Language in Cognitive Development

8/15/2019 Language in Cognitive Development

1/19

gcncc

or

th

C.U..Iigrnatic

Mind

I

Â '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


2/19


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


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-


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


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-


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


8/19

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 .


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-


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


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


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


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),


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.


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


16/19

1 1 ) ; 1

IIII 'S '/I ISI ' 1 i'O iS 'II I III '

1 ), 1 / 1 1 1 ,


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.


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,


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.

Documents

Language in Cognitive Development