Developmental Science 12:3 (2009), pp 426 - PDF document

Developmental Science 12:3 (2009), pp 426Ð437DOI: 10.1111/j.1467-7687.2008.00792.x© 2009 The Authors. Journal compilation © 2009 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA. Blackwell Publishing Ltd Harvard Graduate School of Education, USA Choosing your informant427© 2009 The Authors. Journal compilation © 2009 Blackwell Publishing Ltd.either of these two cases, children should retain theirtrust in a familiar informant even in the face of his orher temporary inaccuracy. Thus, although previousÞndings have clearly shown that preschoolers Ð especially4-year-olds Ð come to mistrust an inaccurate informant,this might apply only to unfamiliar informants and notto familiar informants.Experiment 1 examined two related aspects of informantfamiliarity. First, in four pre-test trials we asked whether3- and 4-year-olds use familiarity as a cue when decidingwhich of two informants to trust. To assess whetherchildren use this strategy, we varied the relative familiarityof two informants by presenting children with Þlms of afamiliar teacher from their preschool site paired with anunfamiliar teacher from an afÞliated preschool site. Inpre-test trials, both teachers served as informants wholabeled novel objects and pantomimed novel functions.Children were recruited from two preschool facilitiessuch that for children from preschool 1, teacher 1 wasfamiliar and teacher 2 was unfamiliar, whereas forchildren from preschool 2, teacher 1 was unfamiliar andteacher 2 was familiar. We anticipated that childrenwould be more likely to trust information provided bythe particular teacher who was familiar to them. Toprobe the scope of that trust, children were given twotypes of trial. On trials, they were invited to indicatewhom they wished to ask for information. On Endorsetrials, having heard each informant make a differentclaim, they were invited to say which claim they agreedwith. As a further probe of the childÕs relationship withthe familiar teacher, we asked the two teachers to ratetheir relationship with each child in their care using theStudentÐTeacher Relationship Scale Ð Short Form(Pianta, 2001). We anticipated that children exhibiting amore close and/or a less conßictual relationship with thefamiliar teacher might be especially prone to seek andendorse information from her as opposed to the relativelyunfamiliar teacher.Our second goal was to assess the extent to whichchildren take an informantÕs recent history of accuracyinto account even in cases where they could, in principle,rely solely on familiarity. Accordingly, after the four pre-test trials, children were presented with four accuracytrials in which they could assess the relative reliability ofthe two teachers. Half of the children viewed a Þlm inwhich the familiar teacher was 100% accurate whereasthe unfamiliar teacher was 0% accurate; the other halfof the children viewed a Þlm in which the familiarteacher was 0% accurate whereas the unfamiliar teacherwas 100% accurate. Then, children received four post-testtrials, equivalent in format to the pre-test trials. Thus,they were presented with Þlms in which the two informantslabeled novel objects. If children prefer to use evidenceof recent accuracy to differentiate between conßictingclaims, they are likely to choose the previously accurateinformant, regardless of familiarity. If children prefer tobe guided by their prior relationship with the teacher,they are likely to choose the more familiar informant,regardless of her recent accuracy.Forty-one children participated in this study: 20 3-year-olds ( = 3;4, = 3 months, range: 3;0Ð3;10, ninemale) and 21 4-year-olds ( = 4;5, = 4 months, range:4;0Ð5;0, 11 male). Twenty-two of the children (10 3-year-olds, 12 4-year-olds) were recruited from the main site ofa preschool in Buffalo, NY and the remaining 19 childrenwere recruited from a separate site of the preschool. Bothpreschools recruit children from a broad socioeconomicrange. The children were primarily White, although arange of ethnicities was represented. Children participatedwith the consent of their parent.All children were tested in three phases. First, childrenparticipated in eight pre-test trials in which they viewedmovies showing a familiar and an unfamiliar informantlabeling and demonstrating the function of novelobjects. ChildrenÕs preference for asking and endorsingthe more familiar informant was measured during theseeight trials. Second, children received four accuracy trialsin which one of the two informants consistently labeledfour familiar objects accurately and the other informantconsistently labeled the objects inaccurately. ChildrenÕsown names for these familiar objects were also elicited.Finally, children received a in which the twoinformants labeled novel objects. ChildrenÕs preferencefor asking and endorsing the familiar informant wasagain measured during four post-test trials. In addition,their explicit judgments regarding the accuracy of thetwo informants were obtained both immediately beforeand after these four post-test trials. Finally, childrenwere probed for their preferred explanation of why oneof the informants made errors. Each phase is describedin more detail below.Each child was tested in two sets of four trials: anovel object label set consisting of four trials, and a novel set consisting of four trials. The order ofsets was systematically varied across participants. Foreach of the two sets, a short Þlm was created comprisingfour clips, one for each trial. All clips featured the sametwo female preschool teachers. The teachers were similarin age and appearance and were recruited from the sitesof the two preschools in Buffalo, NY. Teacher 1 wasfamiliar to the children in the Þrst site (Teacher 2 wasunfamiliar) and Teacher 2 was familiar to the children inthe second site (Teacher 1 was unfamiliar). The two 428Kathleen Corriveau and Paul L. Harris© 2009 The Authors. Journal compilation © 2009 Blackwell Publishing Ltd.female teachers wore differently colored shirts and wereseated at a table. Each clip began with a male teacherstanding behind the two female teachers and placing anovel object on the table between them (e.g. a green rubbertoilet ßapper, a gold and red metallic sprinkler head; seeTable 1 for a full list of objects). Before each clip wasplayed, children were presented with a still photographof the relevant object. The order of trials was maintainedacross participants, as shown in Table 1.To introduce the task, the experimenter pointed to astill frame from the Þlm and said, ÔSee these two people?ThatÕs your teacher C. and sheÕs wearing a pink shirt.ThatÕs S. and sheÕs a teacher at another school and sheÕswearing a black shirt. TheyÕre going to show you somethings and tell you what they are called (what they areused for). I want you to listen very carefully and thenIÕm going to ask you some questions. LetÕs watch.Õ Oneach trial, children were asked about the name or functionof the novel object. After their ignorance had beenestablished, children were invited to ask one of the twoinformants for information (Ask Question). Both inform-then offered conßicting information about the novelobject; the male teacher repeated that information, andinvited children to endorse one of the two informantsFor each trial of Film 1 (Novel Object Labels) and (Novel Object Functions), children saw a still photographof a novel object and a corresponding video clip. On anygiven trial, once they had been shown the still photographof the novel object, children were asked, ÔDo you knowwhat this is called (what this is for)?Õ Children were givena chance to reply and were then presented with the , ÔI bet one of these people can help us Þnd out.Which person would you like to ask, C. with the pinkshirt, or S. with the black shirt?Õ Children who claimedto know the name or function of the novel object weretold, ÔActually, I donÕt think thatÕs what it is called (whatit is for). I bet one of these people can help us Þnd out.Which person would you like to ask, C. with the pinkshirt, or S. with the black shirt?ÕNext, children saw a video clip in which a maleteacher placed the relevant novel object on the table andasked one teacher, ÔCan you tell me what this is called?Õor ÔCan you tell me what this is for?Õ The Þrst femaleteacher responded by producing a novel label (e.g.ÔThatÕs a sneggÕ) or by producing and pantomiming anovel function (e.g. ÔYou use that for feppingquestion was posed to the second female teacher, whoproduced a different, novel label (e.g. ÔThatÕs a produced and pantomimed a different, novel function(e.g. ÔYou use that for rokingÕ). In each Þlm, the order inwhich teachers were asked questions alternated acrossthe four video clips.Endorse Questions were posed after children hadwatched the video clip. The experimenter paused thevideo, reiterated the information supplied by the twoinformants, and asked children what they thought theobject was called (used for). For example, in the novel trials, the experimenter said, ÔC. in the pinkshirt said itÕs a snegg and S. in the black shirt said itÕs a. What do you think itÕs called, a snegg or a Children were requested to give either a verbal (Ôwhat S.said,Õ Ôa sneggÕ) or a nonverbal (pointing) response.Because previous experiments found no systematicdifference in performance between and Endorse, we expected to combine answers to thesequestions in order to achieve more statistical power fromeach individual trial.A short Þlm of four clips was createdfor the accuracy trials. The Þlm featured the same twoinformants (one the familiar preschool teacher, one theunfamiliar preschool teacher) again wearing different,solid-colored shirts seated at a table. As in pre-test trials,each clip began with a male teacher standing behind twofemale teachers and placing an object on the tablebetween them. On all four trials, the objects were familiar(e.g. spoon, bottle; see Table 1 for a full list of objects usedin accuracy trials). The order of trials was maintainedacross participants.Novel objectsInformant 1 labelsInformant 2 labelsNovel labelsGrey rubber squeegeeÔThatÕs a sneggÕÔThatÕs a hoonÕBlue toilet ßapperÔThatÕs a yiffÕÔThatÕs a zazzÕMetal cocktail pourerÔThatÕs a crutÕÔThatÕs a larpÕMetal bathroom hookÔThatÕs a linzÕÔThatÕs a slodÕNovel functionsYellow plastic sprinkler attachmentLook through like a telescopeHold up to mouth and blowWooden orange juicer Roll on tableHammer on tableBlack and grey knee padSnap like a slingshot Use as a hatBlack toilet plungerSpin like a topSquish togetherAccuracy trialsSpoonÔThatÕs a duckÕÔThatÕs a spoonÕBottleÔThatÕs a bottleÕÔThatÕs an appleÕBrushÔThatÕs a brushÕÔThatÕs a plateÕDollÔThatÕs a cupÕÔThatÕs a dollÕPost-test trialsOrange hose attachmentÔThatÕs a ligÕÔThatÕs a joobÕGold and red sprinkler headÔThatÕs a doapÕÔThatÕs a thafÕGreen toilet ßapperÔThatÕs a tarkÕÔThatÕs a chabÕMetal lemon juicerÔThatÕs a neviÕÔThatÕs a mogoÕ Choosing your informant429© 2009 The Authors. Journal compilation © 2009 Blackwell Publishing Ltd.To introduce the task, the experimenter pointed to astill frame from the Þlm and said, ÔNow C. and S. aregoing to show you some more things and tell you whatthey are called. I want you to listen very carefully andthen IÕm going to ask you some questions. LetÕs watch.ÕThe accuracy of the teachersÕ claims was not mentionedin this introduction.On each accuracy trial, children were Þrst presentedwith a picture of a familiar object and then watched avideo clip of the male teacher, the two female teachersand the familiar object. Trials began when the maleteacher placed the object on the table between the twofemale teachers and asked one teacher, ÔCan you tell mewhat this is called?Õ One teacher labeled all four objectscorrectly (100% correct). For example, when presentedwith a brush, the accurate teacher said, ÔThatÕs a brush.ÕThe other teacher labeled all four objects incorrectly (0%correct). For example, when presented with a brush, theinaccurate teacher said, ÔThatÕs a plate.Õ For half of theparticipants, the familiar teacher was 100% correct andthe unfamiliar teacher was 0% correct. For the otherhalf, the unfamiliar teacher was 100% correct and thefamiliar teacher was 0% correct. In each Þlm, the orderin which the teachers were asked to label the familiarobject alternated across the four video clips. In everyclip, the object labels provided by the two teachers werematched for age of acquisition (Fenson, Dale, Reznick,Bates, Thal & Pethick, 1994).Name Checks occurred after viewing each video clip.The experimenter paused the video and asked childrenwhat they thought the object was called. For example,the experimenter said, ÔC. in the pink shirt said itÕs a and S. in the black shirt said itÕs a . What doyou think itÕs called, a or a The post-test consisted of three Explicit Judg-, followed by four post-test trials involvingnovel objects, and a further three Explicit Judgment. To pose the Explicit Judgment Questionsexperimenter referred to a still frame of the video andasked, ÔWas C. in the pink shirt very good or not verygood at answering these questions?Õ The experimenterthen repeated this question in reference to the otherinformant (S. in the black shirt). Finally, children wereasked to make a judgment about the relative accuracy ofthe two informants, ÔWhich person was better at answeringthe questions?Õ These three Explicit Judgment questionswere asked after the fourth and Þnal accuracy trial.The four followed the same format asthe four pre-test trials for novel object labels. Finally,immediately following the fourth and Þnal post-test trial,children were asked the second set of Explicit Judgment. The format was identical to that used forFinally, an Explanation Probe was asked following theExplicit Judgment questions. Children were reminded ofa speciÞc error that one of the informants made duringa familiarization trial and were asked why the error wasmade. For example, children might be asked, ÔRememberwhen S. in the black shirt said that the brush was aplate? Why do you think she said that? Was it becauseshe didnÕt know what it was called, or because she wasjust pretending?Õ The order of the two forced-choicealternatives varied across children.Teacher Relationship Scale Ð Short Form (Pianta, 2001)was used to assess the relationship between the childrenand the familiar preschool teacher who had been presentedin the Þlms. Teachers were asked to read and rate (from1: deÞnitely does not apply, to 5: deÞnitely applies) 15statements about different aspects of their relationshipwith each child at the preschool where they worked.Seven of the questions were used to compile a closenessmeasure. The remaining eight questions were used to measure. Neither measure predictedthe pattern of information-seeking that was observed, asbrießy described in the discussion of Experiment 1.We Þrst examine pre-test trials to check whether childrenpreferred to ask for and endorse information from themore familiar teacher within each of the two trial sets(novel object labels and novel object functions). We thenreport on childrenÕs replies to the name checks duringthe accuracy trials. Next, we examine childrenÕs repliesto the Explicit Judgment questions Ð after they hadreceived accuracy information. We then analyze to whatextent children preferred the more familiar informant inpost-test trials as compared to pre-test trials. Finally, wereport childrenÕs replies to the Explanation Probe.To anticipate, both age groups showed a preferencefor the familiar teacher in pre-test trials. The pattern ofresponding in post-test trials varied with age. Three-year-olds were relatively unaffected by the differentialaccuracy of the two teachers during the accuracy trials.Thus, they maintained their initial preference for thefamiliar teacher, irrespective of how the two teachersbehaved during the accuracy trials. By contrast, 4-year-olds displayed a stronger preference for the familiarteacher if she had proved accurate but no preference ifshe had proved inaccurate.Scores for the Ask and Endorse questions are found inTable 2 together with comparisons to chance via -tests.Scores on questions represent the proportion oftrials on which children asked for information about thenovel objects from the more familiar informant. Both3- and 4-year-olds performed above chance in asking themore familiar informant. Scores on Endorse represent the proportion of trials on which childrenendorsed the label provided by the more familiar 430Kathleen Corriveau and Paul L. Harris© 2009 The Authors. Journal compilation © 2009 Blackwell Publishing Ltd.informant. Both 3- and 4-year-olds performed abovechance in endorsing the more familiar informantÕs label.In summary, on the novel object label trials, both 3- and4-year-olds preferred to ask for and to endorse informa-tion from the more familiar informant.Proportion correct for the Ask and Endorse ques-tions are found in Table 2. Both 3- and 4-year-oldsperformed above chance in asking the more familiarinformant. They also performed above chance in endorsingthe more familiar informant. In summary, as in the novelobject label trials, both 3- and 4-year-olds preferred toask for and endorse information from the more familiarinformant.Every 3- and 4-year-old accurately chose the correctlabel for the familiar objects in all four accuracy trials.Thus, childrenÕs naming was unaffected by the incorrectnames supplied by one of the two teachers, irrespectiveof whether she was familiar.The proportion of that children responded accurately to the Þrst (EJ1) andsecond (EJ2) set of Explicit Judgment questions isshown in Table 3. Inspection of Table 3 reveals that overall4-year-olds gave more correct replies than 3-year-olds. Inaddition, children generally replied correctly if the familiarinformant had been accurate, but were less likely to replycorrectly if the familiar informant had been inaccurate,especially on the second set of questions (EJ2). To checkthese conclusions, a three-way ANOVA with age (3, 4)and condition (familiar 100% correct, familiar 0% correct)as the between-subjects variables and EJ question (EJ1,EJ2) as the within-subjects variable was calculated forthe number of correct replies. The main effect of Agegroup ( = .12) conÞrmed that4-year-olds gave more correct replies than 3-year-olds (3-year-olds, = .59, = .27; 4-year-olds, = .78, = .29). In addition, there were main effects of Condition = .24), and EJ question = .20) and an interactionbetween these variables (To interpret the interaction, the simple effect of Con-dition was calculated for each set of EJ questions. On theÞrst set of questions, preschoolersÕ judgments about thetwo informants were generally correct for both conditions). On the second set of questions, childrenÕsjudgments were less likely to be correct if the familiarinformant had been 0% rather than 100% accurate ( ), as shown in Table 3, childrenwere generally correct on the Þrst set of EJ questions butExperiment 1Experiment 23-year-olds = 204-year-olds = 215-year-olds = 20LabelsAsk.71 (.15)5.68***.60 (.19)2.36*.83 (.17)8.85***Endorse.64 (.24)2.34*.62 (.20)2.68*.71 (.19)5.10***Ask.76 (.21)6.47***.67 (.20)3.84***.78 (.14)8.90***Endorse.65 (.29)2.60*.69 (.16)5.59***.79 (.19)6.90***TotalsLabels Total.66 (.17)4.20***.61 (.16)2.99**.77 (.10)11.83***Functions Total.71 (.22)4.33***.68 (.15)5.62***.78 (.13)9.87***Overall Total.68 (.17)4.95***.64 (.13)5.04***.78 (.08)15.98***Total Site 1.64 (.18)2.57**.65 (.12)3.77***.78 (.07)12.17***Total Site 2.73 (.15)4.63***.65 (.14)3.34**.77 (.08)9.83*** Experiment 1Experiment 23-year-olds4-year-olds5-year-oldsFamiliar 100% correct = 10 = 10 = 10Explicit Judgment 1.78 (.33).90 (.22)1.0 (0)Explicit Judgment 2.70 (.20).93 (.21).93 (.21)Familiar 0% correct = 10 = 11 = 10Explicit Judgment 1.67 (.37).75 (.30).93 (.21)Explicit Judgment 2.27 (.29).54 (.40).90 (.16)TotalFamiliar 100% correct Total.74 (.27).92 (.22).98 (.11)Familiar 0% correct Total.47 (.33).65 (.35).92 (.19) Choosing your informant431© 2009 The Authors. Journal compilation © 2009 Blackwell Publishing Ltd.made errors on the second set, especially when it was thefamiliar informant who had been inaccurate.Explanation Probe, when the familiar informanthad been 0% correct, 60% of 3-year-olds and 54% of 4-year-olds said that she ÔdidnÕt knowÕ the names of thefamiliar objects. The remaining children Ð 40% of 3-year-olds and 46% of 4-year-olds Ð said that she was ÔjustpretendingÕ. When the unfamiliar informant had been0% correct, 80% of 3-year-olds and 80% of 4-year-oldssaid that she ÔdidnÕt knowÕ the names of the familiarobjects. The remaining children Ð 20% of 3-year-olds and20% of 4-year-olds Ð said that she was Ôjust pretendingÕ.Thus, a greater proportion of children attributed ignoranceto the unfamiliar informant.To assess the impact of accuracytrials on childrenÕs preference for the familiar teacher, wecompared childrenÕs scores on the Ask and Endorseprobes during post- versus pre-test trials. These scoresare shown in Table 4.A four-way ANOVA with Age (3, 4) and Condition(familiar 100% accurate, familiar 0% accurate) as thebetween-subjects variables and Question Type (ask,endorse) and Phase (pre-test, post-test) as the within-subjects variables was calculated for the number ofchoices directed at the familiar informant. This revealeda main effect of Condition ( = .37) and a three-way interaction of Age Condition Phase ( = .15). Inspectionof Table 4 suggests that 4-year-olds but not 3-year-oldsaltered their preference from pre- to post-test trials.To check this conclusion, the simple effect of Phase wascalculated for each of the four possible Age Conditioncombinations. Three-year-olds did not show any shift inpreference for the familiar teacher between pre- andpost-test trials, regardless of condition (familiar 100%correct: ; familiar 0% correct: ). In contrast, 4-year-olds showed a greater pref-erence for the familiar teacher after the accuracy trials ifshe had been 100% accurate ( )If the familiar teacher had been 0% accurate, theirpreference for her weakened slightly after the racy trials but this decline was not signiÞcant (In the introduction, we asked whether preschool childrentrust a familiar rather than an unfamiliar informant andwhether that preference is altered if the familiar informantproves accurate or inaccurate. Our Þndings provide aclear answer to the Þrst question. Both age groups pre-ferred to ask for and endorse information about novellabels and functions from the familiar informant ratherthan the unfamiliar informant. This preference emergedin each preschool. Thus, the teacher who was systemat-ically preferred in one preschool was systematicallyless preferred in the other preschool. By implication,childrenÕs selectivity cannot be attributed to a preferencefor the appearance or interactive style of just one of thetwo teachers. Instead, it can be plausibly attributed tothe differential familiarity that children had with the twoteachers in each preschool.We had speculated that childrenÕs trust in a familiarteacher might be moderated by the extent to whichchildren had either a close or a conßictual relationshipwith the familiar teacher (as measured by the StudentÐTeacher Relationship Scale Ð Short Form; Pianta, 2001).We found no effect for either scale when the four-wayANOVA reported above was re-calculated with themeasures of closeness and of conßict included as cov-ariates. Arguably, is the main determinantof childrenÕs selective trust, rather than the particulartype of emotional relationship that they have established.However, two major caveats are warranted. First, bothteachers were experienced preschool workers with astable history of employment at the respective facilities.Hence, they had probably established relatively closerelationships with most, if not all, of the children intheir care. Indeed, scrutiny of the scores for the group ofchildren at each preschool conÞrms that scores wereconcentrated in the upper and lower half of the scalesfor closeness and conßict, respectively. Thus, neitherteacher reported having a distant relation to any child inher care. Second, direct observation of teacherÐchildrelationships might provide a more probing assessmentof whether childrenÕs emotional relationship to aninformant affects their trust in that informant. Forexample, measures derived from attachment theory,especially if they involved an assessment of childrenÕsrelationship to a primary caregiver, might reveal thatcertain emotional aspects of childrenÕs relationship witha given informant do affect their trust.AskEndorse3-year-oldsPre-test Familiar 100% correct.75 (.14).64 (.23)Pre-test Familiar 0% correct.70 (.16).66 (.22)Post-test Familiar 100% correct.70 (.21).61 (.22)Post-test Familiar 0% correct.62 (.13).57 (.16)4-year-oldsPre-test Familiar 100% correct.68 (.13).63 (.13)Pre-test Familiar 0% correct.59 (.17).68 (.18)Post-test Familiar 100% correct.93 (.12).93 (.12)Post-test Familiar 0% correct.45 (.29).52 (.24)5-year-oldsPre-test Familiar 100% correct.84 (.12).71 (.14)Pre-test Familiar 0% correct.76 (.09).79 (.12)Post-test Familiar 100% correct.90 (.18).88 (.18)Post-test Familiar 0% correct.20 (.20).15 (.18) 432Kathleen Corriveau and Paul L. Harris© 2009 The Authors. Journal compilation © 2009 Blackwell Publishing Ltd.We now turn to the second question, namely theextent to which evidence for the inaccuracy of either thefamiliar or unfamiliar informant altered the pattern ofselective trust that children had displayed during thepre-test trials. Three-year-olds were minimally affected bythis accuracy information. First, they were less accuratein their replies to the Explicit Judgment questions than4-year-olds. Second, the analysis of responses in pre-and post-test trials conÞrmed that 3-year-oldsÕ overallpreference for the familiar informant during pre-test trialsremained unaltered during post-test trials, whether thefamiliar informant had been 100% or 0% correct duringthe accuracy trials. For 3-year-olds, therefore, familiarityappears to be a more important heuristic than accuracywhen selecting between two informants. More generally,these Þndings are consistent with a pattern that hasrecurred across several earlier experiments. Three-year-oldsare less sensitive to recent variation in the accuracy oftheir informants than are 4-year-olds (ClŽment Koenig , 2004; Koenig & Harris, 2005, Experiment 1;Pasquini Four-year-olds were affected by exposure to thedifferential accuracy of the two informants during theaccuracy trials. First, they gave more correct replies tothe Explicit Judgment questions than 3-year-olds. Second,the analysis of responses in pre- and post-test trials con-Þrmed that 4-year-oldsÕ initial preference for the familiarinformant tended to intensify or attenuate depending onthe relative accuracy of the two informants. Thus, when thefamiliar informant had been 100% accurate, 4-year-oldsdisplayed a stronger preference for her. By contrast, ifthe familiar informant had been 0% accurate, 4-year-oldsÕinitial preference for her tended to weaken.One Þnal result of Experiment 1 deserves comment. Bothage groups were relatively accurate in their replies to theÞrst set of Explicit Judgment questions but often madeerrors in the second set. In particular, they were likely toclaim that the familiar teacher had been accurate evenwhen she had been consistently inaccurate. A plausibleinterpretation of this result is that children were prone tomisremember information inconsistent with their initialschema of that teacher. Similar results were reported byLeichtman and Ceci (1995). Three- and 4-year-olds withprior expectations regarding a male visitor to their day-carecenter were more likely to misremember the details of hisvisit than children with no prior expectations.The results of Experiment 1 suggest that there is adevelopmental shift. Unlike 3-year-olds, 4-year-olds weighrecent accuracy alongside familiarity when decidingwhom to trust. We might expect the pattern displayedby 4-year-olds in Experiment 1 to emerge even morestrongly among older children. Accordingly, in Experi-ment 2, we tested 5-year-olds using the same procedureTwenty 5-year-olds ( = 5;8, = 3 months, range:5;3Ð6;1, nine male) participated in this study. Ten of thechildren were recruited from one kindergarten classroomof a school in Brookline, MA and the remaining 10children were recruited from another kindergartenclassroom at the other end of the school (the two kinder-garten classrooms did not interact). The kindergartenrecruited children from a broad socioeconomic range.The children were primarily White, although a range ofethnicities was represented. Children participated withthe consent of their parent.The procedure was identical to the procedure in Experi-ment 1. Children participated in pre-test trials in whicha familiar and an unfamiliar teacher labeled and demon-strated the function of novel objects. Next, they receivedaccuracy trials in which one of the two informants con-sistently labeled four familiar objects accurately and theother informant consistently labeled the objects inaccu-rately. Finally, children received a comprisingtwo sets of Explicit Judgment questions, four in which the two informants labeled novel objectsExplanation ProbeAs in Experiment 1, every Þlm shown to the childrenfeatured the same two female kindergarten teachers. Thetwo female teachers were similar in age and appearanceand were recruited from a school in Brookline, MA.Teacher 1 was familiar to the children in the Þrst class-room (Teacher 2 was unfamiliar) and Teacher 2 wasfamiliar to the children in the second classroom (Teacher1 was unfamiliar). The two female teachers wore ferently colored shirts and were seated at a table. Tointroduce the task, the experimenter pointed to a stillframe from the Þlm and said, ÔDo you know who this is?ThatÕs right, thatÕs your teacher X and sheÕs wearing agreen shirt. Do you know who this is? ThatÕs Y andsheÕs a different kindergarten teacher and sheÕs wearingSince the StudentÐTeacher Relationship Scale had notidentiÞed a link between selective trust and childrenÕsrelationship with their teachers, this scale was notAs in Experiment 1, we Þrst examine pre-test trials tocheck whether 5-year-olds prefer to ask for and endorseinformation from the more familiar teacher within eachof the two trial sets (novel object labels and novel objectfunctions). We also report on the accuracy of childrenÕsnaming during the accuracy trials. Next, we examinechildrenÕs performance on the two sets of Explicit Choosing your informant433© 2009 The Authors. Journal compilation © 2009 Blackwell Publishing Ltd.Judgment questions and the Explanation Probe. We thenanalyze to what extent children preferred the morefamiliar informant in post-trials as compared to pre-testtrials. Finally, we provide a statistical analysis of theÞndings across Experiments 1 and 2.The proportion of times that children chose the morefamiliar teacher for the Ask and Endorse questions isshown in Table 2, together with comparisons to chance-tests. Five-year-olds performed above chance inpreferring the more familiar teacher on both andEndorse questions.The proportion of times that children chose the morefamiliar teacher for the Ask and Endorse questions isfound in the right-hand column of Table 2. As in thenovel object label trials, 5-year-olds performed abovechance in asking and endorsing the more familiar teacher.Every 5-year-old accurately chose the correct label forthe familiar objects in all four accuracy trials. Thus, chil-drenÕs naming was unaffected by the incorrect namessupplied by one of the two teachers, irrespective ofwhether she was familiar.The proportion oftimes that children responded correctly to the Þrst (EJ1)and second (EJ2) set of Explicit Judgment questions isshown in the right-hand column of Table 3. Inspectionof Table 3 shows that children almost always repliedcorrectly, irrespective of condition and question set. Asexpected, a two-way ANOVA with condition (familiarcorrect, familiar 0% correct) as the between-subjects variable and EJ question (EJ1, EJ2) as the within-subjects variable produced no signiÞcant main effects orinteractions, conÞrming that 5-year-oldsÕ identiÞcationof the more accurate informant was consistently good.Explanation Probe, when the familiar informanthad been 0% correct, 50% of 5-year-olds said that sheÔdidnÕt knowÕ the names of the familiar objects and theremaining 50% of children said that she was ÔjustpretendingÕ. When the unfamiliar informant had been0% correct, 70% of 5-year-olds said that she ÔdidnÕt knowÕthe names of the familiar objects and the remaining 30%of children said that she was Ôjust pretendingÕ. Thus,as in Experiment 1, a greater proportion of childrenattributed ignorance to the unfamiliar informant. AChi-square test conÞrmed that a marginally greater pro-portion of children (collapsed across Experiments 1 and2) said that the unfamiliar, inaccurate informant did notknow as compared to the familiar, inaccurate informant = .06).As in Experiment 1, we askedhow often children selected the more familiar teacher inthe post-test as compared to the pre-test trials. The pro-portion of times that children chose the familiar teacheris shown in the lower panel of Table 4. Inspection ofTable 4 reveals that 5-year-oldsÕ preference for the familiarinformant persisted, and even intensiÞed, from pre- topost-test trials if the familiar informant had been accurate,whereas they preferred to endorse the unfamiliar (butaccurate) informant on post-test trials if the familiarinformant had been inaccurate. A three-way ANOVA withCondition (familiar 100% accurate, familiar 0% accurate)as the between-subjects variable and Question Type(ask, endorse) and Phase (pre-test, post-test) as the within-subjects variables was calculated on the number of choicesdirected at the familiar informant. This conÞrmed the maineffects of Condition ( = .79) togetherwith the interaction of Condition Phase ( = .88).To interpret this interaction, the simple effect of Phasewas calculated for each condition. Five-year-olds revealeda similar preference for the familiar informant acrosspre- and post-test trials in the 100% accurate condition), but sharply reduced their preferencefor the familiar informant on post-test trials in the 0%accurate condition ( )In conclusion, prior to any accuracy information,5-year-olds were more likely to seek and endorse infor-mation from a familiar rather than an unfamiliarteacher. However, this pattern of trust was signiÞcantlyaffected by information about the relative accuracy ofthe informants. Five-year-olds appropriately identiÞedwhich informant had been more accurate and preferredto ask and endorse that informant, regardless of theirfamiliarity with her. Thus, in post-test trials, 5-year-oldsÕinitial preference for the familiar informant was replacedby a preference for the unfamiliar informant if she hadproved to be the more accurate.The proportion of times thatchildren chose the familiar informant during the post-testof Experiments 1 and 2 is shown in Figure 1 as a func-tion of Age and Condition. Inspection of Figure 1 showsthat differentiation between the two accuracy conditionsincreased with age. To check this conclusion, a three-wayANOVA with Age (3, 4, 5) and Condition (familiar 100%accurate, familiar 0% accurate) as the between-subjectsvariables and Question Type (ask, endorse) as the within-subjects variable was calculated on the number of timesthat children chose the familiar informant. Main effectsof Age ( = .22) and Condition = .69) were found. As 434Kathleen Corriveau and Paul L. Harris© 2009 The Authors. Journal compilation © 2009 Blackwell Publishing Ltd.expected, a two-way Age Condition interaction alsoemerged ( = .49). Thisinteraction is displayed in Figure 1.To clarify the Age Condition interaction, the simpleeffect of Condition was calculated for each age group.Three-year-olds did not show any effect of condition onthe strength of their preference for the familiar teacher). However, both 4- and 5-year-oldsasked and endorsed the familiar teacher more oftenwhen she had been 100% accurate, rather than 0%accurate (4-year-olds: ear- ), unlike 3-year-olds,4-year-olds and particularly 5-year-olds modiÞed theirpreference for the familiar teacher in the wake ofaccuracy information.In order to explore whether childrenÕsability to identify the more accurate informant (ExplicitJudgment performance) affected post-test preference forthe familiar informant, we repeated the above three-wayanalysis of Age, Condition, and Question type butincluded EJ1 performance as a covariate. We found nomain effect of EJ1 (). Nevertheless, weretained the main effects of Age ( = .09) and Condition ( .30) as well as the interaction of Age Condition (= 15.56, = .30) found in the previous ANOVA.Experiment 2 extends the results of Experiment 1 to anolder age group. We asked if 5-year-olds, like 3- and4-year-olds, prefer a familiar to an unfamiliar informantin the absence of any other cues. We also asked if sucha preference for a familiar informant is altered byaccuracy information. Five-year-olds displayed a strongpreference for the familiar informant both when askingabout novel objects and functions and in endorsing theclaims made by the familiar informant.Five-year-oldsÕ initial preference for the familiarinformant was clearly affected by accuracy information.They identiÞed the more accurate informant in both setsof Explicit Judgment questions. When the familiarinformant as opposed to the unfamiliar informantlabeled objects accurately, 5-year-olds displayed a con-tinuing preference for the familiar informant. However,when she named objects incorrectly and the unfamiliarinformant named them correctly, 5-year-olds appropri-ately chose to ask for and endorse information from theunfamiliar, but accurate, informant.Taken together, the results of Experiments 1 and 2support two main conclusions. First, preschool childrenprefer information from familiar as compared to un-familiar informants. Second, older preschoolers are morelikely to moderate that preference in the wake of informa-tion about the relative accuracy of the two informants.We consider each of these conclusions in turn beforeconsidering both the limitations and broader implicationsof the Þndings.The preference for familiar as compared to unfamiliarinformants that was observed in Experiments 1 and 2 iswide-ranging. It applies to childrenÕs informationseeking as shown by responses to the Ask probes and itapplies to their acceptance of information as shown byresponses to the Endorse probes. Moreover, the preferenceis apparent for both object names and object functions.The preference is also stable across age groups. Itemerged among 3-, 4-, and 5-year-olds. Indeed, an omnibusANOVA (in which data from the two experiments werecombined) conÞrmed that the strength of childrenÕs pre-ference for the familiar informant in the pre-test phasedid not vary with age for children in either condition.Preference for a familiar informant might be inter-preted as shyness or wariness about asking questions ofa stranger. However, as just noted, a similar pattern ofresults emerged for Endorse as well as for Ask probes.Although children preferred to direct their questions tothe familiar informant, recall that once both informantshad volunteered information regarding the name of thenovel object or its function the experimenter re-statedtheir claims (e.g. ÔC. in the pink shirt said itÕs a snegg andS. in the black shirt said itÕs a . What do you thinkitÕs called, a snegg or a ?Õ). ChildrenÕs selectiveendorsement of new information cannot be readilyattributed to shyness or wariness because the two namesand the two demonstrations were equally novel and theexperimenter repeated them. Accordingly, it is reasonableto interpret childrenÕs selectivity in terms of differentialtrust in the two informants rather than in terms of strangeranxiety or a desire to interact with a known teacher.On what basis might children come to seek out andtrust the information provided by a familiar teacher? Wemay consider four different possibilities. One possibilityis that repeated exposure is, in itself, sufÞcient to increasetrust. For example, research with adults has shown Figure 1Proportion of times children chose the familiar informant by age group and condition in Experiment 1 (3- and 4-year-olds) and Experiment 2 (5-year-olds). Choosing your informant435© 2009 The Authors. Journal compilation © 2009 Blackwell Publishing Ltd.convincingly that when a stimulus is repeatedly presented,liking for that stimulus increases in a relatively automaticfashion, independent of slower-operating, recognitionprocesses (Zajonc, 1980, 2001). It is conceivable that asimilar process operates among young children. Mereexposure to a person might increase liking. In futureresearch, this could be tested by showing children ashort Þlm of either informant A or informant B andchecking whether children subsequently display greatertrust in the more familiar informant. However, preliminaryevidence against this interpretation has already emergedin ongoing research. Children classiÞed as avoidant at14 months on the basis of the Strange Situation worth, Blehar, Waters & Wall, 1978) did not displaypreferential trust in their mother as compared to anunfamiliar adult when tested at both 50 and 61 monthsin a choice procedure comparable to the pre-test trials ofExperiments 1 and 2. By contrast, both secure andambivalent children did display such a preference(Corriveau, Harris & Nelson, 2007). These Þndings suggestthat the familiarity that ensues from repeated exposureis not sufÞcient to lead to preferential trust becausechildren will ordinarily have repeated exposure to theirmother even if they establish an avoidant relationshipwith her. At the very least, these Þndings show that ifthere is an effect of mere exposure, it can be overriddenby other aspects of the interaction with the informant.A second possibility is that childrenÕs preference forthe familiar informant is based on their experience of herauthority and expertise as a teacher. Note that childrenwere told that the unfamiliar informant was a teacherbut they had no experience of her in that role. The mainweakness of this interpretation is that it misrepresentsthe everyday role of the familiar teacher, particularly forthe 3-year-olds of Study 1. Formal lessons were not partof the preschool curriculum at either of the two sites.Thus, although it is likely that the familiar teacher inboth sites had served as an informal model and informant,it is doubtful that younger children construed her assomeone with any special epistemic authority or expertise.The 5-year-olds tested in Study 2 did receive someformal instruction and it is plausible that they attributedthe expertise and authority that is associated with therole of teacher to the familiar informant. Yet, as notedabove, there was no evidence that 5-year-olds showed astronger preference for the familiar informant as pared to 3- and 4-year-olds.A third possible interpretation of the preference for afamiliar informant is that repeated exposure to a givenadult typically ensures that children hear that personproduce a large number of true or plausible claims andbuild up a deep reservoir of trust. Thus, each time thata familiar informant, be it a parent, a daycare provider,or a preschool teacher, makes a claim about an object orevent that children can conÞrm on the basis of Þrst-handobservation or prior knowledge, their trust in thatinformant might be strengthened if they judge that theclaim is accurate Ð or weakened if they judge that it isinaccurate. Such trust might be especially likely toaccumulate in certain types of distinctive interaction.Thus, Gergely and his colleagues have argued thathuman caregivers are well equipped to cue infants andyoung children to the fact that they are providing newinformation (Gergely, Egyed & Kir‡ly, 2007). In duecourse, such cumulative, accuracy-based trust might beextended to claims that children cannot check. Such aninductive strategy corresponds to the strategy that Hume(1748/1957) believed to operate among adults. Certainly,there is evidence that young children assert the existenceof various entities (e.g. germs, the Tooth Fairy) that theyare told about but cannot ordinarily observe (Harris,Pasquini, Duke, Asscher & Pons, 2006). Taken together,however, the Þndings of Experiments 1 and 2 suggestthat even if young children are capable of monitoringfor accuracy, their trust in familiar informants is notexclusively grounded in accuracy monitoring. First,recall that 3-year-olds showed a clear preference for thefamiliar informant in pre-test trials but they did notmodify that preference in the face of accuracy informa-tion. By implication, preference for familiar informantsis not based on accuracy monitoring. Second, accuracy-trust might be expected to accumulate morerapidly among older preschoolers on the assumptionthat they receive and process more verbal claims thanyounger preschoolers. Yet, as noted, preference forthe familiar informant was no stronger among olderpreschoolers.A Þnal possibility is that repeated exposure to a givenadult typically means that children experience numerousfriendly or cooperative interactions with that person.The ensuing impression is likely to produce a ÔhaloÕeffect: Children come to judge that familiar person morepositively on several dimensions Ð more likeable, morecompetent, and more trustworthy as compared to astranger. Indeed, irrespective of age, childrenÕs replies tothe explanation probe suggested that they were less willingto acknowledge ignorance on the part of the familiar ascompared to the unfamiliar informant. This interpreta-tion suggests that childrenÕs selective trust in familiarinformants is based neither on repeated exposure nor onaccuracy monitoring but primarily on the positiveemotional quality of repeated interaction. Note thatsuch an interpretation has the advantage of being con-sistent with the Þndings for different attachment groupsreported above (Corriveau , 2007). Thus, it is plausiblethat avoidant children have experienced a sufÞcientnumber of negative interactions for their approach andtrust to be undermined. In future research, it should befeasible to introduce preschool children to an unfamiliarperson, to vary the type of interaction that children havewith him or her and to measure the degree of selectivetrust that ensues. Preliminary evidence indicates that thistype of selective trust can emerge quite rapidly. Mascaro(2006) introduced 3-, 4-, and 5-year-olds to a nice animalpuppet that caressed the experimenter and a mean animalpuppet who hit him. When the two animal puppets 436Kathleen Corriveau and Paul L. Harris© 2009 The Authors. Journal compilation © 2009 Blackwell Publishing Ltd.subsequently made conßicting claims about the identityof a hidden object, all three age groups were more likelyto agree with the nice rather than the mean puppet. Insummary, of the four interpretations of the familiarityeffect observed in Experiments 1 and 2, the most plausible,current interpretation is that repeated, positive inter-action with a person renders him or her an attractivesource of information. Further research is certainlyneeded, however, to fully establish this conclusion.The second major Þnding is that older children weremore likely than younger children to alter their preferencefor a familiar informant following information about heraccuracy. This shift emerged to a similar extent for bothask and endorse probes. Such an age change could beattributed to an age change in childrenÕs assessment ofwhether an informant has made true or false claims Ð asindexed by their explicit judgments. Alternatively, it couldbe due to an age change in the tendency to extrapolatefrom Ð or weigh Ð such judgments of prior accuracyinformation when evaluating informantsÕ subsequentclaims. The analyses of childrenÕs explicit judgments pro-vide some preliminary support for the Þrst explanation.Recall that 5-year-olds were generally correct in theirexplicit judgment of the two informants across bothconditions and both sets of questions. By contrast, 3-and 4-year-olds were prone to error when the familiarinformant had been inaccurate, particularly when askedthe second set of Explicit Judgment questions. However,when post-test scores across the two experiments wereanalyzed with correct replies to the Þrst set of ExplicitJudgment questions entered as a covariate, the inter-action of Age and Condition still emerged. Thus, the agechange in sensitivity to informant accuracy as indexedby responses to the Ask and Endorse probes cannot beexplained solely in terms of Explicit Judgment scores.Further support for this conclusion comes from closeinspection of childrenÕs scores on the Þrst set of ExplicitJudgment questions Ð posed just prior to the four testtrials. These were quite high across both conditions andall three age groups.Given these considerations, we conclude that evenwhen children were able to note and explicitly commenton the differential accuracy of the informants, youngerchildren were less likely than older children to takeaccount of such accuracy information in assessing whichinformant to trust. We may consider two different inter-pretations of that age change. One possibility is thatyounger children rarely, if ever, take accuracy informationinto account. A second possibility is that younger childrendo take it into account but only for unfamiliar informants.Thus, in the case of familiar informants they are capableof registering and remembering inaccuracy, but they setit aside or ignore it when deciding whom to trust.When recent Þndings are considered, it is evident thateven 3-year-olds are able to register, remember, and beguided by inaccuracy in the case of unfamiliar informants,even if they are less prone to do so than older preschoolers(Birch , 2008; Pasquini , 2007). Indeed, 3-year-olds remember such accuracy information for severaldays and continue to take it into account when weighingwhom to trust (Corriveau & Harris, 2009). By implica-tion, the age change observed in the present study islikely due to a discounting process: like older preschoolers,younger preschoolers can register and rememberinaccuracy but unlike older preschoolers they are proneto discount or ignore such information if it familiar and ordinarily trustworthy informants. support for discounting by younger children emergedin the second set of Explicit Judgment questions. Recallthat 3- and 4-year-olds, unlike 5-year-olds, oftenjudged that the familiar informant had been good atanswering the questions even when she had consistentlymade mistakes.Two design features of Experiments 1 and 2 warrantdiscussion. First, it could be objected that children willrarely, if ever, encounter an informant who makes severalfalse claims in succession. However, it is important toemphasize that children will certainly hear claims thatthey know or discover to be false. For example, a caregivermay make a mistake about where a toy is located, whois calling on the telephone, or the name of the childÕsclassmate. Indeed, a caregiver may occasionally make aseries of apparently false claims. For example, havingmisidentiÞed which past episode the child is referring to,a caregiver may produce several claims in successionthat the child regards as mistaken. Experiments 1 and 2provided children with an exaggerated version of suchinaccuracy. Note that the presentation of such exagger-ated or Ôsuper-realÕ stimuli has proven a useful researchstrategy in ethology (Lorenz, 1981) even if it has rarelybeen used in developmental psychology. We make theworking assumption that children react to such consistentinaccuracy with a stronger form of the strategy that theyapply to less consistent inaccuracy. In line with thatassumption, recent Þndings indicate that 4-year-oldsshow greater mistrust of an informant who is often Ð butnot consistently Ð inaccurate as compared to an informantwho is only occasionally inaccurate (Pasquini The second design feature calling for discussion is thefact that the experimenter invited children to reßect onthe accuracy of the two informants. Thus, in bothexperiments, children were asked to say whether eachinformant was Ôvery good or not very goodÕ at answeringthe questions, and also to say which informant was better.Because the Þrst trio of Explicit Judgment questions wasasked before the Ask and Endorse probes, these ques-tions may have prompted children to attend to accuracyinformation when responding to the probes. Studies inwhich the Explicit Judgment questions are either includedor omitted offer a way to evaluate this possibility. In themeantime, there is emerging evidence that preschoolersdo monitor for accuracy even in the absence of ExplicitJudgment questions. Birch (2008) and ScoÞeld andMiller (2007) each report that preschoolers were morelikely to learn a new word from an informant who hadproved accurate rather than inaccurate in the past Choosing your informant437© 2009 The Authors. Journal compilation © 2009 Blackwell Publishing Ltd.even when children had not been explicitly asked tojudge the accuracy of the two informants. Similarly,Corriveau and Harris (2009) report that both 3- and 4-year-olds trusted an accurate informant more than aninaccurate informant over several days even in theabsence of explicit judgment questions.Finally, we may consider the broader implications ofthe Þndings for childrenÕs language acquisition andlearning about the world. In the last 20 years, a consider-able body of research has accumulated showing thatchildrenÕs acquisition of language is not divorced fromtheir skill at social cognition. In particular, when childrenlearn the meaning of new words, they make use of variousnon-verbal and verbal cues to interpret and weigh aspeakerÕs claims (Baldwin, 1993; Fusaro & Harris,2008; Jaswal, 2004; Sabbagh & Baldwin, 2001; Tomasello,Carpenter, Call, Behne & Moll, 2005). The present also demonstrate a link between childrenÕs social cogni-tion and their acquisition of new words. Nevertheless,the thrust of the present research is different. Apparently,young children use their social cognitive skills, not justto decode and weigh a speakerÕs current claims but alsoto decide whether to accept or reject the speakerÕs laterclaims. Thus, children use their social cognitive skillsto make a subsequent choice among their informantsand not just to interpret what an informant is currentlysaying.It seems likely that childrenÕs learning about other,non-linguistic aspects of the world is similarly guided bytheir selective trust in particular informants. In line withthis expectation, two recent studies have shown thatchildren use prior accuracy when choosing betweenconßicting claims about how to use a tool (Birch 2008; Koenig & Harris, 2005) or the properties of hidden, 2004). We anticipate that furtherresearch is likely to consolidate the conclusion thatchildrenÕs willingness to learn from others is tempered bycareful monitoring of whom they learn from.Ainsworth, M.D.S., Blehar, M.C., Waters, C., & Wall, S.Patterns of attachment. Hillsdale, NY: LawrenceErlbaum.Baldwin, D.A. (1993). InfantsÕ ability to consult the speakerfor clues to word reference. Journal of Child Language395Ð418.Birch, S., Vauthier, S., & Bloom, P. (2008). Three- and four-year-olds spontaneously use othersÕ past performance toClŽment, F., Koenig, M., & Harris, P.L. (2004). The ontogenesisof trust. Mind and LanguageCorriveau, K.H., & Harris, P.L. (2009). Preschoolers continueto trust a more accurate informant 1 week after exposure toaccuracy information. Developmental ScienceCorriveau, K.H., Harris, P.L., & Nelson, C. (2007, JChildrenÕs trust in testimony. Paper presented at the annualmeeting of the Jean Piaget Society, Amsterdam, NE.Fenson, L., Dale, P.S., Reznick, J.S., Bates, E., Thal, D.J., &Pethick, S.J. (1994). Variability in early communicativedevelopment. Monographs of the Society for Research inChild Development (5).Fusaro, M., & Harris, P.L. (2008). Children assess informantreliability using bystandersÕ non-verbal cues. DevelopmentalGergely, G., Egyed, K., & Kir‡ly, I. (2007). On pedagogy.Developmental ScienceHarris, P.L. (2007). Trust. Developmental Science, 135Ð138.Harris, P.L., Pasquini, E.S., Duke, S., Asscher, J.J., & Pons, F.(2006). Germs and angels: the role of testimony in youngchildrenÕs ontology. Developmental Science, 76Ð96.Hume, D. (1748/1957). An enquiry concerning human understand-. Oxford: Clarendon Press.Jaswal, V.K. (2004). DonÕt believe everything you hear: pre-schoolersÕ sensitivity to speaker intent in category induction.Child DevelopmentJaswal, V.K., & Neely, L.A. (2006). Adults donÕt always knowbest: preschoolers use past reliability over age when learningnew words. Psychological ScienceKoenig, M., ClŽment, F., & Harris, P.L. (2004). Trust in testi-mony: childrenÕs use of true and false statements. PsychologicalKoenig, M.A., & Harris, P.L. (2005). Preschoolers mistrustignorant and inaccurate speakers. Child DevelopmentLeichtman, M.D., & Ceci, S.J. (1995). The effects of stereotypesand suggestions on preschoolersÕ reports. DevelopmentalPsychologyLorenz, K.Z. (1981). The foundations of ethology. New York:Springer-Verlag.Mascaro, O. (2006). Benevolence and competence in childrenÕsadjustment of epistemic vigilance. Paper presented at theNASH workshop, Institut Jean Nicod, Paris.Pasquini, E.S., Corriveau, K.H., Koenig, M.A., & Harris, P.L.(2007). Preschoolers use past reliability in deciding whichinformant to trust. Developmental PsychologyStudentÐTeacher Relationship Scale Ð ShortForm. Lutz, FL: Psychological Assessment Resources, Inc.Sabbagh, M.A., & Baldwin, D.A. (2001). Learning words fromknowledgeable versus ignorant speakers: links betweenpreschoolersÕ theory of mind and semantic development.Child Development, 1054Ð1070.ScoÞeld, J., & Miller, A. (2007). How speaker accuracy effectsword learning. Poster presented at the biannual meeting of theSociety for Research in Child Development, Boston, MA.Tomasello, M., Carpenter, M., Call, J., Behne, T., & Moll, H.Understanding and sharing intentions: the origins of cul-tural cognition. Behavioral and Brain SciencesZajonc, R.B. (1980). Feeling and thinking: preferences need noinferences. American PsychologistZajonc, R.B. (2001). Mere exposure: a gateway to the subliminal.Current Directions in Psychological ScienceReceived: 19 May 2006Accepted: 11 April 2008