American Patheopele sist
‘ognition and Categorization. Eleanor Rosch
nd Barbara B. Lloyd, eds. Hillsdale, NJ.
awrence Erlbaum Associates, 1978. viii + $28
1p $16.50 (cloth).
Eugene S. Hunn
University of Washington
This collection documents an intriguing in-
stsection of anthropological and psychological
tsearch and theorizing. The anthropological
put is represented by Berlin's lead article, an
troduction to his “general principles” for the
gnitive peychologists and paycholinguists at
snding an SSRC conference on the nature and
Finciples of category formation, organized by
ne editors of this volume. The volume docu:
vents the conference proceedings. The ethno:
ence tradition represented here by Berlin is
ognized at key stimulus to the development
fa "new paradigm of reference” in cognitive
sscarch, the central tenet of which is that
categorization, rather than being arbitrary,
tay be predicted and explained” (p. vi).
‘The cognitive anthropological influence is
vost direct. in Rosch’s “natural category”
ssearch, Berlin's recognition of the special
atus of folk generic taxa in folk biological
lasification spurred Rosch's experimental
cumentation of a "basic level of abstraction’
Thuman thought defined a8 “the most in-
sive -- level at which the categories can
fF the structure of attributes perceived in
world” (p. $0). Basic-level categories are not
tricted to plant and animal taxa, such as oak
‘accoon, but may aio include such man-
ide objects as chairs and pianos.
Basic level categories are distinguishable in
perican Sign Language as well as is shown by
‘port and Bellugs In ASL, basic-level
fegories contrast with both superordinate and
ordinate categories by the use of primary
‘ns for the former and various compound sign
2es for the latter
Basic level categories are not constructed by
‘erence to 2 perceptually arbitrary set of
fining feacures (4 la componential analysis)
xtead. the members of such categories have
any comimon features and a high proportion
features not shared by directly contrasting
tegories. They are “polythetie concepts” with
gh information content, and thus efficiently
ode the structure of the world perceived. In-
inces of sich categories are recognized not by
ference to an analytical rule but by compar
in with a prototype of some sort.
‘Tversky and Gati summarize experimental
search demonstrating that perceived similar
the proces underying any comparion of
tens with prototypes is non-Euclidean, thus
porting the authors’ feature:contrast model
‘mnilarty-difference judgments in which "che
hilarity ofa to & is deseribed as a linear com:
By TTS CD
ination. of the measures oftheir comson
nd diaincve features (p. 60). Their demon:
Strain resus on evidence that similarity i not
the precae comers of diference and tht sm
tasty judgments are not aymeteical and ate
Gamer dicusion ares the conan
imposed on the clair response by the sue:
tang of the sialan, He erties the Tack of
Searenes among experimenters in peychology
Of ceria hey dsincions relative to. the
Stimulus, Le “component” verus "whol
ropertcs af a aimulus, the former either
Ffeltares” defined here ab inherently single
valued thes. present or abint—oF “dimen
oe” by conan, muvalued. Garner's
fection of emergent, whole, configual
Stimulus properties quite unexpectedly clarifie
S poine wll hanging in the ehnoscence Tera
tafe, that i he sence of reciprocity" a8 8
fearase of the meaning of Englah in terms
(eg. a in undlenepheve, sce Romney and
D’Andrade, 44 66, No. 8, Part 2, 146-170,
1964), despite the logical priority of underlying
feavures such as “generation” and “lineality.”
Garner notes that emergent configural attr
Dutes (read “reciprocity") are necessarily
logically nonindependent of component proper:
tice (read “generation,” “linealty”) but will be
preferred to-component properties in rapid cog’
nitive procesing
‘Shepp distinguishes the use of “integral” and
“weparable” dimensions in constructing classifi
ceations. He argues that integral dimensions
(eg., “red” as hue + brightnest) “lead to
redundancy gain when correlated and... pro-
duce interference . .. when the [underlying]
dimensions are orthogonal.” under speed of
classification constraints (p. 135), whereas
separable dimensions (e.g.. "red square” as col-
‘or + shape) do not. This allows the operational
‘differentiation of these two grouping strategies
in experiments. He found that the use of separ
able dimensions inereased with age and/or
‘schooling. Curiously, this increased use of
separable dimensions in category formation
reflected not 10 much increased cognitive capac:
fty—for the youngest were capable of construct
ing both kinds of groupings~but. rather. a
changing preference for the separable of
analytical approach, perhaps asa by-product of
school instruction.
‘Much the same contrast is explored by Brooks
in his comparison of “analytical” and “non:
“analytical” processes. He notes that “in studying
‘concept identification, experimental psycholo:
{ists have found it far easier co capture analytic
than nonanalytic processes (p. 168). He notes
research Gndings (eg.. studies of anificial
{grammar learning) indicating that people can
Successfully generalize without an. analytical
basis. His own experiments show that nonanalyt-
ical “analogical” reasoning, in which new in-
stances are inferred on the basis oftheir overall
similarity to specific prior examples, will pro-
duce far better than chance performance on
tests of concept learning. Brooks suggests that
such nonanalytical reasoning is superior to
analytical reasoning when the complexity of the
underlying analytical rule exceeds short-term
memory procesinglimjations, Analogical pro
cesses might go a long! way toward explaining
child language learning without recourse to
elaborate language-specific mental wiring.
Kossyin presents an {ntrospective analysis of
mental imagery and evaluater a computer
simulation designed to account for introepective
and experimental evidgnce on the nature of
mental images. He conchdes that such ima
is quite real and that it provides acco to
memory data not available through verbally
mediated reall: Brown's concluding chapter
Pursues a curious example oilutrate the com
Plesity of linguistically encoded mesnoges and to
Brgue the necesity of inconporaing entena of
Posey wl a of thi the lncon
met uses the meta language of se tory
to argue that the controveis current amoug
prychologse concerning competing model of
eal representation, whether, fer example,
Sr "standard templates” “digalcntrices,
"inary feature tale,” oF “proposional a
are beside the point ince “all are models ofthe
same cognitive theory” (p. 279). He secs the ask
of cognitive prychology ss clearly differentiated
from’ that "of neurophysiclogy. Copaiive
prychology isnot “concerned withthe concrete
Say in which mental representation i 3
omplshed within the brain and the nervous
Sypem”s rather, i “can spite only to informa.
tonal equivalence between its models and ae
tual mental representation inside the head” (p
270,
This diancton clarifies the parallel sue in
cognitive anthropology and lings between
Sods ruth and hous pocar” The goal of
cognitive anthropologists, Hnguiss, nd omni
tive paychologins stem tenilly equivalent
Each in attempting to consiract model of the
human mind’t modts of the world (pp. 273.
776) in order to account forthe information
procesing behavior characteristic of out
Species. The tee feds differ primarily in the
Iethodological context of the invegation
‘The present volume gies strong tetimony to
common purpose shared by thee divert do
tiptines In all, the proportion that “how
humans... proces information i nota very
Simple quesion” (p. 151) is now widely 3p
precited, and the legacy of behavior i
phaccally rected