What Is Cognitive Load Theory? (And Why Reading Hard Things Feels Like Drowning)

Cognitive load theory, proposed by John Sweller in 1988, says working memory has a hard ceiling of roughly four chunks. Any task pulls from that finite pool. When the total mental effort needed exceeds the pool, comprehension collapses. Sweller split the load into three sources: intrinsic (the material's own difficulty), extraneous (difficulty added by presentation), and germane (effort spent building understanding).

You open a paper. Three pages in, you realise you've read the same paragraph four times and you still don't know what it says.

Your eyes are moving. Your finger is tracking the line. The words are landing on your retina in the right order. Nothing is broken on the input side.

But the meaning isn't sticking. It's like pouring water into a cup that's already full. Each new sentence pushes the last one out before it can do anything useful.

You blame yourself. You're tired. You're distracted. Maybe you're not smart enough for this. You close the tab and tell yourself you'll come back when you can focus properly.

You won't. Or you will, and the same thing will happen again.

The moment I stopped blaming my focus was on 20 May 2025. I started building what would become Alexandria that day. Alexandria is the reading platform built for retention, not consumption: synced word-by-word audio, structured knowledge capture, spaced retrieval. The trigger wasn't that I'd discovered a productivity method. It was the realisation, after one too many evenings staring at a dense article and getting nowhere, that the cost of opening any properly difficult piece of writing wasn't a question of willpower. It was a question of what my working memory could carry before it ran out of room. Something about the noise of the design itself, the columns, the unexplained jargon, the buried thesis, was eating capacity I needed for the actual idea on the page.

Here's what's actually happening. There's a name for it. There's a research framework for it. And once you see it clearly, you stop feeling stupid and start reading differently.

It's called cognitive load theory.

What Is Cognitive Load Theory?

Cognitive load theory is a framework from educational psychology that explains why some material is easy to learn and some material flattens you. It was proposed by John Sweller in 1988 in the journal Cognitive Science, in a paper titled "Cognitive Load During Problem Solving."

The core claim is simple. Working memory is small. Anything you're trying to understand has to fit through that small space. When the demand exceeds the space, learning stops.

That sentence sounds obvious until you sit with it. Working memory holds roughly four chunks of information at once. Not seven, which is what most people remember from school. Cowan's 2001 reanalysis revised the number down to four when chunking and rehearsal are properly controlled.

Four.

A single complex sentence can fill that capacity instantly. The subject is one chunk. The verb pattern is another. The conditional clause hanging off the back is a third. An unfamiliar term is a fourth. By the time you reach the full stop, you're at capacity.

Sweller's contribution was to break the load into types. Not all mental effort is the same. Some is unavoidable. Some is wasted. Some is the actual point of learning. If you don't know which is which, you can't reduce it. If you can, hard reading stops feeling like drowning.

Who Was John Sweller, and Why Did He Develop It?

John Sweller is an Australian educational psychologist based at the University of New South Wales. He's been publishing on this since the early 80s and is still active in the field.

The origin story is worth knowing because it grounds the abstract framework in something concrete.

Sweller was studying maths problems. He noticed something strange. Students who could explain every step of a problem in isolation often couldn't solve a problem that combined those steps. They knew the parts but couldn't run them all at once.

The standard explanation at the time was motivation, prior knowledge, or teaching quality. Sweller's hunch was different. He thought the bottleneck was working memory itself.

So he ran experiments. Different problem formats. Different ways of presenting the same content. Small changes in layout produced big changes in whether students could solve the problem. The content was identical. Only the load was different.

That was the insight. The brain isn't a single pool of effort you can scale up by trying harder. It's a system with separate compartments, each with hard limits. Respect the limits and learning happens. Ignore them and it doesn't, regardless of how clever the student is.

The 1988 paper formalised this. Over the next two decades, Sweller and his collaborators built it out into a full framework that now sits at the centre of instructional design, multimedia learning research, and an awful lot of EdTech product decisions.

The reason it matters for reading is that reading is just self-paced learning. Same architecture, same rules. When you open an article on a topic you don't know well, you're a student of that material. The load rules don't switch off because you're on your own.

What Are the Three Types of Cognitive Load?

This is the part that's actually useful. Sweller didn't just say "your brain has a ceiling." He said the load comes from three different places, and only some of it can be reduced.

Load type	What it is	Where it comes from	Can you reduce it?
Intrinsic load	The inherent difficulty of the material	The number of elements that must be held together to make sense of an idea	Only by simplifying the content or building prior knowledge first
Extraneous load	Difficulty added by presentation	Layout, vocabulary, split attention, redundant text, distracting environment	Yes, often dramatically
Germane load	Effort spent building durable understanding	Connecting new ideas to what you already know, forming schemas	This is what you want, but it only fits in the room left over

Total cognitive load is the sum of the three. Working memory caps the sum at roughly four chunks. The maths is brutal. If extraneous load is high, there's no room left for germane load, which is the actual point of reading. You finish the article, you can quote sentences from it, but nothing has bonded to your prior knowledge. The reading happened. The learning didn't.

Take each one in turn.

Intrinsic load is set by the material. A book on quantum field theory has higher intrinsic load than a recipe for pasta. You can't lower it without dumbing the content down. What you can do is build up prior knowledge first. The more you already know about a topic, the more chunks you can compress into a single chunk, which is what experts do without thinking. That's why a physicist can read a quantum paper in an afternoon while a smart non-physicist needs a week. Same paper. Different intrinsic load, because of different starting knowledge.

Extraneous load is the wasted effort. Bad formatting. Walls of text. Citations crammed inside the sentence. Footnotes that need a second pass. PDFs with two-column layouts that force your eye to leap halfway across the page. Articles where the meaning of paragraph three depends on a definition buried in paragraph one. None of that is the content. All of it costs working memory.

This is the load you can attack. Most reading problems are extraneous load problems pretending to be intrinsic load problems.

Germane load is the good kind. It's the effort of saying "ah, this connects to that thing I read last month." That's the work that turns reading into knowledge. The trouble is, germane load only happens if there's room. If extraneous load has already filled the budget, your brain is too busy processing the layout to do the connecting. You read the article and learn nothing, because every chunk of capacity went toward fighting the presentation rather than building understanding.

Sweller revised his position on germane load in 2011, treating it less as a separate type and more as the productive use of whatever capacity is left after intrinsic load is paid. The three-bucket model is still useful as a teaching tool even if the academic argument has moved on.

The reading that exhausts me fastest is dense text when I'm already tired. The morning after a long day, I'll open something I've been meaning to get to, three paragraphs in, and feel my working memory tap out. The intrinsic load is normal. The extraneous load (the design noise I haven't paid attention to) becomes uncrossable. That's the one I notice most. Which probably tells me my reading time should respect the recovery curve more than it does.

Why Does Reading Hard Material Feel So Tiring?

Now translate Sweller's framework into what it actually feels like in your body.

You sit down with an article. The first paragraph lands fine. The budget is balanced. Intrinsic load is moderate, extraneous load is low, germane load is doing its job.

Then a paragraph hits with three new technical terms. Each one forces a slow phonological decode. That's pure intrinsic load eating capacity. The mental model you were building gets paused while your brain processes the words.

You finish the paragraph and try to resume. But the suspension cost something. Your brain reaches for the thread it was holding and finds half of it. So you re-read the previous paragraph. That's a tax. You're paying the same load twice.

Two paragraphs later, the writer makes an inferential leap that depends on something from the first page. You can't access it. Working memory has already paged it out. So you scroll up, find it, scroll back. Now you're juggling the new material plus the retrieval effort plus the visual cost of the scroll. The budget is shot.

This is the moment when the page goes flat. Words still parsing, no meaning forming. Most people call this losing focus. Sweller would call it cognitive overload.

The body knows before your mind does. Your jaw tightens. You look away from the screen. You feel a small flinch of "I'm tired" or "I'm not smart enough for this." You stand up to make tea. The reading session is over even if you don't close the tab for another ten minutes.

Reading fatigue isn't a character flaw. It's working memory exhaustion. The neurotransmitters that maintain prefrontal focus, mainly dopamine and norepinephrine, deplete during sustained cognitive load. Most adults can sustain strong reading comprehension for 20 to 45 minutes before fatigue becomes measurable. Mental fatigue from reading is real, biological, predictable.

The reason hard reading feels like drowning is that overload is binary. It doesn't degrade gradually. It collapses. You're carrying the load fine, then you're not, with very little warning between the two states.

How Does Cognitive Load Apply to Articles and PDFs, Not Just Classrooms?

Most of Sweller's original research was done on instructional material in maths and science education. The framework was built to help teachers design better lessons. So a fair question: does it actually translate to the things you read every day, like Substack essays, news pieces, technical documentation, or research papers?

Yes. With one adjustment.

The classroom version assumes someone is designing the load on your behalf. A teacher has thought about which chunks to introduce when. Reading is the wild version. The writer optimised for their own clarity, not your working memory. Nobody is curating the order in which ideas hit you.

That makes extraneous load even more important. The articles you read online were not load-budgeted. Some publications get it right. Most don't. The classic offenders:

Long paragraphs with no breathing space, which force you to hold the whole block in working memory at once
Pop-up newsletter signups, cookie banners, and embedded videos that compete for the suppression budget
Inline citations and links that pull your attention sideways mid-sentence
Paywalls and "continue reading" buttons that break flow
PDFs designed for print, viewed on a screen too small for two columns
Vocabulary that assumes more prior knowledge than you have
Embedded images that arrive without context, splitting attention between caption and surrounding text

Each of those is extraneous load. None of it is the writer's argument. All of it costs you chunks.

Sweller built the framework in controlled instructional settings. Most reading happens in deliberately uncontrolled environments designed to capture attention rather than build understanding. The internet is, structurally, a bad place to read.

That doesn't mean every article online is too hard. It means the extraneous load is higher than you think. The article you bookmarked at 9pm to read in bed isn't carrying its full intrinsic difficulty. It's carrying its difficulty plus the load of the device, the notifications, the open tabs, and the small voice asking whether you should be doing something else.

For more on the working memory mechanism that sits underneath this, why your brain gives up after 3 paragraphs covers the exact sequence of how the budget collapses.

How Can You Reduce Extraneous Load While Reading?

This is the actionable bit. You can't change the intrinsic difficulty of an article. You can change the environment around it, and that's where most of the gains live.

Lever	What you actually do	Which load it cuts
Close other tabs	One tab, one article. No notifications. Everything else minimised.	Cuts extraneous load from suppression effort
Read in 25 to 30 minute blocks	Stop before fatigue. Take a real break. Resume when the budget refills.	Prevents the budget from running dry mid-article
Match difficulty to prior knowledge	Read an introduction first if the topic is new. Build the schema before tackling the dense piece.	Lowers intrinsic load by raising your starting context
Print or use single-column reader mode	Strip the article down to body text. No sidebar, no related links.	Cuts visual extraneous load
Use synchronised audio with text	Audio handles decoding. Visual handles tracking. The load splits across two channels.	Distributes load instead of concentrating it on the visual channel
Active recall between sections	Stop at the end of each major section. Try to summarise it from memory before continuing.	Converts germane load into actual long-term memory before the budget reloads
Read the same material twice with a gap	First pass for shape. Second pass for detail. The first builds prior knowledge that lowers intrinsic load on the second.	Lowers intrinsic load on the second pass by building schemas on the first

Most of these are obvious individually. The reason people don't do them is that the cost feels low when you're in the middle of struggling. You don't notice the load. You notice the feeling of being slow or tired, and you blame yourself rather than the environment.

"Close other tabs" sounds patronising until you understand what it's doing. Each visible tab is a small ongoing draw on the suppression budget. Your brain keeps track of them even when you're not looking. Five tabs is five small drains. Closing them frees capacity for the actual reading.

The audio trick is worth spending a moment on. Richard Mayer ran 17 separate experiments comparing spoken narration plus visuals against text-only learning. Spoken narration won 17 out of 17. The mechanism: when everything enters through the visual channel, the visual channel overloads. When narration shifts to the auditory channel, the load splits across two channels.

Synchronised highlighting (audio playing word by word with the matching text lit up as it's spoken) is a specific application of this. The audio offloads decoding. The visual handles position-tracking, so your eye doesn't have to search for its place. The reader running near the cognitive ceiling, the ADHD reader, the second-language reader, the tired reader at 9pm, gets the most out of this. Fluent expert readers below their ceiling notice less, because they had room to spare anyway.

This is part of what Alexandria does. The reader app pairs synchronised audio with word-by-word highlighting on the text, which is, in cognitive load terms, a deliberate split-channel design. The product exists to lower extraneous load on hard reading rather than to make hard reading shorter. A 5,000-word essay is still a 5,000-word essay. Whether you can actually carry it depends on how much of your working memory is being spent on decoding the page versus understanding the argument.

What Does This Mean for Choosing What to Read?

The thing nobody tells you is that load is contextual. The same article can be effortless on Saturday morning and impossible on a Tuesday evening. You're not a different reader. The capacity is just different.

This matters for what you choose to read and when.

Save the dense material for the moments your budget is full. First thing in the morning, ideally on paper, with a coffee and no notifications. That's when intrinsic load is most affordable.

Save the easy material for the spent moments. Late evening, on your phone, after a hard day. Trying to read a dense paper at 11pm isn't a discipline failure. It's a budgeting failure. You're trying to lift weight you don't have the room for.

This also reframes the "I want to read more" goal that haunts a lot of people. The number that matters isn't how many articles you finish. It's how many actually make it past extraneous load and become germane. You can read fifty articles a month and remember none of them, because the load was wrong every time. Or you can read four articles a month under good conditions and have all four bond to your existing knowledge. The four-article reader is, in the only sense that matters, reading more.

There's a related concept I keep coming back to. Comprehension Debt is what builds up when you read fast under load and never let germane processing happen. The article got read. The understanding never formed. The debt accumulates. You feel well-read but can't actually use what you've consumed. Phantom Knowledge is what fills the gap, the ghost of an idea that you're sure you encountered but can't actually access when you need it.

Both come from the same source. Reading without enough room left for germane load. The cure isn't faster reading. It's lower extraneous load and longer pauses between sections.

The goal isn't to read more. The goal is to understand more. Sweller's framework is just the maths underneath that statement.

For the encoding/retrieval method that turns reduced load into actual durable memory, how to learn faster without any of the hacks is the four-move companion piece. For the related question of why what you do read disappears so fast even when reading goes well, why you forget articles within a week gets into the forgetting curve and what breaks it. And for the practical retention side, how to actually remember what you read covers active recall, spaced repetition, and dual coding in more depth.

Frequently Asked Questions

What is cognitive load theory in simple terms?

Cognitive load theory, proposed by John Sweller in 1988, says working memory has a hard ceiling of roughly four chunks. Any learning task draws from that finite pool. When the total mental effort required exceeds the pool, comprehension collapses. The theory splits load into three sources: intrinsic (the difficulty of the material itself), extraneous (difficulty added by poor presentation), and germane (effort spent forming durable understanding).

Who is John Sweller?

John Sweller is an Australian educational psychologist at the University of New South Wales. He published the foundational paper on cognitive load theory in 1988 in the journal Cognitive Science. His original work focused on why students struggled with maths problems even when they understood each individual step. The framework has since been applied to instructional design, reading comprehension, multimedia learning, and software interface design.

What are the three types of cognitive load?

Intrinsic load is the inherent difficulty of the material itself, set by how many ideas must be held in mind at once to make sense of it. Extraneous load is difficulty added by how the material is presented: confusing layouts, jargon, split attention between sources, irrelevant decoration. Germane load is the productive effort of connecting new information to existing knowledge so it sticks. Total load is the sum of all three. When the sum exceeds working memory capacity, learning breaks down.

How does cognitive load theory explain reading fatigue?

Reading taxes all three load types simultaneously. Intrinsic load comes from the complexity of the ideas. Extraneous load comes from unfamiliar vocabulary, dense formatting, and the screen environment around the text. Germane load comes from building mental models that connect what you read to what you already know. When the combined load exceeds working memory, your brain switches from processing meaning to just decoding words, which feels like reading without absorbing anything.

How can you reduce extraneous cognitive load while reading?

Close other tabs, silence notifications, and read in a single window. Pick texts at the right level rather than constantly looking up vocabulary. Break long reading sessions into 25 to 30 minute blocks. Use audio synchronised with text so the visual and auditory channels share the work, which research by Richard Mayer found improves transfer in 17 out of 17 studies. Skip text designs that split attention between caption, image, and body.

Is cognitive load theory still considered valid?

Yes. Cognitive load theory is one of the most cited frameworks in educational psychology, with over 30 years of supporting research. The three-load framework has been refined over time. Sweller himself revised the position on germane load in 2011, and some researchers debate exactly where the line sits between extraneous and germane load. The core claim, that working memory is the bottleneck for learning and reducing unnecessary load improves outcomes, is well established.

How does cognitive load theory apply to articles and PDFs, not just classrooms?

The framework applies to any task that asks the brain to process and understand new information. An article is just a self-paced learning environment. The intrinsic load is set by the writer's ideas. The extraneous load is set by formatting, vocabulary, and the device you read on. The germane load is whatever effort you put into connecting it to your prior knowledge. Articles fail readers the same way badly designed lessons fail students: too much extraneous load, too little room left for understanding.

Why does reading feel like drowning sometimes?

When cognitive load exceeds working memory, the brain doesn't slow down gracefully. It collapses. You stop processing meaning and just track words. Re-reading the paragraph rarely helps because the budget is already spent. The drowning feeling is the gap between the load the text is asking you to carry and the capacity you actually have in that moment. The cure isn't more willpower. It's reducing the load.

Sources: Sweller, J. (1988). Cognitive Load During Problem Solving: Effects on Learning. Cognitive Science, 12(2), 257-285. | Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioural and Brain Sciences, 24(1). | Sweller, J., Ayres, P., & Kalyuga, S. (2011). Cognitive Load Theory. Springer. | Mayer, R.E. (2001, 2009). Cognitive Theory of Multimedia Learning. Cambridge University Press. | Mayer, R.E. & Moreno, R. modality principle research, summarised in Pressbooks Learning Environments Design. | PMC2864034: The Magical Mystery Four, Working Memory Capacity. | PMC11253940: Probing sustained attention and fatigue across the lifespan, PLOS One 2023.