8/15/2019 Language in Cognitive Development
1/19
gcncc
or
th
C.U..Iigrnatic
Mind
I
 'C hierarchical category structures natural products of human cogni-
tlon
independent oflanguage? Or do they depend on language for their
dcvelopment? As discussed in Chapters 4 and 5, language is based on
categoríes ofvarious kinds, fromphonemes to sentences toword classes
ind lexical referents. Moreover, as reviewed in Chapter 4, research has
Iocumented that infants categorize objects at varying levelsof generality
Irom global to specific. Perceptual categorization is characteristic, not
only of human infants but of all complex organisms that must distin-
uish between foodand not-food, speciesmembers and others, for exam-
pIe.Thus categorization is basic to cognitivefunctioning ofallkinds, and
certainly to the construction of event representations, mimetic symbols,
basic and complex language, and narrative.
However, hierarchical categorization or taxonomic classification goes
beyond this basic leveI. It operates on basic categories and sets them into
new order relationships with other categories at the same leveI and at
higher and lower levels. These operations are engaged in bypeople in all
cultures (Berlin, 1978),who form folk taxonomies of plants and ani-
maIs, for example. The questions as to when children are capable of
engaging in these kinds of ordered relations, and what cognitive com-
petencies are required (for example, what logic), have long been topics
of study in developmental psychology. As with the assumption of basic
numerical and grammatical principIes, there is now an implicit assump-
tion on the part of many researchers that infants and young children
have basic taxonomic principIes that are displayed under conditions of
learning the appropriate language, or being given appropriate materiaIs
(e.g., MacNamara, 1982;Waxman & Gelman, 1986).This assumption is
bolstered by the observation that naturallanguage (e.g., English) incor-
porates hierarchical category structures, with superordinate and subordi-
nate branchings, and that children learn and use the terms applicable to
223
8/15/2019 Language in Cognitive Development
2/19
8/15/2019 Language in Cognitive Development
3/19
l'l(i 1 / / / lS / l I I S iu ( ' ' 'S /I h ,, : L I/ /I / / I ;
na l conditions bu t is c ss e nt lu lly 11 1 1 11 11 .11\ IIIIHI'HH, 1101
fi
1'01111 t I 1 1 1
antagonistic, mutually exclusivc fOl'íllfl 01 Ihlllklllg.
(p,
157 )
For Vygotsky (unlike Piaget) the move Irom spontancous concopt»
1 1 I
scientific concepts rested essentially on uses of language. In hts vlvw
the most important characteristics of mature thought are (1) gcncrallt
and
(2)
systematicity, and he held that these characteristics are ablwlIl
from the concepts of preschool children, and that scientificconcepts
11I'i'
acquired through formal instruction.
Luria, Vvgotsky's dose colleague in the 1930s,ventured to test thcs«
ideas through studies of peasants and workers on newly collectivizcd
farms in the Soviet Union. These studies, published in English in 1976,
presented category choice problems, such as the following: Ofhammer,
wood, ax, and saw, which does not belong? The results ofthese studics
indicated that uneducated adults group items thematicaIly rather than
categoricaIly.For example, they said that the wood goes with the other
items, because you use the axand the saw to cut wood. Tulviste(199])
reexamined these studies and others in the light of Vygotsky's theory
and Leontiev's (1978)extension of it in terms of activity theory. He ar-
gues that Vygotsky's ideas about the difference between scientific and
spontaneous concepts reflects a basic difference between the use of
words and categories in different activities. The claim is that scientific
thinking often uses the same linguistic terms used in everyday activities,
but that everyday language does not reflect the same logical structures,
which are a product of schooling. This condusion implies that young
children' s thinking does not employ categoriesin the sameway or ofthe
~same type as adults' thinking.
In contrast, the contemporary understanding of young children' s
thought and concepts, as it has emerged over the past decades, is that
they are similar to those of adults, although with a much restricted
knowledge base. An important influence in this direction was Rosch's
(1975)work, which introduced the idea of the
b a s ic
leveI of a natural
language semantic hierarchy, the level at which members of a category
are most similar to each other in terms ofshape and motoric actions and
most differentiated from members of other categories at the same leveI.
According to this proposal, basic leveIcategorieshave higher cue valídíty
in comparison with subordina te categories (which share more features
with other categories at their level) and superordinate categories (in
which members share fewer features). Rosch and her coIleagues estab-
lished that categories at the basic leveI have psychological salience for
both adults and children. They observed that children tend first to learn
A
1 11I ' 1 '1 '
1111111 1 '11 1111 111 / 11 '1 I t 'liI141 U' 1I 111 11t1 -llltHk Ii-v(-I,()I ' (' )(11 1 1 1pll', dog and car,
IIIIIH'I'h u u 111 II 1 1 ,11 1 Hllp('I'ordllllllc(c.g., animal ) 01' lower subordi-
1J1 t1t ·
(t',g
I'ollit''') levei (Anglin, 1977; Rosch et a1., 1976).2Prior to'
l{mwh'H
work, Brown (1958a)suggested that parents tend to label things
un lhe basís of the greatest functional utility for the child, and that, in
',(' leral, words are used at the level ofgreatest utilitywithin a particular
I'ontcxt (ao important and stilllittle recognized principIe). ~
Rosch's work on the internal structure of categories suggested paral-
1 (· 1 8
between the Piagetian and Vygotskian view of conceptual defi-
clcncy in the younger child and the way that adult categories are com-
posed. Rosch (1973,1975,1978)proposed that the internal structure of
natural language categories is organized around prototypes; that catego-
rioshave graded structure, with more central and more peripheral mem-
bers; and that categories are structured in terms of familyresemblances
(Wittgenstein, 1953), that is, overlapping features, none of which are
~ithernecessary or sufficient. Rosch also emphasized that attributes of
objects are not randomly distributed in the world, but form patterns of
orrelated attributes. For example, having feathers is associated with
having wings and with flying, whereas having fur is associated with
different attributes, such as having four legs. Rosch's results implied
that young children are not different from older children and adults in
their basic conceptual structure. Rather, human concepts at aIl ages are
formed around central exemplars or abstracted prototypical features ,.,
(these are two different kinds of prototype models). And although chil-
dren's concepts lack consistent logical structure (as Piaget daimed), so
do adults' concepts. . , . . .J
Certainly one of the major effects ofthis line of work was to highlight
the real conceptual and categorical competencies of very young children
and even infants (e.g., Mervis, 1987; Bauer & Mandler, 1989b). This
research has opened up issues about the basis for children's initial con-
ceptual structures, and the developmental course ofcategorical organiza-
tion. These developments relate to the basic proposals outlined here in
terms ofmental event representations (MERs),relations that began to be
explicated in terms of the functional core hypothesis, introduced briefly
in Chapter 4.
The Funciional Core Hypothesís
Nelson (1974)proposed that the child's interactions with the world deter-
mine the child's viewof the functions of objects, which in turn motivates
8/15/2019 Language in Cognitive Development
4/19
the basis for the forrnation of
COIll'l'pllllllld
t i , 11'1 1111111 /; of t'l Ií '1y
wun]
The functional core hypothesis (FClI) dlllllll'd 11 11 I1 chlldren f01'l1l 1 '011
cepts of objects based on object function and [orm, with function
1 I/l lh l'
core of the concept and form its probabilistic periphery, used to
ItklllHy
instances of the concept. The child's concept becomes the basis for wonl
meaning, with the first words mapped onto already formed conccpts,
The word is then extended to other objects sharing form and function
features, thus recognized as instances of the child's concept. This
hy
pothesis argued against a model of the abstraction ofperceptual features
as the basis for concept formation and word meaning on the following
grounds: Given that any object or set ofobjects can be described in terrn:
of an indefinite number of features, and related to one another on thc
basis of similarity along an indefinite number of dimensions, formíng
concepts by abstracting features from encounters with the world as-
sumes that there is already some motivating factor that makes some
properties relevant and others not. The child's experience with objects
was assumed as the motivating factor. Evidence in favor of the model
carne primarily from an analysis of the words that children learned for
objects that they could do things with. Additional evidence was sought
in experiments that posed form against function as the basis for forming
concepts.
During the 1970sa number of contradictory studies were reported that
aimed to test the predictions of the functional core hypothesis by pitting
function against perception. However, this became recognized as a false
issue (Nelson, 1978b, 1979)on the following grounds. First, the FCH did
not claim that children do not notice or rely on perceptual features, but
rather that child-experience-based functional characteristics are more
central to their concepts of objects. Second, function can be conceptual-
ized in a number of ways.
Intrinsic
function - what something does -
can be a perceptual feature as well (for example, characteristic move-
ment ar noise). Therefore, in these cases there is no difference between
functional and perceptual features of an object, and the argument re-
duces to the claim that some features are more important than others, a
claim empirical tests did not in fact support (e.g., Bowerman, 1976).
E x tr i nsi c function, however, relates an object to something else, pri-
marily to the people who interact with it. Although all interaction in-
volves perception, the relation of person to object cannot be said to be a
perceptual feature of the object; rather, it is a matter of perspective,
intention, or stance. Moreover, extrinsic function may (but need not)
involve the conventional uses of objects, for example, that telephones
1111
11 1 ' tl/'Hd 10 111111111111111 \1111
'
lVI'I' 111111 '. 1 1111 111111'1 \/1, 1 I1
'
1 '1 I11 1H ' 111 restrlcüons on
Ih,'
kl lllW h'dl '.l ' 1111I'11 '0 1 yOllllgchlldrcn, lhelr concept» of objects often do
11 01
IIIWI'por'llll'
such Iunctions. Yct some studies of children's reliance
11 11
Iunctlon
11 1
acquiring concepts and word meanings tested conven-
Ilolllll Iunctions rather than those based on child interactions, and thus
IlIdr reporta ofno effect are not surprising (e.g., Andersen, 1975).
'l'hc FCH was designed to explain infants' spontaneously formed
nbjcct concepts, which might lie behind their earliest naming practices,
IlIld
it was extended to a hypothesis about the formulation of any new
roncept. Now I would recognize more strongly the collaborative nature
IIf the child's conceptual processes, even in the infancy period, which
might lead to concepts based on attributes other than function, as well
IS
those with a functional core. Still, although function is not the only
basis for forming concepts of objects, it has become more widely recog-
nized as a critical component of conceptual structure for both adults
nnd young children in recent years (e.g., Barsalou, 1991;Kemler Nel-
son, 1995).4Moreover, the idea of a core of conceptual properties sup-
plemented with probabilistic features has also gained wide acceptance
(Medin, 1989).The perspective of event cognition emerging from Gib-
sonian theory (Shaw
&
Hazelett, 1986)puts these issues into an adap-
tive perspective that is consonant with the evolutionary perspective
presented in Chapters 2 and 3. In this perspective, action and percep-
tion are two sides of a single schematizing system, in which concepts
emerge from the experience of invariances that exist in the world of
events. Function and structure are then indissolvable abstractions from
adaptive interactions.
Function may play different roles in different levelsofa semantic hierar-
chy. Basic leveI concepts may have a functional core, correlated with
perceptual features that are determined by function, as some ofthe early
research (Nelson, 1979)supported. Bycontrast, superordinate categories
are defined almost entirely in terms offunction (Nelson, 1979,1985;Scrib-
ner, 1974),and subordinate categories are frequently distinguished from
basic Ievelcategories primarily in terms ofdistinctive perceptual features.
For example, children as well as adults learn to distinguish among sub-
types of cars, cereals, and shirts on the basis ofperceptual features rather
than function. This proposal accords with the developmental sequence
that is usually assumed, namely, that basic level concepts (combining
function and form) are acquired first, with some subsets acquired when
perceptual features distinguish among instances, and with superordinate
categories acquired only later when functions themselves become catego-
8/15/2019 Language in Cognitive Development
5/19
rized across objects and CVClllH (M lvIH ,~ l. rlHUn . I')H~;1111I11 1.
Markman, 1980).
Hierarchical Category Organization and Alternatives
Although not all researchers are as conservative as Vygotsky and Piagt'I,
the majority have affirmed that the hierarchical structure represented 111
terms of coordinates, subordinates, and superordinates is inadequatel
grasped by children below the age of about 5 years. Rosch et aI. (197(1)
showed that preschool children can classify at the basic leveI but do nol
succeed at higher levels; this finding has generally held. In languago
studies, a long tradition of empirical research has demonstrated the
younger child' s inability to form inclusive category structures of the kind
represented by X is a kind of Y, where X is a hyponym of Y (e.g., u
robin is a kind of bird ). These relationships require the understanding
ofinclusivity (robins are included in the class ofbirds), asymmetry (rob-
ins are birds, but birds are not robins), and transitivity - attributes of
higher class apply to the lower, but attributes ofa lower category on the
hierarchy cannot be assumed to apply to the higher. In this type of
structure, when
x
isa member ofy and y is a member ofz, then logically
it follows that
x
is a member of z. The logic of inclusion relations is not
evident in classifications in the real world; it is an abstraction realized in
symbolic formo
Grasping the implications ofthese relations is traditionally found tobe
a devélopmental achievement of the early or later school years (Winer,
1980).Greene (1994)found that kindergartners understand transitivity
but not asymmetry, but even this conclusion is constrained by the fact
that the artificial categories used in this study were equivalent to the
basic leveI and lower branches, not to abstract superordinates. Greene
concludes that critical developments in categorical understanding take
place between years 4 and 7. Benelli (1988)reported three studies that
provided evidence that superordinate categorization is essentially a mat-
ter oflinguistic acquisition, that is, that children come to an understand-
ing of semantic organization around the age of 7 years. This proposal is
consistent with the conceptual and language integration hypothesis dis-
cussed in the latter sections of this chapter.
The logical relations of classes do no~necessarily hold for part-whole
or collective structures. For example, a child is a member of a school
class, and the class isa part of the school, but the child isnot a member
of the school, and the attributes of the school (e.g., teachers, books) do
j .- ,
I
Il,tI ill'l'ly I 111• ,llIltI WIr I 'ilHVYI'ltll~lky
II/H/,)
1 lflld{'I'{'d
hc collcc-
Iltlll
I 111'1
r/IIIII/VI' l'OIlCl'p(1I1l1Hll'Ul'llll'l', MUI'kllltlll'S
( 1 9 1 : 3 1 ) interpreta-
Iluu IIldÍt'lIlt'dlhol l'OI ('dions rcprescnt a pcrvasive alternative to inclu-
VI' rllll'gorics for adults as welJ as children. Among other things, she
huwcd that superordinate category terms in different languages often
1t·I'I'rL'nccollections, in which members play differentiated roles [e.g.,
1 /11 /. 1 , (school) c la s s , jorest] rather than true categories in which all mem-
hl'rs have equivalent status.> Markman also demonstrated that young
dllldren use collections more readily than categories in cognitive tasks
uch as class inclusion. Markman's hypothesis is that collections, as
PMl-whole structures, are based on a more natural kind of organizing
principle than that of hierarchical classification, based on inclusion rela-
Ilons, and that young children may be able to use this principIe before
Ihcy grasp the inclusive principIe of true categories. Therefore, collec-
tlons may precede and serve as a foundation from which children can
construct true categories. Bythis hypothesis, categories (classes) might
('rnerge from collections.
In a related vein Mandler (1979)proposed that there is a schematic to
categorical shift in development such that younger children display
knowledge organized in terms of schemas and older children and adults
Iisplay knowledge organized in terms of categories. This hypothesis
scemed to accord with evidence from memory experiments with young
children, which revealed deficiencies in the use of category structures
but proficiency in the use of information about scripts and scenes.
chemas include scenes, scripts for events, and stories that organize
material in terms of part-whole configurations [see Mandler (1983,
I984b);Chapters 4,6, and 7]. Schemas are based on functional organiz-
ing principIes; for example, things belong together in a particular scene
(such as a kitchen) because they have a function in that scene, and they
are related to each other as parts within a functional whole. Thus they
may be related to the development of categorical structures from func-
tionallybased concepts.
Since putting forth the schematic to categorical hypothesis, Mandler
has revised her proposal, and now holds that categorical and schematic
structures develop together (Mandler, 1983).She has carried out a series
of experiments, finding that some of the categories infants and toddlers
recognize are general or global, of the extension that is appropriate for
superordinate categories in the 'adult world (Bauer & Mandler, 1989b;
Mandler & McDonough, 1993;see Chapter 4).6But this raises an impor-
tant issue: Are the categorical processes in infancy similar to or the same
8/15/2019 Language in Cognitive Development
6/19
as those that undcrlic lhe fornuulun II1 1 1 1 \1 11 11 1 11 1 11 1 I1t 'I'III'1 'h lt 'H (:IV I'II 11 1 1 '
explicit inclusive asymmetry and
ll ' l ll H l l l lv ll y
1 1 1
1 1 1 1 '
luüer, IIl ld
t ' v l d l 'II I 'I '
of the difficulty that even young schoolchlklrvn havc with
1 1 1 1 '/ - 1 ( ' 1 '1 · 1 1 1
tions, the answer must be no, but then some way of relating lh(' Iwp
types developmentalIy must be found.
Vygotsky's theory of conceptual development based on the
contrturt
between spontaneous and scientific concepts addressed the problcm ( 11
integrating spontaneously developed conceptual knowledge with scicn
tific or theorétical cultural systems mediated through language. Thl
problem, central to the thesis of this book, is this: how to
reconcik-
índividually and informalIy - often implicitly - constructed knowledge
systems with culturalIy derived, formalIyorganized, explicit systems 01
knowledge, when the two incorporate different concepts and conceptual
relations but refer to the same domain ofknowledge.
The problem in its essential form is pervasive from the beginning of
the child's acquisition of Ianguage; the Ianguage that the child learns
initialIyimplicitly incorporates cultural knowledge systems that are only
partialIy and imperfectly represented - and inimportant cases not repre-
sented at alI - in the child's prelinguistic experientialIybased knowledge
system. The problem is acute for words that denote superordinate cate-
gories. Because superordinate categories do not exist as such in the real
world, but only in the language used to talk about them, the child
cannot have a prelinguistic concept that is the equivalent of the adult
superordinate concept. The child's problem then is to find a way of
forming word meanings for superordinate classes that map the adult
meaning system appropriately. Event representations provide a basis for
such meanings.
Event Representations and Categories:
The Syntagmatic-Paradigmatic System
As is now welI documented, young children have organized knowledge
of familiar reliably structured events, such as having Iunch (Chapter 4).
Thus they have an e ve n t c a te go ry of eating Iunch, a category whose exten-
sion includes alI the successive occasions ofIunch, and whose intension
includes the specification of necessary components - actions, objects,
persons. Events incorporate two types of hierarchies. First, events can
be said to be organized hierarchicalIy, in that they may be composed of
smaller event units. For example, the eating Iunch event may be com-
posed of subevents ofpreparation, eating the main meaI, eating dessert,
Illd 1 '11 '1 1 11 li/'. Itl Vl l~ '
Hnd ,
IIIII 'H I'III 1 1 1 11 11 1 1 1 1 11 1 ' l'II Il/l I I'I'l' d 111 1
ovcnt ln it e
IIIV II 1 1 ,, ,, , wlrh 111 1 '1 '\'11 1 tl uh go1 tlH
r·whOl'dillllll·10
lhe
main goal of eating
1 1 1 1 11 '11 . M( )I'(' ( )V l 'I',
lhe cuung lunch cvcnt ís part of a larger event se-
'1 I'II\ 'C rc -prcscnttn g a typical day. This type of organization can be seen
H I II pnl'l-whole hierarchical structure.
1\ second type of hierarchy involves the combination of two or more
,'V l' I1t S
into a more general category. For example, Iunch can be consid-
I'l'< .I an instance of the meai category, which includes as welI dinner,
hn-nkfast,
British tea, Iate-night snack, and so on. In turn, dinner might
t( , differentiated into the subtypes family, party, restaurant, formal, and
\I
on. Rifkin (LucarielIo
&
Rifkin, 1986) has demonstrated that such
nclusive hierarchies of event categories reflect the same kind ofpsycho-
10gicaIreality a sobject categories foradults. In addition, event categories
provide the functionaI basis for construction ofhigher-order object cate-
\ories, and it is this relation that ismost significant to the developmental
lssues in focus here.
In brief, the claim is that children begin with functionalIy derived
.ategories at the basic leveI, and that these categories are recombined
into larger groups that enter into open slots in events, thus forming
slot-filler categories. These categories are not true taxonomic hierarchi-
al structures, but form the basis for the formation of such structures,
which are constructed in colIaborationwith adult instruction and experi-
ence with the categoricaI structure of the adult Ianguage.
S yn ta g m a ti cs a nd P ar ad ig ma ti cs o [ t he C o nc e p tu al S ys te m:
S lo t- fi ll er C a t eg o rie s
The syntagmatic-paradigmatic axisis one of the three major organizing
principIes oflanguage in Saussure's
(1959/1915)
theory. According to this
principIe, language is structured by:
(1)
combinatory principIes, what
can be sequentially combined with what in a given structure; and (2)
substitutability principIes, indicating which forms are alternatives within
a particular sIot in the structure to be filled. The former principIe is
known as the sy ntag m ati c relation, the Iatter as the p ar ad i g m ati c . Paradig-
matic structures, according to Saussure, are abstractions from experi-
ence, whereas syntagmatic structures are evident in spoken utterances.
The two types are interdependent in that the combinatory structures are
defined in terms of the paradigmatic categories.?
Different lexicaI organization principIes derive from syntagmatic and
paradigmatic relations and are reflected in changes in children's re-
8/15/2019 Language in Cognitive Development
7/19
1 \ - 1 1 .1 1 / 1 , '\ / 1 1 1 , '\ 1 ' 1 11 ( '1 1 , '\ /1 1 1 i l ' l ' / ) 1 / 1 1 1 I 1 1 1 /
sponses
011
word
essoclntlon WI\
111111,,,It1I1II'IlIlyI011l1dd11'1'11111111011
(Nelson,
1977b).
In the word tll:lHOi'lIrlioll11I'11,111'HIII1J('\'IIi 11II'Ii I\'
with a list of words one at a time and Ill:lk('d10 rospond to (11ll'1 11 w]
the first word that comes to mind. In thís task adults respond n10HllIllII
time with words from the same grammatical class, that is, with IHI lIdl
matic responses. Very young children respond primarily with Hy,1I11
matic associat ions, that is, words that come from a different forrn
du
Such responses diminish as children advance in the school years
Ih
syntagmatic-paradigmatic shíft. Many theories have been put forl 10
explain this shift [see Nelson
(1977b)
for review]. The perspectlvo ,,;
event knowledge implies that the finding can best be interpreted in 111•
broader context of the development of categorical structures.
A similar syntagmatic-paradigmatic analysis can be applied to CVI'IlI
representations. Within a real-world event - for example, the event 01
eating lunch - there is a more or less fixed sequence of actions and a It'
fixed set of alternative slot-fillers of possible foods. Thus basic leveI COI1
cepts offoods are represented interms ofthe meal events inwhich particu
lar foods appear (such as sandwiches at lunch). Hierarchical
SLOT-FILLflll
categories then may be formed by combining representations of food in
different events under a single dominating term, the slot-header terrn,
FOOD.
This context-defined conception of slot-fi ller categories contrasu
with the logical conception of food as an abstract superordinate concept
whose intension isroughly anything that can beeaten and whose exten-
sion is alI possible foods. The latter is an abstract notion of a functionally
based and context-free higher-order category. The claim here is that this
conception is a late achievement of categorical development but is based
in the earlier achievement of an understanding of slot-filler hierarchies.
Many familiar object categories exist in relat ion to event categories as
sl o t-f il le r c a t eg o ri e s , that is, i tems that can fil la particular slot in a particu-
lar type of event, for example, foods that can be eaten at lunch. Objects
in events may also be related in terms of contiguity when they occur in
the same event but not in the same slot. For example, plates and cups
occur in the lunch event but not in the EAT x slot. The relation between
plate and, say, spaghetti, is a co m p l e m e n ta ry or th e m ati c relat ion. In the
developmentalliterature on category formation, a contrast is often made
between thematic constructions and categorical (or taxonomic) construc-
tions, with the latter assumed to be more advanced than the former, as
Mandler's schematic to taxonomic
(1979)
hypothesis proposed. A the-
matic associate to
do g
would be
a b o n e
the dog was chewing on; a taxo-
nomic associate would be
h o r se ,
a member of the same taxonomic cate-
, , , . , ,, .
.
'\ 1\' 1111'11111111'1,111111111/1
uv ,'
Iypl ully 111 '/1111111.111111''I'.IIly HIIII~lllllo
I1lIlIt',1'1111.11'1 1.lol IlIh'I'\'IIII'I\\)I'II'H11'('11111111'llltH'Ill:llhemalic assoei-
''''14, 1,,11\'1',dot.fllli.'l' Il(.· '1H-1rc gl'oupcd log~'lllcl' on the basis of similar
1111111111111I1
'ol l 'H ,
not on the basis of complementary roles.
1'111
'
l'xduHivl' focus on objects and object taxonomies in psychology
111114
,tI
to the assumption that the thematic relation is primitive and that
1 1 ,
prcschoolcr must overcome it in arder to achieve the taxonomic
I
HllIlii':ation. In contrast, the event knowledge perspective on concep-
111111l'presentations incorporates complementary (or thematic) as welI as
III,'gorical (or taxonomic) relations, and demonstrates both how they
11'1'olated and why they may be called on in different tasks. The view
111'1'('s that both complementary and categorical configurat ions are
IIIIHCdn the same conceptual organization but display different relations
wilhin that organization. Thus the relation between individual experien-
1IIIIIybased event knowledge and the construction of abstract category
knowledge is a dynamic, constructive one that results in interacting
planes of knowledge organization.
, iu b stitutabi l i ty. The slot-filler category proposal points to an empirical
nnd rational basis for concept formation that has been neglected in both
classical and contemporary studies.
S u bst i tu tabi l i ty
is not included in the
Iraditional accounts of the basis for category formation ar associations -
which include similarity and contiguity - and it has been little discussed
in any literature. Substitutability, unlike contiguity, is not apparent in
ony single real-world context; i tems that substitute for one another in an
event are not necessarily present in the same context at the same time.
Moreover, such items need bear li tt le similari ty to one another; for exam-
ple, bananas, cookies, and pudding may be alternative lunch dessert
slot-fi llers for a child but are not similar in appearance, texture, or taste.
The traditional bases for associat ions of similari ty and contiguity do not
explain the basis for slot-filler categories. Substitutability, a paradigmatic
relation, does.
Consider again the schematic to categorical shift proposed
by
Mandler
(1979).
From the perspective of the syntagmatic-paradigmatic basis for
category formation, this move can be seen as the emergence of one
structural type from the analysis of another, not as a substitution of one
type for another. In this form of the proposal categorical structures are
assumed not to replace the schematic structures, but to complement
them. Schematic structures continue to be relied on by older children
and adults, even for many purposes where categorical structures are
8/15/2019 Language in Cognitive Development
8/19
generally assumed to bc more d'fkh'lll I11,'11I t1 1I1'111goI'yHIIIII'IIII '
are ideal for the logical classification fll'llI'lIl1'Hl'II\1I1H.ll'd by Hl'ltnllllly
work, and particularly by science. They exlst in nonscientific CUIlIIl'I'H1
folk taxonomies (Atran, 1990),and as semantic organizations in ord
i
11111
language use.
Paradigmatic conceptual structures thus parallellinguistic structu 1\'
The child's conceptual system appears designed, not simply to analyz«
patterns that are apparent in perceptual arrays, but also to analyze th«
resulting representations for structural characteristics. Among thes«
characteristics is the alternation of items within specific locations in li
sequence, and the extraction ofpossible items into a category. Given the
evidence that this process is a general strategy of conceptual analysis,
could it not be applied equally well to language categories in terms of
paradigmatic alternatives? The analysis and developmental construction
of grammar may well be similar to the analysis and developmental con-
struction of conceptual representations. Might such a principIe of cate-
gory formation lie behind or generate the structure of language, rather
than arising from it as a unique solution? This question is raised, not to
point to an answer, but to emphasize the bidirectionality of the concept-
language relation at its core.
F ro m E ve nt -b a s e d S lot-fille r C a teg ori es to S em an tic C ateg o rie s
In time the child must somehow acquire a system of shared word mean-
ings; that is, the child's ideas about what a word can refer to must come
to coincide in critical ways with the way the word is used by others in
her linguistic community. Experience with parenta I uses of category
terms isa first step in this direction. Achievement ofan abstract category
language depends upon the further development ofa differentiated - or
abstracted - level of semantic representation in which linguistic terms
are related to one another directly and are not embedded in the particu-
lar experientially based event representation and its derivatives.
In previous research it has been found that parents use basic level
category terms with their beginning talkers, and that children acquire
these terms more or less exclusively in their early vocabularies (e.g.,
Anglin, 1977).Lucariello and Nelson (1986)analyzed the category terms
displayed in maternal talk with 2-year-old children in three different
contexts in a study showing how naturallanguage categories are related
to event structures in discourse contexts. This analysis of the talk be-
tween 10mothers and their 24-month-old children found that, although
,,,
1111I'lh'II'VI'I'IIIHI11II'dlllllllllll 'dHlllllwnlllil\/l' 1111111.1,1111'1'1'l\Hlso UI:iC
111blllll tIlIJH'I'0I'dI1ulIl'I11t1tlhol'lllllllll'II'I'IIIHy holh mothcrs and chil-
111\'11,l'ho 11011bnHIcevei termBtcndcd to uppcar most frequently in the
1lllIll'xlof familiar ovcnts. such aseating lunch orgetting dressed. Moth-
1'/1provided clucs to the use of the hierarchical terms, such as, What
l.lnd of a drink do you want? following with alternative possibilities in
1111.'rink category. Another discourse clue was provided by using basic
lvvclitems in the same syntactic frame as the hígher-level category term,
uch as the following: Let's put on your clothes. Put on your shirt, put
nn your socks. In this context the term clothes can be extracted as a
general slot-filler term for items that are put on.
The development over time of the construction of these relations be-
lween events and object categories can be seen in transcripts of the pre-
bed talk between 2-year-old Emily and her father [seeChapter 4;Nelson
(1989c)]. Alternative
slot-fille rs
were found in Emily's talk about food,
typically embedded in talk about breakfast. An early example from 22
months is the following monologue:
(1) Emmy like cornbread and toast.
I don' t like [?] apples and [?]
I like toast and muffins.
Food I like and [muffins] too.
I don't like anything ... cept for that, that bread daddy
has ...
Here we see that already Emily is using the general term food as an
alterna tive to the basic level terms of specific items.
In the pre-bed dialogues with Emily's father there are many discus-
sions of what she would like to eat for breakfast the next day. For exam-
ple, when Emily was not quite 2 the following discussion took place:
(2) E:What we have on breakfast day? What we have?
F:What will we have for breakfast? You know, tomorrow
morning, you're going to have yogurt, and bananas and
wheat germ, like mommy gave you this morning, re-
member that? Instead of an egg tomorrow we're going
to have yogurt and bananas and wheat germ ...
Later, Emily enters into the dialogue more actively, specifying what she
wants, as in the following dialogue from almost 27months:
(3) F: We'll get up ... and we'Il go down and have breakfast,
you can choose what type ofegg you want,
I .
8/15/2019 Language in Cognitive Development
9/19
• IH {. tII I,li II II,li I' 1 /1 ( 'P811111(11' {IH /I ' ',I /;' /II
E: I want ... a
boilcd
l'gg.
F: Okay. And you can
choos« w h n t Iypl'
of
c(JI'~lIl yOIl WIIIII
you can have either shredded whea t 01 ChCCr iOH .
E: Shredded
wheat
A month and a half later, Emily is entering her own suggestions:
(4) E: And, so now te11me about today
F: We11,today you had a Tanta day also.
E: I
uxm
yogurt.
F: And you want yogurt. I know and I think 1'11have som
raspberries for you tomorrow.
E:And I . . . cereal
F: Today you had strawberries, tomorrow I think you'l1havo
raspberries.
E:Cereal Cereal
F: You'l1have cereal? Okay. Cereal and yogurt? You want
bananas in yogurt, or raspberries in your cereal?
E: Yeah.
F: Okay. That'l1be good.
E:And strawberries in my cereal.
In these discussions, Emily's breakfast food category was highly con-
strained to the particular situation and did not stray from the alternatives
specified by this particular family (e.g., yogurt, cereal, fruit, eggs). It did
not wander into domains ofpizza, hamburgers, or other items appropri-
ate for dinner. That is, her category of alternatives for breakfast was
specific to that event - it constituted a slot-filler category of breakfast
foods.
These examples highlight the critical relation of language to the con-
ceptualization of objects and events. Emily's monologues reveal that her
event representations are not independent oflinguistic formulations and
of particular linguistic input from her parents. Thus even when Emily is
2 years of age we cannot speak of a conceptual system that is indepen-
dent of language; rather, the two evolve together interdependently.
Note, however, that the language that simultaneously expresses and
shapes Emily's representations is not the abstract language ofcategories
but the concrete language of experience. To be sure, her language in-
cludes category terms, but these are particularized to specificexperience.
The coordination of the child's language of categories with that of the
adult takes place in everyday activities and the discourse surrounding
them. The child's MERs reflect the systematization of her experience in
1:1111111111iYflllllll 111111/11111)
- . -
,. . .
ti
t
L
~
~
1
~
do~ > 00
1
1
~
----O
c::>--a--O
Atililt Represenlational System
Adult Presentation (Parole)
Ch il d Semanti c R ep re senlation
Child Conceptual Systern
Figure 8.1. Mediation of the child's semantic system
by
the adult's partial
representation and presentation systemand the child's conceptual system.
S o u r c e : K. Nelson, M aki ng s en s e: Th e a cqu is itia n a f s ha re d m ea nin g 1 985).
Permission to reprint from AcademicPress.
the world, but this system is not the system that organizes the relevant
cultural categories. The conventional cultural system is displayed to the
child, not systematically as in a school text, but in bits and pieces
through adult-child talk. The child acquires a partial system mediated
by
both the adult's partial system and the child's conceptual system.
Figure 8.1illustrates this processoThe important points tobe noted are
(1)that the adult's representation is an incomplete structure in compari-
son with the cultural system as a whole, (2)that its display in speech is a
partial and disconnected representation of the adult system as a whole,
and (3)that what the child gets fromthis display is an incomplete, error-
prone construction based on both prior (nonhierarchical) concepts and
language use. The adult's talk about a category displays certain rela-
8/15/2019 Language in Cognitive Development
10/19
tions, il11pfying 1 . 1 hicrarcht cnl l'flllllltl, 1111.1 , lIvldl'H l i fl'w 01' 11 .••
submembers in the híerarchy.
'l'hc
chlh] IlIlIy ,'c'l'I)g,llzc thcse OH 't'fule.tI
members within her event scheme, and I11tly f0111 a bond betwccn lh('III.
But event talk about the category is inevitably fragmentary and uI1S
tematic (as the example from Emily's talk with her father illustratea],
sampling bits and pieces from the cultural construction of the cat
so that the child's mapping may be distorted and fu11ofholes.
Achievement of an abstract category language depends upon the fur
ther development of a differentiated - or abstracted -leveI of semantic
representation in which linguistic terms and their related concepts are
not embedded in the experientia11yderived event representation system
but constitute a semantic system ofabstract relations. The development
of that leveI makes possible the representation of a true semanti
hierarchy - a taxonomy that is based on hierarchical inclusion relations
and not simply on combinations of event-contexted slots. Its construc-
tion, in co11aborationwith adult informants, is a major development of
the preschool and ear1yschool years.
S lo t- fi ll er , T he ma ti c, a nd T ax on om ic C a teg o ri e s: E vi d e n ce
If, as claimed here, children derive their first category hierarchies from
event structures, we should be able to observe evidence of these struc-
tures in category membership, in what items are closely associated in
memory, and in the use of categories in tasks such as sorting objects
and pictures, remembering lists of category items, and forming new
categories. These tasks have a11been used to evaluate the proposal that
slot-fillers form higher-order categories intermediate between the basic
leveI and conventional superordinates; the evidence is summarized
here.
C at eg o ry P ro du c ti on . Three studies of children's category item produc-
tions have been carried out (Lucariello,Kyratzis, & Nelson, 1992;Nelson
& Nelson, 1990;Yu & Nelson, 1993). In the study by Lucariello and
co11eagues,4-year-olds, 7-year-olds, and adults were asked to provide
category items for five superordinate categories: food, clothes, animaIs,
furniture, and tools. The results were analyzed in a cluster analysis, and
the resulting clusters were examined for their content, specifica11yfor
slot-filler or conventional subcategory clusters. Furniture and tools
emerged as poor1y organized categories for subjects of a11ages. Four-
year-olds produced very few items in these categories, and included a
I'~III
1 1 1 1 1 1
11 1 1 1
1 1 1 1 1 1
1 1 11 , 1 1 '
1 1 1 \1
IIWI
11 1 1
I{
I;()II I
,IIICK I I
H K lf l l
M'JIININ
MIXED
I L- ~ t
10.0
4.0
2.0
o
.0 6.0
Proximity Value
Figure 8.2. Clothing category clusters generated by 4-year-olds.
S o u rc e :
J . Lucariello, A. Kyratzis, and K. Nelson, Taxonomic knowledge: What
kind and when? (1992). Permission to reprint from
Child Deoelopmeni,
University ofChicago Press.
relativelyhigh number of nonmembers. Therefore, these categories did
not seem to have a slot-filler basis. Furniture may be organized more in
terms of collections in certain spatial configurations (bedroom, living
room, for example); tools may remain at a more particular, local func-
tionallevel. These categories warrant further investigation. It should be
noted that although the thesis here is that slot-fillers in MERsform an
important basis for the construction ofhigher-order semantic categories,
they are not the only basis.
In contrast, food, clothes, and animaIs, which were assumed tohave a
slot-fillerstructure, were organized in terms of single-event contexts by
the youngest children; by multiple slot-fillersubcategories for the 7-year-
olds; and by a combination of slot-fillerand conventional subcategory by
the adults. This evidence indicates that preschoolers organize categories
in terms of single events, young schoolchildren combine slot-fillers from
different event contexts under the single-category term, and adults orga-
nize these categories in multiple ways, relying on both event-based slot-
fillers and conventional subcategories. Figures 8.2-8.4 illustrate the clus-
ters for the clothes category at the three age levels.
In two studies with children not from mainstream American groups,
Nelson and Nelson (1990)and Yuand Nelson (1993)reported findings
similar to those above. Nelson and Nelson (1990)asked low-income
inner-city African American children to provide category members in
three categories: food, clothes, and animaIs. In one condition children
were asked toprovide items intwo different event contexts (e.g., break-
fast foods and snack foods); in the other condition they were asked to
8/15/2019 Language in Cognitive Development
11/19
1
I'ANIII
1IIIIIn
IlHUlH
SIIOI S
SOCKS
SWEATER
UNDERSHIRT
UNDERWEAR
SKIRT
t - - - - - - U
BLOUSE
BOOTS
SNEAKERS
JEANS
BATHROBE
SHORTS
BRA
COAT
JACKET
VEST
SNOWSUIT
SCARF
HAT
GLOVES
MITTENS
MORNING
OUTSIDE, MAIN
OUTSIDE
OUTSIDE, ACCESSORV
~
I I I
10.0
8.0
6.0
4.0
2.0 o
Proximity Value
Figure 8.3. Clothing category cIusters generated by 7-year-olds. S o u r ce :
J. LucarielIo, A. Kyratzis, and K. Nelson, Taxonomic knowledge: Whal
kind and when? (1992). Permission to reprint from
C h i ld D eo e lop m e ni ,
University of Chicago ~ress.
respond to the general category term, food. An interaction ofagewith
condition was found, with 5-year-olds providing more category mem-
bers under the event context condition, and 7-year-olds providing more
instances under the general instructions. Again, examination of the con-
tent of clusters indicated that under general instructions items were
clustered by event at both ages, and for the younger children one event
context predominated. Further, of the 5-year-olds, those who had had
preschool experience produced more category members under both con-
ditions than those without such experience.
The results of this experiment, shown in Table8,1, indicate that youn-
ger children may have difficultyaccessing more than one slot-fíllercate-
gory at a time, and that they have not yet begun to combine categories
into larger general conventional structures, as 7-year-olds have. The
difference between the 5-year-old groups indicates that specific cultural
experiences, such as preschool, influence the construction of conven-
tional categories. This is to be expected under the hypothesis that cate-
gory construction is culturally and linguistically determined. Preschool
1'1 11 1
11 1 1 /1
1 1 1 1 1 1 1 1, 1
WI AIII~
IIHI ',ti
IKIIlI
l:nAI
JAt:KI I
OOCK
SIIOI
UNDEI'lWEAR
HORTS
I·SHIRT
STOCKING
SUIT
VEST
SWEATSHIRT
GLOVE ~
HAT
SCARF
MITTEN
PANTYHOSE
BELT
T
BOOT
SNEAKER
2·PIECE SUIT
BRA
PANTIES
SLlP
ROBE
NIGHTGOWN
BATHROBE
SLlPPER
JEANS
SLACKS
TUXEDO
UNIFORM
MOI(NINO, MAIN
/
; ; = t
'0 , 0 '
J t i \ j 00 0 ,
u
MORNING
~
.
.~
J
ACCESSORV
OUTSIDE,
~ ccessosv
ORNING, A
~ IV WOM ' 0
0 f B n o
~ r
I I I
10.0
8.0
6.0
4.0
2.0
o
Proximity Value
Figure 8.4. Clothing category cIusters generated by adults. So u rc e :
J .
LucarielIo, A. Kyratzis, and K. Nelson, Taxonomic knowledge: What
kind and when? (1992). Permission to reprint from C hi ld D e v e l o p r nen t,
University of Chicago Press.
and kindergarten teachers specificallyteach children the discourse forms
for category inclusion (Watson, 1985;Wertsch, 1991).
In a cross-cultural and cross-linguistic study using the same para-
digm, Yu and Nelson (1993)asked Korean-speaking children in Korea
to produce category members in the same three categories used with
American children, and the results were essentially similar to those
found with English-speaking children, although with specific cultural
variations. Thus the category production results indicate that slot-filler
organization is common among young children across cultures and
8/15/2019 Language in Cognitive Development
12/19
1 1 /1 , '0 1 1 1 1 1 , '0 1 1 '
1/
C , , \ ll i fl l 'l
Table
8.1.
M e a n n urn b e r of
ii e m s p r o r ll ll ' I 't l l l I , '0 1 1 1 1 1 / 1 1 ,
n nu lli knt ,
1 / II I I I ' I II I 'S I / I/ 1
Group/Condition
N
Animals
Uothing
Food
Mt'llIl
K1
Taxonomic
8
5.88
4.00
7.88
5.\
SIot-filIer
8
6.25
9.12
10.87
8 .71 1
K2
Taxonomic
7 8.86 7.57 9.43 8.6
Slot-filIer
7 11.29
12.43
19.00
14.211
Grade
2
Taxonomic
8
17.25
13.00
27.63
19.2~
Slot-filIer
8
14.00
10.62
12.25
12.2(
Sourc e :
Nelson
&
Nelson
(1990),
table
1,
p.
435.
Usedwith permissionof the
publisher.
does not disappear as conventional taxonomic categories are integrated
with it, but can stilI be seen in the productions of adults for familiar
everyday categories.
W o r d A s s o c i a t io n s. The word assoeiation paradigm was used with 4-and
7-year-olds by LucarielIo, Kyratzis, and Nelson
(1992),
who presented
slot-filler category items from the categories offood, clothes, and animaIs.
The responses were analyzed in terms of the relation between stimulus
word and response word. Relations could be categorical (conventional or
slot-filler) or thematic or event-based (event, function, complementary).
Older children produced significantly more categorical associates than
younger children, but the great majority of these were slot-filler assoei-
ates. Younger children, in contrast, produced three times as many the-
matic assoeiates as categorical, and significantly more thematic assoeiates
than did the older children. Of these thematic assoeiates, twice as many
were functional responses (e.g., cookie-eat) than complementary (e.g.,
cereal-spoon). These results indicate that younger children's words are
still spontaneously organized in terms of the events in which they are
relevant, rather than in terms of categories, whether slot-filleror conven-
tional; however, asthe category production task showed, they are capable
ofproduction of slot-fillercategory items. By7years ofage, children have
begun to organize their words more in terms of categorical relations
(illl 'IIII H 11 1 I 1 '1 11 l'IIII '}', lItlt 'ld fllld
I
1111 1111 '111 1 11 11 ') 111 1 1 I lIu/J1 oltIH'Ir'ClIle-
1',llIh'lIl I'/I/lIIIIHI'lH~HH'IIIII'H11 '(' HtllI/lI )
l l ll H
'I'III'H ' ('Hullshelp to ex-
1 111 1 111 ' WI,II-('HlllbIlHIH.'c.Jyntugrnatlc-purudlgmtlc shift i nword associa-
IIIIIIHound I thc carly schooJ years. in that most thematic responses
IV/'I'I'rom dlíícrcnt word classes (e.g., noun-verb), and most categorical
I/'HpOIlHCScrc from the same word class (noun-noun). Four-year-olds
I 'ovldc mostly thematic responses; 7-year-olds provide mostly categori-
1 tI
responses.
Scll
(1992)
reported a word assoeiation task with preschool, kindergar-
I , and fourth-grade children, categorizing responses as event-based,
whlch indicated an object's function or purpose, slot-filler or taxonomic.
'l'hcre was a progression with age in this task: preschoolers provided
primarily event-based assoeiations, kindergartners provided slot-fillers
IIS
well, and schoolchildren were able to provide taxonomic associations
1 1 1 addition to slot-fillers. Thus Sell's research replicates and extends the
Indings from Lucariello and colleagues
(1992).
L is t R e c a ll . A standard test ofthe establishment of categorical organiza-
tion in semantic memory is the recall of categorized word lists. In general
it has been found that young children do not recall more from lists
containing members of the same category than from random lists, and
they do not cluster category members together in recall (e.g., Nelson,
1969).
Several experiments have been carried out using this paradigm to
test the psychological reality of slot-filler categories. Two recall experi-
ments (Lucariello & Nelson,
1985)
presented 4-year-old children with
three kinds of lists - conventional category, slot-filler category, and
thematic - under free recall and cued recall, with either slot-filler or
superordinate cues. Briefly, these young children recalled significantly
more from the slot-filler list than from either of the other lists, and the
best recall was found with the slot-filler list and script cues. Clustering
was significantly better with the slot-fillerlist as well.
These results have been replicated with Korean children (Yu & Nel-
son,
1993;
Yu,
1993).
In Yu's studies it was found that 7-year-olds per-
formed as well with the taxonomic listas with the slot-filler list, suggest-
ing that they had integrated the slot-filler categories into a hierarchically
composed taxonomic category. In her last two experiments (Yu,
1993),
the same results emerged under different study conditions. Four-year-
olds did consistently better with the slot-filler list, and they were able to
use strategy instructions with this list but not with the taxonomic list,
1
I1 1
I
I.
II
8/15/2019 Language in Cognitive Development
13/19
whereas 7-year-olds used SlI'ull')\II'Hwlll: 111 1'1'11111111111'IHIIIH
• .
1111,..111'
slot-filler list.
Other Studies.
Slot-filler superiority over taxonornic (or coordinate) 11111
in different memory paradigms has also been reported from othcr labom
tories by Rosner and Smiek (1989), Kraekow and Blewitt (1989),
8/15/2019 Language in Cognitive Development
14/19
Children are asked to find anothe» 011 111
1
MIIII,pl.llldlH 11 H ' 1111 '/\ ''1 1 '1 1 '11 11 '1 '
from the remaining two choice pictures. 111 11 11 '
Ih'HI cxpcrlmcnta, whhh
have defined the paradigm, children werc glvcl1choices
bctwocn
I1 nW 1I 1
ber of the same category (e.g., a cat when the target is a dog)
anel
thematic associate (e.g., abone). They found, similar to the results
frOI1\
the study by Lucariello and colleagues
(1992),
that when preschooJ chll-
dren were asked to choose another one they chose thematic assocíatos
predominantly. However, when children were told that the target had o
nonsense name (e.g., this is a
dax )
and asked to find another of thc
same name (e.g., find another
dax )
they chose the category associate.
From these results researchers have concluded that children obey a taxo-
nomic constraint on word meanings; that is, they assume that words
refer to taxonomic categories and not to groups of thematically related
items. Therefore, when they are given a name for the item, they override
a natural propensity to group things thematically and instead group
them categorically. The extension of this explanation to the picture-
choicetask used in the experiments by Lucarielloand colleagues and Sell
(1992)
suggests that in the latter tasks children were relying on their
natural tendency to see the world in thematic terms, unrelated to lexical
knowledge, which is categorically (taxonomically) organized.
It is obvious that the thematic-taxonomic contrast is related to the
event-category development projected here. However, the proposal of
slot-filler categories based in event representations that become inte-
grated into larger taxonomic structures has the advantage of providing a
more detailed and deeply developmental explanation of the results of
these experiments. It also is consistent with Donald's
(1991)
evolutionary
hypothesis about the development of the hybrid representational sys-
tems of the human mind. It goes beyond the naming effectproposed by
Markman and her associates to suggest that these effects are part of the
emergence of the semantic representational system and the convention-
alization through language of children's categories.
Our experiments (Lucariello et aI.,
1992)
have shown that preschool
children's conceptual organization displays both event (thematic) and
category (slot-filler) bases, and that these persist throughout develop-
ment. Thematic relations appear to be especially compelling when pic-
tures are used in cognitive tasks, perhaps because pictures appear to
represent parts ofevents.t? What changes is the integration of spontane-
ous category structures into conventional, linguistically constructed cate-
gories as the child experiences these in school or through other instruc-
tional contexts.
I1111U,
lUuU' llld C'1I111'1~,,1t1l11
ChIlJlIl.I·
1) 1 1 l
lu- bllHIH 01'
Ilw research rcportcd hcrc, lt appcars that a change in
4llllllllk momory
organization leads to a conceptual shift in children
l t • • twocn t J . and 7 years of age that has broad implications for many
dll
f('rcnl
cognitive tasks. How slot-filler categories become transformed
' 1 110 conventional categories provides important clues to what might be
tnklng
place at a deeper cognitive leveI during the preschool years.
IIrlcfly,development is hypothesized to proceed as follows.
As delineated here, the first assumption is that slot-fillercategories are
.ibstractions
spontaneously derived from event structures, the latter
based on experience with routine events. Because of the way that adults
use category labels and functional statements relating basic and subordi-
nate terms to superordinate terms, children learn to apply the general
superordinate term to their slot-filler categories. This enables them to
respond with slot-filleritems tothe request toproduce category members.
lot-fillers then come to be headed by a conventional category label and,
as the label is used in different script contexts, to be combined into more
general categorical structures mirroring those of the conventional seman-
tiecategory. Thisprocess was examined in an experiment (Kyratzis,
1989)
in which children were taught the names and categories oftools used ina
craft project. Five-year-old children had no problem acquiring these cate-
goriesbut did have problems with the inclusion relation connecting terms
at the basic level to a term at the superordinate leveI.
The process of connecting terms and categories begins early and pro-
ceeds at a different pace for different categories varying on familiarity
and complexity. During the preschool and early school years both par-
ents [Adams
&
Bullock
(1986);
Callanan
(1985);
Lucariello
&
Nelson
(1986);
see previous examples] and teachers (Watson,
1985;
Wertsch,
1991)
engage in labeling practices that emphasize the inclusive structure
of hierarchies. Yet until quite late in the game, young children do not
easily take advantage ofthese inclusion statements (Horton
&
Markman,
1980;
Kyratzis,
1989).
It seems reasonable to conclude that although
category construction begins and proceeds at a moderate pace during
the preschool years, the hierarchical inclusion structures of categories
are not well established, stable, and relatively autonomous within the
semantic system until at least the early school years. 'This change-
reorganization - in semantic memory is then observable in terms ofage
differences in performance on episodic memory tasks that involve se-
mantic categories, such as the listrecall task, as previously documented.
8/15/2019 Language in Cognitive Development
15/19
The contribution of linguilHk
t~XI1l'dl'IH III 1 1 1 1 ' l'IIIIHII'Ih'IIOII 0 1 11 'Pi
conventional naturallanguage hi era rchlen 11 1 IIIII'IH'IIIIII bt.'l'HIIH\.',IHI
'lI'vl
ously noted, items such as different
foods, nnlmule. and clothcs do
lIul
sort themselves into neat hierarchically inclusive structures in lIw
1 '1 ,, ,1
world. Rather, language itself imposes ways ofcategorizing items,
I.'nlt'
gories that may vary from culture to culture and language to languagc, 11
is n e c e s s a ry for the child to be exposed to the particular ways tha: 1 1
language categorizes items in order to learn the compositional structurv
of the categories. Put simply, hierarchies are in the language (and in
conceptual systems), not in the physical or material world [see also
Wierzbicka (1994)].
Early in the preschool period (2to 3 years) the child is still acquiring
basic competence in using a first language for communicative purposcs
(Chapter 5). It is only after this point that one may expect more abstract
organizations such as semantic categories to emerge as pragmatic struc-
tures from language in use. This shift in functioning emerges at about th
same time that other language-dependent structures such aspersonal nar-
ratives and autobiographical memory do. Language uses of the commu-
nity pull the child toward theorganizations inherent inthe linguistic struc-
tures ofthe culture. On their own, children might construct quite complex
semantic organizations, but they do not need to because they have avail-
ablelinguisticmodels. Furthermore, in order tospeak the language ofthe
community they need to acquire the categories and organizations that
that language incorporates. What they must dois to integrate the models
of the language with their own emerging knowledge organizations.
Although children may gain a great deal of knowledge about cultural
categories simply from interacting with adults araund relevant activities,
schooling advances the process explicitly,as the evidence from the Nel-
son and Nelson (1990)category praduction study suggested. Wertsch
(1991)pravides examples from school sessions that illustrate the pracess
that takes place. In the following excerpt a teacher is involved in an
exercise with six first-grade students of finding the one that doesn't
belong among four pictures [similar to the categorization tests carried
out by Luria (1976)and Tulviste (1991)]:
(5) T: Why doesn't the keybelong with the ham and a tomato
and a banana, Mikey?
C:Because the key isn't a fruit.
T: Well, a ham isn't a fruit. What are all those things? Things
you can ...
C : n . lI .
1 : 1 1 1 1 . 'l'ldl\ns
vou l'HIl
\. n l o • o
T: Which
ono are we going toput an Xon, Jessica?
.':' I'hc plant.
T: Thc plantoWhy?Annie.
: Bccause it's not clothes.
T: Jt's not clothes. Good ..
o ••
T: Patrick, which one are you going to e l i mi n a t e in the last
one?
C:Ummm ... The goose.
T: Why?
C: Because it's not something you can sit on or sleep on.
T: Very good. It's not fumiture. right? We can callthat furni-
ture.
o (p.
133)
As Wertsch points out, this excerpt not only reveals the explicit
teaching of everyday category vocabulary that does on in the school,
but also illustrates the mixture of experience language and formal lan-
guage. In the first exchange, the teacher not only accepts but also
elicits the formulation you eat instead of the category term food,
but in the last exchange she switches to the formal verb eliminate
and substitutes the category term (or linguistic object) fumiture for
the informal functional category. This excerpt illustrates the implied
process of integration of everyday knowledge with the structures of
cultural categories.
The inclusion relation that is critical to hierarchical categories is
used to relate scientific concepts in taxonomic structures and has also
infiltrated the common language to a greater or lesser extent. Children
can learn the superordinate language that incorporates the inclusion
relation without yet understanding its implications (e.g., the transitiv-
ity and asymmetry of its relations). The question that has not yet been
adequately addressed is how children may come to understand these
relations, and how the instructional efforts in school advance this un-
derstanding. The issue here is how conceptual reorganization and re-
structuring proceeds with examples that are simply displayed in cul-
tural talk and through explicit instruction that formally lays out how
systems such as taxonomic categories are structured. Conceptual re-
structuring on the basis of both implicit and explicitdemands from the
8/15/2019 Language in Cognitive Development
16/19
1 1 ) ; 1
IIII 'S '/I ISI ' 1 i'O iS 'II I III '
1 ), 1 / 1 1 1 ,
8/15/2019 Language in Cognitive Development
17/19
birds build nests and lay
cggs nn- tlol 111'1111111, 11 11 ' dhll'oV\',y 01 yllll1l)
children but rather are directly taughl
IIH III P IH ,111 H 01 b ln lH, 1 1 111 /1 11 WI'
are to take the theory claims seriously, wc neod
1 0 IIl-1k
wherc the
1 1 1 ( '0 1 1 1 '
come from: On what basis does a child construct a theory?
By
Llw
11111 (' 1 1
child is in grade school it seems quite reasonable to believe that
slw 1111
constructed from various sources, including verbalIy conveycd knowl
edge, mini- or proto-theories that serve to organize and explain
CO I11 111 01 I
phenomena. However, it also seems reasonable that since theorics
0 1'
about phenomena in the real world, presumably, they must be con
structed on the basis of pretheoretical knowledge about the real world,
just as scientific theories are based on pretheoretical knowledge. Prethco
retical knowledge is gained in the ubiquitous process of making senso.
Theory building begíns with a descriptive knowledge base in any field,
The distinction between descriptive theory and explanatory theory
haH
been ma de in different ways by both Chomsky (1965) and Mayr (1982).
A
descriptive theory is concerned with
how
things are in the world; an
explanatory theory is concerned with why they are the way they are. For
the individual child, as for science, description necessarily precedes expla-
nation, and smalI-scale explanation precedes general theory. This is not to
deny the validity of the claim that all observation and description are in
some sense biased by one's interests (and in science by one's theory). But,
simply put, to ask for an explanation of something is already to have
knowledge of that something. If we are not to revert recursively to innate
knowledge we must account for its initial acquisition.
Thus the attribution of theory structures to the knowledge that
young children possess in certain domains is, from the present view, a
questionable practice, especialIy when theorists equate a child' s theo-
ries directly with scientists' theories (Copnik. 1993b, p. 100), even
though the child' s theories are held to be implicit, not explicit, as are
the scientists'. Theories in science are constructed according to well-
understood, systematicalIy organized knowledge domains, in accor-
dance with formal principIes of theory construction, including some
form of hypothetico-deductive logic. Scientific theories define a set of
concepts and order these within an explanatory system of causal rela-
tions. They are the property of a community of scientists and are in
princíple testable against newly discovered facts. A new theoretical
advance may be originated by an individual, but it remains a hypotheti-
cal proposal until or unless it is accepted as valid by the community.
The child's implicit theory has no such validation procedure, and the
implicit theory by definition remains individual and unshared. Any
- '1 1 1 1 , ,1' 0
1 1 \ l 1
I tI
.1' '1' 11 11111 1 lu' do,l\' 1 1 1 .1 til
1 ' ,1 0 '
, 1 , 1 , 1 1 1 1 1 1 11l'l'l'pluI1Cl. of
IIld.,lllll'o
.11
111/llldlll 1 ,, 1 . Ii '
H
1 1 1 IIi'1I ('OIl'I'''' i,ly
HIII*'clivc.
I 1 'tljWl1Si'
to such objcctlons,
some
íhcory theorists claim that
l'IIlldl'('/1'1-I lrnplicit
thcories are the equivalcnt of folk theories, not
1'i,'IILi(k
thcorics. Thcy are the construals of the world that the folk
huld in cornmon. But there are problems here, too. First, adult folk can
Irlkulate their theories, or at least the propositions that derive from
lhern.
Second, young children do not seem to share many of the folk
Ihcories of the adult community; therefore, they must be little folk
thcories. but unshared. Yet if these theories are unshared, what guaran-
Ices
that children arrive at a single common theory; why not many
Iifferent idiosyncratic theories?
Listening closely to young children in their everyday lives does not
suggest systematic organized knowledge within well-delineated do-
mains. Consider the folIowing exchange between a 4-year-old child and
her mother from a maternal diary study [Callanan and Oakes (1992, pp.
221-222)]:
(6) Situation: Bedtime
Child: Why does Daddy, James (big brother), and me have
blue eyes and you have green eyes?
Parent: (Told her she got her eyes from Daddy. Then said
goodnight and left the room.)
Child: (child calIs mother back 5 minutes later) 1like Pee
Wee Herman and 1have blue eyes. Daddy likes
Pee Wee Herman and he has blue eyes. James likes
Pee Wee Herman and he has blue eyes. If you liked
Pee Wee Herman you could get blue eyes too.
Parent: (I told her it would take more than my liking Pee Wee
Herman to make my eyes blue. 1realized that she
didn't understand me, so 1explained that God
gave me this color and that they couldn't be
changed.)
Child: Could you try to like Pee Wee Herman so we could
see if your eyes turn blue?
Parent: (I said 1would think about it, but if my eyes stayed
green it was ok.)
Although this example indicates creative causal thinking, it violates in
the most extreme way the boundary between basic biological and sym-
bolic cultural domains, and does not suggest the possession of systemati-
cally organized, causalIy related knowledge in either domain.
8/15/2019 Language in Cognitive Development
18/19
Theories have an irnportaní
plm'I' 111
lhe' IIl1wl'lh'lIwl
oj
11111111hlllly,hl
at Donald's (1991) fourth level, As oulll'14'1I I ( ' 1 1 1 1 1 '1 c 1
3,
1111'(11'1'111'111
thinking is held to be a product of litcrate soclettes. 'lulvistc
( 1 1 ) < ) 1 )
111
gues in a similar vein that hierarchical taxonomic categorical structur«
are cultural developments useful in scientific activities, but are not inho:
ent aspects of human thinking. For example:
If thinking appropriate to certain types of human activity
W('I'I'
hereditary, we would undoubtedly have to expect a basic simila riI
in the thinking of peoples of any cultures and epochs .... l3ul
there is no basis for assuming that such specific types of thinking,
as for example, formal operations (according to Piaget) or thinking
in scientific concepts (according to Vygotsky) would have to arise
during phylogenesis or anthropogenesis , and then be transmittcd
hereditarily. It is difficult to understand why these types of think-
ing would have to be formed in anthropogenesis if most of human-
ity that has nothing to do with schooling and scientif ic information
even today gets along very well without them. (p. 70 )
I think we should take theories seriously, but not attribute them to
infants and young children without good cause. The attribution of theo-
retical knowledge to the young child is deeply misleading, I believe, in
obscuring the differences between types of human knowledge systems
and thus problems faced in conceptual change and development. In
particular , the attribution of theoríes to the preschooler obscures the
problems of integrating an experientially derived organization with a
cultural system of knowledge. Theories rest on the prior systematization
of knowledge that is culturally organized. Individual representations of
theoretically based knowledge may be imperfectly represented (as Fig-
ure 8.1 suggests), especially if that knowledge is conveyed only implic-
itly through informal discourse. But partial representations may be a
long way from fully established theories.
I I
[ I
M aking T he o re ti ca l S en se
A basic assumption here is that the child is engaged in the process of
m a ki ng s en se ,
a process with many subprocesses that begins with the
dawning of mind. The following scheme reviews the developments cov-
ered thus far in terms ofa sequence of developmental tasks prerequisi te
to theory construction.
• D a ta g at he ri ng:
From the first days oflifethe infant isintent on gathering
data about theworld she inhabits, includingespeciallythe people attend-
L
'If',hl IIlldhe'l ~1I1 CI 'lIll1lfl'~1111 11 f',1,IIII.dlll',ulllnlly (1l'(w
e
l'dH ulong
hlllloglt'lIl1yWI,III'llllIhllt\IrI'dlllll/IIhlll dl'lI II,IIH'whnt klnds of data are
111I1'11I1,.d
0 1I11d
how Iht'y nre Ol'dl'l'pd
111
II1l'1110ry.Ioward this end, the
chlld lI equlpped wlth ways of parsing the world, and capacities for
Intorproung and intcrrclating the pieces. For example, exploring spaces
11Uobjects appears to be as universal and as much a part of an
~pigcncticprogram as eating, sleeping, and smiling.
• M o de l bu il din g:
Fromvery early theinfant appears to construct and torely
on models of relations in the world - what happens when, where, and
in what way - that ís , event or situat ion models. By
3
years of age,
children demonstra te detai ed command of the situational structures of
their worlds, induding expected people roles, objects, and action se-
quences, and can transform these in play.
• Small-scale ex pla na ti on s eeking: When children believe that they under-
stand how things
ar e
they begin to seek explanations for why they are.
They ask questions of adults about the how and why and they project
explanations themselves. beginning at about
3
years. The
3-
and 4-year-
old's persistent why may be precipitated by a dawning ofthe notion
that what ismust have a reason for sobeing. But anyone who attends
dosely to the child's questions must be struck by the limited nature of
her understanding of the world and its causal mechanisms at this age
[see example (6)]. At first her own explanations often reveal a quite
fragmentary and distorted picture of the world and how
it
works, and a
readiness to cross ontological boundaries in seeking explanations, as
example (6)illustrated.
Once some small-scale explanations in a domain have been proposed
and accepted, and entered into the child 's ini tial explanatory model, the
child may begin to order them and generate new hypotheses on their
basis, which can then be articulated and tested for acceptabil ity by oth-
ers. The process of minitheory construction may begin at around age 6
as children are inducted intocultural categories and cultural knowledge
systems and as they integra te these with their prior experiential ly based
knowledge. This proposalleaves open the recognit ion that children
co n-
str u ct
their own knowledge while at the same time recognizing the
so ci al -
cultura l origins
of that knowledge.
The later steps in this progression are largely dependent on the use of
language, and on social-cultural constructions made available to the
child through language.The construction of a theory even in childhood
is thus viewed as fundamental ly a social act , based on culturally shared
understanding of the phenomena in a domain [see Freyd (1983)]. Once a
child can use language as a sense-making system she can begin to use
cultural sources to reconstruct on the individual level what is already
established and revealed on the social-cultural level. This proposal is
consistent with Vygotsky's insight that systematic scientific concepts
are leamed through transactions with, including direct instruction from,
8/15/2019 Language in Cognitive Development
19/19
the socialworld, and must be coordinated with thc spOnl(\IH'ClIIM~IIII
cepts of the individualchild.
Through the construction of cognitive models of her world, 11 11 \ 1 1 11
man infant and young child re-creates in mind experiences in t1wwI .I I.1
and thus can understand - in the sense of predicting - how
thingH
,li
These are descriptive models, not explanatory models. The
cstabltal:
ment of stable world models in infancy and early childhood is thua 1 1
necessary first and basic step toward the establishment ofmore
abstrucl
explanatory knowledge. As the child begins to seek explanations
flu'
why these relations are as they are, the models themselves may
contaln
clues that are used to frame the deeper questions. The child's descriptivv
modeI ofhow things are supports the construction and understandinz 01
reasons for why they are that way.
The process ofprojecting explanations and receiving them from othera
is among the new possibilities inherent in the use of language as [1
representational medium. The causal explanations of the young child
maybe based onwhat others have told her orthey may emerge fromher
own inferencing processes. They emanate not from general theories in a
well-defined domain but rather from small-scale observations. They can
be seen as a step in the direction of the construction of larger, mor
systematic explanatory structures.
As the discussion in this chapter has emphasized, recognition of
knowledge acquisition that is not available in perceptual displays does
not necessarily imply theoretically organized knowledge. Rather than
impose our scientifically derived notions of theories on the minds of
young children, it seems wiser to remain open to different alternative
possibilities when studying the organization of their knowledge. In this
chapter the organization of events, and knowledge derived from and
organized in terms of events, has been shown to be at least one very
salient basis for organization in the young child's knowledge system.
This basis bears little resemblance to the scientists' theories; it is more
plausibly related to the narratively structured knowledge that was out-
lined in the previous chapter. Indeed, it appears that the syntagmatic
structure of narrative forms the basis for the paradigmatic structure of
categories and the theories - folkor scientific- within which these cate-
gories take their place.
Top Related