Reading, the test the internet invented
The arm test, and what it tells you in thirty seconds
Nobody taught women this. They worked it out on their own, in review sections and forum threads, because it is the fastest way to find out whether the number on the screen has anything to do with the sound going into the microphone.
Three public recordings. One is a baby, one is a mother, one has nothing in it, and the engine says which.
Firstsong is not a medical device, it is not a doppler and it does not use ultrasound. Nothing on this page is a way to check on your baby. If you have noticed fewer movements than usual, or a change in the pattern you know, call your midwife or maternity unit now, before you test anything.
What the test actually is
Three placements, one after the other, with the same app and the same settings. It takes less time than reading the description in the App Store.
- Your bare forearm, phone flat against the inside of it, where you can feel your own pulse if you press. There is a genuine heartbeat here, and it is an adult one.
- A cushion or a folded towel. There is nothing in a cushion. Any number that appears here came out of the software, not out of the world.
- A hard table, phone face down, room quiet. Nothing here either, plus the sort of low frequency rumble that fools bad detectors.
An honest app produces an adult rate on the arm, close to the pulse you can count yourself, and says out loud that it looks adult. On the cushion and the table it produces nothing, and tells you it found nothing. Three placements, two of them silent.
What it is actually testing
It separates a measurement from a decoration. A measurement is driven by the sound arriving at the microphone, so changing the sound changes the answer, and removing the sound removes the answer. A decoration is driven by whatever makes the screen look convincing, and it produces the same reassuring range no matter what you point it at.
There are three failure signatures and they mean different things.
- A number on the cushion. This is invention, full stop. There is no rhythm in a cushion to be wrong about, so a rate appearing there was generated rather than detected.
- A fetal looking number on your arm. There is a real rhythm here, so the app is not making it up from nothing, but a number in the 130s or 140s from an adult wrist means it is either doubling your pulse or has no idea which range it is looking in. That doubling problem, in full.
- A number that barely moves between all three. If the arm, the cushion and the table all land in the same comfortable band, the input is not reaching the arithmetic at all.
Run each placement three times rather than once. A real measurement wanders a little and tracks what you did. An invented one clusters in a narrow, pleasant range, because that range is what it was written to produce.
Why this test exists, and why the category earned it
Because the honest answer in home listening is often no, and no does not sell. An app that ends a session with nothing gets one star and a refund request. An app that shows a plausible number gets a screenshot posted to a group chat.
Read the reviews of the biggest apps in this category and the same two sentences come round again and again. It did not work. I think that was my own heartbeat. Women worked out the arm test because those two sentences kept happening to them, and because nobody selling these apps was going to hand them a way to check.
The uncomfortable part for us is that this test is aimed at everybody in the category, and it should be. We are asking you to trust an app you have not installed, sold by people you have never met, about the one subject where a false reassurance costs the most. The correct response to that is not faith. It is a cushion.
What actually helps when you run it
- Use headphones, so that the phone speaker is not feeding sound back into the microphone and inventing a rhythm for you.
- Count your own pulse first. Thirty seconds, doubled. Now you know what the arm should produce, and you can spot a doubled version of it instantly.
- Keep the room the same across all three placements. If the fridge starts up between test one and test three, you have measured your kitchen instead of the app.
- Give each placement the full time the app normally takes. An app that needs thirty seconds on a belly should be given thirty seconds on a cushion before you judge it.
- Screenshot each result. Three screenshots is a complete review, and it is the most useful thing you can leave for the next woman reading the App Store page at midnight.
When to call your midwife
None of this tells you anything about your baby, and it is not meant to. It tells you whether a piece of software is honest.
What tells you about your baby is movement. Fewer movements than usual, or any change in the pattern you know, is worth a call at any hour, as many times as it happens. The NHS guidance is here and Kicks Count exists for this one purpose. Testing an app is a good use of a calm afternoon and a terrible use of a frightening night.
How Firstsong handles this
Firstsong ships the arm test inside the onboarding, before the paywall, on purpose. You are walked through the forearm and then a cushion, in that order, before you are asked for anything. On the arm you get a rate in the adult range and the app says in words that this looks like an adult heart, not a baby. On the cushion you get nothing and the app says nothing was found. Then it asks you to pay. Putting our own refusal in front of the price is the most persuasive thing we own, and it costs us the customers who wanted to be told what they hoped for.
Runs in this tab on public research audio. Nothing is uploaded and there is no account.
One email, the day it opens
Firstsong is not on the App Store yet. Leave an address and you get exactly one message when it is, with the price on the first line. No newsletter, no drip, no partner offers. You can reply to that email and a person reads it.
Stored on our own server, never sold, never shared. Your IP is hashed, not kept. Nothing about your pregnancy is asked for or stored.
The bottom line
The arm test costs nothing, needs no expertise and cannot be argued with. An app that produces a heart rate from a cushion has told you exactly what it is, and no amount of five star reviews changes what you just watched happen on your own kitchen table.
Run it on every app in this category, and run it on ours the day it ships. If Firstsong ever prints a number over a cushion, that is a bug worth telling us about, and we would rather hear it from you than not hear it at all.
Common questions
What should a good app show on my arm?
A rate close to the pulse you can count yourself, labelled as adult rather than fetal. Silence on an arm is acceptable if the app only ever looks in the fetal range, but a number in the 140s from a wrist is not, because that is your own pulse doubled and dressed up.
What if it shows nothing anywhere, including my belly?
Then you have learned that it is not inventing anything, which is the first thing worth knowing, and separately that it is not finding your baby. Those are two different problems. The reasons a phone finds nothing are worth reading before you decide which one you have.
Does a real detector ever produce a number from a cushion?
It should not. Any well built detector has a threshold below which it publishes nothing, and a cushion is far below any sane threshold. If a number appears, either the threshold does not exist or it is set so low that it is decorative.
Is this test unfair to apps that are trying?
No, and the apps that are trying pass it comfortably. It only punishes designs that cannot say no, and being unable to say no is not a small flaw in this category. It is the whole flaw.
Will Firstsong pass its own test?
That is the point of putting it in the onboarding where you can watch. On an arm it reports an adult rate and names it as adult, and on a cushion it reports nothing. If it ever does anything else, the test is right and we are wrong.
Keep reading
- Is that my heartbeat or the baby's. why an arm can produce a number that looks like a baby
- What the silence actually means. the difference between an app that cannot hear and an app that will not lie
- Home dopplers. the objection that has nothing to do with ultrasound
- Hear the difference. the arm test idea, run on real research audio in your browser
Last checked 2026-08-23