experiments: COGN

1. prototypical colours ( centers ) are stable across languages (e.g. "good
   red").

2. focal colours are faster in naming experiments
   ==> suggest that focal colours have greater perceptual an cognitive
   salience

speed with which colour is named:
fastest: Black - then yellow, white, purple, blue, red, pink, brown, green,
and orange. --> not related to Implicational hierarchy.

Some langs without sep colours for blue/green may have others to the right -
e.g. Tsonga has a grey [7 basic colours - 1. ntima: black, 2. rikuma:white,
3. basa:white-beige, 4. tshwuka:red-pink-purple, 5. xitshopana:yellow-orange,
6. rihlaza:green-blue, 7. ribungu:dark-brown-dull-yellow-brown].  Zulu, like
most Bantu lgs, does not distinguish blue/green - but has a term for focal
brown.

Terms for green-blue (Kay/MacDaniel call it grue) - are bifocal - ie. they
have both to focal blue and focal green, rather than just one focal colour.

3. Dani/English experiment COGN

20 speakers of English, and 20 of Dani (Papua/NG lg, with only two colour
terms, mola = focal white + warm colours (red, orange, yellow, pink, purple),
and mili = focal black + cool colours (blue, green).

Task: shown a colour sample for 5 seconds.  After 30 secs, have to identify
the colour from a colour array.

Results: Engl speakers could recog colours more accurately than the Dani.
However, for the focal colours, Dani performed better than for non-focal.
This latter is unlikely to arise due to linguistic effects.

4. Expt 4: long term trainability.  When taught to recog colours, Dani were
   more adept at learning focal colours.

Also [Heider 71] -
- 3/4 year-olds were better at matching focal colours than non-focal
- 3-year olds without the full range of colour vocab - were
  more attentive to focal than to non-focal colours; and also

PHYSIOLOGY OF VISION
beyond the RGB cones of retina, colour vision propagates as:
green/red and yellow-blue --> excite complimentary neural structures.
Along with black/white - these may be more primitive than orange, purple,
etc.

[The evolutionary history of human colour vision is that the red-green
separation was the result of a fairly recent mutation.  That is why
the red and green wavelenghts differ by about 0.1 octave out of the
entire visible range of 1 octave). - (Robert Pollack: Missing Moment, 1999). ]

colour terms - often originate in natural objects - e.g. orange, pink,
violet, burgundy, lime.

--
   IMPURITY OF LANGUAGES: The actual systems called 'languages' in ordinary
   discourse are undoubtedly not 'languages' in the sense of our
   idealizations... [They] might ... be 'impure' in the sense that they
   incorporate elements derived by facilities other than the language
   faculty. - Chomsky 80, p.28

Cog grammar view

   Can we treat language as an independent subject of study? Is there a
   legitimate science of words alone, or phonetics, grammar, and
   lexicography?  Or must all study of speaking lead to the treatment of
   linguistics as a branch of the general science of culture? [Malinowsky
   1937:172]

[Rudzka-Ostyn 88, Topics in Cog Ling, Benjamin, Amsterdam] - four chapters
contributed by Langacker - offer an accessible intro to his ling theory.

Ch 3 Prototype 1

Classical models not enough: Wittgenstein

Labov: cups vs bowls - attributes are not discrete.  Even whether it has a
handle or not can be thought of as a matter of degree - from a full-fledged
loop to a little stub. [44]

[ROSCH 75] cog rep of sem cats, J Exp Psyc 104:192-233
To what extent, on a scale of 1 to 7, would the following be regarded as a good
example of the category "furniture"?  [asked of 200 american college students]

Rank
1.5 chair
1.5 sofa
3.5 couch
3.5 table
5  easy chair
6 dresser
6 rocking chair
8 coffee table
9 rocker
10 love seat
11 chest of drawers
12 desk
13 bed
..
22 bookcase
27 cabinet
29 bench
31 lamp
32 stool
35 piano
41 mirror
42 tv
44 shelf
45 rug
46 pillow
47 wastebasket
49 sewing machine
50 stove
54 refrigerator
60 telephone

No difference between natural kind categories (bird) and artefacts (toy,
vehicle).  Also diff to define a set of features, even for the top
30 members...

Psych:
Degree of membership affects responses to qs like "an X is a Y' - quicker on
"robin is a bird" than for duck.

Priming: when primed with the superordinate category the better exemplars are
activated.  e.g. subject has to identify if two words are the same or not.
Before flashing chair-chair, if "furniture" is flashed, it is quicker, than,
e.g. stove-stove. [Rosch 75]

When asked to name a few exemplars, the more highly ranked items come up.
Prototype effects also for abstracts: to what degree is this narrative
an instance of telling a lie? [Coleman/Kay:1981], verbs like look,
kill, speak, walk [Pulman:83], adjs like "tall" [Dirven/Taylor:88].

Hierarchies

Basic level category.  What is the object I am sitting on?  "A chair".
Not a "furniture".  Draw a furniture - not possible.  chair is
countable, monomorphemic.  subordinates - "desk chair" often composed
with this word,
superordinates often strange, e.g. "*a furniture", "*furnitures".

In German superordinate terms Tier (animal), obst (fruit), gem\"use
(vegetable), and metalal are all neuter, while specific animals etc
are generally masc or fem.

Pulman: verbs: categories -

Do - cause -  kill - murder
     make     cook   execute
     become   boil   assassinate

 ----------------------------->  degree of specificity

artefact - furniture -     chair -  dining chair
	     tool            table    dentist chair
	     dwelling place  bed      kitchen chair

attribute discrete-feature systems fail to capture the notion of a
privileged level in the hierarchy.

why basic level? because they cut up the perception world into
maximally informative categories.  They
 - maximize the num of attributes shared by members of the cat
 - minimize the num of attribs shared with other cats

Basic level concepts

(Ungerer 1994: 194)

* commonly expressed by words which have a simple
  morphological form, which first come to mind and which are first learned by
  children (such as dog);
* basic level concepts have a large number of attributes which are shared by
  category members (e.g. by various types of dog) and are at the same time
  distinct from attributes of other basic level concepts (e.g. 'elephant' and
  'bird');
* basic level concepts are the prime examples of the prototype-cum-periphery
  structure, i.e. they comprise good and bad examples, with the prototype
  functioning as a model for the categorization of the other category
  members;
* members of basic level concepts (e.g. all kinds of dogs) have a
  characteristic overall shape that is readily identified as such; this seems
  to favour holistic (or gestalt) perception;
* basic level concepts refer to objects and organisms which stimulate typical
  actions or motor movements (chairs are connected with the action of sitting
  down, etc).