Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/38021158

Categoricalflexibilityinpreschoolers:Contributionsofconceptualknowledgeandexecutivecontrol

ArticleinDevelopmentalScience·November2009

DOI:10.1111/j.1467-7687.2009.00832.x·Source:PubMed

CITATIONS

17

READS

41

2authors:

AgnesBlaye

Aix-MarseilleUniversité

65PUBLICATIONS1,691CITATIONS

SEEPROFILE

SophieJacques

DalhousieUniversity

22PUBLICATIONS627CITATIONS

SEEPROFILE

AllcontentfollowingthispagewasuploadedbySophieJacqueson14November2014.

Theuserhasrequestedenhancementofthedownloadedfile.Allin-textreferencesunderlinedinbluearelinkedtopublicationsonResearchGate,lettingyouaccessandreadthemimmediately.

PAPER

Categorical flexibility in preschoolers: contributions ofconceptual knowledge and executive control

Agnes Blaye1 and Sophie Jacques2

1. Laboratoire de Psychologie Cognitive, Universit! de Provence, France2. Department of Psychology, Dalhousie University, Canada

Abstract

The current study evaluated the relative roles of conceptual knowledge and executive control on the development of categoricalflexibility, the ability to switch between simultaneously available but conflicting categorical representations of an object. Exper-iment 1 assessed conceptual knowledge and executive control together; Experiment 2 differentiated conceptual knowledge fromcostly executive processes. In Experiment 1, 3- to 5-year-olds were given a three-choice (taxonomic, thematic, and nonassociate)match-to-sample task and asked to match two associates. In Experiment 2, same-aged children were assessed on another match-to-sample task that reduced executive costs by presenting thematic and taxonomic associates on separate trials. By comparingperformance across tasks, age-related changes resulting from conceptual knowledge and executive control indicated that con-ceptual knowledge of superordinate relations showed gains between 3 and 4 years, whereas gains in executive control were seenbetween 4 and 5 years, suggesting a d!calage in the development of conceptual and executive processes underlying categoricalflexibility.

Introduction

Objects can be considered as members of multiple cate-gories depending on the current goal at hand: Hence, adog can be considered as an animal (taxonomic category)or as part of a hunting scene (thematic category), amongmany other categories. The ability to switch betweensimultaneously available, but conflicting categoricalrepresentations of a given object – namely categoricalflexibility – is therefore a critical conceptual achievement.However, the question of when and how categoricalflexibility emerges in development has long been neglectedbecause early theories of conceptual development ini-tially held that young children’s conceptual systems werebased primarily on complementary relations (Inhelder &Piaget, 1959/1964; Nelson, 1977) binding objects thatco-occur in time and or space. In other words, thematiccategories were supposed to emerge first in development.Taxonomic1 categories were believed to emerge onlylater, resulting from a major conceptual reorganizationinstigated by developments in logic (Inhelder & Piaget)or language (Nelson). As a result, given the supposedlimits in conceptual understanding, the question regarding

flexible representation of different conceptual relationsin young children seemed irrelevant.Whereas early research by Inhelder and Piaget

(1959/1964) and others primarily used free-sortingtasks to assess categorical organization, contemporaryresearch on preschool children’s categorization abilitieshave used match-to-sample tasks in which children arepresented with series of pictured objects (or toy replicas)and asked to select the best match for a target (e.g. ananimal [dog]) among competing associates (e.g. a taxo-nomic match [a snail] and a thematic associate [dog-house]). This approach has revealed that preschoolers canacknowledge both thematic and taxonomic pairings(e.g. Blaye & Bonthoux, 2001; Markman & Hutchin-son, 1984; Walsh, Richardson & Faulkner, 1993; Wax-man & Namy, 1997; see Murphy, 2002, for a review).Different contextual variables, however, such as instruc-tional format (e.g. Waxman & Namy) or presence orabsence of labels (e.g. Markman & Hutchinson), mayorient children to select particular types of associates(thematic vs. taxonomic), suggesting that they are sensi-tive to both kinds of organizational networks but thatthey can be biased to favorone over the other.In fact, to assess directly whether children understand

taxonomic categories, a few studies have presentedchildren with match-to-sample tasks that involved only ataxonomic option along with a nonassociate (or severalnonassociates) with no competing thematic associates(e.g. Fenson, Vella & Kennedy, 1989; Nguyen & Murphy,

Address for correspondence: Agn!s Blaye, Universit" de Provence, Laboratoire de Psychologie Cognitive, 29, Av R. Schuman, 13621 Aix enProvence, cedex 1, France; e-mail: Agnes.Blaye@univ-provence.fr

1 Note that when we refer to taxonomic categories, we consider only

superordinate taxonomic categories (Rosch, Mervis, Gray, Johnson &

Boyes-Braem, 1976). Basic-level taxonomic categories are too ambi-

guous to interpret, as these can be identified as a result of perceptual (i.e.

high level of perceptual similarity between two exemplars; e.g. dogs) or

conceptual processing.

! 2009 The Authors. Journal compilation ! 2009 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and350 Main Street, Malden, MA 02148, USA.

Developmental Science 12:6 (2009), pp 863–873 DOI: 10.1111/j.1467-7687.2009.00832.x

2003, Experiment 2; Rosch, Mervis, Gray, Johnson &Boyes-Braem, 1976, Experiment 8). Using this approach,researchers have found evidence of conceptual knowledgeof (superordinate) taxonomic categories in preschoolers,at least with 4-year-olds. For example, Rosch et al. foundthat when presenting a target with a superordinatetaxonomic match and a nonassociate, 96% of 4-year-oldscould identify superordinate taxonomic matches correctly,although only 55% of 3-year-olds could do so.Nonetheless, although the aforementioned studies

suggest that under some conditions, preschoolers canidentify superordinate taxonomic matches, other studieshave revealed that the majority of 4-year-olds are unableto justify such matches (e.g. Greenfield & Scott, 1986;Lucariello, Kyratzis & Nelson, 1992; Tversky, 1985),indicating that conceptual knowledge of taxonomiccategories might still undergo important developmentsduring the preschool years or that verbal requirements injustification studies may be especially demanding. How-ever, in most justification studies, children were requiredto activate successively two representations of the sametarget objects: One that they spontaneously chose first(usually the thematic option) and then the second (thetaxonomic option), which they did not favor, but wereaskedto justify.Moreover, their justificationswererequiredwhile the favored and more salient (thematic) option wasstill present. Consequently, existing justification studiesmight haveunderestimated conceptual knowledgebecausethey inadvertently also required categorical flexibility.In the current study we investigated the development

of categorical flexibility directly by examining whetherpreschool children can switch successively betweenrepresenting objects as part of schemas (henceforth,members of thematic categories) or as members ofsuperordinate taxonomic categories (or visa versa) withinthe same context. To achieve categorical flexibility, onemust have developed sufficient conceptual knowledge ofthematic and taxonomic representations. As mentionedabove, it is possible that in existing justification studies,children were unable to justify taxonomic categoriesbecause they did not have sufficient conceptualknowledge. However, a lack of categorical flexibilitydoes not necessarily imply deficient conceptualknowledge because switching between competingrepresentations as is required in categorical flexibilityalso requires sufficient executive control, and more specifi-cally cognitive flexibility, both of which are known todevelop drastically during the preschool years.Executive control (Miller & Cohen, 2001) is an ill-

defined construct ‘that refers generally to the psycho-logical processes involved in the conscious control ofthought and action’ (Zelazo & M#ller, 2002, p. 445).Executive processes are hence specifically involved whentop-down control is required instead of stimuli-drivenprocessing. Important components of executive controlthat have been proposed include cognitive flexibility,inhibition, and working memory (Diamond, 2006;Garon, Bryson & Smith, 2008; Hughes, 2002). Cognitive

flexibility can be defined as ‘the ability to flexibly switchperspectives, focus of attention, or response mappings’(Diamond, 2006, p. 70); inhibition refers to the abilityto exert inhibitory control on dominant prepotentresponses (Carlson & Moses, 2001); and workingmemory refers to ‘a system for the temporary holdingand manipulation of information during the performanceof a range of cognitive tasks’ (Baddeley, 1986, pp. 33–34).Thus, because categorical flexibility requires that childrenbe able to switch flexibly between different categoricalrepresentations of the same object, it is one manifestationof cognitive flexibility when representations arecategorical in nature.Like conceptual knowledge, there are significant age-

related changes in executive control during the preschoolyears and more specifically, a drastic increase in cognitiveflexibility (e.g. Carlson &Moses, 2001; Chevalier & Blaye,2008; Frye, Zelazo & Palfai, 1995; Gerstadt, Hong &Diamond, 1994; Jacques & Zelazo, 2001; Luria, 1961;M#ller, Zelazo, Hood, Leone & Rohrer, 2004; Smidts,Jacobs&Anderson, 2004;Zelazo, 2006; see alsoDiamond,2006, and Jacques & Zelazo, 2005, for recent reviews). Acommonlyusedtasktoassesscognitive flexibility in thisagerange is the Dimensional Change Card Sort (DCCS)designed by Zelazo and Frye (Frye et al., 1995; Zelazo,2006; Zelazo, M#ller, Frye & Marcovitch, 2003). In thistask, children are presented with cards that vary on twodimensions (color and shape) that they have to sort firstaccording to one dimension (for instance shape) for agivennumber of trials and then by a second dimension (forinstance color) requiring opposite responses to be appliedto the same stimuli. Many studies have revealed thatwhereas3-year-olds can sort correctlyaccording to the firstdimension, most of them fail to switch to the otherdimension. In clear contrast, most 4-year-olds experienceno switching difficulties.The DCCS is a deductive task (Jacques & Zelazo, 2005)

in that children are explicitly told which matching toperform for a given picture. Other tasks provide lessexplicit instructions (see De$k’s Flexible Induction ofMeaning task, for instance in De$k, 2003) and others stillare clearly inductive in that children must identify forthemselves more than one match for each objectsuccessively. The Flexible Item Selection Task (FIST)designed by Jacques and Zelazo (2001) is one such task.On each trial, participants are shown three items that varyon two of three dimensions (shape, color, and size) and apivot item matches one of the other items on onedimension and the remaining item on the otherdimension. Participants are then asked to find two pairsof matching items, and as a result, they have to select thepivot item twice. Jacques and Zelazo found that 3-year-olds have relatively more difficulty than 4- and 5-year-oldson their first selection, suggesting that they have difficultydetecting a similarity between items when these arenonidentical. In contrast, both 4- and 5-year-olds dowell on their first selection. However, 4-year-olds haverelatively more difficulty than 5-year-olds on their second

864 Agn!s Blaye and Sophie Jacques

! 2009 The Authors. Journal compilation ! 2009 Blackwell Publishing Ltd.

selection, suggesting that they have difficulty selectingthe pivot item flexibly (see Jacques, Zelazo & Lourenco,2008, for additional findings with this task). Thus, boththe DCCS and the FIST, as well as many other similartasks, reveal important changes between 3 and 5 incognitive flexibility, although successful performance oninductive tasks like the FIST is generally delayedcompared to the DCCS and other deductive tasks (seeJacques & Zelazo, 2005, for a review).In recent years, research has begun on flexible categor-

ical representations in both infants (e.g. Ellis & Oakes,2006; Mareschal & Tan, 2007) and preschool children(e.g. Blaye & Bonthoux, 2001; De$k, 2003; Nguyen &Murphy, 2003; Waxman & Namy, 1997). However, mostof these studies have not assessed categorical flexibilityper se because they compared (a) categorical choices ofindependent groups of children placed in different settingconditions, (b) the performance of the same children ondifferent sets of objects, or (c) the performance of thesame children on the same set of objects, but onlyafter a delay. Although suggestive, findings using thesemethodologies do not inform us about whether childrenare able to alternate successively between differentcategorical representations of the same objects at any givenmoment.Hence, the current research had two broad aims. First,

in Experiment 1, we developed a categorization task thatallowed us to assess categorical flexibility directly inconditions of conflict between two potentially correctresponse options. One could argue that both the DCCSand the FIST can be construed as measures of categoricalflexibility per se; however, they only assess percep-tual categories (e.g. colors, sizes, or shapes) and as such,they do not assess children’s ability to switch betweendifferent semantic representations. Therefore, we used athree-choice match-to-sample task (hereafter, the DoubleCategorization Task) that included a superordinatetaxonomic associate, a thematic associate, and anonassociate. To assess categorical flexibility, we askedchildren to identify two matches for the target. The firstselection was assumed to correspond to participants’mostsalient match in the context of another potential match,whereas the second selection, presumably a less salientmatch, had to be identified while the first salient selectionwas still present, thereby requiring categorical flexibility.Second, in Experiment 2, we used another match-to-

sample task that allowed us to measure conceptualknowledge independent of executive control involved inthe Double Categorization Task. In this SimpleCategorization Task, we reduced executive costs bypresenting children with thematic and taxonomicassociates on separate trials (cf. Nguyen & Murphy,2003, Experiments 1 and 2). On thematic trials, childrenwere shown targets with a thematic associate and twononassociates, whereas on taxonomic trials, they were shownthesametargetswitha superordinate taxonomicassociateandtwo nonassociates. Consequently, by comparing per-formance on the Double and Simple Categorization

Tasks, we were able to determine whether age-relatedchanges in categorical flexibility measured in theDouble Categorization Task result from increases inconceptualization abilities, efficient executive control, orboth.In short, the purpose of the current experiments was

to evaluate the relative role of both conceptual know-ledge and executive control involved in the developmentof categorical flexibility. On their own, adequateconceptual knowledge and executive control skills areboth necessary but not sufficient conditions forcategorical flexibility. For categorical flexibility toemerge, both need to be operational. Experiment 1assessed conceptual knowledge and executive controltogether, whereas Experiment 2 differentiated conceptualknowledge from costly executive aspects.

Experiment 1

In this experiment, we examined the performance of 3-,4-, and 5-year-old children in a match-to-sample taskthat required children to match a target successively withboth thematic and taxonomic associates. This was doneto determine whether children could identify twomatches for a given target, providing evidence of cate-gorical flexibility. In addition, it allowed us to determinewhether age-related differences in categorical flexibilityemerge at the same time during the preschool period asthey do in research on cognitive flexibility in the execu-tive function literature (e.g. De$k, 2003; Jacques &Zelazo, 2001; Zelazo et al., 2003).

Method

Participants

Participants were 20 3-year-olds (M ! 44 mos, range 38to 47 mos), 20 4-year-olds (M ! 54 mos, range 48 to 59mos), and 20 5-year-olds (M ! 66 mos, range 60 to 71mos). Approximately equal numbers of boys and girlsparticipated at each age. Children were recruited frompreschools in a middle-sized city in the south of Franceand were predominantly from middle-class backgrounds.Parents were provided with a short written descriptionof the experiment and they gave informed consent allow-ing their children to participate.

Stimuli

Black and white line drawings of familiar objects wereused. Each trial was presented on an A4 sheet displayedhorizontally with a target (e.g. a dog) positioned at thetop and centered, and three potential associates posi-tioned together on the same line below the target (seeFigure 1, Panel A). Among these three associates, onewas a thematic associate (e.g. a doghouse), another wasa superordinate taxonomic associate (e.g. another animal,

Conception and flexibility 865

! 2009 The Authors. Journal compilation ! 2009 Blackwell Publishing Ltd.

a snail), and the remaining one was semantically unre-lated to the target (e.g. a telephone). Sixteen item setswere used in test trials including eight artifacts and eightnatural objects as targets. Two additional item sets wereused in demonstration trials. The placement (left, middle,or right) of the three types of associates (thematic,taxonomic, or nonassociate) was counterbalanced acrossthe 18 item sets.Eleven adults aged between 23 and 48 years were asked

to participate in a judgment task to ensure that theyagreed with our pairings of the targets and the threepotential types of associates (taxonomic, thematic, orunrelated). Participants were initially presented withthree example trials in which they were shown threetypes of pairs of pictures (i.e. one thematic, onetaxonomic, and one nonassociate pair of pictures) inorder to define the three categories in which they wereexpected to classify test pairs. The three categories weredefined as (a) ‘the two objects belong to the samecategory, they are the same sort of things’, (b) ‘the twoobjects can often be seen together in a common scene orevent’, and (c) ‘the two objects have no links whatsoever’.Adults were then shown 64 test pairs of picturesinvolving each of the 16 targets presented four times,once paired with its thematic associate, once paired withits taxonomic associate, and once paired with each oftwo nonassociates.2 The 64 pairs were presented in aconstant random order on computer using PowerPointSlideShow mode and participants saw each pair successi-vely in a self-paced manner. They had to decide towhich of the three types of associations (thematic, taxo-nomic, or unrelated) each pair of pictures belonged.Across all pairings, the average degree of agreement

between participants’ judgements and our own a prioricategorization of pairs was 98.5% (range 95.3% [oneparticipant] to 100% [four participants]). For pairings ofnonassociates, there was 100% agreement on all pairsbut one, which produced only one disagreement. For thetaxonomic pairings, 15 pairs produced 100% agreement

and one pair produced one disagreement. Similarly, forthe thematic pairings, 11 pairs produced 100% agreementand four produced only one disagreement. The remain-ing thematic pairing, however, did produce somedisagreement with only 45.5% of the adult sampleagreeing with our pairing. This particular thematicpairing involved a screwdriver (the target) and a child’sbicycle (the thematic match). Adults who disagreedconsidered this pair to be unrelated. Despite the lowagreement from our adult participants, it is worthconsidering that children’s performance in the SimpleCategorizationTask inExperiment 2on this itemwashigh:44outof 53 children (83%) selected thebicycle fromamongthe other two nonassociates. This discrepancy betweenadults’ and children’s judgments may be due to recentpersonal experience: Many young children may have hadtheir bicycle repaired recently (e.g. many children withinthis age range have had the training wheels removed fromtheir bicycle), whereas many of the urban adults in oursamplemaynothaveevenownedabicycle.Overall,then,thedata overwhelmingly confirmed our a priori associationsfor the different pairs (see list of stimuli in the Appendix).

Procedure

Children were tested individually in their preschool. Thetask began with two demonstration trials. In each dem-onstration (and test) trial, children were asked to selecttwo matches for each target. For Selection 1, childrenwere told (in French), ‘Look at this one [Experimenterpointed to the target]. Can you show me, among thesethree, the one that goes best with this one? To show me,put this token beneath the one you choose.’ For Selec-tion 2, they were told, ‘OK, now there are only two left.Can you show me the one that goes best with this one[Experimenter pointed to the target] among these two?[Experimenter pointed to the two remaining items] Hereis another token to indicate your choice.’ If childrenselected the nonassociate for either selection in the dem-onstration trials, they received corrective feedback inthat the token was moved to the correct associate. Orderof the two demonstration trials was counterbalancedacross participants. After the two demonstration trials,

Figure 1 Stimuli display presented in the Double Categorization Task (Experiment 1, Panel A) and the Simple Categorization Task(Experiment 2, Panels B and C).

2 As we used four nonassociates for a given target in the SimpleCategorization Task in Experiment 2, half of the adults saw two non-associates and the other half saw the remaining two.

866 Agn!s Blaye and Sophie Jacques

! 2009 The Authors. Journal compilation ! 2009 Blackwell Publishing Ltd.

16 test trials were presented in which a shortened versionof the instructions was used with no corrective feedback.Two random presentation orders for test trials weredevised and counterbalanced across participants.

Results and discussion

For each target, there were three possible responseoptions (thematic, taxonomic, and nonassociates), onlytwo of which were correct (thematic and taxonomicones). On each trial, each of the two selections wasscored as correct only if children selected the thematic ortaxonomic choice. Although categorical flexibility isdemonstrated only in children’s ability to select bothassociates (thematic and taxonomic) for a given targetcorrectly, we first considered Selection 1 performancealone to ensure that children understood the task and toexamine patterns of performance on Selection 1 thatmight influence categorical flexibility.

Selection 1 performance

Although the type of target objects (artifacts vs. naturalobjects) was controlled across trials, we first analyzedwhether there were any potential effects of this variableon performance. No main effect or interaction with ageor type of associate was found in this experiment orin Experiment 2. Therefore, this variable was not includedin any further analyses. A 3 (Age Group: 3-, 4-, vs. 5-year-olds) · 2 (Associate Type: taxonomic vs. thematicassociates) mixed analysis of variance (ANOVA) withrepeated measures on associate type was then conductedon the proportion of correct choices on Selection 1. TheANOVA detected no reliable effect of associate type, noran interaction between associate type and age group,although descriptively the proportion of thematic firstselections (M ! .48, SD ! .22) was higher thantaxonomic ones (M ! .40, SD ! .19). However, asignificant main effect of age group was detected, F(2, 57)! 8.80, p< .0005, partialg2 ! .24, revealing that 5-year-olds (M ! .97, SD ! .07) differed significantly fromboth other age groups (M ! .80, SD ! .15, for 3-year-olds;M ! .87, SD ! .14, for 4-year-olds), Tukey’s HSDtests, ps < .005, who did not differ from each other. Theproportion of correct trials expected on the basis of chancealone was .67 (two out of three response options), andaccording to this criterion, all age groups performed abovechance on Selection 1, all ts(19) > 4.00, p < .0005.Together, these results suggest that children have suffi-cient conceptual knowledge and control capacities toselect at least one relevant match for a target by 3 yearsof age.The design of the match-to-sample task allows one to

examine if children approach the task on a trial-by-trialbasis providing evidence for a bottom-up, stimulus-based approach or if they approach the task using acontrolled approach and systematically select the samekind of associate across trials providing evidence for a

top-downapproach.Indeed,Scheuner,Bonthoux,Cannardand Blaye (2004) found evidence that children’s selectionsare partly determined by the relative strength of parti-cular associations among potential associates of a giventarget. Such a stimulus-driven approach can result inintra-individual variability in categorical matching acrosstrials (Walsh et al., 1993). However, Waxman and Namy(1997) observed that intra-individual variability tends todecrease between 3 and 4 with older children showingmore consistent patterns of selections across trials. Onour account, an interesting aspect of executive controlmay be revealed by considering the consistency ofSelection 1 responses across different targets.Specifically, our contention is that to approach the taskusing top-down processes, despite uncontrolledvariability in the relative strength of associations ofparticular matching candidates, provides evidence ofexecutive control (cf. Kendler, 1995), which might in turnbe a precursor to categorical flexibility itself. To test thisprediction, we defined a consistent pattern of Selection 1responses as one in which children selected twice as manycorrectmatches of one type or another, whether these werethematic or taxonomic matches. According to thisdefinition, we found that 40% of 3-year-olds, 60% of 4-year-olds, and 70% of 5-year-olds produced a consistentselection pattern for Selection 1. Although there was anincrease in systematicity with age as reported byWaxmanand Namy, only the difference between 3- and 5-year-oldswas significant, v2(1, N ! 40) ! 3.63, p < .057.Alternatively, if we considered the exact dominance ratiobetween the frequencies of the two kinds of responses(ABS [(THfreq ) TXfreq)/(THfreq + TXfreq)]) onSelection 1, a one-way ANOVA also revealed asignificant effect of age, F(2, 57) ! 4.67, p < .02, partialg2 ! .14. Post-hoc comparisons indicated that this effectwas again only significant when contrasting the youngestand oldest age groups, p < .01.

Performance across both selections

Selection 2 scores were not examined independentlybecause the probability of correct responses on Selection2 was entirely dependent on performance on Selection 1.Instead, we considered flexibility scores measured as theproportion of correct double selections; that is, the pro-portion of trials for which children made two correctselections. A one-way ANOVA with age group (3, 4, vs.5 years) as a between-subjects variable revealed a signi-ficant increase in this score with age, F(2, 57) ! 13.81,p < .0001, partial g2 ! .33. As shown in Figure 2 (leftpanel), the increase was reliable between 3- and 5-year-oldsand 4- and 5-year-olds, ps < .002, but not between 3- and4-year-olds, p < .33. The proportion of correct doubleselections expected on the basis of chance alone was .33(p ! .67 for Selection 1 and p ! .50 for Selection 2, thus,p ! .33 for both). According to this criterion, all agegroups performed above chance on both selections,ts(19) > 3.39, p < .003.

Conception and flexibility 867

! 2009 The Authors. Journal compilation ! 2009 Blackwell Publishing Ltd.

We then addressed the question of whether thosechildren who showed a consistent selection pattern forSelection 1, suggesting a systematic matching strategy,also tended to be more flexible. A t-test contrastingchildren who demonstrated a consistent response pattern(either thematic or taxonomic) for Selection 1 with thosewho did not revealed that flexibility scores were higheramong the former group (M ! .73, SD ! .24), t(58) !2.46, p < .02, than among the latter one (M ! .57, SD !.26). However, as the strategic use of categorical relationsincreasedwith age, its influence on flexibility scores mighthave been at least partly confounded with age.Consequently, we also calculated a correlation betweenthe exact dominance ratio of Selection 1 responses andflexibility scores, partialing out age. This correlation wassignificant, r(57) ! .33, p< .011, indicating that controlledresponding on Selection 1 may be predictive of theability to switch between two representations of thesame object.Overall, then,Experiment1revealed importantdevelop-

ments in categorical flexibility during the preschoolperiod, primarily between 4 and 5 years. However, assuggested in the Introduction, success on the DoubleCategorizationTask depends on both access to conceptualknowledge of thematic and taxonomic associates of eachtarget and efficient executive control, both of which areknown to develop during the preschool years.Consequently, it is impossible to determine whether it isinsufficient development in one or the other that bestaccounts for the observed age-related changes incategorical flexibility. Therefore, Experiment 2 was designedto assess the development of children’s conceptualknowledge in a context that minimized the need forexecutive control.

Experiment 2

The aim of Experiment 2 was to isolate conceptualknowledge from executive control costs involved in theDouble Categorization Task used in Experiment 1. TheDouble Categorization Task requires selecting betweentwo conflicting options for Selection 1 and immediatelyswitching to a new representation of the target in thepresence of the first more salient match. Hence, inExperiment 2 we examined the performance of 3- to5-year-olds on another version of the match-to-sampletask – the Simple Categorization Task – in which childrenwere presented twice with the same set of targets. Like theDouble Categorization Task, children also were asked tomatch targets to thematic and taxonomic associates, butthematic and taxonomic options for each particulartarget were presented on different trials (see Nguyen &Murphy, 2003, for a similar procedure). On thematictrials, targets were presentedwith three potential matches,a thematic associate and two nonassociates, whereas ontaxonomic trials, the same targets were presented with asuperordinate taxonomic associate and twononassociates. For half of the targets, thematic trialswere presented first, and for the other half, taxonomictrials were presented first. Consequently, Selection 2responses were asked only after a series of other targetshad been presented. In addition, the relevant associate tobe identified for Selection 2 was presented in thecontext of two new nonassociates.This Simple Categorization Task, then, presented

three main differences with the Double CategorizationTask. First, children made each of their selections in anon-conflictual context because only one associate wassemantically related to the target on a given trial (the

Figure 2 Mean proportions (and standard errors) of correct double selections as a function of age for the Double CategorizationTask in Experiment 1 (left panel) and the Simple Categorization Task in Experiment 2 (right panel).

868 Agn!s Blaye and Sophie Jacques

! 2009 The Authors. Journal compilation ! 2009 Blackwell Publishing Ltd.

other two being unrelated). Second, the second potentialmatch for a given target was requested only after a delay.Third, the second potential match was requested in theabsence of the first match. We hypothesized that thesedifferences, all of which contributed to reducing the executiveload of the task, might help reveal conceptual knowledgethat might otherwise be underestimated on the DoubleCategorization Task and traditional match-to-sampletasks more generally. Moreover, a direct comparisonwith performance on the Double Categorization Task inExperiment 1 allowed us to examine the relative contri-butions of conceptual knowledge and executive controlto categorical flexibility. In addition, this comparisonallowed us to trace differences in developmental changesacross the two tasks, highlighting potentially differentcritical periods for developments in conceptual and execu-tive processes.

Method

Participants

Participants were 18 3-year-olds (M ! 43 mos, range 40to 47 mos), 18 4-year-olds (M ! 55 mos, range 51 to 58mos), and 16 5-year-olds (M ! 66 mos, range 60 to 71mos). Approximately equal numbers of boys and girlsparticipated at each age. Children were recruited in thesame manner as those who participated in Experiment 1,but no child who participated in Experiment 1 parti-cipated in this experiment.

Stimuli

Targets and thematic and taxonomic associates were thesame as those used in Experiment 1.3 However, becausethematic and taxonomic associates were presented ondifferent trials, two nonassociates needed to be presentedon both thematic and taxonomic trials. Thus, three newnonassociates were used with each target in addition tothe one used in Experiment 1. The particular pair ofnonassociates presented on thematic and taxonomictrials was counterbalanced between participants.

Procedure

The task began with two demonstration trials consistingof a taxonomic trial in which a target was presented witha taxonomic associate and two nonassociates (see Figure 1,Panel B) and a thematic trial in which a target waspresented with a thematic associate and two nonassoci-ates (see Figure 1, Panel C). Instructions were similar tothose used in Experiment 1 except that no tokens wereused: Children simply pointed to the best match for eachtarget. Corrective feedback was given on the twodemonstration trials. Participants were then presented

with the two experimental blocks of 16 trials each, cor-responding to the 16 targets used in Experiment 1, andeach target appeared only once within each block.Moreover, within each block, approximately half of thetargets were presented in thematic trials and the otherhalf were presented in taxonomic trials.4 At the outset ofthe second block, children were informed that they weregoing to see the same targets as in the first block, but thatnew pictures were to be displayed with them. They weretold that they again had to select the one that bestmatched the target. Two random orders of test trials weredevised and counterbalanced across participants. Foreach order, however, targets were presented in thesame order across the two blocks to keep the number ofinterspersed trials between first and second selectionsconstant for all targets.

Results and discussion

On a given trial children had to choose among threeresponse options: one correct (taxonomic or thematic)associate and two incorrect nonassociates. In this con-text, then, Selection 1 responses were those responsesthat children made in the first experimental block and acorrect double selection involved selecting the two asso-ciated pictures presented in the first and second trial of agiven target, one in the first block and the other in thesecond block.

Selection 1 performance

Children’s performance was first examined on the pro-portion of correct Selection 1 responses using a 3 (AgeGroup: 3-, 4-, vs. 5-year-olds) · 2 (Associate Type:thematic vs. taxonomic) mixed ANOVA with repeatedmeasures on associate type. It revealed a significant maineffect of age,F(2, 49) ! 18.31, p< .0001, partialg2 ! .43,with the mean proportion of correct Selection 1responses increasing from .73 (SD ! .14) to .88 (SD !.08) to .94 (SD ! .08) in 3-, 4-, and 5-year-olds, respec-tively. Tukey’s HSD analyses indicated that the increasein performance occurred primarily between 3 and 4years of age: Only the youngest group’s performancediffered significantly from that of the other two groups,ps< .001.However,withaprobabilityofp ! .33ofmakinga correct selection on Selection 1 on the basis of chancealone, even 3-years-olds performed above chance level,ts(15 or 17) > 12.3, ps < .0001. There was also asignificant main effect of associate type with childrenperforming better on thematic selections, F(1, 49) !

3 An additional two item sets were presented in this experiment, but

these were not considered here for the sake of comparison with

Experiment 1.

4 Because we considered only the 16 items common to both experi-

ments, the balance between the two types of associates presented first

was not perfect. That is, some children received seven thematic trials as

their first presentation and others received nine. However, in each

age group, there were as many children with more thematic trials first

as children with more taxonomic trials first. In addition, we compared

proportions of correct responses instead of frequencies.

Conception and flexibility 869

! 2009 The Authors. Journal compilation ! 2009 Blackwell Publishing Ltd.

7.23, p < .01, partial g2 ! .13 (M ! .88, SD ! .14, forthematic selections and M ! .81, SD ! .17, for taxo-nomic selections), although performance for both types(taxonomic and thematic) was high. There was nointeraction between age and type of associate, p > .40.

Performance across both selections

Turning to correct double selections, we used a 3 (AgeGroup) · 2 (Presentation Order: thematic then taxo-nomic vs. taxonomic then thematic) ANOVA on theproportions of correct double selections. It revealed noeffect of presentation order or interaction, ps > .50, but asignificant main effect of age, F(2, 49) ! 26.85, p < .0001,partial g2 ! .52. As suggested in Figure 2 (right panel),therewas a significant improvement between 3 and 4 years(M ! .52,SD ! .18, andM ! .78,SD ! .09, for 3- and4-year-olds, respectively), but no significant changesbetween 4 and 5 years (M ! .86 and SD ! .14, for 5-year-olds). Tukey’s post-hoc tests confirmed thatdifferences were reliable only between the youngestgroup and the two older ones, ps < .0001, who did notdiffer from one another, p > .32. However, it is noteworthythat with a mean proportion of .52 correct doubleselections, 3-year-olds’ performance was far abovechance given that the proportion of trials expected to becorrect by chance alone was only .11 (p ! .33 for Selec-tion 1 and p ! .33 for Selection 2, thus, p ! .11 for both).

Comparisons across the Double and SimpleCategorization Tasks

Scores represented in Figure 2 contrasting the propor-tions of correct double selections in the Double andSimple Categorization Tasks, respectively, are notdirectly comparable because the base probability ofselecting a correct double selection for a given targetdiffered across tasks. Consequently, we examined thepercentages of children in each age group who producedflexibility scores (i.e. correct double selections for eachtarget) superior to chance as a function of task usingbinomial probability distributions (i.e. x ‡ 9 for theDouble Categorization Task, binomial ! .05, and x ‡ 5for the Simple Categorization Task, binomial ! .03).Whereas a large majority of children produced flexibilityscores superior to chance from 3 years onwards on theSimple Categorization Task, less than half of 3-year-oldsand approximately half of 4-year-olds did so on theDouble Categorization Task (Figure 3). One had to waituntil age 5 to observe comparable percentages on bothtasks. Indeed, there was a significant effect of age groupfor the Double Categorization Task, v2(2, N ! 60) !11.14, p < .004, but no effect of age group for the SimpleCategorization Task, v2(2,N ! 52) ! 3.93, p > .13. Forthe Double Categorization Task, pairwise comparisonsdetected a significant difference only between the twoyounger age groups and the oldest one, v2(1, N ! 40)! 6.14, p < .02. For the Simple Categorization Task, a

ceiling effect was obtained, even if two 3-year-olds (11%of this age group) failed to perform above chance on thistask. Clearly, the difference in performance across thetwo tasks provides support for the hypothesis that cate-gorical flexibility involves more than only conceptualknowledge of both kinds of associations. We return tothis point in the next section.

General discussion

We explored whether age-related changes during the pre-school years in children’s categorical flexibility occur as aresult of changes in conceptual knowledge or changesin executive control, as categorical flexibility depends onboth. To assess categorical flexibility in Experiment 1,we used the paradigmatic match-to-sample task andadapted it: For each target, children had to select a firstmatch presented in a conflictual context involving twopotential associates and a nonassociate, and they then hadto select a second match. Categorical flexibility in thiscontext required that children access and switchsuccessively between the two simultaneously availableassociates. Thus, aswith research on cognitive flexibility inthe executive function literature and other domains (e.g.Carlson & Moses, 2001; Chevalier & Blaye, 2008; Frye etal., 1995; Gerstadt et al., 1994; Jacques & Zelazo, 2001;Luria, 1961; M#ller et al., 2004; Smidts et al.,2004; Zelazo, 2006; see also Diamond, 2006, Jacques &Zelazo, 2005, for recent reviews), categorical flexibility inour task involved successively switching between available,but conflicting, representations.Experiment 2 focused on assessing conceptual know-

ledge of both kinds of associations for each targetindependent of executive demands. That is, because theexecutive cost involved in the Double CategorizationTask might underestimate conceptual knowledge itself,we assessed children on the Simple Categorization Task.Selecting two associates on the Simple CategorizationTask did not require categorical flexibility as the two

Figure 3 Percentages of children with above-chanceflexibility scores as a function of age and task.

870 Agn!s Blaye and Sophie Jacques

! 2009 The Authors. Journal compilation ! 2009 Blackwell Publishing Ltd.

representations of the objects were not placed directly inconflict with one another. Instead, the Simple Categori-zation Task measured what Nguyen (Nguyen & Murphy,2003; Nguyen, 2007) called cross-classification, namelythe ability to access two representations of the sameobject on two separate occasions. Because accessing bothrepresentations of the same object was done on separateoccasions in this task, the potential interference resultingfrom having represented the first selected representationwas removed. In other words, children did not need toaccess two conflicting representations of the same objectand switch between them.Hence, performance on this taskprovided an index of children’s conceptual knowledge ofboth kinds of associations.Both tasks revealed significant increases in perform-

ance during the preschool period, but with differentdevelopmental trajectories: Whereas correct doubleselections increased between 4 and 5 years of age on theDouble Categorization Task, increases in correct doubleselections occurred earlier, between 3 and 4 years, on theSimple Categorization Task. In fact, our findings with theSimple Categorization Task replicated recent find-ings by Nguyen and Murphy (2003; Nguyen, 2007). Forexample, Nguyen (2007) identified a developmentalpattern similar to ours in that she found significantchanges in conceptual knowledge between 3 and 4, but notbetween 4 and 6 (cf. Rosch et al., 1976). Likewise, Nguyenand Murphy obtained similar findings, but these werelimited to conceptual knowledge of food categories.However, Nguyen and Murphy, and Nguyen did notinclude a task like the Double Categorization Task.In contrast, the period of rapid change in categorical

flexibility on the Double Categorization Task that weobserved between 4 and 5 years coincides well with theperiod of rapid change in cognitive flexibility observedon other tasks, including the FIST (Jacques & Zelazo,2001; see Jacques & Zelazo, 2005, for a review), a similarinductive task that measures cognitive flexibility usingperceptual dimensions (as opposed to categorical rela-tions). As in the Double Categorization Task, measuresof cognitive flexibility in the executive function literatureand other domains also involve conflicting responseoptions and switching between two responses, althoughcertain tasks like the DCCS are solved earlier, perhapsbecause of their deductive nature.Despite these similarities with previous research, how-

ever, our findings are unique in that they are the first todocument contrasting developmental pathways for con-ceptual and executive processes, two complementaryprocesses that are vital for the emergence of categoricalflexibility. More precisely, our findings indicate that thedevelopment of categorical flexibility occurs in two stepsduring the preschool years: First, children need to masterconceptual knowledge of thematic and superordinatetaxonomic categories, something they achieve by the ageof 4. Second, they then need to acquire sufficient executivecontrol skills, in particular cognitive flexibility, that allowsthem to access alternative conceptual representations, even

in circumstances when these conflict with each other.These findings might explain the relatively poor per-formance of preschool children in existing justificationstudies (e.g. Greenfield & Scott, 1986; Lucariello et al.,1992; Tversky, 1985), despite findings from other studiesdemonstrating conceptual knowledge of superordinaterelations in same-aged children (e.g. Rosch et al., 1976).Specifically, in justification studies, researchers generallyasked children to first select a match (for which they oftenopted for the thematic match) and then justify thealternative match (the taxonomic one) in the continuedpresence of the first more salient match, thereforerequiring categorical flexibility. On the basis of ourfindings, difficulties in these justification studies mayhave resulted fromexecutive control difficulties, not lackofconceptual knowledge.In addition, we found that children who tended to

select one kind of associate (either thematic or taxo-nomic) consistently for their first selection across trialson the Double Categorization Task also had higherflexibility scores (double correct selections) than childrenwho were less consistent, even after controlling for age.This finding is remarkable because it suggests that childrenwho use a top-down approach also demonstrate morecategorical flexibility, providing evidence for a linkbetween these two controlled processes. This finding isalso noteworthy because it provides additional evidencethat both conceptual knowledge (i.e. recognizing differentkinds of relations) and executive control skills (i.e.systematically selecting particular relations and cognitiveflexibility) contribute jointly to the development ofcategorical flexibility. Further research is now requiredto gain a better understanding of exactly how conceptualand executive aspects might support one another acrossdevelopment. The fact that gains in conceptual knowledgeprecede gains in cognitive flexibility suggests thatconceptual knowledge is not only a prerequisite forflexibility, but that it may actually support it (see alsoBlaye, Bernard-Peyron, Paour & Bonthoux, 2006; Blaye,Chevalier&Paour, 2007; Chevalier &Blaye, 2008; Jacqueset al., 2008; Maintenant & Blaye, 2005).However, recent research has suggested that categori-

cal flexibility might even be observed in infants (Ellis &Oakes, 2006; Mareschal & Tan, 2007). In fact, comparedto definitions of flexibility proposed byothers (e.g. Blaye&Bonthoux, 2001; De$k, 2003; Nguyen & Murphy, 2003;Ellis & Oakes, 2006; Mareschal & Tan, 2007; Waxman &Namy, 1997), we have restricted our meaning ofcategorical flexibility to a very conservative and narrowone:We propose that demonstrating categorical flexibilityrequires that children be exposed to simultaneouslyavailable conflicting representations of the same objectand be able to switch successively between these. Ourresults clearly demonstrate that the conditions in whichchildren are presented with different categoricalrepresentations of the same object (i.e. simultaneously or atdifferent times) are instrumental in enhancing or pre-venting them from being able to identify all relevant

Conception and flexibility 871

! 2009 The Authors. Journal compilation ! 2009 Blackwell Publishing Ltd.

representations. Indeed, in our study, it was especiallydifficult for children to represent objects flexibly withindifferent conceptual networks in situations that wereassociated with high executive control demands, but itwas relatively easier for them to switch at a younger agewhen these representations were not placed in conflictwith each other and were presented instead at different

times. However, even though our definition of categori-cal flexibility may seem unnecessarily exacting, trueflexibility requires that children be able to alternatebetween different representations at any given moment.As such, conceptual knowledge and executive controltogether must be considered integral parts of categoricalflexibility.

Appendix

List of stimuli used in Experiments 1 and 2

Target Thematic associate Taxonomic associate Nonassociate 1 Nonassociate 2 Nonassociate 3 Nonassociate 4

fish fishing rod fox broom playing cards pencil toothbrushshirt iron swimsuit kangaroo fork ax beetlecot sleep wear table watch helicopter kettle birdfrying pan egg pressure cooker star jacket crib teddy bearhouse key tent pear frog cup shipwood saw logs hammer ring horse sailboat conedog doghouse snail telephone tomato hammer pantscake candle chicken giraffe truck sock rattleflower watering can leaf crash helmet bed sausage guitarplane suitcase motorcycle bone pig violin stoolapple knife banana snowman gift box parasol toothpastemitten skis pyjamas butterfly sofa piano lionscrewdriver bicycle power drill pineapple water jug dress mousebird birdcage dolphin tap T-shirt toy robot excavatortree owl potted plant rubber ring doll glass dustbincarrot rabbit strawberry book bus bedside lamp binoculars

Acknowledgements

This research was funded by a grant (AF No. 03 2 438)from the French Ministry of Research to the first author.We thank Maelle Lechevrel-Loubet and Am"lie Doucetfor assistance with data collection. Part of this work waspresented at the XIth European Conference on Develop-mental Psychology, Milan, Italy, August 2003.

References

Baddeley, A. (1986). Working memory. Oxford: Oxford Uni-versity Press.

Blaye, A., Bernard-Peyron, V., Paour, J.L., & Bonthoux, F.(2006). Categorical flexibility in children: distinguishing re-sponse flexibility from conceptual flexibility; the protracteddevelopment of taxonomic representations. European Journalof Developmental Psychology, 3, 163–188.

Blaye, A., & Bonthoux, F. (2001). Thematic and taxonomicrelations in preschoolers: the development of flexibility incategorization choices. British Journal of DevelopmentalPsychology, 19, 395–412.

Blaye, A., Chevalier, N., & Paour, J.L. (2007). The develop-ment of intentional control of categorization behaviour: astudy of children’s relational flexibility. Cognition, Brain andBehavior, 11 (4), 789–808.

Carlson, S.M., & Moses, L.J. (2001). Individual differences ininhibitory control and children’s theory of mind. ChildDevelopment, 72, 1032–1053.

Chevalier, N., & Blaye, A. (2008). Cognitive flexibility in pre-schoolers: the role of representation activation and mainte-nance. Developmental Science, 11, 339–353.

De$k, G.O. (2003). The development of cognitive flexibilityand language abilities. In R. Kail (Ed.), Advances in childdevelopment and behavior, Vol. 31 (pp. 271–327). San Diego,CA: Academic Press.

Diamond, A. (2006). The early development of executivefunctions. In E. Bialystok & F.I.M. Craik (Eds.), Lifespancognition: Mechanisms of change (pp. 70–95). New York:Oxford University Press.

Ellis, A.E., & Oakes, L.M. (2006). Infants flexibly use differentdimensions to categorize objects. Developmental Psychology,42, 1000–1011.

Fenson, L., Vella, D., & Kennedy, M. (1989). Children’sknowledge of thematic and taxonomic relations at two yearsof age. Child Development, 60, 911–919.

Frye, D., Zelazo, P.D., & Palfai, T. (1995). Theory of mindand rule-based reasoning. Cognitive Development, 10, 483–527.

Garon, N., Bryson, S.E., & Smith, I.M. (2008). Executivefunction in preschoolers: a review using an integrativeframework. Psychological Bulletin, 134, 31–60.

Gerstadt, C.L., Hong, Y.J., & Diamond, A. (1994). The rela-tionship between cognition and action: performance ofchildren 3 1

2 7 years old on a Stroop-like day-night test.Cognition, 53, 129–153.

Greenfield, D.B., & Scott, M.S. (1986). Young children’spreference for complementary pairs: evidence against a shiftto a taxonomic preference. Developmental Psychology, 22, 19–21.

872 Agn!s Blaye and Sophie Jacques

! 2009 The Authors. Journal compilation ! 2009 Blackwell Publishing Ltd.

Hughes, C. (2002). Executive functions and development: whythe interest? Infant and Child Development, 11, 69–71.

Inhelder, B., & Piaget, J. (1964). The early growth of logic in thechild: Classification and seriation (E.A. Lunzer & D. Pepert,trans.). New York: Harper & Row. (Original work publishedin 1959.)

Jacques, S., & Zelazo, P.D. (2001). The Flexible Item SelectionTask (FIST): a measure of executive function in preschoolers.Developmental Neuropsychology, 20, 573–591.

Jacques, S., & Zelazo, P.D. (2005). On the possible roots ofcognitive flexibility. In B.D. Homer & C.S. Tamis-LeMonda(Eds.), The development of social understanding and commu-nication (pp. 53–81). Mahwah, NJ: Lawrence Erlbaum.

Jacques, S., Zelazo, P.D., & Lourenco, S.F. (2008). The role oflabeling on the development of cognitive flexibility. Manu-script under revision.

Kendler, T.S. (1995). Levels of cognitive development. Mahwah,NJ: Lawrence Erlbaum.

Lucariello, J., Kyratzis, A., & Nelson, K. (1992). Taxonomicknowledge: what kind and when? Child Development, 63,978–998.

Luria, A.R. (1961). The role of speech in the regulation of normaland abnormal behavior (J. Tizard, Ed.). New York: PergamonPress.

Maintenant, C., & Blaye, A. (2005, August). The developmentof categorical flexibility in children: Rule-switching but . . .which rules? Paper presented at the XIIth European Con-ference on Developmental Psychology, Tenerif, Spain.

Mareschal, D., & Tan, S.H. (2007). Flexible and context-dependent categorization by eighteen-month-olds. ChildDevelopment, 78, 19–37.

Markman, E.M., & Hutchinson, J.E. (1984). Children’s sensi-tivity to constraints on word meaning: taxonomic versusthematic relations. Cognitive Psychology, 16, 1–27.

Miller, E.K., & Cohen, J.D. (2001). An integrative theory ofprefrontal cortex function. Annual Review of Neuroscience,24, 167–202.

M#ller, U., Zelazo, P.D., Hood, S., Leone, T., & Rohrer, L.(2004). Interference control in a new rule use task: age-relatedchanges, labeling, and attention. Child Development, 75,1594–1609.

Murphy, G.L. (2002). The big book of concepts. Cambridge,MA: MIT Press.

Nelson, K. (1977).The syntagmatic-paradigmatic shift revisited:a review of research and theory. Psychological Bulletin, 84,93–116.

Nguyen, S.P. (2007). Cross-classification and category repre-sentation in children’s concepts. Developmental Psychology,43, 719–731.

Nguyen, S.P., & Murphy, G.L. (2003). An apple is more thanjust a fruit: cross-classification in children’s concepts. ChildDevelopment, 74, 1783–1806.

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

Scheuner, N., Bonthoux, F., Cannard, C., & Blaye, A. (2004).The role of associative strength and conceptual relations inmatching tasks in 4- and 6-year-old children. InternationalJournal of Psychology, 39, 290–304.

Smidts, D.P., Jacobs, R., & Anderson, V. (2004). The ObjectClassification Task for Children (OCTC): a measure ofconcept generation and mental flexibility in early childhood.Developmental Neuropsychology, 26, 385–401.

Smiley, S.S., & Brown, A.L. (1979). Conceptual preference forthematic or taxonomic relations: a nonmonotonic age trendfrom preschool to old age. Journal of Experimental ChildPsychology, 28, 249–257.

Tversky, B. (1985). Development of taxonomic organization ofnamed and pictured categories. Developmental Psychology,21, 1111–1119.

Walsh, M., Richardson, K., & Faulkner, D. (1993). Perceptual,thematic and taxonomic relations in children’s mental repre-sentations: responses to triads.European Journal of Psychologyof Education, 8, 85–102.

Waxman, S.R., & Namy, L.L. (1997). Challenging the notionof a thematic preference in young children. DevelopmentalPsychology, 33, 555–567.

Zelazo, P.D. (2006). The Dimensional Change Card Sort(DCCS): a method of assessing executive function inchildren. Nature Protocols, 1, 297–301.

Zelazo, P.D., & M#ller, U. (2002). Executive functions intypical and atypical development. In U. Goswami (Ed.),Handbook of childhood cognitive development (pp. 445–469).Oxford: Blackwell.

Zelazo, P.D., M#ller, U., Frye, D., & Marcovitch, S. (2003).The development of executive function in early childhood.Monographs of the Society for Research in Child Development,68 (3, Serial No. 274).

Received: 20 August 2007Accepted: 30 July 2008

Conception and flexibility 873

! 2009 The Authors. Journal compilation ! 2009 Blackwell Publishing Ltd.