L&B Blog – Education & Teacher Conferences Skip to main content

Category Archives: L&B Blog

Older posts
Thinking Critically About Critical Thinking
Andrew Watson
Andrew Watson
July 05, 2026

Thinking Critically About Critical Thinking

Read More

Almost any teacher will say: “in our school, we want students to become critical thinkers.” Once we embrace that praiseworthy goal, we’ve got some questions to ask.

How, exactly, do we help students think critically?

Or, here’s a bigger question:

Can we do that? Is critical thinking a skill that can be taught?

This debate frequently rages in educational circles.

  • Team A says: “our society needs critical thinkers! Schools should teach this skill…here’s how.”
  • Team Z says: “critical thinking isn’t a generic skill! People need knowledge with which to think critically…schools should teach that knowledge.”

I myself am sympathetic to Team Z’s claims. For instance, I’ve written a book about critically evaluating “research-based” claims in education. In that book, I repeatedly warn that I’m offering steps to think critically about this one topic; I’m not the right guy to think critically about the history of medieval Russia, or innovations in jet engine design, or a new rugby formation. I just don’t know enough about those topics to guide critical thinking about them.

I recently came across a study looking at this question. This study, helpfully, offers both reasons to be optimistic and reasons to be cautious. Here’s the story.

“Correlation Isn’t …”

A research team looked at 400 students taking college-level philosophy courses. About 115 of them took a course called “Critical Thinking”; the rest took other intro-level courses: “Introduction to Philosophy” and “Moral Problems.”

The critical thinking class focused on common biases or errors in judgment:

  • mistaking correlation for causation
  • honoring sunk costs
  • ignoring regression to the mean
  • forgetting about opportunity costs, and
  • the gambler’s fallacy (“if a fair coin comes up heads 8 times, the next toss is REALLY likely to come up tails”)

The instructor began the course by offering obvious examples that intuitively resonated with students. Over the course of the term, students practiced applying the rules in many different contexts: athletics, romantic relationships, business, war, friendship, and so forth. The class interleaved these examples, and worked to ensure that students looked for the “deep structure” behind them. That is: rather than thinking “this is a claim about capital punishment,” the students learned to think “this is a claim about causation…I should be sure it’s not relying on correlational data.”

At the end of the course, students in the Critical Thinking class made ENORMOUS strides compared to the students in the other philosophy classes. The effect sizes ranged from 0.91 — well into the “large” range — to 2.01 — comfortably in the “I’ve never seen anything like it” range.

Let’s look at some raw data. For one of the fallacies — sunk costs — the control group and the critical thinking group both bought into a “sunk costs” logic at the beginning of the term. On a scale of 1 (“that’s not a good reason at all”) to 7 (“that’s a really good reason”), they gave a sunk-costs argument an average rating of 5.5. That is: they substantially agreed that “past investments should influence future decisions.” (“I’ve read half of this book, so I’ve got to finish it … even though I’m not enjoying it.)

At the end of the term, the control group continued to support this fallacy: they rated sunk costs reasoning at 5.5. The critical thinking students now dropped their average ratings to 1.4: very close to the “that’s not a good reason at all” end of the scale.

Even more impressive, those gains lasted. 25% of the enrolled students returned for a post-test sixteen months later. The effect sizes remained roughly 1.00 — a remarkably high number.

Technically speaking, this was quite a class.

Practicing What We Preach

Having seen the highlights of this critical-thinking study, let’s think critically about its methods and conclusions. A few points stand out.

First: students were not randomly placed in these courses; they opted in. Perhaps it’s unsurprising that students who signed up for a course called “Critical Thinking” improved at critical thinking. In other words: these students might differ in meaningful ways from students in the other courses.

A related point: these are college students taking philosophy courses. That’s a very select group within a very select group. As I’ve often argued, we should be cautious about applying college-student data to K-12 education.

Second: this was one course with one teacher. Maybe this study has identified an unusually effective professor.

Third: I gotta say, those effect sizes give me pause. A wise stats-y friend of mine says: “in education research, an effect size of greater than 1.00 is almost always suspect.”

Fourth: the post-test problems weren’t identical to the class examples…but they were structurally highly similar. We could plausibly call the results “near transfer,” but not “far transfer.”

Remembering Opportunity Costs

When I DO teach this novel, I’m NOT teaching that novel. That absence is the “opportunity cost,” and we should always be on the lookout for them.

In this case, I think we can say:

College students who opted into a critical thinking class got better at spotting four specific reasoning errors. (They didn’t get better at spotting the fifth — the gambler’s fallacy.) Their ability to spot those fallacies lasted an impressively long time: at least 16 months.

We don’t know if other professors are as effective as this one. We don’t know if such a course would work with younger, or with less interested, students. And — crucially — we don’t know if students got better at spotting those errors in their actual lives: when reading advertisements for timeshares, or evaluating claims by political parties, or making decisions about medical treatment.

For the reasons listed in the second paragraph, I worry about summarizing this study by saying “look, critical thinking CAN be taught”; instead, I think it arrives at a much more modest set of claims.

And: I worry about opportunity costs. I’m quite sure that students can’t think critically without substantial amounts of factual knowledge. I’m still unsure if students — even when they do well in a course like the one described above — actually think critically in real life. For that reason, I’m all in favor of talking with students about critical thinking; I certainly offer up examples of doing so in the classroom (and, I hope, on this blog). I’m still not persuaded that the time taken to teach a full course on the topic — especially to younger students — will accomplish our praiseworthy goal.

Bishop, M., Feltz, A., & Conway, P. (2026). Critical thinking classes can reduce common biases: Results from a field experiment. Journal of Experimental Psychology: Applied.

When Five Doesn’t Equal Five: Counting “Slots” in Working Memory
landb
landb
June 28, 2026

When Five Doesn’t Equal Five: Counting “Slots” in Working Memory

Read More

If you’ve been to a Learning and the Brain conference in the last 26 years, you’ve heard at least one speaker talk about the importance of working memory. Your working memory – typically abbreviated WM – allows your mind to hold a few bits of information and then reorganize and combine them into something new.

For example, imagine I say to you “without writing anything down, alphabetize the workdays of the week.”

You start by recalling those days in the order they occur: “Monday, Tuesday, Wednesday, Thursday, Friday.” Now that you’re holding those bits of information, you can reorganize them in your mind: “Friday, Monday, Thursday, Tuesday, Wednesday.” Voila: that was working memory.

You can quickly see that WM is at the core of all academic learning. To learn (almost) anything new, students must use their working memory.

For that reason, discussions of WM typically stress this alarming point: we just don’t have very much working memory. For instance: recall that alphabetizing task I gave you a few paragraphs ago. When I ask teachers to do that in a workshop, most of them succeed. But a few minutes later, when I ask them to alphabetize 10 months of the year, they laugh nervously and give up almost immediately.

In brief: alphabetize 5, no problem. Alphabetize 10, no can do. This realization guides almost everything we do as teachers.

The Size of The Problem

The importance of these insights leads to an obvious question. We know that working memory is small…but just how small? Exactly how narrow is this cognitive bottleneck that constricts students’ thinking?

The best-known answer comes from a well-known article by George Miller, written way back in 1956. His clever title tells us what we need to know: “The Magical Number Seven Plus or Minus Two.” Miller’s formula says, basically, that adults have, roughly, 7 “slots” in WM. Some folks have lower WM capacity – perhaps as low as 5. Others reach higher – up to 9. Voila: 7±2.

From one perspective, this formula makes sense. It puts plausible numbers on two experiences:

  • We don’t have a lot of WM, and
  • Some people have more than others

In fact, Miller’s formula helps explain that WM exercise described at the top of this blog post. Teachers successfully alphabetize the workdays of the week because there are five of them – comfortably at the low end of Miller’s range. But when I ask teachers to alphabetize 10 months, they routinely fail – because the task goes beyond the maximum of 9.

You might even have heard of another bit of support for this formula: “phone numbers were initially limited to seven digits to keep them within average WM capacity.” This claim turns out not to be true. After all, the US phone number system was developed in the 1940s, over a decade before Miller published his paper. But the popularity of this myth suggests that Miller’s argument just makes sense to many people.

Honey, I Shrunk the Memory

Despite all these reasons to adopt the 7±2 mantra, we have at least two good reasons to resist it.

First, more recent research, published by Nelson Cowan in 2001, suggests that working memory may have as few as four slots. If we accept Cowan’s revised formula, we understand even more viscerally why our students struggle with working-memory tasks in school – where almost everything is a working-memory task.

More radically, I want to propose a second reason to resist Miller’s account of seven slots. In fact, I want to move away from the idea of “slots” altogether – whether we’re talking seven or four. In my view, the “slots” explanation of working memory encourages teachers to think about the wrong thing; truthfully, it asks us to do something quite impossible. Let me explain.

When 5 ≠ 5

I suggested above that “alphabetizing five workdays” falls at the low end of Miller’s 7±2 formula. This way of thinking encourages teachers to focus on counting. In other words, we should be asking ourselves: how many specific chunks are students manipulating at this moment? If the answer is “less than seven,” then everything should be fine.

But let’s go back and look at that task. Notice that – to succeed at that task – you need to hold MANY more items in WM.

  • You need to hold on to the instructions I gave you. If you didn’t keep track of the task demands, you couldn’t complete it.
  • You also need to keep track of the order of the alphabet. To decide where “Thursday” fits in your revised list, you’re constantly checking in with that rhyme you learned even before kindergarten.
  • By the way: how many “slots” does the order of the alphabet fill? One? Twenty-six? some other number?

This wider view of working-memory demands leads me to two conclusions:

A: Accomplishing this mental alphabetization task requires holding and processing MANY more than 5 bits of information.

B: More broadly, trying to “count slots” is an entirely futile endeavor. I’ve been using this alphabetizing test for years, and I have no useful notion of how to quantify its WM lift. I can say this: “most teachers succeed at the first; almost no one succeeds at the second.” But trying to assign a numeric value to these tasks leads to frustration and confusion – not to better teaching.

Teaching Implications

In my view, focusing on “slots” distracts us from a more useful and important task: recognizing and solving students’ working memory overload. That is, we should be good at

  • Anticipating the classroom experiences that might result in overload,
  • Recognizing overload when it happens, and
  • Solving – or at least mitigating – those problems.

Yes, having a number like 4 or 7 in mind might be helpful background knowledge. But the real work comes not in counting, but in rethinking the work we do in the classroom. (Here’s a series of blog posts on how to do so.) Teachers will experience our teaching work differently when we start seeing learning from our students’ working-memory perspective.

Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological review, 63(2), 81.

Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioral and brain sciences, 24(1), 87-114.

Beyond “Checking for Understanding”
Andrew Watson
Andrew Watson
June 21, 2026

Beyond “Checking for Understanding”

Read More

Blog posts here typically start with research and then consider classroom application. Today’s post — more speculative than most — begins with a concept and then moves to thinking aloud. In brief: I want to explore the concept of “checking for understanding,” and propose a new category of classroom moves that should precede such checks.

Completing the Subject

As an English teacher, I get paid money to explain the grammatical concept of “subject complement” to my students. Let’s consider these sentences:

Joyce is President. She is decisive.

In these sentences, we know that “is” is a linking verb. When we look at the words “President” and “decisive” we see that they offer more information about Joyce. “President” is another word for “Joyce.” “Decisive” describes Joyce.

So, a subject complement completes the subject. Specifically:

A subject complement follows a linking verb, and renames or describes the subject.

“President” and “decisive” are subject complements because they follow the linking verb “is,” and offer more information about the subject Joyce.

In this explanation, I’ve given my students lots to think about. I should stop and check to see how all this information is landing. Specifically, I’m worried about working memory overload. My students have taken onboard new information, and they have to combine that new information with ideas already in long-term memory. Both of those demands — processing and combining new info — can strain working memory to the breaking point.

I can easily brainstorm several questions I might ask at this point to assess working memory function. Here — in a deliberately jumbled order — are four of my potential questions. Please take a minute to sort these questions by their difficulty: that is, which one places the least demand on students’ WM, and which one the most?

  1. Write a sentence with a subject complement.
  2. In the paragraph below, convert subject complements into appositives.
  3. What is the definition of “subject complement”?
  4. In the sentence below, which word is the subject complement?
[I’m strategically pausing here to give you a chance to think.]

I suspect you came up with this order:

  1. What is the definition of “subject complement”?
  2. In the sentence below, which word is the subject complement?
  3. Write a sentence with a subject complement.
  4. In the paragraph below, convert subject complements into appositives.

The first question simply asks students to repeat the definition I just gave them. The next three questions ask for application, each one with a greater degree of difficulty than the previous. That is: whereas question 3 is a bit more difficult than question 2, question 4 is MUCH more difficult than question 3.

Checking for Something

Let’s go back to look at those questions. Which of them, in your opinion, measures understanding? That is: you would comfortably say: “If my students can answer question X correctly, they understand the idea of subject complement?”

I myself would say: a correct answer to question 4 starts to demonstrate “understanding.” (Reasonable people might argue that #3 also measures “understanding.” I think “understanding” requires more than “simple application,” but the argument I’m about to make will hold wherever you draw the line on that list of questions.)

Of course, the other three questions give me important data:

  • Have my students heard what I said?
  • Have they done some basic processing of the idea?
  • Has my lesson plan prompted initial encoding?
  • Can they apply this new idea in basic ways?

So, I really must ask those first three questions — and others like them — even though they’re not actually helping me know if my students understand.

In my experience, those first three questions would typically fall under the heading “checking for understanding.” However — and here’s the point of my blog post — such questions don’t by themselves verify understanding. For that reason, the phrase “checking for understanding” is misleading. It provides false comfort. Once I’ve asked those questions, I’ve checked for something, but I haven’t yet learned if my students understand.

We need a better label. I have a modest proposal: “checking for uptake.”

Walking before Running

In suggesting this new category, I’m deliberately focusing on very basic parts of teaching. My students can’t capital-l Learn about a topic, can’t capital-u Understand a topic, if they haven’t started basic encoding and processing. “Checking for uptake” doesn’t sound especially glamorous, but the lack of glamour is the point.

Because I’m always focused on working memory, I think checking for uptake questions provide two working-memory benefits.

First: if my students haven’t successfully taken up foundational ideas, they won’t be able to do more advanced cognitive work. I can prevent future WM overload by ensuring that basic ideas are onboard. Checking for uptake helps me solve WM problems before they arise.

In other words: until my students succeed at increasingly challenging checking-for-uptake questions, I shouldn’t even ask checking-for-understanding questions.

In my own teaching, I think one of my greatest failings has been to ask “understanding” questions before my students had fully taken in the ideas I was presenting.

Second: if my students can’t answer checking for uptake questions, I might discover that even my initial explanation overwhelmed working memory. Perhaps my students don’t understand “linking verb” as well as I thought. Perhaps they’re not even clear on the idea of a grammatical subject.

In other words: checking for uptake can both prevent working memory overload in the future and reveal WM overload that might already have happened.

Double Checking

To ensure that we don’t overload students’ WM during direct instruction, we should check for two Us:

Checking for Uptake: did my students hear what I said? Can they say it back to me, or paraphrase it? Can they take baby steps with the concept?

Checking for Understanding: once students have taken up a lesson, can they do something new with it? If yes, can they do something even more challenging? Can they override prior misconceptions? Can they create and analyze?

Good teaching will include both kinds of checking — but Checking for Understanding should follow Checking for Uptake.

Do Captions Turn Screen Time Into Reading Time?
Andrew Watson
Andrew Watson
June 14, 2026

Do Captions Turn Screen Time Into Reading Time?

Read More

A year ago, I found myself drawn to an intriguing question: when showing educational videos, should we turn captions on to help students learn or focus? In my experience, most video services provide caption capabilities, and many turn them on by default. For this reason, I assumed that we have good reason to use captions. But, do we?

The best known research summary answers with an emphatic YES; the title says it all: ““Video Captions Benefit Everyone.” Its first two sentences:

More than 100 empirical studies document that captioning a video improves comprehension of, attention to, and memory for the video.

Video captions, also known as same-language subtitles, benefit everyone who watches videos (children, adolescents, college students, and adults).

I quickly learned that this summary’s confidence lacks much justification. The 100 empirical studies it reviews rely on self-report data, focus on college-age students, and/or look at data for language learners: that is, for example, English speakers learning Spanish by watching captioned videos. (While this last category is certainly interesting, it’s a very specific use of captions; it doesn’t obviously generalize to answer the broader question.)

In my search, I found exactly one study using objective measures to look at neurotypical K-12 students learning from same-language captions. That study suggests — but does not conclude — that non-struggling readers learn FEWER words when they watch videos with captions.

Given the alarming dearth of evidence, I concluded:

  1. We just don’t have enough evidence to make strong claims one way or another, and
  2. I really hope that someone starts reseaching this topic.

Dreams Realized

Even as I wrote that blog post, my hopes were being met by a research team in England. This team, led by Dr. Anastasiya Lopukhina, worked with 127 six- to eight-year-olds in England. All the children spoke English at home; none had diagnoses for special needs.

In some ways, this study was quite straightforward. For six weeks, half of the children watched TV with captions on while at home. The other half watched with captions off. The researchers gathered pre- and post-data on oral fluency: specifically, timed reading of words and non-words, and timed reading of a passage. In an especially thoughtful move, they also used eye-tracking software to test how much time the children spent focusing on captions.

So: did the caption-on children become more (or less) fluent after six weeks? Did they spend more (or less) time focusing on the captions?

In a word: nope. Both groups read more words after six weeks, and read them faster. On both fluency tests, they made roughly the same amount of progress. To be statistically precise: the differences between these groups did not achieve statistical significance.

The eye-tracking data yielded interesting results as well. Here again, both groups spent a bit more time reading the captions…but the changes were roughly the same in both groups. This result strikes me as especially surprising; the captions-on group had been practicing reading captions, while the captions-off group hadn’t. And yet: both groups developed in the same direction at the same rate.

Intriguingly, more than half the time the children didn’t engage with the captions. They fixated on roughly 40% of the captioned words, which means they overlooked roughly 60%.

In brief, I think we can safely say: if the goal is reading fluency, this one study gives us no reason to think that captions help or harm young readers.

Pushing Back

People who champion the use of captions might offer several rejoinders to these conclusions:

  • One study is just one study. We need more research to draw strong conclusions.

I absolutely agree.

  • This study measured reading fluency, not learning. Captions could help fluent readers learn!

Very true. We don’t have research to support (or contradict) that claim, but it could be true.

  • Six weeks isn’t long enough. The benefits won’t show up until more time has passed.

This critique, while possible, strikes me as less plausible. During this time, the children watched — on average — sixty-six hours of TV. If a teaching strategy doesn’t help after 66 hours…well, it’s hard to feel lots of confidence that 120 hours will be different.

In this field, I’m often the guy saying “let’s slow down — that study doesn’t say what people claim it does.” In this case, I’m the guy saying “this study asks an important question, gathers appropriate data, and comes up with a clear and unhyped conclusion. We should take that conclusion seriously.”

At this point, the burden of proof lies squarely on those who claim that captions help children develop reading fluency.

Lopukhina, A., Cooper, H., Hsieh, C. Y., van Heuven, W. J., & Rastle, K. (2026). No evidence that same‐language subtitles improve children’s reading fluency. British Journal of Psychology.

Does a “Growth Mindset” Matter? Evidence from Half a Million Students
Andrew Watson
Andrew Watson
June 07, 2026

Does a “Growth Mindset” Matter? Evidence from Half a Million Students

Read More

Few ideas in education have seen a greater pendulum swing than Dr. Carol Dweck’s concept of Mindset. When she published her book in 2006, it became the must-read text that launched a thousand posters. Even ten years later, practically every school I visited had a wall of growth mindset catch phrases, and a block-capitals sign that read “…YET!”

In 2018, the backlash officially began. A pair of meta-analysis — authored by a team of respected scholars — could be summarized with the word “meh.” The findings could fairly be summarized as: “the effects are much smaller and less reliable than many advocates claimed.” In brief, according to this meta-analysis, the whole Mindset movement was a bust.

This anti-mindset perspective often takes on a harsh tone. I have colleagues who use the phrase “still believes in growth mindset” to mean “still thinks the earth is flat.” Some public critics claim that “mindset interventions work only when Dweck tries them.” This claim — which implies that Dr. Dweck puts her thumb on the research scales — is flatly untrue. (You might disagree with her conclusions, but if you’ve read any of her work you know how meticulous she is.)

A recent study adds to our understanding, and might at last help steady this pendulum.

PISA 2022

Every few years, fifteen-year olds from across the globe take the Programme for International Student Assessment — typically shortened to PISA. Because of its global reach, the number of students who take the test, and the demographic and educational information that it gathers, the PISA regularly offers a rich data-bank for analysis. One group of scholars looked at the most recent PISA to ask

  • Is growth midset a thing? More specifically, is a growth mindset associated with higher scores of mathematical ability?
  • Does it vary by country? If so, how?
  • Does socio-economic status matter? If so, how?

Because more than 500,000 students took the PISA that year, this research team has an ENORMOUS amount of data and can look for hard-to-detect effects. Before we get carried away by the reach of this data set, we should note a few limitations:

  • These data can find correlation, but not causation. That is: scholars can determine if students who have more of a growth mindset score higher on the math section, but they can’t determine if the mindset caused the higher score. (After all: greater ability in math could create a growth mindset.)
  • Scholars learn about the students’ mindset from exactly one question. Students rated their agreement with this statement: “Your intelligence is something about you that you cannot change very much.” The more that students agreed with that statement, the lower their growth mindset score.

These limitations noted, we can still recognize the potential strengths of this study. If a growth mindset correlates with mathematical ability, then this research should be able to detect even modest associations. If socio-economic status has an influence on that correlation, again, this study should recognize that connection.

74 Envelopes, Please

In one sentence, this study concludes: “The answer to those questions depends substantially on the country where the students took the test.” The relationship between growth mindset and mathematical ability varies meaningfully among the 74 participating countries.

In many Anglophone countries — Australia, New Zealand, the UK, Ireland, the US — a higher GM correlated with higher math scores. In other countries — Poland, Greece, Morocco, Saudi Arabia, the Philipines — the relationship between the two was statistically indistinguishable from zero.

The relationship between socio-economic status (SES) and GM also varied from country to country. In Latin America and Southeast Asia, for instance, high SES students with a GM saw greater math benefits than their low SES peers. A small group of countries — Singapore, Austria — saw the opposite pattern: low SES students benefitted more from a GM than their high SES counterparts. In the US (and other countries), socio-economic status didn’t particularly influence the relationship between GM and math achievement.

Local News

This global perspective helps us think more wisely about GM questions, but it also distracts us from the question that US-based teachers would like answered: where should the pendulum be? Here in the US, should we be on the “mindset is a scam!” end of the continuum, or closer to the “mindset posters for everyone!” end?

This study offers a very stats-y answer: “mindset correlates with math scores, but not lots-n-lots.” If you speak stats, you’ll be glad to know that r = 0.28. If you don’t speak stats, that means “growth mindset is one factor among many.” An r of 0.30 is regularly described as “a clear but modest tendency”; of course, 0.28 is just shy of 0.30.

To say all that in everyday words:

  • In the US at least, growth mindset isn’t nothing. It correlates — at least modestly — with math performance.

Based on that finding, I’ve got three suggestions:

First: we don’t know that GM training helps; this study doesn’t consider that question. But we do know that a GM and stronger math scores correlate with each other. For that reason, we should stop belittling mindset researchers.

Second: more specifically, Dan Willingham made this wise point: “we know that a GM seems to help in some circumstances, but not in many circumstances. We should try to understand which circumstances go in which categories.” (To be clear, those are my words, not his.)

Third: as I’ve written before, I think it’s unlikely that telling studetns about mindsets will have much of an effect. Instead, we need to change our school’s policies and procedures to align with growth-mindset thinking.

In sum: teaching is splendid but difficult work. We need all the tools we can get. If mindset can help — and, in the US, it seems to help a bit — then we should be open to and curious about that news. Over time, an incremental benefit can yield important results.

Dweck, C. S. (2006). Mindset: The new psychology of success. Random house.

Sisk, V. F., Burgoyne, A. P., Sun, J., Butler, J. L., & Macnamara, B. N. (2018). To what extent and under which circumstances are growth mind-sets important to academic achievement? Two meta-analyses. Psychological science, 29(4), 549-571.

Charoensilp, P., Kim, H., & Sriutaisuk, S. (2025). Relationships between growth mindsets and math achievement across socioeconomic status in 74 countries: Evidence from PISA 2022. PloS one, 20(11), e0337039.

Do Classroom Jokes Help Students Learn?
Andrew Watson
Andrew Watson
May 31, 2026

Do Classroom Jokes Help Students Learn?

Read More

Back in January, I wrote about a study on classroom uses of humor. The headlines:

  • On-topic humor increased student ratings of the teacher’s likeabililty and of their own motivation (and did a few other good things)
  • Making fun of students harmed relationships with students, and lowered motivation (and did a few other bad things)
  • Off-topic humor — jokes about something other than the class topic — didn’t have much of an effect on the measured variables

All this helpful information, however, came with a drawback: the study relied entirely on students’ ratings of how they felt. Such “self-report” data isn’t nothing, but it’s not the most persuasive kind of data. What would we find if we measured something more objective, like learning?

I find this question especially important because of a claim I occasionally hear at conferences: “research shows that laughter increases learning 44%.” Just imagine if that were true!

If we’ve got data about humor and laughter and learning one way or another, I’d love to see it. (By the way: I have looked at the source behind that specific 44% claim — it doesn’t hold up to even casual scrutiny.)

For all these reasons, I was happy to find a study that does look specifically at humor and learning. Sure enough, it’s a helpful addition to the discussion. As you’ll see, humor can be beneficial…but not 44% beneficial.

Eclipses, Lightning, and Pessimists

This two-part study, led by Dr. Lisa Bender, worked with both college students and middle school students in Germany. Participants watched narrated slide shows about science topics: “how lightning forms,” and “solar and and lunar eclipses.” All of those slideshows covered the same scientific information, but included important differences:

  • The control slideshows simply covered the information.
  • Version B included (non-humorous) examples: “Negative particles are like snowflakes falling to the ground.”
  • Version C included humorous examples: “Negative particles in a cloud are like pessimists — they are always down.”
  • Version D included humor that related to the topic, but didn’t illustrate the ideas: “If an electric car is struck by lightning, is it charged?”

The researchers did collect some of that self-report data I described above: the cognitive load of the slideshows, or the likeability of the instructor. Crucially, for my purpose, they gathered more objective data as well. Specifically, they tested the students to see how much they recalled, and how well they could transfer information.

44%?

Here’s what Bender’s team found:

First: for both the college students and the middle-school students, off-topic humor interfered (a bit) with recall and transfer. Specifically, the college students in the irrelevant humor group scored 9% lower than the control group on transfer questions; middle-school students in the irrelevant humor group scored 7% lower than the on-topic group on recall questions.

The authors suspected that irrelevant humor might function like a “seductive detail” — interesting information that distracts from the main learning goal.

Second: the on-topic humor didn’t make much of a difference either way. On average, students learned roughly the same amount in the control group and the relevant-humor group.

Notice that both of these findings flatly contradict the “laughter increases learning 44%” claim. If students are laughing about off-topic jokes, they’ll probably learn less. If they’re laughing about on-topic jokes, they probably won’t learn dramatically more.

Third: both off-topic humor and on-topic humor increased the students’ ratings of teacher likeability.

We don’t really know why humor had the effects it did; but we have these additional data points about the effects themselves.

You Be You

Before we read too much into these results, we should notice the limitations on this study. (All studies have limitations.) These studies lasted only a short time: four minutes for the college study, twelve for the middle-school study. And — unlike most classroom humor — the “jokes” were scripted and scored. In real classrooms, I suspect, humor typically arises spontaneously in the moment. It builds team spirit, and relies on in-group knowledge. (I had one group of students who wanted me to call them “the rodents” — it would take too long to explain why.)

For all these reasons, I think we can mark out “humor in the classroom” as a topic where research will struggle to provide authoritative classroom advice.

  • Yes, I think it’s straightforwardly wrong to claim that “laughter increases learning 44%”; nothing does. (If any one thing increased learning that much, teachers would have figured it out already.)
  • I likewise think that off-topic humor is probably a bad idea during the middle of a cognitively challenging part of a lesson. However, if we’re welcoming students to class, or transitioning from one part of a lesson plan to another, or — heck — just need a break, throwing in a random humorous observation sounds just fine.

The best teaching advice isn’t “try to be funny,” or really “try to be anything.” We should be a professional version of ourselves. That genuine self-presentation will be the best starting place from which to help students learn.

Bender, L., Renkl, A., & Endres, T. (2026). Punchline with (out) purpose: Integrating research on instructional humour and seductive details. British Journal of Educational Psychology.

AI and Learning: Beyond “Good” vs. “Bad”
Andrew Watson
Andrew Watson
May 24, 2026

AI and Learning: Beyond “Good” vs. “Bad”

Read More

As AI changes the education landscape moment by moment, we can expect a steady flow of studies offering us guidance about its use. We’re happy to have all that research, but it does come with a probable danger. Every month or so, we will likely hear that a study has DEFINITIVELY concluded

  • AI damages our long-term ability to learn, or
  • AI revolutionizes classrooms to benefit students, or
  • AI promotes collaboration while fostering independent thinking, or
  • AI harms neurons and corrupts synapses…

Like Celine Dion’s heart, this list will go on and on.

In this blog post, I want to quickly summarize two AI studies which — unless carefully parsed — reach contradictory conclusions. My point will be not that one is wrong and the other right, but that each study has asked meaningfully different questions and measured results in a substantively different way. BOTH of these studies offer us useful guidance…as long as we resist the temptation to cry: “Look! This study gives us a clear-cut, definitive answer to our question.”

I’ll end by arguing that we should focus not on “AI good” vs. “AI bad,” but on “AI that replaces thinking” vs. “AI that fosters thinking.” Let’s start with those two studies.

Study #1

In a recent study by Dr. Andre Barcaui, a group of 120 undergraduates at a Brazilian university spent two weeks preparing presentations for their classmates. Half of those students did use AI to assist them in this work; the other half didn’t. Then — SURPRISE — 45 days later all students took a follow-up test on the material in their presentations.

The students in the AI group averaged 57.5% on that test; the students who didn’t use AI averaged 68.5%. That’s quite a difference! (If you speak stats, the Cohen’s d was 0.68.) The one sentence summary: “students who study with AI remember less than those who don’t.” If you want a hyped-up version:

Using AI harms learning.

Like all studies, this one has methodological strengths and weaknesses.

Strengths: most education research tests “learning” a few hours later. This one tests learning 45 DAYS later. That schedule is rare, and makes it more reasonable to talk about “learning.” If students recall more information after a month and a half, we can plausibly say they “learned more.”

Weaknesses: To my mind, the biggest weakness here is that we don’t really know what the students DID with, or without, AI. How exactly did they use AI tools in preparing? Did they ask for summaries or scripts or graphs? Equally important, we don’t know that the students in the non-AI group actually avoided AI. They could have used AI back in their dorms…how would the researcher know?

All that being said, Barcaui’s study makes rough-n-ready sense within a cognitive science framework. Basically speaking, we can assume that students in the AI group let ChatGPT or Claude “do more of the thinking for them.” On the other hand, students in the non-AI group did the thinking themselves. No one should be surprised that the students who thought more learned more.

One simple way to highlight this point: students in the AI group averaged 3.2 hours of prep time for their presentations; those in the non-AI group averaged 5.8 hours. If I spend twice as much time thinking deeply about a topic, I’m likelier to learn it. If AI reduces the amount of time I think, I don’t learn as much.

Study #2

A second study, led by Dr. Angel Tsai-Hsuan Chung, seems to reach the opposite conclusion. In this study, more than 1000 Taiwanese high-school students took a five month course in Python (the programming language, not the British comedy troupe). All of the students used an AI tutor to help them solve problems, but that tutor came in two distinct versions. The basic version gave students problems in a consistent order: easy, then medium, then hard.

The enhanced version gave problems depending on the student’s current level of understanding and “productive struggle.” That is: the enhanced AI analyzed the student’s work — the quality of their questions, the correctness of their edits, and so forth — to determine how hard the next question should be. Students who succeeded at one level of difficulty advanced to the next level; those who struggled with a problem got an easier one next. In this way, the enhanced AI tutor kept the challenge in a “desirably difficult” range.

Sure enough, at the end of the course, students who practiced with the enhanced AI tutor scored roughly 0.15 standard deviations higher on the final exam. This number is hard to translate into a non-stats framework, but let’s put it this way: the difference wasn’t huge, but students would certainly be happy they scored those few extra points — or disappointed if they didn’t.

We hyped up that first study by saying “using AI harms learning.” We can hype up this study by saying

AI-enhanced studying will transform education.

Two Conclusions

On first reading, the Barcaui study and the Chung study seem to point in different directions: the first basically “anti-AI” and the second basically “pro-AI.” On eX-Twitter — where I first learned about these studies — they were held up as the last word on the question of AI in education.

  • “This Barcaui study shows why we must ban AI from our classrooms!”
  • “This Chung study shows that AI will be essential for the future of learning!”

I myself draw two different conclusions.

First: both studies reinforce a core conclusion from cognitive science. If we want students to learn, we should do everything we can to cause them to think harder — but not too hard — about that topic. In the Barcaui/class presentation study, students who let AI do the thinking for them remembered less. In the Chung/python study, AI tutors that ramped up the thinking challenge helped students learn more.

In other words: AI is a tool like many others. It can be used badly — to replace thinking — or well — to prompt thinking.

Second: This pair of studies, and the over-hyped conclusions drawn about them, highlight an ongoing danger in our field. I’m confident that, with some frequency, we will hear that a new study provides the last word on AI: we must be all in or all out. Books and newspapers and blogs will make sweeping claims. Teachers and school leaders will be tempted, or encouraged, or required, to make passionate commitments in one direction or the other.

Rather than take an absolute stance, I think we should look carefully for useful, specific guidance.

  • “This classroom use of AI requires students to think more deeply (that’s good!), but has the potential to distract them (that’s bad).”
  • “This AI tool reduces WM load of a complex assignment (that’s good!), but isolates and demotivates students by reducing their connections with each other (that’s bad!)”

Over time, we will learn how a specific version of AI affects particular cognitive functions in certain ages, grades, disciplines, and cultural contexts. That growing, ever-shifting body of guidance will help us decide when AI support enhances learning, and when another tool — a book, a pencil, a mini-whiteboard — better serves our students.

A final note: after I wrote the blog post above, I came across an analysis by my friend Dr. Ian Kelleher whose thinking substantially overlaps with mine. If you’d like to see his more fleshed-out thinking, you can find it here.

Barcaui, A. (2025). ChatGPT as a cognitive crutch: Evidence from a randomized controlled trial on knowledge retention. Social Sciences & Humanities Open, 12, 102287.

Chung, A. T. H., Zhang, B., Kung, L. C., Bastani, H., & Bastani, O. (2026). Effective personalized AI tutors via LLM-guided reinforcement learning. Available at SSRN.

Mini-Whiteboards Work. Participation Is the Point.
Andrew Watson
Andrew Watson
May 17, 2026

Mini-Whiteboards Work. Participation Is the Point.

Read More

If you have a colleague who uses mini-whiteboards, you know the passion that these simple tools inspire. According to the fervent accounts I read on Twitter and hear at conferences, MWBs increase student participation, reduce teacher stress, and cure most cases of bursitis. (I might have made up that last one.)

Up to now, I’ve seen pro-MWB arguments relying on two kinds of support:

  • Experience. “When I switched to MWBs in my classroom, I immediately noticed changes X, Y, and Z.”
  • Common sense: “If all my students answer a question, that’s clearly better than just one of them answering a question. The participation ratio improves to 100%!”

Today, for the first time, I came across a meta-analysis looking at research into the topic. Its conclusions – and limitations – make for helpful reading.

Here’s the story.

What They Asked, What They Found

This research team, led by Dr. Robbie Marsh, identified 29 studies – including just over 400 K-12 participants — exploring the use of MWBs. (Marsh’s team also included studies looking at pre-printed response cards.) They reviewed these studies to answer four questions.

Compared to classes where students raised hands to answer questions, did MWBs

  1. Improve class participation?
  2. Reduce off-task behavior?
  3. Raise scores on quizzes (small, teacher created assessments soon after instruction)?
  4. Raise scores on tests (more substantial assessments given after more time had passed)?

The highlights – which you can find in table 2 – suggest unambiguously good news for Team MWB. On average,

  • Participation was almost 60% higher in the MWB class.
  • Off-task behavior fell by more than 25% (or, on-task behavior rose by more than 25%)
  • Quiz scores increased 18%
  • Test scores increased almost 14%

Marsh and Co. checked to see if the results were the same in both general education and special education settings. The answer: basically yes. More precisely: students in special education settings saw higher academic gains. Those in general education classes saw more behavioral improvement.

If you’re a MWB enthusiast, this meta-analysis gives you many, many reasons to celebrate.

Before We Get Carried Away…

While this meta-analysis clearly supports the use of MWBs, I don’t think its findings require us all to adopt them in our own teaching. I’ve got two stats-y reasons, and one conceptual reason, to resist any such call.

First: You might have noticed that the 29 studies included 405 people. In other words, the average sample size in each study was 15. That’s TINY. We have many studies, but not many people in those studies.

Second: While the averages noted above are compelling, Marsh’s team found a wide range around those averages. In some cases, MWBs helped A LOT. In others, they didn’t make much difference. (Almost none of the studies found that MWBs reduced good outcomes.) We don’t have enough data to know why they helped a lot in some cases and only a little in others. Simply put: your mileage may vary.

(For the stats-minded, I’m commenting on “heterogeneity” here: the I2 clocks in at 98% to 99% on these four categories I’ve listed.)

Third: These statistical cautions aside, there’s a bigger pedagogical point worth keeping in view.

Marsh’s meta-analysis shows that students do better when they respond frequently during instruction. Calling on raised hands doesn’t get that job done.

Mini-whiteboards do — but not because they’re special. They work because they make it easy to get everyone responding, all at once. And they’re not the only way to do that. Turn-and-talks, cold calling, clickers, do-nows—all of these can increase the number of students thinking and responding during a lesson.

So the real lesson here isn’t “use mini-whiteboards.” It’s this:

Move away from hand-raising and toward routines that keep everyone responding.

Once teachers make that shift, MWBs are one good option among many — not the only answer.

Marsh, R. J., Cumming, T. M., Randolph, J. J., & Michaels, S. (2023). Updated meta-analysis of the research on response cards. Journal of Behavioral Education, 32(3), 450-473.

The Problem with “Students Teaching Students”
Andrew Watson
Andrew Watson
May 10, 2026

The Problem with “Students Teaching Students”

Read More

Should our students teach their peers? The obvious answer to this question is: “yes, of course.”

Experience shows that teaching leads to greater understanding for the person who did the teaching. After all, when I figure out how to explain “tragedy” to my sophomores, I end up knowing more about tragedy than I did before. No doubt my students would benefit from such experiences.

In this blog post, I’m going to make a surprising pair of arguments:

  • We should ask students to explain ideas and topics, but
  • We shouldn’t ask students to teach those ideas and topics.

Here’s why.

Theory and Practice

Educators regularly hear that we should ask our students to teach one another.

One often-discussed version of this approach: the “jigsaw method.” In this pedagogical strategy, the teacher divides a topic into several jigsaw pieces. For instance, if I’m teaching the digestive system, I can divide it into several sub-topics: the stomach, the pancreas, the small intestine, and so forth.

Next, I assign these jigsaw pieces to pairs of students. Once each pair has mastered their topic, they all circulate and teach the other pairs. In this way, all my students reassemble the jigsaw by teaching each other.

Over the years, teachers have offered me other examples.

  • When parents asked “how can I help my child study when I don’t understand the math they’re doing,” one teacher suggested: “have your child teach the concept to you. She will understand the concept better because she did.”
  • A middle-school biology teacher explained: “I often begin class with a warm-up exercise: one student teaches a shoulder partner the concepts we studied yesterday.”
  • A kindergarten teacher offered a fun example: “for one unit, my students each chose a community helper — a postal worker, a crossing guard, a firefighter — someone like that. They then taught the whole class about that community helper.”

No doubt you can think of many (many) other examples of students teaching students.

Definitions Matter…

As I’ve written before, our work often benefits when we stop and focus on precise definitions. Let me offer the most basic possible definition of the verb to teach:

“To teach is to cause someone else to learn.”

In the examples above, we’re inviting students to explain a topic to a second person: a parent, a shoulder partner, the whole class.

We are not – by the definition above – inviting students to teach a topic to a second person. In practice, we’re not measuring their success by someone else’s learning.

  • If the parent doesn’t ultimately understand the math concept, that’s okay. The goal was to help the child learn, not the parent.
  • If a kindergartener’s classmates don’t learn about postal workers or firefighters, no worries. The goal was to help each child learn about a specific community helper, not for everyone in the class to learn about all the community helpers. (I know because I asked.)

We have lots of research showing that explaining a topic to another person helps people learn. In fact, I recently read a study (Nestojko 2014) showing that students who thought they were going to explain a topic to another person learned it better – even though they never did the explaining. Simply planning to explain produced modest learning benefits.

If you’ve heard about generative learning, you know that this strategy fits neatly into that category. When students select, organize, and combine ideas, they learn those ideas better. Students might select, organize, and combine by

  • Drawing a picture
  • Creating a mind map
  • Acting out a scene, or
  • Presenting an idea to their classmates

When they do any of this generative mental work, students learn more.

But – to be precise – the goal of that last option is that the student who is presenting learn more about the topic. If their classmates haven’t learned, the presentation still counts as generative learning because the presenter learned.

… and Learning Matters

From one perspective, this distinction might seem merely fussy. If students are explaining ideas to one another, what’s the problem with calling that teaching? Explaining is close enough to teaching, isn’t it?

I myself don’t think so. Simply put, explaining – by itself – isn’t teaching. Simply hearing an explanation does not reliably lead to learning – and “causing someone else to learn” is the definition of teaching.

Let’s consider the jigsaw example above. At the end of that class, all the students will have heard explanations about the digestive system’s components: stomach, liver, large intestine. Those segmented explanations, however, will rarely add up to an understanding of digestion as a system.

Understanding that system requires not simply an organ-by-organ review of facts: “hydrochloric acid in the stomach breaks down proteins and kills off bacteria.” Instead, students need to see the connections among all those organs. Unless this lesson “causes students to learn” how those organs and functions add up to accomplish the goal of digestion, they have not been taught. They have heard explanations, but those explanations weren’t enough.

To be clear: I can help students learn by asking them to explain. That generative work — selecting, organizing, and expressing ideas — has real cognitive benefits.

But explaining isn’t the same as teaching.

  • The goal of explanation is that the speaker learns.
  • The goal of teaching is that someone else learns — and accomplishing that goal requires more than explanation alone.

When I ask students to explain, I’m giving them a powerful way to strengthen their own understanding. When I want their classmates to learn, however, that responsibility ultimately sits with me.

Nestojko, J. F., Bui, D. C., Kornell, N., & Bjork, E. L. (2014). Expecting to teach enhances learning and organization of knowledge in free recall of text passages. Memory & cognition, 42(7), 1038-1048.

It Works, but Is It Right? Incentivizing Sleep
Andrew Watson
Andrew Watson
May 03, 2026

It Works, but Is It Right? Incentivizing Sleep

Read More

Sleep is the wonder drug we can all afford. It reduces stress and depression, increases concentration and academic performance, lowers blood pressure, fosters self-regulation, and provides health benefits too numerous to mention. For all these reasons, we’re excited when we find strategies that reliably help students get more sleep.

For instance, several months ago I wrote a blog post about a sleep-enhancing strategy. When students a) anticipated the problems that might interfere with their sleep, b) created plans to solve those problems in advance, and c) pledged to USE those solutions, they ended up getting more shut-eye. Even better: those changes lasted — even eight months!

I’ve found another study with an alternative approach. I’m impressed with its findings, but — honestly — unsettled by its method. Let me explain.

Gathering Data

Researchers in Pittsburg worked with over 1000 college students; before the study, those students slept an average of 6.6 hours per night.

This study tried several different strategies to extend students’ sleep time:

  • personalized bedtime reminders
  • feedback in the morning
  • a special incentive — either immediate or delayed (more on this later)

And, of course, the researchers tracked all sorts of data:

  • How much sleep did students get? How consistent were they with their bedtime?
  • How much time did they spend on screens before bed?
  • What happened to their grades?
  • Did students’ cognitive function or emotional state change?
  • How long did any changes last?

By the way: students wore a fitbit to track their sleep. In other words, the researchers didn’t rely on self-report — always an unreliable measure.

Exciting News

This research team found one approach that worked. The students who got bedtime reminders and an IMMEDIATE incentive experienced measurable gains.

  • They slept almost 20 minutes more a night. That number isn’t huge, but it meant that these students were noticeably likelier to reach the researchers’ 7-hour-per-night target.
  • Their grades went up a bit. The average improvement was not quite enough to move a B average to a B+ average. Grades in STEM classes rose more — roughly a third of a letter grade.
  • Students reported that they spent less time on screens before bed, and dealt with stress more effectively.
  • Some of these changes lasted, up to a point.
    • Sleep schedules remained more consistent after the study finished.
    • Students slept more than they had before the study began (but not as much as they did during the study).
    • The GPA changes lasted a term — but not two terms.

Now, not all the news was good. When researchers directly measured cognitive function (specifically: math and creativity), they found no changes. And, students’ level of physical activite, depression, and anxiety didn’t change.

Given the vital importance of sleep, we could well be delighted to get even these modest benefits.

Persistent Doubts

So far, I’ve glossed over a key point: how did the researchers “incentivize sleep”?

Simple: cash. Students who slept 7 hours a night — as confirmed by the fitbit — earned $4.75 per night. And the students in the “immediate” incentive group got that money right away.

While I’m impressed with the quality of the study overall, I confess I’m unsettled by a “cash-for-behavior” approach to changing students’ sleep.

Let’s pause to think over other research-informed strategies: retrieval practice, or exercise, or fostering relatedness. We have lots of psychology or neuroscience research to explain their benefits. For that reason, they strike me as ethically uncomplicated. If they work — and especially if they align with our school’s teaching philosophy – we can go ahead and try them out.

This research creates a special case: a strategy with a dollar sign next to it. For that reason, it strikes me as ethically much more complicated — especially for younger students. If I were a school leader, I don’t think I’d be ready to pay students to sleep.

I should admit, however, the weakness in that ethical position. Schools spend substantial sums to encourage better health and study practices: advising programs, merit scholarships, wellness centers. I can predict a reasonable question: “why is it wrong to spend money on students, but not give them money directly? Why does the middleman make the expenditure ethical?” In fact, as the study’s authors point out, this incentive program got bigger results for much less money.

I don’t have training in ethics, and so I don’t have a sophisticated answer to that question. But there is a difference between spending money to help students and giving money to students. At least for now, that difference brings me up short.

Let me make the same point a different way. A few weeks ago, I found research about creating concept maps. The study struck me as persuasive, so I passed along its guidance:

Students benefit when they work on concept maps alone before collaborating with others. (Caveats apply.)

In this case, the study strikes me as persuasive, but I am NOT comfortable encouraging you to embrace its findings. Even though this approach might work, reasonable school leaders may decide it’s not the right thing to do.

To Sum Up

This study offers a strategy to help students sleep more.

At the same time, it raises bigger questions about the ultimate goals of schooling. If we want students who succeed academically and improve their overall health, we can pursue this incentive approach. If we also value students’ ability to steer their own lives and make healthy choices, we may well hesitate to distribute those cash incentives. The time we spend wrestling with that ethical question will ultimately make us better at our work.

Giuntella, O., Saccardo, S., & Sadoff, S. (2024). Sleep: Educational impact and habit formation.

Older posts