The research
Why this works.
Whistlestop makes some specific claims about attention: that you stop hearing a sound you already know, that being interrupted mid task costs more than being interrupted between tasks, that a sound of a real object reaches you faster than a beep. Those are not opinions we arrived at while designing an app. They are findings, and this page is where they are.
For each one: what the study actually showed, what we built because of it, and, where it matters, where the evidence stops and our own reasoning starts.
Your brain stops responding to a sound it already knows
The research
Habituation is specific to the particular stimulus rather than to the category it belongs to, it develops faster the more frequently and rapidly the stimulus repeats, it generalizes to similar stimuli in proportion to how similar they are, and it recovers after a rest.
How we make it happen
Every part of an alarm is drawn separately, so the arrangement is nearly always new even though the individual recordings recur. The biggest change of all comes between snoozes, because that is the one place in an alarm where repetition is genuinely rapid. Each library is also deliberately varied rather than sixteen versions of one noise, since near-identical sounds habituate almost as fast as a single one.
Being interrupted mid task costs far more than being interrupted between tasks
The research
Interruptions delivered at boundaries between subtasks produced less disruption to performance, fewer errors and less annoyance than the same interruptions delivered in the middle of a subtask.
How we make it happen
The preminder. Two toots a few minutes early, so you can reach a boundary you chose instead of being cut off at whatever moment the alarm happened to land on.
Where this stops being evidence
The study used computer based tasks with adult participants and was not about ADHD specifically. It tells us when to interrupt. It does not by itself prove that a warning beforehand helps, which is our own reasonable extension of it.
The hard part is stopping, and that is the part ADHD makes harder
The research
Barkley's model puts behavioral inhibition, the ability to suppress a response already underway, at the foundation of ADHD, with working memory, self regulation of motivation and affect, internalized speech and reconstitution all depending on it.
How we make it happen
The whole shape of the app, honestly. It is why the preminder exists at all, why an alarm asks you for a time and nothing else, and why nothing here adds a puzzle at the exact moment stopping is already hard.
Where this stops being evidence
This is a theoretical model supported by a review of evidence, not a single experiment, and parts of it remain debated. We use it for the one claim it is strongest on and that matches what people describe: stopping is its own demand, separate from whatever you are being asked to do next.
A sound that gets your attention, then words that tell you what it is
The research
Patterson's guidelines for civil aircraft recommend a limited set of distinctive warning sounds whose perceived urgency matches their priority, each followed by a keyword speech warning. The non-verbal sound captures attention; the speech carries the meaning.
How we make it happen
The order of every alarm: a railway bell first, then our conductor naming the stop. It is the same two-part shape, which is why it works at six in the morning.
Where this stops being evidence
Written for pilots in cockpits, which is about as far from a bedroom as a setting gets. What transfers is the structure, not the urgency levels.
Patterson, R. D. (1982). Guidelines for auditory warning systems on civil aircraft. Civil Aviation Authority, London. CAA Paper 82017.
A sound of a real thing beats an abstract beep
The research
Auditory icons, warning sounds that represent the thing they refer to, produced faster and more accurate responses than conventional abstract warning tones.
How we make it happen
The doors open on a pill bottle, a school bell or a kettle instead of a beep, so you know what the alarm is for before you have looked at the phone.
Where this stops being evidence
Demonstrated in driving and collision-avoidance contexts rather than with alarm clocks.
We had this wrong
Our own design notes credited this to Belz, Robinson and Graham. The third author is Casali. Corrected here on 3 October 2026 when the citations were checked against the sources.
Belz, S. M., Robinson, G. S., and Casali, J. G. (1999). A new class of auditory warning signals for complex systems: auditory icons. Human Factors, 41(4), 608-618.
How urgent a sound feels is a property you can design
The research
Perceived urgency varies systematically with measurable parameters of a sound, including pulse rate, fundamental frequency and harmonic content, so urgency can be engineered rather than guessed at.
How we make it happen
Escalation across the snooze cycle is carried by pace and brevity, not by volume and not by making the sound harsher. Each stage is shorter and quicker than the last, and the last call has an engine building under it.
Meaning gets through where sound alone does not
The research
In dichotic listening, people could report essentially nothing from the message they were not attending to. The one thing reliably found to break through that barrier was the listener's own name, which suggests it is material that matters to the person, rather than any acoustic property, that gets through.
How we make it happen
Recording your own voice saying your own reason is a first class feature rather than a novelty, and it is the one thing in the app that does not rely on novelty to keep working.
Where this stops being evidence
A 1959 laboratory study of selective attention, with a small sample by modern standards. The own-name effect has been replicated since, including recent preregistered work, but it is evidence about attention rather than about alarm clocks.
An if-then plan with a reason attached helps people act on what they already know
The research
Children with ADHD who turned a goal into an implementation intention, an explicit if-then plan, improved their inhibition of an unwanted response on a Go/No-Go task to the level of children without ADHD. A second study found implementation intentions plus stimulant medication produced the best performance of all.
How we make it happen
An alarm here is built to carry your reason for setting it and not just the time, which is what a recorded message in your own voice actually is.
Where this stops being evidence
This is children, and it is response inhibition measured on a laboratory task rather than medication taken on time in real life. Our own notes previously summarized it as improving follow through generally, which is stronger than what the study shows.
We had this wrong
Our own notes said this showed improved follow through in ADHD. What it showed was improved response inhibition, in children, on a Go/No-Go task. Narrowed here on 3 October 2026.
An alarm that cries wolf stops being believed
The research
Breznitz's experimental work on false alarms found that each false alarm measurably reduced the credibility of the warning system, and that the effect on trust can be substantial even when false alarms are rare.
How we make it happen
The last call is true. When our conductor says it is the final one, the app really does stop, so you never learn that his announcements can be safely ignored.
Where this stops being evidence
This is about false alarms specifically, warnings for threats that did not materialize. Applying it to an alarm that nags after the moment has passed is our inference, and a reasonable one, but it is ours rather than Breznitz's.
Novel sounds may grab the ADHD brain harder, and the evidence is mixed
The research
Children with ADHD showed more missed responses and a different brain response to unexpected novel sounds during a visual task, which the authors attributed to weaker control over involuntary attention.
How we make it happen
It is a reason for the variety to always carry meaning rather than being novel for its own sake. Every sound here tells you what the alarm is for, and none of it was chosen just to be surprising.
Where this stops being evidence
This is the weakest card in the deck and we are not going to pretend otherwise. It is children, the measure is distractibility rather than usefulness, and the wider literature on novelty and ADHD does not point all one way. Being more captured by a novel sound is not automatically a good thing: it is also what makes an unexpected noise derailing.
What this page is not
It is not a claim that Whistlestop is clinically validated, because it is not. Nobody has run a trial on this app. What is below the hood is a set of design decisions, each one taken because of a real finding about how attention works, and the honest description of that is “built on research” rather than “proven to work”.
It is also not medical advice, and it is not a treatment for ADHD. It is an alarm app, and a good alarm app should not be the only thing standing between you and something that would hurt you to miss. There is more on that in the terms.
How these were checked
Every citation on this page was opened and read against the published source on October 3, 2026, rather than copied from our own notes.
That turned up 2 places where our internal documents had something wrong: a misattributed author, and a finding we had been describing more broadly than the study supported. Both are corrected above, and flagged where they occurred rather than quietly fixed.
If you think something here is overstated, misread or out of date, we would genuinely like to know. Tell us and we will check it and change it.
Whistlestop is nearly here.
Leave your email and you will get one message the day it is out. No newsletter, no drip sequence, and your address goes nowhere else.
