Is it the screens? Or education systems?

The Dark Ages implies television and phones are the main cause of cognitive decline. This fails to explain the patterns in PISA scores. Why did England and Scotland fall so precipitously from 2000 to 2005 whilst America improved? Why did England and Estonia hold steady after 2015 whilst most other OECD countries declined? How have Singapore, Taiwan, Japan avoided decline altogether?

A better explanation is that a country’s education system is more important than its television diffusion.1 East Asian PISA and IQ scores have probably remained constant, or even risen, because of their rigorous education systems and intensive tutoring cultures. The two European countries which avoid PISA-malaise – Estonia and England – have more rigorous education systems than their neighbours. They (more or less) use the knowledge-rich curricula, direct instruction, and systematic phonics – techniques which their more progressive neighbours abandoned between 1975-1990.

Here is much more from Alexander Thompson, recommended.

Comments

Oh great, then just give kids "smartphones" willy nilly. See? Studies show!
At least television couldn't be watched during class, walking down the street, or alone in bed at 3am. Obviously the constant distraction from phones and fragmentation of everyone's attention into a million bits could never cause a decline in PISA scores. They just need more "AI" to be more "intelligent."

Respond

I am begging Tyler to go observe an environment of kids that isn't filled with precocious teenagers who want to delve deep into the arts or start their own NGO by the time they're 18 years old but one where less than 50% of their parents have college degrees. It will radicalize you how those kids use their phones.

Respond

How kids in that situation use their phones doesn't matter. The point is: would they be doing any better without them, or would they just be engaged in some other useless activity?

Respond

The weakness of this hypothesis is that changes in PISA scores are very uneven across countries. If screen time is the main culprit, why is the outcome so uneven?

Respond

Isn't the trivial answer that the change from screens is roughly constant, modulated by culture, and the changes from that constant decline between countries are related to other factors? After all, didn't PISA scores move up and down between countries before smart phones? What was the explanation for scores to change between countries back then? Wouldn't it make sense that those same processes continued, overlaid on the background of increasing phone use?

Respond

I don’t pretend to have a theory of PISA scores across countries. :)

My narrow point is that if changes in screen time are similar across countries—which is just my intuition—while changes in PISA scores differ a lot, screen time cannot have much explanatory power. It may, of course, still partially explain a common decline.

Respond

I could see this argument being right, but the solution that keeps the screens in the schools is not feasible: change your whole education system, which we've been trying to do for how long now? Sometimes economists forget about politics.

Respond

Not as relevant to the longer trends but from what I looked at, it seemed Japan did not actually avoid decline on the 2025 vs 2022 results. Their scores looked about 10.7 points down on average across the Maths+Reading+Science scales, which is pretty comparable to the loss of 10 points in the Netherlands, 10.8 points in Germany or 11.5 points in Spain. Things are a lot better in Japan when you consider 2025 on 2018, because they actually improved in 2022 (which perhaps reflects a difference response to the pandemic conditions, using it for learning?). But the 2025 on 2022 overall decline seems quite similar.

The PISA authors don't make many observations about schooling systems (perhaps they should?) but they do note the following in their report:

"Within the PISA reading test, the most rapidly declining success rates were observed on tasks that used longer texts as stimuli. Between 2018 and 2022, success rates declined in particular on reading tasks involving higher-level processes, such as evaluating and reflecting, which require careful reading, remembering and integrating multiple pieces of information, sometimes across several sources.

In contrast, the least affected tasks were those requiring students to locate a single piece of information. The latter tasks can often be solved through more targeted engagement with the text. Over a longer period (2018 to 2025), declines are more uniform across all types of tasks but difficulties related to text length remained a strong driver of the overall decline (Annex A1).

PISA results also suggest, more broadly, that the decline could be related to students’ capacity or willingness to sustain focused engagement with demanding reading tasks (Annex A1):

- Performance declines are already visible at the very start of the reading assessment, in the fluency items that measure how quickly and accurately students process simple written statements. The proportion of accurate and fluent readers declined by about seven percentage points between 2018 and 2025. The proportion of hasty readers – those who gave fast and incorrect responses – increased by about five percentage points, on average, from almost 7% in 2018 to 11% in 2025.

- The proportion of hasty responses – fast and incorrect – also increased significantly in the rest of the reading assessment over the same period, on average. At the same time, the proportion of non-reached items declined. This suggests that students became more inclined to answer quickly rather than to leave difficult items unanswered.

- Performance declined throughout the reading assessment, but more strongly in the later sections of thetest than in the early sections. This appears consistent with the above two findings: it suggests that students’ attention, persistence or confidence when facing difficult tasks declined over time.

Students’ self-reported effort, in PISA, and their general attitudes towards learning and school, tended to also decline:

- In most countries and economies, the effort that students reported having spent on the PISA test declined in 2025 compared with 2022 and 2018 (Annex A1)."

"Gen Z" is broadly much more cynical about doing tests where there's nothing "in it for me"? Generations are usually fake but its striking the decline happens immediately in the PISA round after the youngest Millennials (born 1996) reach the PISA assessment age.

Respond

Scores from England actually, arguably, were produced by two different systems. Pre-2016 (less rigorous) and post-2016 (more rigorous assessments).

Yet the results from both are almost the same. Almost like all that extra work is just to overcome some “new” environmental factor.

Respond

>Here is much more from Alexander Thompson, recommended.

Much more indeed, thank you. I think you in effect undersell the essay by excerpting this passage.

Best part is the social media post by "Gurwinder" with the graphic animation of how time was spent over the years 1920-2026, with the comment:
"The most underrated change is the loss of time spent 'doing nothing' (i.e. thinking)."

Amen.

But as for the screens, I'm with Marriott (rather than Thompson, or for that matter yourself (TC)). Thompson:

>Could it be television? This would be a little odd since the peak generation was born in 1975, well after television entered our homes.

The problem is actually color television, which took hold around the time the great stagnation started. I don't think black-and-white television damages literacy nearly as much (and radio I suspect actually improves it). OTOH, the Internet, and social media above all, is even worse than color television -- like a more extreme form thereof.

Mind you, these technologies (TV and internet) can be used in a literate way, but it's not the default, so few do. (Whereas with radio, I think it actually is.)

Respond

>I don't think black-and-white television damages literacy nearly as much...
Interesting. My high school English teacher believed this - back in the 1970s. They only had a black and white TV in their house. This is the first time I've heard anybody ever mention it since - cool.

Respond

I would not equate "doing nothing" with "thinking". Furthermore, the graphic seemed to show that TV and other entertainment replaced housework, cooking, and eating rather than replacing "thinking".

Respond

>I would not equate "doing nothing" with "thinking".

Talk to creative people. "Thinking" very often looks like doing nothing, or doing something stupid, from the outside. Terry Tao has some nice anecdotes about this. (Thank goodness these slackers are about to replaced by AI...)

But more to the point: whether you call it "thinking" or not, it's that "nothing" time that trains one to face the existential void and produces the depth of personality necessary to be capable of nontrivial thought in the first place.

Respond

"it's that "nothing" time that trains one to face the existential void"

:) That's impressive sounding. Wrong. But impressive sounding.

What I hear from most people is that they get their best ideas doing *something* - walking, and chopping wood are two activities that come up a lot. Neither takes much brain power. The repetitive activity with low cognitive demand is the trigger.

Respond

Yeah, well those who don't use much brain power chopping wood end up with problems walking.

Respond

ha, yeah, over half a century and I'm still walking :) Good training I guess...

Respond

>Do college students even know that Strunk and White is not a music duo?

If they were, then they would be especially unfamiliar to the current generation...

Respond

It's also terribly dated, like Great Books curricula that were defined 100 years ago... An up-to-date, enlightened take on the humanities is not a heresy.

Respond

Given that the neighbours of England and Estonia abandoned direct instruction etc. in the period mentioned (1975-1990), I should be quite curious to learn how that is relevant to their trajectories post-2000.

Respond

The answer has been floating around on X an is immigration...
The native score of most countries is still good, while the immigrant score (including 2nd gen or 3rd gen) is low. Of course the larger the share of immigrants and sons of immigrant gets, the lower the score will be.

(1) Why England and Estonia hold steady after 2015 whilst most other OECD countries declined? --> England's immigration has been high skilled, Estonia has very little low skilled immigration.

(2) How have Singapore, Taiwan, Japan avoided decline altogether? --> also very little low skilled immigration

Respond

I don't think so, not exactly.

The OECD published a big Excel table (Table I.B1.2d.9) where they give the scores across Maths, Reading and Science of children with migrant and native backgrounds, from 2015-2025.

If I do a weighted average of Europe (all the European countries in PISA), then overall from 2018 to 2025, I find that the overall weighted score of "Europe" declines by 17.3 points in this interval, while native children in Europe decline by 12.2 points.

So the decline would still be ~58% as large absent the contribution of migration.

(This does make the native decline more similar between the US and the European aggregate; -11.5 points in the US over this interval).

You might be able to advance a thesis that it's Covid-19+immigration, and the migration contribution is significant. But I don't think migration on its own does the job.

Respond

True, but this seems to provide little consolation, given that (1) Today's immigrants are tomorrow's non-immigrants, and (2) migrants tend to have higher fertility rates.

Respond

England's immigration has NOT been high skilled, which we will start to see come through into the PISA results if the unskilled immigrants raise families in England.

Respond

“ The native score of most countries is still good, while the immigrant score (including 2nd gen or 3rd gen) is low. Of course the larger the share of immigrants and sons of immigrant gets, the lower the score will be.”

Some immigrants and their children do very well, and you can tell when they get here which ones they’ll be.

For example, immigrants who come here as STEM graduate students.

Check out the names here (of the top 25% on National Mathcounts, which is the most important math contest for middle schoolers):

2026 RTX MATHCOUNTS National Competition
Monday, May 11, 2026, Orlando, FL

https://storage.pardot.com/980762/1778523075mkqGTf4N/2026_Final_Standings_Document__post_CDR_.pdf

Probably only a couple of those 56 kids were not the children of recent immigrants, mostly from Asia.

Respond

Out of those 56 kids, there are two whose names suggest that maybe they weren’t the children (or maybe grandchildren) of recent immigrants:

James Stewart

Catherine Pennington

But it’s quite possible that at least one of their moms was an immigrant, maybe a Chinese tiger mother.

And number 56 is an interesting anomaly:

Arcel Ieuan Garcia

We have A LOT of Hispanics in America, but that’s the only name that seems Hispanic, and I’ll bet the kid’s parents did not come here illegally to pick tomatoes or hang drywall.

Respond

Check out the names of the US IMO team in 2026 and 2025:

https://www.imo-official.org/results/team/year/2026/country/USA/

https://www.imo-official.org/results/team/year/2025/country/USA/

Hannah Fox image here:

https://www.egmo.org/people/person2636/

Hannah’s dad is a Stanford math professor. She looks like her mom is probably Chinese.

There’s also Feodor Yevtushenko in 2025 (Ukrainian?), but the other names are all Asian.

Of course this is a phenomenon from which well-mannered people politely avert their eyes.

Respond

The mean PISA score is not determined by the few superbright students.

Respond

“The mean PISA score is not determined by the few superbright students.”

But as you would expect by the astounding Asian domination at the top level of US math contests, Asians also do quite well on average. See the graphs here (and click on them to enlarge):

https://x.com/cremieuxrecueil/status/1732087511327908128

For example:

https://x.com/cremieuxrecueil/status/1732087511327908128/photo/1

On every test, for US students it’s:

Asians > Whites > Hispanics > Blacks

Respond

“The mean PISA score is not determined by the few superbright students.”

Mean PISA score is important, but when it comes to progress in math and science and technology (which I for one think is a big deal), it’s the superbright ones who matter.

Respond

This is an old debate in America, old but it arose long after I had finished my formal education. For example, students are not taught "long division" (algorithms) or how to diagram a sentence. How can one develop critical reading skills (and its cousin, critical thinking skills) without an understanding of the different parts of a sentence. Is the Harbrace College Handbook even published any more? Do college students even know that Strunk and White is not a music duo? While I place part of the blame for cognitive decline on the screen, the problem goes deeper: the screen addiction is a symptom not the cause. It's the superficial, stupid!

Respond

Have people investigated whether said education system is effectively teaching to test for PISA?

I was educated in 90s and my folks in the 60s. They’re better at wrote type stuff, my siblings and I better at reasoning (imo, anyway)

Respond

It pretty much is this; transference model of education. Very limited benefits outside of standardised testing. More benefits in inquiry-based approaches (which of course include direction instruction elements).

Respond

“ The native score of most countries is still good, while the immigrant score (including 2nd gen or 3rd gen) is low. Of course the larger the share of immigrants and sons of immigrant gets, the lower the score will be.”

Some immigrants and their children do very well, and you can tell when they get here which ones they’ll be.

For example, immigrants who come here as STEM graduate students.

Check out the names here (of the top 25% on National Mathcounts, which is the most important math contest for middle schoolers):

2026 RTX MATHCOUNTS National Competition
Monday, May 11, 2026, Orlando, FL

https://storage.pardot.com/980762/1778523075mkqGTf4N/2026_Final_Standings_Document__post_CDR_.pdf

Probably only a couple of those 56 kids were not the children of recent immigrants, mostly from Asia.

Respond

Culture encompasses both, and much more.

Respond

Maybe it’s not just about having sufficiently well-paid and (thus?) motivated teachers, screen time, etc., but also about competition among students?

Poland has two nationwide standardized exams—after 8th grade and at the end of high school (there used to be three). Same dates, same hours, same questions. The first is a decisive criterion in admission to public high schools; matura scores, from the nationwide exam administered at the end of 12th grade, are the basic criterion for university admission. Students can apply to schools anywhere in the country—a very visible tournament.

Poland / OECD, PISA mathematics:

2003: 490 / 500
2006: 495 / 498
2009: 495 / 496
2012: 518 / 494
2015: 504 / 490
2018: 516 / 489
2022: 489 / 472
2025: 484 / 463

Screens may matter. So may incentives.

PS. Presumably, staff at Anthropic and OpenAI have sizeable screen time. I am scared to think what progress in AI we would see if they did not. :)

Respond

According to Google's AI Overview automated brainiac: PISA scores do not measure or purport to measure "rates of adult literacy" ("literacy outright", basic/functional literacy, sub-literacy [the last two rather fluid categories], and "illiteracy outright").

No: the categorizations I just adduced are not precise--they are simply "human terms" for assessing or appreciating adult reading ability. Quite likely, they do not and would not suit the purposes of academicians or statisticians.

Perhaps the following account could begin to suggest adequate metrical categories:

http://fictionaut.com/stories/strannikov/fire-skates

Even if this account fails to suggest relevant categories for assessing adult literacy, it cannot fail to possess enormous explanatory power for the rates of adult literacy that have been (and, apparently, continue to be) accepted at least here in the US--as such standards have been accepted for several (more than three) decades continuously.

Just think for one moment, brainiacs and brainiac-aspirants: how is it that no American can cite even the roughly accurate percentages on any given day, in any given year, for the standing rates of adult literacy, adult functional literacy, adult sub-literacy, and adult illiteracy? The stark numbers are not advertised with even annual regularity. These data are next to impossible to find or discover--on any given day, in any given year. That they are so difficult of access begins to suggest exactly why they should embarrass our "most educated classes", the very classes of Americans who have corrupted American public education by offering every pre-digested pedagogy and ill-conceived brand of indoctrination--except sound training in reading, writing, and math permitting reasonably accurate measurable outcomes.

Distinctly odd.

Respond

“These data are next to impossible to find or discover…”

One reason is that the data you refer to would be extremely racist.

Respond

Minor cx(x).--". . . the data you refer to would be deemed (or: permitted to be deemed) extremely racist".

That could well be the case, considering how our unelected managers of public discourse and those commandeering our First Amendment behave typically. (They seem to have their own peculiar interests in suppressing adult literacy and numeracy--who'd've thunk?)

Respond

It's the students and the teachers but we've decided to stop talking about that.

Respond
Respond

Add Comment