What to Attend to: From the Lab to the Classroom

Rachel Wu1, Andria Shimi2, Michael Solis1, and Gaia Scerif 3

Abstract ■ Research in adult cognitive addresses the bi- the lifespan. In particular, we review the research program that directional relationship between attentional selection and prior we have developed over the last few years, describe the con- knowledge gained from learning and experience. This research straints that we have faced in integrating adult and developmen- area is ready for integration with developmental cognitive tal paradigms, and delineate suggested next steps to inform neuroscience, in particular with educational neuroscience. We educational neuroscience in more applied ways. Our proposed review one aspect of this research area, learning what to attend path of integration transitions from basic to applied research, to, to propose a path of integration from highly controlled while also suggesting that input from could inform experiments based on developmental and adult cognitive new basic research avenues that may more likely yield out- theories to inform cognitive interventions for learners across comes meaningful for education. ■

INTRODUCTION to adulthood. We highlight the significance of this re- Many basic research areas in adult search question by placing it within the broader field of have the potential to have translational significance and, cognitive neuroscience and summarize our teams’ find- in turn, generate useful avenues for future basic research. ings over the past few years. We present methodological In particular, adult cognitive neuroscience research that limitations and potential solutions when integrating adult dovetails with developmental cognitive neuroscience cognitive neuroscience and developmental psychology research has the potential to inform educational neuro- and provide justifications for translating from the lab to . The goals of educational neuroscience are three- the classroom. fold: (1) to better understand underlying neural and behavioral learning mechanisms to improve educational outcomes for all learners, (2) to develop markers that SIGNIFICANCE OF LEARNING WHAT TO can identify learners who are struggling or who are at ATTEND TO risk, and (3) to develop evidence-based therapeutic One research area in adult cognitive neuroscience that is practices to address issues in education (see Ansari, underutilized for bridging basic and applied research Coch, & De Smedt, 2011; Szűcs & Goswami, 2007, for addresses the bidirectional relationship between atten- reviews). Educational neuroscience faces not only the tional selection and prior knowledge gained from learn- challenges of moving from the lab to the classroom (e.g., ing and experience. For the past few years, both as a differences between basic research aims and translational team of collaborating researchers and independently, research aims; Onken, Carrol, Shoham, Cuthbert, & Riddle, we have been investigating this topic in adult cognitive 2014; Weisz, Ng, & Bearman, 2014) but also integrating neuroscience using developmental theories, with the between different basic research areas, such as adult aim of future application to education. We have focused cognitive neuroscience and developmental psychology. on one aspect in particular: learning what to attend. This article briefly reviews our research team’s efforts At least 40 years of experimentation have revealed the in the past few years to integrate adult cognitive neuro- process by which observers find what they are looking for science and developmental psychology to inform educa- (i.e., top–down visual search): The to-be-searched item is tional neuroscience. In particular, we have investigated represented as an “attentional template,” which is a how observers learn what to attend to from childhood prioritized working memory representation, and this template is matched against the current input. Top– down visual search studies have shown that attentional 1University of California, Riverside, 2University of Cyprus, 3Univer- templates can contain a single feature (e.g., shape or sity of Oxford color) or an object (e.g., a red square; e.g., Eimer,

© 2018 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 30:12, pp. 1749–1756 doi:10.1162/jocn_a_01316 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn_a_01316 by guest on 23 September 2021 2014; Olivers, Peters, Houtkamp, & Roelfsema, 2011; and children, who have to learn what to attend to and Wolfe & Horowitz, 2004; Desimone & Duncan, 1995; what to learn as they acquire knowledge about potential Treisman & Gelade, 1980), and even a category or a rule targets and distractors (e.g., Wu, Gopnik, Richardson, & (e.g., Wu & Zhao, 2017; Moores, Laiti, & Chelazzi, 2003). Kirkham, 2011; Wu & Kirkham, 2010). In all top–down search tasks, the participant has to learn It is critical to understand how learning to attend what to attend to (i.e., the target). In one-feature search to relevant stimuli develops, as the development of at- studies, the target (i.e., the feature, such as red objects) tentional processes and their neural correlates undergo is typically easy to determine, especially if an example of significant change from infancy into childhood and adult- the target is provided. In more complex category search hood (see Amso & Scerif, 2015; Power, Fair, Schlaggar, & studies, the target (i.e., the category, such as any letter) Petersen, 2010, for reviews). Correlated but distinguish- may require more time and effort to learn. Classic visual able attentional and executive control networks can be search studies have demonstrated limitations in search identified from early in childhood (e.g., Rueda et al., efficiencies when searching for more than one feature 2004), but their connectivity is characterized by increas- (one feature vs. two or more features in conjunctive ing segregation and differentiation from childhood search) or object (one object vs. two or more objects; into adulthood that supports more efficient see Olivers et al., 2011, for a review). Category search (e.g., Fair et al., 2008; Grayson et al., 2014; see Power studies have demonstrated that one way around this et al., 2010, for a review). How do these developing limitation is by grouping objects into one unit (i.e., a skills support the ability to identify what is relevant to category; Nako, Wu, & Eimer, 2014; Wu et al., 2013; the task at hand, that is, the ability to learn to attend? see also Moores et al., 2003). Similarly, grouping fea- Our research program on learning what to attend to tures into one object allows for higher search efficien- has focused on searching for abstract categories. Unlike cies for multiple features within that object (e.g., Wu, perceptual categories (i.e., grouping objects based on Pruitt, Runkle, Scerif, & Aslin, 2016). common perceptual features, such as wheels for cars), A parallel and now equally influential focus of research more general or “abstract” categories are created by has investigated how information held in memory guides grouping objects based on rules, associations, or rela- the orienting of attention: A wealth of evidence points to tions. Importantly, there may be high perceptual dissim- the influence of the contents of memory on attentional ilarity within an abstract category (e.g., the numeral “4” selection. The evidence includes both implicit benefits does not look like the numeral “5”), and there may be of repeated learned contexts on visual search (e.g., re- high perceptual similarity between abstract categories peated spatial arrangement of objects facilitates speed (e.g., the numeral “5” shares features with the letter of target selection, albeit without explicit recollection; “S”). Constructing abstract categories (e.g., letters, num- Chun & Jiang, 1998) and explicit memory of information bers, food, toys) is a critical skill for young children to encoded during visual search of complex or naturalistic master because these categories play a prominent role scenes (such as recalling as precisely as possible a previ- in many facets of everyday cognition. In particular, from ously learned target location in a scene), which are preschool to early school age (i.e., 3–6 years of age), chil- effects that depend on the interplay between fronto- dren have to learn categories important for education parietal and hippocampal circuits (e.g., Patai, Doallo, & (e.g., numbers and letters). Incorrect construction of Nobre, 2012; Stokes, Atherton, Patai, & Nobre, 2012; these new categories, both at the level of the precision of Summerfield, Lepsien, Gitelman, Mesulam, & Nobre, exemplars within each category and at the level of the in- 2006). As a whole, this literature highlights that atten- efficient use or manipulation of these categories, may re- tional processes influence and are influenced by what is sult in poorer academic outcomes and more advanced learned by adult observers. learning that has, at its core, those categories as building Decades of cognitive neuroscience research have blocks. For example, a growing literature emphasizes the focused on the mechanisms and nature of visual atten- importance of symbolic representations of number as the tion for known targets by adults and more recently on foundations for more complex arithmetic (e.g., Bartelet, the interplay between attention and memory in adults. Vaessen, Blomert, & Ansari, 2014; Holloway & Ansari, In a complementary but novel fashion, our team’sap- 2009). A now vast body of work suggests that attentional proach has been to focus on how observers’ search for and executive skills are a strong correlate and predictor familiar versus unfamiliar or newly learned targets, as well of emerging mathematics from preschool (e.g., Clark, as how search evolves when target characteristics are Sheffield, Wiebe, & Espy, 2013; Bull, Espy, & Wiebe, learned (e.g., Wu, McGee, Echiverri, & Zinszer, 2018; 2008) into childhood (e.g., Bull & Scerif, 2001) and be- Wu, McGee, Rubenstein, et al., 2018; Wu, Pruitt, Zinszer, yond (see Cragg & Gilmore, 2014, for a comprehensive & Cheung, 2017; Wu et al., 2013, 2015, 2016). This issue review focused on education). Perhaps the ability to dis- is not commonly studied with adults, because adults tinguish relevant from irrelevant items to focus on the often know what the search target is, either from explicit relevant items underlies the successive acquisition of instructions in the lab or from prior knowledge in the real many important skills (Steele, Karmiloff-Smith, Cornish, world. However, this issue is very important for infants & Scerif, 2012).

1750 Journal of Cognitive Neuroscience Volume 30, Number 12 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn_a_01316 by guest on 23 September 2021 SUMMARY OF OUR COGNITIVE costs of target knowledge can emerge over a 1-hr exper- NEUROSCIENCE FINDINGS imental training session as the participant learns what the target is (Wu, McGee, Rubenstein, et al., 2018; Wu et al., In the past few years, our team has conducted a number 2013) or can interfere even when the knowledge is im- of visual search studies with adults and children to better plicit and completely task-irrelevant (e.g., scope of an understand how observers learn what to attend to, in object’s category when searching for the specific object; particular how individuals learn to look for abstract Wu, McGee, Echiverri, et al., 2018). In summary, learning categories. Our ERP (N2pc) studies with adults have what is relevant information helps the learner when the shown not only that categorical representations are situation matches what is learned but may lead to inter- measurable (Bayet, Zinszer, Pruitt, Aslin, & Wu, 2018; ference when aspects of the relevant information change Wu, McGee, Echiverri, et al., 2018; Wu, Pruitt, et al., in a different situation. Although further testing in the 2017; Wu et al., 2013, 2015; Nako, Wu, & Eimer, 2014; classroom is required, these processes may be crucial Nako, Wu, Smith, & Eimer, 2014), even as they are being when learning new information or building and consoli- learned (e.g., Wu, McGee, Rubenstein, et al., 2018; Wu dating previously acquired information. et al., 2013, 2016), but also that, once learned, categories We have begun to gather evidence of the existence of a can guide attention almost as efficiently as searching for a similar neural marker of categorical attentional templates single item, where search is based on perceptual fea- in children (Shimi, Nobre, & Scerif, 2015; Shimi, Kuo, tures (Nako, Wu, & Eimer, 2014; Nako, Wu, Smith, et al., 2014). In adults, the N2pc component is the most et al., 2014). The latter results imply that categorical established and fastest physiological marker of top– templates can be basic units of attention, in addition to down, template-based attentional target selection (e.g., perceptual features to which our visual system is tuned, Eimer, 1996; Luck & Hillyard, 1994). Establishing whether such as line orientation. this component exists in younger children and its nature Moreover, our team has aimed to better understand would allow us to better understand how children learn how knowledge may facilitate or decrease search effi- what to attend to. Moreover, other factors, such as ciency to help or hinder the learner’s ability to identify working memory load and temporal decay, seem to im- relevant information for a given task. Grounded in the pact working memory capacity and search efficiency in cognitive neuroscience we have described, future re- children more than young adults (Shimi & Scerif, 2017; search could investigate even more directly how develop- Shimi, Kuo, Astle, Nobre, & Scerif, 2014). Indeed, chil- ing robust abstract categories may be useful for learning dren’s search efficiency is dependent on their working educationally relevant materials.Incontrasttoinfants memory span (Shimi, Nobre, Astle, & Scerif, 2014). Shimi and young children, young adults have a great deal of and Scerif (2015) also showed that young children had knowledge that allows them to find relevant items and more difficulties than older children and adults in search- information efficiently, such as when looking for a known ing for abstract novel meaningless shapes compared with object (e.g., a sandwich) or a broad category (e.g., some- highly familiar items, such as animals, perhaps due to dif- thing to eat for lunch). Young adults (so-called “peak ficulty in maintaining their representations. Such difficul- performers”) often outperform other age groups on vi- ties could be present in an educational setting and may sual search tasks across the lifespan. Adult N2pc ERP stud- impact learning of new material. It is important to note ies have quantified this efficiency: Targets are typically that our research team does not assume that the N2pc located within 200 msec in a visual search paradigm would be identical in younger children and adults or even (e.g., Eimer, 1996; Luck & Hillyard, 1994). Even when that there would be a one-to-one mapping between the searching for a broad subjective category, such as “any adult and child N2pc components. We are interested in healthy food,” participants are highly efficient (Wu, Pruitt, using the N2pc as a tool to understand how developing et al., 2017). However, efficiency for identifying relevant attentional systems impact different types of learning and information comes at a price. Adults can become dis- vice versa. We could investigate when the N2pc seems to tracted by familiar objects (e.g., an apple) that are not rel- be more “adult-like,” relative to when children’stop– evant to the current task (e.g., searching for oranges) but down search strategies and behavioral performance are related to the target (e.g., apples and oranges are re- resembles that of adults. lated because they are both fruits; Wu, Pruitt, et al., 2017; Here, we focus on work that is specifically centered on Wu et al., 2015; Nako, Wu, & Eimer, 2014; Nako, Wu, the N2pc. However, research investigating the interface Smith, et al., 2014; see also Telling, Kumar, Meyer, & between attention, memory, and learning holds addi- Humphreys, 2009). Wu, Pruitt, et al. (2017) showed that tional promise to developmental cognitive neuroscien- the costs of knowledge on search efficiency may be re- tists. For example, in terms of the dynamic interplay lated to how much experience one has with the search between memory and attention that is now so well re- objects. The study showed that people with increased searched in adults, our research team has begun to dieting experience had higher “costs” of knowledge re- demonstrate that similar dynamics are key to under- lated to healthy and unhealthy food categories compared standing how both children and adults use memory for with people with less dieting experience. The benefits and newly learned information to guide their attention

Wu et al. 1751 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn_a_01316 by guest on 23 September 2021 (Nussenbaum, Scerif, & Nobre, under review). These dren may have less procedural knowledge of how to in- memory- guided effects on attention are reflected in cor- teract with the experimental device itself, and their fine responding modulation of oscillatory activity in the alpha motor skills may still be somewhat more immature rela- range (Doherty, Fraser, Nobre, & Scerif, under review). tive to adults. Given a smaller working memory capacity Using magnetoencephalography, we also have discovered in younger children, they may also have a more difficult that the attentional state of distributed oscillatory net- time holding rule representations or task goals in mind, works during encoding into memory is a significant pre- even if they grasped them initially. One way to address dictor of accurate memory recall in children (Astle et al., this issue is by reminding them what the goal is on every 2015). Other techniques have also focused trial by providing a cue at the beginning of each trial as on the neural correlates of attention and memory inter- well as feedback at the end of each trial. However, a actions, albeit primarily in adults (e.g., Aly & Turk-Browne, drawback resulting from such an implementation is that 2016; Stokes et al., 2012; Summerfield et al., 2006). each trial duration and the overall task duration both in- As a whole, these findings point to the wealth of novel crease. This increase in task duration leads to young chil- empirical questions and findings that can emerge from dren either completing fewer trials than older children investigating the interplay between attention, learning, and adults due to fatigue but with the completed trials and memory in young learners. At the same time, our reflecting their true abilities or completing the same efforts have pinpointed key methodological consider- number of trials but with a higher proportion of errors ations that are important for any cognitive arising due to fatigue and inattention and not due to less embarking in the process of building bridges between developed abilities. The younger the children, the bigger more basic and applied/educational questions. this issue becomes, as children from 6 years of age and above seem to do well with “reminding” instructions and feedback at the beginning and end of brief blocks of tri- METHODOLOGICAL LIMITATIONS AND als, rather than at the beginning and end of each trial POTENTIAL SOLUTIONS WHEN INTEGRATING (when boredom from repetition might instead ensue). ADULT COGNITIVE NEUROSCIENCE AND In our pilot work, we optimized the adult paradigm for DEVELOPMENTAL PSYCHOLOGY children by using highly salient regular rewarding visual Beyond the potential promise of methods geared to as- stimuli (i.e., very clear feedback with a smiley