The Keyboard Nobody Specified

Everything on this desk went through a decision except one thing. The laptop has a memory configuration somebody defended in writing. The monitor has a panel type and a refresh rate that cost a week of discussion and one strong opinion about color. Even the chair had a trial period. The keyboard came in the box with the laptop and nobody has thought about it since.

Rank the same objects by how much of the day your body is in contact with them and the order comes out reversed. The monitor you look at. The chair you sit in and forget. The keyboard is the one you are physically touching, continuously, the whole time you are working, and it is the one nobody was asked to approve.

The subject also attracts a lot of confident nonsense, most of it about switches and much of it written by people who sell switches, so here is the sentence the rest of this hangs on. The interesting thing about keyboards is not that the good ones are better, because on the measure everyone reaches for first they are not. It is that a tool you touch continuously and a tool you touch occasionally deserve very different amounts of thought, and almost nobody allocates it that way, including me for about fifteen years.

A switch is a spring, a stem and two contacts. Press a key on a rubber dome board and the dome collapses onto a membrane. The bottom of the stroke is the switch, so the travel has to be completed for the keystroke to exist. A mechanical switch separates those two events. CHERRY's MX2A RED is published at 45 cN of operating force, 2.0 mm of pre-travel and 4.0 mm of total travel: registered at halfway down, with the remaining two millimeters optional.

Linear, tactile and clicky answer one question. The question is what the stem tells your finger at the moment of registration. A linear switch says nothing and the force curve is smooth. A tactile switch has a bump shaped into the stem so the actuation point is felt. A clicky switch adds a separate mechanism that makes noise. Tactility and sound are different parts doing different jobs, which is why people who ask for clicky usually want tactile.

The entire argument is about fifteen centinewtons. CHERRY publishes the RED at 45 cN, the tactile BROWN at 55 cN and the clicky BLUE at 60 cN, all with four millimeters of travel. That is the mainstream range end to end, a spread your hand adapts to inside a day. The rated lives say more: over 100 million keystrokes for the RED and the BROWN, over 50 million for the BLUE, because the click is an extra moving part and halves the published life. A feature is a trade, printed on the vendor's own page.

The largest look at how people actually type is "Observations on Typing from 136 Million Keystrokes", presented at CHI in 2018 by researchers at Aalto and Cambridge: 168,000 volunteers, 136 million keystrokes, average 51.56 wpm with a standard deviation of 20.2. It is an online study of self-selected participants rather than a controlled sample, and the authors say so, which makes the comparisons within it the useful part.

Formally trained typists came out about 5 wpm ahead of untrained ones, a gap the paper reports at an effect size of 0.27 and calls relatively small. What predicted speed was rollover, correlating at 0.73, plus finger count and hand alternation. The study recorded whether each participant was on a physical or a laptop keyboard, roughly a 44 to 54 split, and keyboard type is not among the predictors it reports. That is an absence rather than a measured null, and I will not push it further.

Rollover there is a behavior: pressing the next key before releasing the last, which the fastest participants did on 40 to 70 percent of keypresses. It is not the hardware specification with the similar name, which is how many simultaneous presses the matrix can report without dropping or inventing one. But they meet. If your hands overlap most of their keystrokes, a board that resolves only a few at once will sometimes disagree with you, which is why Das Keyboard advertises full n-key rollover. A correctness argument, not a speed one.

The number pad sits between your hand and the mouse. Cook and Kothiyal put ten computer users through an editing task in Applied Ergonomics in 1998, measuring shoulder and arm activity by surface EMG. With the mouse beside a keyboard that had no numeric pad, anterior and middle deltoid activity was significantly lower than beside a standard board; trapezius did not differ. Ten subjects is ten subjects. But the mechanism needs no study: the pad holds your mouse hand out to the side all day.

Tenkeyless is the boring correct answer for most people. Remove the number pad and the mouse comes in by the width of the block you removed, which is the largest change you can make to a desk without buying anything unusual. The test takes a week: notice whether you ever touch the pad. If your day is spreadsheets, you use it constantly and none of this applies. If you write code and prose, you have been paying for the space anyway.

The measured posture gain came from the wrist rest. The same group put 35 people on an adjustable split keyboard for one to two weeks and measured wrist angle by motion capture. Mean preferred opening angle was 14 degrees, and ulnar deviation improved only for the twelve who chose the widest angles. Wrist extension was lower on the adjustable board, 17 degrees against 24, which the authors attributed most likely to the palm support rather than the split. The part everyone photographs was not the part that moved.

Split and ortholinear cost weeks, in both directions. A genuinely different layout is a real commitment and the adaptation is measured in weeks of being slower at your job. The part people underplay is that it runs both ways: get fluent on a split ortholinear board and every other keyboard in the building, including your own laptop in an airport, becomes foreign. A fine trade if you type at one desk, and worse than it looks if you do not.

A modern keyboard is a small computer whose only job is to report which keys you pressed, and on a growing number of boards that computer runs QMK, open source firmware with layers, remapping and macros. Layers are the important one: hold a key and the whole board becomes a second board, so cursor keys and brackets sit under fingers already resting on the home row.

Operating systems have remapping tools and they are not the same thing, because the remap lives in the wrong place. Firmware travels with the hardware. It works before anything has booted, in a setup screen, through a KVM, and on the locked-down machine where you may not install a utility. The alternative is a per-machine install, an administrator prompt and a config file to keep in sync across every device you touch.

That is what turned a soldering hobby into a purchase. Boards such as Keychron's Q and V lines ship with QMK and a graphical configurator, so remapping a key is a drag rather than a toolchain. The honest cost is dependency: shape the board around your hands and your hands expect it everywhere, and somebody else's machine is where you find out how much fluency you moved into firmware.

Both wearing parts come out. A hot-swap board holds each switch in a socket rather than a solder joint, so a key that starts chattering is a thirty-second repair with a puller and a spare. Keycaps pull off too, and the shine on a well used board is plastic polishing itself against a fingertip. Compare a laptop, where the keyboard is often assembled into the top case and one failed key takes the whole upper half of the machine with it.

The cost only works if you keep it, which is the point. The blank-legend Das Keyboard 4 Ultimate lists at $199 against a bundled board at nothing, and on the day that difference is entirely real. Spread over a decade it is a couple of dollars a month. But the amortization is not a fact about the price, it is a bet on the denominator, and the bet only pays if the thing is repairable, which is why the sockets came first in this list and the price comes last.

It is the cheapest lever a manager has, and it has no budget line. Laptops have a refresh cycle, a specification and a line item. Peripherals have a cupboard. A keyboard and a decent mouse cost a small fraction of one machine refresh, outlast several of them, and change the physical experience of every working hour. I am not building a policy out of that. It is just the one place where trivial money buys something noticed daily, and the place nobody looks.

Switch feel is preference, and I have found no study showing any mainstream switch type makes anybody measurably faster or more comfortable than another. The nearest thing to evidence cuts the other way. Tittiranonda and colleagues ran an 80-participant, six-month, randomized, placebo-controlled trial of alternative geometry keyboards and reported an improving trend in pain severity for two of them, no matching improvement in clinical findings, and a significant correlation between pain improvement and satisfaction with the keyboard. That last one is exactly the confound you would have worried about in advance.

The externality gets discussed less and is worse. A clicky board in a shared room imposes a sound on people who did not choose it, cannot mute it, and have no graceful way to raise it. The owner hears feedback and everyone else hears an interruption several times a second. "They will get used to it" is the argument of everyone who ever put a noise into somebody else's room. Buy clicky switches for a room you are alone in.

Which shrinks the claim I started with, and it should. Somebody who has typed happily on a laptop keyboard for twenty years is not doing it wrong and is not leaving measurable speed on the table. What survives is duller than the enthusiasm suggests: a layout that keeps the mouse near your centerline, firmware that puts the keys you use where your fingers already are, and a board you can repair. None of it is about how the switch sounds.

I should resist turning this into a rule about equipment generally, because most of a desk does not repay this kind of attention and I have plainly spent more time here than a keyboard deserves. That is arguably evidence for the argument rather than against it. What makes it worth any thought at all is that contact time is a variable almost nobody puts in the equation, and six hours a day is a category with about two members.

Four millimeters, several times a second, for a career. The board in front of me has outlasted three laptops and one confident opinion of mine about switches, and what I would change if I started over is not the switch under my fingers. It is that I would have worked out much sooner that the keys can be moved.