On the Virtue of Scrolling

There is a gesture so universal that we no longer notice it. We perform it hundreds of times a day, on phones and laptops and tablets and the occasional desktop computer that someone’s IT department has not yet replaced. We scroll. We scroll and scroll and scroll. We scroll past things we meant to read, past things we never meant to read, past advertisements for things we mentioned once in a conversation that was not, we were assured, being monitored. We scroll.

This essay is about scrolling. It is also, unavoidably, something you must scroll through. This is intentional.

The word “scroll” is older than the web and older than the computer. It comes from the Latin scrofa, meaning a pig, which became escroe in Old French, meaning a strip of parchment, which became scroll in English, meaning a roll of writing material. The pig connection is disputed by etymologists who prefer cleaner origin stories. The point is that scrolling is ancient. People have been unrolling things to read them since before they had the good sense to fold them into pages.

The codex — the book as we know it, with pages — replaced the scroll not because scrolling was bad but because codices were easier to navigate. You could flip to chapter seven without unrolling everything before it. Scholars of the ancient world had essentially invented random access. They did not know they had invented it. They just wanted to find the good parts faster.

We have now spent decades un-inventing this. The infinite scroll, that cursed design pattern, returned us to the parchment roll. You cannot jump to the middle. There is no middle. There is only more.

I want to be fair to scrolling. Scrolling is not inherently bad. The problem is not the gesture. The problem is what has been designed to exploit it.

The infinite scroll was invented in 2006 by Aza Raskin, who later said he regretted it and estimated it was responsible for 200,000 hours of wasted human time per day. He invented it in about an hour. This is the ratio that haunts software engineering: one hour of effort, millions of hours of consequence. He was trying to make pagination less annoying. He succeeded in ways he did not intend.

Pagination is annoying. Nobody liked clicking “Next Page.” It interrupted the flow. But interruption, it turns out, is sometimes the point. The click gave you a moment to decide whether you wanted to continue. The infinite scroll removed that moment. It replaced choice with momentum. You are not reading; you are falling.

There is a posture associated with serious reading. You sit up. You hold the book at an angle. If you are in a library, you perhaps feel a faint social pressure not to slouch. The physical arrangement of the activity signals to your brain that something is happening, that attention is being paid.

There is a posture associated with scrolling. You know the one. The neck bent at an angle that chiropractors have begun calling “text neck.” The thumb moving in a small, repetitive arc. The eyes slightly glazed. You can scroll for twenty minutes and retain almost nothing, because nothing was ever designed to be retained. It was designed to be scrolled past.

Reading a long essay on a screen, then, is an act of mild defiance. You are using the scroll gesture — the same gesture, the same thumb, the same slightly glazed eyes — but in the service of a single coherent argument that intends to go somewhere. The scroll is repurposed. You are falling, but you know where the ground is.

The web was not supposed to be this. The web was supposed to be documents. Hypertext, in the vision of its early theorists, was a new kind of reading: non-linear, associative, connected. You would follow links not because an algorithm suggested them but because one idea genuinely led to another. The link was a thought, not a trap.

Tim Berners-Lee imagined a system for sharing physics papers at CERN. This is perhaps why his vision did not account for the engagement metrics that would later determine what the web became. Physicists, famously, do not optimize for time-on-site. They optimize for correctness. These are not the same thing.

Let us talk about reading on screens, since we are doing it.

The research is inconclusive, which is the most honest thing I can say. Studies suggest people read more slowly on screens than on paper. Other studies suggest the difference is small and shrinking as people spend more of their lives reading on screens. Still other studies suggest the difference depends heavily on what you are reading and why.

What is less contested is that reading on screens tends toward skimming. We F-pattern our way through web pages: reading the first lines, then the left edges, then stopping. Eyetracking studies have produced heat maps of this behavior that look like the letter F drawn by someone who gave up halfway through. We read the headline. We read the first sentence. We scroll a little. We leave.

This is rational behavior given what most web pages deserve. The web has trained us, over decades, that most content is not worth reading carefully. We have adapted. The adaptation is now poorly suited to the content that does deserve careful reading, but that is not our fault. It is the fault of every listicle, every auto-playing video, every “One Weird Trick” that ever demanded our attention and repaid it with nothing.

I am asking you to read carefully now. I am aware this is a test essay, created to test whether a navigation bar sticks to the top of the screen as you scroll. I am aware that you may be reading this less for its content than to watch the header. This is fine. The header should be watching back.

A sticky navigation bar is a small piece of UX. It costs little to implement and provides a modest convenience: you can see where you are and where you might go without scrolling back to the top. It is the codex principle applied to a single page. You could jump, if there were somewhere to jump to.

Most personal sites do not need much navigation. Mine has essentially one thing: essays. You are reading one. When you are done, you can go back to the list of them. The navigation exists to make that return easy, not to suggest a complexity the site does not have. Restraint in navigation is, I think, a virtue.

Restraint is not a popular value in web design, which is perhaps why it needs defending.

The modern web page is often a triumph of presence over purpose. There are animations, because animations signal that effort was made. There are gradients, because flat colors signal that effort was not made. There is a hero section, because someone wrote a blog post about hero sections. There is a call to action, because every page must want something from you.

What does this page want from you? Approximately nothing. It wants you to read the words, if you feel like it, and then go away. This is the relationship that books have with their readers, and it is a good relationship. The book does not ask for your email address. The book does not send you notifications. The book waits, patiently, for you to return.

A defense of long-form writing on the web is also, necessarily, a defense of the reader’s time. Long writing respects the reader’s time only if it earns the length. A 5,000-word essay that could have been 500 words is not generous; it is inconsiderate. The length must be in service of something: an argument complex enough to require it, a subject rich enough to sustain it, a voice interesting enough to justify the imposition.

I will not claim this essay fully earns its length. It is a test essay. It was requested to be long. I have tried to make it honest about this fact, which is perhaps the best that can be done under the circumstances. An essay that knows it is filling space and says so is at least not pretending. Pretending is worse.

What is the appropriate length of a web page? This is not a question web designers often ask explicitly, because the answer varies enormously by purpose. A homepage is not an essay. A product page is not an article. But there is a general pressure, in web culture, toward shortness — not out of respect for the reader’s attention but out of distrust of it.

The assumption is that readers will not make it to the end. So the important things go first. The less important things come after. The things that don’t fit get cut. This is not wrong as a principle — it is the inverted pyramid that journalists have used for a century. But it becomes pathological when applied to everything, when every piece of writing is structured as if the reader might stop at any moment and must be fed the conclusion immediately.

Some things cannot be front-loaded. Some arguments require setup. Some ideas are not legible without context. If you have decided, as a reader, to read something, you have made a commitment that the writer should be allowed to honor. Trust runs both ways.

We are, by my count, roughly halfway through this essay. If you have made it this far, you have either found it interesting or you are testing navigation behavior very thoroughly. Both are valid.

The scroll has brought you here. You are approximately in the middle of a document about scrolling, which has a certain symmetry. Below you is as much text as above. The navigation bar, if it is working, is still at the top of your screen. You could click away at any moment. You won’t, probably, but you could.

This is the thing about reading: it is voluntary all the way through. You can stop at any sentence. Nobody is stopping you. The text continues whether or not you do. This is different from a conversation, where stopping is rude. It is different from a film, where stopping requires effort. With text, the default is stopping; continuing requires a small, repeated act of will.

You are still here. Good.

The history of reading is also a history of what reading was thought to do to people. In the 18th century, there was a moral panic about novels. Novels, critics worried, were too engrossing. Women, particularly, were thought to be susceptible to confusing fiction with reality, to taking romantic plots as behavioral templates. The reading woman became a stock figure of social concern: head bent over a book, unavailable to the world.

This concern was not entirely wrong in its diagnosis, only wrong in its judgment. Novels do produce identification. Readers do take on the perspectives of characters. This is not a defect; it is the point. The capacity to inhabit another consciousness, even a fictional one, is a form of practice. Empathy, like most things, can be trained.

What replaced this concern about novels? Roughly: concern about television, then video games, then social media. Each new medium is charged with corrupting attention and manufacturing passivity. Each new medium turns out to be, for the most part, fine. People adapt. The mediums that are genuinely harmful turn out to be harmful for specific, identifiable reasons, not simply because they are new.

What, specifically, is harmful about the infinite scroll?

It is not the scrolling. It is the removal of endpoints. Human attention works better with defined tasks than with open-ended ones. You can read for an hour if you know you have an hour. You cannot easily stop after an hour if the feed simply continues, because there is always more, and the algorithm is specifically calibrated to find the more that you will find most difficult to stop consuming.

This is not a flaw in the design. It is the design. The goal is time-on-platform. Time-on-platform is served by minimizing friction and maximizing what is sometimes, in product meetings, called “the next-content pull.” The next-content pull is the feeling that maybe the next thing will be the good thing. It is a slot machine that costs nothing to pull. You will pull it again.

I am being harsh about social media in an essay hosted on a static site with no tracking, no comments, no social features of any kind. This is perhaps too easy. It is simple to be virtuous in the absence of revenue pressures.

If I were trying to build a business around this site, I would be making different choices. I would want to know how long you spent reading. I would want your email so I could tell you when something new went up. I would want to show you related essays, which would keep you here longer, which would eventually become a metric someone would care about. The corruption is not dramatic. It is incremental. Each small concession seems reasonable. The aggregate is the problem.

I am not building a business around this site. This helps.

The static site is an underrated form. It has no database. It has no server-side code. It is, at its core, a folder of files that a web server sends to people who ask for them. This is how the web worked in 1995, and there is something clarifying about returning to it.

The complexity of modern web infrastructure is not without purpose. Databases allow dynamic content. Server-side rendering allows personalization. APIs allow integration with other services. If you need these things, the complexity is justified. If you do not need them, the complexity is overhead that serves primarily to give developers something to maintain.

A site of essays needs to do exactly one thing: display text. It turns out that HTML is quite good at displaying text. CSS can make it look reasonable. A build script can generate the index. Everything else is optional.

There is a pleasure in using simple tools for appropriate tasks. A chef’s knife for vegetables. A fountain pen for signatures. A static site for essays. The pleasure is partly aesthetic — the satisfaction of the right tool — and partly practical. Simple tools fail in simple, understandable ways. Complex tools fail in complex, mysterious ways.

I have encountered this recently with Docker, which is involved in other parts of this infrastructure. Docker is a good tool for running services in isolation. It is a complex tool that fails in complex ways. When something goes wrong with a static site, you look at the HTML. When something goes wrong with a containerized service, you look at logs, then network configurations, then volume mounts, then environment variables, and eventually you find the thing that was misconfigured in a way that seemed fine but wasn’t.

This is not an argument against Docker. It is an argument for using the right tool for the task. The static site does not need Docker. It needs rsync.

Rsync is a program for synchronizing files between computers. It is old. It is not glamorous. It does exactly what it promises to do. When you run it, files appear where you wanted them. This is the whole thing.

The deploy script for this site is essentially a single rsync command. Files go from here to there. The server serves them. This process is so simple that there is almost nothing to debug. When something goes wrong, it is usually because you forgot to build first. This is a good failure mode: obvious cause, obvious solution.

We are approaching the end. You can tell because the text is starting to feel slightly elegiac, and because essays almost always signal their endings before they reach them. The signal is usually a return to the opening theme, or a broadening of scope, or a sentence that feels slightly more conclusive than the ones before it.

We started with scrolling. You have been scrolling. The navigation bar has been (hopefully) sticking to the top of the screen. These things are related.

What does it mean to read to the end of something? In practical terms: you have seen all the words. In experiential terms: something different. To read to the end is to have followed an argument, or a story, or a meditation, wherever it went. It is to have trusted the writer enough to go the distance. It is a form of patience that is rarer than it should be, given how often it is rewarded.

Not every ending rewards the patience. Some essays run out of ideas before they run out of words. Some arguments reach their conclusion and then continue for three more paragraphs for reasons that are never clear. Endings are harder to write than beginnings. Beginnings have novelty on their side. Endings must earn their finality.

The ending I am reaching for is this: the web works best when it respects the reader’s choice to read. A long essay on a static site is a small statement about that respect. There are no tricks here to keep you reading. There is no infinite scroll. There is text, and then there is the end of the text, and then there is a link back to the index.

The navigation bar at the top of the screen is part of this. It says: you know where you are. You can leave whenever you want. The exit is always visible. Read as much as you like, and when you are done, here is how you go back.

That is, I think, the right relationship between a site and its reader. Not captive. Not trapped. Just: here is something, if you want it.

You have scrolled to the end. The end is here.