View in EDS HTML Full Text PDF Full Text

Preschoolers Use Probabilistic Evidence to Flexibly Change or Maintain Expectations on an Active Search Task

Saved in:
Bibliographic Details
Title: Preschoolers Use Probabilistic Evidence to Flexibly Change or Maintain Expectations on an Active Search Task
Language: English
Authors: Brooke C. Hilton (ORCID 0009-0007-2059-751X), Mark A. Sabbagh (ORCID 0000-0003-2938-6288)
Source: Child Development. 2025 96(2):881-890.
Availability: Wiley. Available from: John Wiley & Sons, Inc. 111 River Street, Hoboken, NJ 07030. Tel: 800-835-6770; e-mail: cs-journals@wiley.com; Web site: https://www.wiley.com/en-us
Peer Reviewed: Y
Page Count: 10
Publication Date: 2025
Document Type: Journal Articles
Reports - Research
Descriptors: Preschool Children, Probability, Evidence, Educational Games, Search Strategies, Expectation, Change, Student Behavior
DOI: 10.1111/cdev.14190
ISSN: 0009-3920
1467-8624
Abstract: This study investigated 3- to 5-year-olds' (N = 64, 37 girls, 62.5% White, data collected between 2021-2022) ability to use probabilistic information gleaned through active search to appropriately change or maintain expectations. In an online fishing game, children first learned that one of two ponds was good for catching fish. During a subsequent testing phase, children searched the ponds for fish. Half saw outcomes that were probabilistically consistent with training, and the other half saw outcomes that were probabilistically inconsistent. Children in the Inconsistent condition adapted their search strategies, showing evidence of changing their expectations. Those in the Consistent condition maintained their initial search strategy. Trial-by-trial analyses suggested that children used a combination of heuristic and information integration strategies to guide their search behavior.
Abstractor: As Provided
Notes: https://osf.io/92meh/?view_only=86830b0d6e7143b0925a6c86578a3b8f
Entry Date: 2025
Accession Number: EJ1461456
Database: ERIC
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
    Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwFZCKpPOjS6oFxFOJres-5WAAAA4jCB3wYJKoZIhvcNAQcGoIHRMIHOAgEAMIHIBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDMWbcg2t5X55CG4VpgIBEICBmmpjxRn05A-_38C34yiVzOngV6NXXevfOhggnMRmCrs0tk2n3SGXUcX9ROj6aOSZoqZPuLiJDT4HEfSohc0QuXIXDDM40_K5ST5QAbrQ3rL4_2050-538j93ZnFc6nkpisKw1TbkDiSPh-i5BYhmN1aJ4Cz-fD0ruv0UaLDRrk9-fOR-lacZ2PMAD_B8p-b6xj6fGjruKIbQaU4=
Text:
  Availability: 1
  Value: <anid>AN0183920683;cdv01mar.25;2025Mar24.07:20;v2.2.500</anid> <title id="AN0183920683-1">Preschoolers use probabilistic evidence to flexibly change or maintain expectations on an active search task </title> <p>This study investigated 3‐ to 5‐year‐olds' (N = 64, 37 girls, 62.5% White, data collected between 2021‐2022) ability to use probabilistic information gleaned through active search to appropriately change or maintain expectations. In an online fishing game, children first learned that one of two ponds was good for catching fish. During a subsequent testing phase, children searched the ponds for fish. Half saw outcomes that were probabilistically consistent with training, and the other half saw outcomes that were probabilistically inconsistent. Children in the Inconsistent condition adapted their search strategies, showing evidence of changing their expectations. Those in the Consistent condition maintained their initial search strategy. Trial‐by‐trial analyses suggested that children used a combination of heuristic and information integration strategies to guide their search behavior.</p> <p></p> <ulist> <item> Abbreviations</item> <p></p> <item> AIC Akaike information criterion</item> <p></p> <item> BIC Bayesian information criterion</item> <p></p> <item> CAD Canadian Dollars</item> <p></p> <item> GLME Generalized Linear Mixed‐Effect</item> <p></p> <item> LL log‐likelihood</item> <p></p> <item> WSLS win‐stay, lose‐shift</item> </ulist> <p>In complex environments, individuals may have experiences that run counter to their pre‐existing expectations. Perhaps a teacher who has previously been trustworthy turns out to be inaccurate, or a restaurant that has previously been delicious serves you an overcooked, chewy steak. Sensitivity to these expectation‐inconsistent events is important, as inconsistent information can signal a change environmental structure or inaccuracies in one's current understanding—perhaps the restaurant has changed management and is no longer a source of good food. However, in probabilistic environments, expectation‐inconsistent events may arise occasionally even when one's underlying assumptions about environmental structure are correct—perhaps the teacher was having a bad day and is still a source of good information. In this case, maintaining an accurate representation of the environment requires that one ignores the occasional piece of expectation‐inconsistent information. A primary goal of this study is to provide an initial characterization of preschool‐aged children's abilities to adaptively change or maintain expectations in the face of probabilistic evidence.</p> <p>An emerging empirical and theoretical literature centers the role of children's self‐directed search and exploration as an engine of cognitive development (e.g., Gopnik, [<reflink idref="bib11" id="ref1">11</reflink>]; Schulz, [<reflink idref="bib31" id="ref2">31</reflink>]; Xu, [<reflink idref="bib40" id="ref3">40</reflink>]). Children's capacity for exploration is robust; they explore broadly (Liquin & Gopnik, [<reflink idref="bib19" id="ref4">19</reflink>]), adapt their exploration to the information structure of the task (Ruggeri et al., [<reflink idref="bib29" id="ref5">29</reflink>]; Siegel et al., [<reflink idref="bib33" id="ref6">33</reflink>]), and are at least partially guided by a desire to reduce uncertainty (Blanco & Sloutsky, [<reflink idref="bib5" id="ref7">5</reflink>]; Meder et al., [<reflink idref="bib21" id="ref8">21</reflink>]; Wang et al., [<reflink idref="bib36" id="ref9">36</reflink>]). Children also consider both the strength of their prior expectations and the epistemic quality (e.g., quantity, reliability, representativeness) of the new information when deciding whether to change their mind (Kimura & Gopnik, [<reflink idref="bib13" id="ref10">13</reflink>]; Langenhoff et al., [<reflink idref="bib17" id="ref11">17</reflink>]; Macris & Sobel, [<reflink idref="bib20" id="ref12">20</reflink>]; Schleihauf et al., [<reflink idref="bib30" id="ref13">30</reflink>]). Young children are adept and, at least in some circumstances, selective explorers.</p> <p>Whether preschool‐aged children use expectation‐inconsistent probabilistic information to guide their search behavior is less clear. On one hand, infants show relatively sophisticated reasoning about probabilistic evidence (see Denison & Xu, [<reflink idref="bib8" id="ref14">8</reflink>] for a review) and there is some evidence that they use it to form expectations about likely events in the world. When presented with probabilistic evidence by an experimenter, preschool‐aged children also use that information to revise their expectations, for instance about an object's likely properties (Bonawitz et al., [<reflink idref="bib6" id="ref15">6</reflink>]), or about how to best operate a causal system (Kushnir & Gopnik, [<reflink idref="bib14" id="ref16">14</reflink>], [<reflink idref="bib15" id="ref17">15</reflink>]). During active information search, however, children as young as 6 years old may neglect probability information (Betsch et al., [<reflink idref="bib4" id="ref18">4</reflink>]; Lang & Betsch, [<reflink idref="bib16" id="ref19">16</reflink>]), with young children adjusting their search behavior in response to probabilistic evidence less effectively than adult learners (Plate et al., [<reflink idref="bib25" id="ref20">25</reflink>]). It remains an open question whether, and to what extent, preschool‐aged children use probabilistic information acquired via active search to revise their expectations.</p> <p>A fruitful avenue for assessing children's ability to learn from self‐directed exploration lies in the use of reinforcement learning paradigms. In reinforcement learning paradigms, participants must choose between a series of behavioral options (e.g., slot machines) that vary in terms of their underlying reward structure (von Helversen et al., [<reflink idref="bib35" id="ref21">35</reflink>]). Participants learn which options are most likely to lead to reward by sampling from the options and tracking the observed outcomes over time. A growing body of literature has studied reinforcement learning in children and adolescents (see Nussenbaum & Hartley, [<reflink idref="bib23" id="ref22">23</reflink>] for a review). Adolescents and children over the age of 8 years have been shown to learn from probabilistic evidence, and to use this information to guide exploratory behavior during reinforcement learning (Weiss et al., [<reflink idref="bib38" id="ref23">38</reflink>]; Xia et al., [<reflink idref="bib39" id="ref24">39</reflink>]). To date, however, there is very little research that directly examines how children ages 5 years old and younger learn from reinforcement learning paradigms (Nussenbaum & Hartley, [<reflink idref="bib23" id="ref25">23</reflink>]), and there is especially little work that requires preschool‐aged children to revise expectations based on probabilistic information acquired during reinforcement learning.</p> <p>The current study addresses these gaps in the literature using a paradigm that is similar in premise to the reinforcement learning tasks described above. Preschool‐aged children first completed an expectation‐induction phase in which they were shown two ponds and given evidence that fish might be found in one pond but not the other. Then, in a test phase, they were given 24 trials to search the ponds for fish. Children received probabilistic evidence that was either mostly consistent or mostly inconsistent with what they saw in the induction phase. Our focal, exploratory research question was whether children would use this probabilistic evidence to maintain their initial expectations about where to find fish in the Consistent condition and to change expectations in the Inconsistent condition.</p> <p>There are multiple ways in which children might use probabilistic information to guide search. These strategies vary with respect to their cognitive complexity. By leveraging the multi‐trial design of our paradigm, we then explored the extent to which children's search patterns could be explained with simple heuristic strategies and the extent to which they used a more cognitively complex strategy that required integrating probability information over time. Here, we considered three possible strategies.</p> <p>First, children's choices could be guided by a heuristic win‐stay, lose‐shift (WSLS) strategy, in which they continue fishing from the same pond on successive trials after catching a fish and switch ponds after catching seaweed. This strategy is common in probabilistic reinforcement‐learning environments among both adults and children (e.g., Bonawitz et al., [<reflink idref="bib6" id="ref26">6</reflink>]; Schusterman, [<reflink idref="bib32" id="ref27">32</reflink>]; Weir, [<reflink idref="bib37" id="ref28">37</reflink>]) and is cognitively straightforward because children make choices based solely on their most recent outcome without attending to changes in reward probability over time. Second, children might consider absolute reward frequency, or the total number of fish that they have caught, when deciding where to fish. Using this heuristic, children attend to changes in the total number of rewards over time, but do not consider the rate of "hits" (fish) to "misses" (seaweed) when determining whether a pond is the best place to search. Finally, children might use a more complex cognitive strategy that involves maintaining time‐integrated representations of the relative reward value of each pond. Using this strategy children must actively attend to the <emph>probability</emph> of catching a fish, as compared to the probability of catching seaweed, from each pond and update this value continuously throughout the game. We considered this third approach to be the most sophisticated use of probabilistic information as it is the only strategy that integrates information about both rewards and losses over time.</p> <hd id="AN0183920683-2">METHOD</hd> <p></p> <hd id="AN0183920683-3">Participants</hd> <p>Sixty‐four children ages 42–63 months (F = 37, M = 27, <emph>M</emph><subs>age</subs> = 51.47) were recruited for an online study between January 2021 and September 2022. An additional four children were recruited but not retained for data analysis; 1 experienced technical difficulty, and 3 children did not complete the study. Children were recruited from a database of families located in a mid‐size city in southeastern Ontario and via online advertisements that targeted a broader community (Table 1).</p> <p>1 TABLE Participant demographics.</p> <p> <ephtml> <table><thead valign="bottom"><tr><th align="left">Demographic characteristic</th><th align="left">%</th><th align="left"><italic>n</italic></th></tr></thead><tbody valign="top"><tr><td align="left">Ethnicity</td></tr><tr><td align="left">White</td><td align="char" char=".">62.5</td><td align="char" char=".">40</td></tr><tr><td align="left">Asian</td><td align="char" char=".">10.9</td><td align="char" char=".">7</td></tr><tr><td align="left">White and Asian</td><td align="char" char=".">9.4</td><td align="char" char=".">6</td></tr><tr><td align="left">White and Indigenous/Inuit/Metis</td><td align="char" char=".">7.8</td><td align="char" char=".">5</td></tr><tr><td align="left">White and Latino/Latina</td><td align="char" char=".">3.1</td><td align="char" char=".">2</td></tr><tr><td align="left">White and Other/unspecified</td><td align="char" char=".">3.1</td><td align="char" char=".">2</td></tr><tr><td align="left">Black and Asian</td><td align="char" char=".">1.6</td><td align="char" char=".">1</td></tr><tr><td align="left">Other/unspecified</td><td align="char" char=".">1.6</td><td align="char" char=".">1</td></tr><tr><td align="left">Country of residence</td></tr><tr><td align="left">Canada</td><td align="char" char=".">90.6</td><td align="char" char=".">58</td></tr><tr><td align="left">United States</td><td align="char" char=".">7.8</td><td align="char" char=".">5</td></tr><tr><td align="left">Missing</td><td align="char" char=".">1.6</td><td align="char" char=".">1</td></tr><tr><td align="left">Family income</td></tr><tr><td align="left"><50,000 CAD</td><td align="char" char=".">6.3</td><td align="char" char=".">4</td></tr><tr><td align="left">50,000–74,000 CAD</td><td align="char" char=".">6.3</td><td align="char" char=".">4</td></tr><tr><td align="left">75,000–99,000 CAD</td><td align="char" char=".">14.1</td><td align="char" char=".">9</td></tr><tr><td align="left">100,000–149,000 CAD</td><td align="char" char=".">29.7</td><td align="char" char=".">19</td></tr><tr><td align="left">>150,000 CAD</td><td align="char" char=".">42.2</td><td align="char" char=".">27</td></tr><tr><td align="left">Missing</td><td align="char" char=".">1.6</td><td align="char" char=".">1</td></tr><tr><td align="left">Parent 1 education</td></tr><tr><td align="left">High school/general education diploma (GED)</td><td align="char" char=".">3.1</td><td align="char" char=".">2</td></tr><tr><td align="left">Trades certification</td><td align="char" char=".">1.6</td><td align="char" char=".">1</td></tr><tr><td align="left">Some college (no degree)</td><td align="char" char=".">3.1</td><td align="char" char=".">2</td></tr><tr><td align="left">Associate degree</td><td align="char" char=".">6.3</td><td align="char" char=".">4</td></tr><tr><td align="left">Bachelor's degree</td><td align="char" char=".">39.1</td><td align="char" char=".">25</td></tr><tr><td align="left">Master's degree</td><td align="char" char=".">25</td><td align="char" char=".">16</td></tr><tr><td align="left">Professional degree</td><td align="char" char=".">6.3</td><td align="char" char=".">4</td></tr><tr><td align="left">Doctorate</td><td align="char" char=".">15.6</td><td align="char" char=".">10</td></tr><tr><td align="left">Parent 2 education</td></tr><tr><td align="left">High school/GED</td><td align="char" char=".">10.9</td><td align="char" char=".">7</td></tr><tr><td align="left">Trades certification</td><td align="char" char=".">18.8</td><td align="char" char=".">12</td></tr><tr><td align="left">Some college (no degree)</td><td align="char" char=".">6.3</td><td align="char" char=".">4</td></tr><tr><td align="left">Associate degree</td><td align="char" char=".">7.8</td><td align="char" char=".">5</td></tr><tr><td align="left">Bachelor's degree</td><td align="char" char=".">28.1</td><td align="char" char=".">18</td></tr><tr><td align="left">Master's degree</td><td align="char" char=".">9.4</td><td align="char" char=".">6</td></tr><tr><td align="left">Professional degree</td><td align="char" char=".">6.3</td><td align="char" char=".">4</td></tr><tr><td align="left">Doctorate</td><td align="char" char=".">10.9</td><td align="char" char=".">7</td></tr><tr><td align="left">Missing</td><td align="char" char=".">1.6</td><td align="char" char=".">1</td></tr></tbody></table> </ephtml> </p> <hd id="AN0183920683-4">Procedure</hd> <p>Prior to their online appointment, participants completed a demographic questionnaire (see Supporting Information), as well as two other questionnaires designed to assess the comparability of the two conditions: the Developmental Vocabulary Assessment for Parents (Libertus et al., [<reflink idref="bib18" id="ref29">18</reflink>]), and the Children's Social Understanding Scale (Tahiroglu et al., [<reflink idref="bib34" id="ref30">34</reflink>]). Parents also consented to have their child participate in a subsequent Zoom appointment. Questionnaires were administered through Qualtrics<sups>XM</sups> (Qualtrics, Provo, UT).</p> <p>At the time of the appointment, families were asked to join a Zoom call with an experimenter using a desktop, laptop, or tablet with a webcam. Following a brief warm‐up, children were introduced to the fishing game. The game was designed in Psychopy<sups>3</sups> (Peirce et al., [<reflink idref="bib24" id="ref31">24</reflink>]) and shared using the Zoom screen sharing feature. Parents were asked to sit with their children during the study, but not provide instructions or interfere with their children's performance. Children had the option to make selections either by verbally indicating which pond they wished to choose to the experimenter (e.g., "<emph>the square one</emph>") who subsequently clicked on the chosen pond on the child's behalf, or by clicking on the pond using a mouse/trackpad. Choices were registered on the experimenter's computer. Data were recorded automatically by the Psychopy<sups>3</sups> program for each participant. The experimenter monitored the session in real time to prevent parental interference. Sessions were videorecorded through Zoom.</p> <hd id="AN0183920683-5">Expectation‐induction phase</hd> <p>Task stimuli are shown in Figure 1. In an expectation‐induction phase the experimenter demonstrated fishing three times from each pond. One pond produced three fish (henceforth, the "trained pond"), while the other (henceforth the "alternate pond") produced three pieces of seaweed. Which pond (i.e., "square" or "circle") served as the trained pond was counterbalanced between participants. The desirability of each outcome was vocally reinforced (e.g., <emph>Amazing, it's a fish! Oh no, it's seaweed!</emph>). At the end of the induction phase, the experimenter reiterated the outcomes from each pond (e.g., <emph>We caught three fish from the circle pond and three seaweeds from the square pond</emph>). To ensure that the induction phase was successful, children were then asked to indicate which pond was best for finding fish. 90.6% answered correctly. Those who answered incorrectly were reminded of the outcomes from each pond and asked again. All children except one, whose response could not be verbally determined, answered this second question correctly. Removing this child from analyses did not change the pattern of results, and thus the child has been retained (see Open Science Link).</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/CDV/01mar25/cdev14190-fig-0001.jpg?ephost1=dGJyMNHX8kSepq84v%2bvlOLCmsE6epq5Srqa4SK6WxWXS" alt="cdev14190-fig-0001.jpg" title="1 Illustration of the first trial of the fishing game. (a) Children selected a pond. (b) Fisherman Fred appeared onscreen and fished from the chosen pond. (c) A results screen highlighted the outcome of the fishing decision. (d) Outcomes from past trials were stored above their respective pond." /> </p> <p></p> <hd id="AN0183920683-7">Testing phase</hd> <p>Following the induction phase, children were told that Fisherman Fred only wanted to catch fish, and that the child's selections would determine which pond he would try. Across 4 blocks of 6 trials each (24 trials total), children selected a pond to fish from. The experimenter commented on each outcome's desirability. Between blocks, children were reminded of their task objectives (<emph>Remember, he only wants fish ... so try to catch as many fish as you can</emph>).</p> <p>The critical manipulation concerned which of the two ponds was optimal for catching fish during the testing phase. Children were randomly assigned to one of two conditions. In the <emph>Consistent</emph> condition, the optimal pond to search was the trained pond from the expectation‐induction phase. In the <emph>Inconsistent</emph> condition, the optimal pond to search during the testing phase was the pond that had previously produced only seaweed. In both cases, the optimal pond produced fish probabilistically, 75% of the time, while the non‐optimal pond produced fish on 25% of trials. The sequence of fish and seaweed that could be caught from each pond was the same for all participants (see Supporting Information), though, because participants made their own choices about where to fish, their specific experiences varied substantially.</p> <hd id="AN0183920683-8">End‐of‐task belief assessment</hd> <p>After 24 trials of the fishing game, children were informed that Fisherman Fred could only try one more pond and that he <emph>really</emph> wanted to catch the last fish. Children's pond selection on this final trial was thought to reflect their end‐of‐task beliefs about which pond was best for finding fish.</p> <hd id="AN0183920683-9">RESULTS</hd> <p>Conditions did not differ in age, vocabulary, or social understanding score (<emph>p</emph>s > .05, see S1). The data and analytic code necessary to reproduce these analyses have been made available at the URL: https://osf.io/92meh/?view_only=86830b0d6e7143b0925a6c86578a3b8f.</p> <hd id="AN0183920683-10">Preliminary analyses</hd> <p>Binomial tests revealed that both conditions chose the trained pond at above‐chance levels on the first trial (Con: 29/32, <emph>p</emph> < .001, Incon: 24/32, <emph>p</emph> = .007) and the distribution of choices did not significantly differ by condition (<emph>χ</emph><sups>2</sups>(<reflink idref="bib1" id="ref32">1</reflink>) = 1.76, <emph>p</emph> = .19). By the end‐of‐task belief assessment, children's choice of the trained pond differed across conditions (<emph>χ</emph><sups>2</sups>(<reflink idref="bib1" id="ref33">1</reflink>) = 20.27, <emph>p</emph> < .001). Children in the Consistent condition continued to select the trained pond at above‐chance levels (Con: 25/32, <emph>p =</emph> .002) while children in the Inconsistent condition did so at below‐chance levels (Incon: 6/32, <emph>p</emph> < .001). Unsurprisingly, given that the available evidence supported their initial expectations, children in the Consistent condition also caught more fish than those in the Inconsistent condition (<emph>M</emph><subs>Con</subs> = 14.66, <emph>M</emph><subs>Incon</subs> = 13.22, <emph>t</emph>(<reflink idref="bib62" id="ref34">62</reflink>) = −3.93, <emph>p</emph> < .001) (Figure 2).</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/CDV/01mar25/cdev14190-fig-0002.jpg?ephost1=dGJyMNHX8kSepq84v%2bvlOLCmsE6epq5Srqa4SK6WxWXS" alt="cdev14190-fig-0002.jpg" title="2 Number of children selecting the trained pond. Dashed line marks chance‐level responding." /> </p> <p></p> <hd id="AN0183920683-12">Modeling children's choices</hd> <p>To gain further insights into children's performance, we used a Generalized Linear Mixed‐Effects Model (GLME) with the <emph>glmer</emph> function from the <emph>lme4</emph> package (Bates et al., [<reflink idref="bib1" id="ref35">1</reflink>]) in R (R Core Team, [<reflink idref="bib28" id="ref36">28</reflink>]) and RStudio (Posit Team, [<reflink idref="bib27" id="ref37">27</reflink>]) to characterize how children's choices changed over time in the two conditions. Using a standard Rescorla‐Wagner algorithm, we first calculated the expected "reward value" of each pond on any given trial given the child's previous choices and outcomes (see Open Science Linkfor code). Then, for every trial, we coded whether participants chose the pond with the highest expected reward value (Figure 3). We first assessed a full model in which children's tendencies to choose the highest‐value pond were predicted by the fixed effects of Trial (centered), Condition (effects‐coded), Age (centered), and their interactions, with random effects of Trial for each participant. This random effects structure resulted in a singular fit, and so the error term was simplified to include only random intercepts for each participant. Following visual inspection of the data, we modeled Trial with linear and quadratic functions. Including Age and its interactions did not significantly improve the fit of the model, and so it was excluded.</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/CDV/01mar25/cdev14190-fig-0003.jpg?ephost1=dGJyMNHX8kSepq84v%2bvlOLCmsE6epq5Srqa4SK6WxWXS" alt="cdev14190-fig-0003.jpg" title="3 Children's probability of choosing the optimal value pond as a function of trial and condition. (a) The observed proportions of children making optimal value choices on each trial in the Consistent (blue) and Inconsistent (red) conditions. (b) The model predicted probabilities of optimal value choices. Lighter lines represent individual participants' model fits, and the bold lines represent the average model prediction." /> </p> <p></p> <p>This parsimonious model (Akaike information criterion [AIC]: 1883.7, Bayesian information criterion [BIC]: 1921.0, log‐likelihood [LL]: −934.8) showed a significant main effect of Condition (<emph>b</emph> = 0.56, SE = 0.09, <emph>z</emph> = 6.37, <emph>p</emph> < .001), significant main effects for the linear (<emph>b</emph> = 10.48, SE = 2.29, <emph>z</emph> = 4.58, <emph>p</emph> < .001) and quadratic (<emph>b</emph> = 7.04, SE = 2.27, <emph>z</emph> = 3.09, <emph>p</emph> = .002) Trial terms, and a significant interaction between Condition and the linear effect of Trial (<emph>b</emph> = −7.09, SE = 2.29, <emph>z</emph> = −3.10, <emph>p</emph> = .002). Follow‐up analyses for the interaction showed that children's selection of optimal value ponds changed very little over trials in the Consistent condition (<emph>b_</emph><subs>linear</subs> = 2.49, SE = 2.45, <emph>z =</emph> 1.02, <emph>p</emph> = .31), but increased over trials in the Inconsistent condition (<emph>b_</emph><subs>linear</subs> = 12.02, SE = 2.12, <emph>z =</emph> 5.66, <emph>p</emph> < .001). The raw data and the model predictions are presented in Figure 3a,b.</p> <hd id="AN0183920683-14">Mechanisms accounting for change over trials</hd> <p>For the next set of analyses, we aimed to determine whether children's fishing behavior was guided by simple heuristic strategies and whether children were actively maintaining complex, time‐integrated representations about the probability of catching a fish from each pond. We assessed three possible proximal factors. The first of these factors was a WSLS heuristic where children stayed at the same pond after catching a fish and switched ponds after catching seaweed. Second, children's choices could have been guided by absolute reward frequency, where they attended solely to the number of times that they had caught a fish from the trained pond when determining where to search. Finally, children might have used a more complex cognitive strategy that involved maintaining time‐integrated representations of the relative reward value, or probability of catching a fish, from each pond.</p> <p>To assess each of these possibilities, we first created variables that predicted children's pond choices if they were following any of these three strategies and submitted them all as predictors in a GLME analysis. The WSLS predictor was calculated as the interaction between a child's pond selection on the previous trial (1 = trained, −1 = alternate) and its resulting outcome (1 = reward, −1 = no reward). Multiplications resulting in a value of 1 predict selecting the trained pond (either by win‐stay or lose‐shift) on the current trial. The reward frequency predictor was coded as the total number of fish that children had caught from the trained pond and accumulated from trial‐to‐trial. Finally, reward probability was coded as the difference between the relative reward value of the trained pond and that of the alternate pond as generated by the reinforcement‐learning model described previously. Positive values for this variable represent higher reward values for the trained, as opposed to the alternate, pond, while negative values represent higher reward values for the alternate pond as compared to the trained pond.</p> <p>The GLME was conducted with each of these three factors (WSLS, Reward Frequency, Reward Probability), Age, and Age by factor interactions as fixed effects, and participant intercepts entered as a random effect. The model (AIC = 1800.4, BIC = 1848.5, LL = −891.2) showed evidence for significant independent main effects of WSLS (<emph>b</emph> = 0.44, SE = 0.06, <emph>z</emph> = 6.88, <emph>p</emph> < .001), Reward Frequency (<emph>b</emph> = 0.09, SE = 0.02, <emph>z</emph> = 4.66, <emph>p</emph> < .001), Reward Probability (<emph>b</emph> = 1.03, SE = 0.18, <emph>z</emph> = 5.86, <emph>p</emph> < .001), and Age (<emph>b</emph> = 0.03, SE = 0.02, <emph>z</emph> = 2.06, <emph>p</emph> = .04). There was also a significant interaction between WSLS and Age (<emph>b</emph> = 0.05, SE = 0.01, <emph>z</emph> = 3.42, <emph>p</emph> < .001). The full model significantly outperformed all simpler models when compared using a likelihood ratio test (see Open Science Link for code). Further investigation of the Age and WSLS interaction revealed that the tendency to win‐stay at the trained pond (<emph>b</emph> = 0.13, SE = 0.05, <emph>z</emph> = 2.63, <emph>p</emph> = .009), but not the tendency to lose‐shift or the tendency to win‐stay at the alternate pond (<emph>p</emph>s > .05, see Open Science Link) was related to age (Figure 4).</p> <p> <img src="https://imageserver.ebscohost.com/img/embimages/rdk/CDV/01mar25/cdev14190-fig-0004.jpg?ephost1=dGJyMNHX8kSepq84v%2bvlOLCmsE6epq5Srqa4SK6WxWXS" alt="cdev14190-fig-0004.jpg" title="4 Model‐predicted probability of demonstrating a win‐stay or lose‐shift strategy by age. Gray lines show the model‐predicted probability of choosing the trained pond given a participant's last choice (trained or alternate pond) and last outcome (fish or seaweed). Dashed lines indicate whether responses on the subsequent trial represent a tendency to "shift" options or to "stay" at the same pond." /> </p> <p></p> <p>The relative role of each factor on children's choice behavior was assessed with an exploratory descriptive analysis using model comparisons. For each factor (WSLS, Reward Probability, and Reward Frequency), a full model containing all main effects was compared to a reduced model excluding the target factor. Results suggested that inclusion of WSLS had the largest impact on model fit, followed by Reward Probability, and Reward Frequency (Table 2).</p> <p>2 TABLE Change in model fit statistics by main effect.</p> <p> <ephtml> <table><thead valign="bottom"><tr><th align="left" /><th align="left">ΔAIC</th><th align="left">ΔBIC</th><th align="left">ΔLL</th><th align="left"><italic>χ</italic><sup>2</sup></th><th align="left"><italic>p</italic>‐Value</th></tr></thead><tbody valign="top"><tr><td align="left">Win‐stay, lose‐shift</td><td align="char" char=".">−44.1</td><td align="char" char=".">−38.8</td><td align="char" char=".">23.06</td><td align="char" char=".">46.12</td><td align="char" char="."><.001***</td></tr><tr><td align="left">Reward prob.</td><td align="char" char=".">−31.8</td><td align="char" char=".">−26.5</td><td align="char" char=".">16.87</td><td align="char" char=".">33.75</td><td align="char" char="."><.001***</td></tr><tr><td align="left">Abs. frequency</td><td align="char" char=".">−18.8</td><td align="char" char=".">−13.5</td><td align="char" char=".">10.37</td><td align="char" char=".">20.75</td><td align="char" char="."><.001***</td></tr></tbody></table> </ephtml> </p> <p>1 <emph>Note</emph>: Values represent improvements in Akaike information criterion (AIC), Bayesian information criterion (BIC), and log‐likelihood (LL) fit statistics when the factor of interest is added to the model as compared to a main effect model with the target factor excluded. Larger negative values for ΔAIC and ΔBIC represent greater improvements in model fit. Larger positive values for ΔLL represent greater improvements in model fit. *** indicates <emph>p</emph> <.001, two‐tailed.</p> <hd id="AN0183920683-16">DISCUSSION</hd> <p>The present study used a reinforcement learning task to evaluate children's ability to use probabilistic information to guide their exploratory behavior. Our results suggest that challenges associated with probabilistic evidence were readily negotiated by preschool‐aged children who used a combination of probability integration and heuristic strategies to guide their search behavior.</p> <hd id="AN0183920683-17">Maintaining and changing expectations in response to probabilistic evidence</hd> <p>Past work has suggested that while young children are sensitive to probability information (Denison & Xu, [<reflink idref="bib8" id="ref38">8</reflink>]), they neglect to use this information during search (Betsch et al., [<reflink idref="bib4" id="ref39">4</reflink>]; Lang & Betsch, [<reflink idref="bib16" id="ref40">16</reflink>]; Plate et al., [<reflink idref="bib25" id="ref41">25</reflink>]). The current study demonstrates that preschoolers as young as 3.5 years of age adapt behavior and change expectations in response to probabilistic evidence uncovered via their own searches. Children in the Consistent condition saw multiple pieces of evidence that could have undermined their initial expectation that the trained pond was the best place to find fish, yet they nonetheless persisted in selecting the trained pond more frequently than the alternate. In the Inconsistent condition, however, children changed their search strategy to systematically search the alternate pond.</p> <p>Methodological differences may explain the discrepancy between these results and those of past literature. The reward values between ponds in this study were quite disparate (i.e., 75% vs. 25% chance of catching fish). Other studies have used probability values that are less disparate (e.g., 50% vs. 67%), and/or search environments with more than two options (Betsch et al., [<reflink idref="bib4" id="ref42">4</reflink>]; Lang & Betsch, [<reflink idref="bib16" id="ref43">16</reflink>]; Plate et al., [<reflink idref="bib25" id="ref44">25</reflink>]). If children's ability to consider probability during search is still emerging during the preschool period, then it is possible that children will only show this ability on relatively simple probability tasks. More complex methodologies, such as those with more than two search options or those with less disparate reward values across options, may exceed young children's capacity for probability integration during search.</p> <hd id="AN0183920683-18">Mechanisms of change in search behavior</hd> <p>When deciding where to search, children relied, at least in part, on a cognitively complex probability integration strategy that required them to maintain up‐to‐date representations of the likelihood that they would catch a fish from each pond. Previous work has shown that preschool‐aged children adapt their exploratory strategies to the statistical structure of a task (Ruggeri et al., [<reflink idref="bib29" id="ref45">29</reflink>]; Siegel et al., [<reflink idref="bib33" id="ref46">33</reflink>]). Our results extend this work by revealing that children use a combination of heuristic and probability‐integration strategies when deciding where to search.</p> <p>In addition to probability‐integration, children also relied on heuristic strategies such as reward frequency and WSLS. Indeed, exploratory model comparisons revealed that WSLS may have had the largest overall effect on children's search decisions. We initially reasoned that these heuristics might be a less sophisticated use of probabilistic information. Nonetheless, these strategies may have been employed by children because they are cognitively efficient. In noisy environments, such as the testing phase of the current study, WSLS in particular may allow individuals to efficiently balance the demands for behavioral flexibility and the desire for rewarding outcomes (e.g., Nowak & Sigmund, [<reflink idref="bib22" id="ref47">22</reflink>]; Posch, [<reflink idref="bib26" id="ref48">26</reflink>]). Using heuristic strategies, children can efficiently glean rewards in a probabilistic environment without expending the additional cognitive resources that are required to maintain up‐to‐date probabilistic representations.</p> <hd id="AN0183920683-19">Limitations and future directions</hd> <p>When using WSLS, older children were more likely than younger children to stay at the trained pond after a win. This is consistent with past literature which has found that children, and adults under cognitive load, tend to switch following a loss more often than they stay after a win (e.g., Berman et al., [<reflink idref="bib2" id="ref49">2</reflink>]; Garon & English, [<reflink idref="bib10" id="ref50">10</reflink>]; Ivan et al., [<reflink idref="bib12" id="ref51">12</reflink>]). Some researchers have speculated that "switching" between response options might be a default response which children need to inhibit to search successfully (Ivan et al., [<reflink idref="bib12" id="ref52">12</reflink>]). The protracted development of cognitive control abilities (Best & Miller, [<reflink idref="bib3" id="ref53">3</reflink>]; Carlson, [<reflink idref="bib7" id="ref54">7</reflink>]) may make it especially difficult for young children to inhibit the desire to switch responses. Future work may wish to consider the role of individual differences in cognitive control as they pertain to preschool children's use of WSLS.</p> <p>The present study required children to play an online computer game. It is plausible that access to computerized games affected children's performance. Participants in this study were sampled from English‐speaking, majority‐White, and middle socioeconomic status populations. These factors have been shown to impact children's exposure to at‐home computers or tablets (e.g., see Dolan, [<reflink idref="bib9" id="ref55">9</reflink>] for a review). Future work is required to replicate these findings in non‐WEIRD and lower SES populations to further examine the generalizability of these results.</p> <p>An assumption of this paradigm is that children's performance on laboratory tasks is informative about how they change expectations and explore in the real world. An open question in this study is whether children's search behavior and their tendency to change search strategies in the Inconsistent condition, reflects an underlying change in children's beliefs about reward structure. On one hand, children's use of heuristic search strategies does not require them to hold broader beliefs about the reward structure of the system. On the other hand, children's use of a probability integration strategy suggests that they were attending to changes in reward values over time. Anecdotally, children also sometimes verbalized beliefs spontaneously (e.g., by saying "<emph>The circle one has more fish</emph>"). Future work may wish to examine the relationship between children's expectations and their search decisions by combining behavioral tasks with physiological measures (e.g., brain electrophysiology, pupil dilation) that can more directly index the extent to which children's search decisions are driven by changes in their expectations.</p> <hd id="AN0183920683-20">CONCLUSION</hd> <p>The current study provides evidence that preschool‐aged children use probabilistic evidence uncovered via active search to maintain expectations that are often supported, and to change expectations that are broadly unsupported, by new information. When searching for information in a probabilistic environment, children relied on a combination of heuristic and cognitively complex strategies that required them to reason about changes in reward probability. Results provide evidence that even preschool‐aged children can integrate up‐to‐date information about the probability of outcomes during active search and that they will use this information to maintain or change an expectation about where to search for rewards.</p> <hd id="AN0183920683-21">AUTHOR CONTRIBUTIONS</hd> <p>Testing and data collection were performed by Brooke C. Hilton. Brooke C. Hilton performed the data analysis and interpretation under the supervision of Mark A. Sabbagh. All authors were involved in the drafting and revision of the manuscript, and all authors approved the final version of the manuscript for submission.</p> <hd id="AN0183920683-22">ACKNOWLEDGMENTS</hd> <p>We thank participating families for their participation in this study. We also wish to thank members of the Early Experience Laboratory at Queen's University for their thoughtful comments and support during the drafting of this manuscript.</p> <hd id="AN0183920683-23">FUNDING INFORMATION</hd> <p>This work was funded by the Natural Sciences and Engineering Research Council of Canada (Grant Number: RGPIN‐2018‐05200) and the Alexander Graham Bell Canada Graduate Doctoral Scholarship (Award Number: CGSD3‐559518‐2021).</p> <hd id="AN0183920683-24">CONFLICT OF INTEREST STATEMENT</hd> <p>The authors report no known conflicts of interest.</p> <hd id="AN0183920683-25">DATA AVAILABILITY STATEMENT</hd> <p>The data and analytic code necessary to reproduce the analyses presented here are publicly accessible. Data and code are available at the following URL: https://osf.io/92meh/?view%5fonly=86830b0d6e7143b0925a6c86578a3b8f. The materials necessary to attempt to replicate the findings presented here are publicly accessible. Materials can be found in the corresponding S1 file. The analyses presented here were not preregistered.</p> <p>GRAPH: Data S1.</p> <ref id="AN0183920683-26"> <title> REFERENCES </title> <blist> <bibl id="bib1" idref="ref32" type="bt">1</bibl> <bibtext> Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed‐effects models using lme4. Journal of Statistical Software, 67, 1 – 48. https://doi.org/10.18637/jss.v067.i01</bibtext> </blist> <blist> <bibl id="bib2" idref="ref49" type="bt">2</bibl> <bibtext> Berman, P. W., Rane, N. G., & Bahow, E. (1970). Age changes in children's learning set with win‐stay, lose‐shift problems. Developmental Psychology, 2, 233 – 239. https://doi.org/10.1037/h0028753</bibtext> </blist> <blist> <bibl id="bib3" idref="ref53" type="bt">3</bibl> <bibtext> Best, J. R., & Miller, P. H. (2010). A developmental perspective on executive function. Child Development, 81, 1641 – 1660. https://doi.org/10.1111/j.1467‐8624.2010.01499.x</bibtext> </blist> <blist> <bibl id="bib4" idref="ref18" type="bt">4</bibl> <bibtext> Betsch, T., Lehmann, A., Lindow, S., Lang, A., & Schoemann, M. (2016). Lost in search: (mal‐)adaptation to probabilistic decision environments in children and adults. Developmental Psychology, 52, 311 – 325. https://doi.org/10.1037/dev0000077</bibtext> </blist> <blist> <bibl id="bib5" idref="ref7" type="bt">5</bibl> <bibtext> Blanco, N. J., & Sloutsky, V. M. (2021). Systematic exploration and uncertainty dominate young children's choices. Developmental Science, 24, e13026. https://doi.org/10.1111/desc.13026</bibtext> </blist> <blist> <bibl id="bib6" idref="ref15" type="bt">6</bibl> <bibtext> Bonawitz, E. B., Denison, S., Gopnik, A., & Griffiths, T. L. (2014). Win‐stay, lose‐sample: A simple sequential algorithm for approximating Bayesian inference. Cognitive Psychology, 74, 35 – 65. https://doi.org/10.1016/j.cogpsych.2014.06.003</bibtext> </blist> <blist> <bibl id="bib7" idref="ref54" type="bt">7</bibl> <bibtext> Carlson, S. M. (2005). Developmentally sensitive measures of executive function in preschool children. Developmental Neuropsychology, 28, 595 – 616. https://doi.org/10.1207/s15326942dn2802_3</bibtext> </blist> <blist> <bibl id="bib8" idref="ref14" type="bt">8</bibl> <bibtext> Denison, S., & Xu, F. (2019). Infant statisticians: The origins of reasoning under uncertainty. Perspectives on Psychological Science, 14, 499 – 509. https://doi.org/10.1177/1745691619847201</bibtext> </blist> <blist> <bibl id="bib9" idref="ref55" type="bt">9</bibl> <bibtext> Dolan, J. E. (2015). Splicing the Divide: A Review of Research on the Evolving Digital Divide Among K–12 Students. Journal of Research on Technology in Education, 48 (1), 16 – 37. https://doi.org/10.1080/15391523.2015.1103147</bibtext> </blist> <blist> <bibtext> Garon, N. M., & English, S. D. (2022). Heterogeneity of decision‐making strategies for preschoolers on a variant of the IGT. Applied Neuropsychology: Child, 11, 811 – 824. https://doi.org/10.1080/21622965.2021.1973470</bibtext> </blist> <blist> <bibtext> Gopnik, A. (2020). Childhood as a solution to explore–exploit tensions. Philosophical Transactions of the Royal Society B, 375, 20190502. https://doi.org/10.1098/rstb.2019.0502</bibtext> </blist> <blist> <bibtext> Ivan, V. E., Banks, P. J., Goodfellow, K., & Gruber, A. J. (2018). Lose‐shift responding in humans is promoted by increased cognitive load. Frontiers in Integrative Neuroscience, 12, 9. https://doi.org/10.3389/fnint.2018.00009</bibtext> </blist> <blist> <bibtext> Kimura, K., & Gopnik, A. (2019). Rational higher‐order belief revision in young children. Child Development, 90, 91 – 97. https://doi.org/10.1111/cdev.13143</bibtext> </blist> <blist> <bibtext> Kushnir, T., & Gopnik, A. (2005). Young children infer causal strength from probabilities and interventions. Psychological Science, 16, 678 – 683. https://doi.org/10.1111/j.1467‐9280.2005.01595.x</bibtext> </blist> <blist> <bibtext> Kushnir, T., & Gopnik, A. (2007). Conditional probability versus spatial contiguity in causal learning: Preschoolers use new contingency evidence to overcome prior spatial assumptions. Developmental Psychology, 43, 186 – 196. https://doi.org/10.1037/0012‐1649.43.1.186</bibtext> </blist> <blist> <bibtext> Lang, A., & Betsch, T. (2018). Children's neglect of probabilities in decision making with and without feedback. Frontiers in Psychology, 9, 191. https://doi.org/10.3389/fpsyg.2018.00191</bibtext> </blist> <blist> <bibtext> Langenhoff, A. F., Engelmann, J. M., & Srinivasan, M. (2023). Children's developing ability to adjust their beliefs reasonably in light of disagreement. Child Development, 94, 44 – 59. https://doi.org/10.1111/cdev.13838</bibtext> </blist> <blist> <bibtext> Libertus, M. E., Odic, D., Feigenson, L., & Halberda, J. (2015). A developmental vocabulary assessment for parents (DVAP): Validating parental report of vocabulary size in 2‐ to 7‐year‐old children. Journal of Cognition and Development, 16, 442 – 454. https://doi.org/10.1080/15248372.2013.835312</bibtext> </blist> <blist> <bibtext> Liquin, E. G., & Gopnik, A. (2022). Children are more exploratory and learn more during an approach‐avoid task. Cognition, 218, 104940. https://doi.org/10.1016/j.cognition.2021.104940</bibtext> </blist> <blist> <bibtext> Macris, D. M., & Sobel, D. M. (2017). The role of evidence diversity and explanation in 4‐ and 5‐ year‐olds' resolution of counterevidence. Journal of Cognition and Development, 18, 358 – 374. https://doi.org/10.1080/15248372.2017.1323755</bibtext> </blist> <blist> <bibtext> Meder, B., Wu, C. M., Schulz, E., & Ruggeri, A. (2021). Development of directed and random exploration in children. Developmental Science, 24, e13095. https://doi.org/10.1111/desc.13095</bibtext> </blist> <blist> <bibtext> Nowak, M., & Sigmund, K. (1995). A strategy of win‐stay, lose‐shift that outperforms tit‐for‐tat in the Prisoner's dilemma game. Nature, 364, 56 – 58. https://doi.org/10.1038/364056a0</bibtext> </blist> <blist> <bibtext> Nussenbaum, K., & Hartley, C. A. (2019). Reinforcement learning across development: What insights can we draw from a decade of research? Developmental Cognitive Neuroscience, 40, 100733. https://doi.org/10.1016/j.dcn.2019.100733</bibtext> </blist> <blist> <bibtext> Peirce, J. W., Gray, J. R., Simpson, S., MacAskill, M. R., Höchenberger, R., Sogo, H., Kastman, E., & Lindeløv, J. (2019). PsychoPy2: experiments in behavior made easy. Behavior Research Methods, 51, 195 – 203. https://doi.org/10.3758/s13428‐018‐01193‐y</bibtext> </blist> <blist> <bibtext> Plate, R. C., Fulvio, J. M., Shutts, K., Green, C. S., & Pollak, S. D. (2018). Probability learning: Changes in behavior across time and development. Child Development, 89, 205 – 218. https://doi.org/10.1111/cdev.12718</bibtext> </blist> <blist> <bibtext> Posch, M. (1999). Win‐stay, lose‐shift strategies for repeated games—Memory length, aspiration levels and noise. Journal of Theoretical Biology, 198, 183 – 195. https://doi.org/10.1006/jtbi.1999.0909</bibtext> </blist> <blist> <bibtext> Posit Team. (2023). RStudio: Integrated development environment for R. Posit Software, PBC. <ulink href="http://www.posit.co/">http://www.posit.co/</ulink></bibtext> </blist> <blist> <bibtext> R Core Team. (2023). R: A language and environment for statistical computing. R Foundation for Statistical Computing. https://www.R‐project.org/</bibtext> </blist> <blist> <bibtext> Ruggeri, A., Swaboda, N., Sim, Z. L., & Gopnik, A. (2019). Shake it baby, but only when needed: Preschoolers adapt their exploratory strategies to the information structure of the task. Cognition, 193, 104013. https://doi.org/10.1016/j.cognition.2019.104013</bibtext> </blist> <blist> <bibtext> Schleihauf, H., Herrmann, E., Fischer, J., & Engelmann, J. M. (2022). How children revise their beliefs in light of reasons. Child Development, 93, 1072 – 1089. https://doi.org/10.1111/cdev.13758</bibtext> </blist> <blist> <bibtext> Schulz, L. (2012). The origins of inquiry: Inductive inference and exploration in early childhood. Trends in Cognitive Sciences, 16, 382 – 389. https://doi.org/10.1016/j.tics.2012.06.004</bibtext> </blist> <blist> <bibtext> Schusterman, R. J. (1963). The use of strategies in 2‐choice behavior of children and chimpanzees. Journal of Comparative and Physiological Psychology, 56, 96 – 100. https://doi.org/10.1037/h0040054</bibtext> </blist> <blist> <bibtext> Siegel, M. H., Magin, R. W., Pelz, M., Tenenbaum, J. B., & Schulz, L. E. (2021). Children's exploratory play tracks the discriminability of hypotheses. Nature Communications, 12, 3598. https://doi.org/10.1038/s41467‐021‐23431‐2</bibtext> </blist> <blist> <bibtext> Tahiroglu, D., Moses, L. J., Carlson, S. M., Mahy, C. E., Olofson, E. L., & Sabbagh, M. A. (2014). The children's social understanding scale: Construction and validation of a parent‐report measure for assessing individual differences in children's theories of mind. Developmental Psychology, 50, 2485 – 2497. https://doi.org/10.1037/a0037914</bibtext> </blist> <blist> <bibtext> von Helversen, B., Mata, R., Samanez‐Larkin, G. R., & Wilke, A. (2018). Foraging, exploration, or search? On the (lack of) convergent validity between three behavioral paradigms. Evolutionary Behavioral Sciences, 12, 152 – 162. https://doi.org/10.1037/ebs0000121</bibtext> </blist> <blist> <bibtext> Wang, J., Yang, Y., Macias, C., & Bonawitz, E. (2021). Children with more uncertainty in their intuitive theories seek domain‐relevant information. Psychological Science, 32, 1147 – 1156. https://doi.org/10.1177/0956797621994230</bibtext> </blist> <blist> <bibtext> Weir, M. W. (1964). Developmental changes in problem‐solving strategies. Psychological Review, 71, 473 – 490. https://doi.org/10.1037/h0041785</bibtext> </blist> <blist> <bibtext> Weiss, E. O., Kruppa, J. A., Fink, G. R., Herpertz‐Dahlmann, B., Konrad, K., & Schulte‐Rüther, M. (2021). Developmental differences in probabilistic reversal learning: A computational modeling approach. Frontiers in Neuroscience, 14, 536596. https://doi.org/10.3389/fnins.2020.536596</bibtext> </blist> <blist> <bibtext> Xia, L., Master, S. L., Eckstein, M. K., Baribault, B., Dahl, R. E., Wilbrecht, L., & Collins, A. G. E. (2021). Modeling changes in probabilistic reinforcement learning during adolescence. PLoS Computational Biology, 17, e1008524. https://doi.org/10.1371/journal.pcbi.1008524</bibtext> </blist> <blist> <bibtext> Xu, F. (2019). Towards a rational constructivist theory of cognitive development. Psychological Review, 126, 841 – 864. https://doi.org/10.1037/rev0000153</bibtext> </blist> </ref> <aug> <p>By Brooke C. Hilton and Mark A. Sabbagh</p> <p>Reported by Author; Author</p> </aug> <nolink nlid="nl1" bibid="bib11" firstref="ref1"></nolink> <nolink nlid="nl2" bibid="bib31" firstref="ref2"></nolink> <nolink nlid="nl3" bibid="bib40" firstref="ref3"></nolink> <nolink nlid="nl4" bibid="bib19" firstref="ref4"></nolink> <nolink nlid="nl5" bibid="bib29" firstref="ref5"></nolink> <nolink nlid="nl6" bibid="bib33" firstref="ref6"></nolink> <nolink nlid="nl7" bibid="bib21" firstref="ref8"></nolink> <nolink nlid="nl8" bibid="bib36" firstref="ref9"></nolink> <nolink nlid="nl9" bibid="bib13" firstref="ref10"></nolink> <nolink nlid="nl10" bibid="bib17" firstref="ref11"></nolink> <nolink nlid="nl11" bibid="bib20" firstref="ref12"></nolink> <nolink nlid="nl12" bibid="bib30" firstref="ref13"></nolink> <nolink nlid="nl13" bibid="bib14" firstref="ref16"></nolink> <nolink nlid="nl14" bibid="bib15" firstref="ref17"></nolink> <nolink nlid="nl15" bibid="bib16" firstref="ref19"></nolink> <nolink nlid="nl16" bibid="bib25" firstref="ref20"></nolink> <nolink nlid="nl17" bibid="bib35" firstref="ref21"></nolink> <nolink nlid="nl18" bibid="bib23" firstref="ref22"></nolink> <nolink nlid="nl19" bibid="bib38" firstref="ref23"></nolink> <nolink nlid="nl20" bibid="bib39" firstref="ref24"></nolink> <nolink nlid="nl21" bibid="bib32" firstref="ref27"></nolink> <nolink nlid="nl22" bibid="bib37" firstref="ref28"></nolink> <nolink nlid="nl23" bibid="bib18" firstref="ref29"></nolink> <nolink nlid="nl24" bibid="bib34" firstref="ref30"></nolink> <nolink nlid="nl25" bibid="bib24" firstref="ref31"></nolink> <nolink nlid="nl26" bibid="bib62" firstref="ref34"></nolink> <nolink nlid="nl27" bibid="bib28" firstref="ref36"></nolink> <nolink nlid="nl28" bibid="bib27" firstref="ref37"></nolink> <nolink nlid="nl29" bibid="bib22" firstref="ref47"></nolink> <nolink nlid="nl30" bibid="bib26" firstref="ref48"></nolink> <nolink nlid="nl31" bibid="bib10" firstref="ref50"></nolink> <nolink nlid="nl32" bibid="bib12" firstref="ref51"></nolink>
Header DbId: eric
DbLabel: ERIC
An: EJ1461456
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Preschoolers Use Probabilistic Evidence to Flexibly Change or Maintain Expectations on an Active Search Task
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Brooke+C%2E+Hilton%22">Brooke C. Hilton</searchLink> (ORCID <externalLink term="https://orcid.org/0009-0007-2059-751X">0009-0007-2059-751X</externalLink>)<br /><searchLink fieldCode="AR" term="%22Mark+A%2E+Sabbagh%22">Mark A. Sabbagh</searchLink> (ORCID <externalLink term="https://orcid.org/0000-0003-2938-6288">0000-0003-2938-6288</externalLink>)
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="SO" term="%22Child+Development%22"><i>Child Development</i></searchLink>. 2025 96(2):881-890.
– Name: Avail
  Label: Availability
  Group: Avail
  Data: Wiley. Available from: John Wiley & Sons, Inc. 111 River Street, Hoboken, NJ 07030. Tel: 800-835-6770; e-mail: cs-journals@wiley.com; Web site: https://www.wiley.com/en-us
– Name: PeerReviewed
  Label: Peer Reviewed
  Group: SrcInfo
  Data: Y
– Name: Pages
  Label: Page Count
  Group: Src
  Data: 10
– Name: DatePubCY
  Label: Publication Date
  Group: Date
  Data: 2025
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: Journal Articles<br />Reports - Research
– Name: Subject
  Label: Descriptors
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Preschool+Children%22">Preschool Children</searchLink><br /><searchLink fieldCode="DE" term="%22Probability%22">Probability</searchLink><br /><searchLink fieldCode="DE" term="%22Evidence%22">Evidence</searchLink><br /><searchLink fieldCode="DE" term="%22Educational+Games%22">Educational Games</searchLink><br /><searchLink fieldCode="DE" term="%22Search+Strategies%22">Search Strategies</searchLink><br /><searchLink fieldCode="DE" term="%22Expectation%22">Expectation</searchLink><br /><searchLink fieldCode="DE" term="%22Change%22">Change</searchLink><br /><searchLink fieldCode="DE" term="%22Student+Behavior%22">Student Behavior</searchLink>
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1111/cdev.14190
– Name: ISSN
  Label: ISSN
  Group: ISSN
  Data: 0009-3920<br />1467-8624
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: This study investigated 3- to 5-year-olds' (N = 64, 37 girls, 62.5% White, data collected between 2021-2022) ability to use probabilistic information gleaned through active search to appropriately change or maintain expectations. In an online fishing game, children first learned that one of two ponds was good for catching fish. During a subsequent testing phase, children searched the ponds for fish. Half saw outcomes that were probabilistically consistent with training, and the other half saw outcomes that were probabilistically inconsistent. Children in the Inconsistent condition adapted their search strategies, showing evidence of changing their expectations. Those in the Consistent condition maintained their initial search strategy. Trial-by-trial analyses suggested that children used a combination of heuristic and information integration strategies to guide their search behavior.
– Name: AbstractInfo
  Label: Abstractor
  Group: Ab
  Data: As Provided
– Name: Note
  Label: Notes
  Group: Note
  Data: https://osf.io/92meh/?view_only=86830b0d6e7143b0925a6c86578a3b8f
– Name: DateEntry
  Label: Entry Date
  Group: Date
  Data: 2025
– Name: AN
  Label: Accession Number
  Group: ID
  Data: EJ1461456
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ1461456
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1111/cdev.14190
    Languages:
      – Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 10
        StartPage: 881
    Subjects:
      – SubjectFull: Preschool Children
        Type: general
      – SubjectFull: Probability
        Type: general
      – SubjectFull: Evidence
        Type: general
      – SubjectFull: Educational Games
        Type: general
      – SubjectFull: Search Strategies
        Type: general
      – SubjectFull: Expectation
        Type: general
      – SubjectFull: Change
        Type: general
      – SubjectFull: Student Behavior
        Type: general
    Titles:
      – TitleFull: Preschoolers Use Probabilistic Evidence to Flexibly Change or Maintain Expectations on an Active Search Task
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Brooke C. Hilton
      – PersonEntity:
          Name:
            NameFull: Mark A. Sabbagh
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 03
              Type: published
              Y: 2025
          Identifiers:
            – Type: issn-print
              Value: 0009-3920
            – Type: issn-electronic
              Value: 1467-8624
          Numbering:
            – Type: volume
              Value: 96
            – Type: issue
              Value: 2
          Titles:
            – TitleFull: Child Development
              Type: main
ResultId 1