Writing

Somatic UX: designing for the body that sees the screen

Every interface sets a rhythm. Design it on purpose.

Nasal breathing, one participant, played back at recording speed (25 samples a second). Watch it for half a minute. data Ph.D. lab recording, raw, not public

Before your users read a word, your product has already changed them.

Seeing changes state

What we see changes our state. By state I mean the body's current setting: how fast you breathe, how alert or relaxed you are, how narrow or wide your attention is. Every screen moves that setting somewhere, whether or not anyone designed it to, and often without the reader noticing. As a neuroscientist, I treat that as a design surface, to be set as deliberately as color, type and copy.

Watching the breathing above may already have moved yours. People who watch a body breathe drift toward its pace. In virtual reality, participants who watched their own avatar breathe slower or faster shifted their own breathing partway toward it, by about half a breath a minute over three minutes, without being told to follow it (Czub & Kowal, 2019). A related effect shows up at a desk. Thirteen people did their own work with a small bar at the edge of the screen that, they were told, showed a target breath rate. While it was on, they breathed about two breaths a minute slower, and went back to their usual rate when it turned off (Moraveji et al., 2011).

None of this is new. Researchers in human-computer interaction (HCI) already have a name for it: somatic interaction design asks what an interface does to the body before it asks what it says (Höök, 2018; Thecla Schiphorst). Breathing and meditation apps design for state every day. In my experience, it rarely makes the checklist anywhere else. It belongs in every product, not only in the ones built to calm us down.

State changes what is seen

The effect runs the other way too: the body's state changes perception. Even the phase of a single breath matters: people judged visuospatial problems more accurately when a trial began on an inhale than on an exhale (shoutout to Perl et al., 2019).

Put together, an interface that shifts state also shifts what information enters the system and what it meets there. The same sentence, read by a rushed body and by a settled one, is not the same input.

Breath is the lever

Breath is the part of the autonomic nervous system we can most easily steer: it runs on its own, but we can take it over at will. Slowed to about six breaths a minute, it raises heart rate variability, a marker of a calmer body (Lehrer & Gevirtz, 2014). In a month-long study of about 100 people, five minutes a day of deliberate, exhale-weighted breathing improved mood (Balban et al., 2023). That makes breath the lever a screen can most plausibly reach.

However, adding a breathing exercise to the site is exactly the wrong move. People resist influence they can see aimed at them. A message you overhear can move you more than the same message said to your face (Walster & Festinger, 1962). Saying "let's take a breath" to someone who came to work will make them object, and once the mind objects, it may block the body's response. Trying to relax on command can even backfire: people told to relax while under mental load showed higher skin conductance, a sign of arousal, not lower (Wegner, Broome & Blumberg, 1997).

The alternative is ambient, not hidden. Put the rhythm in the text, the motion and the pauses, without asking for attention to it. Calm technology asked for this decades ago: technology that informs from the periphery of attention instead of demanding its center (Weiser & Brown, 1996).

Sound may be a stronger channel than sight. In one study, faster music sped up listeners' breathing, and silent pauses slowed it below where it started (Bernardi, Porta & Sleight, 2006). Most sites are silent, so vision carries the rhythm, but a product that already uses sound has a second lever to test.

Decide the state

If state shapes what people take in, the first design question is which state the product needs, and it differs by task. Checking work needs alert vigilance; thinking an idea through needs a slower, settled mind. Pick the target, then set the rhythm that gets people there. Make this a deliberate decision. If nobody asks, the pace gets set by accident: text streams as fast as the model produces it, and animations run at whatever speed the code library ships with.

Left to chance: AI chat

AI chat shows what happens when nobody decides. In a survey of about 1,500 US workers, people whose work was mostly overseeing AI output (reviewing it, correcting it, checking whether it is right) reported 12% more fatigue and 19% more information overload than people who used AI to take tasks off their plate. It is a self-report survey, so it shows an association, not a cause. The authors call it "AI brain fry" (Bedard, Kropp & Hsu, 2026). The body is off rhythm at the screen even without AI: in informal measurements of a few dozen people, Linda Stone found that most of them held their breath or breathed shallowly while they answered email, and named it screen apnea.

For anyone designing AI products, this is the case that matters most. A chat interface streams long, confident answers faster than people read, and the user's job becomes checking them: the oversight work the survey linked to fatigue. Whether the speed of the stream adds to that load has not been tested. But how fast the stream runs, and where it pauses, is a design decision, and it affects how alert or rushed the user is while checking.

If your users spend their day checking AI output, make their job easier. Default to a pace that suits the state the task needs, pause where sentences end, and let them stop, skip to the end or set the tempo themselves.

Chosen on purpose: a small example

In Idea-to-Product, the goal is a state: the slow, deliberate thinking that testing an idea needs. Rituals are an old tool for this; they help people regulate emotion and settle into a goal state (Hobson et al., 2018). Each session opens with the same short ritual before the conversation, so a returning founder comes back to that state, not only to their notes. The text types itself at 40 ms a character, pausing 600 ms at commas, 1.4 s at periods and 2 s between lines. Behind it, the background shifts on a 6-second cycle. There is no timer, and the founder presses Begin when the text is done.

4 design principles

Design for state, then comprehension. State comes first and shapes what gets understood. Ask what the screen does to the person seeing it before what it says.
Ambient, not hidden. Set the rhythm without asking for attention to it.
Choose the rhythm on purpose. Ambient motion slow enough to breathe with: slow breathing runs about 10 s per cycle, around six breaths a minute, on a sine or ease-in-out curve. Linear reads as mechanical.
Punctuation and layout are timing. Commas and periods are natural places for a pause. Line length sets pace too: on screen, very short lines are read more slowly than long ones (Dyson & Kipping, 1998).

What we know, and what we don't

The pieces have support: what we see shifts how we breathe, breathing shifts the body's state, and the body's state shapes perception. To the best of my knowledge, what has not been tested is the whole chain inside a product, whether a paced screen shifts the reader's breathing and, through it, what they take in. That needs its own experiment: breathing measured while people read a paced and an unpaced version of the same screen, followed by a test of what they took in.

Until then, the surface is there either way. Every interface sets a state. The only choice is whether you choose it.

References

Balban, M. Y., et al. (2023). Brief structured respiration practices enhance mood and reduce physiological arousal. Cell Reports Medicine. link
Bedard, J., Kropp, M., & Hsu, M. (2026). When using AI leads to "brain fry". Harvard Business Review / Boston Consulting Group. Survey, self-report. link
Bernardi, L., Porta, C., & Sleight, P. (2006). Cardiovascular, cerebrovascular, and respiratory changes induced by different types of music in musicians and non-musicians: the importance of silence. Heart, 92(4), 445–452.
Czub, M., & Kowal, M. (2019). Respiration entrainment in virtual reality by using a breathing avatar. Cyberpsychology, Behavior, and Social Networking, 22(7), 494–499. link
Dyson, M. C., & Kipping, G. J. (1998). The effects of line length and method of movement on patterns of reading from screen. Visible Language, 32, 150–181.
Höök, K. (2018). Designing with the Body: Somaesthetic Interaction Design. MIT Press.
Hobson, N. M., Schroeder, J., Risen, J. L., Xygalatas, D., & Inzlicht, M. (2018). The psychology of rituals: an integrative review and process-based framework. Personality and Social Psychology Review, 22(3), 260–284. link
Lehrer, P. M., & Gevirtz, R. (2014). Heart rate variability biofeedback: how and why does it work? Frontiers in Psychology, 5, 756. link
Moraveji, N., et al. (2011). Peripheral paced respiration: influencing user physiology during information work. UIST '11. n = 13. link
Perl, O., Ravia, A., Rubinson, M., Eisen, A., Soroka, T., Mor, N., Secundo, L., & Sobel, N. (2019). Human non-olfactory cognition phase-locked with inhalation. Nature Human Behaviour, 3, 501–512. link
Stone, L. (2014). Are you breathing? Do you have email apnea? Informal measurements, not peer-reviewed. link
Walster, E., & Festinger, L. (1962). The effectiveness of "overheard" persuasive communications. Journal of Abnormal and Social Psychology, 65, 395–402. link
Wegner, D. M., Broome, A., & Blumberg, S. J. (1997). Ironic effects of trying to relax under stress. Behaviour Research and Therapy, 35, 11–21.
Weiser, M., & Brown, J. S. (1996). Designing calm technology. PowerGrid Journal, 1(1).