The 10 Most Significant Education Studies of 2025

After installing over 1,000 wooden boxes across 10 college campuses in India, researchers embarked on a massive, randomized controlled trial to determine the impact of cell phone bans on academic performance. Over the course of a semester, almost 17,000 students either marched into class with their phones or deposited them in the new wooden receptacles and—suddenly, heartbreakingly Instagram-less—took their seats.

At scale, phone-free classrooms produced better academic outcomes, especially among new and struggling students. Phone bans “can substantially improve student grades” and “narrow achievement gaps,” the researchers concluded, providing conclusive evidence of a causal link between cell phone restrictions and better grades. A 2025 study of phone bans in Florida high schools, which appeared a few months later, also reported academic improvements and significant reductions in unexcused absences—though troubling, short-term patterns of racialized discipline were initially reported as the bans got off the ground.

The will to ban phones in U.S. schools has been slow to coalesce but has recently gained momentum. In 2025, according to Ballotpedia, 22 states passed new laws restricting phone use in schools.

More states should listen to teachers and follow suit. Beyond improved grades, the researchers in India reported “fewer instances of disruptive behavior” in classrooms, “less peer-to-peer conversation unrelated to course material,” and better student-teacher relationships. Kids who attended class without phones, meanwhile, became converts: As a group, they were “significantly more supportive of phone-use restrictions” going forward—signaling a “convergence of academic performance” and “increased student receptivity” that highlights the potential of phone bans to transform school cultures, the researchers say.

In everything from recipes to social media algorithms, important mathematical principles are at work below the surface of everyday life. But even simple math word problems that attempt to capture this reality can derail students as they wade through a tangle of numbers, operations, and story details.

After analyzing 1,000 solutions to middle school math problems, researchers found that students often stumbled when trying to translate narrative text into manageable, computable steps. What may seem like a straightforward scenario—calculating the gas, food, and lodging costs for a family road trip, for example—can exert “high demands on working memory” as information outstrips cognitive bandwidth, the researchers explain in a 2025 study.

In the study, the most common technique—highlighting key elements of a word problem—was only marginally helpful. But when used as the first step in a broader “organizational and elaborative” approach that included sketching diagrams, categorizing information, and annotating the problems with arrows or labels, students were better able to see how the pieces fit together. Kids who used these note-taking strategies were 29 percent more likely to solve a problem than those who simply highlighted details, the study found.

What’s the secret behind the strategies? Math word problems often present more information than students can hold in working memory. “Capable problem solvers” offload information to sketch pads and margin notes and reintegrate it later, researchers explain in a different 2025 study—allowing savvy students to refocus their attention on a smaller set of factors as they work to translate a story into solvable math.

Sustained student attention, which researchers often refer to as “vigilance,” is the horse that never wins. In a 2025 study of attentional limits in several UK classrooms, in fact, researchers detected the first signs of wobbly student focus a mere five minutes into a lecture. Attention then steadily declined for the rest of the lesson.

In search of possible solutions, the researchers enrolled 253 sophomore psychology undergraduates in 90-minute lectures that were interrupted by 90-second microbreaks every 10 minutes, or by a single 10-minute break at the halfway point. Microbreaks were casual affairs, consisting of activities like “closing your eyes, quietly speaking with fellow classmates, stretching, or drinking water.” On quizzes that measured attention at regular lecture intervals, students in the microbreak group outperformed their counterparts by significant margins—up to 76 percent better—virtually every step of the way.

The attempt to master challenging material always leads to mind-wandering—there are no cures. Historical studies of attention spans place the limits at various thresholds, from eight to 10 to 25 minutes. Instead of trying to “overcome these constraints,” the researchers say, educators should “acknowledge the theoretical impossibility of perfect sustained attention” and choose strategies that are compatible with “inherent neural, biological, and cognitive limitations.”

When in doubt, chunk learning into smaller parts, incorporate regular movement or chat breaks, or change learning modalities to allow kids to rest and reset, perhaps as often as every 10–15 minutes.

In the debate over screens versus paper, new research tips the scale dramatically, revealing that writing by hand—but not typing—helps build the cognitive framework young students rely on to decode letters and recognize words.

In a study published earlier this year, researchers asked 5-year-old pre-readers to study new letters and two-syllable words, then practice them through handwriting or typing. After several short training sessions, the children were tested on how well they could name the letters, write them from dictation, and recognize and decode the new words.

Across nearly every measure, the children who wrote letters by hand demonstrated superior alphabetic and orthographic skills. When asked to name letters, handwriters achieved 92 percent accuracy, compared to 75 percent for typers, and the discrepancy was even greater for letter writing: Children who practiced by hand wrote the new letters accurately more than twice as often—64 percent of the time, compared to 28 percent for their typing peers.

In recent years, neuroscientists also peeked inside the brains of older students as they used pencils and keyboards. A 2020 study of seventh graders revealed telltale traces of deeper learning when kids wrote words instead of typing them, and a study published earlier this year confirms that handwriting is an “important tool for learning and memory retention” that benefits students across all ages, including middle and high school. 

As screens increasingly claim space in children’s daily routines, the studies argue for a return to older technologies. For the youngest readers and writers, the need for a steady diet of pencil and paper work is inarguable. Meanwhile, middle and high school students can move between tools like Google Docs and old-fashioned paper notebooks, gaining crucial experience with modern technologies while periodically slowing down to engage in methodical, embodied thinking.

Feeling competent is crucial to well-being, but for young learners it tends to come at a price: a dose of (healthy) frustration. 

Now a new study confirms that when adults spot a struggling student and intervene too quickly, it can signal that solutions are beyond the child’s ability—and dampen their confidence and willingness to take intellectual risks when new challenges arise.

As early as age 5, children across a range of studies reviewed by the researchers became “less motivated to persist on a difficult task” after an adult stepped in to help solve a puzzle, suggesting that even “well-intentioned behavior toward children can backfire” and sabotage important skill-building opportunities. Likewise, when 6- to 11-year-old girls received “unsolicited help” from an adult during a paper-folding exercise, they reported feeling “less smart afterward.” 

The impulse to rescue students from confusion and frustration is hard to ignore, but real learning often happens in the difficult moments just before we do. When tempted to step in, educators might consider other scaffolds like “providing hints or asking questions,” the researchers suggest, pointing kids in the right direction without doing the thinking for them. Offering a few useful stepping stones in lieu of answers, the research reveals, can preserve independence and foster self-sufficiency.

Special education teachers often face an overwhelming volume of paperwork, from drafting IEPs to logging weekly data on student progress and tracking learning accommodations. 

That’s precious time that could be reallocated to working directly with kids. In a 2025 study, a team of researchers asked experienced K–12 special education teachers to write an IEP goal based on a brief description of a student’s disability, past performance, and areas of need. The same teachers then used ChatGPT to generate an IEP goal by providing basic information about a student’s learning differences. 

After analyzing both sets of goals on six dimensions including clarity, measurability, and timeliness, the researchers found “no statistically significant difference in quality” between the AI-generated versions and those written entirely by teachers. The teachers, however, had a more favorable impression: Most said that the ChatGPT goals were “either of the same or better quality than they think a special education teacher… would have written,” and viewed AI as a tool they could use to improve efficiency.

Recalling the hours she spent drafting documents for her students, Danielle Waterfield—the study’s lead author and a former special education teacher we interviewed for this article—said that her time with AI convinced her of the tech’s potential to relieve that administrative burden and shift special education teachers from paperwork to what really counts: “face-to-face time with their students.”

Decades ago, according to a June 2025 study, the U.S. had a “simple philosophy” on school recess that recognized outdoor play as “essential for healthy and happy children” and honored the principle by setting aside 60 minutes for daily recess. In the ensuing years, the researchers say, a creeping tide of academic expectations led to more seat time and testing, undermining the quality and quantity of free play in schools.

The cumulative impact is no joke. To test a theory linking a lack of playtime to chronic stress in kids, researchers conducted a novel experiment, comparing 130 fourth-grade volunteers who received either 30 or 45 minutes of daily recess during the academic year. Hair samples of the children were then analyzed for cortisol levels, providing a unique biological measure of chronic rather than short-term stress. Students in the 45-minute play group had 68 percent less cortisol stored in hair strands—from head to toe, they were chill.

Despite the surprising results, the researchers—who say that recess should be frequent, unstructured, and outdoors—aren’t really going out on a limb. 

More independent play yields happier, more socially competent children, according to an ever-expanding body of research. An exhaustive 2023 study that combed through 50 years of historical records, for example, concluded that kids in the past spent more time outdoors and derived long-term benefits from opportunities to “play, roam, and engage in other activities independent of direct oversight and control by adults.” On the question of play, it might be time to turn back the clock.

From kindergarten through high school, students spend roughly 15,000 hours with teachers, making the quality of those relationships a crucial factor in learning. This year, two studies covering millions of school-aged kids resoundingly affirm the point.

In the first, a team of researchers analyzed 70 years of studies encompassing more than 2.6 million K–12 students and concluded that trusting, supportive student-teacher relationships were linked to a wide range of benefits across grade levels: higher academic achievement, improved behavior, better executive function and self-control, and greater feelings of belonging, motivation, and well-being. Positive relationships were equally important for girls and boys, the researchers noted, but may have outsized benefits “for students in middle and high school [when] compared to younger children.”

Meanwhile, a 2025 meta-analysis of 40 studies found that comprehensive social and emotional programs in grades 1–12 have a clear impact on academics, enough to improve overall academic achievement by 8.4 percentile points. “Students who feel a sense of belonging within the school community are more successful academically,” the researchers note, pointing to the crucial role that teachers play in creating a culture that helps students reach their full potential.

Relationships before rigor holds true, then. The two studies are a powerful reminder that social cohesion in classrooms—the sense of belonging that flourishes when teachers cultivate trust with students, maintain consistent support, and hold high expectations—may be one of the most reliable levers for improving learning outcomes.

Thrust into the chaos of real classrooms, pre-service teachers may look back at their training and wonder if too much time was spent learning about conceptual models—and not enough time practicing everyday teacher moves.

Researchers wondering the same thing compared “traditional” teacher prep programs, which emphasize reading and discussion of theoretical frameworks, with “practice-based” approaches that focus on expert observation and role-playing in simulated classroom environments. Watching videos of master teachers and then “rehearsing” in the presence of coaches might be especially beneficial for pre-service teachers, the authors hypothesized, because the approach allows for “feedback in the moment” along with quick hints on “how to elevate instruction.”

In the end, practice trumped theory. When attempting to elicit and respond to correct and incorrect answers in elementary math—by helping adult actors playing students to solve single-digit problems, for example—prospective teachers who had analyzed teacher moves and then actively rehearsed them with coaches were more proficient than those who had merely discussed possible strategies.

Teaching dozens, or even hundreds, of students is mind-numblingly complicated: Kids process information at different speeds, possess wildly disparate skills in reading and math, and sometimes come to school grumpy, fidgety, or even desperately hungry. To get learning off the ground under those circumstances, as so many teachers do every day, theory is insufficient. Regular practice, access to inspiring mentors, time for planning, and plenty of encouragement and patience from administrators and peers is the path to improving one of the most challenging jobs in the world.

Buried in the appendix of a study Edutopia covered last year was a troubling detail about how roughly 1,000 high school students had used AI to complete grade-level math problems. Given free rein with ChatGPT, ninth, 10th, and 11th graders had engaged in only superficial conversations with the software; among the most frequent student queries were “can u solve this question?” and “what is the answer?”

The results were unsurprising: AI users performed well in practice sessions, but then quickly forgot most of what they’d learned and bombed a closed-book test on the material. 

Now a 2025 MIT study of older students wired up to EEG machines as they wrote essays reveals a similar pattern. College-aged students who were given access to ChatGPT as they responded to provocative questions tended to “follow the thinking” of the machine, produced “statistically homogenous essays,” and exhibited brain activity that was localized and poorly coordinated. Stunningly, minutes later, only 17 percent of the ChatGPT users could recall a single sentence from their essays. Students who used search engines or wrote essays without any assistance fared much better, recalling sentences at rates of 83 and 89 percent, respectively.

It’s not as cut-and-dried as it sounds; the how and when of AI usage seems to matter a great deal. Several recent studies conclude that AI tutors designed to withhold answers and ask probing questions, for example, make excellent study partners. And in the MIT writing study, the researchers found a silver lining: Authors who initially wrote their own essays managed to use AI effectively during a later round of revision. To help students get better results, educators should consider “combining AI tool assistance with tools-free learning phases,” the researchers suggest.

Despite the sense of inevitably around AI, surrender isn’t the answer. ELA teachers are right to ask which writing skills must be safeguarded—from brainstorming to outlining to writing good transitions—and they are perfectly right to limit or prohibit the use of AI as students practice those skills. If students never do it, they’ll never learn it.

Hero image by Ben Denzer. Additional illustrations by Maria Hahne and Chelsea Beck.