Children’s Beliefs About Everyday Reasoning

Jennifer AmsterlawUniversity of Washington

Two studies investigated children’s metacognition about everyday reasoning, assessing how they distinguishreasoning from nonreasoning and ‘‘good’’ reasoning from ‘‘bad.’’ In Study 1, 80 1st graders (6 – 7 years), 3rdgraders (8 – 9 years), 5th graders (10 – 11 years), and adults (181 years) evaluated scenarios where people (a)used reasoning, (b) solved problems with nonreasoning approaches, or (c) reacted appropriately but auto-matically to events. All age groups distinguished reasoning from type (b) nonreasoning cases, but age-relatedimprovement occurred for type (c) cases. Study 2 tested 160 1st, 3rd, 5th graders’ and adults’ evaluation of goodand bad reasoning processes, finding 2 developmental changes: initial improvement in discriminating thinkingprocesses by 3rd grade, and emergence of an adult-like, process-focused (vs. outcome-focused) concept ofthinking quality by 5th grade.

Children’s progress toward more sophisticated un-derstandings of the thinking mind is described by alarge and varied literature that includes research onchildren’s early ‘‘theories of mind’’ (Flavell & Miller,1998; Wellman, 2002), their beliefs about the natureand sources of knowledge, or epistemology (Mont-gomery, 1992; Pillow, 1999), and their metacognitiveknowledge about the goals, strategies, and demandsof various cognitive tasks (Kuhn, 2000b). The currentresearch focuses on children’s knowledge abouteveryday reasoning, an important domain of meta-cognitive development that has not been studiedpreviously.

Reasoning can be broadly defined as ‘‘mental ac-tivity that consists of transforming given informationin order to reach conclusions’’ (Galotti, 1989, p. 335).It is a staple of our everyday cognition. Planning,evaluating arguments, making decisions, drawinginferences, testing hypothesis, and solving logicpuzzles all involve reasoning. Understandably, agreat deal of research in cognitive science has been

devoted to characterizing and explaining humanreasoning processes (for reviews, see Baron, 2000;Galotti, 1989), tracking their development over thelife span (e.g., DeLoache, Miller, & Pierroutsakos,1998; Jacobs & Klaczynski, 2002; Moshman, 1998),and attempting to improve people’s reasoning skills(e.g., Baron & Sternberg, 1987).

Research on children’s knowledge about reason-ing has the potential to inform our understanding ofboth theory of mind development and reasoningdevelopment. From a theory of mind perspective,more detailed information about the global organi-zation of children’s concepts of thinking processes,and about the development of their ‘‘specific theo-ries’’ (Wellman, 1990) of particular mental functions,such as reasoning, is necessary to adequately de-scribe development. As Flavell, Green, and Flavell(1995a) note, a focus on children’s knowledge aboutmental states (e.g., beliefs, desires, intentions) has leftimportant questions concerning children’s knowl-edge about mental processes and activities (e.g.,learning, reasoning, paying attention) unanswered.This is especially true for the older child’s theory ofmind, as developments beyond the preschool yearsare not as well studied. According to Wellman andGelman (1998), theory of mind development in-volves both integration and differentiation of keyconcepts. Thus, the organization of children’s con-cepts of mental activity can reveal important quali-tative changes in their theories of mind: for example,by reflecting older children’s emerging awareness ofthe mind’s constructive and interpretive functions(Schwanenflugel, Fabricius, & Alexander, 1994). Ifdevelopments observed in children’s concepts of

r 2006 by the Society for Research in Child Development, Inc.All rights reserved. 0009-3920/2006/7702-0014

This research served as part of the author’s doctoral dissertationin psychology at the University of Michigan, Ann Arbor.

Portions of the data were presented at the biennial meeting ofthe Cognitive Development Society, October 2003, in Park City, UT,and at the biennial meeting of the Society for Research in ChildDevelopment, April 2005, in Atlanta, GA.

Support during research and writing was provided by NIHgrant HD-22149 and NSF grant SBE-0354453.

I would like to thank the schools, parents, and children whoparticipated in the studies, and Sandra Iler and Melanie Stevensonfor assistance with data collection and coding. I am grateful toHenry Wellman and Andy Meltzoff for helpful comments on thepaper.

Correspondence concerning this article should be addressed toJennifer Amsterlaw, Institute for Learning and Brain Sciences,University of Washington, Box 357988, Seattle, WA 98195. Elec-tronic mail may be sent to amsterj@u.washington.edu.

Child Development, March/April 2006, Volume 77, Number 2, Pages 443 – 464

reasoning parallel those in other areas of theory ofmind, this suggests a coherent system of conceptualchange.

From the perspective of reasoning development,some researchers hypothesize a link between meta-cognitive knowledge and the development of rea-soning skillsFan idea that reflects the thrust ofearlier metamemory research (for a review, see Joy-ner & Kurtz-Costes, 1997). Moshman (1994, 1998), forexample, has argued that children’s ability to reasonaccording to the norms of rational thought dependsin large part upon their skills of ‘‘metareasoning,’’which includes knowledge about logical principlesand beliefs about appropriate reasoning strategies.Kuhn (2000a, 2000b; Kuhn, Katz, & Dean, 2004) hassimilarly posited a strong relationship betweenchildren’s emerging metacognitive knowledgeand the development of strategic reasoning duringmiddle childhood, calling this metalevel ‘‘thelocus of developmental change’’ Kuhn (2000a, p.179). Although there is preliminary evidence thatmetacognition and reasoning are related develop-mentally (e.g., Kuhn & Pearsall, 1998), research di-rectly investigating developmental changes inchildren’s knowledge about reasoning has beensparse. Such evidence will be necessary to refine andevaluate metacognitive accounts of reasoning de-velopment.

The goal of the current research was to investigatetwo aspects of developmental change in children’sknowledge about everyday reasoning: (1) how chil-dren distinguish reasoning from other kinds ofthinking and (2) how they differentiate ‘‘good’’ rea-soning from ‘‘bad’’ reasoning.

Reasoning Versus Nonreasoning

The first research question concerns children’s fun-damental concept of reasoningFthat is, how theyunderstand reasoning as a mental activity. Being ableto recognize instances of reasoning and distinguishthem from other kinds of thinking constitute someof the most basic knowledge about reasoning chil-dren might be said to possess. Earlier work withU.S. college-age adults (Amsterlaw, in preparation)shows that when given descriptions of peoplethinking in everyday situations, adults draw cleardistinctions between cases that involve reasoning(e.g., weighing the pros and cons of a choice, infor-mal hypothesis testing) and various nonreasoningcases, such as using shortcut strategies (e.g., flippinga coin), responding to events automatically (e.g.,pulling one’s hand from a hot stove), or engaging inpurposeful thinking of some other kind (e.g., learn-

ing from a book, rehearsing a grocery list). Foradults, reasoning/nonreasoning distinctions arerooted in judgments about underlying cognitiveprocesses; reasoning cases are regarded as involvingmore thinking, mental effort, and logic, more use ofstrategies, clearer goals, and less automatic re-sponding than nonreasoning cases. Study 1 asks: Doschool-age children use these same process featuresto distinguish between reasoning and nonreasoning?

It may be that even young children distinguishreasoning from nonreasoning as adults do. Prior re-search has found that children aged 5 – 7 years al-ready possess important insights about complexcognitive activities such as memory (e.g., Kreutzer,Leonard, & Flavell, 1975). Studies of children’s un-derstanding of inference, a concept closely related toreasoning, suggest that children first recognize in-ference as a source of knowledge at around 5 or 6years of age (Gopnik & Graf, 1988; O’Neill & Gopnik,1991; Sodian & Wimmer, 1987). Thus, we might ex-pect children to have basic knowledge about rea-soning by this point. However, other researchindicates that key metacognitive developments occuronly later in childhood. For example, developmentalresearch on children’s and adults’ concepts of mentalactivities (e.g., Schwanenflugel et al., 1994;Schwanenflugel, Henderson, & Fabricius, 1998) findsthat children as old as 8 – 10 years do not use infer-ence or constructive processing as organizing di-mensions in their categories of mental activities asadults do. Although children respect similarities inmental activities based on memory and attention(e.g., linking ‘‘making a list at home of all the kids inyour new class’’ with ‘‘saying Happy Birthday on theright day to your friend’’), they do not respect un-derlying similarities in constructive mental processeslike inferring (e.g., ‘‘seeing a puddle on the groundand realizing it must have rained last night’’),guessing (e.g., ‘‘filling in a question when you don’tknow the answer’’), and reasoning (e.g., ‘‘pointingout why joining the club is better than not joiningit’’). More generally, in research with 5 – 13-year-olds,Flavell and his colleagues report late-emergingcomprehension of phenomena such as the control-lability of mental states (Flavell & Green, 1999;Flavell, Green, & Flavell, 1998), attentional focus(Flavell et al., 1995a), and distinctions between con-scious and unconscious mentation (Flavell, Green,Flavell, & Lin, 1999). Developmental differences inchildren’s and adults’ ontologies of mental func-tioning may well be reflected in their conceptsof reasoning, with younger children showingless differentiated categories of reasoning and non-reasoning.

444 Amsterlaw

‘‘Good’’ Reasoning Versus ‘‘Bad’’ Reasoning

The second question deals with children’s ability todistinguish between good and bad reasoning: Givena description of a person’s thinking process, canchildren appropriately assess its quality? This ques-tion targets children’s beliefs about the norms andstandards of rational thought, the kind of metacog-nitive knowledge Moshman (1994, 1998) describes asclosely tied to reasoning development. The issue ofwhat constitutes ‘‘good thinking’’ has been a con-tentious one (for a discussion, see Stanovich & West,2000, and replies). In this work, I draw on distinc-tions given by Baron (2000), Perkins, Jay, and Tish-man (1993b), and Kuhn (1996), among others, whoemphasize that good thinking is typically strategic,thorough, and unbiased. Previous research on adults’beliefs about thinking (Amsterlaw, in preparation;Lizotte & Amsterlaw, 2001) has found similar themesin their characterizations of good versus bad rea-soning: Good reasoning involves a search for infor-mation or evidence to draw conclusions, theconsideration of multiple viewpoints, being alert tobiases, and high amounts of thinking and mentaleffort, whereas bad reasoning involves automaticand emotional responses, incomplete or biased ap-praisals of information, and minimal thinking,strategy, or effort. Study 2 asks whether children alsouse these criteria to distinguish between good andbad reasoning.

Prior research in this area has focused on chil-dren’s metacognition in formal reasoning situations,such as solving logic problems (e.g., Moshman &Franks, 1986) or testing scientific hypotheses (e.g.,Kuhn & Pearsall, 1998). A small literature on chil-dren’s acquisition of ‘‘metalogic,’’ or explicitknowledge of logical reasoning principles, such asinferential validity and logical coherence (see, e.g.,Morris, 2000; Moshman & Franks, 1986; Tunmer,Nesdale, & Pratt, 1983), indicates that the ability tomeaningfully discriminate logical and illogical formsof argument emerges relatively late in childhood,around the age of 10 years. However, this researchhas not fully considered what children might knowabout good and bad thinking in everyday contextswith which they are more familiar. There appears tohave been only one study directly investigatingchildren’s awareness of flaws in everyday thinking(Perkins, Tishman, Richart, Donis, & Andrade, 2000,Study 2). In that study, sixth graders read stories thatdescribed people showing two common shortcom-ings in thinkingFneglecting alternative options andfailing to seek reasons on both sides of a caseFandhad to describe any problems they noticed. In an-

other condition, problems were highlighted ahead oftime and children generated solutions to them.Perkins et al. found that children’s ability to detectflawed thinking was quite low; however, they werebetter at generating solutions. These results implythat distinguishing between good and bad reasoningcan be difficult even for older children. Yet becausethe study did not examine developmental effects, wedo not know whether younger children would haveperformed still more poorly, or whether adultsFwho may also ignore flaws in thinking (see Jacobs &Klaczynski, 2002)Fwould have fared any better.The current work compares the performance of chil-dren at three ages, as well as adults, on multiplejudgments about thinking.

A related question is whether children can dis-tinguish a good thinking process from simply gettinga good outcome. In general, good thinking processesare expected to yield good outcomes (and badprocesses, bad outcomes). Yet this is not always so. Ifa person’s initial information is incorrect or incom-pleteFnot an unusual event in many real-life situa-tionsFany well-intentioned efforts to determine thebest solution may fall short. Similarly, correct solu-tions can be obtained fortuitously, not necessarilyas a result of good thinking. Adult judgmentsof thinking quality in such cases tend to prioritizethinking processes over final outcomes (e.g., badthinking that results in a good outcome is still badthinking)Falthough outcomes do bias ratings to-ward the valence of the outcome (e.g., Baron &Hershey, 1988). Outcome biases have not been in-vestigated in young children previously (but seeKlaczynski, 2001, for evidence that the bias declinesfrom early to middle adolescence). Conceivably,young children who are naıve to process-baseddistinctions in reasoning quality will show extremeoutcome biases in their evaluations of thinking.A second condition of Study 2 evaluates thishypothesis.

Study 1: Development of the Reasoning/Nonreasoning Distinction

Study 1 investigated children’s ability to makeprocess-based distinctions between reasoning andnonreasoning cases. Specifically, two kinds of non-reasoning cases were assessed. The first involvesinstances of shortcut problem solvingFcases where aperson uses a shortcut nonreasoning approach tosolve a problem that is typically solved with rea-soning (e.g., flipping a coin to make a decision). Thesecond involves instances of automatic respondingFcases where a person responds appropriately to

Beliefs about Everyday Reasoning 445

some stimulus in the environment, but does so uncon-sciously, without any real thinking at all (e.g., removingone’s hand upon touching a hot stove). Both cases areespecially clear examples of nonreasoning in adults, butthey offer interesting test cases for children because theycontain situational featuresFthe presence of a problemto solve, generating an appropriate responseFthat areoften associated with reasoning. Young children withless developed knowledge about thinking may in factassume a reasoning response is occurring just because aperson needs to solve a problem or because they re-spond to an event appropriately. Such a focus on salientcharacteristic features, in advance of appropriate atten-tion to the defining features that reliably determinecategory membership, is often reported in conceptualdevelopment research (Keil, 1989).

‘‘Reasoning’’ itself is not a term that children can beexpected to know. Schwanenflugel et al.’s (1998) re-search on cognitive verb extensions suggests thatthird and fifth graders do not apply the verb ‘‘reason’’in the same way adults do. This poses obvious chal-lenges for studying children’s ideas about the cogni-tive process the term references. One way to judgewhether children’s concepts of reasoning are likethose of adults, however, is to examine whether theymake similar distinctions over key features of theunderlying thinking process. In this study, questionsabout how much and how hard people are thinkingprovide measures of children’s sensitivity to processfeatures that distinguish reasoning from nonreason-ing, although the term ‘‘reasoning’’ itself is not used.To check children’s understanding of the other cog-nitive terms that are used, a separate measure ofchildren’s cognitive word knowledge is included.

Method

Participants

A total of 60 children (20 in each of first, third, andfifth grades) and 20 adults from a small Midwestern cityparticipated. There were 11 females and 9 males fromthe first grade (M age 5 6.5 years, SD 5 0.4); 9 femalesand 11 males from the third grade (M age 5 8.8 years,SD 5 0.6); 11 females and 9 males from the fifth grade(M age 5 10.7 years, SD 5 0.3); and 10 female and 10male adults (M age 5 19.3 years, SD 5 2.1). Childrenwere recruited from public elementary schools throughletters to parents ( � 45% acceptance rate). Adult par-ticipants were college students who participated tofulfill an Introductory Psychology course requirement.The final sample was approximately 65% Caucasian,13% African American, 9% Hispanic, 5% Asian, and 8%other or unknown backgrounds. No direct measures of

socioeconomic status were obtained, but the popula-tions were generally middle to upper-middle class.

Measures and Procedures

Children were tested individually in a quiet areaof their school, receiving the reasoning/not-reason-ing task followed by the cognitive word task. Ses-sions lasted about 20 min and were audio-taped.Adults received the tasks as a written questionnaire.

Reasoning/not-reasoning task. Participants receivednine scenarios portraying everyday thinking eventsthat participants of all ages were likely to under-stand. Scenario order was randomized within agegroups (see the Appendix for scenarios). There werethree types of scenarios: (1) reasoning scenarios de-scribed cases where the character faced a problemthat required reasoning and used reasoning to solveit (identifying a secret admirer, figuring out whyplants are wilting, deciding which bike to buy); (2)shortcut problem-solving scenarios described caseswhere the character faced a problem that typicallyrequired reasoning to solve, but he or she solved it insome other way (setting an alarm clock by trial-and-error, deciding which summer camp to go to byflipping a coin, simply recalling the answer to a mathproblem); and (3) automatic action scenarios de-scribed cases where the character responded to astimulus in the environment appropriately but au-tomatically, without thinking (removing hand from ahot stove, jumping away from an oncoming truck,eating some appetizing cookies).

For children, a 500 � 800 drawing of a child’s facewas shown with each scenario. The researcher readthe scenarios aloud and asked children to respond toa series of questions for each one: (1) Was X think-ingFyes or no?; (2) How much was he or shethinkingFa whole lot, a medium amount, or notreally thinking at all?; (3) How hard was he or shethinkingFvery hard, kind of hard, or not hard atall?; (4) How long did it take him or herFa longtime, a medium amount of time, or no time at all?; (5)In that story, did X have some kind of problem tofigure out? Did he or she really have one, kind ofhave one, or not have one at all?; and (6) How smartwas what he or she didFvery smart, kind of smart,or not smart at all? In addition, directly after ratingthe amount of thinking in each scenario, participantswere asked to explain their ratings (e.g., ‘‘Why didyou say he was thinking a lot?’’). Questions aboutamount of thinking and mental effort were usedbecause these are dimensions on which adultsstrongly distinguish between reasoning and non-reasoning cases, and because they reference concepts

446 Amsterlaw

that even young children understand. Additionally,a question about length of time tested whether chil-dren might understand that reasoning typically takeslonger than nonreasoning, even if they could notdistinguish them by thinking processes per se. Thefinal two questions tested whether children distin-guished the cases on the basis of two salient non-process featuresFpresence of a problem, andsmartness of the response. It was expected that rea-soning cases would receive high ratings on both ofthese dimensions. Explanations yielded additionaldata for interpreting developmental findings.

An 800 � 1000 drawing showing three circles of in-creasing size was used to help children respond toquestions along a 3-point scale (scored 0 – 2 foranalyses). The researcher explained the scale bytelling children to point to the small circle to say notvery much or none at all, to the medium circle to say amedium amount, and to the large circle to say a wholelot. For the first few trials (and then on any subse-quent trials where children did not readily respondto questions), the researcher pointed to and labeledeach circle with its corresponding response option.To control for possible order effects, the options weregiven in reverse order for half the children in eachgroup. Two warm-up items trained children how touse the scale. The researcher first asked children toimagine that she had eaten 100 hamburgers for lunchyesterday, asking ‘‘How hungry was IFvery hun-gry, kind of hungry, or not hungry at all?’’ and en-couraged them to use the scale to respond. She thenhad them imagine that she had eaten only one pieceof lettuce, again asking how hungry she was. Noneof the children had difficulty with the warm-upitems.

Cognitive word task. This task, based on Astingtonand Olson (1990), tested children’s knowledge ofwords describing various cognitive activities, both toensure comprehension of key terms used in thereasoning/not-reasoning task and to provide a moreglobal measure of metacognitive knowledge. Chil-dren were shown a drawing of a girl, ‘‘Sarah,’’ andwere asked to hear about ‘‘some different ways Sarahcan use her mind’’ and to say ‘‘which way you thinkshe is using it.’’ There were 12 multiple-choice itemsthat the researcher read aloud to the child, eachconsisting of a brief description of a cognitive ac-tivity followed by three terms (one target term andtwo distractors) from which the child could select(see the Appendix for items). The task began with aneasy warm-up item (listen) to introduce the format.In the few cases where children failed the warm-upitem, it was repeated. No child needed it repeatedmore than once.

On the basis of research documenting children’sacquisition of cognitive words (see Astington & Ol-son, 1990; Booth & Hall, 1994, 1995), the 12 itemsreflected a range of difficulty levelsFsome easy foreven the youngest children (e.g., remember), somedifficult for even the oldest children (e.g., interpret),and the rest intermediate. Several items (remember,compare, wonder, and decide) were used in the rea-soning/not-reasoning task itself. Distractor termswere chosen so that one term was somewhat similarto the target term and the other was quite dissimilar(e.g., distractor terms for conclude were predict andwish). The position of the target term in the choice setwas counterbalanced across items. Item order wasrandomized within age groups.

Explanation coding. Participants’ explanations fortheir amount of thinking ratings in the reasoning/not-reasoning task were coded for whether they cit-ed target features of the characters’ thinking processor other aspects of the scenario. For reasoning sce-narios, responses were credited with citing targetthinking processes if they explicitly mentioned fea-tures of the cognitive process that was used (e.g.,‘‘because he wrote down everything and comparedit’’), or otherwise expressed that the person wasthinking hard to solve the problem (e.g., ‘‘becauseshe really wanted to figure it out’’ or ‘‘he didn’tknow what it was, so he really had to think’’). Forshortcut problem-solving cases, credit was given forresponses that downplayed the characters’ thinking,emphasizing that they were just pressing buttons,flipping a coin, or remembering the answer (e.g.,‘‘the coin was doing the thinking’’; ‘‘she was justpressing random buttons’’; ‘‘he was using memory,not thinking’’). For the memory scenario, it alsoseemed appropriate to give credit for stating thatremembering inherently involves thinking (e.g., ‘‘hehad to think to remember’’). For automatic actionscenarios, credit was given for responses that men-tioned instincts, automatic responding, or describedcharacters as acting without thinking (e.g., ‘‘he justsaw the cookies and went yum’’). Other responsesdid not receive credit. Two independent raters codedthe data using written transcripts that concealedparticipants’ age, race, and gender. Inter-rateragreement on 20% of the data was 92%.

Results

Reasoning/Not-Reasoning Task

Analyses tested children’s ability to make the twokey contrasts between reasoning and shortcut prob-lem solving and between reasoning and automatic

Beliefs about Everyday Reasoning 447

action. In each case, overall age effects wereexamined with a series of 4 (grade) � 2 (scenariotype: reasoning or nonreasoning) ANOVAs on eachof the five rating dimensions (amount of thinking,mental effort, length of time, problem, and smart-ness). Scenario type was a within-subjects factor. Inaddition, planned paired t tests assessed whetherindividual age groups successfully made the con-trasts.

There were no significant effects of gender orquestion order (i.e., whether the response scale waspresented as increasing or decreasing), and no sig-nificant interactions with grade, on task perfor-mance. Thus, all analyses collapse across gender andquestion order. Preliminary analyses found that rat-ings for one automatic action scenario, describing aboy who impulsively eats some cookies, differedsignificantly from ratings for the other two, whichdescribed people reacting to dangerous situations.Child (but not adult) participants rated the cookiescenario significantly lower on amount of thinking,M 5 0.23, SD 5 0.54, t(59) 5 9.98, mental effort,M 5 0.28, SD 5 0.58, t(59) 5 8.29, and smartness,M 5 0.23, SD 5 0.53, t(59) 5 18.27, all pso.001. Chil-dren’s explanations revealed that many thought theboy was misbehaving, an interpretation that wasunintended. This scenario was therefore excludedfrom subsequent analyses.

Figure 1 displays participants’ dimension ratingsby scenario type and grade. (Note that the maximumscore on each dimension is 2.) Responses to the initialquestion for each scenario (asking simply whetherthe character was thinking or not) were redundant

with the amount of thinking ratings and were notanalyzed separately.

Reasoning versus shortcut problem-solving. Asshown in Figure 1, all age groups successfully dis-tinguished between reasoning and shortcut problemsolving in their ratings. Main effects for scenario typewere found for amount of thinking, F(1, 76) 5 244.6;mental effort, F(1, 76) 5 231.3; length of time,F(1, 76) 5 137.6; presence of a problem, F(1, 76) 5

53.2; and smartness of response, F(1, 76) 5 221.5, allpso.001, with reasoning cases receiving significantlyhigher ratings on all measures. Paired t tests con-firmed that these effects held for all age groups,pso.05. Significant Grade � Scenario Type interac-tions were found only for length of time,F(3, 76) 5 4.12, po.01, and smartness judgments,F(3, 76) 5 2.96, po.05. In both cases, these simplyreflected an increasing divergence of ratings withage. To reiterate, even first graders significantlydistinguished between reasoning and shortcutproblem solving on these dimensions.

Reasoning versus automatic action. In contrast, therewere clear age differences in distinguishing reason-ing from automatic action. Specifically, SignificantGrade � Scenario Type interactions were found forboth amount of thinking, F(3, 76) 5 7.77, and mentaleffort, F(3, 76) 5 9.97, pso.001. Paired t tests revealedthat whereas older children and adults rated rea-soning cases as involving more thinking and mentaleffort than automatic action cases, first graders ratedthem as involving similarly high amounts of think-ing, t(19) 5 1.22, and mental effort, t(19) 5 0.62,ps4.05. By third grade, children appropriately rated

0

1

2

1 5 Adult

Reasoning

Automatic Action

Shortcut Problem-Solving

0

1

2

1 5 Adult

0

1

2

1 5 Adult0

1

2

Rat

ing

Rat

ing

Mental Effort

0

1

2

1 5 Adult

Length of Time Presence of Problem Smartness

Amount of Thinking

Grade Grade

Grade Grade Grade

3 3

3 3 1 5 Adult3

Figure 1. Dimension ratings for reasoning/not-reasoning task scenarios, Study 1. Note. Error bars are standard errors of the means.

448 Amsterlaw

automatic action cases lower on both amount ofthinking, t(19) 5 2.86, and mental effort, t(19) 5 3.12,pso.05.

There was a Significant Grade � Scenario Typeinteraction for length of time ratings; older ages re-ported larger time differences between scenariotypes, F(3, 76) 5 6.83, po.001. However, even firstgraders appropriately rated reasoning cases as in-volving more time, t(19) 5 3.37, po.01. Thus, firstgraders could distinguish the cases by length of time,although not by amount of thinking or mental effort,a finding consistent with the expectation that chil-dren would more readily distinguish the cases byovert behavioral features.

For problem and smartness judgments there wereno significant interactions with grade. However,paired t tests indicated that although all age groupsgave higher problem ratings to reasoning cases thanto automatic action cases, ts(19)43.05, pso.01, thisdifference did not reach significance for first graders,t(19) 5 1.74. Also, while first graders showed theadult pattern of rating both reasoning and automaticaction scenarios high on smartness, their ratings forautomatic action cases were significantly higher thanthose for reasoning, t(19) 5 2.57, po.05, whereasolder groups’ ratings did not differ.

In sum, as expected, adults and older childrencharacterized reasoning cases as involving extended,effortful thinking, the presence of problems to solve,and smart responses. In most instances, they ratednonreasoning cases much lower on these dimen-sions. (Automatic action cases were also viewed asinvolving smart responses, albeit ones achievedwithout much thinking.) Even the youngest childrenendorsed similar ideas about key contrasts betweenreasoning and nonreasoning, as they successfullydistinguished between reasoning and shortcutproblem-solving cases. Notably, children departedfrom the adult pattern not by mischaracterizingreasoning, but by overgeneralizing its cognitive fea-tures to automatic actions.

Explanations for amount of thinking judg-ments. Participants’ explanations were coded forwhether or not they referenced target thinkingprocess features described in the scenario. Table 1reports the percentage of process-based explanationsin each age group. As shown in the table, referencesto target process features increased with age for allthree scenario types.

For reasoning cases, most older children gave ap-propriate process-based explanations for their rat-ings, although only about half of first graders did.The rest gave various other responses, includingoverly vague explanations (e.g., ‘‘because he was

thinking a little,’’ ‘‘he was only thinking aboutbikes’’) and references to other story events (e.g., ‘‘hisparents should have been there’’), with no consistentpattern to these responses. (One exception was thewilting plant story; some children said the boy wasthinking a lot because he was helping his plants.)Similar results held for the shortcut problem-solvingcases, where about half of the first graders reportedthat the characters were actively trying to solve theirproblems (e.g., ‘‘because she had to figure out whattime’’ or ‘‘he couldn’t decide which one to go to so heflipped a coin’’).

For automatic action cases, none of the first grad-ers referred to instincts, automatic responding, ormentioned that the characters acted quickly withoutthinking. Such responses were more frequent amongthird and fifth graders, but were common onlyamong adults. Most children instead focused on theidea that the characters’ actions were smart or nec-essary (e.g., ‘‘if she didn’t move her hand, she wouldget burnt’’). Some even described the characters ashaving such thoughts before acting. Adults rarelygave high ratings to these cases, although some be-lieved that thinking of some kind had occurred (e.g.,‘‘she had to process how close she was to the truck’’).To illustrate how frequently younger children refer-enced the smartness of people’s actions as evidenceof thinking, across the two automatic action stories,roughly 95% of first graders, 68% of third graders,60% of fifth graders, but only 8% of adults’ expla-nations mentioned the smartness or necessity ofpeople’s actions. Thus, young children’s tendency toreport high amounts of thinking and mental effort inautomatic action cases was closely related to theirbelief that these actions were smart or appropriate tothe situation.

Table 1

Participants’ References to Target Thinking Process Features in

‘‘Amount of Thinking’’ Explanations, Study 1

Scenario type

Grade

1 3 5 Adult

Reasoning 55 95�� 95 100

Shortcut problem solving 55 80w 95 90

Automatic action 0 35�� 45 90��

Note. Values represent the percentage of participants referring totarget thinking processes features in explanations for at least twoout of three scenarios per type. Because one automatic actionscenario was dropped from analyses, values for automatic actioncases reflect the percentage of participants who referred to targetthinking processes for at least one of two total scenarios. Signifi-cance was determined by w2(1, N 5 40) against the next youngerage group.wpo.10. ��po.01.

Beliefs about Everyday Reasoning 449

Cognitive Word Task

The cognitive word task measured children’s un-derstanding of specific cognitive words used in thereasoning/not-reasoning task (decide, remember,wonder, and compare) as well as additional words ofvarying difficulty. Initial analyses showed no sig-nificant main effects for gender, and no significantinteractions with grade, on cognitive word taskperformance. As expected, performance improvedwith age, F(3, 76) 5 35.49, po.001. On average, firstgraders got 7.1 of 12 items correct (SD 5 2.5), thirdgraders got 9.5 items correct (SD 5 1.6), fifth gradersgot 11.1 items correct (SD 5 1.0), and adults got 11.8items correct (SD 5 0.4).

Table 2 shows performance on individual items bygrade. Note that even first graders correctly under-stood most of the terms from the reasoning/not-reasoning task. (Some first graders had difficultywith compare, which was used in the ‘‘bike’’ reason-ing scenario. However, children did not show lowerperformance on this scenario relative to others.)Nearly all third and fifth graders understood all theterms from the task, as well as additional, moredifficult items. Thus, it seems unlikely that devel-opmental differences on the reasoning/not-reason-ing task were due to younger children’s difficultywith any of the specific cognitive terms used.

For children, better performance on the cognitiveword task was correlated with greater differentiationof reasoning and nonreasoning on the reasoning/not-reasoning task (where amount of differentiation

was calculated by subtracting children’s mean‘‘amount of thinking’’ ratings for nonreasoning casesfrom their mean ratings for reasoning cases),r(60) 5 .34, po.01. Age was also correlated with per-formance on both the cognitive word task r(60) 5 .70,and the reasoning/not-reasoning task, r(60) 5 .33.Linear regression results indicated that age alone ac-counted for roughly 10% of the variance in reason-ing/not-reasoning task performance, F(1, 58) 5 7.13,po.01, and cognitive word task performance did notsignificantly increase the model fit, F(1, 57) 5 1.60,p4.05. Thus, development of the reasoning/nonrea-soning distinction was generally associated withmetacognitive developments assessed by the cogni-tive word task; however, performance on the reason-ing/not-reasoning task was not directly attributableto differences in children’s metacognitive wordknowledge, after controlling for age.

Discussion

Study 1 found developmental differences in chil-dren’s ability to distinguish reasoning from nonrea-soning mental activities. Specifically, first gradersdid not respect a basic reasoning/nonreasoningdistinction that is critical to adultsFthat reasoning(i.e., deliberate, effortful thinking) does not occurduring quick, automatic responding. Not until thirdgrade did children begin to recognize this. Evidencefrom children’s explanations suggested that a pri-mary issue was younger children’s confusion of ap-propriate responding with extended, effortfulthinking: young children cited the smartness ofcharacters’ actions to justify statements that charac-ters were thinking a lot. (Also suggestive is the factthat the excluded cookie story, which childrenviewed as a case of misbehavior, and therefore notsmart, was the only automatic action case rated lowon thinking and mental effort.) These results indicatequalitative differences in children’s understandingof when thinking is reasoning-like.

Children did successfully distinguish reasoningfrom shortcut problem solving, and did not assumethat people were reasoning whenever they facedproblems to solve. Thus, even the first graders ac-knowledged some basic distinctions between rea-soning and nonreasoning. Previously, Flavell, Green,and Flavell (1993, 1995b) found that first-gradechildren readily attribute ongoing thinking to peoplepresented with problems to solve (although they areless likely to do so when people lack an obvioustarget for cognition). The current results furtherdemonstrate that children are aware of importantthinking process features that underlie problem

Table 2

Performance on Cognitive Word Task Items by Grade, Study 1

Item

Grade

1 3 5 Adult

Decide 90� 100� 100� 100�

Figure out 80� 90� 95� 95�

Guess 80� 95� 95� 100�

Remember 75� 95� 100� 100�

Discover 75� 95� 100� 100�

Wonder 70� 85� 100� 95�

Understand 65� 95� 90� 100�

Compare 60� 85� 100� 100�

Get a hunch 40 80� 100� 100�

Predict 25 75� 90� 90�

Conclude 30 15 65� 100�

Interpret 15 45 70� 100�

Note. Values represent percentages of correct responses by grade(N 5 20). Percentages marked with an asterisk are significantlyabove the chance expectation of 33% according to one-tailed bi-nomial test.�po.05.

450 Amsterlaw

solving (e.g., mental effort, length of time) and canuse these features to distinguish various thinkingstrategies. In particular, children knew that reason-ing requires more mental activity than flipping acoin, using random trial-and-error, or recalling an-swers from memory. These results, along with thosefrom Flavell et al., indicate that children’s knowledgeabout problem solving is considerably more devel-oped than their knowledge about uncontrolled,unconscious mental functioning (e.g., stream-of-consciousness, reflexes)Fperhaps because thinkingduring problem solving involves more consciouscontrol and is more available to introspection.

It is also noteworthy that first graders associatedmore and more effortful thinking with smarter re-sponses, as prior work has shown that young chil-dren link thinking effort with task success (e.g.,Heyman, Gee, & Giles, 2003). Further work is neededto clarify the connections young children are makingbetween extended thinking, smart responses, and‘‘good’’ behavior. For example, in Study 1 it is un-clear whether children viewed extended thinkingand appropriate behavior as causally related or asmerely associated by moral valence. In addition, amore stringent test of children’s understandingmight have controlled for the effects of story content;in the current study, specific stories and the prob-lems they described covaried with scenario type.This raises the possibility that children were re-sponding to story elements other than the targetthinking process features. However, including threedifferent stories for each scenario type helps to allayconcerns about the effects any one particular storymight have had on the final results.

The present age effects are generally consistentwith accounts of later theory of mind developmentgiven by Flavell et al. (1999) and by Schwanenflugelet al. (1994, 1998), which describe increasing differ-entiation in children’s ontologies of conscious/un-conscious and constructive/nonconstructive mentalprocessing during middle childhood. That task suc-cess was associated with a measure of metacognitiveknowledge also suggests that the reasoning/non-reasoning distinction is connected to broader devel-opments in theory of mind, although it must benoted that the effects of metacognitive knowledgein the current study were not separable from thoseof age.

Study 2: Development of Beliefs about ‘‘Good’’ and‘‘Bad’’ Reasoning

Study 2 assessed developmental changes in chil-dren’s ability to distinguish good reasoning pro-

cesses from bad reasoning processes. Performancewas assessed under two conditions: (1) when sce-narios included only information about thinkingprocess and no information about outcome and (2)when scenarios included information about bothprocess and outcome, but outcome valence was in-consistent with process quality. The first conditiontests whether children are sensitive to basic qualitydistinctions in thinking processes at all, whereas thesecond offers a stronger test of the belief (endorsedby adults) that the features of a person’s thoughtprocess, more than the final outcome, determinewhether it is judged ‘‘good’’ or ‘‘bad.’’

Method

Participants

A total of 120 children (40 in each of first, third,and fifth grades) and 40 adults from a small Mid-western city participated. Children were publicschool students recruited through letters to parents( � 42% acceptance rate). Adults were college stu-dents who participated to fulfill an IntroductoryPsychology course requirement. Mean ages for first,third, fifth grade, and adult groups, respectively,were 6.2 years (SD 5 0.3), 8.3 years (SD 5 0.3), 10.6years (SD 5 0.3), and 18.7 years (SD 5 0.8). Therewere 10 males and 10 females per age group andcondition. The sample was 68% Caucasian, 14%Asian, 8% African American, 3% Hispanic, and 7%unknown or other backgrounds. No direct measuresof socioeconomic status were obtained, but thepopulations were generally middle to upper-middleclass.

Measures and Procedures

Children were tested individually in a quiet areaof their school, receiving the good/bad thinking taskfollowed by the cognitive word task. Sessions lasted20 – 25 min and were audio-taped. Adults receivedthe tasks as a written questionnaire.

Good/bad thinking task. Participants received 14scenarios describing people thinking in everydaysituations and were asked to rate whether they were‘‘using good thinking or bad thinking’’ and to ex-plain why (see the Appendix for scenarios). Scenar-ios were intended to portray everyday situations thatall ages would understand (e.g., deciding whether ornot to get a pet, figuring out who ate all the candy).Content was based on theoretical criteria for goodand bad thinking (see Baron, 2000), as well as beliefsabout good and bad reasoning that adults in prior

Beliefs about Everyday Reasoning 451

research had endorsed (Amsterlaw, in preparation;Lizotte & Amsterlaw, 2001). There were four types ofscenarios: (1) strategy scenarios contrasted using athoughtful decision strategy with using an arbitraryselection method (‘‘eenie – meenie’’), (2) alternativesscenarios contrasted considering alternative possi-bilities with jumping to an immediate conclusion, (3)evidence scenarios contrasted gathering evidencewith acting on a hunch, and (4) pros – cons scenarioscontrasted considering both pros and cons of achoice with considering only the pros. A good versionand a bad version for each scenario type yielded eightfocal scenarios. Versions were matched on peripheralfeatures, such as subject matter and length, so thatthe critical difference was in the thinking process.(Subtle differences minimized repetitiveness; pilottesting with adults and children confirmed that thesedid not affect ratings.) In addition, good and badversions of a control scenario were used to verify thatchildren could follow the basic task format. Thecontrol scenario contrasted an obviously bad prob-lem-solving strategy of not looking, not listening,and not trying with an obviously good strategy oflooking, listening, and really trying.

Participants were randomly assigned to either theno outcome or the mismatch task condition. In the nooutcome condition, scenarios included only infor-mation about people’s thinking processes, and didnot say whether the final outcome was good or bad(i.e., whether people solved problems correctly orwere happy with their choices). In the mismatchcondition, a final line was added to each scenario todescribe the outcome. Critically, for the eight focalscenarios, outcome valence was pitted againstthinking process quality; good processes were pairedwith bad outcomes and bad processes were pairedwith good outcomes. (Control scenarios were pre-sented with matched outcomes in the mismatchcondition and with no outcomes in the no outcomecondition.) In addition, four match scenarios, one foreach scenario type, where good processes led togood outcomes and bad processes led to bad out-comes, acted as filler scenarios for the mismatchcondition. Although not of direct interest, these wereincluded to ensure that the task did not seemoverly contrived to participants. (match scenarioswere also presented, without outcome information,in the no outcome condition to maintain similaritiesacross conditions.) Scenario orders were random-ized within age groups; the two constraints werethat scenarios of the same type did not followeach other and that the task always began with amatch scenario to provide a warm-up to the taskformat.

For children, a 500 � 800 drawing of a child’s facewas presented with each scenario. The researcherread each scenario aloud and asked the child to re-spond verbally to the key question, ‘‘Was X usinggood thinking or bad thinking?’’ (To control forpossible order effects, the order of ‘‘good’’ and ‘‘bad’’was counterbalanced between subjects.) After thechild responded, the researcher asked the follow-up,‘‘Was it very – very good/bad or just pretty good/bad?’’ Responses were later converted to a 0 (verybad) to 3 (very good) scale. Children were also asked toexplain their ratings (e.g., ‘‘Why do you think it wasvery bad thinking?’’).

Adults received the task in written form. Afterreading each scenario, they responded to ‘‘Was Xusing good thinking or bad thinking?’’ on a ratingscale with points labeled 0 (very bad), 1 (pretty bad), 2(pretty good), and 3 (very good). They were also askedto explain their ratings.

Cognitive word task. As before, the cognitive wordtask was included to test children’s comprehensionof key terms used in presenting the good/badthinking task (key terms were decide, figure out,wonder, compare, and get a hunch) and to provide amore global measure of metacognitive knowledge.The task was administered as in Study 1.

Explanation coding. Participants’ explanations fortheir ratings on the good/bad thinking task werecoded for appropriate reference to the thinkingprocess features targeted in each scenario. For ex-ample, for strategy scenarios, explanations for thegood version had to mention that it was good thatthe character made a list, compared, or looked forwhat she wanted, and for the bad version that it wasbad that the character used ‘‘eenie – meenie,’’ did notthink about what he wanted, or should have donesomething else to decide, such as test out the items.Responses that did not clearly reference target-thinking processes (see below) were not counted asevidence of understanding. Two raters independ-ently coded the data using transcripts that concealedparticipants’ age, race, and gender. Inter-rateragreement was 95%; disagreements were resolvedthrough discussion. (One adult gave no explanationsand so was excluded from the explanation analyses.All other participants had complete or near-completedata.)

For the mismatch condition, additional analysesassessed whether participants focused on thinkingprocesses, outcomes, or both, in justifying theirquality judgments. Explanations were coded intoone of three categories: (1) outcome explanations citedonly outcome information as a basis for the qualityjudgment (e.g., ‘‘He really liked the bike he got’’);

452 Amsterlaw

(2) process explanations cited only process informa-tion, regardless of whether the process aspect de-scribed was the target process feature for thatscenario type or if it was some other process feature(e.g., ‘‘She didn’t think about the possibilities thatcould have happened’’); and (3) mixed explanationsmentioned both outcomes and processes, seeminglywithout privileging one over the other (e.g., ‘‘Hethought about the good things and the bad things,but he turned out not to like it’’). Some explanationsthat mentioned both outcomes and processes wereconsidered process-focused. These included explicitstatements that thinking processes were central re-gardless of the final outcome (e.g., ‘‘Even if he got itright, it was still bad thinking because . . .’’), andresponses where an inconsistent outcome was usedto infer attributes of the thinking process (e.g., ‘‘Shemust not have tested it thoroughly if she picked thewrong thing’’). A fourth category included ‘‘I don’tknow,’’ off-topic, or unintelligible responses, as wellas responses that did not clearly refer to either out-comes or thinking processes. Uncodable responseswere most frequent among first graders, totaling 41out of 160 responses (26%), and were less frequentamong third graders (13%), fifth graders (2%), andadults (1%). Again, two independent coders ratedthe explanations. Inter-rater agreement was 96%;disagreements were resolved through discussion.

Results

Good/Bad Thinking Task

For each condition, developmental changes indistinguishing good and bad versions of each sce-nario type were examined with a 4 (grade) � 2(version: good or bad) ANOVA, where version was awithin-subjects factor. Preliminary analyses for ef-fects of question orderFwhether the question ‘‘WasX using good thinking or bad thinking?’’ was askedin that order or with the options reversedFfound noeffects, and no significant interactions with grade, ontask performance in either condition. There were alsono main or interaction effects for gender in eithercondition. Thus, all analyses collapse across thesevariables.

No outcome condition. As expected, no significantage effect was found for performance on the controlscenarios, F(3, 76) 5 2.24, p4.05. All age groups ap-propriately distinguished the good version(M 5 2.72, out of a maximum score of 3, SD 5 0.48)from the bad version (M 5 1.09, SD 5 0.33),F(1, 76) 5 1416.9, po.001 (confirmed for all agegroups with paired t tests, all pso.001). The top

panel of Figure 2 displays participants’ mean think-ing quality ratings for the focal ‘‘good process’’ and‘‘bad process’’ scenario versions in the no outcomecondition. The larger graph on the left gives themeans (and standard errors) collapsing across sce-nario type, and the four smaller graphs on the rightshow results for each type separately. Analyses onthe collapsed means found a Significant Grade �Version interaction, F(3, 76) 5 29.3, po.001, indicat-ing increasing differentiation of good and badthinking processes with age. This same effect wasobserved for all four scenario types when testedseparately: for strategy scenarios, F(3, 76) 5 32.08;evidence scenarios, F(3, 76) 5 7.16; alternatives sce-narios, F(3, 76) 5 12.60; and pros – cons scenarios,F(3, 76) 5 7.71, all pso.001.

Specifically, for strategy scenarios, first graders didnot differentiate between a thoughtful decisionprocess and a chance-based one, giving them equallyhigh ratings, t(19) 5 0.90, p4.05. By third grade,children rated the bad process significantly lowerthan the good one, t(19) 5 7.31, po.001. Similarly, foralternatives scenarios, first graders did not differen-tiate between a good thinking process that con-sidered alternatives from a bad one that involvedgoing with one’s very first thought, t(19) 5 1.02,p4.05. By third grade, however, children rated theprocess that involved considering alternatives asbetter than the one involving immediate conclusions,t(19) 5 12.25, po.001. First graders did distinguishbetween good and bad versions of the pros – consscenarios, rating a process that considered both sidesof the issue higher than one considering a single side,t(19) 5 3.16, po.01. They also appeared to distin-guish evidence scenarios, rating a hypothesis-testingprocess as better than using a hunch, t(19) 5 4.29,po.001, with these tendencies becoming more pro-nounced with age.

Individual response patterns showed similartrends. For strategy scenarios, 25% of first gradersrated the good version higher than the bad version,whereas 90% of third graders, 95% of fifth graders,and 100% of adults did so. For alternatives scenarios,45% of first graders rated scenarios in the expecteddirection, whereas 90 – 100% of third graders, fifthgraders, and adults did so. The figures were 60%,90%, 95%, and 100% for pros – cons scenarios, and65%, 75%, 80%, and 95% for evidence scenarios. Allages were near ceiling (95 – 100%) for control sce-narios.

Children’s tendency to explain their thinkingquality ratings by referencing target thinking processfeatures increased with age, particularly betweenfirst and third grades (see Table 3). Considering

Beliefs about Everyday Reasoning 453

whether children gave appropriate process-basedexplanations to at least one of two versions (eithergood or bad) for each scenario type, only about halfof first graders gave appropriate process-based ex-planations for ratings on strategy and alternativesscenarios, with somewhat better performance onpros – cons scenarios and poorer performance onevidence scenarios. In contrast, nearly all thirdgraders, fifth graders, and adults appropriately ref-erenced target thinking processes in their explana-tions for all four scenario types.

The explanations given by children who did notreference target thinking processes varied by sce-nario. For good strategy scenarios, most of thesewere nonspecific endorsements of the character’sthinking or the purchase itself (e.g., ‘‘because that’s agood way to do it’’; ‘‘she can take pictures with acamera’’), and for bad versions that eenie – meeniewas a good way to decide. For pros – cons scenarios,most other responses to the good version said it wasbad not to ask one’s parents about getting a pet ormentioned possible bad outcomes (e.g., pet getting

0

1

2

3

1 5 Adult

Good ProcessBad Process

0

1

2

3

1 3 5 Adult

Good ProcessBad Process

0

1

2

3

1 3 5 Adult

0

1

2

3

1 3 5 Adult

0

1

2

3

1 3 5 Adult

0

1

2

3

0

1

2

3

1 3 5 Adult

0

1

2

3

1 3 5 Adult

1 3 5 Adult 1 3 5 Adult

0

1

2

3

1 Adult

0

1

2

3

Qua

lity

Rat

ing

Qua

lity

Rat

ing

Qua

lity

Rat

ing

Qua

lity

Rat

ing

Qua

lity

Rat

ing

Qua

lity

Rat

ing

Grade

Grade

Grade Grade

Grade Grade

Qua

lity

Rat

ing

Qua

lity

Rat

ing

Qua

lity

Rat

ing

Qua

lity

Rat

ing

No Outcome Condition

Mismatch Condition

Mean of All Scenarios

Strategy Scenarios Pros-Cons Scenarios

Evidence ScenariosAlternatives Scenarios

Strategy Scenarios Pros-Cons Scenarios

Evidence ScenariosAlternatives Scenarios

Mean of All Scenarios

3

53

Figure 2. Thinking quality ratings for ‘‘good process’’ and ‘‘bad process’’ versions of scenarios from the good/bad thinking task, Study 2.Note. Error bars are standard errors of the means.

454 Amsterlaw

sick), and for the bad version that it was good tohave a pet. For alternatives scenarios, responses toboth versions typically indicated agreement with thecharacter about who had committed the crime (e.g.‘‘because older brothers do that’’). Similarly, re-sponses to evidence scenarios indicated agreementor disagreement with the character about the sourceof the problem (e.g., ‘‘if a plant is drooping it doesn’tneed to be trimmed’’).

Because ratings and explanations were coded in-dependently, children could get credit for appropri-ate explanations even if their ratings did notdistinguish between good and bad thinking, andvice versa. However, in most cases, children whogave higher ratings to good scenario versions alsogave appropriate process-based explanations (seeTable 4). The only exception was for evidence sce-narios, where only 3 of 13 first graders who gavecorrect ratings also gave process-based explanations.This confirms a general correspondence betweentask success and knowledge of key process featuresfor most scenario types. (It also indicates that forevidence scenarios, first graders’ rating data likelyoverestimated their understanding of the contrast.)

Mismatch condition. As before, there was no sig-nificant age effect for performance on the controlscenarios, F(3, 76) 5 1.78, p4.05. Participants of allages rated the good version (M 5 2.76, SD 5 0.46)higher than the bad version (M 5 0.14, SD 5 0.38),

F(1, 76) 5 1369.1, po.001 (confirmed for all agegroups by paired t test, all pso.001). Similarly, allages rated good versions of the match scenarios(M 5 2.38, SD 5 0.48) higher than bad versions(M 5 1.00, SD 5 0.50), F(1, 76) 5 354.5, po.001. Therewas also a significant Grade � Version interaction,indicating increasing divergence of ratings with age,F(3, 76) 5 6.50, po.01.

The lower panel of Figure 2 displays participants’thinking quality ratings for good and bad processversions in the mismatch condition. (Recall that forthe focal scenarios in this condition, good processversions were always presented with bad outcomesand bad process versions with good outcomes.) Asindicated in the figure, there was a significant GradeX Version crossover interaction, F(3, 76) 5 29.8,po.001, with first graders showing a reversed ratingpattern from that of adults. This interaction occurredfor all four scenario types when tested separately: forstrategy scenarios, F(3, 76) 5 22.8; evidence scenar-ios, F(3, 76) 5 12.4; alternatives scenarios, F(3, 76) 5

22.3; and pros – cons scenarios, F(3, 76) 5 7.56, allpso.001.

Specifically, for strategy scenarios, first gradersrated the bad process version higher than the goodprocess version, t(19) 5 4.01, po.01, and third grad-ers rated them similarly, t(19) 5 0.81, p4.05. Onlyfifth graders showed the adult pattern of rating thegood version significantly higher than the bad ver-sion, t(19) 5 2.77, po.05. For alternatives scenarios,first graders also rated the bad process versionhigher than the good process version, t(19) 5 4.87,po.001. By third grade, however, children rated thegood version higher, t(19) 5 2.45, po.05. For pros –cons scenarios, it was not until fifth grade that chil-dren rated the good version higher than the badversion, t(19) 5 2.46, po.05. First graders rated them

Table 3

Participants’ References to Target Thinking Process Features in Thinking

Quality Explanations, Study 2

Scenario type

Grade

1 3 5 Adult

No outcome condition

Strategy 55 95�� 100 100

Alternatives 55 100�� 90 100

Pros – cons 70 100�� 95 100

Evidence 15 85��� 85 100w

Control 100 100 100 100

Mismatch condition

Strategy 25 75�� 90 100

Alternatives 20 95��� 95 100

Pros – cons 35 90��� 85 90

Evidence 10 35w 65w 95�

Control 90 95 100 100

Note. Values represent the percentage of participants referring totarget thinking process features in explanations for at least one oftwo scenario versions (either good or bad) per type. Significancewas determined by w2(1, N 5 40) against the next younger agegroup.wpo.10. �po.05. ��po.01. ���po.001.

Table 4

Performance on Thinking Quality Explanations as a Function of Per-

formance on Thinking Quality Ratings, Study 2 (No Outcome Condi-

tion)

Scenario type

Grade

1 3 5 Adult

Strategy 4 (5) 18 (18) 19 (19) 19 (19)

Alternatives 8 (9) 20 (20) 18 (18) 19 (19)

Pros – cons 10 (12) 18 (18) 19 (19) 19 (19)

Evidence 3 (13) 13 (15) 15 (16) 19 (19)

Control 19 (19) 20 (20) 20 (20) 18 (18)

Note. Values represent the number of participants referring totarget thinking processes in their quality explanations, out of thetotal number (shown in parentheses) who rated the ‘‘good-pro-cess’’ version higher than the ‘‘bad-process’’ version.

Beliefs about Everyday Reasoning 455

similarly, t(19) 5 1.45, as did third graders,t(19) 5 0.93, ps4.05. Finally, only adults distin-guished between good and bad evidence scenarios,t(18) 5 3.92, po.01. Third graders still rated the badversion higher than the good version, t(19) 5 2.44,po.05, while fifth graders rated them similarly,t(19) 5 0.42, p4.05.

Individual rating patterns reflected these sametrends. Adults overwhelmingly rated good-process/bad-outcome scenarios higher than bad-process/good-outcome scenarios; 90% did so for strategyscenarios, 85% for alternatives and pros – cons sce-narios, and 63% for evidence scenarios. When devi-ating from this pattern adults rarely rated scenariosin the opposite direction (4 of 160 total judgments),instead tending to rate scenarios equally. By contrast,first graders’ dominant response was to rate bad-process/good-outcome scenarios higher, with 70%doing so for strategy scenarios, 65% for alternativesscenarios, 45% for pros – cons scenarios, and 65% forevidence scenarios. Third graders and fifth graderstended to rate the scenarios either equally or ac-cording to the adult pattern.

Although adults continued to see thinking qualityas a function of underlying thinking processes evenwhen outcomes were mismatched, outcome infor-mation did affect their ratings, as earlier outcomebias research has shown (e.g., Baron & Hershey,1988), particularly in bad outcome cases. Relative tothe no outcome condition, adults ratings for goodprocess scenarios in the mismatch condition were0.63 points lower on average when outcomes werebad, t(38) 5 4.30, po.001, and ratings for bad-processscenarios were 0.34 points higher on average whenoutcomes were good, t(38) 5 3.06, po.01. Whenperformance across the two conditions was directlycompared, all age groups, even adults, were morelikely to rate good thinking processes higher thanbad thinking processes in the no outcome condition.Assigning each participant a score (0 – 4) for thenumber of scenarios on which they showed the ex-pected pattern, first graders averaged 1.95(SD 5 1.40) expected ratings in the no outcome con-dition, but only 0.40 (SD 5 0.82) in the mismatchcondition, t(38) 5 4.28, po.001. Third graders aver-aged 3.55 (SD 5 0.61) expected ratings in the nooutcome condition, compared with 1.40 (SD 5 1.31)in the mismatch condition, t(38) 5 6.65, po.001, andfifth graders averaged 3.60 (SD 5 1.00) versus 2.10(SD 5 1.33), t(38) 5 4.03, po.001. Adults showed thesmallest margin of differenceF3.95 (SD 5 0.22) inthe no outcome condition versus 3.15 (SD 5 0.80) inthe mismatchFbut the difference was still signifi-cant, t(38) 5 3.22, po.01.

As in the no outcome condition, children’s ten-dency to explain their ratings by citing target think-ing processes increased with age, with childrenshowing clear improvements between first and thirdgrades on most scenario types (see Table 3). Addi-tional results on the frequency of process- versusoutcome-focused explanations, presented in Figure3, showed that for all four scenario types, exclusivelyprocess-focused explanations increased significantlybetween first and fifth grades, w2(3, N 5 160)430.1,all pso.001, whereas exclusively outcome-focusedexplanations significantly decreased in three of fourcases, w2(3, N 5 160)417.8, all pso.001. (For pros –cons scenarios, outcome-focused explanations weresimilarly low among all age groups, w2(3,N 5 160) 5 3.60, p4.05.) Adults focused almost ex-clusively on process features in explaining their rat-ings for all scenario types.

Cognitive Word Task

Initial analyses showed no significant main effectsfor gender or condition (no outcome vs. mismatch),and no significant interactions of either variable withgrade, on cognitive word task performance, allps4.05. As before, performance improved with age,F(3, 156) 5 112.4, po.001. On average, first gradersgot 6.9 of 12 total items correct (SD 5 2.0), thirdgraders got 10.1 items correct (SD 5 1.1), fifth grad-ers got 10.8 items correct (SD 5 0.8), and adults got11.8 items correct (SD 5 0.4). Performance on indi-vidual items was comparable with results fromStudy 1. Most children correctly understood all thecognitive terms used in the good/bad thinking task.The majority (72% or more) of first graders knew theterms decide and figure out, whereas 60% knew won-der, 50% knew compare, and 38% knew get a hunch(performance on all except hunch was significantlyabove the chance expectation of 33% correct, ac-cording to one-tailed binomial tests, pso.05). Most(80% or more) third and fifth graders understood allthe terms used in the task, as well as additional,harder items.

For children, better performance on the cognitiveword task was associated with greater differentiationof good and bad thinking on the good/bad thinkingtask (where amount of differentiation was calculatedby subtracting children’s mean ratings for badprocess versions from their mean ratings for goodprocess versions) in both the no outcome,r(60) 5 0.67, and mismatch conditions, r(60) 5 0.57,pso.001. Age was also correlated with performanceon the good/bad thinking task, r(57) 5 0.58 (no out-come) and r(59) 5 0.61 (mismatch), as well as with

456 Amsterlaw

performance on the cognitive word task,r(116) 5 0.69, pso.001. Linear regression results in-dicated that age alone explained 33% of the variancein task performance in the no outcome condition,F(1, 55) 5 27.93, po.001. Cognitive word task per-formance accounted for an additional 14%,F(1, 54) 5 15.68, po.01, for a total of 47% varianceexplained, F(2, 54) 5 25.53, po.001. In the mismatchcondition, age accounted for 36% of the variance,F(1, 57) 5 34.26, po.001. Cognitive word task per-formance did not add to this significantly,F(1, 56) 5 4.00, p4.05. Thus, for basic distinctionsbetween good and bad thinking (no outcome con-dition), both being older and having better meta-cognitive word knowledge contributed to tasksuccess, whereas for prioritizing processes overoutcomes (mismatch condition), only being olderexplained better performance. These resultsmay mean that outcome bias declines with age in-dependent of other metacognitive developments.Alternatively, more sensitive metacognitive assess-ments may be necessary to test the relationship.

Discussion

Study 2 observed developmental changes in chil-dren’s differentiation of good and bad reasoning,both in their basic ability to discriminate good andbad processes and in their prioritization of processversus outcome information in global evaluations of

thinking quality. In the no outcome condition, firstgraders were sensitive to quality distinctions forpros – cons and control scenarios only. For otherscenarios, it was not until third grade that childrenreliably distinguished good and bad thinking pro-cesses. In the mismatch condition, only fifth gradersconsistently showed the adult pattern of privilegingprocess over outcome information for most cases.First graders tended to privilege outcomes overprocesses, whereas many third graders appeared toweight them equally. These results identify impor-tant age-related changes in children’s ability toevaluate thinking processes vis-a-vis adult culturalnorms of rationality.

Previously, a single study by Perkins et al. (2000)found that sixth graders have trouble detectingthinking flaws in everyday thinking, such as ne-glecting alternatives and failing to seek reasons onboth sides of a case. The current work demonstratesthat even first graders have some basic intuitionsabout the difference between good and bad thinking,and that third and fifth graders’ knowledge aboutthese topics is in fact fairly well developed. Thesetwo studies may have reached different conclusionsbecause of methodological differences: Perkins etal.’s (2000, p. 276) arguably more stringent assess-ment used a paper-and-pencil measure where chil-dren provided written explanations that were laterevaluated for elements such as ‘‘elaboration, crea-tivity, and the presentation of a range of ideas,’’

0%

20%

40%

60%

80%

100%

0%

20%

40%

60%

80%

100%

0%

20%

40%

60%

80%

100%

1 5 Adult

0%

20%

40%

60%

80%

100%

1 5 AdultProcess

Mixed

Outcome

Strategy Scenarios Pros-Cons Scenarios

Alternatives Scenarios Evidence Scenarios

Grade

Grade

Exp

lan

atio

ns

Exp

lan

atio

ns

Exp

lan

atio

ns

Exp

lan

atio

ns

3 1 5 AdultGrade

3

3 1 5 AdultGrade

3

Figure 3. Frequency of outcome, process, and mixed explanations types in the mismatch condition of Study 2.

Beliefs about Everyday Reasoning 457

whereas this study used a verbal format, included aratings-based assessment as well as explanations,and restricted evaluations of children’s responses toconceptual content, rather than additional features.

In the current study, similar patterns of develop-ment were observed across multiple types of sce-narios, which suggests the possibility of a basicdevelopmental change in children’s ability to eval-uate thinking processes. One exception worth notingis that children’s ability to distinguish evidence-based conclusions from plausible hunch-based con-clusions developed more slowly than did otherprocess-based distinctions. This echoes results fromprior research on children’s scientific reasoning skills(e.g., Kuhn, Amsel, & O’Loughlin, 1988), showingthat the distinction between theory and evidence canbe especially difficult for children.

Results from the cognitive word task showed thatchildren who had more sophisticated metacognitiveword knowledge were better able to distinguish be-tween good and bad thinking. This suggests thatdevelopments in children’s metacognition aboutreasoning are related to broader metacognitivechanges, or at least to increased comprehension ofmental terms. It also raises the question of whetheryounger children performed more poorly on the tasksimply because they had difficulty with specificterms. Although it is possible that this issue affectedsome younger children, it cannot account for thelarger age effects. Some first graders did have diffi-culty with the terms compare and get a hunch (used instrategy and evidence scenarios, respectively);however, first graders who understood these termswere no more likely to distinguish between good andbad versions of the relevant scenarios than werechildren who did not, w2(1, N 5 20)o0.1, ps4.05).Further, age effects were similar across multiplescenario types, including those (such as alternativesscenarios) that contained no problematic terms. Fi-nally, third graders, who showed excellent compre-hension of all terms from the good/bad thinkingtask, still differed from older ages in their ratings,most notably in the mismatch condition. Somedeeper conceptual difference must explain theseeffects.

A likely reason for outcome biases in children’sevaluations of thinking is simply that they lack themetacognitive knowledge necessary for makingprocess-based judgments. (Indeed, results from theno outcome condition show that first graders wereoften unable to distinguish between good and badthinking based on only process information.) If thisis the case, we might expect first graders to showrating reversals in the mismatch condition only for

scenario types they failed to distinguish in the nooutcome condition (strategy, evidence, alternatives),and not for those they did distinguish (pros – cons).The current study used a between-subjects designand so offers only an indirect test of this relationship;however, the data do show this pattern. Pros – consscenarios are the only type for which first gradershad no significant rating reversal.

This suggests that a main contributor to outcomebias is children’s limited ability to discriminate goodand bad thinking processes. That is not the wholestory, though, because even when children can ap-propriately distinguish thinking processes, youngerchildren still overweight outcomes in their qua-lity judgments. (Third graders, for example, distin-guished between good and bad thinking in the nooutcome condition, but failed to privilege processesover outcomes in the mismatch condition.) Furtherresearch is necessary to determine the sources of thiseffect. One possibility is that outcome bias declinesin older children and adults not only because theyare more knowledgeable about specific thinkingprocesses and strategies, but also because they arebetter able to distinguish between outcomes that arecausally attributable to actors’ strategy choices andthose that are not (e.g., see Nicholls & Miller, 1985).When outcomes are inconsistent with process qual-ity, older children and adults may discount them asdue to factors beyond the actor’s control (e.g., badluck, lack of available information) while youngerchildren hold actors equally accountable for bothstrategy choices and outcomes.

General Discussion

These two studies reveal meaningful developmentalchanges in how children conceptualize reasoningand evaluate its quality. Between the ages of about 6years to about 10 years, children’s categories of rea-soning versus nonreasoning became increasinglydifferentiated, their ability to distinguish betweengood and bad reasoning processes improved, andtheir concepts of thinking quality became increas-ingly process-based.

These developments can be interpreted in thecontext of broader metacognitive changes takingplace during middle childhood. Prior research hasdescribed the ‘‘interpretive’’ or ‘‘constructive’’ theoryof mind (Carpendale & Chandler, 1996; Schwanen-flugel et al., 1994) as a key developmental achieve-ment of this period. That is, children come torecognize that subjective mental processes, such asinterpretation and inference, intervene between theworld and our representations of and responses to it.

458 Amsterlaw

This emphasis on human cognitive processes assources of knowledge seems to entail an increasingsensitivity to a variety of process-based distinctionsin children’s concepts of thinking and knowing, in-cluding distinctions between states of consciousnessand unconsciousness (Flavell et al., 1999), construc-tive and nonconstructive cognitive processing(Schwanenflugel et al., 1998), and sources of moreand less certain knowledge (Pillow, Hill, Boyce, &Stein, 2000). The current results indicate that thesechanges are equally reflected in children’s develop-ing concepts of reasoning, as they learn to distin-guish between reasoning and other kinds of thinkingand develop ideas about the effectiveness of variousreasoning practices.

Some theories posit that increasing metalevelknowledge about thinking contributes to develop-mental change in reasoning (Kuhn, 2000b; Moshman,1998). Applied to the present findings, children’s useof appropriate reasoning practices, such as gatheringevidence to inform conclusions, may crucially de-pend on their having knowledge of them in the firstplace. In theory, such relationships make sense.When children encounter problems, surely theymust appeal to some metalevel knowledge abouthow to proceed. If they did not, it is unclear howthey would regulate their reasoning attempts at all.(To the extent that children do try to regulate theirthinking, then, we might assume they have someminimal sense of standards to apply.) Nonetheless,the relationship between metacognition and reason-ing is not straightforward. First, despite obviousmetacognitive improvements of the sort demon-strated here, children’s reasoning performance doesnot always improve with age (see Jacobs &Klaczynski, 2002). Although knowledge of appro-priate strategies and standards might be considerednecessary for effective reasoning, it may not be at allsufficient. (Adults’ failure to live up to standards ofnormative judgment is a case in point.) ‘‘Disposition-al’’ factors that describe individuals’ tendencies todeploy their knowledge and skills in specific situa-tions may prove just as important as metacognitiveknowledge (Perkins et al., 1993a, 1993b). Second,there is accumulating evidence that development inchildren’s reasoning knowledge and skills is notlimited to the acquisition of adaptive practicesand strategies. Even as children are learning strate-gies that enable them to reason appropriately, itseems they are also acquiring heuristics and short-cuts that can lead to biased or incorrect judgments,actually causing some aspects of reasoning perfor-mance to decline with age (e.g., Jacobs & Potenza,1991; Klaczynski & Narasimham, 1998; Reyna &

Ellis, 1994). This raises the possibility that children’s(as well as adults’) metacognitive knowledge in-cludes maladaptive, as well as adaptive, beliefs.Developmental theories must take these factors intoaccount.

What leads to developmental change in children’sknowledge about reasoning? Two likely mechanismsare (1) children’s introspective reflection about theirreasoning experiences and (2) social learning in in-teraction with adults (Moshman, 1994). With regardto experience and reflection, children’s reasoningstrategies and their metastrategic knowledge haveboth been found to improve during extended en-gagement with reasoning tasks (Kuhn & Pearsall,1998), presumably because children learn from taskfeedback. Young children’s heightened attention tooutcomes in reasoning situations may thus servethem well, by bootstrapping knowledge about whichstrategies reliably obtain desired outcomes. At thesame time, children’s earliest forays into problem-solving and decision-making occur in social collab-oration with more expert individuals (e.g., parentsand teachers), and much of the specific knowledgechildren acquire, such as beliefs about good and badreasoning practices, is likely transmitted to them viateaching, both direct (e.g., instruction about relevantconcepts and strategies) and indirect (e.g., modelingpractices, providing feedback, talking about think-ing). (Of course, faulty beliefs and heuristics mayalso be acquired this way.)

Previous research has focused almost exclusivelyon children’s understanding of reasoning in formalcontexts, such as scientific experimentation andsolving logic problems. However, given young chil-dren’s limited experience with such cases, beliefsabout everyday informal reasoning may well be theplace where their earliest and most developedknowledge exists. The current work asks newquestions about children’s ideas about reasoningin everyday contexts (e.g., making choices aboutpurchases, assigning blame to others), and findsthat in responding to such cases, children readilyrecognize the legitimate problems these situationspresent, are interested in questions about the statusof various problem-solving approaches, and have anemerging set of important insights about these is-sues. Relative to prior work, these results speak moredirectly to the kind of metalevel knowledge childrenmight invoke when they encounter reasoning situa-tions in their real lives. A crucial next step in thisresearch path will be to evaluate developmental re-lationships between children’s metacognition aboutreasoning and their reasoning behavior in natural-istic settings.

Beliefs about Everyday Reasoning 459

References

Amsterlaw, J. (in preparation). Naıve rationality: Folktheories of everyday reasoning.

Astington, J. W., & Olson, D. R. (1990). Metacognitiveand metalinguistic language: Learning to talk aboutthought. Applied Psychology: An International Review, 39,77 – 87.

Baron, J. (2000). Thinking and deciding (3rd ed.). New York:Cambridge University Press.

Baron, J., & Hershey, J. C. (1988). Outcome bias in decisionevaluation. Journal of Personality and Social Psychology, 54,569 – 579.

Baron, J. B., & Sternberg, R. J. (Eds.) (1987). Teachingthinking skills: Theory and practice. New York: W. H.Freeman.

Booth, J. R., & Hall, W. S. (1994). Role of the cognitive in-ternal state lexicon in reading comprehension. Journal ofEducational Psychology, 86, 413 – 422.

Booth, J. R., & Hall, W. S. (1995). Development of theunderstanding of the polysemous meanings ofthe mental-state verb know. Cognitive Development, 10,529 – 549.

Carpendale, J. I., & Chandler, M. J. (1996). On the distinc-tion between false belief understanding and subscribingto an interpretive theory of mind. Child Development, 67,1686 – 1706.

DeLoache, J. S., Miller, K. F., & Pierroutsakos, S. L. (1998).Reasoning and problem solving. In D. Kuhn & R. Siegler(Eds.), Handbook of child psychology: Vol. 2. Cognition,perception, and language development (5th ed., pp. 801 –850). New York: Wiley.

Flavell, J. H., & Green, F. L. (1999). Development of intu-itions about the controllability of different mental states.Cognitive Development, 14, 133 – 146.

Flavell, J. H., Green, F. L., & Flavell, E. R. (1993). Children’sunderstanding of the stream of consciousness. ChildDevelopment, 64, 387 – 398.

Flavell, J. H., Green, F. L., & Flavell, E. R. (1995a). Thedevelopment of children’s knowledge about attentionalfocus. Developmental Psychology, 31, 706 – 712.

Flavell, J. H., Green, F. L., & Flavell, E. R. (1995b). Youngchildren’s knowledge about thinking. Monographs of theSociety for Research in Child Development, 60 (1, Serial No.243).

Flavell, J. H., Green, F. L., & Flavell, E. R. (1998). The mindhas a mind of its own: Developing knowledge aboutmental uncontrollability. Cognitive Development, 13,127 – 138.

Flavell, J. H., Green, F. L., Flavell, E. R., & Lin, N. T. (1999).Development of children’s knowledge about uncon-sciousness. Child Development, 70, 396 – 412.

Flavell, J. H., & Miller, P. H. (1998). Social cognition. In D.Kuhn & R. S. Siegler (Eds.), Handbook of child psychology:Vol. 2. Cognition, perception, and language development (5thed., pp. 851 – 898). New York: Wiley.

Galotti, K. M. (1989). Approaches to studying formal andeveryday reasoning. Psychological Bulletin, 105, 331 – 351.

Gopnik, A., & Graf, P. (1988). Knowing how you know:Young children’s ability to identify and remember thesources of their beliefs. Child Development, 59, 1366 –1371.

Heyman, G. D., Gee, C. L., & Giles, J. W. (2003). Preschoolchildren’s reasoning about ability. Child Development, 74,516 – 534.

Jacobs, J. E., & Klaczynski, P. A. (2002). The development ofjudgment and decision making during childhood andadolescence. Current Directions in Psychological Science,11, 145 – 149.

Jacobs, J. E., & Potenza, M. (1991). The use of judgmentheuristics to make social and object decisions: Adevelopmental perspective. Child Development, 62,166 – 178.

Joyner, M. H., & Kurtz-Costes, B. (1997). Metamemorydevelopment. In N. Cowan (Ed.), The development ofmemory in childhood (pp. 275 – 300). Hove, England:Psychology Press.

Keil, F. (1989). Concepts, kinds, and cognitive development.Cambridge, MA: MIT Press.

Klaczynski, P. A. (2001). Analytic and heuristic processinginfluences on adolescent reasoning and decision-mak-ing. Child Development, 72, 844 – 861.

Klaczynski, P. A., & Narasimham, G. (1998). Representa-tions as mediators of adolescent deductive reasoning.Developmental Psychology, 34, 865 – 881.

Kreutzer, M. A., Leonard, C., & Flavell, J. H. (1975). Aninterview study of children’s knowledge about memory.Monographs of the Society for Research in Child Develop-ment, 40 (1, Serial no. 159).

Kuhn, D. (1996). Is good thinking scientific thinking? InD. R. Olson & N. Torrance (Eds.), Modes of thought:Explorations in culture and cognition (pp. 261 – 281). NewYork: Cambridge University Press.

Kuhn, D. (2000a). Metacognitive development. CurrentDirections in Psychological Science, 9, 178 – 181.

Kuhn, D. (2000b). Theory of mind, metacognition, andreasoning: A life-span perspective. In P. Mitchell (Ed.),Children’s reasoning and the mind (pp. 301 – 326). Hove,England: Psychology Press.

Kuhn, D., Amsel, E., & O’Loughlin, M. (1988). The devel-opment of scientific reasoning skills. Orlando, FL: Aca-demic Press.

Kuhn, D., Katz, J., & Dean, D. (2004). Developing reason.Thinking & Reasoning, 10, 197 – 219.

Kuhn, D., & Pearsall, S. (1998). Relations between meta-strategic knowledge and strategic performance. Cogni-tive Development, 13, 227 – 247.

Lizotte, D., & Amsterlaw, J. (2001) Reasoning about reason-ing: A descriptive study of young adults’ everyday theories ofcognition. Poster presented at meeting of the CognitiveDevelopment Society, Virginia Beach, VA.

Montgomery, D. E. (1992). Young children’s theoryof knowing: The development of a folk epistemology.Developmental Review, 12, 410 – 430.

Morris, A. K. (2000). Development of logical reasoning:Children’s ability to verbally explain the nature of the

460 Amsterlaw

distinction between logical and nonlogical forms of ar-gument. Developmental Psychology, 36, 741 – 758.

Moshman, D. (1994). Reasoning, metareasoning, and thepromotion of rationality. In A. Demetriou & A. Efklides(Eds.), Intelligence, mind, and reasoning: Structure and de-velopment (pp. 135 – 150). Amsterdam: North-Holland/Elsevier.

Moshman, D. (1998). Cognitive development beyondchildhood. In D. Kuhn & R. S. Siegler (Eds.), Handbook ofchild psychology: Vol. 2. Cognition, perception, and languagedevelopment (5th ed., pp. 947 – 978). New York: Wiley.

Moshman, D., & Franks, B. (1986). Development of theconcept of inferential validity. Child Development, 57,153 – 165.

Nicholls, J. G., & Miller, A. T. (1985). Differentiation of theconcepts of luck and skill. Developmental Psychology, 21,76 – 82.

O’Neill, D. K., & Gopnik, A. (1991). Young children’sability to identify the sources of their beliefs. Develop-mental Psychology, 27, 390 – 397.

Perkins, D. N., Jay, E., & Tishman, S. (1993a). Beyondabilities: A dispositional theory of thinking. Merrill-Palmer Quarterly, 39, 1 – 21.

Perkins, D. N., Jay, E., & Tishman, S. (1993b). New con-ceptions of thinking: From ontology to education. Edu-cational Psychologist, 28, 67 – 85.

Perkins, D., Tishman, S., Richart, R., Donis, K., & Andrade,A. (2000). Intelligence in the wild: A dispositional viewof intellectual traits. Educational Psychology Review, 12,269 – 293.

Pillow, B. H. (1999). Epistemological development in ado-lescence and adulthood: A multidimensional frame-work. Genetic, Social, & General Psychology Monographs,125, 413 – 432.

Pillow, B. H., Hill, V., Boyce, A., & Stein, C. (2000). Un-derstanding inference as a source of knowledge: Chil-dren’s ability to evaluate the certainty of deduction,perception, and guessing. Developmental Psychology, 36,169 – 179.

Reyna, V. F., & Ellis, S. C. (1994). Fuzzy-trace theory andframing effects in children’s risky decision-making.Psychological Science, 5, 275 – 279.

Schwanenflugel, P. J., Fabricius, W. V., & Alexander, J.(1994). Developing theories of mind: Understandingconcepts and relations between mental activities. ChildDevelopment, 65, 1546 – 1563.

Schwanenflugel, P. J., Henderson, R. L., & Fabricius,W. V. (1998). Developing organization of mental verbsand theory of mind in middle childhood: Evidence fromextensions. Developmental Psychology, 34, 512 – 524.

Sodian, B., & Wimmer, H. (1987). Children’s understandingof inference as a source of knowledge. Child Development,58, 424 – 433.

Stanovich, K. E., & West, R. F. (2000). Individual differencesin reasoning: Implications for the rationality debate?Behavioral and Brain Sciences, 23, 645 – 726.

Tunmer, W. E., Nesdale, A. R., & Pratt, C. (1983). The de-velopment of young children’s awareness of logical in-

consistencies. Journal of Experimental Child Psychology, 36,97 – 108.

Wellman, H. M. (1990). The child’s theory of mind. Cam-bridge, MA: MIT Press.

Wellman, H. M. (2002). Understanding the psychologicalworld: Developing a theory of mind. In U. Goswami(Ed.), Blackwell handbook of childhood cognitive development(pp. 167 – 187). Malden, MA: Blackwell.

Wellman, H. M., & Gelman, S. A. (1998). Knowledge ac-quisitions in foundational domains. In D. Kuhn & R. S.Seigler (Eds.) Handbook of child psychology: Vol. 2. Cogni-tion, perception, and language development (5th ed., pp.523 – 573). New York: Wiley.

Appendix

Scenarios for Reasoning/Not-Reasoning Task (Study 1)

Reasoning Scenarios

On Sylvia’s birthday, somebody sent her somebeautiful flowers and a birthday card. But the persondidn’t sign their name; it just said ‘‘from your secretfriend.’’ Who could it be? Whoever sent the flowersknew that Sylvia loved daisies. Sylvia thought aboutwhich friends of hers might know that. ‘‘I was justtelling Chris how much I like daisies,’’ she said. ‘‘It’sprobably Chris who sent them to me.’’

Kyle’s parents are getting him a bike for hisbirthday this year. So Kyle went to the bike store topick one out. There were lots of different bikes tochoose from. Kyle wrote down a list of the importantthings he wanted in a bike, and then he comparedthe bikes in the store to his list. He looked for the bikethat was the closest match to the things on his list.

Jeremy has a garden. One day Jeremy saw thatsome plants in his garden were wilting and turningbrown. He wondered what was going on. Jeremytried changing different things, like how much waterand plant food the plants got, and he kept track ofwhat helped the plants and what did not. On thebasis of that, he decided they needed more plant food.

Shortcut Problem-Solving Scenarios

Hugh was excited because this summer he got togo to summer camp. There were two different campshe could go to and Hugh’s parents said he could pickthe one he wanted. So Hugh got out a quarter and hesaid, ‘‘If it’s heads it’s this camp, and if it’s tails it’sthat camp.’’ Hugh flipped the quarter and it landedon tails. So he went to the second camp.

One night Fabio and his family were watching TV.They were watching a certain game show where the

Beliefs about Everyday Reasoning 461

people on the show have to answer lots of questions.One of the questions was a big math problem withlots of parts. It was pretty tricky! But Fabio had seenthis same question on the show before. He just re-membered that the answer was 15.

Bianca got a new alarm clock, but she didn’t knowhow to set the time. There were all these funnybuttons on the back of the clock. Bianca didn’t knowwhat they were for. Without really watching whatshe was doing, Bianca started pressing differentbuttons. She pressed a few over here, and she pres-sed a few over there, just pressing buttons. After that,she saw that the clock was on the right time.

Automatic Action Scenarios

Henry loves chocolate. It’s his favorite flavor. Oneday Henry’s mom made a big batch of chocolate chipcookies. She left them sitting on the kitchen counterto cool. A little while later Henry got home fromschool. He saw the cookies sitting on the counter.And as soon as he saw them, he walked right overand ate three of them.

Lois likes to help her mom and dad cook dinner.One night they were making spaghetti for dinner.Lois was stirring the pot of spaghetti when she ac-cidentally put her hand down on the stove. The stovewas very, very hot! As soon as she touched the hotstove, Lois jerked her hand away.

When Marcy walks to school, she has to cross avery busy street with lots of traffic. One day Marcystarted to cross the street when suddenly a hugetruck came speeding down the road. At first she didnot see it coming, but at the last minute she lookedup. The truck was almost right to her! Marcy jumpedback to the side of the road.

Items for Cognitive Word Task (Study 1 and Study 2)

Listen (warm up). Let’s say Sarah is on the tele-phone. She has the phone right up to her ear. It’s hermom on the other end. Sarah’s mom is telling her astory. Is Sarah looking at her mom, listening to hermom, or touching her mom?

Decide. Let’s say Sarah just got out of school at theend of the day and now she is thinking about what todo next. She can either go to the playground with hersister or she can go to the library and read books. Shecan’t do both, so she has to think about which oneshe would rather do. Is Sarah dreaming, learning, ordeciding?

Figure out. Let’s say Sarah is leaving her classroomat the end of the day when she sees somebody’s note-book lying on the ground. There is no name on the

front, only the initials AJ. Whose is it? Sarah thinks ofthe names of all the people in her class. There is onlyone name with those initialsFit’s Andrew James.Sarah thinks it must be his notebook. Was Sarah re-membering it, making it up, or figuring it out?

Guess. One day Sarah’s dad played a game withSarah. He put his hands behind his back so Sarahcouldn’t see and he said, ‘‘I have some candy in oneof my hands, but I’m not going to tell you which one.If you can pick which hand it is, I’ll give you thecandy.’’ Sarah picked one, but she didn’t knowwhere the candy really was. Was Sarah guessing,forgetting, or planning?

Remember. Let’s say Sarah’s mom sent her to thegrocery store to get some things. She didn’t giveSarah a list. She just told her what things to get. NowSarah is at the store and she’s thinking about whatthings her mom wanted her to get. Is Sarah learning,wishing, or remembering?

Discover. Let’s say Sarah is riding her bike over toher friend Joe’s house. It usually takes her a while tobike over to Joe’s. This time, when Sarah is riding herbike, she sees a little path on one side of the road thatshe never saw before. It turns out to be a secretshortcut over to Joe’s house. Did Sarah interpret thepath, imagine the path, or discover the path?

Wonder. Let’s say one day there was a big snow-storm. When Sarah sees all the snow, she startsasking herself some questions in her head. ‘‘Wheredoes snow come from?’’ she thinks ‘‘How is it madeup in the clouds?’’ Is Sarah guessing, wondering, ordeciding?

Understand. Sarah is in her science class and herteacher is giving them directions about how to do thehomework assignment. Sarah is following what herteacher says, and she feels that she knows exactly whatto do. Is Sarah understanding, wondering, or exploring?

Compare. Let’s say Sarah is at the zoo and she islooking at some different animals. She is looking atan elephant and a rhinoceros, and she is thinkingabout how they are alike and how they are different.Is Sarah listening, comparing, or understanding?

Get a hunch. Let’s say Sarah is driving along in hercar when the car starts making a funny noise. Sarahdoesn’t know for sure why the car is making a funnynoise, but she has a feeling that it has something todo with the fan. Is Sarah getting a hunch, making apromise, or having a dream?

Predict. Sarah and her friend Michelle are plan-ning to go on a picnic. They want to choose a niceday. One morning they wake up early. Michelle says,‘‘Should we go today?’’ Sarah looks out the windowand she says, ‘‘Yes. It will be sunny at lunchtime to-day.’’ Was Sarah pretending, predicting, or concluding?

462 Amsterlaw

Conclude. Sarah woke up one morning and lookedoutside. She saw that the ground was all wet. Therewere puddles everywhere. ‘‘It must have rained lastnight,’’ she said. Did Sarah predict that it rained,conclude that it rained, or wish that it rained?

Interpret. Sarah has a cat named Fluff. When Sarahwent away on vacation for a few weeks, she left Fluffat home with one of her friends. When Sarah cameback and saw Fluff for the first time, Fluff startedmeowing and meowing at her. ‘‘He’s telling me hemissed me,’’ Sarah said. Did Sarah interpret the me-owing, discover the meowing, or calculate the meowing?

Scenarios for Good/Bad Thinking Task (Study 2)

Note that information about outcomes (shownbelow in parentheses) was presented in the mismatchcondition but omitted in the no outcome condition.

Control Scenarios

Good version. One day Lucy was at school in mathclass when her teacher gave the class a subtractionproblem to work on. The teacher told the class howto start the problem and then he said, ‘‘OK, nowfigure out the answer and write it down in yournotebook.’’ So Lucy had to figure out the answer tothe problem. Here’s what Lucy did. Lucy tried reallyhard to figure out the answer. She really wanted todo it right! She was listening when her teacher talkedto the class, she looked at the problem very carefully,and she really thought about it. So that’s how Lucycame up with her answer. She looked, she listened,and she really tried to figure it out. (It turned out thatLucy got the answer right.)

Bad version. One day Shawn was at school in mathclass when his teacher gave the class a multiplicationproblem to do. The teacher told the class how to startthe problem and then she said, ‘‘OK, now figure outthe answer and write it down on your worksheet.’’So Shawn had to figure out the answer to the prob-lem. Here’s what he did. Shawn didn’t even try tofigure out the answer. He didn’t listen when histeacher was talking. He didn’t even look at theproblem. Shawn just wrote something down for ananswer and he didn’t care if it was right. So that’swhat Shawn did. He didn’t listen, he didn’t look, andhe didn’t even try to figure it out. (It turned out thatShawn got the answer wrong.)

Strategy Scenarios

Good version. Christina got some money for herbirthday to buy herself a camera. So she went to the

store to look for one. When she got there, there weremany different kinds to choose from. So Christinahad to figure out which camera to get. This is whatshe did. Christina wrote down a list of the importantthings that she wanted in a camera. Then she com-pared the cameras at the store to her list to see whichone was the closest match. So that’s how she pickedwhich camera to get. She compared them to her list.(It turned out that Christina didn’t really like thecamera she got. It didn’t work very well.)

Bad version. Eric got some money for his birthdayto buy himself a new bike. So he went to the bikestore to look for one. When he got there, there weremany different kinds to choose from. So Eric had tofigure out which bike to get. This is what he did. Ericwent down the row of bikes and he said ‘‘eeniemeenie miney moe, catch a tiger by the toe.’’Whichever one his finger landed on was the bike hegot. So that’s how he picked which bike to get. Heused ‘‘eenie meenie.’’ (It turned out that Eric reallyliked the bike he got. It worked really well.)

Match version. Joyce got some money for herbirthday to buy herself a CD player. So she went tothe store to look for one. When she got there, therewere many different kinds to choose from. So Joycehad to figure out which CD player to get. This iswhat she did. Joyce wrote down a list of the impor-tant things that she wanted in a CD player. Then shecompared the CD players at the store to her list to seewhich one was the closest match. So that’s how shepicked which CD player to get. She compared themto her list. (It turned out that Joyce really liked theCD player she got. It worked really well.)

Alternatives Scenarios

Good version. Marvin had a nice big box of choco-lates. But one day, Marvin came home after school tofind that somebody had eaten all of his chocolates. Ohno! Who was it? So Marvin had to figure out who hadeaten his chocolates. Here’s what he did. Marvin’svery first thought was that his sister did it. But he didnot go blame it on her right away. No, Marvin tried tothink if there was any way anybody else could haveeaten them. So Marvin didn’t just go with his first idea.He tried to think of all the possibilities, and that’s howhe decided it was his sister. (It turned out that Marvinwas wrong. His brother really ate all the chocolates.)

Bad version. Trisha got some special colored pencilsfor her art class. One day, Trisha came home to findthat somebody had been using her colored pencils andnow some of them were broken. Oh no! Who was it?So Trisha had to figure out who had used her pencils.Here’s what she did. Trisha’s very first thought was

Beliefs about Everyday Reasoning 463

that her brother did it. Trisha did not try to think ifthere was any way anybody else could have done it.She just went with her first thought, and that’s howshe decided it was her brother. (It turned out that shewas right. Her brother had been using her pencils.)

Match version. Erin loves oatmeal for breakfast.But one morning she woke up to find that someonehad eaten up the last of the oatmeal. Oh no! Who ateit? So Erin had to figure out who had eaten the oat-meal. Here is what she did. Erin’s very first thoughtwas that her little sister did it. Erin did not try tothink if there was any way anybody else could havedone it. She just went with her first thought, andthat’s how she decided it was her sister. (It turnedout that Erin was wrong. Her mother really ate allthe oatmeal.)

Pros – Cons Scenarios

Good version. James was sitting on his front porchone afternoon when a girl walked by with some cutelittle kittens in a basket. ‘‘My cat had kittens,’’ shesaid to him, ‘‘and now the kittens are getting big andthey are ready to be adopted. Would you like to havea pet kitten?’’ So James had to figure out whether ornot he should get a kitten. This is what he did. Jamesthought about all the things that would be fun abouthaving a pet kitten. He also thought about all thehard work it would be to take care of the kitten. SoJames thought about both thingsFthe good thingsand the bad things, and that’s how he decided to geta kitten. (It turned out that James wasn’t so happy hegot the kitten. It was a lot of work.)

Bad version. Ruby was walking home one daywhen she saw a sign in front of her neighbor’s house.It said ‘‘Free Puppies.’’ She went up to the house andthere were all these cute, furry little puppies playingin the yard. The man asked Ruby: ‘‘Do you want totake a puppy home with you?’’ So Ruby had to fig-ure out whether or not she should get a puppy. Thisis what she did. Ruby thought about all the thingsthat would be fun about having a puppy. She didn’tthink about any things that would be hard abouttaking care of a puppy. So Ruby only thought aboutthe good things, and that’s how she decided to get apuppy. (It turned out that Ruby was really happy shegot the puppy. It was a lot of fun.)

Match version. Tom was walking by the pet storeone day when he saw a sign that said ‘‘Lizards forSale.’’ Tom went into the pet store and there weresome little green lizards jumping around in a bigtank. The lady at the store asked Tom if he wanted toget one. So Tom had to decide whether or not heshould get a lizard. This is what he did. Tom thought

about all the things that would be fun about having apet lizard. He also thought about all the hard work itwould be to take care of the lizard. So Tom thoughtabout both thingsFthe good things and the badthings, and that’s how he decided to get a lizard. (Itturned out that Tom was really happy he got thelizard. It was a lot of fun.)

Evidence Scenarios

Good version. Liza planted some tomato plants inher garden this year, but they weren’t doing verywell. They started wilting and turning all yellow.Liza wondered what was going onFwhy were herplants getting all yellow like that? She thought of afew different things it could be: Maybe they neededmore sunshine, maybe they needed more water, ormaybe they needed some plant food. So Liza had tofigure out what was wrong with her plants. This iswhat she did. Liza tried changing first one thing andthen the other and she kept track of what helped theplants and what didn’t, and that’s how she decidedthe plants needed more plant food. So Liza usedchanging things and keeping track to decide. (Itturned out that she was wrong. What the plants re-ally needed was more sun.)

Bad version. Robert has a pet fish that lives in a bigfish tank. One day Robert saw some yucky greenstuff growing inside the fish tank. Why was that stuffgrowing in there?, he wondered. He knew of a fewdifferent things it could be: Maybe the water in thetank was too warm, maybe it was getting too muchsunlight, or maybe he needed to clean the filter. SoRobert had to figure out why that green stuff wasgrowing in his fish tank. This is what he did. Roberthad a hunch. He just had a feeling that the tankshouldn’t be getting so much sun. So that’s how hedecided. Robert used a hunch. (It turned out that hewas right. The tank was getting too much sun.)

Match version. Adam has a lilac bush planted in hisyard. One day Adam saw that the lilac bush waslooking kind of funny. The leaves didn’t look veryhealthy and the flowers were drooping. Adamwondered what was going on. He knew of a fewdifferent things it could be: Maybe the lilac bushneeded more water, maybe it needed to have itsbranches trimmed, or maybe it needed fertilizer. SoAdam had to figure out what was wrong with hislilac bush. This is what he did. Adam had a hunch.He just had a feeling that the lilac bush needed tohave its branches trimmed. So that’s how he decided.Adam used a hunch. (It turned out that he waswrong. What the lilac bush really needed was morefertilizer.)

464 Amsterlaw