The Cognitive Development Society conference is comprised of pre-conference workshops on Thursday October 17th followed by two days of conference proceedings. The conference will include two plenary speakers, invited symposia, contributed symposia and oral papers as well as poster sessions.

08:30 – 17:00: Pre-conference workshops

17:30 – 19:00: Welcome reception
Grand Foyer of the Galt House Hotel

Join us for appetizers and a cash bar to catch up with old friends and make new acquaintances! 

19:00 – 22:00: Student Pub Night
Patrick O’Sheas – Loft Space
123 W. Main Street

Open to all students and students at heart!  Make sure you have your name badge in order to receive a discount on food and drinks.


08:00 – 08:30: Coffee and registration

08:30 – 09:00: Opening remarks

09:00 – 10:00: Plenary Address 1: Michael Tomasello, Duke University

Becoming Human: A theory of Ontogeny

Over the past two decades, my colleagues and I have documented many differences in the cognitive and social abilities of human children and their nearest great ape relatives. In this talk, I attempt to bring these studies together into a coherent theory of the ontogeny of uniquely human psychology. For each of eight uniquely human developmental pathways – four cognitive and four sociomoral – there are two key transitions: one at 9 months (joint intentionality) and one at 3 years (collective intentionality). The hypothesis is that these transitions result from the new kinds of social coordination, sociocultural experience, and social self-regulation that the maturation of the capacities for joint and collective intentionality make possible.


10:00 – 10:30: Refreshment break

10:30 – 12:00: Plenary Symposium 1: Religious Cognition

Kathleen Corriveau, Boston University - The role of religious exposure in children's conceptualization of the invisble and the impossible

Children acquire knowledge through direct personal experience and exploration in both the scientific and religious domains; however, they also acquire knowledge indirectly through the testimony of others. Indeed, testimony is especially important when acquiring knowledge in domains where first-hand experience is limited or impossible, such as ordinarily unobservable scientific (e.g., bacteria) and religious phenomena (e.g., God). In this talk, I focus on individual and developmental differences in children’s ability to make ontological judgments based on environmental and cultural influences such as religious and scientific exposure. I present findings from children growing up in the United States, China, and Iran, and highlight the role of community consensus in children’s use of parental testimony when determining the existence of things they cannot readily see for themselves.

Deb Kelemen, Boston University - Implications of early intuitions about nature for religion and science

Spirits are everywhere. In nearly every culture, mythologies and religious beliefs abound of disembodied souls that populate or design nature, persist after death, and influence our lives. But where do these beliefs come from?One classic answer is that they are solely the products of cultural learning; they derive from testimony and narratives that get told and re-told to each societal generation. However, in this talk, I will suggest that this is not the whole story. Perhaps such beliefs also recur and persist because of a suite of cognitive tendencies that human minds are predictably inclined to develop. If this is so, people across cultures should display early intuitions or later spontaneous gut assumptions about the existence of designs, purposes, and eternal souls in nature. I will present evidence that this is the case and consider the implications of these tendencies for not only religious but scientific knowledge acquisition.

Larisa Heiphetz, Columbia University - Connections Between Religious and Moral Cognition

Many American adults report that it is necessary to believe in God in order to be moral. How does this idea emerge? If adults learn to link religion and morality based on decades of social learning (e.g., teachings they might hear in places of worship), then children – who have far less social experience than adults – may not show a strong association between these concepts. However, several lines of work converge to show that the mental association between religion and morality is already present during the early elementary school years. Children view God as a moral agent and perceive those who share their own religious beliefs in more positive ways (e.g., as more pro-social) than religious out-group members. These findings suggest that children readily form judgments of others based on invisible mental states and that religion serves an important role in the development of social inferences.

Jon Lane, Vanderbilt University - Constructing ideas of the supernatural

The proliferation of research on religious concepts and the development of religious concepts is wonderful, and necessary for a fuller account of human psychology. But there seems to be a tendency to exotify religious and supernatural concepts; to treat them as distinct or special. Thus, some scholars have argued that religious concepts are especially “natural”, others have argued that they are largely social artifacts, and others simply dismiss these ideas as “immature” or “childish.” I argue that religious or ‘supernatural’ concepts are no more strange or special than ‘natural’ concepts. Just like natural concepts, supernatural concepts are constructed upon and constrained by other robust naïve theories—e.g., of mind, biology, and physics. In my own work, my colleagues and I have examined how children come to understand god-like minds, particularly all-knowing minds. We have found that young children have great difficulty appreciating this sort of mind, but through a protracted developmental process older children and adults eventually approach an understanding of complete omniscience. Based on these and other data, I argue that constructivism makes possible some supernatural ideas that are typically not realized until late childhood or adulthood, thus arguing against views that religious concepts are simply ‘natural’ or ‘childish’. As well, I highlight that, to the extent that belief hinges upon mental representation of the relevant ideas, a constructivist framework is necessary to fully account for children’s (and adults’) belief in supernatural notions.


12:00 – 13:15: Lunch on own or Lunch workshops

13:15 – 14:30: Poster session 1 & Exhibitors

14:30 – 16:00: Parallel Session 1 – 5

S1           Stages of predictive processing in infants and toddlers: Forming expectations, experiencing prediction error and what it means for learning and memory
Chair: Felicia Zhang, Princeton University
Discussant: Lauren L Emberson, Princeton University
Prediction, or the ability to use past experiences to generate expectations about future sensory input, is believed to be crucial to adult cognition. However, the role of prediction in cognitive development has only recently become an important topic of research. This symposium presents cutting-edge findings on prediction in infants, toddlers, and preschoolers. Recent work in adults has parsed prediction into three phases: 1) Form an expectation that is compared to sensory input. If the expectation matches sensory input, there is facilitation of processing. 2) Any mismatch between the expectation and sensory information results in a prediction error. 3) Prediction errors spark attention and learning that change expectations to be more accurate in the future. Each of the three talks in this symposium present research investigating one of these three phases of prediction from a developmental perspective. The first presentation examines whether the process of forming expectations is beneficial for causal learning in toddlers. The second presentation focuses on prediction error in 14-months-olds by measuring anticipatory eye movements and pupil size. The last presentation directly tackles whether predictability influences learning and memory in preschoolers. Taken together, these studies provide important fuel to address the fundamental question of whether or not predictability, as it has been found in adults, supports cognitive development or whether prediction develops along with cognitive development. Moreover, breaking prediction down into these three phases provides a model in which to integrate findings from different tasks and populations.

Emily Daubert, Rutgers University - Why does puppy have a tummy ache?: Facilitating expectation using pedagogical questions to promote learning in preschoolers

Emily Daubert¹, Yue Yu², Milagros Grados³, Patrick Shafto³, Elizabeth Bonawitz³

¹University of Hawaii at Manoa, ²National Institute of Education Singapore, ³Rutgers University – Newark

Prompting children to make predictions about causal relations can serve as an important tool for early learning (Legare, 2012). Specifically, pedagogical questions, questions asked about relevant information with the intent to teach, elicit explanations (Walker & Nyhout, in press), which in turn encourage children to generalize their understanding using both observational evidence and their prior beliefs to novel situations (Walker et al., 2016). Relatedly, prompting for explanations may also reduce prediction bias (Hirt et al., 2004), especially in domains in which children have relatively entrenched beliefs. Indeed, children who are prompted to make predictions about causal phenomena generate simpler hypotheses (Walker et al., 2017) with greater explanatory virtue (Lombrozo, 2016). We tested the effect of pedagogical questioning, compared to direct instruction (providing relevant information in the form of statements with the intent to teach), which, does not prompt prediction, but crucially, explicitly teaches children about relevant concepts. Specifically, 72 preschoolers (Mage=49 mo; 50% female), from a variety of socioeconomic backgrounds, participated in a two-week storybook intervention aimed at improving their understanding of the relations between psychological causes and bodily effects, or psychosomatic understanding. This domain was selected because it is an area in which children’s beliefs are relatively entrenched, and has broader implications for health and well-being. Children were included in the study if they initially endorsed a plausible within-domain (physical) cause rather than the statistically likely, but cross domain (psychological) cause in a pretest book. Children were then randomly assigned to one of three conditions: Pedagogical Questions (PQ), Direct Instruction (DI), and Control (CT). During storybook reading, children in the PQ condition heard questions about the underlying mechanisms of psychosomatic phenomena, children in the DI condition were taught about the mechanisms, and the children in the CT condition read similar books for the same amount of time but received no information or questions about psychosomatic events. At posttest, on a measure of near transfer (a modified storybook similar to the pretest screening book), children’s responses significantly differed across conditions, F(2,69)=4.09, p=.021, such that children in the PQ condition were more likely to identify the psychological cause (38%) than children in the DI (13%), who in turn, were more likely to recognize a psychological cause than children in the CT (8%) condition, p=.017. On a measure of far transfer (an open-ended storybook that prompted an explanation about a novel situation), children’s significantly differed across conditions, F(2,69)=5.07, p=.009, such that children in the PQ condition were more likely to identify the psychological cause (33%) than children in the DI (25%), who in turn, were more likely to recognize a psychological cause than children in the CT (0%) condition, p=.003. Lastly, children’s memory for storybook details significantly differed across conditions, F(2,69)=3.21, p=.047, such that children in the PQ condition remembered significantly more details (M=7.33) about the storybooks than children in the DI (M=5.67) and CT (M=6.17) conditions, p=.018. Results will be discussed in terms of the role of pedagogical questions in facilitating prediction, and the implications for learning, transfer, and memory.

Felicia Zhang, Princeton University - Prediction and prediction error in 14-month-old infants

Felicia Zhang¹, Lauren Emberson¹

¹Princeton University

Research with adults has shown that predictions can be correct or incorrect, and each has the potential to support cognitive changes. When your prediction is correct, it allows you to use what you have learned to support more efficient processing of your sensory input (Kersten & Yuille, 2003; Kutas, DeLong & Smith, 2011; Wei & Stocker, 2015). When your prediction is incorrect, this can generate prediction error signals in the brain (Hollerman & Schultz, 1998; Rao & Ballard, 1999; Friston, 2005), which helps you update your predictions and promotes learning. Recent work in developmental psychology, using different methods such as EEG (Kouider et al., 2015) and fNIRS (Emberson et al., 2015), have shown that infants also experience top-down sensory prediction in the brain suggesting markers of prediction are also present in infants. But whether or not infants use prediction and prediction error in a more sophisticated, adult-like way has yet to be answered. To study the relationship between prediction and prediction error early in development, we used anticipatory eye movements (AEM) to index prediction and pupil change, to index prediction error. Infant learning has never been addressed using this unique combination of behavioral measures. In our learning task, 14-months-old infants are presented with a target on either the left or right side of the screen for 8 trials (pre-switch). Then, the same target appeared in the other position for 8 (post-switch) trials. This pattern was repeated with new targets and locations for 3 blocks. In pre-switch trials, we hypothesized that infants would predict the target location and generate AEMs to that location. Post-switch trials will allow us to examine how predictions (AEMs) relate to prediction error (pupil) and then subsequent predictions. Our results show that infants who learn better (i.e. have higher percentage of correct AEMs in pre-switch) make more mistakes in the form of incorrect AEMs when the target switches (r = 0.35, p = 0.06). Infants who learn better also have larger pupil change during the first trial of post-switch, which was designed to elicit the strongest prediction error, compared to infants who did not perform as well in the task (r = 0.37, p = 0.05). We will also model the trial-by-trial dynamics of prediction and prediction error in relation to canonical computational learning models that include prediction and attention (e.g., Rescorla-Wagner and Pearce-Hall models). Taken together, these results suggest that more prediction is correlated with larger prediction error, as evidenced through infants’ eye movements and pupil size and, like adults, prediction and prediction error are intertwined.

Viridiana Benitez, Arizona State University - Preschoolers remembering and learning from predictable and unpredictable events

Viridiana Benitez¹, Martin Zettersten², Jenny Saffran²

¹Arizona State University, ²University of Wisconsin-Madison

Young children are adept at tracking the predictable regularities in their environment (Aslin, 2014; Gomez & Gerken, 2000; Saffran, Aslin, & Newport, 1996). Further, recent research suggests that the predictability of events may influence how young children learn from those events (Benitez & Smith, 2012; Benitez & Saffran, 2018; Stahl & Feigenson, 2015; 2017). In particular, events that are predictable facilitate toddlers’ word learning in comparison to events that violate predictions (Benitez & Saffran, 2018). In three experiments (N=179), we examine if predictability also affects memory and learning in preschool-aged children. We presented 3- to 5-year-old children with a tablet task that displayed 3 locations on the screen. When 1 of the 3 locations turned green, children had to touch that location to reveal a hidden animal. The order of locations to touch followed a predictable spatiotemporal pattern (1, 2, 3), such that, over trials, children could learn to predict which location would light up next to reveal an item. Critically, we presented novel items (Exp. 1), novel sound-item associations (Exp. 2), or novel word-item associations (Exp. 3) during the sequence, half from events consistent with the spatiotemporal pattern (Predictable: 1, 2, 3), and the other half from events that violated the spatiotemporal pattern (Unpredictable: 1, 2, 1). In all experiments, children showed slower reaction times to Unpredictable events than to Predictable events [F(1,13.76)=247.26, p<0.001], demonstrating learning of the predictable sequence. The key question was whether children remembered or learned differently from Predictable or Unpredictable events. In Experiment 1 (N=28, Mean Age= 4.6, SD=0.89), we examined if predictability affected children’s memory for the novel items. Results showed that children remembered items accurately when presented during both Predictable [M=0.71, SD=0.21; b=0.97, STE=0.22, z=4.39, p<0.001] and Unpredictable events [M=0.71, SD=0.25; b=1.06, STE=0.27, z=3.92, p<0.001]. There was no effect of predictability on accuracy scores [b=0.09, STE=0.28, z=0.33, p=0.74]. In Experiment 2 (N=31, Mean Age= 4.6, SD=0.83), we examined children’s learning for sound-item associations presented during Predictable or Unpredictable events. Results again showed learning and equal performance [b=0.19, STE=0.28, z=0.68, p=0.50] for sound-item associations presented during Predictable [M=0.62, SD=0.33; b=0.57, STE=0.28, z=2.05, p=0.04] and Unpredictable [M=0.63, SD=0.34; b=0.76, STE=0.30, z=2.51, p=0.012] events. Similarly, Experiment 3 (N=120, Mean Age= 4.47, SD=0.86) again showed that children learned novel word-item associations well regardless of predictability [Predictable: M=0.59, SD=0.33, b=0.49, STE=0.15, z=3.21, p=0.001; Unpredictable events: M=0.61, SD=0.33, b=0.52, STE=0.14, z=3.65, p<0.001; Predictable vs. Unpredictable: b=0.03, STE=0.18, z=0.21, p=0.84]. Across the three experiments, preschool-aged children remembered and learned equally well from predictable and unpredictable events. Unlike toddlers, children in our task were able to robustly encode novel associations even when they were presented during prediction-violating events. Given our large dataset, we are currently conducting exploratory analyses to examine individual differences in whether children benefit from predictability. The results together should shed light on the role of predictability for young children’s remembering and learning.

S2           Naive epistemology: Children’s intuitive theories of knowledge and informativeness
Chair: Rosie Aboody, Yale University
Discussant: Melissa Koenig, University of Minnesota
Much of what we learn comes from others, making it critical to distinguish those who are knowledgeable and informative from those who are not. Prior work shows that even young children select informants by tracking cues that reveal knowledge, such as accuracy and confidence. Are such decisions based on an intuitive theory of knowledge? Or do these judgments reflect isolated social learning strategies, learned independently? This symposium brings together work that characterizes different dimensions of children’s naive epistemology. Our goal is to combine different perspectives to advance a unified account of children’s epistemic reasoning. Paper 1 focuses on the experiences that license knowledge, exploring whether preschoolers have normative expectations about when agents are justified to share what they know, or what they think they know. Paper 2 focuses on the actions that reveal knowledge, showing that children’s epistemic inferences are best explained by mental-state reasoning rather than by special-case rules and heuristics. Paper 3 extends the scope of these inferences beyond others’ knowledge about the world, showing that children also reason about the informativeness of others’ praise based on their past actions, and use it to learn about their own performance. Across all experiments, children evaluated agents’ potential to inform by considering why agents selected an action or made a claim. Collectively, these studies suggest that even preschoolers reason about others’ informativeness based on a naive epistemology. Our discussant, an expert on the development of epistemic trust and social learning, will synthesize these findings to provide further insights into children’s developing ability to navigate the social world. Altogether, these results suggest that children’s ability to reason about others’ knowledge may serve as an “epistemic compass”, guiding their learning about the world, about others, and even about themselves.

Lucas Butler, University of Maryland - Investigating children's developing understanding of integrity in others' epistemic practices

Lucas Butler¹, Hailey Gibbs¹

¹University of Maryland

We live in a society increasingly inundated with information, with eroding and often insufficient checks on what is authoritatively true, facing a crisis of honesty and integrity both in science and in the public sphere. Thus a pressing issue facing society is ensuring that as children develop the mental habits that will guide how they engage with the world, they learn to approach information with an evaluative lens. Fundamentally, this means understanding and fostering young children’s ability to assess whether others’ claims about the world are made in an intellectually honest and valid manner. This is especially true against a backdrop of a crisis of honesty and integrity in science and beyond. This talk presents a framework for investigating children’s developing ability to reason about whether others are engaging in epistemic practices with integrity. Specifically, we propose three broad principles underlying integrity in epistemic practice: (1) Don’t See Something? Don’t Say Something. This is the principle that we ought to only make claims to knowledge that we actually have, and the corollary that if do not have the relevant knowledge to make a claim, we ought not to make it. (2) Be Honest About What You Know. This is the principle that, we ought to be judicious in tailoring our claims to the evidence we have, and not claim to know something we are not sure of. (3) Use Evidence to Help, not to Deceive. This is the principle that when communicating evidence, we ought to do it in a way that conveys what we believe to be true, and to do so for prosocial reasons, rather than to manipulate another’s belief for our own gain. We will then present several studies investigating these principles. In one series, we find that 3- to 7-year-olds (N=120) evaluate unverified claims as less acceptable than verified ones based on sufficient knowledge. In second series of nearly-completed studies, we are finding that children understand that one’s sampling behavior ought to be tailored to ones empirical question. We will present a third series of studies, currently ongoing, exploring whether children can recognize when individuals are using evidence for prosocial or deceptive purposes. Finally, we will present a roadmap and some preliminary data supporting further investigations of children’s understanding of each of these principles underlying honest and transparent epistemic practice.

Rosie Aboody, Yale University - Ignorance = doing what is reasonable: Children expect ignorant agents to act based on prior knowledge

Rosie Aboody¹, Julian Jara-Ettinger¹

¹Yale University

We frequently encounter novel situations, but even when we lack knowledge, we’re rarely at a loss. We can often draw on our prior experiences to guide our actions, and we expect others to do the same, assuming that those who lack knowledge will apply relevant prior experiences to novel situations. It is unclear, however, whether children share these expectations. In some contexts, children judge that ignorance will lead to failure (Ruffman, 1996; Saxe, 2005); in others, success (Friedman & Petrashek, 2009; German & Leslie, 2001). In the current work, we test whether four- to six-year-olds expect ignorant agents to act according to their prior experience. In Experiment 1, children (Age: 4.05-6.99 N = 72) and two puppets were introduced to a novel toy, and learned which of three buttons activated it (see Figure 1a). Next, each agent was asked to activate another (outwardly identical) toy. One agent acted consistently with his prior knowledge, pressing the same button that activated the original toy. The other acted inconsistently, pressing a different button. Both agents succeeded. Last, children were asked to judge which agent had already known how to make all the toys go. Experience with the initial toy perfectly explains the consistent agent’s choice: he simply transferred his experience with the initial toy to the novel one. But this experience does not explain the inconsistent agent’s choice, and thus children may assume this agent acted based on some additional knowledge. In fact, five- and six-year-olds (but not four-year-olds) judged that the inconsistent agent was the one who already knew how to make all the toys go (see Figure 1b). This suggests that by five years, children expect both knowledgeable and ignorant agents to act according to their prior knowledge. In Experiment 2, children (Age: 3.99-6.92 N = 72) and puppets learned how the same initial toy worked. Next, each agent was asked to activate another (outwardly identical) toy. Again, one pressed the same button he had seen make the original toy go, and one pressed a different button. But this time, both failed to activate the toy. Children were asked to judge which agent knew more (but crucially, not all) about the novel toy. As before, experience with the initial toy does not explain the inconsistent agent’s choice. So children may again judge that this agent acted according to some (incomplete) additional knowledge. In fact, six-year-olds (but not younger children) judged that the inconsistent agent knew more about the toy (see Figure 1b). This suggests that by age six, children not only expect ignorant agents to act according to their prior knowledge, but also understand that knowledge is not binary: that agents can know more or less about any given topic. In two experiments, we demonstrate that by age five, children expect ignorant agents to act based on prior knowledge, attributing additional knowledge to agents whose actions are inconsistent with their prior experience. And by age six, children understand that knowledge is continuous, attributing greater knowledge even to agents whose failed actions are inconsistent with their observable prior experience. These results suggest that young children expect ignorant agents to act based on relevant prior experience, and that children use this expectation to infer what others know from what they do.

Mika Asaba, Stanford University - Children can use statistical information to infer the informativeness of others' praise

Mika Asaba¹, Emily Hembacher¹, Michael Frank¹, Hyowon Gweon¹

¹Stanford University

Learning about the self means more than observing our own actions and outcomes; we also seek others’ evaluative feedback about the self (e.g., praise, criticism) and learn from what others think about our actions and outcomes. Importantly, even the same feedback (“That’s a great drawing!”) can be perceived differently depending on the evaluator; praise from someone who usually praises indiscriminately is less informative than praise from someone who selectively praises high-quality work. The ability to understand who provides more informative feedback is a critical skill for young children as they begin to construct early self-concepts. Prior work suggests that young children are sensitive to the informativeness of others’ testimony about object labels (e.g., Koenig & Harris, 2005; Sobel & Kushnir, 2013) or demonstrations of causal functions (e.g., Gweon & Asaba, 2017; Gweon, Pelton, Konopka, & Schulz, 2014) based on agents’ history of accuracy or informativeness. Furthermore, children can use minimal covariation information to draw rich inferences about people (e.g., Gweon & Schulz, 2011; Seiver, Gopnik, & Goodman, 2013). Building on previous work, here we ask whether young children can infer the informativeness of others’ praise from minimal covariation information about the correspondence between an agent’s praise and the quality of the product being praised. We hypothesized that children can detect the statistical pattern in others’ praise and use it to inform their own decisions about whose praise to endorse. In Study 1, children (preregistered; Age: 4.0-5.9, N=80) made two tracings and placed them into separate envelopes. Children then watched videos of two teachers providing feedback on another student’s tracings (6 tracings total: 3 good, 3 bad). The Selective Teacher praised just the 3 good tracings whereas the Overpraise Teacher indiscriminately praised all 6. Then, participants received feedback on their own tracings: The Selective Teacher praised the drawing in one envelope, while the Overpraise Teacher praised the drawing in the other envelope. When children were asked to choose the best tracing, children chose the envelope with the tracing that was praised by the Selective Teacher (72.5%, p = 0.006, Binomial), but when asked to choose which teacher was trying to be nice, they chose the Overpraise Teacher (82.5%, p < .001). In Study 2 (preregistered; Age: 4.0-4.9, N=24), we asked whether children valued the Selective Teacher’s praise because she praised more rarely; children may have simply used rarity as a cue for informativeness. The procedure was identical to Study 1, except that children were presented with the Selective Teacher (who praised only the 3 good tracings) and the Selective-Incongruent Teacher (who praised only the 3 bad tracings). When asked to choose the best tracing, again children selected the envelope that was praised by the Selective Teacher (75%, p = 0.02, Binomial), suggesting that they specifically trust a teacher who selectively praises higher quality work. Ongoing work is asking who young children approach for feedback and further, whether they differentially expend effort depending on who might later evaluate their work. These results suggest that praise is more than positive reinforcement for preschool-aged children: They are attuned to the underlying meaning of the praise and understand whose praise is more informative, and use it to make sense of the relative quality of their own work.

S3           Applying cognitive principles to children’s learning in educational contexts
Chair: Caroline Hornburg, Virginia Polytechnic Institute and State University
Discussant:  Lisa Fazio, Vanderbilt University
Developmental scientists have long been interested in cognitive mechanisms of children’s learning of new information. Although children’s cognitive processes have been examined extensively in highly-controlled laboratory contexts, these experiments and materials are often simple, contrived, and not naturalistic. Thus, results of these investigations may not always apply to children’s learning from real-world contexts, which is of urgent interest to educators, caregivers, and policymakers. This symposium presents new research on how children learn information in educational contexts. In particular, it presents examinations of how two learning mechanisms, retrieval practice and comparison, impact children’s memory for and generalization of new information learned in applied settings using educational materials. Three papers will be presented which identify cognitive mechanisms underlying children’s learning – or failure to learn – from storybooks, educational textbooks, and a supplemental curriculum. Paper 1 will describe how implementing retrieval practice into children’s storybooks impacts preschoolers’ recognition and retrieval memory for novel words. Paper 2 will demonstrate how elementary-school children’s learning from science textbooks depends on both the nature and frequency of retrieval practice. Paper 3 will show how high-school students’ learning of algebra is impacted by comparison of problem-solving strategies presented with or without students’ names to label strategies. The goal of this symposium is to examine how well-understood learning mechanisms, such as retrieval practice and comparison, translate from laboratory experiments to naturalistic and educational settings. The topics presented will spark discussions of the successes and failures of applying these cognitive principles to improve children’s learning in educational contexts.

Haley Vlach, University of Wisconsin - Children's recognition and retrieval memory for words learned via storybook reading

Catherine Bredemann¹, Haley Vlach¹

¹University of Wisconsin-Madison

Research demonstrates that children’s storybooks can be facilitative for word learning (e.g., Ninio, 1983); what remains unclear is how the structures of storybooks contribute to children’s success in language comprehension and production. It could be that incorporating elements that engage children in retrieval practice benefits children’s word learning moreso than passive listening. Retrieval practice has been demonstrated to improve adults’ learning over simple restudying (e.g., Roediger & Karpicke, 2006); however, it is unclear whether children benefit from the same type of retrieval practice. This study investigated the effect of retrieval practice, via recognition testing or cued recall, on children’s language comprehension and production during and after storybook reading. Across three experiments, preschool children (N=167; Age Range=26-68 months) were repeatedly read a storybook introducing them to ten novel words and objects. Experiment 1 examined children’s language comprehension. Across two conditions, children engaged in passive word mapping or retrieval practice in the form of a recognition memory task. After a delay, children’s comprehension was assessed using a recognition memory task, in which they were asked to select and point to the label’s referent from among an array of objects. Results revealed high performance at the post-test for recognition memory in both conditions. Experiment 2 examined children’s language production. Across two conditions, children engaged in passive word mapping or recall practice in the form of a production task. After a delay, children’s production was assessed using a recall memory task, in which they were asked to produce the words for target novel objects. Results revealed floor performance for recall of novel words in both conditions. Experiment 3 provided additional mapping and retrieval opportunities during storybook reading. Half of the children received mapping practice followed by recall, while half received recall practice followed by mapping. At a delayed production test, all children again displayed poor memory for novel words. Thus, retrieval practice in the form of recognition testing during storybook reading may be beneficial for children’s comprehension of language, but recall practice may not benefit children’s production of language.

Caroline Hornburg, Virginia Polytechnic Institute and State University - Harnessing the benefits of retrieval practice for children's learning through implementation of open-book and closed-book activities

Caroline Hornburg¹, William Aue³, Stephanie Karpicke², Jeffrey Karpicke²

¹Virginia Polytechnic Institute and State University, ²Purdue University, ³Wright State University

Retrieval practice is a powerful learning technique that can boost long-term retention of educational material (Karpicke, 2017). For optimal learning and retention, there needs to be a balance between initial retrieval success and retrieval practice difficulty (Bjork, 1994; Karpicke, Lehman, & Aue, 2014). Less research has been conducted with elementary students in this domain. In a series of experiments covering science material with fourth and fifth grade children, we manipulated the nature of the practice activity (open book or closed book; fill-in-the-blank, short answer, or cued recall) and the number of practice activities (one, two, three, or four practice phases). We assessed differences in initial performance and delayed performance on the same content after one week. Practice activity types were compared to each other and to control conditions of either simply reading the text passage or studying a series of questions with answers (akin to a completed study guide). We also measured children’s judgments of learning of the materials at the end of the first session. Based on previous work with college students (Agarwal et al., 2008), we predicted that open-book practice would yield better initial performance than closed-book practice, but that this benefit would not persist after a delay. No prior study had implemented both open-book and closed-book practice; however, we expected higher performance after a delay when mixing text support compared to only practicing with open-book activities. As hypothesized, results indicated a benefit of the presence of text support (i.e., open-book practice) on initial performance, but generally lower performance in open-book conditions (i.e., more forgetting) after a delay. Regardless of the question format or order of practice, mixing of practice with both open- and closed-book activities tended to result in higher retention than practice with one or two phases of solely open-book or closed-book activities. Finally, children’s judgments of learning tended to be higher whenever learning activities included open-book practice, as well as when children practiced three or four phases. Based on these results, we suggest that one way for teachers and parents to aid in children’s learning may be to provide opportunities to answer questions both with the text present and without the text present to boost both initial success and harness the benefits of retrieval practice. If time is limited, it may be best spent allowing children to answer questions with the text present rather than studying questions and answers or simply rereading a text passage. Overall, results contribute to the larger body of research aimed at integrating cognitive science principles within educational practice.

Abbey Loehr, Washington University in St. Louis - Does calling it 'Morgan's way' reduce adoption and generalization of the strategy?

Abbey Loehr¹, Bethany Rittle-Johnson², Kelley Durkin², Jon Star³

¹Washington University in St. Louis, ²Vanderbilt University, ³Harvard University

When describing how to solve a problem, it is common to give an example of how someone else solved it. For example, middle-school mathematics textbooks often use hypothetical students to present problem-solving strategies (e.g. Morgan solved the problem using distribution; Riggs, Alibali, & Kalish, 2015). Presenting strategies as being implemented by a specific individual (i.e., person-presentation) is meant to make educational materials more interesting and engaging. However, person-presentation reduced strategy generalization in a controlled, short-term setting (Riggs et al., 2015, 2017). In the current study, we tested whether person-presentation harms strategy generalization when strategies were compared and integrated into classroom instruction. Specifically, ninth graders learned through comparison and explanation of multiple strategies. Comparison and explanation both support learning by focusing attention to the underlying structure of problems and away from less relevant surface features (Gentner & Medina, 1998; McEldoon, Durkin, & Rittle-Johnson, 2011; Siegler & Chen, 2008). Therefore, comparison and explanation of multiple strategies could protect against the negative effects of person-presentation on strategy generalization. During a unit on solving linear equations, five Algebra I teachers and their 174 students used 9 worked example pairs (WEPs) in which strategies were either presented with pictures and names of hypothetical students (e.g., “Morgan’s way”; person-presentation condition) or with a strategy label (e.g., “add up way”; strategy-label condition), with teachers randomly assigned to condition. After comparing and explaining the two strategies in each WEP, students answered generalization questions by rating how likely another student, teacher, or themselves would be to use each strategy in the future on a 5-point scale. Students completed a pretest and posttest at the beginning and end of the unit assessing their conceptual and procedural knowledge of the strategies as well as their procedural flexibility. Students’ scores on the assessment improved from pretest to posttest (M’s = .34 to .53). Students in the person-presentation condition performed similarly on the posttest compared to students in the strategy-label condition, controlling for pretest scores (M = .53 vs .52). Their generalization ratings were also similar (M = 3.73 vs 3.70). Thus, in a long-term classroom context we found no negative effects of comparing and explaining strategies presented with pictures and names of hypothetical students. Learning different ways to solve problems and comparing strategies helps children become flexible problem-solvers (Rittle-Johnson & Star, 2007). Our findings suggest no negative (or positive) effects of teachers using students’ names to label strategies, a method recognized by The National Council of Teacher of Mathematics (NCTM, 2000). Why previously found negative effects of person-presentation may not have extended to this context are discussed, including (a) comparison and explanation directing focus away from surface-features of the examples such as character names, and (b) affordances of person-presentation aiding implementation of the materials.

S4           How do young children infer what the norms are?
Chair: Sydney Levine, MIT
Discussant: Marjorie Rhodes, New York University
There are some norms (such as harm-based norms) that even pre-verbal children seem to understand at a very young age. It is possible that these sorts of norms are universal. However, some norms are unique to a certain context or social group. How do young children infer what the norms are? What kind of information is relevant to social-norm inference? What mechanisms are available to children to make norm inferences in novel contexts? This symposium presents three answers to this set of questions. The first paper looks at the role of punishment in norm inference. The authors show that when a novel non-harmful action is consistently punished, that action is considered to be just as wrong as a novel action that involved harm. The second paper looks at the role of statistical regularity in norm inference. The authors show that both an internalist and a structural construal of statistical regularities lead to norm endorsement. The third paper looks at the role of logical consistency in norm inference. The authors show that even an action with a positive outcome can be inferred to be wrong if the action can’t be “universalized” (done by everyone). A discussant will consider how this body of work relates to research on generics, stereotypes and moral development. Together, this body of work describes the diverse mechanisms behind norm inference in early childhood.

Sophie Arnold, Yale University - Punishment as a signal of wrong: How children's judgements of novel actions are swayed by the presence, or absence, of punishment

Sophie Arnold¹, Yarrow Dunham¹

¹Yale University

In a world filled with ambiguity, how do we learn right from wrong? While some behaviors may seem clearly wrong, many others are ambiguous – only wrong in certain situations (e.g. talking without raising your hand first) or cultures (e.g. eating beef). Thus, the moral value of at least some actions must be culturally transmitted. How does this occur? Here we explore our hypothesis that the presence or absence of punishment is a powerful signal as to the rightness or wrongness of actions, with punished actions seen as more wrong and unpunished actions seen as less wrong. Punishment is a plausible moralizing force because children experience it from a young age (e.g. time outs) and even engage in it (e.g. tattling); they also expect familiar wrongs to be punished and express desire to live in a world in which regimes of punishment are present rather than absent (Bregant et al., 2016). These considerations motivate our hypothesis that children will also use the presence or absence of punishment to infer the moral valence of novel actions. We investigated children’s intuitions about the signaling capacity of punishment across two studies. In each, we showed participants six novel actions performed by cartoon children. Actions differed by type and punishment information. To understand children’s intuitions across a range of potential behavior, we included two types of actions, solo and dyadic. For solo actions, an actor preformed a novel action on their own while a bystander watched. For dyadic actions, the novel action negatively affected a patient. In terms of punishment information, children were told that the novel action was always punished, never punished, or given no extra information (baseline). Children were asked to judge whether each novel action was “good, bad or neither.” In order to explore the specific kind of information that punishment conveys, in Study 2 children were also asked to judge whether the novel action was harmful or gross. We found that both action type and punishment information affected children’s ratings of novel actions. Solo actions were neutral at baseline and unaffected by the lack of punishment, but became strikingly bad when always punished. In contrast, dyadic actions were rated negatively at baseline; but, with effects of approximately equal magnitude, dyadic actions became less bad in the absence of punishment and more bad in the presence of punishment. Study 2 replicated these effects and added additional ratings of harm and grossness. Beginning with harm, always punished actions were judged more harmful than baseline actions and never punished actions were judged as less harmful than baseline actions. However, perceived harmfulness did not entirely account for the relationship between punishment and judgements of wrongness, suggesting that the signaling value of punishment goes beyond a narrow signal of harm. In contrast, the presence or absence of punishment did not have a strong effect on children’s evaluations of how gross novel actions were. Taken together, these studies show that punishment predicts judgements of both harm and wrongness. Strikingly, punishment can turn neutral actions bad and reduce the badness of harmful actions. This suggests that punishment could be a central channel by which children learn culturally specific wrongs, thereby representing a powerful force for the moralization of previously unfamiliar behaviors.

Tania Lombrozo, Princeton University - Do structural (versus internalist) construals of social categories support normative judgments?

Nadya Vasilyeva¹, Alison Gopnik², Tania Lombrozo¹

¹Princeton University, ²University of California Berkeley

Statistical regularities between social categories and properties (e.g., between “girls” and “liking pink”) can be interpreted in many ways. Vasilyeva, Gopnik, and Lombrozo (2018) found that even preschool-aged children can construe such regularities in internalist terms (e.g., as a product of the inherent characteristics of category members, such as a preference for a particular color), or in structural terms (e.g., as a product of stable external constraints shaping the options available to category members, such as the availability of pink clothing for girls vs. boys). Here we explore whether these construals influence normative judgments, specifically about the wrongness of violating the category norm (e.g., giving something pink to a boy). We predicted that an internalist construal would license strong normative judgments: if the association between gender and color is seen as reflecting inherent properties of boys versus girls, then norm violation should generate strong disapproval. The predictions for a structural construal, however, are less clear. One possibility is that because a structural construal suggests that category-property associations would be different under different structural conditions (Vasilyeva, Gopnik, & Lombrozo, 2018), norm violations should be tolerated. Another possibility is that even a structural construal will have normative implications: Vasilyeva, Gopnik, and Lombrozo (2018) found that a structural construal licensed formal explanations (“X did Y because she is a girl”), which are closely tied to generic statements (“girls do Y”; Prasada & Dillingham, 2006; Vasilyeva, Gopnik & Lombrozo, forthcoming) that have been shown to communicate prescriptive norms (Roberts, Ho, & Gelman, 2017). In a novel study with 269 3- to 8-year old children, we induced either an internalist or structural construal of a category-property association, and we examined judgments of a norm violation across these two conditions, as well as a baseline condition. Participants learned about a school where girls picked yellow fruit, and boys picked red fruit. Under the internalist framing, schoolchildren were free to pick any fruit, inviting an inference about a preference for each color. Under the structural framing, the school assigned the girls and boys to two different fields, which later turned out to bear fruits of different colors. In the baseline condition, the outcome was attributed to accidental factors. A set of explanation-generation and evaluation measures confirmed that the target construals were successfully induced. We measured norm endorsement by asking participants to evaluate how “ok” or “not ok” (1-4 scale) it would be for someone to gift a red fruit to a girl (violating the gender norm). Both the internalist and structural conditions produced higher norm endorsement (i.e., more “not ok” judgments) than the baseline condition (Mint=2.82, Mstr=3.09, Mbase=2.37, ps≤.009); the difference between the internalist and structural conditions was not significant (p=.098). This confirms our prediction that an internalist construal supports normative judgments, and additionally suggests that a structural construal can do so as well. We consider mechanisms that could have produced this pattern, alongside their implications for norm construal, norm endorsement, generic language, and stereotyping.

Sydney Levine, MIT - What if everyone did that?: Young children universalize actions to make moral judgments

Sydney Levine¹, Max Kleiman-Weiner¹, Laura Schulz², Joshua Tenenbaum², Fiery Cushman³

¹MIT & Harvard, ²MIT, ³Harvard University

Some things are harmless when one person does them, but harmful in aggregate: Things like feeding a wild animal, leaving a faucet running, or failing to vote. Kant famously proposed these are wrong because the violate a categorical imperative; put simply, he asked, What if everybody did that? Asking this question — which we call “universalization” — is a way to determine what actions are morally acceptable by thinking about what everyone would agree to. Theories of moral cognition typically explain how we make judgments of right and wrong using either a rule-based or an outcome-based approach. We propose that there are some moral judgments that are better explained by universalizing an action. Our model of universalization is as follows: The moral acceptability of some actions is proportional to how bad the outcome would be in a hypothetical world in which all the people who were willing and able to do the action did the action, regardless of whether those people are actually going to do the action. We test this model with experiments in adults and young children. Adults (N=1000) read stories involving actions that are harmless if one person does them, but lead to a bad result if many people do. For example, one person can use a newly powerful fishing method in a small pond, but if everyone did that, the fish population would go extinct. It was emphasized that if one person did the action, no one else would know about it, so one person’s action wouldn’t lead to many other people doing it. This was important to ensure that subjects judgments of the action were based on the hypothetical world where everyone did the action, not an expected utility calculation of what will happen in the actual world. Half the subjects were asked the Collective Action question: the effect if N people took the action (where N was randomly sampled between 0 and 100, the number of people who could possibly take the action). Half the subjects were asked the Moral Judgment question: whether acting in that way was morally acceptable if N other people wanted to do the action. If our model is correct, then the Collective Action question would predict Moral Judgments with a high degree of accuracy and better than a model that uses only outcomes in the actual world or rules. We modeled moral judgments using a sigmoid transform of the Collective Action judgments and found a good fit (R2 = 0.90), significantly better than any lesioned version of the model or the outcome or rule-based model. Young children (n=40, ages 3-6 years old) were told stories with a similar structure to those told to adults. For instance, it’s fun if one person splashes in the pool, but if everyone does, then there would be no water left in the pool. Half of the subjects were asked the Moral Judgment question: whether the action was OK or not OK when many kids, 8 kids, 3 kids, and 1 kid want to do the action. The other half the subjects were asked the Collective Action question: what effect would be if many, 8, 3, or 1 kid did the action. Again, we found a good fit between the Collective Action and Moral Judgment data (R2 = 0.85), which out-performed an outcome-only or rule-only approach. In sum, these studies show that children and adults use universalization distinct from outcome-based or rule-based methods to make moral judgments. Our model explains the fine-grained structure of these judgments.


Oral Papers I

O1.1       How consequential and retributive motivations shape costly third-party punishment in young children

Presenter: Julia Marshall

Julia Marshall¹, Daniel Yudkin¹, Molly Crockett¹

¹Yale University

People punish moral transgressions to satisfy a variety of motivations. For example, people may be motivated by retributive concerns (Kant, 2011): individuals want antisocial others to receive their “just deserts.” On the other hand, people may punish because of consequentialist concerns (Bentham, 1789/2000): individuals wish to deter future harms by teaching the transgressor a lesson. Research with adults has examined which type of motivation underlies punitive behavior by manipulating whether punishment is “communicative”. Specifically, punishment is either known to the transgressor and can communicate a norm (i.e., communicative punishment) and thus satisfies a consequentialist motivation. Or, punishment is unknown to the transgressor (i.e., non-communicative punishment), thereby eliminating any potential prosocial benefit from punishment (and thus only satisfies a retributive motivation) (Crockett, Ozdemir, & Fehr, 2014; Nadelhoffer, Heshmati, Kaplan, & Nichols, 2013). These studies, among many others (Carlsmith, Darley, & Robinson, 2002), find that participants are motivated to punish in both communicative and non-communicative contexts. Here, we are concerned with whether young children, like adults, are also motivated by both consequential concerns and pure retribution. Much research has documented that children at least possess an appetite to punish others–preverbal infants prefer individuals who punish antisocial targets relative to those who are nice to antisocial targets (Hamlin, Wynn, Bloom, & Mahajan, 2011). And, children engage in third-party punishment as young as three (Yudkin, Van Bavel, & Rhodes, 2019). It is unclear though whether this punitive behavior is motivated by retributive or consequential concerns. We examined this question here with 4- to 7-year-old children (N = 113). Specifically, in our study, children could punish an antisocial child who ripped up another child’s drawing by preventing them from playing on an iPad, but doing so meant participants could not play themselves. Punishment either exclusively satisfied retributive motives by only inflicting harm upon the transgressor (“non-communicative punishment”) or additionally satisfied consequentialist motives by teaching the transgressor a lesson (“communicative punishment”). We found that children punished significantly more in the non-communicative condition than in a control condition where the target of punishment behaved neutrally (~40% versus ~13%), and they punished almost twice as much in the communicative condition (~80%). See Figure 1. In an ongoing study (projected N = 138; current n = 101), we are seeking to replicate this effect and also to examine what mechanisms underlie children’s desire to engage in retribution. Taken together, we find clear evidence that children’s punitive motivations are characterized as a mixture of retributive and consequentialist. Our work coheres with past research documenting that children are willing to engage in costly third-party punishment (McAuliffe et al., 2015), even children as young as four. These findings generate a clearer understanding of the early emergence of behaviors implicated in the evolution of cooperation in humans (e.g., Balliet et al., 2011), and show that both retributive and consequentialist motives drive punishment in young children and rule in the possibility that the taste for retribution emerges early in development.

O1.2       Being rich or poor: How inequality affects who children give to in experimental games

Presenter: Kelly Kirkland

Kelly Kirkland¹, Jolanda Jetten¹, Mark Nielsen¹

¹University of Queensland

Economic inequality is an uneven division of resources, where a majority of wealth is possessed by a minority of the population (Wilkinson & Pickett, 2009). Research has revealed a link between high inequality and lower prosocial behavior in adults (e.g., Côté, House & Willer, 2015). However, little research has uncovered how children experience and understand macro-inequality, and how this affects their behavior. One study by Kirkland, Jetten and Nielsen (2019) directly examined how four-year-old children experience inequality in a novel experimental paradigm. After playing a series of games with puppets, the players received tokens which could be exchanged for stickers. The distribution of tokens was either A) highly unequal or B) highly equal. Importantly, this was non-meritocratic and across both conditions children received tokens that placed them in the middle with regard to earnings. After experiencing high inequality, children donated fewer stickers to a child in need in compared to the children who experienced low inequality. However, when asked to give more tokens to the puppets, they did not give more to those who received less. Instead, they divided the points equally irrespective of prior earnings. This research uncovered that, even in an experimentally fabricated setting, high inequality can alter children’s behavior. However, research suggests that an individual’s position (i.e., whether you’re wealthy or poor) can have a marked impact on the experience of inequality (Côté et al., 2015). Thus in the current study, we explored how children react when they are high or low earners, and how this intersects with their experience of high and low inequality. To explore this, we adapted the paradigm from Kirkland et al. (2019) in two experiments (N = 120) to uncover the role of high and low inequality, respectively. Across both experiments, 4- to 6-year-olds played a game with several puppets where each accrued tokens over time. Importantly, the association between the players’ behavior in the games and the tokens awarded to them was random. In both experiments, children were either A) a high earner, or B) a low earner. However, in Experiment 1 children experienced high inequality, and in Experiment 2 children experienced low inequality. Across both experiments, children’s external donation rates did not differ after being a high or low earner. However, we observed an unexpected pattern in their resource division behavior. After experiencing high inequality (E 1), children who were high earners gave more to puppets who received less. However, low earning children in high inequality gave more to high earning puppets. After experiencing low inequality however (E 2), there was no difference in children’s resource division behavior after being a high or a low earner. These results suggest that, when outcomes are unequal, young children are highly sensitive to where they sit relative to others. This may result in children avoiding giving to those who earn a similar amount to themselves as this threatens their position relative to others. These results provide initial evidence that young children’s experience of economic inequality can impact how they treat others. Most of the world’s children will experience wealth disparities in some way, some more intensely and consistently than others. It is critical that research such as this be undertaken to better understand the social and cognitive impact this experience has.

O1.3       What you should have done: Children’s moral judgments incorporate representations of inaction

Presenter: Jonathan Beier

Jonathan Beier¹, Brandon Terrizzi², Amanda Woodward¹, Jonas Ventimiglia¹

¹University of Maryland, College Park, ²Cincinnati Children?s Hospital Medical Center

Moral theories are frameworks for judging how others uphold the principles that bind communities together. Their roots appear early: Even infants view helping as good and hindering/harming as bad. Yet morality involves more than judgments about the rightness or wrongness of actions. It also prescribes what one ought to do; in some contexts, inaction itself is impermissible. The cognitive foundations of children’s early reasoning about moral obligations are not well understood. Accounts of moral development typically take representations of a person’s actions as their starting point. But reasoning about moral obligations further requires the capacity to consider situations in which a person did not act at all. This poses two challenges. First, cases of morally relevant inaction may be unmarked. Second, representations of such inaction must be capable of entering into computations similar to those that support judgments about actions. We investigated children’s reasoning about one’s prosocial obligation to help others. Experiment 1 asked whether preschool-age children spontaneously notice one person’s failure to help another person in need, and whether they evaluate such unhelpful people negatively. Three- to 6-year-olds (N=128) viewed two videos each. Children in the Helpful condition saw one actor help another access an out-of-reach object; those in the Unhelpful condition saw her observe the other’s reach but do nothing. With age, compared to actors in no-help-needed baseline videos, children increasingly evaluated the helper more positively, and the non-helper more negatively. This was evident in their play choices and actor niceness ratings from 5 years, and their selective projection of positive (e.g., “always keeps promises”) and negative (e.g., “never shares”) qualities by 6 years. Experiments 2 and 3 asked whether 4- to 6-year-old children recruit representations of unhelpful inaction for more complex moral reasoning. Experiment 2 (N = 80) asked whether exculpatory factors lessen children’s negative evaluations of unhelpful inaction. Evaluations of inaction were unaffected by the presence of a barrier making it difficult for the actor to help; however, 8- to 10-year-olds were more forgiving of inaction when helping would be effortful. Experiment 3 (N = 96) asked whether unhelpful inaction–even if temporary–lessens children’s positive evaluations of helpful action. Children viewed hesitating helpers as less praiseworthy than immediate helpers (but only by 6 years, when controlling for the duration of expressed need). Experiment 4, with 3- and 4-year-olds, uses a simplified procedure to better compare emerging evaluations of helping and not helping. This ongoing project also assesses judgments about the permissibility of unhelpful inaction, to test the hypothesis that children’s explicit moral theories about helping precede their spontaneous attention to unhelpful inaction. In sum, results thus far demonstrate that preschoolers spontaneously detect instances of morally relevant inaction–a disposition critical for identifying violations of obligations in a dynamic world. Moreover, representations of helpful action and unhelpful inaction support social evaluations, personal preferences, and trait projections along similar developmental trajectories; and during early-to-middle childhood, representations of unhelpful inaction participate in computations just like those supporting children’s moral reasoning about overt actions.

O1.4       The development of beliefs about censorship

Presenter: Rajen Anderson

Rajen Anderson¹, Katherine Kinzler², Kayla Young¹

¹Cornell University, ²University of Chicago

What are children’s intuitions about the nature of censorship? Across four studies, we examined how 5-10-year-old children think about censorship and the social and moral norms around media and communication more broadly. In all studies, we presented children with vignettes describing scenes from eight fictional movies and asked them to decide which movies should or should not be watched. The eight vignettes featured acts across four different domains (harming a human, harming an animal, harming property, and purity violations); each domain was presented once as intentional and once as accidental. In Study 1, we found that children, in deciding both for themselves and for another child, censor movies with harmful content more often than allow those same movies. Older participants also censored less often than younger participants. In addition, children censored more movies with intentional harms than accidental harms, suggesting that children are attending to the moral content of the movies. In Study 2, we examined the role of the age of the audience in censorship decisions. For each movie, participants decided whether three potential audiences (a little kid, a big kid, and a grown-up) could watch the movie or not. We found that children censor more often for younger audiences than older audiences. We also replicated our intentionality effect from Study 1, but also found that intentions matter less when picking movies for a grown-up than for either of the kids. This interaction between audience and intentionality of harm suggests that children understand that norms regarding media consumption differ based on the age of the target. In Studies 1 and 2, children often spontaneously explained their decisions based on avoiding modeling negative behavior for the audience, suggesting that children censor based on the potential effects on the audience. We hypothesized that intentional actions may be more easily modeled and copied than accidental actions and thus motivations to avoid modeling negative behavior should attend to the intentionality of the act. In Studies 3 and 4, we directly examined this motivation by assigning participants to receive a motivation to either avoid making another kid feel sad or do something bad after watching the movies. We found that the role of intentionality on the number of movies allowed was greater when participants were trying to avoid making the other kid do something bad than when trying to avoid making the other kid feel sad. Together, our results indicate that children from age 5 make sophisticated decisions regarding censorship, not only engaging in censorship of film but also demonstrating motivations to avoid modeling harmful behavior for audiences based on the intentionality of the depicted harms. Our findings have implications for both theory on children’s understanding of moral and social norms but also in understanding the development of attitudes about censorship.

O1.5       Sociolinguistic development in a diverse, multilinguistic society: Evidence from 7- to 14-year-old children in Gujarat, India

Presenter: Ruthe Foushee

Ruthe Foushee¹, Mahesh Srinivasan¹

¹University of California, Berkeley

How does the presence of diversity in children’s linguistic environments shape social associations regarding the languages one speaks? Prior research on sociolinguistic development has focused on contexts in which a small number of linguistic distinctions (e.g., speaking one English dialect versus another) map predictably onto distinct social features (e.g., growing up in the South versus North, or being judged ‘nice’ versus ‘smart;’ Kinzler & DeJesus, 2013). The present study traced the development of sociolinguistic associations in a more complex, multilingual environment than previous studies. Specifically, we worked with children in Gujarat, India, who spoke an average of 3.9 languages, including languages corresponding to their parents’ geographic origins, languages used as common languages outside the home, and languages learned in school. 68 3rd (n=25), 5th (n=14), and 7th (n=23) graders responded to four counterbalanced blocks of questions regarding 8 languages or dialects that crossed regional, political, ethnic, and religious boundaries (Gujarati, Marathi, Hindi, Urdu, Tamil, Indian English, Standard American English, Mandarin). For all questions, students were given the option of not answering, to avoid forcing stereotypical responses. Speaker Associations. Children heard unlabeled audio samples of speakers of each language, and judged speakers’ wealth, geographic origin, and religion. While 3rd-graders responded unsystematically, selective associations appeared with age (e.g.,between being Muslim and speaking Urdu; being Hindu and speaking Gujarati/Hindi/Marathi; and being Christian and speaking English; all ps<.05). The age where robust geographic associations first appeared varied substantially by language, suggesting that children were basing their responses on experience in their environments. Hindi/Gujarati voices, for example, were reliably assumed to be from Gujarat before English/Mandarin voices were reliably assumed to be foreign, or before Tamil voices were assumed to be Indian. Stereotypical Speakers. Children guessed who ‘could be speaking’ from among five stereotypical faces (Hindu North Indian, Muslim, South Indian, Han Chinese, and White) in response to audio language samples or language names. Consistent with prior research, children’s face selections became more systematic with age (all ps<.05). Interestingly, children showed more stereotypical responding when languages were referred to by name (e.g.,who speaks Tamil?) compared to when only audio samples were played (p<.001), suggesting that children’s stereotypes about speakers come from culturally transmitted knowledge in addition to their own experiences seeing people speak. Essentialism and Learning. Children predicted how well each face would ‘come to speak [LANGUAGE],’ if the person had never heard it before and had just begun studying it. Despite increased stereotyping with age (see previous section), essentialist responses regarding the learnability of different languages based on speakers’ ethnicities decreased (p<.05), suggesting that older children understand the relation between language and ethnicity is non-deterministic. Faced with great linguistic and social variability, children track language-specific regularities, rather than generalizing across languages. This slow, selective process may be supplemented or even overridden by input from the surrounding culture, where associations may be conventionalized into stereotypes.

16:00 – 16:15: Transition time

16:15 – 17:45: Parallel Session 6 – 10

S5           How children’s understanding of social relationship guides their learning about others
Natalia Vélez, Stanford University
To find their place within the social world, children have to learn about the people around them: What do others know and like, who is friends with whom, and whom should they trust? If children learned about each person in isolation, these questions would be difficult to answer. Social relationships–e.g., kinship, friendship, group membership–highlight similarities between people and provide a window into their knowledge, preferences, and past experiences. This symposium explores how children’s understanding of social relationships influences their learning about, and behavior towards, new individuals. Study 1 finds that children attribute deep meaning even to social relationships based on minimal group membership; 5- to 8-year-olds show ingroup biases even when it is stressed that group membership is arbitrary. Study 2 explores how children use social relationships to infer further similarities between individuals. Between 4-9 years, children’s inferences become increasingly attuned to the nature of the relationship; for example, they expect close friends, but not distant compatriots, to share secrets. Study 3 takes a computational approach to examine how children make rich, graded inferences about others, based on minimal data on how preferences are distributed across groups. Study 4 considers the real-world consequences of representations of social groups. Both US preschoolers and adults have a default representation of friendship as being between members of the same race; in adults, this representation is associated with outcomes such as diversity in friendship networks. Collectively, these studies suggest that inferences based on social groups are a double-edged sword. On one hand, they allow children to efficiently learn about new people; on the other hand, these fast yet robust inferences can perpetuate divisions between groups. Presenters will together discuss the implications of these findings for understanding the developmental roots of prejudice.

Yarrow Dunham, Yale University - Minimal but meaningful: Probing the limits of randomly assigned social identities

Yarrow Dunham¹, Xin Yang¹

¹Yale University

People cluster into kinds and categories based on a wide range of identities, including sex/gender, race, and nationality. Some such groups are plausibly grounded in evolved differences (e.g., sex) or rich social meaning (e.g., nationality), but others seem much less meaningful because they are not clearly related to deep and important properties (e.g., teams picked randomly on the playground). It is intuitive to think that people develop stronger ingroup biases towards more meaningful groups, especially those that reflect important and stable distinctions between people. However, little developmental work has probed this intuition. Here we show that it may well be wrong. The present studies (total n = 151, age range 5 to 8) experimentally manipulated meaningfulness in randomly assigned novel social groups (the “Green Group” and the “Orange Group”) and measured the resulting ingroup biases (including preference, similarity, and resource allocation). To experimentally manipulate the level of meaningfulness we experimentally varied the group assignment procedure. Children placed their hand on a machine that we told them would assign them to one of two groups. Half of the children were told that the machine could look deep inside them to reveal their true category membership (the maximal condition), while the other half of the children were told that the machine assigned them to a group randomly (the minimal condition). Study 1 showed that even in the arbitrary and presumptively meaningless minimal condition, 5- to 8-year-olds developed equally strong ingroup biases as did children in the more meaningful maximal condition, despite the groups being well-differentiated on manipulation check items relating to meaningfulness. While we entered this line of work with the intuition that increasing the meaning associated with the maximal condition would increase ingroup bias, the results from Study 1 instead suggest that differentiating the two conditions would require decreasing the meaning of the minimal condition. This became the goal of Study 2. Study 2 explored the lengths required to effectively reduce ingroup biases by further emphasizing the arbitrariness of the grouping dimension in the minimal condition. We did this in two ways. First, we incorporated a training procedure involving coin flips to emphasize the unpredictability of random processes such as the group assignment procedure. Second, and perhaps more importantly, after the machine assigned them to a group the experimenter switched their group assignment for an explicitly external reason, claiming they had run out of wristbands of the initially assigned color. Now bias in the minimal condition was attenuated on preference and similarity measures, though it persisted in resource allocation behavior. These results indicate that it is surprisingly easy to create powerful group affiliations, and that one has to go to great lengths to counteract the tendency to imbue newly encountered social groups with rich affiliative meaning. These findings have critical relevance to the broader project of understanding the early emergence of prejudice and discrimination, and bolster the contention that mere self-categorization into a group is sufficient to induce robust ingroup biases in children (Dunham, 2018).

Zoe Liberman, UC Santa Barbara - (Un)common knowledge: Children use social relationships to determine who knows what

Zoe Liberman¹, Emily Gerdin², Katherine Kinzler³, Alex Shaw³

¹University of California Santa Barbara, ²Yale University, ³University of Chicago

Knowledge is power. Possessing information allows humans to competently interact with the complex world around them. For social agents, it is not only important to possess knowledge oneself, but it is also critical to know who is “in the know” and who is not. Indeed, effectively navigating social and communicative interactions requires understanding that some people know information that others may not. For example, if Alice is deciding whether to discuss Sam’s secret with Erica, it is important for Alice to know whether Erica already knows Sam’s secret. Exposing another’s private information can have devastating social consequences, so discussing the secret may be okay if Erica already knows it, but not if she doesn’t. One reasonable cue that people could use to predict other people’s knowledge is their social standing and their relationships with others. In our example, you may infer that if Erica and Sam are friends (rather than enemies), then Erica would be likely to know Sam’s secret. Here, we investigate if and when children begin to use social relationships to predict who has access to the same knowledge (“shared knowledge”), and whether their predictions change based on the type of social relationship and/or the type of knowledge. We also compare children’s reasoning about shared knowledge (e.g., a secret) to their reasoning about “common knowledge,” which everyone should reasonably be expected to know (e.g., the wrongness of moral violations). In three studies, we told 4-to 9-year-olds (N=227) that a target child knew three things: that hitting was wrong (moral knowledge), a personal secret (personal knowledge), and how to celebrate a novel holiday, Festivus (conventional knowledge). Then, we asked whether other character’s, such as the target’s friend (Studies 1-3), schoolmate (Study 1), national groupmate (Study 2), and sibling (Study 3) knew each type of knowledge. In all three studies, children accurately used relationships to infer what other people knew. In particular, they expected moral knowledge to be shared widely (e.g., to be “common knowledge”), but expected personal to be shared selectively with friends, and conventional knowledge to be shared with groupmates and siblings. Moreover, with age children increasingly considered both the type of knowledge and an individual’s social relationships when reporting who knew what. Our results provide support for a “Selective Inferences” hypothesis. That is, rather than relying merely on closeness, children recognize that the type of social relationship that connects two people is predictive of the type(s) of knowledge the two agents are likely to share. Overall, these studies suggest that children’s early attention to social relationships facilitates an understanding of how knowledge transfers–an otherwise challenging cognitive process.

Natalia Vélez, Stanford University - Preschoolers use minimal information about social groups to infer individuals' group membership and preferences

Natalia Vélez¹, Hyowon Gweon¹

¹Stanford University

Categories organize a jumble of experiences into flexible, generalizable knowledge. For example, if a child learns that a “blicket” squeaks, she might expect the next blicket she encounters to also squeak. Just as categories like “blicket” support children’s learning about objects, social groups support learning about people. While some classes of social groups have existed over evolutionary time (e.g., kin), others share deep, meaningful similarities in their knowledge and interests despite being a recent invention (e.g., cognitive scientists). The meaning attached to these groups thus stems, in part, from regularities among the people within it. In this work, we combined computational and behavioral methods to examine how children construct their knowledge of social groups from minimal patterns of evidence. In particular, we examined whether children can use statistical information about how preferences are distributed across groups to constrain their inferences about the preferences and group membership of new individuals. Children (N = 97, age 4?5) observed the preferences of novel agents (Gazorps) in two minimal social groups (blue team, red team). Children saw four members of each group choose their favorite of two novel fruits, bubba or kiki. Children then met a new agent (the target; Fig. 1a). In the Infer Group condition, children saw which fruit the target likes and were asked whether it is on the red or blue team. In the Infer Preference condition, children saw which group the target is in and were asked whether it likes kiki or bubba. Each child saw one of three possible distributions of fruits across the two teams, which differ in how strongly they support children’s inferences about the target’s group membership and preferences (Fig. 1b, left). We used a Bayesian beta-binomial model to formalize our hypotheses about children?s beliefs about each team and their expectations of the target. Overall, we find that children use this minimal evidence to make quick, flexible inferences about the target. In the Infer Group condition, children expected the target agent to belong to the group that was more likely to share its preferences–even in subtle cases where the target’s preference was not deterministically associated with a single group. In the Infer Preference condition, children expected the target to share its teammates’ preferences when preferences were consistent in both groups (Consistent & Diagnostic) and were split when preferences were inconsistent in the target’s group (Diagnostic). However, children were also split when preferences were consistent in the target’s group, but inconsistent in the other group (Consistent). Ongoing work is currently testing the possibility that children’s predictions are based not only on hypotheses about groups, but also overhypotheses across groups. These results are cause for both hope and concern. On one hand, these results suggest that children can build rich, generalizable representations of social groups from just a few observations and, critically, that these representations reflect the evidence they’ve observed. On the other hand, children in our task learned about fairly neutral properties of social groups that they’re not a part of. In naturalistic contexts, these inferences might easily go awry, supporting the formation of biases and stereotypes. Our results are thus relevant to understanding how prejudice forms–and how it might be corrected with evidence.

Arianne Eason, Washington University in St. Louis - The intergroup consequences of representing friendship as same-race

Arianne Eason¹, Lori Markson²

¹University of California Berkeley, ²Washington University in St. Louis

The United States’ racial landscape continues to be marred by pervasive patterns of structural segregation. In other words, our societal structure–that is, the patterns of relationships between entities within society and the arrangement of groups within society–continues to reflect the separation of Black and White individuals. Why have these patterns of structural segregation persisted over legal, economic, attitudinal, and cultural shifts? I propose that one possibility is that people mentally represent important relationships, particularly friendship, as being between same-race individuals as opposed to cross-race individuals. Towards this aim, I investigate whether this mental representation develops early (Study 1), is evident in adulthood (Study 2), and has the potential to maintain patterns of structural segregation (Study 2). Study 1. White children were presented with one target and four potential friends and asked to indicate whom the target should befriend (N=96; 8 trials; Study 1a: indicate 1 friend/trial; Study 1b: indicate up to 4 friends/trial). Children were more likely to indicate that same- (vs. cross-) race peers should be friends (Study 1a: same-race-75%, cross-race-25%, p<.001; Study 1b: same-race-90%, cross-race-53%, p<.001). Study 2. Using a highly repeated measures, within subjects design, a diverse set of college students (N=125) were exposed to 40 randomized pairs of people presented side by side (20 same race & 20 cross-race pairs), and in a forced choice paradigm asked to indicate whether that pair of people was “friends” or “not friends”. Participants’ racial attitudes and warmth towards each individual in the pair was measured. Overall, people were more likely to indicate that same-race pairs were friends (73%) compared to cross-race pairs (66%), t(4838)=5.329, p<.001. Importantly, the extent to which people mentally represented friendship as same-race was independent of people’s own racial attitudes, and perceptions of each individual pictured in the task, all |rs|<.096, all ps>.295. With the results of Study 1, these results suggest that even early in development people robustly represent friendship as same-race. Moreover, such representations are conceptually distinct from racial attitudes and warmth, suggesting an additional mechanism by which structural segregation may persist. Intergroup consequences of representing friendship as same-race. The more an individual represented friendship as same (vs. cross) race: the less interested they were in befriending novel individuals of a different race (p=.015); the less diverse their friendship network (p=.036); the less they reported believing that neighborhood segregation was a problem that needed to be rectified (p=.043); and the less diversity they included in a map of their ideal neighborhood (p=.005). All of these results remained significant even when controlling for people’s racial attitudes. Thus, representing friendship as same-race has both interpersonal and structural consequences, shaping friendship networks and residential neighborhood design. Discussion. Whereas prior research has traditionally focused on understanding people’s racial attitudes, or perceptions of others’ racial attitudes, the present results provide new evidence of the importance of understanding how people reason about relationships. Representing friendship as same-race may indeed perpetuate the pervasive patterns of structural segregation found in the U.S.


S6           Understanding individual differences in mathematics knowledge
Chair: Bethany Rittle-Johnson, Vanderbilt University
Cognitive development research often focuses on averages across groups of children. However, there are substantial individual differences in children’s knowledge, and these individual differences are predictive of children’s future academic success (e.g., Duncan et al., 2007). This is particularly true of math knowledge, as children vary widely in their math knowledge even in preschool. Further, weaker knowledge at school entry largely explains weaker math knowledge later in elementary school and accounts for lower math achievement among low-income children compared to their more affluent peers (Jordan et al., 2009). Identifying sources of these individual differences provides insights for both developing theories of learning and for designing effective interventions to improve the outcomes for all children. The current symposium brings together experts who consider varied sources of individual differences in math knowledge for children in preschool through middle school from diverse backgrounds. The first talk will focus on how individual differences in parents’ math anxiety impact math input in the home for two different age groups, highlighting the role of individual differences in math input children receive. The second talk will focus on individual differences in young children’s understanding of a core numeracy concept and relations to their broad math knowledge. The third talk will focus on the importance of individual differences in attention, visual-spatial working memory and ANS acuity for predicting math knowledge among preschoolers at high risk for math difficulties. The fourth talk with describe sex, SES and age differences in the relations between spatial skill and mathematics knowledge. Together, these talks will advance our understanding of how a wide range of individual difference factors influence developing math knowledge, including input from parents, knowledge of core concepts, cognitive skills, and child characteristics.

Julianne Herts, University of Chicago - The effect of individual differences in parent math anxiety on home math support

Julianne Herts¹, Deena Bernett¹, Sian Beilock², Susan Goldin-Meadow¹, Susan Levine¹

¹The University of Chicago, ²Barnard College

Individual differences in children’s math skills are often attributed to variations in intrinsic abilities. However, research suggests that math input that children receive in the home has important consequences for the development of these differences (Levine et al., 2010; Berkowitz et al., 2015). We explore how individual differences in parents’ math anxiety- fear or worry about math- impact math input in the home, which can affect children’s math development (Ashcraft, 2002). For instance, highly math anxious parents who provide a lot of homework help have children who underperform in math (Maloney, et al., 2015). Research suggests that when high math anxious parents are provided with structured math input, their kids’ math performance can improve (Berkowitz et al., 2015). Yet even in structured settings individual differences in parent math anxiety may impact parent-child math interactions. By examining parent-child math interactions across two different age groups, we found that highly math anxious parents provide lower quality math support than parents with lower levels of math anxiety. In study 1, we coded videos of 29 first graders (34% non-white) and their parents (76% female) doing math word problems from the Bedtime Math app together. We used a modified version of the Reformed Teacher Observation Protocol (RTOP; Sawada, 2001) scale to rate the quality of the instruction that parents offered their children. Math instruction was rated from 0-4 for each of six items: respecting the child’s prior knowledge, providing opportunities for exploration, promoting conceptual understanding, understanding of the math content, promoting critical thinking, and being patient. Families’ scores on this scale were higher when parent math anxiety, measured by the Short Math Anxiety Rating Scale (sMARS; Alexander & Martray, 1989) was lower (β = -.361, t=-2.145, p=.041). Instructional quality was also higher when child math ability, measured by the Woodcock-Johnson Applied Problems, was higher (β = .354, t=2.101, p=.045). In Study 2, we coded videos of 50 7th grade children (38% non-white) and parents (90% female) doing math problems together, and problems concerning the arts. Again, high math anxious parents were less supportive than low math anxious parents. While working on the math problems, high math anxious parents generated a smaller proportion of the dyad’s solution strategies than low math anxious parents (95% Highest Posterior Density Interval: -0.72 to -0.13 on a logit scale). In addition, higher math anxious parents disagreed less often with their children’s ideas (95% HPDI: -0.55 to 0.02). Furthermore, high, but not low, math anxious parents expressed more confusion about the math problems than they did about the arts problems (95% HPDI: 0.05 to 1.21). These differences were not observed in the context of the arts problems. High math anxious parents were found to be more hands off than low math anxious parents when working on math, but not arts, problems with their children. Across two grade levels, we find that parent math anxiety impacts parent/child math interactions. While interacting with younger children, math anxious parents provide lower quality support. While interacting with older children, math anxious parents provide less support. These studies provide insight into how individual differences in parents’ math attitudes relate to their math support, which sheds light on how variations in math achievement may arise

Bethany Rittle-Johnson, Vanderbilt University - 16 is one more than 15: The role of the successor principle in building mathematics knowledge

Bethany Rittle-Johnson¹, Erica Zippert¹, Ashli-Ann Douglas¹

¹Vanderbilt University

A key conceptual insight about counting, integers and arithmetic is the successor principle (cardinality for each count word is cardinality of previous count word plus one; Gelman & Gallistel, 1978; Sarnecka & Carey, 2008), and children understand it well after they become cardinal principle knowers (Cheung, Rubenson & Barner, 2017). However, little is known about individual differences in understanding the successor principle across a wide range of numbers or in both directions (adding vs. subtracting one), especially in children from diverse backgrounds, nor how these individual differences relate to math knowledge beyond counting. In Study 1, children in the Spring of Kindergarten (N = 65, mean age = 5.5 years, 52% non-white, 52% received financial assistance) displayed successor principle knowledge for numbers ranging from 15 to 116, using a unit task (extended from Cheung et al., 2017). At least 80% of children were successful on each item and 59% of children were successful on all 10 items. Knowledge was a bit lower for children who received financial assistance or who were from minority groups, but not significantly so. Most children (63%) could also count to 100 and these children also had significantly higher successor principle knowledge than those who could not count to 100, M = 74% vs 95% correct. Importantly, children’s successor principle knowledge was strongly related to their concurrent general numeracy and broad math knowledge, r(59)’s = .45 and .29, respectively, even after controlling for age and general cognitive skills (i.e., verbal and visual-spatial working memory and verbal ability). In Study 2, pre-K children (N = 212; mean age = 4.71 years, 47% non-white) displayed some successor principle knowledge for numbers ranging from 3 to 34 (M = 67% correct). Performance was above chance for every number but 12, including for new items that involved subtracting an object; 9% of children were successful on all items. Knowledge was significantly lower for children from minority groups (M = 64% vs. 71% correct), with income data forthcoming. Children’s successor principle knowledge was correlated with their concurrent general numeracy and broad math knowledge, controlling for age, r’s = .43 and .48, as well as with two measures of children’s oral counting knowledge, r’s = .48 and .50. Finally, some children received 5 training sessions that included activities focused on the successor principle for set sizes 1-10. However, these children did not gain greater successor principle knowledge than children who did not receive training (M = 67% vs. 68% correct). Overall, individual differences in children’s understanding of the successor principle persist through Kindergarten, with some children mastering the principle even for numbers above 99 by the end of Kindergarten. Children from minority groups may show some delays in this knowledge in pre-K. Importantly, these individual differences in successor principle knowledge are related to children’s general numeracy and broad math knowledge. This supports the theory that successor principle knowledge is core to developing math knowledge (Sarnecka & Carey, 2008; Cheung, et al., 2017). However, our training on the successor principle was not successful, nor was a past effort (Spaepen et al., 2018), indicating the need to develop effective instructional methods and to test for causal relations between successor principle knowledge and broader math knowledge.

Marcia Barnes, Vanderbilt University - Individual differences in attention uniquely predict math outcomes in preschoolers at high risk for math difficulties

Marcia Barnes¹, Alice Klein², Greg Roberts³, Anna-Mari Fall³, Bruce McCandliss⁴

¹Vanderbilt University, ²WestEd, ³University of Texas at Austin, ⁴Stanford University

Individual differences in domain-general skills such as executive functions and language predict growth in math in pre-kindergarten. Approximate number system (ANS) acuity, a domain-specific skill, may also contribute to individual differences in math; however, the strength and nature of this relationship has been questioned. This study tested what domain-specific and domain-general abilities at the beginning of pre-kindergarten uniquely predict math outcomes at the end of pre-k. We used a group selected for very low math knowledge; included attention (ability to sustain attention and ignore/inhibit distracting information) given its role in the development of executive skills and for all levels of information processing; tested ANS acuity in the presence of those domain-general skills (working memory, inhibition) thought to account for the relation of ANS acuity and math; and employed autoregressive models controlling for math at the beginning of pre-k. Children were in pre-k programs for children with SES disadvantage. From this general “at risk” sample, only children entering pre-k with the lowest math knowledge (541 of 1700 screened) were consented and randomized to a math plus attention training intervention, a math only intervention, or a business-as-usual control condition (Barnes et al., 2016). The sample was 47% female, 72% Hispanic, and 18% African American. Assessments and interventions were in the child’s first language (English or Spanish). Hypotheses for this talk were: 1) With the inclusion of the auto-regressor, domain-general cognitive abilities (attention, working memory, phonological awareness) will account for significant additional variance in math outcomes; 2) ANS acuity will not predict growth in mathematics in the presence of visual-spatial working memory and inhibition; 3) Attention will account for unique variance in math with other predictors in the models. Hierarchical regression analyses were conducted for the Test of Early Mathematics (TEMA-3) and the Child Math Assessment (CMA). For end of pre-k CMA, the model with time 1 CMA, condition and the cognitive predictors accounted for about 50% of the variance (30% for CMA pretest; 8% for condition; 12% for other predictors, with attention at 7% and working memory at 3%). Regardless of the order of entry, attention and visual-spatial working memory explained most of the variance in CMA scores among the cognitive measures. For end of pre-k TEMA-3, time 1 TEMA-3 scores accounted for 30% of the variance, and cognitive variables added an additional 9% (condition added less than 1%). Regardless of order of entry, attention showed the largest effects (8%) followed by visual-spatial working memory (1%) and phonological awareness (1%). ANS acuity explained minimal or no variance across measures and analyses. Findings suggest that attention and visual-spatial working memory are important for growth in mathematics in pre-k for children who start school with particularly low mathematical knowledge. Implications are as follows: 1) Given the strength of the effects for attention, attention could be an important addition to early assessment for risk; and 2) Intervention research might manipulate aspects of attention (of the instructional math environment and of the learner) as a means for improving math learning in young children at high risk for math difficulties.

Kelly Mix, University of Maryland - SES and sex differences in spatial skill and mathematics

Kelly Mix¹, Susan Levine², Alexander Burgoyne³, Tessa Johnson¹, Christopher Young²

¹University of Maryland, ²University of Chicago, ³Michigan State University

Children of different sexes and different socioeconomic status (SES) perform differently on cognitive and academic measures (e.g., Levine, Vasilyeva, Lourenco, Newcombe & Huttenlocher, 2005; Noble, McCandliss, & Farah, 2007). Understanding these differences is important to addressing achievement gaps and improving child outcomes. Whereas main effect differences have been investigated, few studies have examined differences in relations among skills. In the present study, we tested whether the relations between spatial skill and mathematics were moderated by individual differences in SES, sex, or both, as a secondary analysis of a previously published cross-sectional dataset (Mix et al., 2016; Mix et al., 2017). The dataset included 1592 children (K, 3rd, and 6th grade) tested with 14 measures of spatial skill and mathematics from a diverse sample of communities ranging from impoverished to upper middle class. Preliminary analyses of the combined dataset, presented previously, revealed that boys outperformed girls on both spatial (F(1, 1083) = 19.458, p < .001, η2p = .018) and mathematics tasks (F(1, 1083) = 5.009, p = .025, η2p = .005), and children from higher SES families also outperformed children from lower SES families on both spatial (F(5, 1083) = 21.898, p < .001, η2p = .092) and mathematics tasks (F(5, 1083) = 23.501, p < .001, η2p = .098), and no significant interactions. Using multi-group confirmatory factor analyses (MGCFA), we also observed configural invariance across subgroups, suggesting that the latent structure underlying performance on spatial and mathematics tasks was the same across sex and SES. However, further analyses have revealed subtle differences among subgroups. Hierarchical clustering of the item correlation matrices indicated that certain spatial tasks clustered with mathematics performance tasks for some groups but not others. For example, among kindergarten girls, the ability to perceive scenes from multiple perspectives was clustered with mathematics tasks but not with other spatial tasks. Perspective-taking was also clustered with mathematics items for low-SES but not high-SES kindergarteners. Given that kindergarten girls and those from low SES families also performed worse on mathematics than groups without this same clustering, perspective-taking skill may relate to a compensatory strategy that leads to better performance within these lower performing groups. The measurement invariance models of the MGCFA also indicated that certain items functioned differently between groups. Specifically, achieving partial invariance for the kindergarten subgroups required relaxing the intercepts for visual motor integration (high SES girls) and word problems (low SES girls). Taken together, these findings suggest that kindergarten girls may use abilities other than mathematics or space to answer visual motor integration and word problems items or that these items are biased in favor of girls at this developmental stage. In sum, males and high SES students outperformed females and low SES students on both spatial and mathematical tasks. However, the magnitude of these differences varied across age groups, and there were subtle differences in the way specific spatial measures related to mathematics across sex and SES. These differences might point to instructional approaches that could help close sex and SES gaps in academic outcomes.


S7         How parent authoritarianism and cultural upbringing shape children’s learning and proto-political cognition
Chairs: Annelise Pesch, University of Minnesota & Samuel Ronfard, University of Toronto at Mississauga
Authority powerfully impacts the thoughts and behaviors of adults and children. Relative to its impacts on adults, we know very little about how it shapes how children learn and develop. Children learn about authority -how to respond to it and who wields it- from their parents and their broader community. The present symposium examines how these messages influence how children learn, who they think they can become, as well as their thoughts about power relations between social groups. Papers 1 and 2 focus on the impact of parental authoritarianism on the decisions children make when learning from other people. Paper 1 examines how parental authoritarianism in a lower SES sample shapes American 3-year-old children’s selective trust across different types of informants. Paper 2 examines the relation between parental authoritarianism and 4- to 8-year-old children’s willingness to empirically verify an adults’ counterintuitive claim in the USA and China. Papers 3 and 4 go on to examine how children’s perceptions of who has authority shape their political aspirations and evaluations of social deviance, their proto-political cognition. Paper 3 finds that gender stereotypes about authority relevant traits lead 5- to-11-year-old American boys and girls to have different political aspirations. Paper 4 finds that cultural differences in parent authoritarianism impact 4- to-8-year-old Israeli children’s opinions about deference to authority figures and evaluations of social deviance. Together, this work suggests that the messages children receive about authority from their parents and their culture shapes how they learn and whom they believe should have political power.

Annelise Pesch, University of Minnesota - An examination of how individual differences in parent authoritarian values and economic experiences impact 3-year-olds' inferences about speakers

Annelise Pesch¹, Pearl Han Li¹, Katherine Ridge¹, Dante Cicchetti¹, Melissa Koenig¹

¹University of Minnesota

A large body of work has documented preschool-aged children’s selective preference for information from individuals with a history of reliability or competence. Recent work in this field has begun to examine how individual differences in culture, parenting and economic experiences impact children’s extensions of selective trust (Lucas, Lewis, et al., 2014; Reifen-Tagar et al., 2014), as well as how different types of agent characteristics impact both learning and practical decisions (Pesch & Koenig, 2018). These efforts have been conducted primarily with WEIRD populations. The current study aims to address this gap in the literature by examining how selective trust is extended by children from low socioeconomic (SES) households and how parent authoritarian values are related to these decisions. In the present study, 269 3-year-olds and their mothers were recruited. All families came from low SES homes. Children completed three versions of a selective trust task: accurate vs. inaccurate, benevolent vs. malevolent, and mom vs. stranger. In each case, children were given information about the agents (e.g. “she is smart/nice/your mom” vs. “she is not smart/mean/not your mom”) and asked to make epistemic decisions (who would you like to learn from) and practical decisions (who would you like to play with, share with). Mothers completed an Authoritarianism Questionnaire, which asked them to select values they believed to be more important for their child to have (e.g. “independence vs. respect for elders”). For epistemic decisions, we found a significant effect of question type such that children were more likely to ask the smart/nice/mom agent for information than endorse the information she offered. We also found a significant effect of trust task, with children more likely to ask and endorse information provided by their mother relative to the other two trust tasks. For practical decisions, there was a significant effect of trust task, with children more likely to ask for help and share coins in the Mom selective trust task. A median split on parent authoritarianism revealed that children whose parents reported having higher authoritarian values were slightly less likely to ask for information, ask for help, and share coins with the smart agent (See Fig 1). Taken together, these data offer the first comprehensive examination of how a diverse sample of children perform on different versions of the standard selective trust task and how parent authoritarian values relates to those decisions. The results suggest that for this sample, children’s epistemic decisions dissociate – they are less likely to adopt beliefs from those that they initially seek information from. In addition, we found a greater selective preference for mom compared to competent or benevolent agents who were less familiar. Children whose parents had high authoritarian values were less likely to distinguish between accurate and inaccurate agents across both epistemic and practical decisions. We will offer interpretations of these findings, taking into consideration how diverse experiences and parenting values are reflected in children’s epistemic and practical decisions.

Samuel Ronfard, University of Toronto at Mississauga - American and Chinese children growing up in more authoritarian homes are less likely to empirically verify a counter-intuitive claim

Samuel Ronfard¹, Eva Chen², Paul Harris³

¹University of Toronto, ²Hong Kong University of Science and Technology, ³Harvard University

Although children often rely on what they have been told, they sometimes check counter-intuitive claims through observation and experimentation. We tested the hypothesis that American and Chinese children growing up in more authoritarian households are less likely to test such claims. We presented 167 preschool children (USA: 68; China: 103) and 132 elementary school children (USA: 64; China: 64) with five different-sized Russian nesting dolls (that were painted and unrecognizable as dolls). After asking them to indicate the heaviest one (with all children choosing the biggest doll), we randomly assigned children into two testimony conditions: in the counter-testimony condition, the experimenter falsely stated that the smallest doll was the heaviest and that the biggest was the lightest, while in the confirming-testimony condition, the experimenter confirmed children’s intuitions that the biggest doll was heaviest. The experimenter again asked children which doll was the heaviest and then left, giving children an opportunity to seek empirical evidence by picking up the dolls in her absence. Before leaving, the experimenter provided half of the children with a mild prompt to explore: “I’ll move the dolls a bit closer to you.” Most children endorsed the experimenter’s claim, even when it was counter-intuitive: 78% of American children and 84% of Chinese children endorsed the testimony that smallest=heaviest. To investigate children’s exploration of the dolls, we coded the number of times they picked up each doll during the experimenter’s absence. Younger children picked up the same number of dolls, regardless of the testimony or prompt they received. In contrast, older children picked up a greater number of dolls when they received counter-testimony and were prompted to explore (counter-prompt condition). The number of dolls they picked up did not differ in the confirming-prompt, confirming-no prompt, and counter-no prompt conditions, Testimony X Prompt X Age Group, chi2 = 4.87, p = .03. Children who gathered disconfirming evidence reverted to their initial belief that biggest=heaviest. Children who did not gather such evidence continued to endorse smallest=heaviest. Parents were asked to complete a short survey assessing their endorsement of authoritarian and nonauthoritarian values (Feldman & Stenner, 1997). Parents’ endorsement of authoritarian beliefs was unrelated to preschoolers’ exploration of the dolls. Parents’ endorsement of authoritarian values predicted the number of dolls elementary school children picked up when they received counter-testimony and a prompt to explore, r = -.45, p = .023. Elementary school children whose parents endorsed more authoritarian values picked up fewer dolls than children whose parents endorsed fewer authoritarian values. To summarize, there were no cultural differences in the impact of counterintuitive testimony. Consistent with prior research (Ronfard, Chen, & Harris, 2018), only older children sought empirical evidence in response to such testimony; and of those children, those with less authoritarian parents explored more compared to those with more authoritarian parents. These results show that individual differences associated with authoritarianism are expressed in early childhood shaping not only whose information children value (Reifen Tagar et al., 2014) but also their willingness to verify what they have been told.

Reut Vraneski-Shachnai, Cornell University - Are presidents bossy? Boys' and girls' concepts of presidents differentially predict political aspirations

Reut Vraneski-Shachnai¹, Rachel Leshin², Andrei Cimpian²

¹Cornell University, ²New York University

Women in the United States are underrepresented in leadership positions across a range of professional domains. A notable example is the political sphere: since our country’s inception in 1776, all 45 elected presidents have been male, and as of 2019, women account for only 23.7% of the US Congress. Why do women continue to be underrepresented in positions of political authority? Here, we test the idea that disparities in political authority can be traced back, in part, to the stereotypes about gender and authority that children are exposed to. Authority figures are typically portrayed as both agentic (e.g., assertive, dominant) and competent – traits that are, across cultures, associated with men more than women. In contrast, traits of warmth and compassion, which are often associated with women more than men, are unrelated or even negatively related to notions of authority (Fiske et al., 2002). Given that agency and competence are central to the concept of a political leader (e.g., Eagly & Karau, 2002; Heilman, 2012), gender stereotypes that associate these traits with males may signal to girls that they are not suited for positions of political authority. In the present study, we investigated the types of traits children ascribe to presidents (i.e., agentic vs. competent vs. warm) and whether endorsement of these traits as important to being president predicts children’s interest in being a president when they grow up. For example, do girls who think presidents have to be agentic or competent display less interest in becoming one? To assess concepts of presidents, we asked children (N = 148, 52.7% girls; age range = 5 to 11; Mage = 7.87, SDage = 1.74) “What do presidents have to be like?” and coded all open-ended responses on the dimensions of agency, competence, and warmth. We also presented children with a series of traits that tapped these three dimensions (e.g., “likes to be in charge”, “smart”, “nice”), and asked them whether they think presidents should or can have these traits. Children’s scores on each dimension were combined across these two measures and could range between −100 to 100 (higher scores = the trait is more central to children’s concepts of presidents). To assess interest in political leadership, we asked, “Do you want to be the president when you grow up?” Affirmative answers were followed up with “Do you want to be president a little, some, or a lot?” Children’s responses were combined across these two questions and ranged from 0 (do not want to be president) to 3 (want to be president a lot). In line with our hypothesis, thinking that presidents are agentic and competent differentially predicted boys’ and girls’ aspirations: As Figure 1 (panels A and B) reveals, the more girls thought presidents needed to be high in agency and competence, the less interested they were in this role; in contrast, the more boys thought presidents needed to be high in these traits, the more interested they were. None of these patterns varied by age. With respect to warmth, thinking of presidents as warm increased with age and positively predicted both boys’ and girls’ interest in being president (Figure 1C). These findings provide preliminary support for the idea that gender disparities in politics may have roots in children’s concepts of political authorities as competent and agentic. Interventions to promote women’s involvement in political leadership should consider targeting young children.

Michal Reifen Tagar, Baruch Ivcher School of Psychology - Cultural differences in deference to authority and social convention are apparent already in childhood and contingent on parent authoritarianism

Michal Reifen Tagar¹, Aaron Kuri¹, Tahl Frenkel¹

¹IDC Herzliya

Individual and cross-cultural differences in sociopolitical authoritarianism have traditionally been assumed to only manifest in young adulthood. Recent work, however, suggests that related individual differences are apparent already in young children in a manner consistent with parent sociopolitical authoritarianism. In the current study we extend this work by focusing on children’s explicit expressions of authoritarian orientation – preference for deference to authority and social convention, and examining whether cross-cultural differences in authoritarianism among adults are also evident already in childhood. To test this, we recruited 170 mother-child (4-8-year-old) dyads from two cultural groups in Israeli society that differ in their level of authoritarianism: ultra-orthodox Jews and secular Jews. We expected that to the extent that cultural differences are apparent already in childhood, ultra-orthodox children would be higher both in their endorsement of deference to societal authority and in their intolerance of deviance from social convention relative to secular children; and that such patterns would correspond to differences in mothers’ level of sociopolitical authoritarianism. Mothers completed a survey assessing their sociopolitical authoritarianism. Children completed two tasks both presented in the form of a children’s story: in one they evaluated the legitimacy and desirability of a character making individual choices independent of the recommendations of ‘the king who rules the land’ (preference for deference to authority); and one evaluating a group member who defies a group convention (intolerance of deviance). Children’s responses were recorded on a six-point scale. As expected, we found that ultra-orthodox mothers were significantly higher in authoritarianism than secular mothers. Importantly, ultra-orthodox children where significantly higher both in their endorsement of deference to authority and in their intolerance of social deviance relative to secular children. Furthermore, across the two groups, both child deference to authority and child intolerance of deviance was strongly associated with mothers’ sociopolitical authoritarianism. Associations were similarly strong across cultural group as indicated by non-significant interactions between mother authoritarianism and group, for both child variables. These findings show that cultural differences in the endorsement of authoritarian attitudes are apparent already in young children and correspond to parent differences in sociopolitical authoritarianism. The findings suggest that parent authoritarianism shapes not only their children’s sensitivity to cues related to authority and convention as earlier work has demonstrated, but also to children’s proto-political orientation.


S8           Pretense, counterfactuals, and future hypotheticals: Relating different abilities to reason about possibilities in development
Chairs: Angela Nyhout & Daphna Buchsbaum, University of Toronto
Discussant: Ori Friedman, University of Waterloo
Children imagine and reason about alternatives to reality from early in development. They may pretend one object is another (e.g., Harris & Kavanaugh, 1993), consider how a past event could have turned out differently (e.g., Harris et al., 1996), or imagine a future state that differs from a current one (Atance & O’Neill, 2001). How are abilities to reason about alternatives in the form of pretense, counterfactuals, and future hypotheticals related to one another? Do they comprise the same cognitive process, or are they qualitatively different abilities? The three papers in this symposium present work bridging these different capacities to reason hypothetically. Paper 1 considers the relation between children’s ability to consider counterfactual outcomes and pretend outcomes, and suggests that pretense plays a role in the development of causal counterfactual inference. Cross-cultural evidence from Peruvian and American preschoolers shows a consistent relation between counterfactual conditional reasoning and pretense. Paper 2 examines the relation between children’s causal understanding of 3-variable physical systems and their ability to consider how the system would work under past counterfactual or future hypothetical interventions. Preschoolers answered causal questions correctly, but struggled to answer questions about counterfactual and future hypothetical interventions. Paper 3 considers the role of episodic future thinking (EFT) in children’s reasoning about anticipated, real rewards and hypothetical rewards, finding that EFT was only a predictor of children’s decisions about hypothetical rewards. These findings suggest differences in reasoning about a certain versus imagined future. Finally, our discussant, whose research expertise includes children’s pretend play and reasoning about fantasy and reality, will connect these findings in the context of the symposium’s overarching question.

Daphna Buchsbaum, University of Toronto - Causal learning, counterfactual reasoning and pretend play: A cross-cultural comparison of Peruvian and U.S. children

Adrienne Wente¹, María Fernández Flecha², Teresa Garcia³, Alison Gopnik³, Daphna Buchsbaum⁴

¹University of California, Berkeley, ²Pontificia Universidad Católica del Perú, ³UC Berkeley, ⁴University of Toronto

One important role for childhood pretense may be in the development of causal inference. Previously, Buchsbaum, Bridgers, Weisberg, and Gopnik (2012) demonstrated middle class U.S. children’s ability to maintain a newly learned causal relationship within a pretend scenario, and also found a correlation between children’s causal counterfactual reasoning abilities and their causal pretend play. These results could suggest that pretend play is related to the development of causal inference and counterfactual reasoning. However, while pretense is characteristic of human children across cultures, aspects of it have been found to vary by culture, and exactly how and what children learn via pretend play is still poorly understood. Middle class North American children engage in pretend play more often than children from other backgrounds, and some researchers suggest that extensive pretend play may not play a significant, or universal, role in learning (Lillard et al., 2013). We replicated the Buchsbaum et al. study in a population of low-income Peruvian children (N = 62), where pretend play may be less ubiquitous. Children learned a novel causal relationship (A causes C; B does not cause C.) Then they answered counterfactual questions about the causal relationship, and finally they engaged in pretend play and answered questions about the causal relationships in the pretend scenario. As in the North-American children, Peruvian children’s ability to consider counterfactuals about the causal system was significantly correlated with their ability to reason about it during pretense, r = 0.35, p = 0.005. Interestingly, Peruvian children were significantly better at answering questions about hypothetical causal outcomes during pretense than when asked counterfactually, t(61) = 3.95, p < 0.001. In addition, Peruvian children provided less consistent responses to both the counterfactual, t(120)= 3.54, p=.001, and pretense, t(120)= 2.417, p=.017, questions than did U.S. children. However, though the pretense and counterfactual scores differed across cultures, the relationship between the two measures was the same across cultures. Across cultures, children who provided accurate responses to the counterfactual questions also tended to provide answers during pretense that were consistent with the real world causal relationship. There was no difference between U.S. and Peruvian children’s pretense performance when counterfactual performance was taken into account, F(1, 118)= 1.6, p=.209, as well as no interaction between culture and pretense performance F(1, 118)= .278, p=.599. Taken together, this indicates that the relationship between pretense and counterfactual reasoning is similar across cultures, even though children’s average ability to answer these questions differed across cultures. Follow up work with a Low-SES US sample is ongoing, but preliminary data show a pattern similar to Peruvian low-SES children, and different from higher-SES children in the same geographic region. This is the first cross-cultural evidence for a relationship between causal counterfactual reasoning and pretense. Future research will explore more specifically how cultural variance in play may relate to the development of causal reasoning while engaged in pretense and when reasoning.

Angela Nyhout, University of Toronto- Children's counterfactual and future hypothetical inferences about different causal structures

Angela Nyhout¹, Hilary Sweatman², Patricia Ganea¹

¹University of Toronto, ²McGill University

Whether and when children can contemplate counterfactual possibilities (e.g.,”if X hadn’t happened, would Y have happened?”) is the subject of significant debate. Some research suggests that children can reason counterfactually in the preschool years (Harris et al., 1996; Nyhout & Ganea, 2019), while other work shows limits to this ability (Rafetseder et al., 2013). In particular, the relation between causal and counterfactual reasoning has been debated, with some researchers arguing that counterfactuals are integral to causal inference (Lewis, 1973; Woodward, 2003) and others arguing that causal knowledge is primary (Edginton, 2011). Several accounts predict that children should be able to reason counterfactually about an event if they understand its causal structure. For instance, according to Bayesian learning models, children’s judgments about counterfactual interventions should be compatible with their representation of causal structure (Gopnik & Schulz, 2007). The developmental evidence for this to date is mixed (Burns & McCormack, 2011; Frosch et al., 2012; Schulz et al., 2007). Here we asked (1) whether children who show a robust understanding of various causal structures will also be able to consider the effects of counterfactual interventions to these systems, controlling for their complexity in terms of the number of variables and causal relations; and (2) whether children would be better able to consider possibilities when framed as future hypotheticals, given debates about differences between counterfactual reasoning and future thinking (Beck et al., 2006). Children learned about 3 physical systems in 3 causal models (within-subjects): causal chain (A causes B, which causes C), common cause (A causes both B + C), and common effect (A + B both separately cause C). Stimuli were presented on a touch screen. Children interacted with each system by activating switches on the screen. They were asked causal questions, followed by counterfactual (CF) or future hypothetical (FH) questions about each system. In Study 1, 3-, 4-, and 5-year-olds (N=72) answered causal questions with a high degree of accuracy, binomial tests, ps<.001. Three-, 4-, and 5-year-olds’ performance was above chance for the common effect (overdetermined) model only, p<.034. In Study 2, 4- and 5-year-olds (N=38, target N=48) answered CF or FH questions about the 3 causal models. Preliminary GEE analyses indicate that children’s performance is not significantly different between CF and FH conditions, Wald X2(1)=0.53, p=.466, nor between models, Wald X2(2)< 2.4, p>.13. In both conditions, children’s performance was better than chance for the common effect model (p=.001), but not for the common chain or common cause models. This is surprising in light of previous findings that children have more difficulty answering CF questions about overdetermined (common effect) events than single-cause events (Rafetseder et al., 2013). The results indicate that children can reason causally before being able to reason about hypothetical interventions, and point to commonalities in counterfactual and future hypothetical reasoning about the same events. Follow-up studies will examine the relation between counterfactual and future reasoning within-subjects and using different types of causation (e.g., psychological).

Teresa McCormack, Queen's University Belfast - Saving for the future: Episodic future thinking and delay of gratification for real versus hypothetical rewards

Teresa McCormack¹, Patrick Burns¹, Patrick O’Connor¹, Cristina Atance²

¹Queen’s University Belfast, ²University of Ottawa

Boyer (2008) has argued that episodic future thinking (EFT) facilitates prudent decision-making, suggesting the developmental hypothesis that the emergence of EFT helps children become better at delaying gratification. However, virtually all of the studies to date on EFT and decision-making have been with adults and have used entirely hypothetical monetary rewards. We will report a developmental study that examined the relation between (EFT) and children’s ability to delay gratification for either real or hypothetical rewards. Participants were 131 children aged 7-to-11-years. They completed a Delay of Gratification (DoG) task, a Delay Discounting (DD) task, an EFT Interview, and a subjective temporal distance judgment task, as well verbal and non-verbal IQ measures. The DoG task involved deciding whether to take an immediately available small real reward (e.g., stationery items, candy), or to wait for a larger reward after a delay of 1 day, 1 week, or 1 month (4 trials per delay). The computerized DD task involved purely hypothetical monetary rewards, with participants deciding to wait for a larger reward over various delay periods ranging from 1 week to 3 months. Thirty-one participants were removed who did not produce systematic DD data (Johnson & Bickel, 2008). For the remaining participants, model-based Area Under the Curve was calculated; this serves as a measure of performance on the DD task. In the EFT interview, children were asked to describe particular future events that they were going to participate in, two for each of three periods of time in the future (tomorrow, next week, a few months’ time). The extent to which children described particular future episodes was coded using the episodicity scoring system described by Coughlin, Lyons and Ghetti (2014). The subjective temporal distance estimation task involved participants making judgments about how far away a variety of time points in the future felt to them. We found that performance on the DoG task was strongly correlated with that on the DD task even when controlling for age and IQ (rpartial = 41, p < .01). Episodicity scores for the three future time periods were only weakly correlated, and showed differential relations with other measure. Regression analyses explored the predictors of DoG or DD performance. The only significant predictors of DoG performance were age and the IQ measures, whereas neither of these measures significantly predicted DD performance. Rather, DD performance was predicted by episodicity scores for the few months’ time period and by performance on the subjective time estimation task. Our findings suggest that although there is a relation between children’s ability to wait for real versus hypothetical rewards, these appear to be separable skills that are related to different processes. The ability to engage in EFT was only a predictor of performance on the DD task in which children made decisions about hypothetical rewards. These findings cast doubt on the extent to which EFT plays a role in facilitating children’s delay of gratification in situations in which the nature of the future larger reward is concrete and visually available. Rather, EFT may be more important in facilitating the imagining of an abstract future.


Oral Papers II

O2.1       Do children think that scientists are smart? The influence of gender on kindergartners’ understanding and use of descriptors about science and intelligence

Kelly Runyon

Kelly Runyon¹, Vanessa Diaz¹, Cameron Smith¹

¹Virginia Tech

Prior research shows a change after kindergarten in children’s stereotypes about scientists. In the Draw-A-Scientist task, 5-year-olds drew scientists of roughly equal gender ratios, but the ratio grew to 4-to-1 favoring males once in elementary school (Miller, Nolla, Eagly & Uttal, 2018). In a 2017 study from Bian et al., kindergarten children were asked to choose a ‘smart person’ out of gender balanced picture arrays. At the end of kindergarten girls were less likely than boys to choose members of their own gender, and began to reject activities said to be only for smart children. Thus gender stereotypes are not only present even before the start of elementary school, but they are already affecting which activities girls choose. To our knowledge, no research has been conducted on children’s understanding of the relation between science and intelligence. The current study sought to answer the following questions: Do kindergarten age children associate being smart with being a scientist? Are children’s gender stereotypes about science and intelligence affecting which activities they choose for themselves? What factors may be related to gender stereotypes for boys and girls? 48 5-year-olds (M=68 mo., F=25, M=23) attending public kindergarten in rural SW Virginia participated. Tasks were adapted from Bian et al. (2017) to also include questions about science. When asked to choose the smart one or the scientist out of gender balanced arrays, girls and boys differed significantly with each group overwhelmingly choosing pictures of their own gender (smart: t(46)=3.35, p=.001; scientist: t(46)=2.27, p=.01). Boys also associated the two questions (r(23)=.45, p=.05), while girls did not. When asked if they would like to play games for smart children and for little scientists, girls were more likely to want to play the smart game compared to the one for little scientists, t(24)=2.99, p=.01. There was no difference for boys, indicating a more conceptual disconnect for girls. Finally, when examining which factors influenced children’s answers, we found that parents’ occupation had a specific gendered influence. While boys’ likelihood to choose men as being smart was correlated with their dad’s occupational prestige (the level of education their job requires, r(23)= -.47, p=.05), for girls, whether the mom was employed at all was positively related to girls wanting to play both the game for ‘smart kids’, r(25)=.54, p=.01, and for ‘little scientists’, r(25)=.45, p=.05. Girl’s desire to play the game for ‘smart kids’ was positively correlated with mother’s occupational prestige, r(25)=.47, p=.05. No cross-gender parental occupation effects were found. In conclusion, despite the disconfirming messages that girls receive about women in science such as in children’s media, girls were still more likely to choose women instead of men as scientists, and did not differ from boys in willingness to engage in activities for ‘smart kids’ and ‘little scientists’. It is possible that this may change once they enter elementary school (Miller et al. 2018). In addition, two of our tasks confirmed that boys associated science and intelligence more than girls did. Finally, parental occupation influenced boys and girls differently along clear gender lines. In this rural low SES sample, we see a compounding of socioeconomic risk whereby the restricted professional opportunities of the parents is having a direct impact on the possibilities children see for themselves.

O2.2       Anticipation of social backlash and girls’ interest in leadership

Presenter: Andrea Vial

Andrea Vial¹, Andrei Cimpian¹

¹New York University

Throughout the world, top decision-making positions and leadership roles are dominated by men, and research indicates that this disparity is due at least in part to gender bias. Adult women are often met with resistance and viewed as interpersonally hostile when they behave assertively (Rudman & Phelan, 2008). These assertive behaviors are seen as required for leaders and valued in men, but contradict the traditional feminine role (Heilman, 2001). Women anticipate and often seek to avoid social backlash by refraining from assertive behaviors and authority roles (Moss-Racusin & Rudman, 2010), contributing to gender gaps in leadership. The current research examined whether awareness of (and the tendency to avoid) social backlash in response to leader-like behaviors emerges at an early age (Haun & Tomasello, 2011). Specifically, we tested whether girls (more than boys) anticipate social backlash for leader-like behaviors, and whether they actively seek to avoid such backlash by eschewing opportunities to lead. Over time, girls’ early avoidance of leader roles as a way to avoid social backlash could contribute to the underrepresentation of women in leadership. In two studies, we examined the development of beliefs about the social costs of leadership. Study 1 tested whether girls compared to boys anticipate stronger social backlash from peers for stepping up to be in charge. We read children (n = 99; 49 girls; ages 5-10) four stories, each featuring either a boy or a girl (between participants) who stepped up to be in charge of a group activity. Each story happened in a different setting (e.g., at the park), and each story featured a different group of children, either: (a) a small same-gender group (e.g., two girls); (b) a large same-gender group (e.g., 10 girls); (c) a small mixed-gender group; or (d) a large mixed-gender group. After each story, we asked participants four questions to gauge their anticipation of social backlash against the leader (e.g., “Would the other children like him/her more or less?”). Both girls and boys anticipated stronger backlash for other-gender leaders compared to own-gender leaders (p = .032), but–importantly–girls anticipated stronger social backlash than boys overall (p = .010). Study 2 (data collection ongoing) was designed to test whether girls in particular avoid being in charge due to anticipated social penalties. We are presenting children (ages 5-10) with a group game in which a player may choose to be the “boss,” and measuring participants’ interest in being the boss (i.e., the leader), their leadership self-efficacy (i.e., how effective they think they would be as leaders), and their anticipation of social penalties and resistance from other players. Preliminary results (n = 61; 34 girls) suggest that anticipated social penalties may have a unique influence on girls’ (vs. boys’) leadership aspirations. For boys, interest in being in charge was associated only with leadership self-efficacy (p = .001). In contrast, for girls, interest in being in charge was associated not only with self-efficacy (p < .001), but also with anticipated social penalties (p = .009) and peer resistance (p = .050). As a whole, these studies provide initial evidence that fear of social backlash can influence girls’ interests from an early age, discouraging the pursuit of leader-like activities and potentially contributing to the persistence of gender gaps in leadership.

O2.3       Children’s intergroup attitudes: Insights from Iran

Presenter: Haleh Yazdi

Haleh Yazdi¹, David Barner¹, Gail Heyman¹

¹University of California, San Diego

Children generally favor individuals in their own group over others, and some developmental theories argue that merely being a member of the outgroup is sufficient to induce a negative bias. However, it remains unclear whether the specific nature of the outgroup matters. This issue was investigated among 7-8 and 11-12-year-old Iranian children (N = 71). We tested whether children are equally biased against different outgroups or if feelings toward outgroup members are driven by (a) their perceived similarity to the ingroup (Similarity Hypothesis), (b) social and political relations between the ingroup and outgroups (Intergroup Relations Hypothesis) or (c) the relative social status of the ingroup and outgroups (Social Status Hypothesis). We explored these hypotheses by assessing children’s desire to affiliate with members of different social groups and whether these desires extend to other social judgments, including whether group members are perceived as trustworthy and worthy of loyalty. Participants evaluated ingroup members and three different outgroups: Iranian children from another school, Arab children, and children from the United States. Children’s evaluations closely aligned with the predictions of the Social Status Hypothesis, with Americans ranked as the highest social status group and evaluated as positively as ingroup members, while Arabs were ranked as the lowest social status group and evaluated the most negatively. These patterns were evident on measures of affiliation, trust, and loyalty. These findings, which provide some of the first insights into the social cognition of Iranian children, point to the role that social status can play in the formation of intergroup attitudes.

O2.4       Pragmatic reasoning leads children to draw inferences about unmentioned categories from generic language

Presenter: Kelsey Moty

Kelsey Moty¹, Marjorie Rhodes¹

¹New York University

Adults frequently use generic language to communicate information about social groups to children (e.g., “boys play sports”). Generics make explicit claims about the nature of kinds (e.g., that boys generally play sports; Carlson, 1977), but may also implicitly convey information about other unmentioned groups (e.g., that girls do not play sports). While previous research speaks to how children interpret information explicitly conveyed by generics (e.g., Cimpian & Markman, 2009; Rhodes, Leslie, Bianchi, & Chalik, 2017), little is known about whether generic language leads to these kinds of implicit inferences about unmentioned categories. Studies 1 and 2 tested whether and at what age children draw inferences about unmentioned categories from generic claims. In Study 3, we tested whether pragmatic reasoning (i.e., using social context to evaluate the speaker’s intended meaning) serves as a mechanism underlying these inferences. We preregistered our hypotheses, procedures, and analyses on OSF. In Study 1, we introduced children ages 4 to 6 (N = 236) to a novel social dichotomy between zarpies and gorps (marked by clothing color). Across 4 trials, participants heard either generic (e.g., “zarpies are good at climbing trees”) or specific statements (e.g., “this zarpie is good at climbing trees”). Participants then rated whether 2 additional individuals (one zarpie and one gorp) were also good at climbing trees. We found that across all ages, children who heard generic statements generally responded that other zarpies were good at climbing trees but that gorps were not (across 64% of trials). However, children who heard specific statements did not share this intuition, responding in this pattern across only 38% of trials (GEE binomial model: Language: χ2(1) = 27.98, p < .001). The tendency to draw this inference from generics began to emerge in our youngest participants (4s: generic 50% vs. specific 32%; p = .007) and increased with age (see Figure 1a; Language x Age (continuous): χ2(1) = 4.05, p = .04). In Study 2 (N = 60 4- to 6-year-olds), we replicated these findings using a truncated introduction to the novel social dichotomy between zarpies and gorps (Language: χ2(1) = 37.2, p < .001), suggesting that the tendency to make inferences about unmentioned groups from generics is present even for groups that children know relatively little about. In an ongoing Study 3 (N = 108 4- to 7-year-olds), we examined whether pragmatic reasoning underlies these inferences about unmentioned groups. To test this, we manipulated the informativeness of the generics that children heard (i.e. whether the speaker was knowledgeable or not). Preliminary findings (see Figure 1b) suggest children make these inferences when the generic was said by a knowledgeable speaker (53.8% of trials) but do so less when said by an unknowledgeable speaker (30.9% of trials; Condition: χ2(1) = 14.16, p < .001). The tendency to make these inferences from knowledgeable but not unknowledgeable speakers increases with age (Condition x Age (continuous): χ2(1) = 6.62, p = .01). Together, these studies provide evidence that a) children infer implied meanings from generics that go beyond what was explicitly stated and that b) children’s pragmatic reasoning serves as a mechanism by which children make these inferences. We will conclude by discussing the role that generic language may play in inadvertently communicating social stereotypes to young children.

O2.5       Learning about the social world through pragmatic inference

Presenter: Mahesh Srinivasan

Mahesh Srinivasan¹, Nadya Vasilyeva², Monica Ellwood-Lowe¹

¹UC Berkeley, ²Princeton University

Children readily learn information about the social world from language. For example, hearing generic language about a social group–e.g., “Boys are good at math”–leads children to essentialize it and form stereotypes about its members (Gelman & Roberts, 2017; Rhodes, Leslie, & Tworek, 2012). However, information about social groups can also be communicated in more indirect ways that go beyond literal sentence meanings (Grice, 1975). If a teacher has just observed a classroom of boys and girls and says “The boys are good at math,” they may express that the girls are worse than the boys at math, because they could have instead said “The boys and the girls are good at math”. Such inferences critically depend on the situational context: no inference about the girls would arise if the teacher had only observed a classroom of boys before making their statement. Recent research suggests that, although preschoolers struggle to make scalar implicatures (e.g., Huang & Snedeker, 2009), they readily make pragmatic inferences when they are supported by the context (Stiller, Goodman, & Frank, 2015), and that this provides a mechanism for learning more about the world (Horowitz & Frank, 2016). Here, we apply these ideas to explore whether children can learn about the social world from what a speaker could have said–but chose not to say–about a social group. We introduced 46 3-5 year-olds and 102 adults (data collection ongoing) to two novel alien groups on a faraway planet, Stripeys and Dotties. Participants heard a story about an astronaut who learned about the planet from a resident scientist, and made claims about how good the aliens were at different activities. For example, the astronaut stated that “the Stripeys are pretty good at building chairs”, and the participants were asked to rate how good they thought the Dotties’ chairs are (unmentioned alien), how good the Stripeys’ chairs are (mentioned alien), who makes better chairs, and rate their certainty. Our critical manipulation concerned the context preceding the astronaut’s claim. In the broad context, the astronaut made the claim after having looked at folders containing pictures of objects made by both Stripeys and Dotties. In contrast, in the narrow context, the astronaut had looked only at the folder with pictures made by the mentioned alien type. We expected that in the broad context, participants would be more likely to make an inference that the mentioned aliens are more skilled at the target activity than the non-mentioned aliens, compared to in the narrow context. This is because in the broad context there is a salient, more inclusive alternative utterance–e.g., “The aliens are good at building chairs”–making the astronaut’s choice to exclude one alien group informative. Our results confirm this prediction: children showed this pattern in their ratings of the objects built by the mentioned vs. unmentioned aliens, and, less reliably, in forced choice judgments (“who’s better at making chairs?”), but not in certainty ratings; adults showed this pattern across choices and certainty ratings. These findings suggest that even preschoolers are capable of learning about the social world by “reading between the lines”, perhaps drawing stronger inferences than adults.

17:45 – 19:00: Poster session 2 & Exhibitors



08:00 – 08:30: Coffee and registration

08:30 – 09:00: Announcements and awards

09:00 – 10:00: Plenary Address 2: Adriana J. Umana-Taylor, Harvard University, Harvard Graduate School of Education

Socio-developmental conceptions of adolescents' ethnic-racial identity and the potential for intervention”

Identity formation is a fundamental developmental process that has significant consequences for youth adjustment during adolescence and beyond. Consistent with these notions, findings indicate that among ethnic-racial minority youth in the U.S., exploring their ethnic-racial identity and gaining a sense of clarity regarding this aspect of their identity can serve a protective function and promote positive youth development. In this session, I will introduce an intervention curriculum (i.e., the Identity Project) grounded in developmental theory and focused on engaging adolescents in the processes of ethnic-racial identity exploration and resolution. I will present findings from a randomized controlled trial in which 12-week, 18-week, 1-year, and 2-year follow-up data provide support for program efficacy and its cascading effects on positive youth development. Thus far, the Identity Project intervention is demonstrating promising results that have the potential to significantly shape how we work with youth in school settings to promote identity formation and, in turn, psychosocial adjustment. I will end with a brief discussion of next steps, including exploring the potential universal nature of the program, which is motivated by the premise that identity formation is a developmental task that confers psychosocial benefits to all youth and the fact that ethnic-racial identity is an increasingly salient social identity to young people in multiple regions of the world.

10:00 – 10:30: Refreshment break

10:30 – 12:00: Plenary Symposium 2: Relating to others: Implications for cognitive development

Nameera Akhtar, University of California, Santa Cruz - Stretching the social

Human infants are characterized as social by nature. Early behaviors taken as evidence of their innate sociality generally include eye contact, smiles, and positive vocalizations in dyadic interactions with others. As infants mature, they begin to alternate gaze between objects of interest and their social partners (joint attention) and show signs that some people are special to them by becoming distressed when separated from them (attachment). My goal in this talk is to highlight other less frequently studied ways in which infants and young children show their social interest and attempts to connect with others, drawing on anthropological studies of infants in understudied communities, a recent study that examined how parents of non-speaking autistic children report connecting with their children, and studies of autistic intersubjectivity. By attending to alternate ways of displaying sociality, we can profitably broaden the notion of social behavior beyond what is typically assumed on the basis of studies of WEIRD neurotypical children and their families.    

Dima Amso, Brown University - Social relevance as a cue to learning and attention in infancy

For human infants, social  relationships  are a  source of primary biological rewards such as food and warmth, as well as the most likely providers of emotional and social rewards such as comfort and responsiveness. We have examined mechanisms underlying both infants’ learning from relative social reward and infants’ learning in order to maximize the procurement of toy reward in social contexts. In one study, we presented 7-month-old infants with video stimuli that parametrically increased in social-emotional value (male stranger, female stranger, mother) or in visual attention value as a control (static image, slowed silent cartoon, dynamic cartoon). After establishing infants’ baseline pupil and eye blink responses to these stimuli, we paired the videos with arbitrary shape cues in an associative learning task. Infants showed superior learning from their own mother’s video and a heightened anticipatory arousal pupil response to the mother-associated cue following learning.  These data indicate that relative social value is important for attention and learning and transfers to arbitrary associated stimuli. In another study, we again asked how social contexts might organize infant learning and behavior, this time using a hierarchical rule learning framework. In particular, we examined whether infants’ perseverative errors on the classic A-not-B task might reflect an adaptive attempt to procure the toy reward by learning how the the experimenter organizes toy-well-action rules. Results from 9-month-old infants indicated that  reaching accuracy was better on trials that were consistent with the experimenter-governed rule structure, even if these trials required infants to inhibit a prepotent motor response.  These results are discussed as mechanistically relevant to the broader hypothesis that social stimuli may serve as important organizing cues for learning and attention mechanisms in infancy.

Natasha Cabrera, University of Maryland - Father-child interactions and language development in the early years

Research on the aspects of social interactions that promote language development has highlighted the importance of parents but also the active role that children play in this process. Overall, this research has not typically included fathers and has not explored the variation in the quality of parental and children’s input in low-income families. In this talk, I first show that there is a large variability in the quantity and quality of language use by low-income fathers during interactions with their children, and that this variation is accounted by levels of education and depression. I then present evidence that during social interactions with their fathers, children learn communication skills in ways that may be different from mothers. Compared to mothers, fathers read less often but use more dialogical strategies, which predict to language skills because it increases the child’s interest in reading. When examining child-directed interactions, we find that fathers and mothers responded contiguously to about half of infants’ vocalizations bids. The majority of parents’ responses to infants’ nonverbal bids are noncontiguous and non-semantically related to the child’s interest at that moment. Therefore, providing support for low-income fathers and encouraging them to engage in dialogic reading and to respond contiguously and contingently to children’s communicative attempts may be an important way to address the language gap we see in children from low-income families compared to their wealthier peers.

Meredith Rowe, Harvard University - The importance of conversations for preschool children’s language development and learning

Caregiver input shapes children’s language development, yet the child plays an important role in this process. Research is converging on the importance of back-and-forth conversations for language development and learning. In this talk I will provide evidence that during the preschool years, children who engage in more conversations, and specifically more conversations about abstract or decontextualized (removed from the here and now) topics, are found to have better later oral language skills as well as skills in other cognitive domains. Thus, encouraging extended conversations between caregivers and preschoolers may be a powerful means for promoting both language and cognitive development.

12:00 – 13:15: Lunch on own or Lunch workshops

13:15 – 14:30: Poster session 3 & Exhibitors

14:30 – 16:00: Parallel Session 11 – 15

S9           Children’s understanding of social hierarchies and interventions to reduce status prejudice
Chair: Xin Yang, Yale University
Over the last 30 years income gaps favoring America’s wealthiest have grown, creating an increasingly stratified backdrop against which development occurs. How do children come to understand these inequalities and what can we do to reduce their potential negative impact? Four talks leverage a range of approaches to explore these questions. First, Yang reveals increasing nuance in children’s representations of the wealthy. As children age they increasingly associate the rich with high social power but also come to see wealth inequalities as unfair and begin to evaluate the wealthy more negatively on some dimensions. Second, Terrizzi finds that children also expect people with normative authority to be indifferent to others’ needs, suggesting that they associate social power with a lack of prosociality. At the same time, children reliably associate helpfulness with niceness, again suggesting dissociations between status and valence. Third, Mandalaywala explores race as a particular manifestation of inequality. The study explores how racial and economic characteristics of communities contribute to associations between race and status. She finds that neighborhood income inequality and racial diversity predict children’s use of race as a cue to status, while income inequality also leads children to overestimate their own economic status. Fourth, Liu shifts gears to focus on ameliorating the negative consequences of high-status preference by testing interventions to reduce children’s negative attitudes toward low-status groups. She finds that messages focusing on between-group similarities and within-group differences, as well as explanations focusing on contextual and historical reasons for group differences, have positive effects, whereas messages promoting learning about and appreciating group differences did not. Taken together, we shed new light on children’s understanding of social hierarchies and their psychological consequences in early to middle childhood.

Xin Yang, Yale University - Powerful but mean: Developing a nuanced conceptualization of the wealthy

Xin Yang¹, Yarrow Dunham¹

¹Yale University

Past work suggests that children have an overly rosy view of rich people: they prefer the rich, attribute competence, warmth, and popularity to the rich, and expect them to share more with others (e.g., Ahl & Dunham, 2017; Horwitz, Shutts, & Olson, 2014; Li, Spitzer, & Olson, 2014; Shutts, Brey, Dornbusch, Slywotzky, & Olson, 2016). Such effects seem consistent across development (from age 4 to 9; or even to age 14, e.g., Sigelman, 2012), but many could stem from valence-matching or halo effects, in which children simply associate the rich with positive attributes without having any deep understanding of wealth or the wealthy (e.g., they merely think that those who have nicer stuff are nicer). Also, past work has often employed questions closely related to the possession of resources (e.g., popularity or resource-sharing), meaning that we know much less about how judgments of the wealthy extend to other domains. In two studies (total N = 164), we introduced 4- to 12-year-old children to novel social groups that differed in wealth and measured attitudes, stereotypes, evaluations, behavioral expectations, perception of inequality, and understanding of the link between wealth and social power. We provide the first evidence that children develop a nuanced conceptualization of wealth as they age. Results showed 1) an age-related change in pro-rich attitudes with strong pro-rich biases in 4- to 5-year-olds disappearing by ages 9 to 12; 2) a domain-sensitive evaluation system, with 4- to 5-year-olds thinking the rich are both smarter and nicer, 7- to 8-year-olds thinking the rich are smarter but not necessarily nicer (responding at chance), and 9- to 12-year-olds no longer thinking the rich are smarter and thinking that the poor are nicer; 3) resource-specific behavioral expectations: children expected the rich to donate more material resources, but with age they expected the poor to donate more time; 4) an increase in perception of inequality: 4- to 5-year-olds rated the wealth gap as fair but older children rated it as unfair; and 5) a developing understanding of the wealth-power link: starting around age 5 and increasing with age, children associated the rich with power and the poor with a lack of power (in both interpersonal interactions and control over broader, societal issues). These findings suggest that as children age they develop a more nuanced conceptualization of wealth that goes beyond mere positivity. Discussion will draw connections between these findings and previous work on children’s pro-rich attitudes and expectations, and will explore how our findings might inspire new work on children’s understanding of wealth inequalities and power dynamics.

Brandon Terrizzi, Cincinnati Children's Hospital Medical Center - Young children and adults associate social power with indifference to others' needs

Brandon Terrizzi¹, Amanda Woodward, Jonathan Beier²

¹Cincinnati Children’s Hospital Medical Center, ²University Of Maryland

In hierarchical societies, what do we expect from people at the top? Early in life, children use horizontal social relationships (e.g., affiliation) to predict selectivity in prosocial behavior. But it is unknown whether they believe that asymmetries in prosocial behavior are characteristic of vertical relationships (e.g., differences in social power) as well. In children’s views, will powerful people look out for the less fortunate, or does high status bring indifference to others’ needs? We investigated the development of children’s intuitions about the prosocial responsiveness of authority figures. Four- to 7-year-old children (N=64) and adults (N=32) viewed two videos, featuring different actor pairs, in counterbalanced order. In both videos, one woman struggled to obtain an out-of-reach object. In the Helpful video, a second woman handed the object to her. In the Unhelpful video, the second woman observed the first woman’s reach, but did not help. After each video, participants answered two questions about the two actors in it: Which woman is “in charge” (i.e., gets to make all of the rules), and which one is “nicer”? How the second woman responded to the first woman’s instrumental need influenced participants’ judgments about their relative authority, Χ²= 16.67, p < .001. In the Helpful condition, neither children nor adults had consistent intuitions about who was in charge (both ps < .11). However, in the Unhelpful condition, both age groups viewed the unresponsive actor as holding authority over the person whom she did not help (46/64 children, p < .001; 28/32 adults, p < .001). Children’s responses did not vary with age. The second woman’s prosocial responsiveness also influenced participants’ attributions of relative niceness, Χ² = 5.07, p = .024, though this tendency was greater for children relative to adults, Χ² = 7.38, p = .025. In the Helpful condition, children viewed the helpful actor as nicer (49/64 children, p < .001), but adults did not (21/32 adults, p = .110). In the Unhelpful condition, both age groups viewed the needy actor as nicer than the unresponsive one (54/64 children and 29/32 adults, both ps < .001). These findings establish that, by at least the preschool years, children’s intuitive theories include expectations for links between power and prosociality. Further, the judgments produced by these theories appear to be stable across development. Building from these observations, we will discuss possible trajectories of change in the specificity of children’s theories–i.e., whether they are specific to the association between rule-making authority and instrumental need or whether they also predict associations between other notions of power, status, prosociality, and concern for others. We will also detail new research investigating whether these theories describe only the expected frequencies of behaviors or whether they also incorporate norms about the permissibility of an authority figure’s indifference to others.

Tara Mandalaywala, University of Massachusetts Amherst - It's the economy stupid: Economic characteristics of children's neighborhoods predict race-status covariance

Tara Mandalaywala¹, Marjorie Rhodes²

¹University of Massachusetts Amherst, ²New York University

Between ages three to seven, children begin to use race to predict social status (e.g., expecting Black Americans to have fewer economic resources than White Americans: Mandalaywala, Tai, & Rhodes, 2017; Shutts et al., 2016), yet we know little about how these beliefs develop. In particular, the extent to which the racial and economic characteristics of children’s communities contribute to the development of their beliefs about status has rarely been examined (Odgers & Adler, 2018). In the present study, we tested whether neighborhood-level economic or racial factors predicted: (1) whether children (3.5 – 7 years old, Mage = 4.98) used race as a cue to status (N = 46), and (2) children’s evaluations of their own status (N = 74). We measured the extent to which children across New York City used race to predict wealth-based status (using the Houses Task: children were presented with a White and a Black child and a nicer house and a rundown house, and asked which child lived in which house), as well as whether they used race to predict social power (using the Rope Task: children were introduced to a rope with 6 pegs that could be used to assign status to target stimuli based on both access to resources and decision-making power, and were asked to place a White and a Black child on the rope). Children also indicated their own status on the second task. We extracted neighborhood-level data from the 2017 American Community Survey for each participant’s residential zip-code, examining two economic factors: neighborhood income distribution (i.e., the extent to which the distribution of household incomes was skewed or normally distributed within the neighborhood) and neighborhood wealth (i.e., whether the median neighborhood income fell above or below the median income for New York City); two racial factors (percent of White inhabitants and percent of Black inhabitants); and one racial-economic factor (difference in median income between White households and Black households). Only income distribution and percent of Black inhabitants predicted children’s use of race as a cue to status, and only on the Houses Task. First, children living in neighborhoods with a skewed income distribution were more likely to use race to predict status than those in neighborhoods where income was normally distributed, β = -1.88, SE = 0.87, z = -2.15, p = .032 (Fig. 1a). Second, children living in neighborhoods with fewer Black inhabitants were more likely to use race to predict status than children living in neighborhoods with more Black inhabitants, β = -7.67, SE = 3.90, z = -1.97, p = .049. Finally, only income distribution predicted evaluation of children’s own status; children in neighborhoods with a skewed income distribution rated themselves higher in status than children from neighborhoods with more normally distributed incomes, β = -0.53, SE = 0.23, t = -2.27, p = .026 (Fig. 1b). Surprisingly, this was true regardless of neighborhood wealth (e.g., whether income distribution was skewed toward affluence or poverty). These results suggest that the relative distribution of incomes across a community are more important than absolute wealth for the development of children’s beliefs about race and status, as well as for their beliefs about their own status. Future studies should carefully investigate this relation further, to understand exactly what information children extract from income distribution, both about others as well as about themselves.

Vivian Liu, New York University - What can we tell children to improve their attitudes toward low-status groups

Vivian Liu¹, Andrei Cimpian¹

¹New York University

As children develop an understanding of the status hierarchies in their society, they also begin to prefer higher-status groups (e.g., Hailey & Olson, 2013). Since these attitudes form the basis for discrimination toward low-status groups, it is important to investigate how they can be modified. The current research compares three theoretically-grounded interventions intended to reduce children’s negative attitudes toward low-status groups. The first involves “colorblind” messages, which de-emphasize group membership by focusing on similarities between groups and drawing attention to individual differences within groups (Wolsko et al., 2000). The second intervention involves “multicultural” messages, which promote learning about group differences and appreciation of these differences (Rosenthal & Levy, 2010). The third intervention consists of extrinsic explanations (that is, explanations that appeal to contextual and historical reasons), which have been shown to prompt children to judge status disparities between groups as unfair (Hussak & Cimpian, 2015, 2018). The present research examines how these different messages affect children’s attitudes towards low- and high-status groups. We presented a diverse sample of 4- to 7-year-old children (N = 110; 38% White, 19% Asian, 17% Latinx, 8% Black, and 17% other) with two novel groups that differed in both non-status traits (e.g., preferences) as well as status (e.g., wealth). Children then heard one of three messages about these differences: 1) colorblind messages that downplayed the differences (e.g., “differences don’t really matter”, “deep down, they are all the same,” “everyone is special in their own way”), 2) multicultural messages that celebrated the differences (e.g., “differences are what make them special,” “it is important to learn about these differences,” “it is a good thing that they are different”), and 3) extrinsic explanations for the differences (e.g., “they moved from a different country,” “they had to start over in this new place”). Instead of an intervention message, children in a fourth (control) condition simply heard the description of the disparities for a second time. We examined the effects of the three types of messages with a broad set of measures: attitudes toward the two groups (i.e., warmth), perceptions of homogeneity, biological essentialism, attitudes toward the status disparity, and resource allocation behavior. Overall, the two interventions that had the largest positive impact on children’s attitudes toward the low-status group appeared to be those involving colorblind messages and extrinsic explanations. For instance, both of these messages increased liking of the low-status (vs. the high-status) group compared to the control condition (ps = .010 and .008) and multicultural condition (ps = .022 and .019). However, these two messages were not uniformly superior to multicultural messages; the scores for the three interventions were similar, for example, on the homogeneity and resource allocation measures. This research provides a promising first step toward devising interventions that improve children’s attitudes toward low-status groups.

S10         The symbol-grounding problem in numerical cognition: Insights from developmental psychology
Chair: Dan Kim, The Ohio State University
How do children learn the exact meaning of numerical symbols and number words? In recent years, there has been an on-going debate regarding the mechanisms by which symbolic numbers acquire their semantic meaning–i.e., the symbol-grounding problem. Some argue that the meaning of symbolic numbers is learned by associating symbols with physical quantities, whereas others argue that the representation of symbolic numbers develops through inferential and structural learning without perceptual grounding. The symposium is comprised of presenters that investigate the symbol-grounding problem along different developmental stages, providing valuable insight into the ontogenesis of mathematical concepts in human development. Dr. J. Cantlon will compare the perceptual processing of numerosity and logical rule learning in monkeys with human children and adults from different cultures. Dr. D. Hyde will discuss the relationship between number words and physical quantities in young children and their neural activations. D. Kim and Dr. J. Opfer will discuss the limit and capacity of the analog magnitude representation as the basis for symbolic-number representation. Lastly, Dr. L. Yuan, Dr. L. Smith, and Dr. K. Mix discuss the relational and syntactic pathway to number learning that is not built upon physical quantities. With this diverse input by leading scholars of various backgrounds, the current symposium will convene an excellent set of researchers to discuss important issues in numerical cognition, its development, and the emergence of symbolic concepts.

Jessica Cantlon, - Universal numerical rules in primates and people

Jessica Cantlon¹

¹Carnegie Mellon University

Primitive logical and perceptual processes form the basis of human cognitive development. Humans expand on those primitive processes in profound ways attributable to unique features of their genome, development, and culture. Comparative studies of people and animals, cross-cultural comparisons, and developmental analyses can reveal the relative contributions of evolution, culture, and development to human cognition. We present two studies in which we compare the performance of monkeys, human children, and adults from the US and Amazon cultures on the same numerical tasks and stimuli.

In one study we investigated the representation of numerosity comparatively in monkeys, children, and US and Amazonian adults. Monkeys, 4-year-old children, and adults categorized visual arrays in an unconstrained task wherein categorization by either the numerical or spatial dimensions of the stimuli was correct. All groups showed a spontaneous bias to categorize the stimuli based on the numerical dimension but cultural variables affected human categorization – exposure to mathematics education increased the numerical bias in the Amazonian and child groups. The results suggest a universal bias to represent numerical value across humans and non-human primates, as well as a psychological interaction between formal education and numerical perception in humans.

In a second study we investigated the capacity for recursion during sequential rule learning. Monkeys, 3- to 4-year-old children, and adults were trained to order four parentheses following a center-embedded rule of openA-openB-closeB-closeA, as in the sequence “[ { } ]”. Following training, subjects were tested with novel stimuli to measure generalization of a center-embedded rule (eg., a “( < > )” sequence). Human children and adults from the US and Amazon cultures spontaneously generalized the center-embedded rule whereas monkeys did not. Monkeys were given additional exposure to the training rule and showed evidence of recursion after 20 sessions. Working memory capacity was related to the degree of recursive rule use in humans. The results suggest that humans spontaneously apply recursive rules independently of culture and with limited experience – an ability unique to humans and not observed equally in non-human primates. However, the capacity to learn simple recursive structures is present in non-human primates. Large working memory capacity may be a developmental and evolutionary precursor to the ubiquity of recursion in human psychology, including in the domains of mathematics and language.

The spontaneous representation of ‘number’ as an actionable property of the world and the capacity to learn complex rules are universal cognitive capacities shared across primate species, stages of human development, and different human cultures. Humans are unique from other primates in the degree to which they spontaneously extract ‘number’ from other quantitative dimensions and in their proclivity for extracting complex recursive structure from sequences. We think that these unique features of human cognition are related to their opportunities to observe numerical rules from their culture and also to their ability to “observe” rules internally in a sufficiently large working memory.

Daniel Hyde, University of Illinois at Urbana-Champaign - Neural sensitivity to number word meaning before and after learning to count

Daniel Hyde¹, Ilaria Berteletti², Yi Mou³, Selim Jang¹

¹University of Illinois at Urbana-Champaign, ²Gallaudet University, ³Sun Yat-sen University

Despite decades of scientific inquiry and debate, it is still unclear how children learn the meanings of number words (see Carey, 2009 for a review). At the center of the current debate is whether children initially associate non-verbal approximate numerical magnitudes with number words to gain leverage on their meaning (e.g., Dehaene, 1997; Wagner & Johnson, 2011; Gelman & Gallistel, 2000) or whether this association occurs only later in the developmental process and therefore is not consequential for initial learning (e.g., Le Corre & Carey, 2007; Carey, 2009). To assess sensitivity to number word meaning, we measured the neural response as 3-4-year-old children (n = 67) engaged in a cross-modal number word-quantity comparison task. Specifically, children were asked to indicate whether a given array of 1-3 animals was more or less numerous than a number word (one, two, or three) presented immediately beforehand as event-related potentials (ERPs) were measured from the scalp. We reasoned that if children had associated approximate numerical meanings to number words, then the ERP response to the quantity should be influenced by its numerical ratio relationship to the number word, an established signature of approximate numerical processing (e.g., Hyde & Spelke, 2009; Pinhas et al., 2014). Preliminary analyses show neural sensitivity to the numerical relationship between the quantity and the number word, even in children at the earliest stages of number word acquisition (i.e., subset-knowers). Further analyses suggest that the brain response of children who demonstrated a more extensive understanding of number words (i.e., counting principle/cardinal principle knowers) was modulated by both the numerical magnitude and ordinal relationship (greater than/less than) between number word and quantity. Contrary to some accounts (e.g., Carey, 2009; Barner, 2017) and in support of others (e.g., Spelke, 2017; Piazza, 2010), these results suggest that approximate numerical meanings are associated with number words earlier in the learning process and, thus, could provide a basis by which children come to understanding the meanings of other number words. Although preliminary, these results also suggest that an important qualitative difference between children who understand more number words and those who understand less may be a sensitivity to the ordinal relationships between numbers.

Dan Kim, The Ohio State University - Visuospatial factors in numerosity representation: Development of math concepts from perception

Dan Kim¹, John Opfer¹

¹The Ohio State University

The analog representation of numerosity has been viewed too noisy to ground the precise representation of number, but its limits have not been systematically examined. Despite the inevitable covariance between numerical and non-numerical dimensions (e.g., a positive correlation between numerosity and total area), research on numerosity representation has focused on minimizing the correlation. However, controlling for non-numerical correlates of numerosities increases visual entropy of the display, which may extrinsically increase the noise (i.e., Weber fraction) in numerosity representation. If this is the case, the noise is not an invariant constant, but changes with non-numerical factors. In the two studies presented here, we increase the extent to which non-numerical dimensions correlate with the numerical dimension and examine the effects of three ancillary, visuospatial dimensions (normalization, concatenation, and alignment) to assess the contribution of irrelevant perceptual properties to numerical comparison. In Study 1, we investigated the influences of the visuospatial dimensions by asking 46 adults to compare two numerosities with or without controlling for the visuospatial features as well as two numerals in nine number comparison tasks. We found that the ratio effects on speed and accuracy decreased considerably with visuospatial manipulation, as did the representational noise index, the Weber fraction, for numerosity. When all visuospatial dimensions supported the numerical dimension, the noise in numerosity comparison was as small as that in numerical comparison, and the largest discriminable successors in numerosity comparison were as large as those in numeral comparison. The results suggest that the noise in adults’ numerosity comparison is mostly extrinsic rather than intrinsic. In Study 2, 44 4- to 7-year-old children completed four numerosity comparison tasks, where visuospatial manipulations led to the most improvement in Study 1, along with a numeral comparison task. Children performed more slowly and erroneously and exhibited greater noise than adults in all conditions, indicating the noise-to-noiseless developmental change in numerical comparison. More importantly, the visuospatial dimensions had similar, but more substantial effects on numerosity comparison in children: children compared numerosities faster and more accurately than numerals when visuospatial properties were controlled. In addition, the Weber fraction in numerosity comparison decreased significantly with the visuospatial controls, to the degree that the noise in numerosity comparison was smaller than that in numeral comparison, and the largest successive numerosities that children could accurately compare were larger than the largest successive numerals. These findings imply that the representation of numerosity is less noisy than that of numeral in young children when the extraneous noise from non-numerical dimensions is removed. Together, we demonstrated that the noise in the analog representation of numerosity comes largely from external properties that reduce the signal-to-noise ratio and increase perceptual demands for processing of the numerical dimension. This was especially true for young children who have not gained a full understanding of symbolic numbers. Thus, under optimal conditions, there is nothing about the perceptual system preventing humans from learning the meaning of symbols through association with a large set of objects.

Lei Yuan, Indiana University - Learning numbers as a system of symbols and their relations

Lei Yuan¹, Linda Smith¹, Kelly Mix²

¹Indiana University, ²University of Maryland

Numbers, like words, refer to things in the world. But also like words, numbers get their meaning not just through the represented quantities but through their relations to other numbers and the syntactic rules of the symbol system. Research on early numerical development has focused on how number concepts are grounded in the perception of referred to physical quantities. Formal teaching has also focused on how mathematical concepts can be learned via concrete manipulatives. In three studies, we provide evidence for an independent pathway to numerical knowledge that is not grounded in physical objects but is built upon the relational, syntactic structure of number symbols. Study 1: Distinct developmental trajectories for number symbols and physical quantities Studies have shown the predictive relation between perceptual discrimination of quantities and mathematics achievement. However, advanced mathematics at its core is not about specific quantities but the systems of relations among quantities as variables. We can generate a number (e.g., ten-gazillions) that is vastly beyond human perception of quantities because the syntax of the number system is recursive and affords infinite generalization. Using both between- and within-subject designs, we asked 475 preschoolers to identify written multi-digit numbers (or dot arrays) with their spoken names or to indicate the larger quantity between two written numbers (or two dot arrays). Preschoolers could reliably map spoken number names to written forms and compare the magnitudes of two written numbers. However, these abilities were not related to their non-symbolic representation of quantities. Study 2: Learning numbers based on the symbols themselves To provide evidence for the learnability of symbolic numbers without grounding to physical quantities, we conducted a week-long training study (N = 165) during which preschoolers saw written multi-digit numbers and heard corresponding number words embedded in casual learning activities (e.g., book reading). Participants not only learned the numbers appeared during the training but generalized how multi-digit numbers are named to novel numbers and showed a better understanding of numerical magnitudes. To further demonstrate this learning in “learners” with no prior knowledge, we trained a state-of-art Recurrent Neural Network with limited pairs of number words and written symbols. The model performed at human-level accuracy on both trained and novel numbers. Study 3: Learning numbers with relational-based manipulatives Different from dot arrays, mathematics manipulatives (e.g., base-10 blocks) have perceptual structures that highlight the structures in the number system. However, studies on the efficacy of manipulatives in teaching show mixed results. To investigate the right context for symbol grounding, 226 kindergarteners participated in a training study using various manipulatives. The effectiveness of manipulatives depends on whether the relational similarities between manipulatives and number symbols are highlighted, while extraneous features of the manipulatives are downplayed. In contrast, training with number symbols avoids such tradeoff and led to robust improvement. Overall, number knowledge can be bootstrapped from the symbol system itself–without grounding to physical objects–written numbers and spoken number words carry corresponding relational structures that contribute to the meaning of symbolic numbers.

S11         Metacognitive development in early childhood: Mechanisms and implications
Chair: Christopher Gonzales, University of California – Davis

Children’s metacognitive ability has long been held as a driving force in cognitive and social development (Flavell, 1979, Koriat & Goldsmith, 1996). Children’s ability to monitor and act upon their subjective feelings of uncertainty and knowledge states is crucial for optimal self-guided learning, help-seeking behavior and general curiosity about the world (Dunlosky & Metcalfe, 2009, Ronfard et. al., 2017). Whereas early accounts posited stark limitations in young children’s metacognitive ability (Lyons & Ghetti, 2010), more recent research has elucidated the development of metacognition during the preschool years (Coughlin et al, 2014, Lyons & Ghetti 2013). Researchers have also posited new theoretical accounts (e.g., Roebers & Feurer, 2016) linking children’s early metacognitive ability to other higher order cognitive skills (e.g., executive functions, perspective taking, etc.) and highlighting clear implications for educational outcomes. The current symposium consists of four presentations which investigate the mechanisms and implications of children’s early metacognitive development. The first presentation covers the associations between metacognitive assessments, executive functioning and academic achievement in the transition from preschool to kindergarten. The second presents data from an intervention study investigating the relation between children’s metacognitive judgements and numerical cognition. The third talk presents recent work examining how metacognition and help seeking are related in the content of memory retention. The fourth talk presents two studies examining the domain-generality of children’s uncertainty monitoring across different perceptual modalities. Together, these presentations offer new insight into how young children’s metacognitive abilities relate to broad array of both cognitive and educational outcomes and also explore its developmental mechanisms.

Carolyn Baer, University of British Columbia - Perceptual certainty representations are domain-general in childhood

Carolyn Baer¹, Darko Odic¹

¹University of British Columbia

From an early age, children can reason about the certainty of their perceptual experiences, such as differentiating between being sure vs. unsure of how many cookies are on a plate. Two competing theories have emerged about the representational nature of perceptual certainty. Under one, certainty is represented as a mathematical transformation of the perceptual signal itself (e.g., the standard deviation of a neuronal tuning function; Maniscalco & Lau, 2014), and is therefore highly domain-specific. Under the other theory, certainty is represented in a more domain-general format (e.g., the probability of an outcome being true), allowing otherwise independent perceptual dimensions to be compared in their certainty (De Gardelle & Mamassian, 2014). While work with adults has frequently found evidence for domain-general perceptual certainty (De Gardelle & Mamassian, 2014), the developmental picture is limited and so far inconsistent with this work (e.g., Vo et al., 2014, Geurten et al., 2018).

Here, in two experiments, we provide evidence that perceptual certainty is domain-general in children as young as six. In the first experiment, 6 – 9 year-old children perform either three perceptual discrimination tasks (Number: which side has more dots; Area: which blob is bigger; and Emotion: which face is happier), or a relative certainty judgement, in which children see two discrimination trials varying in difficulty (e.g., an easy and a hard number trial), and had to judge which of the two trials they are more certain of answering correctly. We find that while children’s perceptual discrimination accuracy is unrelated across the three dimensions, that children show strong correlations across the three certainty tasks, providing evidence for shared perceptual certainty representations.

In the second experiment, we conduct a more stringent test of domain-generality by asking children to make within- vs. between-domain comparisons of their certainty. If certainty is truly domain-general, children should be able to compare their certainty states across perceptually distinct dimensions just as effectively as they can compare it within these dimensions. Forty-eight 6 and 7-year-olds were presented with Number, Area, and Emotion perceptual discriminations. Questions were presented in pairs – either Within-Dimension (e.g., number/number) or Between-Dimensions (e.g., area/emotion). After making the discrimination decision for each of the two trials, children reported their relative certainty in the questions (‘which answer are you more sure you got right?’). Consistent with a domain-general account, children were able to compare their certainty between different perceptual task types, t(47) = 3.73, p < .001, d = .54, which was not different from their performance on comparisons within the same perceptual task type, F(1, 47) = .075, p = .39.

Together, these results suggest that – by age six – children’s perceptual certainty is represented in a domain-general format.

Christopher Gonzales, University of California - Davis - Uncertainty monitoring predicts academic achievement at the transition to kindergarten

Christopher Gonzales¹, Alexis Merculief², Isabella Sciuto², Alexis Tracy², Jasmine Karing², Megan McClelland²

¹University of California – Davis, ²Oregon State University

Metacognitive development, along with other higher-order executive functions, is a driving force of children’s increasing ability to engage optimal decision making. In school settings, children’s ability to monitor their subjective feelings of uncertainty i.e., engage in uncertainty monitoring, when faced with difficult test questions or ambiguous instructions, is a key mechanism of self-guided learning and knowing when to seek help from a teacher or peer. Research on children’s metacognition has demonstrated a robust ability by middle childhood (e.g., Lockl & Schneider, 2004) with connections to academic outcomes (Zimmerman, 1990). Until recently, the dominant position was that children’s metacognitive ability was severely limited in early childhood, but newer empirical work has demonstrated that young children demonstrate uncertainty monitoring, quantified as reporting higher confidence for correct compared to incorrect responses in cognitive tasks (Coughlin et al, 2014, Lyons & Ghetti 2013). Although uncertainty monitoring in young children has been linked to judicious help seeking and response withholding, relations with early academic outcomes has yet to be established. In the current study, we sought to explore the relation between young children’s uncertainty monitoring, executive functioning and academic abilities during the transition to kindergarten. Participants (N = 100, 62 female, Mean age = 67 months) came from a larger ongoing study of school readiness which recruited children enrolled in Head Start programs the year prior to kindergarten. In the fall of kindergarten, children completed a perceptual judgement uncertainty monitoring task (Lyons & Ghetti, 2013) and standardized measures of early numeracy, literacy and expressive vocabulary (Woodcock et. al., 2001). Additionally, they completed battery of executive functioning tasks, Day-Night Stroop (Gerstadt, et al., 1994), Dimensional-Change Card Sort (Zelazo, 2006), Head-Toes-Knees-Shoulders (McClelland, et. al., 2014), and Auditory Working Memory (Woodcock et. al., 2001). Moderate correlations were found between children’s uncertainty monitoring and their performance on the DCCS as well as with early numeracy and expressive vocabulary (r = .22 – .26, p < .05). In a multiple regression, children’s uncertainty monitoring uniquely predicted children’s DCCS performance (ß = .27, p = .008), early numeracy (ß = .24, p = .018) and expressive vocabulary (ß = .27, p = .004) above and beyond their age, ELL-status, and overall judgement accuracy on the uncertainty monitoring task. When additional measures of EF were included in the model, children’s uncertainty monitoring was a unique predictor of children’s expressive vocabulary (ß = .19, p = .024), but not early numeracy (ß = .10, p = .177) above and beyond other measures of EF. Together, these findings begin to illustrate how children’s early metacognitive ability is related to broader measures of executive functioning as well as their academic abilities. Collection of six-month follow-up data is about to be completed, which will allow for examining the longitudinal relations between these measures during the transition to more formal educational environments.

Clarissa Thompson, Kent State University - Children can monitor and control their number line estimates

Clarissa Thompson¹, John Dunlosky¹, William Merriman¹

¹Kent State University

Metacognitive processes, such as monitoring and control, are integral to educational outcomes (Dunlosky & Metcalfe, 2009) because they are foundational for effective self-regulated learning (e.g., judging one’s progress, requesting help, slowing down on difficult problems). Little is known about children’s metacognitive abilities in mathematics, such as how accurately they judge their estimation precision. Estimation precision correlates with mathematics achievement, arithmetic performance, and memory accuracy (Booth & Siegler, 2008; Siegler, Thompson, & Schneider, 2011; Thompson & Siegler, 2010; Thompson & Opfer, 2016), which are critical aspects of math proficiency. Across three studies, Wall, Thompson, Dunlosky, and Merriman (2016) showed that 1st-4th graders accurately monitored and controlled their estimates. Children estimated numbers in a small (0-10, 100, or 1,000) and large numerical range (0-100, 1,000, or 100,000). Then, they made a confidence judgment (CJ) after each trial (i.e., not so sure, kind of sure, totally sure; Hembacher & Ghetti, 2014) and decided whether to submit their estimates for scoring (i.e., a control judgment). Children were more confident in their small- than large-scale judgments, F(2,51)=8.94, p<.01. Further, the majority of 1st (77%), 2nd (76%), and 4th (88%) graders submitted their small- and not large-scale packet of estimates for scoring. Control judgments for individual estimates were more strongly related to the child’s confidence than to actual estimation precision, F(1,40)=89.84, p<.01. As a follow up to Wall et al. (2016), we trained 1st and 2nd graders in a pretest-number line intervention-posttest design and evaluated whether increasing estimation precision also increased children’s CJs (Morehead, Thompson, Dunlosky, & Buerke, in preparation). Corrective feedback in the 0-1,000 range (i.e., You said 150 goes here. Actually, it goes here. You showed me where N goes.) results in rapid increases in estimation precision (Opfer & Siegler, 2007; Opfer & Thompson, 2008; Siegler et al., 2009; Thompson & Opfer, 2008, 2010, 2016) as compared to additional practice without corrective feedback (i.e., control group). It was unknown whether correct and incorrect worked examples (McGinn, Lange, & Booth, 2015) might also improve estimation precision. Children’s estimation precision improved from pretest-to-posttest for the feedback group (d=.56) relative to the control group, replicating previous work. Further, correct worked examples (d=.53; Another child said 150 goes here. She was right!), but not incorrect worked examples (i.e., Another child said 150 goes here. She was wrong!), were just as effective at improving estimation precision as was feedback. Most important, children’s metacognition was dissociated from their performance; across conditions, CJs were not significantly different from pretest (M=2.29) to posttest (M=2.31). We speculate that as children’s estimates become more precise, they realize how difficult it is to be exact. Future research can investigate whether a higher dosage of distributed feedback can increase confidence. In summary, elementary-aged children showed above-chance monitoring and control, however there is still room for metacognitive improvements across development. Interestingly, children were more confident for easier/small than more difficult/large problems. Future research will investigate the cues children monitor during estimation (e.g., fluency/difficulty/familiari

Diana Selmezcy, University of California - Davis - Should I ask for help? How children weigh their confidence and available evidence

Diana Selmeczy¹, Alireza Kazemi¹, Simona Ghetti¹

¹University of California, Davis

Children are regularly faced with situations where they can seek helpful information from others (e.g., teachers, siblings, parents) to support their memory decisions. Although asking for help may be beneficial, available helpers may not always be perfectly accurate. If one already knows the answer, asking for help may even be detrimental. Therefore, appropriate help-seeking requires evaluating the quality of one’s own memory (i.e., metamemory) and comparing it against a potentially helpful suggestion. Previous research demonstrates that metamemory develops throughout middle childhood (Fandakova et al. 2017), but little is known about how children actively seek help, and whether the accuracy of the information received influences how children use it. Five- (N=26), 7- (N=27), and 9-year-olds (N=24) completed a yes/no recognition memory test during which they had the option to ask for help that was reliable most of the time. The help provided was in the form of a cue indicating “Likely Yes” or “Likely No” that was valid (correct) 75% of the time and invalid (incorrect) 25% of the time. Thus, children decided whether or not to seek help and whether or not to follow the suggestion. After their recognition decision, children provided confidence judgments. Additionally, during some trials no help was available in order to assess baseline memory. Across ages, we found a negative correlation between proportion of help-seeking and baseline confidence (r=-.30, p=.008) demonstrating that children who were the least confident in their answers when help was not available were the most likely to ask for help. This suggests that subjective uncertainty may contribute to individual differences in help-seeking. For trials in which help was asked, we found a significant age by cue validity interaction on recognition memory, F(2,57)=5.09, p=.009, η2=.15. Following valid cues, percent correct was similarly high across age groups (5-year-olds M=.89, SD=.10; 7-year-olds M=.84, SD=.10; 9-year-olds M=.86, SD=.21). However, following invalid cues, percent correct was significantly lower in 5-year-olds (M=.28, SD=.29) compared to 7- (M=.60, SD=.24, p<.001) and 9-year-olds (M=.54, SD=.32, p=.01). Thus, younger children do not seem to readily identify invalid cues when they occur in a generally reliable context. We found a significant main effect of cue validity on confidence ratings, F(1,57)=4.60, p=.04, η2=.08, such that confidence was higher following valid (M=1.20, SD=.52) compared to invalid (M=1.11, SD=.61) cues. Additionally, there was a significant main effect of age, F(2,57)=3.84, p=.03, η2=.12, such that confidence was significantly lower in 9- (M=0.90, SD= 0.56) compared to 7- (M=1.25, SD=0.42, p=.02) and 5-year-olds (M=1.33, SD=0.59, p=.03). These results suggest that children’s metamemory was sensitive to cue validity and they appropriately decreased their confidence following invalid cues. However, 9-year-olds had significantly lowered confidence overall suggesting they recognized that their confidence should be generally low in situations where they needed to seek help. Overall, older children are more likely to appropriately consider their own memory against a reliable, but not perfectly accurate, suggestion compared to younger children. Additionally, only older children lower their confidence after asking for help, suggesting that they recognize that needing help signals risk for error.

S12         Is that so? How children evaluate claims and conjectures
Chair: Junyi Chu, MIT
Tomer Ullman, MIT
Children preferentially learn from knowledgeable, reliable sources and accept verified claims conveyed with confidence over claims that are unverified or tentative. But considering unverified conjectures can be critical: Claims whose truth value is unknown or unknowable can provide insight into subjective states about which there is no fact of the matter, and can underlie genuine innovation. This symposium, by three emerging leaders in the field, looks at the developmental, social, and cognitive factors that modulate the evaluation of claims and conjectures, focusing on whether children consider not just evidence and their prior beliefs, but also data-independent criteria including the relationship between the form of the conjecture and the question at hand. Paper 1 asks whether individual differences in 6 and 7-year-olds’ epistemological understanding (reasoning about the nature of knowledge) are linked to cognitive skills, parent factors, or context (whether discussions are about matters of fact, interpretation, or personal taste). Furthermore, an “epistemological intervention” is employed to test for effects of adult scaffolding on children’s epistemological development. Paper 2 looks at four to six-year-olds’ ability to distinguish diverse kinds of explanation (mechanistic, teleological, etc.), and its relationship to individual differences in children’s domain-specific knowledge, verbal IQ, and EF. Paper 3 looks at contexts in which adults and children (ages 4-7) might override their preference for accurate and credible information in favor of speculative conjectures, finding that participants prefer conjectures that answer questions to true but irrelevant facts, even when the conjectures contain unverified information, are preceded by uncertainty markers, and contradict prior beliefs. A rising star in computational modeling of intuitive theories will discuss the studies’ relationship to models of idea generation and on-the-fly common sense reasoning.

Sarah Suárez, Boston University - The development of epistemological understanding: Exploring individual differences and potential mechanisms of change

Sarah Suarez¹, Melissa Koenig²

¹Boston University, ²University of Minnesota

A learner’s epistemological understanding–or reasoning about the nature of knowledge–has important implications for critical thinking (Greene, Sandoval, & Bråten, 2016). Children’s intuitive epistemology is often described as strictly absolutist, i.e. they regard knowledge as simple, certain, and objectively true (Kuhn et al., 2000). However, recent work calls this view into question. Parents’ epistemological understanding predicts children’s emphasis on evidence in scientific conversations (Luce et al., 2013) and their evaluations of reasoners’ competence (Suárez & Koenig 2015; 2016). Thus: 1) How varied and sophisticated is children’s epistemology (Studies 1 & 2), 2) What cognitive and parental factors are associated with variations in children’s epistemology (Study 1) and 3) Might children’s epistemological understanding develop within the context of conversation with an adult (Study 1)?

Study 1: Fifty-three 6-7-year-olds participated in a within-subjects study that characterized individual differences in epistemological beliefs at baseline and tested if an adult’s “epistemological scaffolding” intervention would promote more sophisticated beliefs at posttest. In each phase children saw 3 vignettes in which 2 characters disagreed about an issue with no clear answer and answered questions about the cause of the disagreement, how it might be resolved, and if there could be truth to the claims. The issues were one resolvable question of fact (“Objective”; e.g. are there dark clouds outside), one issue of interpretation (“Mixed”; e.g. does an exotic animal make for a good pet), and one potentially unresolvable matter of taste (“Subjective”; e.g. is a painting pretty). Children’s responses to each issue were coded as Absolutist (truth is objective & claims are either right or wrong), Multiplist (claims are subjective opinions which cannot be right or wrong), or Evaluativist (knowledge is a subjective mind’s interpretation of objective facts). During the intervention an adult scaffolded children’s ability to reason in an Evaluativist manner by offering alternative ways of thinking and prompting children to reflect on their reasoning.

Children adjusted their responses based on the issue (Fig 1): Objective issues elicited more Absolutist responses, Subjective issues elicited more Multiplist responses, and Mixed issues elicited more Evaluativist responses. Notably, ~40% of children held Evaluativist perspectives for subjective issues, despite findings that even adults struggle to acknowledge how matters of taste can be evaluated on objective criteria (e.g. judging music based on pitch; Kuhn, et al. 2000). Furthermore, even after controlling for children’s verbal IQ, children’s tendency to make Evaluativist judgments was negatively related to parents’ Absolutist judgments and SES. The intervention significantly reduced Absolutist judgments and increased Evaluativist judgments in real time, but these effects did not persist at posttest (Fig 2). However, individuals who were less Evaluativist at baseline were more likely to show increases in their Evaluativism.

Study 2: Pilot data on 4-5-year-olds suggest that preschoolers tend to be more Absolutist than 6+ -year-olds, but are less Absolutist regarding matters of personal taste.

Thus, children are not strict objectivists. Rather, there is considerable inter- and intra-individual variation in their philosophies of knowledge reflecting age, cognitive skills, parent beliefs, and context.

Kaitlin Sands, University of Texas, Dallas - How do fish breathe underwater?: Young children's ability to discriminate between different quality explanations regarding biological phenomena

Kaitlin Sands¹, Candice Mills¹

¹The University of Texas at Dallas

By the preschool years, children know some facts about biology (e.g., cheetahs run really fast), but less is known regarding what children understand about the mechanisms that support those facts (e.g., how it is that cheetahs can run really fast). Three experiments examined the kinds of explanations children accept in response to how questions about animals. In Experiment 1, 4-, 5-, and 6-year-old children (N = 59) were asked how questions about biological processes in animals and were asked to rate four responses to those questions: mechanistic explanations, teleological explanations, circular explanations, and non-explanations. Results suggested that only 6-year-olds, but not 4- and 5-year-olds, distinguished between the explanations, with mechanistic explanations rated as better at answering questions than the other explanation types (ps < .007). Further, 6-year-olds recognized that non-explanations were worse at answering questions than the other explanation types (ps < .01). Children’s verbal intelligence and biological knowledge (Mechanistic Explanations: p = .17, Non-Explanations: p = .31) did not predict their ratings of different explanation types over and above the effect of age (Mechanistic Explanations: p = .001, Non-Explanations: p < .001). To better understand the reasons for the poor performance of 4- and 5-year-olds in the first study, Experiment 2 focused on examining whether young children would better evaluate explanations in a different task and/or in a different domain. Four- and 5-year-olds (N = 45) were asked to sort mechanistic explanations and non-explanations into boxes that were labeled “Helpful” and “Not Helpful” in response to causal how questions about biological processes about real-world situations. Results demonstrated that 4- and 5-year-old children were accurate in sorting biological mechanistic explanations (p < .001), situational mechanistic explanations (p = .05), and situational non-explanations (p = .005). Children had difficulty, however, accurately sorting non-explanations in the biological domain (p = .25). Results suggested that children who were higher in verbal intelligence and biological knowledge were more accurate at sorting non-explanations in the biological domain (ps < .05). There were domain differences in children’s sorting ability in that children were more accurate in sorting mechanistic explanations in the biological domain and more accurate in sorting non-explanations in the situational domain. A third ongoing experiment (current N = 32) follows up on these findings, adding in a measure of executive function skills to better understand the role of individual differences in how children evaluate explanations. In addition, children are engaged in a free recall memory test to examine whether they remember more detail from the mechanistic explanations or the non- explanations they are given. Results from this experiment will provide a clearer picture of how individual differences play a role in children’s explanation evaluation abilities in the biological domain and will provide information regarding more long-term impacts of children’s explanation evaluation abilities. Overall, this work helps inform us about how preschoolers make sense of and reason about scientific information. Implications for formal and informal science learning will be discussed.

Junyi Chu, MIT - Children flexibly evaluate facts and conjectures

Junyi Chu¹, Laura Schulz¹


What epistemic practices support the discovery of new ideas? Although young children use many criteria to evaluate claims (e.g. prior probability, evidence, simplicity, informants’ credibility, etc.), the preeminent criteria of any hypothesis is that it provides an answer — at least in principle — to the question at hand. Here we look at whether children prefer unverified conjectures even to known, established facts when the former answers an otherwise unresolved question and the latter does not. In Experiment 1, four to seven-year-olds (N=98) were presented with questions that either could or could not be answered by appealing to facts provided in a story; children of all ages appropriately preferred facts for questions with known answers and unverified, but appropriate, conjectures for questions with unknown answers. Because the forced choice design could not tell us whether children preferred conjectures for questions with unknown answers or succeeded simply by rejecting irrelevant facts, in Experiment 2 four to five-year-olds (N=96) were asked to rate each response individually. We also included a condition where conjectures were prefaced with “I don’t know, but maybe …”, thus emphasizing their speculative nature. Replicating Experiment 1, children rated facts higher for questions with known answers and rated conjectures higher for questions with unknown answers. In contrast to previous research in selective trust, adding uncertainty markers did not change these ratings. Thus, by age four, children reliably evaluate conjectures based on how well they answer the question. In Experiments 1 and 2, the conjectures, although unverified, did not contradict any information provided in the story. In real life however, relaxing a commitment to our prior beliefs might help us answer otherwise unresolved questions. Indeed, it is precisely such contexts in which unverified conjectures may be most able to help us shift perspectives and entertain genuinely new ideas. In Experiment 3, we tested adults’ (N=35) and four to five-year-olds (N=21) willingness to accept conjectures that were explicitly in tension with information in the story. Participants rated four conjectures: a claim that was probable and satisfying given the story, a satisfying (but improbable) claim, and two unsatisfying responses varying in probability. Adults endorsed the satisfying but improbable claim as often as the probable one (while correctly rejecting the inappropriate answers); by contrast, preschoolers endorsed the probable claim that answered the question but rejected the improbable claim that answered the question along with the inappropriate answers. An ongoing follow-up (N=25) finds similar results among preschoolers even when the most likely conjecture is not included — and finds that older children (6-7-year-olds) accept the unlikely but satisfying answer at similar rates to adults, suggesting a developmental change in the ability to prioritize question-answer fit over prior beliefs when evaluating conjectures. Collectively, the results suggest that while children recognize the value of reliable information, they may put an even higher value on answering questions. Such “cognitive pragmatism” may be critical to innovation and discovery.

Oral Papers III

O3.1       Semantic transferability rather than perceptual sparseness may underlie the advantage of simple objects in young children’s relational transfer

Presenter: Youjeong Park

Youjeong Park¹, Jinwook Kim²

¹Seoul National University, ²Myongji University

Preschool children are able to generalize relational structures (e.g., spatial relation, and mathematical relations) from a single instance to another instance, but their ability is significantly influenced by the type of objects used to depict the relations. Prior work has shown that children better transfer spatial relations (ON, IN, and above) when shown a sample consisting of abstract geometric shapes than realistic objects (Park & Casasola, 2017). In the present study, we aimed to better understand the advantage of abstract geometric shapes by exploring exactly where the advantage comes from. Specifically, we focused on two characteristics of abstract geometric shapes. First, they are perceptually sparse in that they have less surface features (colors, patterns, and parts). Second, they are semantically transferable in that they are more flexible in terms of their referents. For example, a realistic table can refer to tables only; it can’t refer to a door. In contrast, a drawing of a rectangle can refer to many things such as a door, a table, and a piece of paper. We asked which characteristic of these is responsible for the advantageous effect of geometric shapes. Four- to six-year-old children (N = 160) in Korea were tested with a modified spatial analogies task, in which children were asked to transfer the spatial relations of on, in, and above from a single sample to one of three options. Children were randomly assigned to one of the three conditions: (1) Plain Geometric Shapes, (2) Decorated Geometric Shapes, and (3) Realistic Objects conditions. In the Plain Geometric Shapes condition, the objects in the samples were black-and-white line drawings of geometric shapes, which were not only perceptually sparse but also semantically highly transferable. In the Decorated Geometric Shapes condition, the objects in the sample were geometric shapes that were decorated with multiple colors and patterns, thus perceptually rich but semantically highly transferable. Lastly, the Realistic Objects condition presented the sample consisting of objects that were perceptually rich and semantically less transferable. If the advantage of abstract geometric shapes stems from their perceptual simplicity, then perceptually rich geometric shapes would not be advantageous, failing to promote young children’s transfer of spatial relations. If the advantage stems from semantic transferability of geometric shapes, then even perceptually rich geometric shapes would be advantageous as the perceptually sparse shapes. We found lower level of transfer performance in the Realistic Objects condition than in the Decorated shapes condition. In addition, there was no difference in children’s transfer performance between the Decorated and Plain Geometric Shapes conditions. Thus, the results suggest that semantic transferability but not perceptual sparseness of abstract geometric shapes may be mainly responsible for the advantage of abstract geometric shapes in promoting children’s relational transfer.

O3.2       Gestures facilitate word learning in shared storybook reading

Presenter: Yayun Zhang

Yayun Zhang¹, Chen Yu¹

¹Indiana University

Shared storybook reading is one of the most common everyday activities children experience in western societies and it is an important context in which children learn words. From the young word learners’ perspective, to learn the correct word-referent mappings from books, they need to link what they see with what they hear. However, given multiple objects on a book page, it is not clear how young learners direct their attention to the right referent when hearing a to-be-learned word named by parents. The aim of the current study is to examine how correct word-object mappings may be established through gestures generated by both social partners during book reading. We hypothesize that young learners most often do not attend to the objects named by parents without additional social cues. However, their abilities to use their own gestures to direct the other social partner’s attention and to follow the other’s gestures to redirect their own attention could be critical for early word learning. Method. Sixteen parent-child dyads participated (Mage = 19.03 m.o.). As shown in Figure 1, parents and toddlers sat next to each other at a table. They were provided with 5 storybooks and parents were instructed to read naturally to their children for 15 minutes. We used a dual eye-tracking paradigm to record moment-by-moment parents’ and children’s eye gaze. In total, we collected 45 book reading sessions, providing 2.5 hours of video footage for coding and analyses. 1. Speech and Gesture: On average parents produced 17 utterances per min and each utterance was about 2-second long. Parents on average gestured 9.52 times/minute and children gestured 3.43 times/minute. 2. Coupling of naming and gesture: We then examined the coupling of naming and gesture to see how they are temporally aligned. We found that gestures (either from parents or toddlers) were highly coupled with parents’ naming. About 74% of the naming moments were associated with gestures. Parents gestured (56%) more with naming compare to children (18%). There were also more matched gestures (point at the named object, 56%) than mismatched gestures (18%). 3. Effect of gesture on attention during naming: By investigating the effect of gesture on attention, we found that there were different types of learning situations in book reading contexts. Without gestures, 60% of the time, the child never looked at the target object. But in cases with parent matched gestures, the referential ambiguity was significantly reduced, meaning that there were much fewer instances (45%) that the child never looked at the target. We also found an increase of target look compare to no gesture and mismatched gesture cases. We observed a very similar pattern in cases with child gestures. Gestures created by toddlers themselves can also reduce referential ambiguous by driving their own attention to the correct target. Our findings show that given that there are many objects on a book page, children do not necessarily look at the objects named by the parents. Although individual naming instance can be highly ambiguous or even misleading, gesture cues from both parents and children can significantly reduce the degree of referential uncertainty and provide more informative data for young learners. In other words, gestures from both parents and children are critical for building correct word-object mappings in shared storybook reading context.

O3.3       Disrupting development: The influence of maternal depression on parent-child interaction and child expressive language

Presenter: Brianna McMillan

Lillian Masek¹, Brianna McMillan¹, Verna Rasing¹, Sarah Paterson¹, Roberta Golinkoff, Kathy Hirsh-Pasek¹

¹Temple University

Parent-child interactions play a major role in the language learning environment during infancy and early childhood. An important aspect of high-quality interaction is fluency and connectedness, or the balance and flow of conversation between the caregiver and child (Hirsh-Pasek et al., 2015). Maternal depression can negatively impact both child language development and caregiver behaviors (Stein et al., 2008), however little research has examined how maternal depression affects fluency and connectedness, and hence, child language outcomes, or whether these associations differ across socioeconomic strata. Here, we examine the relations between maternal depression, fluency and connectedness, and expressive language in a low- and high-income sample, hypothesizing that maternal depression will relate to the fluency and connectedness of the interaction and children’s subsequent language scores. Furthermore, we predict that socioeconomic status will moderate these relations. 120 mother-infant dyads were selected from a community sample (NICHD Study of Early Child Care and Youth Development). Families were categorized into income groups based on their income-to-needs ratio at 24 months. 60 participants were selected from each income group for this study. One participant was excluded due to missing data (n=119). Maternal depressive symptoms were assessed at 6-months using the Center for Epidemiologic Studies Depression (CES-D) scale, parent-child interaction at 24-months was assessed using fluency and connectedness (FCC; Adamson et al., 2016), and expressive language was assessed at 36-months using the Reynell Developmental Language Scales (Reynell 1991). The FCC of the interaction was rated on a 1 to 7 scale, with high levels of FCC indicating a more balanced and dynamic conversation, whereas lower levels of FCC indicate the conversation was dominated by one partner, or the dyad were not engaged with one another. Across income groups, higher levels of maternal depressive symptoms at 6-months were related to poorer child expressive language at 36-months (β=-0.312, SE=0.164, CI, -0.0730, -0.079). For low-income families, the relation between maternal depressive symptoms and child’s expressive language was accounted for by FCC at 24-months (β=-0.208, SE=0.092, CI=-0.429, -0.055), but for high-income families, it was not (β=0.053, SE=0.120, CI=-0.115, 0.371). These results suggest that children of mothers experiencing depressive symptoms, especially in low-income families, are at-risk for developing poorer language skills due in part to poorer-quality interactions. Parent-child interactions lay a foundation for language development, but this foundation is disrupted when mothers experience depressive symptoms, especially in low-income families. Higher-income families may have access to protective factors unavailable to their lower-income counterparts. Acute bouts of depressive symptoms may relate to parent-child interactions differently than chronic symptoms and clinical levels of depression may relate differently than subclinical levels. Therefore, in future analyses we will examine the trajectory of depressive symptomatology as well as clinical thresholds to further elucidate the relation between maternal depression, parent-child interactions, and child language outcomes. Altogether, these results highlight how environmental risk factors relate within diverse environments to contribute to children’s cognitive development.

O3.4       What is a good question-asker better at? From unsystematic generalization, to overgeneralization, to adult-like selectivity across childhood

Presenter: Costanza De Simone

Costanza De Simone¹, Azzurra Ruggeri²

¹Max Planck Institute for Human Development, ²Max Planck Institute for Human Development and School of Education, Technical University Munich

Prior research showed that young children are selective social learners and take into account others’ knowledge and expertise to decide from whom to learn. Recent research suggests that young children are also sensitive to the process by which others have acquired knowledge and prefer to seek help from actors who have demonstrated active learning competence. What inferences do children make based on the ability to search effectively, for example by asking informative questions? Are good question-askers smarter, more knowledgeable, or better at solving problems? Do adults make similar inferences and generalizations? In this project we explored across three experiments to what extent adults and children generalize the ability to ask informative questions to more or less-related abilities or characteristics. In Study 1 (adults, N = 30) and 2 (3- to 9-year-olds, N = 120), we presented participants with a storybook depicting two monsters: one monster always asked informative questions and the other always asked redundant, uninformative questions. Participants were then asked to choose which monster they thought was most likely to possess/was better at 12 different characteristics/abilities. Our results show a clear developmental trend: Three- and 4-year-olds drew unsystematic inferences from the monsters’ question-asking expertise, showing no preference for the good question asker when evaluating abilities, traits and characteristics that both adults and older children deemed strongly related to question asking (e.g., being good at treasure hunting). At the same time, they showed a strong preference for the good question asker on some clearly irrelevant questions (e.g., having siblings). Five- and 6-year-olds identified the good question asker as more likely to have nearly every presented ability/characteristic. Seven- to 9-year-olds showed adult-like response patterns, selectively associating question-asking competency to some, relevant abilities and characteristics (e.g., being good at school, being clever), but not to others (e.g., kicking a ball further, liking ice cream). In Study 3 we disentangled the inferences that 5- to 9-year-old children (N = 59) and adults (N = 20) draw based on an agent’s strategic competence versus epistemic knowledge. We presented participants with two agents: Yellow was an expert in a specific domain (i.e., fish), but was not good at asking questions; Blue was good at finding out things by asking questions, but did not know anything about fish. During the test phase, participants had to ask one of the two agents for help to answer three questions: one targeting the domain of expertise of Yellow (fish), one targeting a similar domain (animals) and one targeting an unrelated domain (houses). Preliminary results show again a clear developmental trend: Older children (7- to 9-year-olds) and adults were selective when asking for help, showing a preference for the good question asker on the similar and unrelated domain, but selected the fish expert when the question was about fish. Younger children (5- to 6-year-olds) were equally likely to ask the two agents for help on the similar and unrelated domains, although they already preferred the fish expert when the question was about fish. This project is a first step in understanding whether and how children use their sensitivity to others’ active learning competence to navigate the social world, identifying good role models to learn and to learn how to learn from.

O3.5       Automaticity of reading continues to develop into adulthood

Presenter: Joshua Hartshorne

Joshua Hartshorne¹

¹Boston College

Research on cognitive development has largely focused on childhood – especially early childhood – for good reason: that is when the bulk of learning and development happens. However, it has long been known that at least some cognitive abilities continue to be refined well into adulthood. For instance, people’s vocabularies show substantial continued growth well into sixth decade of life (Hartshorne & Germine, 2015). Grammar – often thought of as being largely acquired by kindergarten – also shows continued improvement until about 30 years of age (Hartshorne, Tenenbaum, & Pinker, 2018). Not only do people continue to acquire knowledge, the ability to learn may itself change in adulthood. For instance, the ability to learn new faces peaks at around 30 years old (Germine, Duchaine, & Nakayama, 2011). Just how much this long unfolding of learning and development should affect cognitive development theory is unclear. Until recently, it was widely believed that lifespan development was well-characterized by a clear dichotomy between fluid intelligence, which involves flexible thinking and peaks by late adolescence, and crystalized intelligence, which involves marshalling accumulated knowledge and peaks in late adulthood (Salthouse, 2003). However, researchers are increasingly finding cognitive abilities that do not fit either of these developmental trajectories (Hartshorne & Germine, 2015), and indeed even the fluid/crystalized dichotomy itself has come under question (Hampshire, Highfield, Parkin, & Owen, 2012). We present additional evidence from the automaticity of reading. 10,647 native English speakers completed a modified Stroop task and a flanker task. The subjects were volunteers who were recruited online using a well-established platform that has been extensively validated [citation suppressed for anonymity]. In the modified Stroop task, subjects saw the words ORANGE or WHITE on a computer screen with orange or white lettering. They were instructed to press ‘o’ or ‘w’ on the keyboard to indicate whether the font color was orange or white, respectively. The task consisted of 14 congruent trials (the font color matched the word) and 14 incongruent trials (the font color and word mismatched). In the analogous flanker task, subjects identified the direction of the central arrow in displays like >><>> and >>>>>. Analysis focused on ages for which there were at least 10 subjects, resulting in the exclusion of 109 subjects younger than 10 or older than 77. Interference was measured as the log-transformed reaction time for congruent trials minus incongruent trials: negative numbers indicate greater interference. Stroop interference increased across the lifespan, with a slight plateau between 30 and 50 (Figure). The increased interference from 10-40 cannot be attributed to attentional control per se: flanker interference decreased during childhood and did not begin to increase again until 40 (Figure). This qualitative description was confirmed using optimal break-point estimation (see Hartshorne et al., 2018). Thus, increased Stroop interference through 40 is likely due to learning. The Stroop results contrast with some prior, smaller studies (Comalli, Wapner, & Werner, 1962; Schiller, 1966). We discuss possible explanations due to sample size, statistical analysis, and method. We also discuss these findings in the context of theories of learning and development.

16:00 – 16:15: Transition time

16:15 – 17:45: Parallel Session 16 – 20

S13         Contributions of naps to sleep-dependent memory consolidation in infancy and early childhood
Chair: Rebecca Gomez, The University of Arizona
A growing body of work in early childhood demonstrates a crucial role for naps in memory consolidation, an off-line process that stabilizes memories, leaving them less vulnerable to interference. Importantly, exactly how sleep supports memory is thought to change as children develop and as their brains support more mature memory processes. This symposium brings together researchers who examine changes in sleep-dependent memory formation across infancy and early childhood. The to-be-presented studies demonstrate the use of nap designs and the cutting-edge tools of cognitive neuroscience and sleep physiology to investigate the development of memory, a fundamental cognitive process. The first talk in the symposium reports developmental change in nap-dependent consolidation across 15 and 18 months that demonstrates an increase in the precision of a nap-dependent memory consolidation. The second talk reports developmental changes in nap-dependent source memory for distinct and similar targets across 3.5 to 5.5 years finding that naps play a crucial role in stabilizing source memories for similar targets. The third talk investigates the role of electrophysiological markers of sleep-dependent memory consolidation in 3-6-year-olds, demonstrating markers found in adult sleep-dependent memory consolidation. The fourth talk finds associations in hippocampal volume and spindle activity, a marker of sleep-dependent memory consolidation across ages 3-5 years, supporting the role of the hippocampus in preschoolers in sleep-dependent memory consolidation. Together the talks in this symposium chart the development of sleep-dependent memory formation and associated neural mechanisms across infancy and early childhood.

Lucia Sweeney, The University of Arizona - Changes in sleep-dependent consolidation in infancy with development

Lucia Sweeney¹, Rebecca Gomez¹

¹The University of Arizona

Previous research on sleep-dependent memory formation in infancy suggests that naps protect memories from greater loss than occurs during wakefulness but do not stabilize specific details of the memory. Here we investigate whether naps contribute to greater stabilization of a memory in older, more mature infants. In prior work, 15-month-olds heard phrases from an artificial grammar containing two nonadjacent dependencies (NADs), e.g. “vot-kicey-jic, pel-wadim-rud, pel-puser-rud,” where “vot-X-jic” forms one NAD, and “pel-X-rud” forms the second (Gomez et al, 2006). In the 4 hours after training, infants either napped or remained awake then were tested using the HTP Procedure. On half of the test trials, infants heard familiar phrases from training (e.g. vot-X-jic, pel-X-rud). On the other half of the test trials, infants heard unfamiliar phrases from another grammar, violating the NADs from training (e.g. vot-X-rud, pel-X-jic). Infants in the No-nap groups discriminated familiar from unfamiliar strings after 4 hours (Gomez et al., 2006) but not 24-hours later (Hupbach et al., 2009). The nap groups showed generalization, looking longer to trials consistent with their first test trial versus trials from the other grammar 4 and 24-hours later, suggesting loss of the specific vocabulary making up the NADs from familiarization but not that there was a relationship between the 1st and 3rd words in phrases. To assess how retention of NADs develops, we tested older 18-month-olds using the same procedure. Now infants who nap discriminated familiar phrases (M = 7.75 sec) from unfamiliar ones (M = 6.58, p < .05) indicating that at 18 months, sleep stabilizes memory and prevents loss of detail. We also investigated whether infants could extend their knowledge of the learned NADs to completely novel vocabulary. Infants first listened to a NAD language with phrases such as guf-X-zam/miv-X-fop and napped in the four hours between learning and test. On half of the trials, infants heard phrases containing vot-X-jic/pel-X-rud NADs. The other half of test trials contained phrases from another grammar with opposite NADs (vot-X-rud/pel-X-jic). If infants could extend their knowledge of NADs to the grammar they hear in the first test trial, then they would learn the dependency in that trial and use it to guide their looking for the remaining trials. Indeed, 18-month-olds looked longer to trials consistent with the first test trial (M = 6.99s) versus trials inconsistent with that grammar (M = 5.85s p < .05). Thus, by 18 months, sleep stabilizes memory such that infants remember the specific NADs heard during training. Moreover, infants can recognize the same grammatical rule presented in completely novel vocabulary. Comparing these findings with our published data on 15-month-olds, we see a shift in sleep-dependent memory formation. At 15-months, sleep promotes generalization potentially due to loss of detail. At 18-months, sleep promotes a stable and specific representation, flexible enough to apply to new vocabulary.

Ji-Soo Kim, University of Arizona - The role of naps in source memory for similar and distinct objects in preschool age children

Ji-Soo Kim¹, Rebecca Gomez¹

¹University of Arizona

Naps play an important role in retention with much work investigating memory for distinct items. However, the difficulty children encounter in keeping source memories separate varies for similar and distinct objects with sources for distinct objects easier to remember. We investigated the extent to which naps preserved source memory as a function of age for similar and dissimilar targets. Children ages 3.5, 4.5, and 5.5 years learned associations between specific puppets (Mr. Raccoon and Mr. Bunny) and highly similar or highly distinct objects to a criterion of 75% accuracy. They then participated in a memory test for the associations after 15 minutes or after a 24-hour delay. Some children napped in the 4-hour interval after training, before experiencing nighttime sleep. Children exhibited equally high source memory for distinct objects when tested 15 minutes after training. Age group means ranged from 84.85% to 100%. In contrast, source memory for similar objects ranged from 61.43% to 79.03, with 3.5-year-olds performing at chance, t(34) = 1.49, p = .15. Children exceeded chance responding by 4.5 years, t(32) = 4.16, p = .00. Having established baseline 15-minute delay retention at 3.5 years, we next asked how naps at this age supported long-term retention. Children napped or not in the 4-hour interval after training and were tested 24 hours later. Children who napped showed excellent source memory for similar (90%) and distinct objects (M=95%). Children who did not nap showed much greater loss of source memory for similar (59%) compared to distinct objects (81%). Thus, the nap protected source memory for similar and distinct targets over a 24-hour delay. Finally, we investigated source memory after a 24-hour delay in 4.5- and 5.5-year-olds who no longer napped on a regular basis. Although children showed robust source memory for distinct objects at ages 4.5 (M=73%) and 5.5 (M=89%), source memory for similar objects did not exceed chance at 4.5 (M=60%), t(25)=1.04, p=.31, and 5.5 years (M=67%), t(17)=1.46, p=.16. In sum, naps play an important role in stabilizing preschooler’s source memories. Although children who do not nap after learning retain fairly robust source memory for distinct objects regardless of age, naps played a crucial role in source memories for similar objects throughout the preschool-age period.

Sanna Lokhandwala, University of Massachusetts - Physiological mechanisms supporting the benefit of naps on preschool learning

Sanna Lokhandwala¹, Rebecca Spencer¹

¹University of Massachusetts

Sleep supports learning in preschool children. For instance, using a visuospatial task, we demonstrated that performance following a nap was greater than performance following an equivalent interval awake (Kurdziel et al., 2013). Moreover, this improvement was specifically associated with sleep spindles, bursts of EEG activity that may reflect underlying memory replay. While it is possible that spindles in naps may support other forms of declarative memories, visuospatial memories may be unique in that the spatial component makes the memories more receptive to hippocampal replay mechanisms. Thus, this spindle benefit may be specific to spatially-dependent declarative tasks. To begin to examine whether sleep spindles benefit other forms of declarative memories, we created a storybook learning task. At encoding, children were read four 10-page storybooks describing activities typical for a preschool child (e.g., baking cookies). Subsequently, during immediate recall, children were given picture cards depicting scenes from the story and asked to place the story in order without feedback. Polysomnography (PSG) was then applied to the child who then either napped (nap condition) or spent an equal amount of time awake engaged in quiet activities (wake condition). Following the nap/wake period, children participated in delayed recall, in which they were once again given the picture cards and asked to put the story back in the correct order. Children engaged in an additional recall session 24 hours after encoding. All children participated in both the sleep and wake conditions (order counterbalanced, separated by one week). In 23 children (36-71 months), performance following the nap was greater than when the same interval was spent awake (p = .004). This nap-dependent benefit persisted 24-hr later (p = .002). Importantly, performance did not correlate with sleep spindle density as previously observed. Rather, we found a significant positive association between time spent in slow wave sleep and recall (r = .472, p = .027). The role of slow wave sleep in declarative memory consolidation has previously been observed in studies of adults. These results further support sleep’s active role in declarative memory consolidation in children. Moreover, these results suggest that slow wave sleep and the underlying hippocampal system are sufficiently developed by this age. In a second study, we are considering whether this benefit extends to another storybook task, an emotional storybook task. Current results suggest that naps do not confer an equivalent, immediate benefit on memories. As we will discuss, underlying physiological processes may drive this difference. Considering that storybooks are used by both parents and educators as a method of learning in young children (Horst, 2015), uncovering mechanisms underlying early learning warrants attention. This understanding will be important for considering sleep-based interventions to better support learning.

Tracy Riggins, University of Maryland - Sleep-dependent memory consolidation and hippocampal development in preschoolers

Tracy Riggins¹, Benjamin Weinberg¹, Arcadia Ewell¹, Tamara Allard¹, Sanna Lokhandwala², Morgan Botdorf¹, Rebecca Spencer³

¹University of Maryland, ²University of Massachusetts, Amherst, ³University of Massachusetts

Sleep promotes memory consolidation, the off-line processing of memories that yields them stronger and less vulnerable to interference. Sleep-dependent memory consolidation is thought to reflect hippocampal-neocortical transfer of memories, a process supported by physiological events in non-REM sleep. Previous behavioral research has demonstrated sleep-dependent consolidation in young children (e.g., Kurdziel, Duclos, & Spencer, 2013). Specifically, in habitual nappers, learning was shown to be protected by an afternoon nap compared to a similar period awake. This “nap benefit” was associated with density of sleep spindles present in EEG recorded during the nap. However, how this nap-benefit and specific sleep architecture relate to development of the hippocampal-cortical network supporting memory has yet to be examined. This gap is problematic as memory, sleep, and the hippocampal-cortical network are all undergoing significant developmental change during early childhood. The objective of the current study is to examine the role of sleep and brain maturation on memory in early childhood, specifically when children transition out of naps. The central hypothesis of this work is that maturation of the hippocampus during this period results in more information being retained without interference, reducing the need for frequent consolidation, and ultimately the transition out of napping. Participants included typically developing 3- to 5-year-old children. Each child completed a visuospatial memory task in which they learned locations of objects in a grid and were assessed immediately to index initials levels of recall. Then children were either wake- or nap-promoted (within subjects, order counterbalanced between subjects) and subsequently tested via delayed recall. During nap-promotion, sleep physiology was assessed via polysomnography (a montage of EEG, EMG, and EOG). T1-weighted MRI scans (.9mm3) were obtained for each participant. These scans were used to extract hippocampal subregions volumes (head, body, tail) bilaterally and intracranial volume (ICV) using a rigorous combination of automated and manual processing tools. Preliminary analyses include data from 18 participants (9 male, Mage=45 months, SD = 5 months). Results suggested, after controlling for effects of age and sex, learning was shown to be protected by an afternoon nap compared to a similar period awake and that this “nap benefit” was associated with density of sleep spindles, r(10) = .47, p = .12. Although this finding currently fails to reach conventional levels of statistical significance, it is consistent with previous work (Kurdziel et al., 2013) and is from a small number of the total expected participants (expected n = 32). Results also showed relations between sleep spindle density and both left and right hippocampal body volume, r(7) = -.91, -.86, ps < .001 respectively. This novel finding supports the suggestion that maturation of the hippocampus results in more information being retained without interference, reducing the need for frequent consolidation, as indexed by sleep spindle density. Data collection is ongoing and will be complete by CDS 2019. This is the first investigation to show relations between sleep spindles and the hippocampus in early childhood. Future analyses will examine relations between memory, sleep, and hippocampal volume in a larger sample.

S14         The scope and roots of children’s surprise-based learning
Chair: Erin Anderson, Northwestern University
Observing a surprising event facilitates exploration even in infants (Stahl & Feigenson, & 2015), suggesting that surprise-based learning may be a lifelong learning mechanism. While this study focused on violations of core object knowledge, prior work has found that older children explore more when other events are inconsistent with their prior beliefs, such as about a causal system or support relations (Legare, et al., 2010; Bonawitz et al., 2012). These results raise the possibility that surprise-based exploration supports early learning in a range of domains beyond core object knowledge, even early in life. This symposium brings together recent findings that demonstrate infants’ abilities to detect violations of prior beliefs across a range of physical and social events and that guide their exploration accordingly. Talk 1 finds that infants react similarly to non-solid substances that violate their expectations as they do to objects, even though non-solids are less constrained than objects. Talk 2 demonstrates that an expectation that is violated does not have to be accurate to lead infants to increase their exploration: A proportion of contact rule that only predicts violations of support in some cases, nevertheless leads infants to explore longer in all cases. Talk 3 tests the hypothesis that infants and preschoolers can even harness others’ expressions of surprise to guide their explorations. Finally, Talk 4 examines the mechanism behind children’s increased exploration after a surprise, testing whether infants are actively searching for an explanation for the violation they witnessed. Together, this work suggests that young children form predictions outside core knowledge domains and that seeking more information when these predictions are violated might be a domain-general mechanism that drives children to acquire new knowledge.

Erin Anderson, Northwestern University - Expectations and learning from non-solid substances

Erin Anderson¹, Natasha Zeigler¹, Susan Hespos¹, Lance Rips¹

¹Northwestern University

Research has confirmed that infants have knowledge of how objects behave and interact through studies using looking paradigms (Spelke, et al., 1992) and action tasks (Hespos & Baillargeon, 2006, 2008). Stahl and Feigenson (2015) demonstrate something new: Infants not only have early object expectations, but this knowledge can motivate them to explore and to learn about the objects. In this study, we test whether this is also true in a new domain – non-solid substances. As adults, we have different reactions to, and expectations about, solids and non-solids. For example, you will have different results if you try to pick up a piece of yogurt than when you pick up a piece of toast. But how do we end up with these expectations? Initial studies using looking paradigms suggest that 5-month-old infants can anticipate how substances will behave and interact (Anderson et al. 2018; Hespos et al., 2009; 2016). Do infants also use violations of this substance knowledge to guide exploration? This study investigated how infants act on substances and what infants do when the behavior of a substance contradicts their current beliefs. We presented 12- to 14-month-olds four items: a solid object (solid consistent), a similar-looking non-solid item (solid inconsistent), a bowl of water (liquid consistent), and a bowl filled with gel beads that look like water (liquid inconsistent). We asked: (1) Do infants plan distinct actions for solids and non-solids? (2) Does violating these expectations lead to an opportunity for learning, with infants spending more time exploring an item that behaves inconsistently? We found that knowledge of both object and substance properties manifests in distinct infant actions. On their first approach, infants’ responses were based on an item’s visible object or substance properties. For example, when infants received a shape made of kinetic sand that looked solid but fell apart when touched, 53 of 64 infants used a solid appropriate action (e.g., pincer grip) on their first touch. When presented with a bowl of gelatinous “water beads” 51 of 60 infants approached it with a substance-appropriate first touch (e.g., splash or dip). These findings answer the first question: infants had distinct expectations for solid objects and substances. The answer to the second question is also yes: infants spent significantly more time exploring stimuli that violated their expectations, whether solid or non-solid, F (1, 57) = 24.00, p < .001, ηp² = .29. Infants spent more time interacting with the solid inconsistent item than with the solid consistent one, and they spent more time interacting with the liquid inconsistent item than with the liquid consistent one. These results provide a more complete understanding of how we manage the complexities of our physical world – non-solids included.

Yu Zhang, University of California, Santa Cruz - Violation to infant faulty knowledge induces object exploration by 7.5-month-olds in support events

Yu Zhang¹, Su-hua Wang¹

¹University of California, Santa Cruz

Infants are exceptional learners; they acquire rules about likely outcomes of physical events and revise these rules when events do not unfold as expected, allowing them to better predict outcomes in the future. At 7.5 months, infants acquire the proportion-of-contact rule for support events and expect an object to remain stable on top of a platform only when half or more of its bottom surface is supported on the platform. Infants responded with prolonged looking to plausible events in which a wide box remained balanced on a narrow platform when released (Wang, Zhang, & Baillargeon, 2016). The present experiments examined 7.5-month-old infants’ exploration of objects after they saw the objects violated their current rule in support events. Building on previous evidence for surprised-induced exploration (Stahl & Feigenson, 2015), we hypothesized that infants would preferentially explore an object that violated the rule even if it is a faulty rule. Specifically, we predicted infants to choose to explore first and spend more time exploring the “surprising” object. In addition, we also explore and examined the types of behaviors that infants exhibit during the exploration. In this experiment, 7.5-month-old infants saw two pairs of support events; each pair included an expected and an unexpected event (see Figure 1), with the order of presentation counterbalanced across infants. In Pair 1, the middle 33% bottom surface of the box was supported; the box remained stationary when released (unexpected event for 7.5-month-olds) or when still held by a hand (expected event). In Pair 2, the right 33% bottom surface of the box was supported; the box remained stationary when released (unexpected) or when held (expected). Each pair of the events was followed by an exploration trial in which infants were offered the two boxes used in the preceding events to explore; the boxes were placed wide apart, allowing infants to make a choice to explore. Paired sample t-tests showed that, on average, infants spent significantly longer time acting on the box of unexpected (M = 20.1 sec, SD = 12.7) than expected events (M = 14.2 sec, SD = 7.9; t(25) = 1.99, p =.05). We analyzed infant’s exploration behaviors, infant exhibited significantly more behaviors that are relevant to testing object support when acting on the box of unexpected events. They showed more sliding to drop (Unexpected: M=4.3, SD=3.6; Expected: M=3.0, SD=3.3), and holding (Unexpected: M= 2.4, SD=2.9; Expected: M= 0.7, SD=1.2) behaviors on the box of unexpected than expected event Sliding to drop: t(25)= 2.74, p<.05; holding: t(25)=2.82, p<.01 ). Together, the results in the present experiments support that infants’ expectations of physical events, even if they are faulty, drive their actions to gather information from the exemplars to fill their current knowledge gap.

Yang Wu, Stanford University - Others' surprise as vicarious prediction error: Young children use others' expressions of surprise to guide their own attention and exploration

Yang Wu¹, Hyowon Gweon¹

¹Stanford University

Prediction error – the discrepancy between expected and actual outcomes – plays a central role in learning. In humans, detection of prediction error often manifests as an expression of surprise, leading to a range of behavioral consequences. For instance, when infants observe a surprising event, they not only look longer than when they observe an unsurprising event, but also explore more and learn better (Stahl & Feigenson, 2015). However, without relevant prior knowledge to generate expectations, a naive learner would fail to recognize an observed event as unexpected or surprising. Building on prior work on early social learning and sensitivities to others’ emotional expressions, here we use looking time in infants (Study 1) and exploratory play in preschoolers (Study 2) to ask whether young learners can harness other people’s expressions of surprise as “vicarious prediction error”. Study 1 asks whether infants use others’ surprise to infer unobservable states of the world. First we conceptually replicated prior work (Xu & Garcia, 2008; n=28, 12.0-17.9 months) to ensure that infants look longer at an improbable sampling outcome (a red ball randomly drawn from a box that contains mostly white and just a few red balls) than a probable outcome (a white ball from the same box). In the main study (n=28), the experimenter drew a sample from the box, peeked at the outcome, and expressed either happiness (Happy condition) or surprise (Surprise condition) before revealing the outcome. As predicted, infants’ looking time at the probable and improbable outcomes was modulated by the experimenter’s emotional expression (condition x emotion interaction: F(1,21)=12.85, p=.002, See Fig 1A); infants in the Happy condition tended to look longer at the improbable than the probable outcome (t(3.09)=2.40, p=.094, two-tailed) but those in the Surprise condition showed a reversed trend (t(13)=-2.11, p=0.055). This effect was observed only on the first trial. While preliminary, these results suggest that observing others’ surprise can modulate infants’ expectations about event outcomes, and even reverse the pattern of looking time. Study 2 asks whether preschoolers consider others’ surprise and their epistemic states to guide their own exploration and discovery. Children (n=56; 3.0-4.9 years) saw an experimenter discover an interesting function of a novel toy. Then either the same experimenter or a naïve confederate expressed surprise while playing with the toy behind an occluder. Children explored the toy more broadly in search of a hidden function following the experimenter’s surprise than the confederate’s surprise, both in the first 30 seconds of play (t(54.0)=2.97, p=.004) and the overall play (t(51.6)=3.31, p=.002). See Fig 1B. Rather than indiscriminately treating surprise as an indicator of novelty, children selectively used others’ surprise to draw inferences about causal functions depending on their epistemic states. Taken together, this study reveals that even young children can make sophisticated use of others’ expressions of surprise to guide their own attention and exploration. The results demonstrate remarkably sophisticated abilities of infants and preschoolers to harness others’ emotional expressions as various prediction error. Our work synthesizes perspectives from literature on prediction error, social learning, and emotion understanding, towards a more comprehensive science of learning.

Jasmine Perez, Johns Hopkins University - Violations of expectation drive infants to search for explanations

Jasmin Perez¹, Lisa Feigenson¹

¹Johns Hopkins University

Decades of research find that infants look longer at impossible than possible events. More recently, research has revealed that infants preferentially explore objects involved in these surprising events (Stahl & Feigenson, 2015). However, it remains unknown why infants exhibit this heightened exploration. One possibility is that surprising events increase general arousal, thereby increasing infants’ post-surprise activity. Another possibility is that this exploratory behavior reflects infants’ search for explanations for surprising events. Here, in three experiments with identical dependent measures we tested the hypothesis that infants’ post-surprise exploration reflects explanation-seeking. In Experiment 1 we showed 20 12-month-olds a surprising event involving object solidity. Infants saw a truck pass through an apparently solid wall. Next, the experimenter rotated the wall so infants could see it fully–providing additional evidence that the wall was solid. Infants’ looking at the wall was measured. Finally, infants got the opportunity to freely explore the familiar truck that had just defied solidity, versus a novel ball. As predicted, we found that infants preferentially explored the surprising object, t (19) = 2.70, p = .014, confirming that surprise motivates targeted exploration. But were infants exploring in search of an explanation for the surprising event? To find out, in Experiment 2 we showed 20 12-month-olds the same surprising event, in which a truck appeared to pass through the wall. This time, the experimenter then rotated the wall to reveal that it had a large opening in its face–thereby providing a plausible explanation for the surprising event. Infants were then given the truck and a novel ball to explore. In contrast to Experiment 1, infants did not preferentially explore the surprising truck, t (19) = .36, p = .72. Moreover, individual differences in infants’ exploratory preferences were predicted by their earlier interest in the visual explanation (the wall with the opening). We found that infants who had looked longest at the explanatory information (wall with opening) demonstrated the greatest preference to explore the novel distractor object (ball) that was unrelated to the surprising event. This sharply contrasts with Experiment 1, in which infants who looked longest at the solid wall (which offered no explanation) demonstrated the greatest preference for the familiar object (truck) that had behaved surprisingly, F (36) = 3.45, p = .07 (Figure 1). Hence exploration of an object that violated expectations was abolished by providing infants with an explanation of the violation. Experiment 3 served as a control, in which 20 12-month-old infants witnessed the truck become stopped by the solid wall (an expected, non-surprising event). Here, as predicted, infants demonstrated no exploratory preferences, t (19) = .78, p = .44 (Figure 2). Additionally, unlike the previous two experiments, individual differences in looking at the wall after the event did not predict subsequent exploration, F (18) = .34, p = .57. Together, these findings suggest that surprise-induced exploration reflects infants’ search for explanations. Additionally, infants differ in the degree to which they are able to recognize an explanation for a surprising event.

S15           New insights on the origins of self in early childhood: links between beliefs, behaviors, and experience
Tamar Kushnir, Cornell University
One of the most common misconceptions about young children is that they have limited conceptual knowledge of the self. For years it was thought that preschool children understand themselves only via superficial possessions ( “I have a pet turtle”) or simple social categories (“I am a girl”), lack accurate awareness of their own abilities and traits, and are prone to overconfidence. Because of this misconception, the potential role of self-concept in driving important behavioral achievements of childhood- the ability to self-regulate, to learn in academic contexts, to behave morally towards others – has been largely ignored. This symposium brings together four talks that challenge these misconceptions. We show that developing self-beliefs are abstract and accurate, are shaped in important ways by input and experience, and are linked to motivation, self-regulation, and moral behavior. The first talk presents a new method showing that 4- to 12-year-olds across socio-economic and cultural groups have abstract beliefs about intelligence, and these beliefs link directly to their motivation to learn. The second talk shows that how 4- to 8-year-olds hear others praised for effort or ability influences their evaluations of their own behaviors and can undermine confidence. The third talk shows that 4- to 8-year-old children have beliefs about their ability to control their own desires and impulses, and that these beliefs may be shaped by developmental improvements in inhibitory control in certain cultural contexts. The final talk shows that preschoolers who engage in sharing behaviors by personal choice develop conceptions of self that influence subsequent moral actions. These talks together offer new methodological tools and theoretical insights that can be used in further uncovering the origins of self-knowledge in early childhood.

Melis Muradoglu, New York University - Mindsets about intelligence in early childhood

Melis Muradoglu¹, Nim Tottenham², Andrei Cimpian¹

¹New York University, ²Columbia University

Children’s beliefs about the nature of intelligence have important consequences for their motivation and performance in school. Students who believe that intelligence is malleable (a growth mindset) fare better in school than students who believe intelligence can’t be changed (a fixed mindset)–e.g., they are more resilient after experiencing failure and more likely to seek out academic challenges. Despite recent evidence that young children are capable of representing intelligence as either malleable or fixed (e.g., Muradoglu & Cimpian, under review), most research on mindsets has focused on older children, adolescents, and college students (e.g., Romero et al., 2014). The current research sought to (1) develop a valid and reliable scale for measuring mindsets in young children and (2) examine how young children’s mindsets relate to important cognitive variables (e.g., IQ) and academic achievement. Samples in these studies were diverse and international, being drawn from the US in Studies 1 and 3 and from South Africa in Study 2. Further, because growth mindsets are particularly beneficial for children at risk for academic underachievement (e.g., Claro et al., 2016), Study 3 involved a sample of adversity-exposed children. In Study 1, we developed a new scale to measure mindsets in young children and assessed its basic psychometric properties in a sample of 112 4- to 6-year-old children (56 girls; Mage = 5.50). To assess test-retest reliability, 23 children completed the scale again roughly one month after the initial test. Items in the scale were internally consistent (12 items, α = .88) and scores were reliable over time (r = .51, p = .01). Further, children with growth (vs. fixed) mindsets, as measured by our new scale, were more likely to seek academic challenges (i.e., they chose a difficult puzzle that they could learn from but might also get wrong; r = .29, p = .002), which provided evidence for the validity of the scale. Study 2 replicated this pattern in a sample of 331 South African children in grades 2-5. For example, children with growth (vs. fixed) mindsets were more likely to choose challenging tasks (r = .29, p < .001). Additionally, children with growth (vs. fixed) mindsets in this sample earned higher grades in mathematics (r = .16, p = .019). Study 3 extended this work by examining how mindsets relate to important cognitive variables in 6- to 12-year-old children (N = 249). A subset of these children had been previously exposed to caregiver adversity (e.g., foster care; n = 144); the others had not (n = 105). In addition to our mindset scale, children completed the Wechsler Abbreviated Scale of Intelligence (an IQ measure) and measures of executive functioning (the flanker test) and academic achievement (the Wechsler Individual Achievement Test). Children exposed to adversity and children in the control group did not differ in their mindsets (p = .66). Importantly, regardless of adversity exposure, children with growth mindsets had higher IQs (β = .24, p < .001), better inhibitory control (β = .19, p < .001), and higher reading achievement (β = .14, p < .05) than children with fixed mindsets. Together, these results suggest that our new scale is a promising tool for measuring mindsets about intellectual ability in young children. Additionally, these findings identify important correlates of mindsets in diverse groups of school-age children, including IQ, executive functioning, and academic achievement.

Lin Bian, Cornell University - Preschoolers use praises as social comparative cues

Lin Bian¹, Lining Sun², Michelle Wang², Steven Roberts²

¹Cornell University, ²Stanford University

As adults, we are constantly comparing ourselves to others and incorporating the evaluative judgments into our self-concepts (e.g., Festinger, 1954). However, it is unclear whether young children can form abstract self-concepts based on behavioral evidence. For example, Ruble and her colleagues (1980) found that children do not consider social comparative information when evaluating their own competence until age seven. Here, we devise a simple paradigm to assess preschoolers’ social comparison ability. Given that preschoolers understand praises as cues indicating satisfying performances (e.g., Cimpian, Arce, Markman, & Dweck, 2005), we ask if preschoolers make inferences about their competence based on the praises offered to their peers (“other-oriented” praises). Additionally, our second goal is to examine whether subtle linguistic cues in the other-oriented praises affect children’s self-evaluations. Prior work has demonstrated that praising children’s trait (e.g., “You are a good girl/boy!”), but not their specific behavior (e.g., “You did a good job drawing”), undermines their motivation when they encounter setbacks (Kamins & Dweck, 1999). In the present work, we investigate whether the content of the other-oriented praises affects children’s self-evaluations. Forty-two 4- to 5-year-old children (55% girls) and sixteen 6- to 8-year-old children (62% girls) from a small university town were invited to play a drawing game. After the drawing game, an experimenter told the participant that another child drew the same picture that day. Next, the experimenter took out the other child’s picture without displaying it to the participant. After examined it carefully, she said “Wow, s/he (gender-matched with the child) is a great drawer!” (Trait-praise condition), “Wow, s/he did a great job drawing!” (Instance-praise condition) or provided no feedback (Baseline condition). Then, she looked at the participant’s drawing again and provided no feedback. Children were asked four questions that assessed their evaluations of their drawing ability (e.g., “Do you think you were good or not good at this drawing game?”). We found that both younger and older children became less confident about their drawing abilities after being exposed to other-oriented praises, ps < .05 (see Figure 1). This condition effect was not moderated by age significantly, suggesting a continuity in using praises as social comparison cues across development. More specifically, children in the trait-praise and instance-praise conditions evaluated their drawing ability lower than children in the baseline condition, ps < .05, but children’s self-evaluations in the trait-praise and instance-praise conditions did not differ significantly, p = .974, suggesting that the two types of other-oriented praises exhibit similar negative effects on undermining children’s self-evaluations. In sum, children as young as age 4 engage in social comparisons to draw inferences about their own abilities, which provides evidence challenging the claim that young children’s self-concepts are concrete and insensitive to specific behavioral information. Our findings also indicate that hearing praises of peers’ specific episodes or their general traits has equivalent detrimental effects on children’s self-efficacy. Ongoing research is investigating ways to frame other-oriented praises (e.g., “S/he worked very hard on this drawing!”) that would undermine these negative consequences.

Alice Zhao, Cornell University - The relationship between beliefs about self-control and self-control behaviors in childhood

Xin (Alice) Zhao¹, Adrienne Wente², Alison Gopnik², Tamar Kushnir¹

¹Cornell University, ²UC Berkeley

Research has revealed important developmental and individual differences in children’s self-control during early childhood (e.g., Carlson & Moses, 2003). Self-control ability measured in early childhood has also been shown to predict major outcomes later in life, including academic performance and social functioning in elementary school (Blair & Razza, 2007; Eisenberg et al., 2001), educational achievement and emotional coping skills in adolescence (e.g., Mischel, Shoda, & Rodriguez, 1989), as well as finance outcomes and health in adulthood (Moffitt et al., 2011; Ayduk et al., 2000). Research has also identified influences that can facilitate children’s self-control. For example, strategies such as distraction, reframing and distancing (see Mischel & Ebbesen, 1970; Mischel, 2014; White & Carlson, 2016) and regular intervention activities and curricula (e.g. martial arts and mindfulness practices) (Diamond & Lee, 2011) have been shown to improve self-control in childhood. Recent research has started to investigate another type of potential influence: children’s beliefs about self-control. Research has found that beliefs about free will, particularly the ability to willfully “choose to” act against or inhibit strong desires, develop during early childhood (e.g., Kushnir et al, 2015). The current studies investigate the relationship between self-control beliefs and children’s emerging self-control abilities. In Study 1, 54 4- to 8-year-olds completed a questionnaire measuring their self-control beliefs. Children also completed four inhibitory control tasks. As expected, children’s inhibitory control improved across this age range, r(52) = .53, p < .001. Children’s beliefs also changed with age: they became increasingly optimistic about their self-control abilities, r(52) = .61, p < .001. Critically, controlling for age, children who held more optimistic self-control beliefs performed better in the self-control tasks, r(51) = .42, p = .002. In Study 2, we investigated the causal direction of the link between self-control behaviors and beliefs. One hundred and forth-nine 4-and 5-year-olds completed two self-control tasks and a free will belief questions in a randomly assigned order (behaviors first or beliefs first). Controlling for age, results replicated the significant positive correlation between self-control beliefs and behaviors, r(144) = 1.73, p = .036. Results also revealed a causal influence of behavior on beliefs: we found significant correlation for the children who completed the self-control tasks first, r(72) = .333, p = .004; but not for children who answered the belief questions first, r(69)= -.05, p=.677. A closer look showed that the correlation was driven by self-control failure – children who failed both self-control tasks were most likely to say they could not “choose to” against or inhibit their own strong desires. These findings reveal a relationship between beliefs about self-control and self-control behaviors in U.S. children. Immediate first-person experience with self-control can shape children’s beliefs about their self-control abilities. Our follow-up research shows that this relationship may be moderated by cultural contexts: The relationship was only apparent in the U.S. sample, but not in other cultures tested so far (i.e., Singapore, China, Peru). We will discuss how cultural context may play a role in the relationship between beliefs about self-control and self-control behaviors.

Nadia Chernyak, University of California, Irvine - Where do I fit in: Self-other overlap and resource distribution in preschool-aged children

Nadia Chernyak¹

¹UC – Irvine

Recent work in developmental psychology has made great strides in understanding the structure, mechanisms, and developmental timeline of children’s sense of fairness. However, most theories to date have neglected the role of the self in children’s acquisition of fairness. In the first set of studies, I present work showing that children are more likely to act prosocially when the actions they’ve engaged in are personally costly (that is, after incurring a cost to the self). Across two studies (Study 1 N = 48; Study 2 N = 49), preschool-aged children (3-4-year-olds) who gave away valuable objects (as opposed to non-valuable objects) were more likely to give again in the future. This work suggests that children evaluate prior actions with respect to how self-relevant they may have been – actions that were personally costly are more likely to be perceived as such, and thus motivate subsequent prosocial behavior. Thus, acting on sympathy for others is fueled by the understanding that the self must incur a cost in order to act. The next set of studies will investigate the extent to which prosocial behavior is motivated by understanding that the self is similar to others. This work will use 3 measures of social-relatedness: 1) an emotion similarity measure in which children will be shown silhouettes depicting the “self” and “other” asked to point to where each one experiences various emotions. Data will be coded for the degree of distance children place between themselves and another, with the expectation that lower degrees of distance show that children integrate the self with another; 2) a self-other body similarity measure in which children’s gestures between self and other will be analyzed for degree of overlap; and 3) a physical proximity measure in which children will be asked to place the self and other in 2 seats at a table. The work will validate these measures and also investigate the relationships between self-other overlap and prosocial behavior. Results will be discussed in terms of underlying theories about fairness and self.


S16         Young children’s scientific theory building and knowledge acquisition
Chairs: Nicole Larsen & Vaunam Venkadasalam, University of Toronto
Discussant: Samuel Ronfard, University of Toronto at Mississauga
Early in development, children’s intuitive theories about various scientific domains are built through direct observation and informal conversations with others. These naïve theories are often at odds with correct scientific theories and incorporate misconceptions that can interfere with the acquisition of scientific knowledge (Carey, 1991; Kelemen, 2019; Vosniadou, 2013). Given that some of these misconceptions persist and co-exist alongside correct theories (Shtulman, 2017), early intervention is critical: It allows children to develop robust scientifically accurate concepts that can outcompete these misconceptions and promotes science achievement in later school years (Morgan et al., 2016). Not all students have access to the same high-quality science education, nor to enriched science experiences in informal settings. Picture books are an accessible resource for all students that allows educators to deliver coherent mechanistic explanation of hard-to-learn scientific concepts. Unfortunately, little is known about how such explanations can best be integrated into picture books to support learning in formal and informal settings. This symposium addresses this question. Paper 1 examines the potential of realistic and fantastical elements in stories to hinder and support conceptual change. Paper 2 examines the impact of integrating parent-child picture book reading on the development of preschoolers’ scientific concepts in an informal setting. Paper 3 explores how picture book reading can be used in a formal classroom setting and combined with experiential learning to promote the best learning outcomes. An expert in children’s scientific learning from picture books will conclude by discussing the implications of this work.

Emily Hopkins, University of Scranton - When do fantastical stories benefit young children's learning?

Emily Hopkins¹, Deena Weisberg²

¹University of Scranton, ²Villanova University

Educational stories often contain both possible and fantastical information. How do young children decide what to learn? Research suggests that children are more likely to learn from a realistic story than a fantastical one (e.g., Richert et al., 2009; Walker et al., 2015), since the similarity between the realistic story and reality facilitates transfer. However, other work hints that fantasy may lead to superior learning under certain circumstances (Weisberg et al., 2015; Stahl & Feigenson, 2017). In four studies, we examined children’s learning from realistic vs. fantastical stories to determine when each type of story may be most beneficial to learning. Studies 1 and 2 confirmed that children do not always learn better from realistic stories. In both studies, preschoolers (Mage = 4;9) were presented with a novel fact embedded in a story that was either fantastical or realistic. In Study 1, children who heard the fantastical story were marginally more likely to say the novel fact was true than children who heard a realistic story: χ2(1, N = 40) = 2.85, p = .09, Φ = 0.27. Study 2 also found that children who heard the fantastical story were marginally more likely to say the novel fact was true: χ2(1, N = 72) = 3.50, p = 0.06, Φ = 0.22. These studies suggest that fantasy elements can help learning. Why might this be the case? Work with infants suggests that physically impossible events may lead to better learning by increasing attention and prompting learners to seek explanations (Stahl & Feigenson, 2015). To test these claims, Study 3 examined children’s learning from stories that contained events that violated either physical or biological laws. Preschoolers (Mage = 4;10) heard a story with information about two science principles (inheritance and balance) and were pre- and post-tested on each. The stories contained events that were physically impossible (e.g., walking through walls; n = 34), biologically impossible (e.g., aging backwards; n = 33), or possible (n = 34). The balance principle was difficult for children to learn; scores on this problem were significantly lower than the inheritance principle at both pre- and post-test. Children showed significant learning of this difficult principle only when the story contained no impossible events: t(33) = 2.36, p = .02, d = 0.40. However, for the inheritance principle, children showed significant learning only when the story contained physically impossible events: Pre/post-test difference scores in this condition were significantly different from zero, t(29) = 3.91, p < .001, d = 0.71. Study 4 (current n = 67) aims to replicate this effect while also bringing the difficulty of the balance principle in line with the inheritance principle. These studies add to a growing body of work showing that some types of fantasy – physically impossible events – can benefit children’s learning. Further, they suggest that fantasy elements benefit learning only in particular circumstances. When a problem is difficult and children are uncertain about it, they may default to quarantining information from fantastical sources to avoid learning inaccurate information. However, for information that is within children’s zone of proximal development, physically impossible events may heighten their attention and lead to deeper processing.

Kathryn Leech, Harvard University - How shared book-reading can boost children's scientific discourse and understanding

Amanda Haber¹, Youmna Jalkh¹, Kathleen Corriveau¹, Kathryn Leech²

¹Boston University, ²University of North Carolina – Chapel Hill

Children’s understanding of unobservable scientific entities (e.g., electricity, germs) largely depends on testimony from other people, as this knowledge cannot be acquired solely through firsthand observation (Harris et al., 2018). One form of testimony is parental explanations that highlight the mechanism underlying a scientific entity. Mechanistic explanations are particularly helpful in promoting children’s conceptual understanding and importantly, help learning transfer to novel situations (Lombrozo et al., 2018). However, mechanistic explanations are rare in parent-child conversation (Authors, under review), raising the question of how best to encourage these conversations. The current study aimed to boost parent-child mechanistic conversation by embedding mechanistic explanations about electricity into storybooks. We were particularly interested in whether this manipulation increased mechanistic language during a subsequent scientific activity, and whether it also improved children’s understanding of electricity’s mechanism. Four- to five-year-old children (N=58) and their parents participated at a science museum and preschools in the United States. During the book-reading phase, parents read a researcher-developed storybook to their children, with eight explanations about electricity embedded within the story. Half of the dyads read books with explanations that referenced mechanisms of electricity (It’s a kind of energy that make things move, light up, or get hot), while the other half read books containing non-mechanistic explanations (It’s a kind of energy that we can’t always see but is very powerful). Otherwise, stimuli in each condition were matched for plot, length, and linguistic complexity. Next, during the dyadic phase, the dyad worked collaboratively to assemble electrical components of a circuit toy so that a light would turn on. Dyadic speech was transcribed using CLAN software (MacWhinney, 2000). Coders then identified whether speaker utterances contained evidence of mechanistic reasoning (e.g., We need to connect these pieces in a circuit). During the learning phase, children first independently assembled a novel circuit, and second, answered comprehension questions to gauge their understanding of electricity’s mechanism. During the dyadic phase, participants assigned to the mechanistic condition (M=94%) were more successful at completing the circuit than the non-mechanistic condition (M=65%), B=3.01, SE=1.12, p=.007. During this phase, mechanistic dyads also produced significantly more mechanistic language (M=7.0% of their total utterances) than non-mechanistic dyads (M=4.0%), F(1,54)=4.35, p=.04, d=.08. Significant condition differences were also observed during the learning phase such that children in the mechanistic condition were more successful at the independent circuit task (M=70%) than children in the non-mechanistic condition (M=38%), χ2(1)=7.60, p=.008. There was no main effect of condition on children’s comprehension, B=-0.30, SE=0.50; p=.54. However, we observed a significant interaction between condition and dyadic mechanistic language on comprehension: children in the mechanistic condition were more likely to successfully comprehend the mechanism if they participated in more mechanistic conversation during the dyadic phase, B=-32.91, t=2.28 p=.02. These results suggest that everyday routines such as book-reading can help support children’s scientific discourse and understanding.

Vaunam Venkadasalam, University of Toronto - Science in the classroom: Addressing science misconceptions in the early years through books and play

Nicole Larsen¹, Vaunam Venkadasalam¹, Patricia Ganea¹

¹University of Toronto

In adopting teaching methods for young children it is important to capitalize on activities that are both enjoyable to children and effective for learning. This is particularly important for science education, as children can be especially resistant to revising science misconceptions (Pine, Miesser, & St. John, 2001). Revision of incorrect beliefs is a gradual process and conceptual change requires prolonged and diverse experiences with the concept in question (Vosniadou, 2013). Joint picture book-reading is an activity that children enjoy and also learn from (Venkadasalam & Ganea, 2018). Play can also be used in the classroom as a way of learning new information. In this research we investigated whether combining book reading with two types of child-focused play, either guided or free-play, leads to differential learning outcomes. In free play, the child has complete autonomy, whereas in guided play, the adult scaffolds learning activities through prompts. We recruited two classes of kindergarten children (ages from 3 to 6) to participate in the intervention. We developed books and activities on two different physics concepts: falling objects and sinking and floating. Children in each class were randomly assigned to either a Picture Book & Guided Play condition or a Picture Book & Free Play condition for each concept, and their condition was switched for the second concept. The intervention for each concept took place in small groups (5 children) twice a week for two weeks. Each week, children read one book (Days 1 and 3) and completed one activity (Days 2 and 4). During the guided play the experimenter provided children with prompts and explanations. For the free play, children were given the same materials, reminded of the picture book and given a learning goal but no guidance as to what to do with the materials. Children were individually pre-, post- and delay-tested to determine learning over time. The post-test occurred on Day 4 and the delay-test 6-to 8-weeks later. In each of the 3 test phases children were shown 4 pairs of objects. Within each set of 4 pairs, two pairs of objects were the same weight, and two were different weights. For each test phase, children received a different object set, but the order of these sets was counterbalanced. Children were asked to make a prediction (for falling pairs ‘Would the objects fall at the same time or one faster than the other?’; for sinking and floating pairs ‘Which object would sink and which would float?’) and to explain their predictions. The outcome measure was children’s explanations scored on a 3-point scale. A one-way ANOVA revealed no differences in pretest scores between the two conditions (p =.414). A 2 (condition) x 3 (test phase) mixed ANOVA revealed no main effect of test phase (p =.097) or condition (p =.143). However, there was a significant interaction between test phase and condition (p =.026). Results showed that at post-test, children revised their physical science misconceptions through both conditions (p =.399), but children in the guided play condition retained their knowledge at a two-month delay significantly better than those in the free play condition (p =.024). We conclude that combining picture books with guided play is best for long lasting learning.

Oral Papers IV

O4.1       Developing a measure of young children’s self-perceptions of cognitive control skills

Presenter: Robbie Ross

Robbie Ross¹, Dare Baldwin²

¹University of South Carolina, ²University of Oregon

Cognitive control skills in early life are vital to success throughout the lifespan. Such skills have been positively linked to a host of important short- and long-term outcomes across many diverse domains (e.g., Moffitt et al., 2011; Schmidt, Pratt, & McClelland, 2014; Son, Lee, & Sung, 2013). Similarly, self-perceptions such as self-efficacy, implicit beliefs about cognition, and self-concept have all been shown to predict meaningful variance in important outcomes like academic performance, even when controlling for direct measures of cognitive ability (e.g., Bong & Skaalvik, 2013; Dweck & Leggett, 1988; Lyons & Ghetti, 2011; Lyons & Ghetti, 2013). To what extent do young children have a self-perception of their own level of cognitive control functioning? To date, no measures have been developed to assess young children’s perceptions of their own skills in this regard. In the current study, we aimed to develop and validate a scale capturing children’s self-perceptions of their cognitive control skills via a puppet interview (Ablow & Measelle, 1993). 125 children aged 4- through 7-years participated in an interview in which two identical puppets, voiced by the experimenter, each presented a contrasting pole on a series of 26 self-perception scale items and then asked participants to indicate which one they most identified with. Participants were free to respond in a number of developmentally appropriate ways to indicate their endorsement and responses were translated by trained coders into a 7-point Likert scale. Participants also completed a battery of lab-based behavioral measures and self-reports of other self-perception constructs. As well, parents completed a battery of parent-report measures of their child’s self-regulatory functioning. Scale analyses of these interviews suggest the scale elicits responses that cluster around two correlated, but separable components: 1) Self- and Emotion-Regulation, and 2) Attention Modulation. Responses on these two subscales showed moderate to strong internal consistency, were strongly correlated with parent reports of similar skills, and self-reports of related constructs, but showed no systematic relation to behavioral tasks measuring executive functioning abilities. The findings suggest that young children are capable of reflecting and reporting coherently on their own cognitive control skills, and that these children’s responses to the new scale’s items correspond to parent reports of similar abilities. Further scale refinement and targeted validation efforts are needed; however, these encouraging early results suggest the new scale holds potential to uncover how children’s self-perceptions influence their learning success.

O4.2       Flexible attention to numerical and spatial magnitudes and children’s development of math skills in preschool

Presenter: Natalie Sheeks

Natalie Sheeks¹, Yiqiao Wang², Victoria Bartek², Elizabeth Gunderson², Mary Fuhs¹

¹University of Dayton, ²Temple University

The Flexible Attention to Magnitudes (FAM) account proposed here argues that a fundamental challenge young children face in early math–that is not addressed by current interventions–is disentangling numerical and spatial dimensions of magnitude and making decisions about which dimensions are relevant in solving certain early math problems. Previous studies showed that children have difficulty making numerical magnitude judgments in cases where numerical magnitudes (discrete and countable objects) and spatial magnitudes (continuous dimensions such as surface area) are incongruent – for example, determining which box has more rocks when one box has ten small rocks and another has two large rocks. We argue that FAM ability is a key predictor of overall math ability, including tasks that require the flexible attention to both spatial and numerical dimensions of magnitude (e.g., number line estimation). Children’s FAM ability is also proposed to be related to children’s executive functioning skills, as children must learn when to inhibit attention to irrelevant dimensions of magnitude and when to shift between them. We hypothesized that children’s FAM ability would be significantly associated with children’s growth in math skills across the preschool year, controlling for executive functioning skills, language, and other covariates (age, gender, family income). Preschool-aged children (N = 213; 53.1% female; Mage = 54.2 months; 47.4% of families had incomes below $25,000) were recruited in the fall of their preschool year and were assessed in both the fall and spring on a battery of measures, including a newly developed measure of FAM ability, executive functioning skills (Minnesota Executive Function Scale), math achievement (Woodcock-Johnson Tests of Early Cognitive and Academic Development Number Sense Scale), and language (Woodcock-Johnson Tests of Early Cognitive and Academic Development Picture Vocabulary Scale). In the FAM ability task, children were first asked to choose which of two boxes contained bigger stars (6 trials), then they were asked to switch rules and choose which of two boxes contained more stars (6 trials), and finally, they were asked to shift between choosing which of two boxes contained bigger and which of two boxes contained more stars (12 trials). Trials were all incongruent with respect to spatial and numerical magnitudes such that the box containing more stars always had smaller stars, while the box containing bigger stars always had fewer stars. Children were not prevented from counting on the task, and the number of stars in each box ranged from 1 – 10. Results indicated that children’s FAM ability in the fall was a significant predictor of their spring math achievement (B = .17, p = .019), controlling for fall math achievement, executive functioning skills, language skills, and other covariates. Results will be discussed in terms of plans to improve the FAM ability measure and potential strategies for testing the malleability of FAM ability in common early childhood math tasks.

O4.3       The neural basis of selective and flexible dimensional attention

Presenter: Aaron Buss

Aaron Buss¹, Anastasia Kerr-German²

¹University of Tennessee – Knoxville, ²Boys Town National Research Hospital

Between the ages of 3 and 5, children develop greater control over the allocation of attention to visual dimensions. For example, children develop the ability to flexibly shift attention between visual dimensions and to selectively process information within a visual dimensions of an object. Previous proposals have suggested that selective and flexible attention are developmentally related to one another; however, the relationship between flexibility and selectivity has not been systematically probed at the behavioral and neural levels. Previously, we developed a dynamic neural field model to explain behavioral and neural on the development of flexible attention in the context of the dimensional change card sort task (DCCS; see Figure 1a). The DCCS requires children to switch from sorting cards by shape or color to sorting cards by the other dimension. Typically, the majority of 3-year-olds fail to switch rules in the task but the majority of 4-year-olds can switch. The model operates based on neural activation dynamics within populations of simulated neural units that are tuned to dimensions of perception and action. The model is composed of an object representation system that binds visual features such as shape and color to spatial locations and a label learning system that associates labels such as “color” or “shape” with visual features. Performance in the DCCS task relies upon the quality of associations between labels and visual features: with weak associations the model perseverates on the habits accumulated during the initial sorting phase, but with strong associations the model switches rules. Thus, with stronger associations between labels and visual features, the model can flexibly shift attention between dimensions. We have previously demonstrated that this model explains both behavioral and neural data on the development of performance in the DCCS task (Buss & Spencer, 2014; 2018). In this report, we generalize this model to also explain development of attentional selectivity. Specifically, we generated behavioral and neural predictions on the association between performance on the DCCS task and a task that requires selective attention, the triad classification (TC) task (Smith & Kemler, 1977; see Figure 1a). The TC requires children to selectively attend to a single dimension to pick matching objects. Whereas the DCCS requires explicit attention, the TC task requires implicit attention and determining which dimension is relevant based on the configuration of stimuli. We tested model predictions by administering the DCCS and TC tasks to 3.5- and 4.5-year-olds while functional near-infrared spectroscopy data were recorded from bilateral frontal cortex. Results supported model predictions. First, children who demonstrated flexibility in the DCCS task also demonstrated higher levels of selectivity in the TC task (see Figure 1b). Second, performing well on either task was associated with increased activation in right frontal cortex as indexed by changes in HbO (see Figure 1c). These results suggest that dimensional attention development is a common process that drives developmental changes across tasks that require the controlled allocation of attention to visual dimensions across differing task demands. Moreover, development in right lateral frontal cortex is responsible for the development of dimensional attention, which we propose is involved in the formation of representations that bind visual features to auditory labels.

O4.4       Children don’t really think about rational numbers as being equivalent in size

Presenter: Lauren Schiller

Lauren Schiller¹, Robert Siegler¹

¹Columbia University

The importance of fractions for math outcomes (Siegler et al., 2012) and evidence supporting a theory of numerical development that integrates all rational numbers (Siegler, Thompson, Schneider, 2011) has been documented, but little is known about the relations among notations. This is troubling because understanding involves incorporation of concepts into an internal network, whereby the degree of understanding is determined by the strength and accuracy of connections between related concepts (Hiebert & Carpenter, 1992). Therefore, fluid understanding of the connections among fractions, decimals, and percentages indicates deeper understanding and superior performance (Moss & Case, 1999). While there’s some information about comparison across fractions and decimals (e.g. Hurst & Cordes, 2016), much less is known about comparison across all notations including percentages. Before debating which notation is best to learn first (Tian & Siegler, 2017), we must first understand how children conceptualize the relation among all notations. Thus, we tested middle school children (N=70) to explore how they perform on magnitude comparisons across all three notations; half of the items compared values with identical or nearly identical digits (e.g. compare 4/5 versus 45%) and half were matched for magnitude across all notations with small, medium, and large differences between compared values (e.g. compare .40 versus 25%, 2/5 versus .25, .4 versus 1/4, etc.). Results showed that children do not actually think about equivalent rational numbers as being equivalent; instead our findings showed that students have a bias towards perceiving percentages as larger than fractions/decimals and fractions as larger than decimals. Accuracy was 92.3% for items where the percentage was larger than fraction/decimal, as compared to 60.7% accurate for items where the percentage was smaller than fraction/decimal. Similarly, overall accuracy was 82% for items where the fraction was larger than the decimal, as compared to 54% accurate for items where the fraction was smaller. Evidence for the bias is reflected in statistically significant differences between mean scores for items where Percent>Fraction and Fraction>Percent (p<.001, t(69)= 7.069), Percent>Decimal and Decimal>Percent (p<.001, t(69)= 6.538), and Fraction>Decimal and Decimal>Fraction (p<.001, t(69)= 6.452). After number line instruction with all notations, this percentage is greater than fraction/decimal bias persisted, though it weakened somewhat. Accuracy was 80% for Percent>Fraction as compared to 70.7% accurate for Fraction>Percent (p=.02, t(69)= 2.346) and 95% accurate for Percent>Decimal as compared to 76% accurate for Decimal>Percent (p<.001, t(69)=5.187). The instruction was fruitful in reducing the fraction is greater than decimal bias, as there was no difference in performance when the fraction was larger or smaller than the decimal at posttest (p=.874, t(69)=.159). We are currently testing the replicability of this finding in another sample, but the fact that post-test results following targeted number line instruction demonstrate greater accuracy when the percentage was larger suggests that the percentage is greater than fraction/decimal bias is real. Finally, to our knowledge, this study is the first to examine magnitude comparisons across all three rational number notations. The initial finding of a skewed perception of size warrants further research in this area.

O4.5       The dynamic nature of children’s strategy use after receiving feedback in decimal comparisons

Presenter: Kexin Ren

Kexin Ren¹, Elizabeth Gunderson¹

¹Temple University

When students start learning decimals, they may incorrectly apply features of their prior numerical knowledge (e.g., whole-number or fraction rules) when reasoning about decimals. However, since whole numbers, fractions, and decimals are fundamentally different, these whole-number and fraction strategies often lead to incorrect solutions. We examined whether receiving immediate feedback while comparing decimal pairs that are either congruent with whole-number rules (e.g., treating numbers with more digits as larger in magnitude) or fraction rules (e.g., treating numbers with fewer digits as larger in magnitude), would lead students to change their decimal-comparison strategy. We hypothesized that, given feedback on their decimal-comparison accuracy, students would reduce their use of whole-number rules, temporarily increase their use of fraction rules (i.e., try an “opposite” strategy), and ultimately increase their use of an accurate decimal-comparison strategy. We also hypothesized that improvement in decimal comparisons would transfer to decimal comparisons involving different numbers of digits. Sixth- to 8th-graders (N=171) were randomly assigned to an experimental or control condition. The conditions differed only in the initial training block, in which they compared 84 decimal pairs either with feedback (experimental condition) or without feedback (control condition). All training pairs included a 1-digit versus 2-digit decimal; half were congruent with the whole-number rule (e.g., 0.3 vs. 0.86) and half were incongruent (e.g., 0.8 vs. 0.34). All participants then completed a posttest without feedback, which included 28 decimal comparisons of the same format as the training block, and decimal comparisons involving different numbers of digits than the training block (e.g., 0.2 vs. 0.356 and 0.75 vs. 0.142). See for preregistration. Consistent with our hypotheses, generalized estimating equations (GEEs) showed that in the experimental condition (with feedback), the percent of students using a whole-number rule decreased over time during the training block (Slope B=-.12, SE = .03, p < .001), and the percent of students showing decimal accuracy (high accuracy on congruent and incongruent pairs) increased over time (Slope B=.10, SE = .02, p < .001). As expected, in the control condition, the percentage of students who have whole-number bias or decimal accuracy did not significantly change over the training trials (ps > .360). However, our hypothesis of a temporary increase in fraction bias during training was not supported (ps > .493). On all posttest measures, whole-number bias was lower in the experimental than control group (ps ≤ .001), and decimal accuracy was higher in the experimental condition than the control (ps ≤ .005). Thus, students’ whole-number bias in decimal comparisons was reduced after receiving feedback on decimal pairs where the whole-number rule was not applicable half of the time. Students quickly learned the correct decimal strategy without necessarily trying out the incorrect fraction strategy. Additionally, students’ high accuracy on posttest decimal comparisons with different numbers of digits indicates that they did not simply learn a procedural rule but a more comprehensive understanding of decimal place value. Overall, the current study provides insights for reducing students’ misconceptions in learning decimals, which could be beneficial to their long-term mathematical development.

17:45 – 19:00: Poster session 4 & Exhibitors