research paper on play based learning

Play-Based Learning: Evidence-Based Research to Improve Children’s Learning Experiences in the Kindergarten Classroom

• Published: 31 October 2019
• Volume 48 , pages 127–133, ( 2020 )

Cite this article

• Meaghan Elizabeth Taylor 1 , 2 &
• Wanda Boyer 1  

32k Accesses

44 Citations

3 Altmetric

Explore all metrics

With a heavy increase in academic expectations and standards to be learned in the early years, educators are facing the challenge of integrating important academic standards into developmentally appropriate learning experiences for children in kindergarten. To meet this challenge, there is a need to become familiar with the role of play in the classroom with an emphasis on developmentally appropriate practices such as play-based learning (PBL). PBL is child-centered and focuses on children’s academic, social, and emotional development, and their interests and abilities through engaging and developmentally appropriate learning experiences. This paper explores the definition of play-based learning (PBL), the theoretical frameworks and historical research that have shaped PBL, the different types of play, the social and academic benefits of PBL, and the ways in which educators can facilitate, support, assess, and employ technology to enhance PBL. The authors will conclude by reflecting on how teaching practices can be informed by evidence-based research to improve children’s learning experiences in the kindergarten classroom.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA) Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Rent this article via DeepDyve

Institutional subscriptions

Similar content being viewed by others

Play in Kindergarten: An Interview and Observational Study in Three Canadian Classrooms

Teaching play skills to children with disabilities: research-based interventions and practices.

Play-Based Learning Within the Early Years

Bodner, G. M. (1986). Constructivism: A theory of knowledge. Journal of Chemical Education, 63 (10), 873–878. https://doi.org/10.1021/ed063p873 .

Article   Google Scholar  

Bruce, T. (2012). The whole child. Play. In T. Bruce (Ed.), Early childhood practice: Froebel today (pp. 5–16). London, UK: SAGE Publication Ltd.

Chapter   Google Scholar  

Goldstein, L. S. (1997). Between a rock and a hard place in the primary grades: The challenge of providing developmentally appropriate early childhood education in an elementary school setting. Early Childhood Research Quarterly, 12 (1), 3–27. https://doi.org/10.1016/s0885-2006(97)90039-9 .

Hoskins, K., & Smedley, S. (2019). Protecting and extending Froebelian principles in practice: Exploring the importance of learning through play. Journal of Early Childhood Research, 17 (2), 73–87. https://doi.org/10.1177/1476718X18809114 .

MacKay, S. H. (2019). Teaching and learning: Starting with story workshop. Retrieved from https://opalschool.org/starting-with-story-workshop/

Miller, T. (2018). Developing numeracy skills using interactive technology in a play-based learning environment. International Journal of STEM Education, 5 (1), 39. https://doi.org/10.1186/s40594-018-0135-2 .

Montessori, M. (1984). The absorbent mind . A laurel book New York, NY: Dell Publishing Company.

Google Scholar  

Mukherji, P., & Albon, D. (2018). Research methods in early childhood. An introductory guide (3rd ed.). Thousand Oaks, CA: SAGE.

Nicol, J., & Taplin, J. T. (2018). Understanding the Steiner Waldorf Approach. Early years education in practice (2nd ed.). London: Routledge. https://doi.org/10.4324/9781315456577 .

Book   Google Scholar  

Nolan, A., & Paatsch, L. (2017). (Re)affirming identities: Implementing a play-based approach to learning in the early years of schooling. International Journal of Early Years Education, 26 (1), 42–55. https://doi.org/10.1080/09669760.2017.1369397 .

Piaget, J. (1962). Play, dreams and imitation in childhood (C. Gattegno & F. M. Hodgson, Trans.). New York, NY: Norton & Company Inc.

Pyle, A., & Alaca, B. (2018). Kindergarten children’s perspectives on play and learning. Early Child Development and Care, 188 (8), 1063–1075. https://doi.org/10.1080/03004430.2016.1245190 .

Pyle, A., & Bigelow, A. (2014). Play in kindergarten: An interview and observational study in three Canadian classrooms. Early Childhood Education Journal, 43 (5), 385–393. https://doi.org/10.1007/s10643-014-0666-1 .

Pyle, A., & Danniels, E. (2017). A continuum of play-based learning: The role of the teacher in play-based pedagogy and the fear of hijacking play. Early Education and Development, 28 (3), 274–289. https://doi.org/10.1080/10409289.2016.1220771 .

Pyle, A., & DeLuca, C. (2017). Assessment in play-based kindergarten classrooms: An empirical study of teacher perspectives and practices. The Journal of Educational Research, 110 (5), 457–466. https://doi.org/10.1080/00220671.2015.1118005 .

Steiner, R. (1996). Lectures One -Eight by Rudolph Steiner Sunday April 15–Sunday April 22, 1923 (R. Everett & R. Everett Trans. & Ed.). The child’s changing consciousness: As the basis of pedagogical practice . Hudson, NY: Anthroposophic Press. Retrieved from https://steiner.presswarehouse.com/sites/steiner/research/archive/childs_changing_consciousness/childs_changing_consciousness.pdf

Vogt, F., Hauser, B., Stebler, R., Rechsteiner, K., & Urech, C. (2018). Learning through play—Pedagogy and learning outcomes in early childhood mathematics. European Early Childhood Education Research Journal, 26 (4), 589–603. https://doi.org/10.1080/1350293X.2018.1487160 .

Vukelich, C. (1994). Effects of play interventions on young children’s reading of environmental print. Early Childhood Research Quarterly, 9 (2), 153–170. https://doi.org/10.1016/0885-2006(94)90003-5 .

Vygotsky, L. S. (1966). The role of play in development. In the published work by Lev Semyonovich Vygotsky entitled: Problems of psychology. In M. Cole, V. John-Steiner, S. Scribner, & E. Souberman (Eds.), Mind in Society. The Development of Higher Psychological Processes (1979, pp. 92–104). Retrieved from http://ouleft.org/wp-content/uploads/Vygotsky-Mind-in-Society.pdf (Original Harvard University Press 1979)

Vygotsky, L. (1967). Play and its role in the mental development of the child. Soviet Psychology, 5 (3), 6–18. https://doi.org/10.2753/RPO1061-040505036 .

Download references

Author information

Authors and affiliations.

Department of Educational Psychology and Leadership Studies, University of Victoria, 3800 Finnerty Road, Victoria, P.O. Box 1700, STN CSC, BC, V8W-2YC, Canada

Meaghan Elizabeth Taylor & Wanda Boyer

Sooke School District #62, Victoria, BC, Canada

Meaghan Elizabeth Taylor

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Wanda Boyer .

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and permissions

About this article

Taylor, M.E., Boyer, W. Play-Based Learning: Evidence-Based Research to Improve Children’s Learning Experiences in the Kindergarten Classroom. Early Childhood Educ J 48 , 127–133 (2020). https://doi.org/10.1007/s10643-019-00989-7

Download citation

Published : 31 October 2019

Issue Date : March 2020

DOI : https://doi.org/10.1007/s10643-019-00989-7

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Play-based Learning
• Play-based assessment
• Developmentally appropriate practice
• Kindergarten
• Find a journal
• Publish with us
• Track your research

• NAEYC Login
• Member Profile
• Hello Community
• Accreditation Portal
• Online Learning
• Online Store

Popular Searches:   DAP ;  Coping with COVID-19 ;  E-books ;  Anti-Bias Education ;  Online Store

The Power of Playful Learning in the Early Childhood Setting

You are here

Play versus learning represents a false dichotomy in education (e.g., Hirsh-Pasek & Golinkoff 2008). In part, the persistent belief that learning must be rigid and teacher directed—the opposite of play—is motivated by the lack of a clear definition of what constitutes playful learning (Zosh et al. 2018). And, in part, it is motivated by older perceptions of play and learning. Newer research, however, allows us to reframe the debate as learning via play—as playful learning.

This piece, which is an excerpt from Chapter 5 in  Developmentally Appropriate Practice in Early Childhood Programs Serving Children from Birth Through Age 8, Fourth Edition (NAEYC 2022), suggests that defining play on a spectrum (Zosh et al. 2018, an idea first introduced by Bergen 1988) helps to resolve old divisions and provides a powerful framework that puts  playful learning —rich curriculum coupled with a playful pedagogy—front and center as a model for all early childhood educators. ( See below for a discussion of play on a spectrum.)

This excerpt also illustrates the ways in which play and learning mutually support one another and how teachers connect learning goals to children’s play. Whether solitary, dramatic, parallel, social, cooperative, onlooker, object, fantasy, physical, constructive, or games with rules, play, in all of its forms, is a teaching practice that optimally facilitates young children’s development and learning. By maximizing children’s choice, promoting wonder and enthusiasm for learning, and leveraging joy, playful learning pedagogies support development across domains and content areas and increase learning relative to more didactic methods (Alfieri et al. 2011; Bonawitz et al. 2011; Sim & Xu 2015).

Playful Learning: A Powerful Teaching Tool

This narrowing of the curriculum and high-stakes assessment practices (such as paper-and-pencil tests for kindergartners) increased stress on educators, children, and families but failed to deliver on the promise of narrowing—let alone closing—the gap.  All  children need well-thought-out curricula, including reading and STEM experiences and an emphasis on executive function skills such as attention, impulse control, and memory (Duncan et al. 2007). But to promote happy, successful, lifelong learners, children must be immersed in developmentally appropriate practice and rich curricular learning that is culturally relevant (NAEYC 2020). Playful learning is a vehicle for achieving this. Schools must also address the inequitable access to play afforded to children (see “Both/And: Early Childhood Education Needs Both Play and Equity,” by Ijumaa Jordan.) All children should be afforded opportunities to play, regardless of their racial group, socioeconomic class, and disability if they have been diagnosed with one. We second the call of Maria Souto-Manning (2017): “Although play has traditionally been positioned as a privilege, it must be (re)positioned as a right, as outlined by the  United Nations Convention on the Rights of the Child, Article 31” (785).

What Is Playful Learning?

Playful learning describes a learning context in which children learn content while playing freely (free play or self-directed play), with teacher guidance (guided play), or in a structured game. By harnessing children’s natural curiosity and their proclivities to experiment, explore, problem solve, and stay engaged in meaningful activities—especially when doing so with others—teachers maximize learning while individualizing learning goals. Central to this concept is the idea that teachers act more as the Socratic “guide at the side” than a “sage on the stage” (e.g., King 1993, 30; Smith 1993, 35). Rather than view children as empty vessels receiving information, teachers see children as active explorers and discoverers who bring their prior knowledge into the learning experience and construct an understanding of, for example, words such as  forecast  and  low pressure  as they explore weather patterns and the science behind them. In other words, teachers support children as active learners.

Importantly, playful learning pedagogies naturally align with the characteristics that research in the science of learning suggests help humans learn. Playful learning leverages the power of active (minds-on), engaging (not distracting), meaningful, socially interactive, and iterative thinking and learning (Zosh et al. 2018) in powerful ways that lead to increased learning.

Free play lets children explore and express themselves—to be the captains of their own ship. While free play is important, if a teacher has a learning goal, guided play and games are the road to successful outcomes for children (see Weisberg, Hirsh-Pasek, & Golinkoff 2013 for a review). Playful learning in the form of guided play, in which the teacher builds in the learning as part of a fun context such as a weather report, keeps the child’s agency but adds an intentional component to the play that helps children learn more from the experience. In fact, when researchers compared children’s skill development during free play in comparison to guided play, they found that children learned more vocabulary (Toub et al. 2018) and spatial skills (Fisher et al. 2013) in guided play than in free play.

Self-Directed Play, Free Play

NAEYC’s 2020 position statement on developmentally appropriate practice uses the term  self-directed play  to refer to play that is initiated and directed by children. Such play is termed  free play  in the larger works of the authors of this excerpt; therefore, free play is the primary term used in this article, with occasional references to self-directed play, the term used in the rest of the DAP book.

Imagine an everyday block corner. The children are immersed in play with each other—some trying to build high towers and others creating a tunnel for the small toy cars on the nearby shelves. But what if there were a few model pictures on the wall of what children could strive to make as they collaborated in that block corner? Might they rotate certain pieces purposely? Might they communicate with one another that the rectangle needs to go on top of the square? Again, a simple insertion of a design that children can try to copy turns a play situation into one ripe with spatial learning. Play is a particularly effective way to engage children with specific content learning when there is a learning goal.

Why Playful Learning Is Critical

Teachers play a crucial role in creating places and spaces where they can introduce playful learning to help all children master not only content but also the skills they will need for future success. The science of learning literature (e.g., Fisher et al. 2013; Weisberg, Hirsh-Pasek, & Golinkoff 2013; Zosh et al. 2018) suggests that playful learning can change the “old equation” for learning, which posited that direct, teacher-led instruction, such as lectures and worksheets, was the way to achieve rich content learning. This “new equation” moves beyond a sole focus on content and instead views playful learning as a way to support a breadth of skills while embracing developmentally appropriate practice guidelines (see Hirsh-Pasek et al. 2020).

Using a playful learning pedagogical approach leverages the skill sets of today’s educators and enhances their ability to help children attain curricular goals. It engages what has been termed active learning that is also developmentally appropriate and offers a more equitable way of engaging children by increasing access to participation. When topics are important and culturally relevant to children, they can better identify with the subject and the learning becomes more seamless.

While educators of younger children are already well versed in creating playful and joyful experiences to support social goals (e.g., taking turns and resolving conflicts), they can use this same skill set to support more content-focused curricular goals (e.g., mathematics and literacy). Similarly, while teachers of older children have plenty of experience determining concrete content-based learning goals (e.g., attaining Common Core Standards), they can build upon this set of skills and use playful learning as a pedagogy to meet those goals.

Learning Through Play: A Play Spectrum

As noted previously, play can be thought of as lying on a spectrum that includes free play (or self-directed play), guided play, games, playful instruction, and direct instruction (Bergen 1988; Zosh et al. 2018). For the purposes of this piece, we use a spectrum that includes the first three of these aspects of playful learning, as illustrated in “Play Spectrum Showing Three Types of Playful Learning Situations” below.

The following variables determine the degree to which an activity can be considered playful learning:

• level of adult involvement
• extent to which the child is directing the learning
• presence of a learning goal

Toward the left end of the spectrum are activities with more child agency, less adult involvement, and loosely defined or no particular learning goals. Further to the right, adults are more involved, but children still direct the activity or interaction.

Developmentally appropriate practice does not mean primarily that children play without a planned learning environment or learn mostly through direct instruction (NAEYC 2020). Educators in high-quality early childhood programs offer a range of learning experiences that fall all along this spectrum. By thinking of play as a spectrum, educators can more easily assess where their learning activities and lessons fall on this spectrum by considering the components and intentions of the lesson. Using their professional knowledge of how children develop and learn, their knowledge of individual children, and their understanding of social and cultural contexts, educators can then begin to think strategically about how to target playful learning (especially guided play and games) to leverage how children naturally learn. This more nuanced view of play and playful learning can be used to both meet age-appropriate learning objectives and support engaged, meaningful learning.   

In the kindergarten classroom in the following vignette, children have ample time for play and exploration in centers, where they decide what to play with and what they want to create. These play centers are the focus of the room and the main tool for developing social and emotional as well as academic skills; they reflect and support what the children are learning through whole-group discussions, lessons, and skills-focused stations. In the vignette, the teacher embeds guided play opportunities within the children’s free play.

Studying Bears: Self-Directed Play that Extends What Kindergartners Are Learning

While studying the habits of animals in winter, the class is taking a deeper dive into the lives of American black bears, animals that make their homes in their region. In the block center, one small group of children uses short lengths and cross-sections of real tree branches as blocks along with construction paper to create a forest habitat for black bear figurines. They enlist their friends in the art center to assist in making trees and bushes. Two children are in the writing center. Hearing that their friends are looking for help to create a habitat, they look around and decide a hole punch and blue paper are the perfect tools for making blueberries—a snack black bears love to eat! Now multiple centers and groups of children are involved in making the block center become a black bear habitat.

In the dramatic play center, some of the children pretend to be bear biologists, using stethoscopes, scales, and magnifying glasses to study the health of a couple of plush black bears. When these checkups are complete, the teacher suggests the children could describe the bears’ health in a written “report,” thus embedding guided play within their free play. A few children at the easels in the art center are painting pictures of black bears.

Contributed by Amy Blessing

Free play, or self-directed play, is often heralded as the gold standard of play. It encourages children’s initiative, independence, and problem solving and has been linked to benefits in social and emotional development (e.g., Singer & Singer 1990; Pagani et al. 2010; Romano et al. 2010; Gray 2013) and language and literacy (e.g., Neuman & Roskos 1992). Through play, children explore and make sense of their world, develop imaginative and symbolic thinking, and develop physical competence. The kindergarten children in the example above were developing their fine motor and collaboration skills, displaying their understanding of science concepts (such as the needs of animals and living things), and exercising their literacy and writing skills. Such benefits are precisely why free play has an important role in developmentally appropriate practice. To maximize learning, teachers also provide guided play experiences.

Guided Play

While free play has great value for children, empirical evidence suggests that it is not always sufficient  when there is a pedagogical goal at stake  (Smith & Pellegrini 2008; Alfieri et al. 2011; Fisher et al. 2013; Lillard 2013; Weisberg, Hirsh-Pasek, & Golinkoff 2013; Toub et al. 2018). This is where guided play comes in.

Guided play allows teachers to focus children’s play around specific learning goals (e.g., standards-based goals), which can be applied to a variety of topics, from learning place value in math to identifying rhyming words in literacy activities. Note, however, that the teacher does not take over the play activity or even direct it. Instead, she asks probing questions that guide the next level of child-directed exploration. This is a perfect example of how a teacher can initiate a context for learning while still leaving the child in charge. In the previous kindergarten vignette, the teacher guided the children in developing their literacy skills as she embedded writing activities within the free play at the centers.

Facilitating Guided Play

Skilled teachers set up environments and facilitate development and learning throughout the early childhood years, such as in the following:

• Ms. Taglieri notices what 4-month-old Anthony looks at and shows interest in. Following his interest and attention, she plays Peekaboo, adjusting her actions (where she places the blanket and peeks out at him) to maintain engagement.
• Ms. Eberhard notices that 22-month-old Abe knows the color yellow. She prepares her environment based on this observation, placing a few yellow objects along with a few red ones on a small table. Abe immediately goes to the table, picking up each yellow item and verbally labeling them (“Lellow!”).
• Mr. Gorga creates intrigue and participation by inviting his preschool class to “be shape detectives” and to “discover the secret of shapes.” As the children explore the shapes, Mr. Gorga offers questions and prompts to guide children to answer the question “What makes them the same kind of shapes?”

An analogy for facilitating guided play is bumper bowling. If bumpers are in place, most children are more likely than not to knock down some pins when they throw the ball down the lane. That is different than teaching children exactly how to throw it (although some children, such as those who have disabilities or who become frustrated if they feel a challenge is too great, may require that level of support or instruction). Guided play is not a one-size-fits-all prescriptive pedagogical technique. Instead, teachers match the level of support they give in guided play to the children in front of them.

Critically, many teachers already implement these kinds of playful activities. When the children are excited by the birds they have seen outside of their window for the past couple of days, the teachers may capitalize on this interest and provide children with materials for a set of playful activities about bird names, diets, habitats, and songs. Asking children to use their hands to mimic an elephant’s trunk when learning vocabulary can promote learning through playful instruction that involves movement. Similarly, embedding vocabulary in stories that are culturally relevant promotes language and early literacy development (García-Alvarado, Arreguín, & Ruiz-Escalante 2020). For example, a teacher who has several children in his class with Mexican heritage decides to read aloud  Too Many Tamales  (by Gary Soto, illus. Ed Martinez) and have the children reenact scenes from it, learning about different literary themes and concepts through play. The children learn more vocabulary, have a better comprehension of the text, and see themselves and their experiences reflected. The teacher also adds some of the ingredients and props for making tamales into the sociodramatic play center (Salinas-González, Arreguín-Anderson, & Alanís 2018) and invites families to share stories about family  tamaladas  (tamale-making parties).

Evidence Supporting Guided Play as a Powerful Pedagogical Tool

Evidence from the science of learning suggests that discovery-based guided play actually results in increased learning for all children relative to both free play and direct instruction (see Alferi et al. 2011). These effects hold across content areas including spatial learning (Fisher et al. 2013), literacy (Han et al. 2010; Nicolopoulou et al. 2015; Hassinger-Das et al. 2016; Cavanaugh et al. 2017; Toub et al. 2018; Moedt & Holmes 2020), and mathematics (Zosh et al. 2016).

There are several possible reasons for guided play’s effectiveness. First, it harnesses the joy that is critical to creativity and learning (e.g., Isen, Daubman, & Nowicki 1987; Resnick 2007). Second, during guided play, the adults help “set the stage for thought and action” by essentially limiting the number of possible outcomes for the children so that the learning goal is discoverable, but children still direct the activity (Weisberg et al. 2014, 276). Teachers work to provide high-quality materials, eliminate distractions, and prepare the space, but then, critically, they let the child play the active role of construction. Third, in guided play, the teacher points the way toward a positive outcome and hence lessens the ambiguity (the degrees of freedom) without directing children to an answer or limiting children to a single discovery (e.g., Bonawitz et al. 2011). And finally, guided play provides the opportunity for new information to be integrated with existing knowledge and updated as children explore.

Reinforcing Numeracy with a Game

The children in Mr. Cohen’s preschool class are at varying levels of understanding in early numeracy skills (e.g., cardinality, one-to-one correspondence, order irrelevance). He knows that his children need some practice with these skills but wants to make the experience joyful while also building these foundational skills. One day, he brings out a new game for them to play—The Great Race. Carla and Michael look up expectantly, and their faces light up when they realize they will be playing a game instead of completing a worksheet. The two quickly pull out the box, setting up the board and choosing their game pieces. Michael begins by flicking the spinner with his finger, landing on 2. “Nice!” Carla exclaims, as Michael moves his game piece, counting “One, two.” Carla takes a turn next, spinning a 1 and promptly counting “one” as she moves her piece one space ahead. “My turn!” Michael says, eager to win the race. As he spins a 2, he pauses. “One . . . two,” he says, hesitating, as he moves his piece to space 4 on the board. Carla corrects him, “I think you mean ‘three, four,’ right? You have to count up from where you are on the board.” Michael nods, remembering the rules Mr. Cohen taught him earlier that day. “Right,” he says, “three, four.”

Similar to guided play, games can be designed in ways that help support learning goals (Hassinger-Das et al. 2017). In this case, instead of adults playing the role of curating the activity, the games themselves provide this type of external scaffolding. The example with Michael and Carla shows how children can learn through games, which is supported by research. In one well-known study, playing a board game (i.e., The Great Race) in which children navigated through a linear, numerical-based game board (i.e., the game board had equally spaced game spaces that go from left to right) resulted in increased numerical development as compared to playing the same game where the numbers were replaced by colors (Siegler & Ramani 2008) or with numbers organized in a circular fashion (Siegler & Ramani 2009). Structuring experiences so that the learning goal is intertwined naturally with children’s play supports their learning. A critical point with both guided play and games is that children are provided with support but still lead their own learning.

Digital educational games have become enormously popular, with tens of thousands of apps marketed as “educational,” although there is no independent review of these apps. Apps and digital games may have educational value when they inspire active, engaged, meaningful, and socially interactive experiences (Hirsh-Pasek et al. 2015), but recent research suggests that many of the most downloaded educational apps do not actually align with these characteristics that lead to learning (Meyer et al. 2021). Teachers should exercise caution and evaluate any activity—digital or not—to see how well it harnesses the power of playful learning.

Next Steps for Educators

Educators are uniquely positioned to prepare today’s children for achievement today and success tomorrow. Further, the evidence is mounting that playful pedagogies appear to be an accessible, powerful tool that harnesses the pillars of learning. This approach can be used across ages and is effective in learning across domains.

By leveraging children’s own interests and mindfully creating activities that let children play their way to new understanding and skills, educators can start using this powerful approach today. By harnessing the children’s interests at different ages and engaging them in playful learning activities, educators can help children learn while having fun. And, importantly, educators will have more fun too when they see children happy and engaged.

As the tide begins to change in individual classrooms, educators need to acknowledge that vast inequalities (e.g., socioeconomic achievement gaps) continue to exist (Kearney & Levine 2016). The larger challenge remains in propelling a cultural shift so that administrators, families, and policymakers understand the way in which educators can support the success of all children through high-quality, playful learning experiences.

Consider the following reflection questions as you reflect how to support equitable playful learning experiences for each and every child:

• One of the best places to start is by thinking about your teaching strengths. Perhaps you are great at sparking joy and engagement. Or maybe you are able to frequently leverage children’s home lives in your lessons. How can you expand practices you already use as an educator or are learning about in your courses to incorporate the playful learning described in this article?
• How can you share the information in this chapter with families, administrators, and other educators? How can you help them understand how play can engage children in deep, joyful learning?

This piece is excerpted from NAEYC’s recently published book  Developmentally Appropriate Practice in Early Childhood Programs Serving Children from Birth Through Age 8,  Fourth Edition. For more information about the book, visit  NAEYC.org/resources/pubs/books/dap-fourth-edition .

Teaching Play Skills

Pamela Brillante

While many young children with autism spectrum disorder enjoy playing, they can have difficulty engaging in traditional play activities. They may engage in activities that do not look like ordinary play, including playing with only a few specific toys or playing in a specific, repetitive way.

Even though most children learn play skills naturally, sometimes families and teachers have to teach children how to play. Learning how to play will help develop many other skills young children need for the future, including

• social skills:  taking turns, sharing, and working cooperatively
• cognitive skills:  problem-solving skills, early academic skills
• communication skills:  responding to others, asking questions
• physical skills:  body awareness, fine and gross motor coordination

Several evidence-based therapeutic approaches to teaching young children with autism focus on teaching play skills, including

• The Play Project:  https://playproject.org
• The Greenspan Floortime approach: https://stanleygreenspan.com
• Integrated Play Group (IPG) Model: www.wolfberg.com

While many children with autism have professionals and therapists working with them, teachers and families should work collaboratively and provide multiple opportunities for children to practice new skills and engage in play at their own level. For example, focus on simple activities that promote engagement between the adult and the child as well as the child and their peers without disabilities, including playing with things such as bubbles, cause-and-effect toys, and interactive books. You can also use the child’s preferred toy in the play, like having the Spider-Man figure be the one popping the bubbles.

Pamela Brillante , EdD, has spent 30 years working as a special education teacher, administrator, consultant, and professor. In addition to her full-time faculty position in the Department of Special Education, Professional Counseling and Disability Studies at William Paterson University of New Jersey, Dr. Brillante continues to consult with school districts and present to teachers and families on the topic of high-quality, inclusive early childhood practices.  

Photographs: © Getty Images Copyright © 2022 by the National Association for the Education of Young Children. See Permissions and Reprints online at  NAEYC.org/resources/permissions .

Alfieri, L., P.J. Brooks, N.J. Aldrich, & H.R. Tenenbaum. 2011. “Does Discovery-Based Instruction Enhance Learning?” Journal of Educational Psychology 103 (1): 1–18.  

Bassok, D., S. Latham, & A. Rorem. 2016. “Is Kindergarten the New First Grade?” AERA Open 2 (1): 1–31. doi.10.1177/2332858415616358. 

Bergen, D., ed. 1988. Play as a Medium for Learning and Development: A Handbook of Theory and Practice . Portsmouth, NH: Heinemann Educational Books. 

Bonawitz, E.B., P. Shafto, H. Gweon, N.D. Goodman, E.S. Spelke, & L. Schulz. 2011. “The Double-Edged Sword of Pedagogy: Instruction Limits Spontaneous Exploration and Discovery.” Cognition 120 (3): 322–30. 

Cavanaugh, D.M., K.J. Clemence, M.M. Teale, A.C. Rule, & S.E. Montgomery. 2017. “Kindergarten Scores, Storytelling, Executive Function, and Motivation Improved Through Literacy-Rich Guided Play.” Journal of Early Childhood Education 45 (6): 1–13. 

Christakis, E. 2016. The Importance of Being Little: What Preschoolers Really Need from Grownups . New York: Penguin Books.  

Duncan, G. J., A. Claessens, A.C. Huston, L.S. Pagani, M. Engel, H. Sexton, C.J. Dowsett, K. Magnuson, P. Klebanov, L. Feinstein, J. Brooks-Gunn, K. Duckworth, & C. Japel. 2007. “School Readiness and Later Achievement.” Developmental Psychology 43 (6): 1428–46. https://doi.apa.org/doi/10.1037/0012-1649.43.6.1428 .  

Fisher, K.R., K. Hirsh-Pasek, N. Newcombe, & R.M. Golinkoff. 2013. “Taking Shape: Supporting Preschoolers’ Acquisition of Geometric Knowledge Through Guided Play.” Child Development 84 (6): 1872–78. 

García-Alvarado, S., M.G. Arreguín, & J.A. Ruiz-Escalante. 2020. “Mexican-American Preschoolers as Co-Creators of Zones of Proximal Development During Retellings of Culturally Relevant Stories: A Participatory Study.” Journal of Early Childhood Literacy : 1–22. https://doi.org/10.1177%2F1468798420930339 . 

Gray, P. 2013. Free to Learn: Why Unleashing the Instinct to Play Will Make Our Children Happier, More Self-Reliant, and Better Students for Life . New York: Basic Books.  

Han, M., N. Moore, C. Vukelich, & M. Buell. 2010. “Does Play Make a Difference? How Play Intervention Affects the Vocabulary Learning of At-Risk Preschoolers.” American Journal of Play 3 (1): 82–105. 

Hannaway, J., & L. Hamilton. 2008. Accountability Policies: Implications for School and Classroom Practices . Washington, DC: Urban Institute. http://webarchive.urban.org/publications/411779.html . 

Hassinger-Das, B., K. Ridge, A. Parker, R.M. Golinkoff, K. Hirsh-Pasek, & D.K. Dickinson. 2016. “Building Vocabulary Knowledge in Preschoolers Through Shared Book Reading and Gameplay.” Mind, Brain, and Education 10 (2): 71–80. https://doi.org/10.1111/mbe.12103 . 

Hassinger-Das, B., T.S. Toub, J.M. Zosh, J. Michnick, R. Golinkoff, & K. Hirsh-Pasek. 2017. “More Than Just Fun: A Place for Games in Playful Learning.” Infancia y aprendizaje: Journal for the Study of Education and Development 40 (2): 191–218. https://doi.org/10.1080/02103702.2017.1292684 . 

Hirsh-Pasek, K., & R.M. Golinkoff. 2008. “Why Play = Learning.” In Encyclopedia on Early Childhood Development [online], eds. R.E. Tremblay, M. Boivin, & R.D. Peters, topic ed. P.K. Smith, 1–6. Centre of Excellence for Early Childhood Development and Strategic Knowledge Cluster on Early Child Development. www.child-encyclopedia.com/play/according-experts/why-play-learning . 

Hirsh-Pasek, K., H. S. Hadani, E. Blinkoff, & R. M. Golinkoff. 2020. A new path to education reform: Playful learning promotes 21st-century skills in schools and beyond . The Brookings Institution: Big Ideas Policy Report. www.brookings.edu/policy2020/bigideas/a-new-path-to-education-reform-playful-learning-promotes-21st-century-skills-in-schools-and-beyond . 

Hirsh-Pasek, K., J.M. Zosh, R.M. Golinkoff, J.H. Gray, M.B. Robb, & J. Kaufman. 2015. “Putting Education in ‘Educational’ Apps: Lessons from the Science of Learning.” Psychological Science in the Public Interest 16 (1): 3–34. 

Isen, A.M., K.A. Daubman, & G.P. Nowicki. 1987. “Positive Affect Facilitates Creative Problem Solving.” Journal of Personality and Social Psychology 52 (6): 1122–31. 

Kearney, M.S., & P.B. Levine. (2016, Spring). Income, Inequality, Social Mobility, and the Decision to Drop Out of High School . Washington, DC: Brookings. www.brookings.edu/bpea-articles/income-inequality-social-mobility-and-the-decision-to-drop-out-of-high-school . 

King, A. 1993. “From Sage on the Stage to Guide on the Side.” College Teaching 41 (1): 30–35.  

Lillard, A.S. 2013. “Playful Learning and Montessori Education.” American Journal of Play 5 (2): 157–86. 

Meyer, M., J.M. Zosh, C. McLaren, M. Robb, R.M. Golinkoff, K. Hirsh-Pasek, & J. Radesky. 2021. “How Educational Are ‘Educational’ Apps for Young Children? App Store Content Analysis Using the Four Pillars of Learning Framework.” Journal of Children and Media . Published online February 23. 

Miller, E., & J. Almon. 2009. Crisis in the Kindergarten: Why Children Need to Play in School . College Park, MD: Alliance for Childhood. https:// files.eric.ed.gov/fulltext/ED504839.pdf . 

Moedt, K., & R.M. Holmes. 2020. “The Effects of Purposeful Play After Shared Storybook Readings on Kindergarten Children’s Reading Comprehension, Creativity, and Language Skills and Abilities.” Early Child Development and Care 190 (6): 839–54. 

NAEYC. 2020. “Developmentally Appropriate Practice.” Position statement. Washington, DC: NAEYC. www.naeyc.org/resources/position-statements/dap .  

Neuman, S.B., & K. Roskos. 1992. “Literacy Objects as Cultural Tools: Effects on Children’s Literacy Behaviors in Play.” Reading Research Quarterly 27 (3): 202–25.  

Nicolopoulou, A., K.S. Cortina, H. Ilgaz, C.B. Cates, & A.B. de Sá. 2015. “Using a Narrative- and Play-Based Activity to Promote Low-Income Preschoolers’ Oral Language, Emergent Literacy, and Social Competence.” Early Childhood Research Quarterly 31 (2): 147–62. 

Pagani, L.S., C. Fitzpatrick, I. Archambault, & M. Janosz. 2010. “School Readiness and Later Achievement: A French Canadian Replication and Extension.” Developmental Psychology 46 (5): 984–94.  

Pedulla, J.J., L.M. Abrams, G.F. Madaus, M.K. Russell, M.A. Ramos, & J. Miao. 2003. “Perceived Effect of State-Mandated Testing Programs on Teaching and Learning: Findings from a National Survey of Teachers” (ED481836). ERIC. https://eric.ed.gov/?id=ED481836 . 

Ravitch, D. 2010. “Why Public Schools Need Democratic Governance.” Phi Delta Kappan 91 (6): 24–27. 

Resnick, M. 2007. “All I Really Need to Know (About Creative Thinking) I Learned (by Studying How Children Learn) in Kindergarten.” In Proceedings of the 6th ACM SIGCHI Conference on Creativity & Cognition , 1–6. New York: Association for Computing Machinery. 

Romano, E., L. Babchishin, L.S. Pagani, & D. Kohen. 2010. “School Readiness and Later Achievement: Replication and Extension Using a Nationwide Canadian Survey.” Developmental Psychology 46 (5): 995–1007.  

Salinas-González, I., M.G. Arreguín-Anderson, & I. Alanís. 2018. “Supporting Language: Culturally Rich Dramatic Play.” Teaching Young Children 11 (2): 4–6. 

Siegler, R.S., & G.B. Ramani. 2008. “Playing Linear Numerical Board Games Promotes Low-Income Children’s Numerical Development.” Developmental Science 11 (5): 655–61. 

Siegler, R.S., & G.B. Ramani. 2009. “Playing Linear Number Board Games—but Not Circular Ones—Improves Low-Income Preschoolers’ Numerical Understanding. Journal of Educational Psychology 101 (3): 545–60. 

Sim, Z., & F. Xu. 2015. “Toddlers Learn from Facilitated Play, Not Free Play.” In Proceedings of the 34th Annual Conference of the Cognitive Science Society , Berkeley, CA. https://cognitivesciencesociety.org/past-conferences . 

Singer, D.G., & J.L. Singer. 1990. The House of Make-Believe: Children’s Play and the Developing Imagination . Cambridge, MA: Harvard University Press.  

Smith, K. 1993. “Becoming the ‘Guide on the Side.’” Educational Leadership 51 (2): 35–37.  

Smith P.K., & A. Pellegrini. 2008. “Learning Through Play.” In Encyclopedia on Early Childhood Development [online], eds. R.E. Tremblay, M. Boivin, & R.D. Peters, 1–6. Centre of Excellence for Early Childhood Development and Strategic Knowledge Cluster on Early Child Development. https://www.child-encyclopedia.com/pdf/expert/play/according-experts/learning-through-play . 

Souto-Manning, M. 2017. “Is Play a Privilege or a Right? And What’s Our Responsibility? On the Role of Play for Equity in Early Childhood Education.” Foreword. Early Child Development and Care 187 (5–6): 785–87. www.tandfonline.com/doi/full/10.1080/03004430.2016.1266588 . 

Toub, T.S., B. Hassinger-Das, K.T. Nesbitt, H. Ilgaz, D.S. Weisberg, K. Hirsh-Pasek, R.M. Golinkoff, A. Nicolopoulou, & D.K. Dickinson. 2018. “The Language of Play: Developing Preschool Vocabulary Through Play Following Shared Book-Reading.” Early Childhood Research Quarterly 45 (4): 1–17.  

Weisberg, D.S., K. Hirsh-Pasek, & R.M. Golinkoff. 2013. “Guided Play: Where Curricular Goals Meet a Playful Pedagogy.” Mind, Brain, and Education 7 (2): 104–12. 

Weisberg, D.S., K. Hirsh-Pasek, R.M. Golinkoff, & B.D. McCandliss. 2014. “Mise en place: Setting the Stage for Thought and Action.” Trends in Cognitive Science 18 (6): 276–78. 

Zosh, J.M., B. Hassinger-Das, T.S. Toub, K. Hirsh-Pasek, & R. Golinkoff. 2016. “Playing with Mathematics: How Play Supports Learning and the Common Core State Standards.” Journal of Mathematics Education at Teachers College 7 (1): 45–49. https://doi.org/10.7916/jmetc.v7i1.787 . 

Zosh, J.M., K. Hirsh-Pasek, E.J. Hopkins, H. Jensen, C. Liu, D. Neale, S.L. Solis, & D. Whitebread. 2018. “Accessing the Inaccessible: Redefining Play as a Spectrum.” Frontiers in Psychology 9: 1–12. https://doi.org/10.3389/fpsyg.2018.01124 . 

Jennifer M. Zosh, PhD, is professor of human development and family studies at Penn State Brandywine. Most recently, her work has focused on technology and its impact on children as well as playful learning as a powerful pedagogy. She publishes journal articles, book chapters, blogs, and white papers and focuses on the dissemination of developmental research.

Caroline Gaudreau, PhD, is a research professional at the TMW Center for Early Learning + Public Health at the University of Chicago. She received her PhD from the University of Delaware, where she studied how children learn to ask questions and interact with screen media. She is passionate about disseminating research and interventions to families across the country.

Roberta Michnick Golinkoff, PhD, conducts research on language development, the benefits of play, spatial learning, and the effects of media on children. A member of the National Academy of Education, she is a cofounder of Playful Learning Landscapes, Learning Science Exchange, and the Ultimate Playbook for Reimagining Education. Her last book, Becoming Brilliant: What Science Tells Us About Raising Successful Children (American Psychological Association, 2016), reached the New York Times bestseller list.

Kathy Hirsh-Pasek, PhD, is the Lefkowitz Faculty Fellow in the Psychology and Neuroscience department at Temple University in Philadelphia, Pennsylvania.  She is also a senior fellow at the Brookings Institution. Her research examines the development of early language and literacy, the role of play in learning, and learning and technology. [email protected]

Vol. 77, No. 2

Print this article

Learning through play: what the science says

Do you know that play builds brains? We look at research that shows how playful learning experiences lay the foundations for brain development and develops 21st century skills

Our brains literally change as we learn

In a sense, what we do is who we are. We now have a large – and growing – pool of evidence to show that learning through play is the best way to support learning. Children are natural scientists – they come into the world ready to experiment and learn through play. And they use what they discover to not only adapt the structure of their brains, but also strengthen the skills they need to continue being engaged, flexible learners for their whole lives.

The evidence keeps mounting that play is the best way for children to learn – and thrive

From our earliest days, play is how we relate to the world, and to each other. When children have plenty of opportunities to learn playfully, they do what they do best: pursue their natural curiosity. And, as they do, they build skills and aptitudes they’ll keep for life. There’s a wealth of science behind our understanding of learning through play: studies in teaching and learning, play, and neuroscience. Here are three key things to take from the research.

Five key characteristics unlock playful learning

Research shows that people learn best from experiences that are joyful, that meaningfully connect the play to their lives, actively engaging, allow testing things iteratively and are socially interactive. Children won’t always experience all of those characteristics at the same time – and that’s fine. But it’s another reason children need lots of different kinds of play. Each strand helps them weave a strong and flexible tapestry of skills to use throughout their lives.

How the five characteristics of playful learning experiences help children grow and thrive

Here’s a quick look at how these qualities help children build their brains as they play

Access the full research papers

Play unlocks essential skills

Our world never stops changing, so how do we prepare children to navigate it? We let them play. Children thrive on play. It’s also perfect practice for tomorrow. Given the chance to think, negotiate, adapt to new rules and try again when things don’t go to plan, children develop essential skills that’ll last a lifetime.

Dive into our latest news, research, blogs and project stories

Ready to play?

When children play, they learn. They solve problems, think strategically, relate to others, and manage life’s ups and downs. Play helps children learn how to learn – and love learning. We've gathered some of our favourite games. You can play them anywhere – using things you find at home.

Ages: 6-9 years

Duration: 60+ minutes

Act Like an Animal

Ages: 0-3 years

Duration: 15-30 minutes

Cotton Colours

Duration: 30-60 minutes

Colour Hunt

Ages: 3-6 years

Duration: Under 15 minutes

Funny Statistics

More research

Help | Advanced Search

Computer Science > Machine Learning

Title: brain storm optimization based swarm learning for diabetic retinopathy image classification.

Abstract: The application of deep learning techniques to medical problems has garnered widespread research interest in recent years, such as applying convolutional neural networks to medical image classification tasks. However, data in the medical field is often highly private, preventing different hospitals from sharing data to train an accurate model. Federated learning, as a privacy-preserving machine learning architecture, has shown promising performance in balancing data privacy and model utility by keeping private data on the client's side and using a central server to coordinate a set of clients for model training through aggregating their uploaded model parameters. Yet, this architecture heavily relies on a trusted third-party server, which is challenging to achieve in real life. Swarm learning, as a specialized decentralized federated learning architecture that does not require a central server, utilizes blockchain technology to enable direct parameter exchanges between clients. However, the mining of blocks requires significant computational resources, limiting its scalability. To address this issue, this paper integrates the brain storm optimization algorithm into the swarm learning framework, named BSO-SL. This approach clusters similar clients into different groups based on their model distributions. Additionally, leveraging the architecture of BSO, clients are given the probability to engage in collaborative learning both within their cluster and with clients outside their cluster, preventing the model from converging to local optima. The proposed method has been validated on a real-world diabetic retinopathy image classification dataset, and the experimental results demonstrate the effectiveness of the proposed approach.

Submission history

Access paper:.

• HTML (experimental)
• Other Formats

References & Citations

• Google Scholar
• Semantic Scholar

BibTeX formatted citation

Bibliographic and Citation Tools

Code, data and media associated with this article, recommenders and search tools.

• Institution

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs .

Facility for Rare Isotope Beams

At michigan state university, frib researchers lead team to merge nuclear physics experiments and astronomical observations to advance equation-of-state research, world-class particle-accelerator facilities and recent advances in neutron-star observation give physicists a new toolkit for describing nuclear interactions at a wide range of densities..

For most stars, neutron stars and black holes are their final resting places. When a supergiant star runs out of fuel, it expands and then rapidly collapses on itself. This act creates a neutron star—an object denser than our sun crammed into a space 13 to  18 miles wide. In such a heavily condensed stellar environment, most electrons combine with protons to make neutrons, resulting in a dense ball of matter consisting mainly of neutrons. Researchers try to understand the forces that control this process by creating dense matter in the laboratory through colliding neutron-rich nuclei and taking detailed measurements.

A research team—led by William Lynch and Betty Tsang at FRIB—is focused on learning about neutrons in dense environments. Lynch, Tsang, and their collaborators used 20 years of experimental data from accelerator facilities and neutron-star observations to understand how particles interact in nuclear matter under a wide range of densities and pressures. The team wanted to determine how the ratio of neutrons to protons influences nuclear forces in a system. The team recently published its findings in Nature Astronomy .

“In nuclear physics, we are often confined to studying small systems, but we know exactly what particles are in our nuclear systems. Stars provide us an unbelievable opportunity, because they are large systems where nuclear physics plays a vital role, but we do not know for sure what particles are in their interiors,” said Lynch, professor of nuclear physics at FRIB and in the Michigan State University (MSU) Department of Physics and Astronomy. “They are interesting because the density varies greatly within such large systems.  Nuclear forces play a dominant role within them, yet we know comparatively little about that role.” 

When a star with a mass that is 20-30 times that of the sun exhausts its fuel, it cools, collapses, and explodes in a supernova. After this explosion, only the matter in the deepest part of the star’s interior coalesces to form a neutron star. This neutron star has no fuel to burn and over time, it radiates its remaining heat into the surrounding space. Scientists expect that matter in the outer core of a cold neutron star is roughly similar to the matter in atomic nuclei but with three differences: neutron stars are much larger, they are denser in their interiors, and a larger fraction of their nucleons are neutrons. Deep within the inner core of a neutron star, the composition of neutron star matter remains a mystery. 

  “If experiments could provide more guidance about the forces that act in their interiors, we could make better predictions of their interior composition and of phase transitions within them. Neutron stars present a great research opportunity to combine these disciplines,” said Lynch.

Accelerator facilities like FRIB help physicists study how subatomic particles interact under exotic conditions that are more common in neutron stars. When researchers compare these experiments to neutron-star observations, they can calculate the equation of state (EOS) of particles interacting in low-temperature, dense environments. The EOS describes matter in specific conditions, and how its properties change with density. Solving EOS for a wide range of settings helps researchers understand the strong nuclear force’s effects within dense objects, like neutron stars, in the cosmos. It also helps us learn more about neutron stars as they cool.

“This is the first time that we pulled together such a wealth of experimental data to explain the equation of state under these conditions, and this is important,” said Tsang, professor of nuclear science at FRIB. “Previous efforts have used theory to explain the low-density and low-energy end of nuclear matter. We wanted to use all the data we had available to us from our previous experiences with accelerators to obtain a comprehensive equation of state.”   

Researchers seeking the EOS often calculate it at higher temperatures or lower densities. They then draw conclusions for the system across a wider range of conditions. However, physicists have come to understand in recent years that an EOS obtained from an experiment is only relevant for a specific range of densities. As a result, the team needed to pull together data from a variety of accelerator experiments that used different measurements of colliding nuclei to replace those assumptions with data. “In this work, we asked two questions,” said Lynch. “For a given measurement, what density does that measurement probe? After that, we asked what that measurement tells us about the equation of state at that density.”   

In its recent paper, the team combined its own experiments from accelerator facilities in the United States and Japan. It pulled together data from 12 different experimental constraints and three neutron-star observations. The researchers focused on determining the EOS for nuclear matter ranging from half to three times a nuclei’s saturation density—the density found at the core of all stable nuclei. By producing this comprehensive EOS, the team provided new benchmarks for the larger nuclear physics and astrophysics communities to more accurately model interactions of nuclear matter.

The team improved its measurements at intermediate densities that neutron star observations do not provide through experiments at the GSI Helmholtz Centre for Heavy Ion Research in Germany, the RIKEN Nishina Center for Accelerator-Based Science in Japan, and the National Superconducting Cyclotron Laboratory (FRIB’s predecessor). To enable key measurements discussed in this article, their experiments helped fund technical advances in data acquisition for active targets and time projection chambers that are being employed in many other experiments world-wide.   

In running these experiments at FRIB, Tsang and Lynch can continue to interact with MSU students who help advance the research with their own input and innovation. MSU operates FRIB as a scientific user facility for the U.S. Department of Energy Office of Science (DOE-SC), supporting the mission of the DOE-SC Office of Nuclear Physics. FRIB is the only accelerator-based user facility on a university campus as one of 28 DOE-SC user facilities .  Chun Yen Tsang, the first author on the Nature Astronomy  paper, was a graduate student under Betty Tsang during this research and is now a researcher working jointly at Brookhaven National Laboratory and Kent State University. 

“Projects like this one are essential for attracting the brightest students, which ultimately makes these discoveries possible, and provides a steady pipeline to the U.S. workforce in nuclear science,” Tsang said.

The proposed FRIB energy upgrade ( FRIB400 ), supported by the scientific user community in the 2023 Nuclear Science Advisory Committee Long Range Plan , will allow the team to probe at even higher densities in the years to come. FRIB400 will double the reach of FRIB along the neutron dripline into a region relevant for neutron-star crusts and to allow study of extreme, neutron-rich nuclei such as calcium-68. 

Eric Gedenk is a freelance science writer.

Michigan State University operates the Facility for Rare Isotope Beams (FRIB) as a user facility for the U.S. Department of Energy Office of Science (DOE-SC), supporting the mission of the DOE-SC Office of Nuclear Physics. Hosting what is designed to be the most powerful heavy-ion accelerator, FRIB enables scientists to make discoveries about the properties of rare isotopes in order to better understand the physics of nuclei, nuclear astrophysics, fundamental interactions, and applications for society, including in medicine, homeland security, and industry.

The U.S. Department of Energy Office of Science is the single largest supporter of basic research in the physical sciences in the United States and is working to address some of today’s most pressing challenges. For more information, visit energy.gov/science.

AI + Machine Learning , Announcements , Azure AI , Azure AI Studio

Introducing Phi-3: Redefining what’s possible with SLMs

By Misha Bilenko Corporate Vice President, Microsoft GenAI

Posted on April 23, 2024 4 min read

• Tag: Copilot
• Tag: Generative AI

We are excited to introduce Phi-3, a family of open AI models developed by Microsoft. Phi-3 models are the most capable and cost-effective small language models (SLMs) available, outperforming models of the same size and next size up across a variety of language, reasoning, coding, and math benchmarks. This release expands the selection of high-quality models for customers, offering more practical choices as they compose and build generative AI applications.

Starting today, Phi-3-mini , a 3.8B language model is available on Microsoft Azure AI Studio , Hugging Face , and Ollama . 

• Phi-3-mini is available in two context-length variants—4K and 128K tokens. It is the first model in its class to support a context window of up to 128K tokens, with little impact on quality.
• It is instruction-tuned, meaning that it’s trained to follow different types of instructions reflecting how people normally communicate. This ensures the model is ready to use out-of-the-box.
• It is available on Azure AI to take advantage of the deploy-eval-finetune toolchain, and is available on Ollama for developers to run locally on their laptops.
• It has been optimized for ONNX Runtime with support for Windows DirectML along with cross-platform support across graphics processing unit (GPU), CPU, and even mobile hardware.
• It is also available as an NVIDIA NIM microservice with a standard API interface that can be deployed anywhere. And has been optimized for NVIDIA GPUs . 

In the coming weeks, additional models will be added to Phi-3 family to offer customers even more flexibility across the quality-cost curve. Phi-3-small (7B) and Phi-3-medium (14B) will be available in the Azure AI model catalog and other model gardens shortly.   

Microsoft continues to offer the best models across the quality-cost curve and today’s Phi-3 release expands the selection of models with state-of-the-art small models.

Azure AI Studio

Phi-3-mini is now available

Groundbreaking performance at a small size 

Phi-3 models significantly outperform language models of the same and larger sizes on key benchmarks (see benchmark numbers below, higher is better). Phi-3-mini does better than models twice its size, and Phi-3-small and Phi-3-medium outperform much larger models, including GPT-3.5T.  

All reported numbers are produced with the same pipeline to ensure that the numbers are comparable. As a result, these numbers may differ from other published numbers due to slight differences in the evaluation methodology. More details on benchmarks are provided in our technical paper . 

Note: Phi-3 models do not perform as well on factual knowledge benchmarks (such as TriviaQA) as the smaller model size results in less capacity to retain facts.  

Safety-first model design 

Responsible ai principles

Phi-3 models were developed in accordance with the Microsoft Responsible AI Standard , which is a company-wide set of requirements based on the following six principles: accountability, transparency, fairness, reliability and safety, privacy and security, and inclusiveness. Phi-3 models underwent rigorous safety measurement and evaluation, red-teaming, sensitive use review, and adherence to security guidance to help ensure that these models are responsibly developed, tested, and deployed in alignment with Microsoft’s standards and best practices.  

Building on our prior work with Phi models (“ Textbooks Are All You Need ”), Phi-3 models are also trained using high-quality data. They were further improved with extensive safety post-training, including reinforcement learning from human feedback (RLHF), automated testing and evaluations across dozens of harm categories, and manual red-teaming. Our approach to safety training and evaluations are detailed in our technical paper , and we outline recommended uses and limitations in the model cards. See the model card collection .  

Unlocking new capabilities 

Microsoft’s experience shipping copilots and enabling customers to transform their businesses with generative AI using Azure AI has highlighted the growing need for different-size models across the quality-cost curve for different tasks. Small language models, like Phi-3, are especially great for: 

• Resource constrained environments including on-device and offline inference scenarios.
• Latency bound scenarios where fast response times are critical.
• Cost constrained use cases, particularly those with simpler tasks.

For more on small language models, see our Microsoft Source Blog .

Thanks to their smaller size, Phi-3 models can be used in compute-limited inference environments. Phi-3-mini, in particular, can be used on-device, especially when further optimized with ONNX Runtime for cross-platform availability. The smaller size of Phi-3 models also makes fine-tuning or customization easier and more affordable. In addition, their lower computational needs make them a lower cost option with much better latency. The longer context window enables taking in and reasoning over large text content—documents, web pages, code, and more. Phi-3-mini demonstrates strong reasoning and logic capabilities, making it a good candidate for analytical tasks. 

Customers are already building solutions with Phi-3. One example where Phi-3 is already demonstrating value is in agriculture, where internet might not be readily accessible. Powerful small models like Phi-3 along with Microsoft copilot templates are available to farmers at the point of need and provide the additional benefit of running at reduced cost, making AI technologies even more accessible.  

ITC, a leading business conglomerate based in India, is leveraging Phi-3 as part of their continued collaboration with Microsoft on the copilot for Krishi Mitra, a farmer-facing app that reaches over a million farmers.

“ Our goal with the Krishi Mitra copilot is to improve efficiency while maintaining the accuracy of a large language model. We are excited to partner with Microsoft on using fine-tuned versions of Phi-3 to meet both our goals—efficiency and accuracy! ”    Saif Naik, Head of Technology, ITCMAARS

Originating in Microsoft Research, Phi models have been broadly used, with Phi-2 downloaded over 2 million times. The Phi series of models have achieved remarkable performance with strategic data curation and innovative scaling. Starting with Phi-1, a model used for Python coding, to Phi-1.5, enhancing reasoning and understanding, and then to Phi-2, a 2.7 billion-parameter model outperforming those up to 25 times its size in language comprehension. 1 Each iteration has leveraged high-quality training data and knowledge transfer techniques to challenge conventional scaling laws. 

Get started today 

To experience Phi-3 for yourself, start with playing with the model on Azure AI Playground . You can also find the model on the Hugging Chat playground . Start building with and customizing Phi-3 for your scenarios using the  Azure AI Studio . Join us to learn more about Phi-3 during a special  live stream of the AI Show.  

1 Microsoft Research Blog, Phi-2: The surprising power of small language models, December 12, 2023 .

Let us know what you think of Azure and what you would like to see in the future.

Provide feedback

Build your cloud computing and Azure skills with free courses by Microsoft Learn.

Explore Azure learning

Related posts

AI + Machine Learning , Azure AI , Azure AI Content Safety , Azure Cognitive Search , Azure Kubernetes Service (AKS) , Azure OpenAI Service , Customer stories

AI-powered dialogues: Global telecommunications with Azure OpenAI Service   chevron_right

AI + Machine Learning , Azure AI , Azure AI Content Safety , Azure OpenAI Service , Customer stories

Generative AI and the path to personalized medicine with Microsoft Azure   chevron_right

AI + Machine Learning , Azure AI , Azure AI Services , Azure AI Studio , Azure OpenAI Service , Best practices

AI study guide: The no-cost tools from Microsoft to jump start your generative AI journey   chevron_right

AI + Machine Learning , Azure AI , Azure VMware Solution , Events , Microsoft Copilot for Azure , Microsoft Defender for Cloud

Get ready for AI at the Migrate to Innovate digital event   chevron_right