The role of associative strength and conceptual relations in matching tasks in 4- and 6-year-old children

Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/242524300

Theroleofassociativestrengthandconceptualrelationsinmatchingtasksin4-and6-year-oldchildren

ArticleinInternationalJournalofPsychology·August2004

DOI:10.1080/00207570344000394

CITATIONS

23

READS

31

4authors,including:

FrançoiseBonthoux

UniversitéPierreMendèsFrance-Grenoble2

31PUBLICATIONS303CITATIONS

SEEPROFILE

ChristineCannard

UniversitéPierreMendèsFrance-Grenobl…

22PUBLICATIONS268CITATIONS

SEEPROFILE

AgnesBlaye

Aix-MarseilleUniversité

65PUBLICATIONS1,691CITATIONS

SEEPROFILE

AllcontentfollowingthispagewasuploadedbyChristineCannardon28March2014.

Theuserhasrequestedenhancementofthedownloadedfile.Allin-textreferencesunderlinedinblueareaddedtotheoriginaldocumentandarelinkedtopublicationsonResearchGate,lettingyouaccessandreadthemimmediately.

The role of associative strength and conceptualrelations in matching tasks in 4- and 6-year-old

children

Nelly Scheuner, Francoise Bonthoux, and Christine Cannard

Agnes Blaye

Universite Pierre Mendes France, Grenoble, France

Universite de Provence, Aix en Provence, France

T he associative strength between target and associates, a factor assumed to be critical but generally not

controlled, and the type of conceptual relation (thematic and taxonomic) were manipulated independently

in a matching to sample task to determine their respective effects on the matching behaviour of 4- and

6-year-old children. Perceptual similarity between target and associates was controlled and maintained at a low

level. A preliminary task was designed to assess the associative strength between targets and several associated

pictures. These judgments served to construct for each child the sets of stimuli used in the matching task. Exp. 1

opposed a strong and a weak associate with the target in different configurations: the sets included a target and

two thematic associates, two taxonomic associates or one associate of each type. Children were asked to choose

the picture that ‘‘went well’’ with the target. Data revealed the role of associative strength on matching choices.

This factor interacted sometimes with the greater availability of thematic relations in 4- and 6-year-old children.

In Exp. 2, two other configurations were tested. Thematic and taxonomic associates were both either strongly or

weakly related with the target. Results replicated those of Exp. 1 and extended them. They showed that younger

children were biased towards thematic relations only when these relations corresponded to strong associations.

Thus, increasing experience with objects appears to reinforce both associative strength and thematic orientation.

Finally, in Exp. 3, instructions orienting toward taxonomic choices modified responses in 6-year-olds only.

Altogether, these results show the influence of specific instances and suggest that preschoolers’ matching

decisions are partly stimulus driven.

L a force d’association entre une cible et des associes, facteur suppose critique mais generalement non

controle, et le type de relation conceptuelle (thematique et taxonomique) sont manipules independamment

dans une tache d’appariement afin de determiner leur role respectif sur les conduites d’appariement d’enfants de

4 ans et 6 ans. La similarite perceptuelle entre cible et associes est controlee et maintenue a un niveau faible.

Une tache preliminaire permet d’evaluer la force d’association entre des cibles et plusieurs images associees. Ces

jugements servent a construire les ensembles de stimuli utilises dans la tache d’appariement. L’experience 1

oppose un fort et un faible associe dans differentes configurations: les ensembles incluent une cible et deux

associes thematiques, deux associes taxonomiques ou un associe de chaque type. Il est demande aux enfants de

choisir l’image qui « va bien » avec la cible. Les donnees revelent l’influence de la force d’association sur les

choix d’appariement. Ce facteur interagit parfois avec la plus grande facilite d’acces aux relations thematiques a

4 ans et 6 ans. Deux autres configurations sont testees dans l’experience 2. Les associes thematique et

# 2004 International Union of Psychological Science

http://www.tandf.co.uk/journals/pp/00207594.html DOI: 10.1080/00207590344000394

Int Jnt Psychology (gamma) IJP35365.3d 29/1/04 17:08:22 Rev 7.51n/W (Jan 20 2003)

The Charlesworth Group, Huddersfield 01484 517077

Correspondence should be addressed to Dr Francoise Bonthoux, Laboratoire de Psychologie et NeuroCognition, CNRS UMR

5105, Universite Pierre Mendes France, PO Box 47, F-38040 Grenoble cedex 9, France (E-mail: Francoise.Bonthoux@

upmf-grenoble.fr).

This research was supported by grants from the University Pierre Mendes France of Grenoble, the ACI Cognitique 2000, the

Centre National de la Recherche Scientifique, and from the University of Provence (Aix en Provence). Authors thank the

preschoolers and their teachers who participated in these studies. We thank Yvette Hatwell for helpful comments on an earlier

version of this paper.

INTERNATIONAL JOURNAL OF PSYCHOLOGY, 2004, 39 (00), 000–000

taxonomique sont tous les deux, soit fortement, soit faiblement associes a la cible. Les resultats repliquent ceux

de l’experience 1 et les etendent. Ils indiquent que les choix des plus jeunes enfants sont biaises vers les relations

thematiques, mais uniquement lorsque celles-ci correspondent aux associations fortes. Ainsi, une experience

accrue avec les objets semble renforcer a la fois la force d’association et l’orientation thematique. Enfin, dans

l’experience 3, une consigne censee orienter vers les choix taxonomiques n’entraıne une modification des

performances que chez les enfants de 6 ans. En conclusion, ces resultats montrent l’influence des instances

specifiques et suggere que les decisions d’appariement des jeunes enfants sont en partie dirigees par les stimuli.

S e manipulo de manera independiente la fuerza en la asociacion entre el blanco y los asociados, un factor

que se presupone crıtico, pero que no se controla comunmente, y el tipo de relacion conceptual (tematica y

taxonomica) en una tarea de igualacion a la muestra para determinar sus respectivos efectos sobre la conducta

de igualacion de ninos de 4 a 6 anos de edad. Se controlo y se mantuvo en un nivel bajo la similitud perceptual

entre el blanco y los asociados. Se diseno una tarea preliminar para evaluar la fuerza de la asociacion entre los

blancos y varias ilustraciones asociadas. Estos juicios sirvieron para construir, para cada nino, las series de

estımulos que se emplearon en la tarea de igualacion. En el experimento uno, se opuso una asociacion fuerte y

una debil contra el blanco en configuraciones diferentes: las series incluyeron un blanco y dos asociados

tematicos, dos asociados taxonomicos o uno de cada tipo. Se solicito a los ninos que seleccionaran la ilustracion

que correspondiera con el blanco. Los datos revelaron el papel que desempena la fuerza de la asociacion en las

elecciones de igualacion. Este factor interactuo a veces con la disponibilidad en aumento de las relaciones

tematicas en los ninos de cuatro a seis anos de edad. En el experimento dos, se sometieron a prueba otras dos

configuraciones. Tanto los asociados tematicos como los taxonomicos se relacionaban fuerte o debilmente con

el blanco. Los resultados repitieron los del experimento uno y los extendieron. Mostraron que los ninos mas

jovenes presentaban sesgos hacia las relaciones tematicas solo si estas relaciones correspondıan a asociaciones

fuertes. Por lo tanto, al parecer, la mayor experiencia con los objetos refuerza tanto la fuerza de la asociacion

como la orientacion tematica. Finalmente, en el experimento tres, las instrucciones que orientaban hacia las

elecciones taxonomicas modificaron solo las respuestas de los ninos de seis anos. En conjunto, estos resultados

muestran la influencia de las instancias especıficas y sugieren que las decisiones de igualacion de los ninos en

edad preescolar estan guiadas por el estımulo.

Taxonomic relations refer to representations of

objects of the same kind belonging to a semantic

category (e.g., animals, vehicles), whereas thema-

tic relations correspond to an organization of

knowledge in terms of familiar scenes or events

(e.g., playing football includes elements of differ-

ent kinds such as a football player, a ball, boots,

etc.). In matching to sample tasks contrasting

taxonomic and thematic associates, children’s

performances have often been interpreted in

term of a ‘‘conceptual preference’’ (Greenfield &

Scott, 1986; Smiley & Brown, 1979) for one type

of categorical relation. ‘‘Conceptual preference’’

was deduced from the observation of a majority

of taxonomic or thematic choices when children

were instructed to match one of two objects with

a target (‘‘which one goes with the target?’’). In

fact, it was supposed to reflect the accessibility of

one type of relation at a given age.

Taxonomic and thematic relations are available

early (Bauer & Mandler, 1989; Daehler, Lonardo,

& Bukatko, 1979; Scott, Serchuk, & Mundy,

1982). Yet some authors reported a thematic

preference in children aged from 4 to 6 years,

followed by a taxonomic preference at around 10

years of age (Smiley & Brown, 1979), whereas

others found a thematic preference at all ages

from 3 to 15 years (Greenfield & Scott, 1986).

More recently, preference has been shown to vary

(1) among children (Dunham & Dunham, 1995)

and (2) across situations (e.g., Baldwin, 1992;

Deak & Bauer, 1995; Golinkoff, Shuff-Bailey,

Olguin, & Ruan, 1995; Imai, Gentner, & Uchida,

1994; Markman & Hutchinson, 1984; Waxman

& Namy, 1997). Additionally, other research

(Osborne & Calhoun, 1998; Walsh, Richardson,

& Faulkner, 1993) underlined the influence of the

stimuli themselves on taxonomic and thematic

choices. Matching seems to depend strongly on

the specific exemplars selected by the researcher to

instantiate each type of relation (see also Lin &

Murphy, 2001, for similar results in adults).

For instance, previous research showed the critical

role of perceptual similarity between exemplars. In

matching studies in which this factor was not

controlled, it was concluded that young children

generalized new names on the basis of taxonomic

membership (e.g., Markman & Hutchinson, 1984).

But in following experiments in which perceptual

similarity and taxonomic membership were decor-

related, authors reported that children relied on both

factors to generalize new names (e.g., Baldwin,

1992). Analogously, some ‘‘category-specific’’ defi-

cits disappeared when the characteristics of the

presented items such as frequency, familiarity, or

visual complexity were controlled (e.g., Caramazza

& Shelton, 1998).

The effect of associative strength on recall,

Int Jnt Psychology (gamma) IJP35365.3d 29/1/04 17:08:45 Rev 7.51n/W (Jan 20 2003)

The Charlesworth Group, Huddersfield 01484 517077

2 SCHEUNER ET AL.

clustering, recognition, or priming has been

previously studied in adults (e.g., Cramer &

Eagle, 1972; La-Heij, Dirkx, & Kramer, 1990;

Mathews, Maples, & Elkins, 1981) and in children

(e.g., Bjorklund & de Marchena, 1984; Frankel

& Rollins, 1985; Krackow & Gordon, 1998;

McCauley, Weil, & Sperber, 1976; Nation &

Snowling, 1999). For instance, Nation and Snowling

tried to dissociate pure semantic priming

(activation of semantic information only) from

‘‘associate’’ priming for taxonomically and func-

tionally (i.e., thematically) related items in 10-

year-olds. Children were presented with a verbal

priming task in which prime and target words

were either highly associated or not (verbal

association task). Results showed that the priming

effect depended on the level of associative

strength for both relations. Hence, associative

strength seems largely involved in automatic

priming effects in children for taxonomic and

thematic relations (see also McCauley et al., 1976,

for priming with pictorial stimuli in 6- and 8-year-

olds). Yet its effects in categorization tasks in

which less automatic comparison processes are

likely to be required remain unexplored.

The aim of this research is to give an account of

the influence of associative strength on matching

behaviour apart from the conceptual relations

shared by the objects. If an associate is more

strongly related to the target than to the other

object, its activation might be more automatic

and might compete or interfere with the more

controlled comparison process that is supposedly

required when choosing between several associ-

ates (Crowley, Shrager, & Siegler, 1997).

Thematic relations among objects come mainly

from individual experience with specific episodes

in which these objects were involved. Therefore,

association strength should be highly variable

depending on the specific presented objects. Alter-

natively, taxonomic relations are likely to be

formed on multiple bases: By extracting common

properties among objects, perceptual and more

abstract ones, and by noticing generic names. Yet,

one can assume that the extraction of some common

properties also derives from episodic encoding. It

might be the case for perceptual properties and

for some functional properties that might emerge

because objects play the same role in a given event or

scene (i.e., slot-filler categories such as things eaten

at breakfast; Nelson, 1983; 1986). The potential

episodic basis of taxonomic relations was simulated

by connectionist networks (Mareschal & French,

2000; Quinn, French, & Mareschal, 2000). With

perceptual attributes as inputs, networks performed

taxonomic-like categorization, and reproduced

categorization behaviour of 3-monTh–old infants

(Quinn, Eimas, & Rosenkrantz, 1993). In simula-

tions like those in infants, the extension of cate-

gories depended on the level of perceptual similar-

ity. Hence, perceptual similaritymightmodulate the

associative strength between taxonomically related

objects in infants. It might also be the case in young

children, since perceptual similarity still influences

categorization behaviour later in childhood (e.g.,

Baldwin, 1992; Jones & Smith, 1993; Landau,

Smith, & Jones, 1988). The direct consequence is

that associative strength should probably depend on

the hierarchical level of categories, exemplars from

basic-level categories being more strongly asso-

ciated than exemplars from superordinate ones

because of their greater perceptual similarity

(Rosch, Mervis, Gray, Johnson, & Boyes-Braem,

1976). Even perceptually dissimilar objects are likely

to be more strongly associated when they belong to

a basic-level category because they also share more

nonperceptual common properties than exemplars

from superordinate categories (Rosch et al., 1976).

It is assumed that exemplars from slot-filler

categories should be associated at an intermediary

level: As superordinate categories, they share few

common properties with the target, but unlike them

they belong to a common context, which should

increase associative strength.

In a very different theoretical framework,

Nelson’s conception also illustrates the experien-

tial (or episodic) basis of taxonomic categories,

which might derive from event schemas (Nelson,

1983, 1986). Taxonomic categories would first

appear as contextualized groupings with a limited

extension—slot-filler categories are composed of

objects of the same kind that belong to a given

schema—before becoming decontextualized group-

ings, i.e., superordinate categories. Hence, associa-

tive strength between taxonomically related objects

might vary depending on whether they share

thematic relations or not. It should be the case in

both children and adults because schemas still play

an important role in categorization later in life (Lin

& Murphy, 2001; Ross & Murphy, 1999; Vallee-

Tourangeau, Anthony, & Austin, 1998). All these

arguments contribute to emphasize the link that

might exist between taxonomic representations and

episodic encoding in all individuals.

As previously noted, performances in matching

tasks are generally reported as mean percentages

of taxonomic or thematic choices computed

across subjects and items. From these means,

conclusions are drawn regarding the accessibility

of taxonomic and thematic relations as a function

Int Jnt Psychology (gamma) IJP35365.3d 29/1/04 17:08:45 Rev 7.51n/W (Jan 20 2003)

The Charlesworth Group, Huddersfield 01484 517077

MATCHING BY ASSOCIATIVE STRENGTH IN CHILDREN 3

of age and/or situation. But associative strength

between targets and associates are never con-

trolled. Some authors checked by mean of a

control group that children properly identified the

presented associations (Blaye & Bonthoux, 2001;

Waxman & Namy, 1997), but this might be

insufficient since it is unlikely that (1) two

associates would be equivalently related to a

given target, (2) all the associates of a same type

(taxonomic or thematic) would be equivalently

related across targets, and (3) the associative

strength of a given associate would be judged

equivalent by different children. Altogether, these

considerations lead us to study the relative effects

of associative strength and type of relation,

taxonomic or thematic, in determining children’s

matching choices.

In the following experiments, associative

strength between target and associates, a factor

assumed to be critical but which is generally not

controlled, and type of relation (thematic and

taxonomic) were manipulated independently to

determine their respective effects in matching

tasks; perceptual similarity was controlled and

maintained at a low level.

EXPERIMENT 1

Performances of 4- and 6-year-old children were

assessed in a matching task after they had judged,

in a previous session, the associative strength

between targets and several associated pictures.

These judgments served to construct the sets of

stimuli used in the matching task for each child. It

is worth noting that in memory studies, associative

strength corresponded to the production frequency

of words in verbal association or exemplar genera-

tion tasks, and thus was a measure of lexical

association. Here, since the matching task involved

pictorial stimuli, judgments of associative strength

were made on pictures.

To show the influence of associative strength

and conceptual relation on matching, each target

was presented along with two associated pictures

of opposite associative strength in two types of

configurations. In homogeneous configurations,

both associates shared the same conceptual

relation with the target (i.e., a strong and a

weak thematic associate, Thz and Th2, or a

strong and a weak taxonomic associate, Taz and

Ta2). In heterogeneous configurations, concep-

tual relations differed (i.e., a strong thematic and

a weak taxonomic associate, Thz and Ta2, or a

strong taxonomic and a weak thematic associate,

Taz and Th2). Children were required to choose

the best match with the target (‘‘which one goes

best with?’’). This instruction was selected because

it was nonconstrained. Since it does not specify

whether the child must choose an object of the

same kind as the target (taxonomically related)

or an object of a different kind, which belongs

to the same event or scene as the target

(thematically related), it allows us to observe

how children spontaneously interpreted the task

(Deak & Bauer, 1995).

Hence, if matching choices under noncon-

strained instructions were predominantly guided

by associative strength, strong associates should

be more frequently chosen than weak associates in

all configurations. Alternatively, if both associa-

tive strength and conceptual relation had an

influence on matching choices, this pattern should

still be observed in homogeneous configurations

but not in heterogeneous configurations. More pre-

cisely, because in previous research young children

demonstrated thematic preference in conditions of

nonconstrained instructions (i.e., ‘‘best match,’’ see,

for example, Smiley & Brown, 1979), they might

choose thematic associates even when these associ-

ates are weakly related to the targets.

Method

Participants

Twenty-two French children participated in the

main study: eleven 4-year-olds (second year in

nursery school, mean age 4;9 years; range 4;4 to

5;2) and eleven 6-year-olds (first grade in primary

school, mean age 6;11 years; range 6;7 to 7;2).

Written parental consent for the children’s

participation was obtained. Two children (one

4-year-old and one 6-year-old) were not included

in the analyses due to poor understanding of the

associative strength judgment task.

Twelve additional children (six from each age

group) participated in a pretest phase. They were

administered a verbal association task which

helped to choose items. Ten adults rated the

perceptual similarity between the targets and their

corresponding associates.

Stimuli

Seventy-seven black-and-white drawings of

objects included in a 769 cm rectangle were

used as stimuli: There were11 targets and, for

each target, 3 taxonomic and 3 thematic associ-

ates (see AppendixA). Perceptual similarity

between each associate and its corresponding

target was low. Fourteen targets were initially

Int Jnt Psychology (gamma) IJP35365.3d 29/1/04 17:08:45 Rev 7.51n/W (Jan 20 2003)

The Charlesworth Group, Huddersfield 01484 517077

4 SCHEUNER ET AL.

selected but three of them were excluded because

the mean similarity rating of at least one pair of

target and associate exceeded 4 on a scale from 1

(perceptually dissimilar) to 9 (perceptually similar).

The three taxonomic and the three thematic

associates of a given target were chosen to

correspond a priori to three levels of associa-

tive strength (strong, medium, and weak, see

AppendixA). Taxonomic associates corresponded

to three hierarchical levels. For instance, for the

target ‘‘dog,’’ the supposedly strongest associate

(Ta1) was another dog, i.e., an exemplar from the

same basic-level category; the supposedly medium

associate (Ta2) was a guinea-pig, i.e., an exemplar

from the same slot-filler category of pets; and the

supposedly weakest associate (Ta3) was a snake,

i.e., an exemplar from the same superordinate

category of animals.

Thematic associates were selected from a verbal

association task. During a pretest phase, children

were asked to say the first words that came to

their mind in response to target names. The most

frequent names produced at the beginning of the

task were considered a priori as the strongest

associates (Th1), the least frequent names pro-

duced later as the weakest associates (Th3), and

the medium associates (Th2) chosen from names

that meet neither of these criteria. It was expected

that children would judge Ta1, Ta2, Ta3, and

Th1, Th2, Th3 as being sufficiently different in

associative strength to be able to extract a strong

(z) and a weak (2) associate for each target and

type of relation (Ta and Th) for each child.

Procedure

Children were run individually in a quiet room of

their school. All of them participated in two

successive sessions separated by a 1-week delay.

First, they judged the associative strength of all

the pairs targets and associates. Ratings were then

analysed to extract for each child and each target

a strong and a weak thematic associate (Thz and

Th2) and a strong and a weak taxonomic

associate (Taz and Ta2). Second, children

performed a matching task in which strong and

weak associates were contrasted.

Session 1:Judgments of associativestrength

Children made two successive series of judg-

ments. They first judged the associative strength

on a scale from 0 to 10 for all the pair targets and

associates (6 judgments—Ta1, Ta2, Ta3 and Th1,

Th2, Th3—for each of 11 targets). The target was

presented above the associate. To adapt the task

for young children, the scale was analogous to the

pain scale used by physicians. The child was

asked: ‘‘show me with the little hand (the cursor)

if both pictures go together very strongly (show-

ing the highest levels at the top of the scale),

moderately strongly (showing the levels at the

middle of the scale) or not strongly (showing the

lowest levels at the bottom of the scale).’’ At

the start of the session, an example of judgment

was provided; details were added as to the

differentiation between degrees for each level—

strongly, moderately strongly, and not strongly

associated items. Then, the child made a series of

12 judgments (2 targets 6 6 associates) to learn to

use the scale. During these familiarization trials

(see AppendixA), which are not included in the

analyses, additional explanations were sometimes

given. Finally, the child judged the remaining

pairs (9 targets 6 6 associates), which were

presented in a fixed pseudorandom order (a given

target never appeared on two successive trials).

This spontaneous judgment phase was followed

by a more constrained one to ensure that a strong

and a weak associate could be extracted for each

child, target, and type of relation. The three

taxonomic or three thematic associates were

shown simultaneously with each corresponding

target. The child was required to order the three

pictures as a function of their associative strength

with the target.

Spontaneous associative strength judgments

were analysed first. For each child, when the

three judgments (three taxonomic or three

thematic associates for a given target) differed

between each other by at least 1 point, weak and

strong associates were chosen on the sole basis of

the spontaneous judgments. Alternatively, con-

strained judgments were used (1) when the three

associates were judged to be equivalent and (2)

when two associates were judged to be equiva-

lent and the third was clearly distinct. In this case,

if both spontaneous and constrained judgments

indicated the same strong or weak associate,

constrained judgments served to select among the

two others. If, however, spontaneous and con-

strained judgments were not consistent regarding

the distinct associate, the spontaneous judgment

was retained for it and the choice between the two

others was made at random (this last method

concerned less than 10% of the judgments at each

age). From these analyses, a strong and a weak

taxonomic associate (Taz and Ta2) and a strong

and a weak thematic associate (Thz and Th2)

Int Jnt Psychology (gamma) IJP35365.3d 29/1/04 17:08:46 Rev 7.51n/W (Jan 20 2003)

The Charlesworth Group, Huddersfield 01484 517077

MATCHING BY ASSOCIATIVE STRENGTH IN CHILDREN 5

were selected for each child and each target to be

used in the following matching task.

Session 2: Matching task

A week later, children performed a picture-

matching task. On each trial, the target was

shown first, followed by two comparison pictures

associated to the target. Associates were placed

side by side below the target. After pointing to the

target (saying ‘‘see this one?’’), the experimenter

asked, pointing successively to the two compar-

ison pictures: ‘‘which one goes best with it?’’ (in

French: ‘‘lequel va le mieux avec?’’)

Children were asked to choose among two

associates, one strongly and the other weakly

associated to the target. They were shown four

configurations for each target. In two configura-

tions, called homogeneous, either two taxonomic

associates (TazTa2) or two thematic associates

(ThzTh2) were contrasted. In two other con-

figurations, called heterogeneous, a taxonomic

and a thematic associate were contrasted

(ThzTa2 and TazTh2). As in the judgment

session, two targets served in familiarization trials

and nine targets in test trials (four configurations

for each). There were therefore 8 familiarization

triads and 36 test triads; half were homogeneous

and half were heterogeneous configurations. The

nine test targets appeared in a different random

order in each of four blocks with the four types of

configurations roughly counterbalanced across the

blocks. The spatial position of the two associates

(strongly and weakly related, and taxonomic and

thematic for heterogeneous configurations) was

counterbalanced across items in each block.

Results

As judgment of associative strength is an original

methodology, the first analyses are intended to

validate the scale. The following analyses are

aimed at testing the effects of associative strength

and type of relation in the matching task.

Judgments of associative strength

At a descriptive level, mean judgments of

associative strength reflected fairly the degrees

established prior to the experiment. For taxo-

nomic associates, associative strength covaried

with category level: 66% of the strong taxonomic

associates (Taz) were basic-level associates (Ta1)

and 66% of the weak taxonomic associates (Ta2)

were superordinate associates (Ta3); slot-filler

associates (Ta2) corresponded equally often to

strong and weak associates (respectively, 24% and

23%). For thematically related pictures, associa-

tive strength was more variable but reflected

roughly the levels established during the pretest:

strongly related pictures (Thz) were predomi-

nantly Th1 associates (59%) and weakly related

pictures (Th2) were predominantly Th3 associ-

ates (54%). This means that the task was generally

well understood and that judgments of associative

strength appeared to be valid. To check that

strong and weak associates were well differ-

entiated, mean judgments for each configuration

of the subsequent matching task were com-

pared (Taz vs. Ta2, Thz vs. Th2, Thz vs.

Ta2, and Taz vs. Th2). All t-tests comparisons

were significant in 4- and in 6-year-old children

(pv.001).

Matching task

The following analyses concerned the matching

task. The dependent variable was the number of

strong associates (‘‘zchoices’’) chosen by age

(4 vs. 6 years) and configuration (ThzTh2,

TazTa2, ThzTa2, TazTh2, see Table 1). To

test the global effect of associative strength, t-tests

were first performed against chance for homo-

geneous configurations (ThzTh2 and TazTa2)

at each age. As indicated in Table 1, all percen-

tages significantly exceeded 50%. These results

show that associative strength reliably influenced

choices in young children. Then, two alternative

hypotheses about the importance of associative

strength could be differentiated. If associative

strength was the main determinant of matching

performances, then ‘‘zchoices’’ should be still

predominant in heterogeneous configurations.

Alternatively, if both associative strength and

type of conceptual relation influenced choices,

then ‘‘zchoices’’ should vary in heterogeneous

configurations depending on which associate was

the most strongly related to the target. Because it

has been shown that young children preferred

thematic relations under nonconstrained instruc-

tions (Greenfield & Scott, 1986; Smiley & Brown,

1979; Waxman & Namy, 1997), their ‘‘zchoices’’

might be predominant in ThzTa2 configurations

but not in TazTh2 configurations.

Results are not consistent with the first hypo-

thesis and they support the alternative one, since

‘‘zchoices’’ were predominant in ThzTa2 con-

figurations (t-tests against 50% were significant at

both ages), but did not differed from chance level

in TazTh2 configurations (Table 1). Hence,

associative strength is not the only factor affecting

Int Jnt Psychology (gamma) IJP35365.3d 29/1/04 17:08:46 Rev 7.51n/W (Jan 20 2003)

The Charlesworth Group, Huddersfield 01484 517077

6 SCHEUNER ET AL.

young children’s performances in the matching

task; conceptual preference also plays a role.

An analysis of variance was then performed on

the number of ‘‘zchoices’’ with age (4 vs. 6) as a

between-subjects factor, configurations (homo-

geneous vs. heterogeneous) and type of relation of

the strong associate (Th vs. Ta) as within-subjects

factors. There was a main effect of age, F(1, 18)~

9.91, pv.01, and an interaction between age and

type of relation, F(1, 18)~5.21, pv.05. The

‘‘zchoices’’ produced by 6-year-old children

exceeded those of 4-year-olds (89% and 71%,

respectively) but the superiority of 6-year-olds’

‘‘zchoices’’ was greater when the strong associate

was thematic rather than taxonomic. The analysis

also revealed main effects of configuration,

F(1, 18)~21.13, pv.01, and type of relation,

F(1, 18)~6.47, pv.05, and an interaction between

both factors, F(1, 18)~39.73, pv.01. Homo-

geneous configurations leaded to more ‘‘zchoices’’

than heterogeneous ones and ‘‘zchoices’’ were

more often thematic than taxonomic. More impor-

tantly, regarding the influence of associative

strength and conceptual relation, the strong

associate was more frequently chosen in homo-

geneous than in heterogeneous configurations

when it was taxonomic, HSD Tukey test,

pv.05, but equivalently chosen in both con-

figurations when it was thematic, HSD Tukey

test, pw.10.

Discussion

The original methodology used here revealed that

valid assessments of associative strength could be

collected in 4- and 6-year-old children and that

this factor played an important role in matching

tasks.

Judgments of associative strength were col-

lected for taxonomically and thematically related

pairs of pictures for which perceptual similarity

was low. Regarding taxonomic relations, strong

associations corresponded mainly to basic-level

pairs (e.g., a dog with the target ‘‘dog’’) and weak

associations to superordinate pairs (e.g., a snake

with the target ‘‘dog’’). Hence, the order of

associative strength between picture pairs

reflected roughly taxonomic hierarchies in 4-

and 6-year-old children, independently of percep-

tual similarity. Concerning thematic relations,

stimuli were selected from a verbal association

task since no a priori hierarchy could be

established apart from the children’s experience

with objects. Here too, children’s estimates

reflected the predicted association levels. Hence,

associative strength between pictures covaried

with hierarchical levels for taxonomic relations

and with verbal associations for thematic rela-

tions. These results contribute to validate the

judgment scale of associative strength in young

children.

Individual judgments were then used to build

triads in which a target was presented along with

a strong and a weak associate. Results in

homogeneous configurations (two thematic or

two taxonomic associates) indicated that decision

was based on the associative strength at both ages

since ‘‘zchoices’’ were always predominant.

However, associative strength was not the

unique decision basis because ‘‘zchoices’’

varied in heterogeneous configurations depending

on whether the strong associate was taxonomi-

cally or thematically related to the target: thema-

tic ‘‘zchoices’’ were predominant (ThzTa2

configurations) whereas taxonomic ‘‘zchoices’’

were not (TazTh2 configurations).

Finally, at the developmental level, effects of

both associative strength and thematic preference

seem to increase with age, as revealed by (1)

the superiority of older children’s ‘‘zchoices’’

Int Jnt Psychology (gamma) IJP35365.3d 29/1/04 17:08:46 Rev 7.51n/W (Jan 20 2003)

The Charlesworth Group, Huddersfield 01484 517077

TABLE 1

Mean number of ‘‘zchoices’’ (mean percentages in parentheses) observed in Experiments 1 and 3 as a function of configurations

and age; values of t-tests against 50% and corresponding p levels are indicated below the corresponding means

Experiment 1 Experiment 3

Homogeneous configurations ‘‘goes

best’’ instructions

Heterogeneous configurations

‘‘goes best’’ instructions

Heterogeneous configurations

‘‘same kind’’ instructions

ThzTh2 TazTa2 ThzTa2 TazTh2 TazTh2 ThzTa2

4-year-olds 5.90 (66%) 6.90 (77%) 5.90 (66%) 4.70 (52%) 4.80 (53%) 6.30 (70%)

t (9)~2.32 (9)~5.24 (9)~2.32 (9)v1 (9)v1 (9)~2.24

p v.05 v.001 v.05 .052

6-year-olds 7.90 (88%) 8.20 (91%) 8.80 (98%) 4.80 (53%) 7.80 (87%) 6.60 (73%)

t (9)~8.36 (9)~9.52 (9)~32.3 (9)v1 (9)~5.96 (9)~2.70

p v.001 v.001 v.001 v.001 v.05

MATCHING BY ASSOCIATIVE STRENGTH IN CHILDREN 7

in homogeneous configurations and (2) the

interaction between age and type of relation,

which showed that ‘‘zchoices’’ were even more

frequent in 6- than in 4-year-old children for

thematic rather than for taxonomic associates. In

accordance with the results of Smiley and Brown

(1979) and Greenfield and Scott (1986), these data

suggest that thematic preference does not dis-

appear between 4 and 6 years of age. On the

contrary, it is assumed that increasing experience

with objects reinforces both associative strength

and thematic links.

To test further the influence of associative

strength and type of conceptual relation with age,

complementary data were collected. The goal was

to directly observe the variation of a given choice

as a function of the associative strength level of

the alternative option (e.g., the variation of

thematic choices as a function of the associative

strength level of taxonomic associates). Therefore,

the performances obtained in heterogeneous

configurations (ThTa) of Experiment 1, in which

the associates were of opposed strength (z and

2), were compared to those of equivalent strength

(z and z or 2 and 2) in identical instructions

conditions.

EXPERIMENT 2

Experiment 1 revealed the role of associative

strength on matching performances. This

factor interacted sometimes with the greater

availability of thematic relations in 4- and

6-year-old children. To further support this

interpretation and to better qualify the processes

that might underlie choices, performances

obtained in the heterogeneous configurations,

with either two strong or two weak associates

(ThzTaz and Th2Ta2), were compared with

those obtained in the heterogeneous configura-

tions of Experiment 1, where the two associates

had different associative strengths (ThzTa and

TazTh2).

If associative strength was the main deter-

minant of choices, then ‘‘zchoices,’’ either

thematic or taxonomic, should increase when

the associative strength of the alternative

option decreases. In contrast, no variation of

‘‘zchoices’’ with associative strength of the

alternative option would be observed if this

factor had no impact. In addition, a difference

between age groups would show developmental

differences in the weighting of associative strength

and conceptual relations in this situation.

Method

Participants

The participants were twenty children: ten

4-year-olds (second year in nursery school,

mean age 4;9 years; range 4;8 to 5;0) and ten

6-year-olds (first grade, mean age 6;11 years;

range 6;7 to 7;4). They came from schools located

in the same neighbourhood as those of Exp. 1. A

written parental consent was required. Each child

performed the tasks individually in a quiet room

of his/her school.

Stimuli

The pictures were those used in Experiment 1

(see AppendixA).

Procedure

The procedure was almost identical to the one

used in Exp. 1: Children made two successive

series of judgments of associative strength and, 1

week later, performed a matching task with ‘‘goes

best’’ instructions. As previously, the triads used

in the matching task were constructed for each

child as a function of his/her preceding judgments.

The only difference with Exp. 1 was that each

target appeared only twice in the familiarization

and test trials of the matching task since only two

heterogeneous configurations (TazThz and

Th2Ta2) were proposed. There were therefore

4 familiarization triads (2 targets) and 18 test

triads (9 targets). The nine test targets appeared in

a different pseudorandom order in each of two

blocks with the two configurations roughly

counterbalanced across the blocks. The spatial

position of taxonomic and thematic associates

was counterbalanced across items in each block.

Results

Judgments of associative strength

Results were similar to those of Exp. 1. for

judgments of associative strength. For taxonomic

associates, associative strength covaried with

category level: 56% of the strong taxonomic

associates were basic-level associates (Ta1) and

55% of the weak taxonomic associates (Ta2)

were superordinate associates (Ta3); slot-filler

associates (Ta2) corresponded equally often to

strong and weak associates (respectively, 28.5%

and 25.5%). For thematic associates, responses

correspond roughly at the levels established

during the pretest: Strongly related pictures

Int Jnt Psychology (gamma) IJP35365.3d 29/1/04 17:08:47 Rev 7.51n/W (Jan 20 2003)

The Charlesworth Group, Huddersfield 01484 517077

8 SCHEUNER ET AL.

were predominantly Th1 associates (54.5%) and

weakly related pictures were often Th3 associates

(47%).

Matching task

In the matching task, thematic choices were

predominant in ThzTaz configurations: 66%

and 63% in 4- and 6-year-olds, respectively. The

comparison with chance level (i.e., 50%) was

significant in both groups, respectively, t(9)~4.02,

pv.01, and t(9)~2.32, pv.05. Thematic choices

were also predominant in Th2Ta2 configura-

tions in 6-year-olds (76%), t(9)~7.04, pv.001,

but not in 4-year-olds (53%), t(9)~1.03, pw.10.

Performances in heterogeneous configurations

when both options were differently associated

(Exp. 1) or equivalently associated (Exp. 2) were

then compared first on taxonomic choices and

second on thematic choices1. Taxonomic choices,

either strongly or weakly associated, were

expected to be more frequent when they were

opposed to weak rather than to strong thematic

associates (i.e., in TazTh2 than in TazThz

configurations and in Ta2Th2 than in Ta2Thz

configurations). To test these comparisons, an

analysis of variance was performed on the number

of taxonomic choices, with age (4 vs. 6 years) and

type of configuration (different or equivalent

strength) as between-groups factors, and with

strength of Ta (z and 2) as a repeated measure.

The planned comparisons contrasting types of

configurations (different and equivalent strength)

for Taz and Ta2 confirmed the predictions.

Taxonomic choices decreased from TazTh2 to

TazThz configurations (parts a1 and a2 of

Figure 1), F(1, 36)~4.16, pv.05 and F(1, 36)~

3.66, p~.06 at 4 and 6 years, respectively, and

from Ta2Th2 to Ta2Thz configurations (parts

b1 and b2 of Figure 1), F(1, 36)~4.21, pv.05,

and F(1, 36)~13.93, pv.001.

An analogous prediction was made for thematic

choices: It was expected that they would be

modified as a function of the associative strength

level of the corresponding taxonomic associates.

The analysis of variance performed on thematic

choices revealed the expected effects in 6-year-

olds: Thematic choices were more frequent in

ThzTa2 than in ThzTaz configurations,

F(1,36)~24.89, pv.001 and in Th2Ta2 than

in Th2Taz configurations, F(1, 36)~13.10,

Int Jnt Psychology (gamma) IJP35365.3d 29/1/04 17:08:47 Rev 7.51n/W (Jan 20 2003)

The Charlesworth Group, Huddersfield 01484 517077

1It is worth noting that these analyses were indepen-dent: For example, TzTh2 configurations werecompared with TzThz configurations on taxonomicchoices whereas they were compared with T2Th2configurations on thematic choices.

Figure 1. Mean percentages of (a) taxonomic choices (z or 2) as a fUnction of associative strength of thematicassociates in 4-year-olds (al) and in 6-year-olds (a2); and (b) thematic choices (z or 2) as a fUnction of associativestrength of taxononiic associates in 4-year-olds (a2) and in 6-year-olds (b2).

MATCHING BY ASSOCIATIVE STRENGTH IN CHILDREN 9

pv.001 (Figure 1). However, 4-year-olds’ thema-

tic choices (either strongly or weakly related to

the target) did not vary as a function of

associative strength of the corresponding taxo-

nomic associate (Fsv1): They were predominant

when strongly associated (in ThzTaz and

ThzTa2 configurations) but did not differ

from chance level when weakly associated (in

Th2Taz and Th2Ta2 configurations), indepen-

dent of the associative strength of the taxonomic

associate.

Discussion

Regarding the associative strength judgment

session, results closely replicated those obtained

in Experiment 1, in which the procedure was

identical. Therefore this experiment adds further

validity to the rating scale of associative strength

used with 4- and 6-year-old children.

As to the matching task, data provide further

evidence that children’s performances were depen-

dent on associative strength. In Exp. 1, this

conclusion was essentially drawn from the pre-

dominance of ‘‘zchoices’’ in both homogeneous

configurations, ThzTh2 and TazTa2. This

conclusion can now be extended since taxonomic

associates were chosen more or less frequently as

a function of the associative strength level of the

corresponding thematic associates at both ages. A

similar pattern of results was observed for

thematic choices in 6-year-olds, but not in

4-year-olds. Except for this last issue, which will

be commented on later, these data confirm the

influence of associative strength in matching

tasks.

Data of Exp. 2 also support the role of young

children’s conceptual orientation in this task.

Although the associative strength of the two

associates was equated at the individual level (two

strong or two weak associates), thematic choices

were predominant in 6-year-old children in both

configurations (ThzTaz and Th2T2) and in

4-year-old children in ThzTaz configurations.

In summary, further evidence is provided that

associative strength and greater availability of

thematic relations interacted in young children’s

choices, as was already shown in the first

experiment.

Returning to developmental differences, the

performances obtained by 4-year-old children

differed from those of 6-year-olds on two aspects:

(1) there was no variation of thematic choices as a

function of associative strength of the correspond-

ing taxonomic associates, and (2) there was no

preference in Th2Ta2 configurations. In fact,

whatever the associative strength of the taxo-

nomic associates, children chose predominantly

the thematic associates when they were strongly

associated to the target but they distributed their

choices equivalently when these associates had a

low associative strength. The first result suggests

that younger children relied less consistently on

associative strength than older ones, as was

already found in Exp. 1. The second result

shows that they were biased towards thematic

relations only when these relations corresponded

to strong associations, whereas responses were at

chance in the case of weak thematic associations.

In other matching studies using identical

instructions (i.e., ‘‘goes best’’), a thematic pre-

ference was observed in 4- and 6-year-olds (Smiley

& Brown, 1979, 4- and 6-year-olds; Waxman &

Namy, 1997, 4-year-olds). Yet this preference

disappeared with ‘‘goes with’’ instructions and

reversed totally when children were required to

find ‘‘another one’’ (Waxman & Namy, 1997).

Analogously, instruction effects were observed in

a matching task contrasting two associates, one

perceptually and the other taxonomically related

to the target (Deak & Bauer, 1995): 4-year-old

children produced predominant perceptual

choices when they were asked to choose ‘‘the

most like’’ as the target but predominant

taxonomic choices were observed when children

were asked to choose ‘‘the same kind of thing’’ as

the target. These findings show that instructions

can orient children towards taxonomic choices

and that the expression of conceptual preference

is context-dependent.

Consequently, the effect of instructions was

studied in the case of strong thematic links,

in which matching seems to be predominantly

stimulus-driven. If it were the case, constrained

instructions orienting children towards taxo-

nomic relations would conflict with associative

strength in ThzTa2 configurations and the

thematic preference would disappear. Addi-

tionally, this type of instruction should act in

the same direction as associative strength and

reinforce its effect in TazTh2 configurations.

These predictions were tested in the following

experiment.

EXPERIMENT 3

To determine whether children’s matching beha-

viour was modified by instructions, another group

of 4- and 6-year-old children performed the same

tasks as in Exp. 1 and 2 (judgments of associative

Int Jnt Psychology (gamma) IJP35365.3d 29/1/04 17:09:08 Rev 7.51n/W (Jan 20 2003)

The Charlesworth Group, Huddersfield 01484 517077

10 SCHEUNER ET AL.

strength and matching). The matching task was

proposed in a constrained condition of instruc-

tions intended to orient towards taxonomic

relations. Children were required to choose the

picture that was the ‘‘same kind’’ as the target

(Deak & Bauer, 1995). Only heterogeneous con-

figurations with contrasted associative strength

(ThzTa2 and TazTh2) were presented. It

was expected that the ‘‘same kind’’ instructions

would lead to increased taxonomic choices as

compared with the ‘‘goes best’’ instructions

used in Exp. 1. Moreover, comparing instruction

conditions might also reveal developmental differ-

ences and help to specify the basis of matching

behaviour in 4- and 6-year-old children.

Method

Participants

Twenty children participated in the experiment:

ten 4-year-old (second year in nursery school,

mean age 4;7 years; range 4;4 to 4;11) and

ten 6-year-old children (first grade, mean age

6;11 years; range 6;7 to 7;4). They came from

schools located in the same neighbourhood

as those of Exp. 1 and 2. Written parental

consent was required. Each child performed the

tasks individually in a quiet room of his/her

school.

Stimuli

The stimuli were those used in Exp.1 and 2.

Procedure

The procedure was the same as in the previous

experiments: Children made two successive series

of judgments of associative strength and, 1 week

later, performed a matching task in which the

presented triads were constructed for each child as

a function of his/her preceding judgments. The

main difference with Exp. 1 and 2 concerned the

instructions used in the matching task: Children

were asked to choose among two associates ‘‘the

same kind of thing’’ (i.e., in French ‘‘la meme

sorte de chose’’) as the target (instead of the ‘‘goes

best’’ picture). Each target appeared twice in

familiarization and test trials since only two

heterogeneous configurations with associates of

contrasted strength (TazTh2 and ThzTa2

configurations) were presented. There were there-

fore 4 familiarization triads (2 targets) and 18 test

triads (9 targets). The nine test targets appeared in

a different pseudorandom order in each of two

blocks with the two configurations roughly

counterbalanced across the blocks. The spatial

position of both associates was counterbalanced

across items in each block.

Results

The results obtained in the judgment session

replicated fairly those of Exp.1 and 2. Regarding

the matching task, ‘‘zchoices’’ (Table 1) were still

predominant in ThzTa2 configurations at both

ages (70% and 73% in 4- and 6-year-old children,

respectively); despite the ‘‘same kind’’ instruc-

tions, which were intended to orient choices

toward taxonomic relations, t-tests against

chance were at least marginally significant

(Table 1). By contrast, in the TazTh2 config-

urations, taxonomic ‘‘zchoices’’ became largely

predominant in 6-year-olds (87%) as predicted,

but remained at chance in 4-year-olds (53%).

These results are similar to those obtained in

Exp. 1.

Performances in this instruction condition

(‘‘same kind’’) were then compared with those

obtained in Exp. 1 (‘‘goes best’’ instruction) for

the configurations that were presented in both

experiments (ThzTa2 and TazTh2). An analy-

sis of variance was performed on the number of

‘‘zchoices,’’ with age (4 vs. 6 years) and

instruction as between-groups factors and type

of configuration (ThzTa2 vs. TazTh2) as a

repeated measure. The difference between instruc-

tion conditions was significant in 6-year-olds for

both configurations, F(1, 36)~7.07, pv.05, and

F(1, 36)~8.91, pv.01, in ThzTa2 and

TazTh2 configurations, respectively. As pre-

dicted, the ‘‘same kind’’ instruction as compared

to the ‘‘goes best’’ instruction increased

‘‘zchoices’’ in TazTh2 configurations (from

53% to 87%) and lowered them in ThzTa2

configurations (from 98% to 73%). However, this

was not the case in 4-year-olds (F valuesv1 for

both configurations). Instructions did not modify

younger children’s choices.

Discussion

The main result was that the ‘‘same kind’’

instructions, as compared with the ‘‘goes best’’

instructions, modified choices in 6-year-olds only:

As expected, ‘‘zchoices’’ in this group increased

with ‘‘same kind’’ instructions in TazTh2

configurations and decreased in ThzTa2 con-

figurations. Hence, the taxonomic orientation

stemming from instructions conflicted successfully

Int Jnt Psychology (gamma) IJP35365.3d 29/1/04 17:09:08 Rev 7.51n/W (Jan 20 2003)

The Charlesworth Group, Huddersfield 01484 517077

MATCHING BY ASSOCIATIVE STRENGTH IN CHILDREN 11

with the spontaneous thematic orientation observed

under nonconstrained instructions. Percentages of

taxonomic choices were particularly high (87%)

when associative strength acted in the direction of

instructions, as in TazTh2 configurations.

However, instructions were not fully successful

when they were in conflict with both associative

strength and thematic orientation, since in

ThzTa2 configurations, taxonomic choices did

not differ from thematic ones. Finally, data show

again the greater availability of thematic relations,

at least in the case of strong associations in

6-year-old children.

Contrary to the findings of Waxman and Namy

(1997), the modification of instructions did not

alter the matching behaviour of 4-year-old

children. Several factors might explain this

discrepancy. First, the instructions that were

intended to orient towards taxonomic relations

differed between experiments: Children were

asked to ‘‘find another one’’ in the Waxman

and Namy study and to find ‘‘the same kind of

thing’’ in this experiment. The first instructions

might be more efficient than the others because

they refer more directly to the category name.

More importantly, Waxman and Namy’s pro-

cedure began with three pretest trials aimed at

showing the two different kinds of relations:

For thematic relations, the experimenter per-

formed an action with both objects, whereas

for taxonomic relations, he/she only tapped them

together. In contrast, children’s attention was

not drawn to both conceptual relations. Famil-

iarization trials were identical to test trials

and served only as a warming-up phase. One

might consider that the Waxman and Namy

demonstration helped children to discover that

both choice options were differently related to the

target. This procedure is likely to have increased

the activation level of taxonomic relations,

leading to the observed difference between

experiments. Finally, these results might also

result from a linguistic difficulty. ‘‘The same

kind of thing’’ which is a usual form in English,

was translated by ‘‘la meme sorte de chose,’’

which is less usual in French. It is thus possible

that some 4-year-old children did not properly

understand this form.

GENERAL DISCUSSION

The goal of these experiments was to examine the

impact of associative strength and conceptual

preference in young children’s matching

behaviour in the absence of perceptual similarity.

To this end, 4- and 6-year-old children first

rated the associative strength between percep-

tually dissimilar picture pairs, and then per-

formed a matching task in which associative

strength and type of conceptual relation were

manipulated.

The first contribution is to show that valid

assessments of associative strength can be col-

lected in young children. The data revealed that

associative strength covaried with the hierarchical

level for taxonomically related associates, inde-

pendently of perceptual similarity, and with

verbal association for thematically related associ-

ates. Due to the consistency across experiments,

these results justify the collection of association

norms between pictures of objects as a function of

age. These norms will allow improvement of the

control of the stimuli used in young children’s

categorization studies, particularly when studying

categorical flexibility (Blaye & Bonthoux, 2001;

Bonthoux, Blaye, Cannard, Petit, & Seraphin,

2001). However, individual variability was quite

high, which has at least two implications: (1) it

seems necessary to include, besides the group

results, individual patterns of performances when

studying categorization skills in young children,

as, for example, did Golinkoff et al. (1995) or

Waxman and Namy (1997), and (2) diverse

individual patterns in a given situation support

pluralistic developmental models rather than

unitary ones (Lautrey, in press).

The most important result is the evidence

showing the role of associative strength in

matching tasks. Results of Experiments 1 and 2,

in which nonconstrained instructions were used

(‘‘goes best’’), support this claim. In Exp. 1,

young children reliably selected the most strongly

associated picture when both associates were

related to the target in the same way (either

taxonomically or thematically). In Exp. 2, the

choices of strong taxonomic associates varied as a

function of the associative strength of the

thematic alternatives at both ages (the analogous

effect for strong thematic associates was signifi-

cant in 6-year-olds only). Hence, young children’s

matching behaviour depends on associative

strength of both choice options and, thus, seems

partly stimulus driven. These results add further

evidence for the impact of specific instances in

matching tasks, as has already been shown by

other researchers either indirectly (Walsh et al.,

1993), or directly (Osborne & Calhoun, 1998).

Although this issue has rarely been raised in the

developmental literature on categorization, it is

not surprising since adult judgments of similarity

Int Jnt Psychology (gamma) IJP35365.3d 29/1/04 17:09:08 Rev 7.51n/W (Jan 20 2003)

The Charlesworth Group, Huddersfield 01484 517077

12 SCHEUNER ET AL.

and thematic relatedness are still influenced

by stimuli characteristics (Wisniewski & Bassok,

1999).

Associative strength conflicted with thematic

orientation (Exp. 1 and 2) since at both ages (1)

choices were equivalently distributed among

strong taxonomic and weak thematic associates

(Exp. 1), and (2) strong thematic choices exceeded

strong taxonomic choices (Exp. 2; weak thematic

choices also exceeded weak taxonomic choices but

only in 6-year-olds). This means that the proces-

sing of conceptual relations was biased towards

thematic associates in this task. This result was

already reported by others (Smiley & Brown,

1979; Waxman & Namy, 1997) with the same

instructions (‘‘goes best’’). However, data addi-

tionally revealed a greater availability of thematic

relations even when the associative strength of

both thematic and taxonomic associates was

controlled.

However, because of the thematic bias observed

with nonconstrained instructions, the data suggest

a stronger activation of thematic than taxonomic

relations. For instance, when the strong associate

was thematically related to the target, ‘‘zchoices’’

were equally frequent in ThzTh2 configurations

in which the associative strength was the unique

basis of choice and in ThzTa2 configurations

in which both associative strength and type of

relation could influence choices (Exp. 1).

When the strong associate was taxonomically

related to the target, however, ‘‘zchoices’’

dominated only in TazTa2 configurations, not

in TazTh2 configurations; in these latter con-

figurations, the taxonomic associate strongly

related to the target was selected as frequently

as the thematic associate weakly related to it. This

factor leads us to consider that the activation

threshold required to select one associate would

be reached earlier for thematic than for taxo-

nomic associates, independently of the level of

associative strength.

The effects of both associative strength and

thematic orientation were greater in 6- than in

4-year-old children. This implies that the weight-

ing of choice criteria used in matching tasks

change during development, with older children

relying more consistently than younger ones on

associative strength and showing a stronger

preference for thematic relations. One can

assume that, because of their greater experience

with objects, 6-year-old children are more likely

than 4-year-olds to judge the pairs target and

associate as strongly related and to identify the

thematic associates quickly.

Moreover, older children were affected by the

modification of instruction whereas younger

children were not (Exp. 3). The ‘‘same kind’’

instructions, intended to bias responses towards

taxonomic matching, successfully modified

choices in 6-year-olds only. This latter result

suggest that when the instructions required it,

these children were more capable than 4-year-olds

of modifying their spontaneously activated choice

and choosing the other associate. Given the

instruction effects observed by other researchers

in 4-year-old children (Waxman & Namy, 1997),

it is assumed that the wording used and the

absence of any demonstration of both conceptual

relations were not sufficient to modify 4-year-olds’

selection criteria.

In summary, these experiments emphasize the

influence of associative strength and they could

help to reconcile some divergent findings. More

specifically, a lack of homogeneity in associative

strength across items and/or individuals could

partly explain the inter- and intra-individual

variability frequently mentioned in matching

tasks. According to this logic, various patterns

of responses among children (Dunham &

Dunham, 1995) might partially result from

individual differences in the judgments of asso-

ciative strength of a given pair of objects, whereas

within-children variability (Osborne & Calhoun,

1998; Walsh et al., 1993) would stem from

differences across items.

Manuscript received January 2003

Revised manuscript received October 2003

REFERENCES

Baldwin, D. A. (1992). Clarifying the role of shapein children’s taxonomic assumption. Journal ofExperimental Child Psychology, 54, 392–416.

Bauer, P. J., & Mandler, J. M. (1989). Taxonomies andtriads: Conceptual organization in one- to two-year-olds. Cognitive Psychology, 21, 156–184.

Bjorklund, D. F., & de Marchena, M. R. (1984).Developmental shifts in the basis of organizationin memory: The role of associative versus cate-gorical relatedness in children’s free recall. ChildDevelopment, 55, 952–962.

Blaye, A., & Bonthoux, F. (2001). Thematic andtaxonomic relations in preschoolers: The develop-ment of flexibility in categorization choices. BritishJournal of Developmental Psychology, 19, 395–412.

Bonthoux, F., Blaye, A., Cannard, C., Petit, D., &Seraphin, J. (2001). Categorical flexibility in youngchildren: The role of verbal and visual interference.Paper presented at the meeting of the Jean PiagetSociety, Berkeley, CA.

Caramazza, A., & Shelton, J. R. (1998). Domain-specific knowledge systems in the brain: The

Int Jnt Psychology (gamma) IJP35365.3d 29/1/04 17:09:08 Rev 7.51n/W (Jan 20 2003)

The Charlesworth Group, Huddersfield 01484 517077

MATCHING BY ASSOCIATIVE STRENGTH IN CHILDREN 13

animate–inanimate distinction. Journal of CognitiveNeuroscience, 10, 1–34.

Cramer, P., & Eagle, M. (1972). Relationship betweenconditions of CrS presentation and the category offalse recognition errors. Journal of ExperimentalPsychology, 94, 1–5.

Crowley, K., Shrager, J., & Siegler, R. S. (1997).Strategy discovery as a competitive negotiationbetween metacognitive and associative mechanisms.Developmental Review, 17, 462–489.

Daehler, M. W., Lonardo, R., & Bukatko, D. (1979).Matching and equivalence judgments in very youngchildren. Child Development, 50, 170–179.

Deak, G., & Bauer, P. J. (1995). The effects of taskcomprehension on preschoolers’ and adults’categorization choices. Journal of ExperimentalChild Psychology, 60, 393–427.

Dunham, P., & Dunham, F. (1995). Developmentalantecedents of taxonomic and thematic strategies at3 years of age. Developmental Psychology, 31,483–493.

Frankel, M. T., & Rollins, H. A. Jr (1985). Associativeand categorical hypotheses of organization in thefree recall of adults and children. Journal ofExperimental Child Psychology, 40, 304–318.

Golinkoff, R. M., Shuff-Bailey, M., Olguin, R., &Ruan, W. (1995). Young children extend novelwords at the basic level: Evidence for the principle ofcategorical scope. Developmental Psychology, 31,494–507.

Greenfield, D. B., & Scott, M. S. (1986). Youngchildren’s preference for complementary pairs:Evidence against a shift to a taxonomic preference.Developmental Psychology, 22, 19–21.

Imai, M., Gentner, D., & Uchida, N. (1994). Children’stheories of word meaning: The role of shapesimilarity in early acquisition. CognitiveDevelopment, 9, 45–75.

Jones, S. S., & Smith, L. B. (1993). The place ofperception in children’s concepts. CognitiveDevelopment, 8, 113–131.

Krackow, E., & Gordon, P. (1998). Are lions and tigerssubstitutes or associates? Evidence against slot-filleraccounts of children’s early categorization. ChildDevelopment, 69, 347–354.

La Heij, W., Dirkx, J., & Kramer, P. (1990).Categorical interference and associative priming inpicture naming. British Journal of Psychology, 81,511–525.

Landau, B., Smith, L. B., & Jones, S. S. (1988). Theimportance of shape in early lexical learning.Cognitive Development, 3, 299–321.

Lautrey, J. (in press). A pluralistic approach tocognitive differentiation and development. In R. J.Sternberg, J. Lautrey, & T. Lubart (Eds.), Models ofintelligence: International perspectives. Washington,DC: APA Press.

Lin, E. L., & Murphy, G. L. (2001). Thematic relationsin adults’ concepts. Journal of ExperimentalPsychology: General, 130, 3–28.

Mareschal, D., & French, R. (2000). Mechanisms ofcategorization in infancy. Infancy, 1, 59–76.

Markman, E. M., & Hutchinson, J. E. (1984).Children’s sensitivity to constraints on wordmeaning: Taxonomic versus thematic relations.Cognitive Psychology, 16, 1–27.

Mathews, R. C., Maples, R. C., & Elkins, R. (1981).Semantic judgments as encoding operations in recall:The encoding of task-relevant and task-irrelevantsemantic attributes of words. The Journal of GeneralPsychology, 105, 311–320.

McCauley, C., Weil, C. M., & Sperber, R. D. (1976).The development of memory structure as reflectedby semantic-priming effects. Journal of ExperimentalChild Psychology, 22, 511–518.

Nation, K., & Snowling, M. J. (1999). Developmentaldifferences in sensitivity to semantic relations amonggood and poor comprehenders: Evidence fromsemantic priming. Cognition, 70, B1–B13.

Nelson, K. (1983). The derivation of concepts andcategories from event representations. In E. K.Scholnick (Ed.), New trends in conceptualrepresentation: Challenges to Piaget’s theory?(pp. 129–149), Hillsdale, NJ: Lawrence ErlbaumAssociates Inc.

Nelson, K. (1986). Event knowledge: Structure andfunction in development. Hillsdale, NJ: LawrenceErlbaum Associates Inc.

Osborne, J. G., & Calhoun, D. O. (1998). Themes,taxons, and trial types in children’s matching tosample: Methodological considerations. Journal ofExperimental Child Psychology, 68, 35–50.

Quinn, P. C., Eimas, P. D., & Rosenkrantz, S. L.(1993). Evidence for representations of perceptuallysimilar natural categories by 3-month-old and 4-month-old infants. Perception, 22, 463–475.

Quinn, P. C., French, R. M., & Mareschal, D. (2000).A connectionist account of asymmetric categorylearning in early infancy. Developmental Psychology,36, 635–645.

Rosch, E., Mervis, C. B., Gray, W. D., Johnson, D. M.,& Boyes-Braem, P. (1976). Basic objects in naturalcategories. Cognitive Psychology, 8, 382–439.

Ross, B. H., & Murphy, G. L. (1999). Food forthought: Cross-classification and categoryorganization in a complex real-world domain.Cognitive Psychology, 38, 495–553.

Scott, M. S., Serchuk, R., & Mundy, P. (1982).Taxonomic and complementary picture pairs:Ability in two- to five-year-olds. InternationalJournal of Behavioral Development, 5, 243–256.

Smiley, S. S., & Brown, A. L. (1979). Conceptualpreference for thematic or taxonomic relations: Anonmonotonic age trend from preschool to old age.Journal of Experimental Child Psychology, 28,249–257.

Vallee-Tourangeau, F., Anthony, S. H., & Austin,N. G. (1998). Strategies for generating multipleinstances of common and ad hoc categories.Memory, 6, 555–592.

Walsh, M., Richardson, K., & Faulkner, D. (1993).Perceptual, thematic, and taxonomic relations inchildren’s mental representations: Responses totriads.. European Journal of Psychology ofEducation, 8, 85–102.

Waxman, S. R., & Namy, L. L. (1997). Challenging thenotion of a thematic preference in young children.Developmental Psychology, 33, 555–567.

Wisniewski, E. J., & Bassok, M. (1999). What makes aman similar to a tie? Stimulus compatibility withcomparison and integration. Cognitive Psychology,39, 208–238.

Int Jnt Psychology (gamma) IJP35365.3d 29/1/04 17:09:09 Rev 7.51n/W (Jan 20 2003)

The Charlesworth Group, Huddersfield 01484 517077

14 SCHEUNER ET AL.

Int Jnt Psychology (gamma) IJP35365.3d 29/1/04 17:09:09 Rev 7.51n/W (Jan 20 2003)

The Charlesworth Group, Huddersfield 01484 517077

APPENDIX A

Stimuli used in Experiments 1, 2, and 3 (targets 1 and 2 served in familiarization trials)

Targets Basic-level (T1) Slot-filler (T2) Superordi-nate (T3) Thematic (Th1) Thematic (Th2) Thematic (Th3)

Saucepan Frying pan Colander Glass Stove Milk Matchbox

Rabbit 1 Rabbit 2 Bear Scorpion Carrot Rabbit hutch Tractor

Dog 1 Dog 2 Guinea pig Snake Bone Kennel Shotgun

Cake Pie Chocolate Sausage Candle Present Flour

Seagull Eagle Ostrich Dolphin Egg Nesting box Scarecrow

Ball Rugby ball Roller-skates Card Soccer player Basketball board Whistle

Bee Fly Butterfly Cat Beehive Flower 2 Sting

Lamp Chandelier Torch Headlight Light bulb Plug Desk

Flower 1 Rose Houseplant Tree Vase Snail Rake

Suit jacket Casual jacket Tie Swimming costume Coat hanger Suitcase Reel of thread

Axe Chain saw Saw Pliers Wood Play house Carpenter

MATCHING BY ASSOCIATIVE STRENGTH IN CHILDREN 15