Neaptide opens with a circle. You hold it for about six seconds, and a session starts. In those seconds the app looks at a handful of small things about how your thumb behaves: how much it trembles, how far it drifts, how long it stays completely still, how your grip shifts, and how quickly you reached for the screen. It uses those to suggest a direction, Settle if you seem wound up, Surface if you seem flat.
That is a big thing to ask of a thumb. So this piece covers what the research supports, where it stops, and why the suggestion is built as a guess you can overrule rather than a verdict.
Why not just ask how you feel?
Asking is the obvious design. Most people would expect a screen that says how are you feeling right now? and offers a few options.
The trouble is that knowing your own internal state is a skill, and it varies a great deal between people. Researchers who study interoception, the sensing of what is going on inside the body, separate it into three different things: how accurately you can detect a bodily signal, how good you believe you are at it, and whether your confidence actually tracks your accuracy.
In a study of 80 people, Garfinkel and colleagues found those three were distinct and dissociable (“Knowing your own heart,” Biological Psychology, 2015, PMID 25451381). Accuracy on a heartbeat-detection task was only partly predicted by how good people thought they were. The three measures lined up meaningfully only in the group with the best accuracy.
That is a study of heartbeat perception, not of stress, and it does not show that people cannot describe their mood. What it does show is more modest and still useful: feeling sure about what is happening in your body and being right about it are separate things. On a hard day, when choosing already feels like effort, a self-label may not be the most reliable place to start.
Can a touch carry information about your state?
There is a real, if small, body of research here, and it is worth being precise about it.
Tremor is the clearest piece. The body has a normal physiologic tremor that is fine and rapid, around 8 to 13 cycles per second, and usually barely perceptible. The Merck Manual’s clinical reference notes that it “becomes noticeable in many people during physical or mental stress,” and lists anxiety among the triggers (Merck Manual, Professional Edition: Tremor). So a thumb held against glass is not a neutral instrument. It picks up some of what your nervous system is doing.
Touch patterns have also been linked to emotional state. Gao, Bianchi-Berthouze and Meng asked 15 people to play a touchscreen game across 300 sessions and analysed their finger strokes (“What does touch tell us about emotions in touchscreen-based gameplay?”, ACM Transactions on Computer-Human Interaction, 2012). Pressure helped distinguish frustration. Speed and directness helped distinguish arousal. A model separated higher from lower arousal correctly about 89% of the time.
A later field study took this out of the lab. Ghosh and colleagues tracked how 22 people typed and swiped on their own phones for three weeks, asking them regularly how they felt (“Emotion detection from touch interactions during text entry on smartphones,” International Journal of Human-Computer Studies, 2019). Personalised models distinguished happy, sad, stressed and relaxed states with an average AUC of 73%.
What those studies don’t show
This is where writing about apps tends to overreach, so here are the limits plainly.
The samples are small. Fifteen and twenty-two participants are useful for showing that a signal exists. They are not enough to say how well it works across everyone.
The tasks are different. Swiping through a game and typing messages give a model a lot of touch data. A six-second hold gives far less.
The best results were personalised. Ghosh’s models were trained on each person’s own data over weeks. Neaptide keeps no history of your sessions, so it cannot learn you that way, and that is a deliberate privacy choice.
Tremor has many causes. The same clinical reference lists fatigue, exercise, sleep deprivation and caffeine alongside anxiety. A shaky thumb after two espressos and a short night is not telling you much about stress.
Put together, the research supports a modest claim: the way you touch a screen carries some information about your state. It does not support the idea that a few seconds of touch can tell anyone reliably how you feel.
So why use the hold at all?
Because a rough starting point that asks nothing of you is worth something, provided it is honest about being rough.
The hold gives you a way to begin without a decision. It gives the app a small amount of evidence to make a first suggestion. And you stay the authority on which direction is right. If Neaptide suggests Settle and you know you are foggy rather than wired, you switch. Neaptide describes; it does not diagnose.
That design follows from something we covered in Why does deep breathing make me feel worse?. Feeling bad comes in two broad directions, wound up and shut down, and a technique that helps one can make the other worse. Picking the wrong direction is common, and it is not a personal failure. The practical answer is to guess, notice, and correct.
The hold is that same approach built into the first six seconds. It makes a guess, and you correct it if it’s wrong.
What stays on your device
A reading like this is personal, so it is worth saying exactly what happens to it. The reading from your hold is temporary and is discarded when you close the app. Neaptide works offline, has no analytics or tracking, and keeps no record of your sessions. The only thing it remembers is the session length you chose. The privacy policy sets this out in plain language.
The short version
- Your thumb carries some information about your state, through signals like tremor and how it moves.
- Research shows the signal exists, but the studies are small, and the stronger results used much more data than a single hold.
- Caffeine, tiredness and exercise can look like stress to a sensor.
- So the hold gives you a starting suggestion, not an assessment. You can change direction at any point.
Sources
- Garfinkel, S.N., Seth, A.K., Barrett, A.B., Suzuki, K., & Critchley, H.D. (2015). Knowing your own heart: Distinguishing interoceptive accuracy from interoceptive awareness. Biological Psychology, 104, 65–74. PMID 25451381. https://pubmed.ncbi.nlm.nih.gov/25451381/
- Rajput, A., & Noyes, E. Tremor. Merck Manual, Professional Edition (reviewed April 2026). https://www.merckmanuals.com/professional/neurologic-disorders/movement-and-cerebellar-disorders/tremor
- Gao, Y., Bianchi-Berthouze, N., & Meng, H. (2012). What does touch tell us about emotions in touchscreen-based gameplay? ACM Transactions on Computer-Human Interaction. https://doi.org/10.1145/2395131.2395138
- Ghosh, S., Hiware, K., Ganguly, N., Mitra, B., & De, P. (2019). Emotion detection from touch interactions during text entry on smartphones. International Journal of Human-Computer Studies, 130, 47–57. https://www.sciencedirect.com/science/article/abs/pii/S1071581918304889
Neaptide is not treatment, and not a substitute for care from a professional. If you are in crisis, contact your local emergency number.