How to check a strange audio recording

Speakers colour sound. They add room echo, mask quiet detail, and can make noise from one source sound like it is coming from another direction. Use headphones — any headphones — and listen to the full recording from the beginning, not just the section you are interested in.

As you listen, note the exact timestamp of the anomaly. Write it down. You will need it in the next steps. Also note what was happening immediately before and after — context matters more than you might expect.

Audacity is a free audio editor available at audacityteam.org for Windows, Mac, and Linux. Download it, install it, then open your file: File → Open → select your recording.

The main view shows the waveform — a graph of the sound's volume over time. Look at the section around your timestamp.

• A spike in the waveform means a sudden loud sound occurred.
• A flat or nearly flat line means the sound is very quiet, buried in background noise.

If your anomaly shows up as a clear spike, it is a real sound the microphone captured. If the area looks flat and featureless, the sound may be extremely faint — which matters when you reach the pareidolia test below.

In Audacity, click the small arrow next to the track name (the dropdown next to the track title on the left side of the screen). Select Spectrogram. The view changes to show frequency content over time — different types of sounds appear as different coloured bands.

Now look at your anomaly and ask:

Is there a constant horizontal line at the very bottom of the display? That is mains hum — electrical interference from wiring or appliances. In the UK and Europe it sits at exactly 50 Hz. In the US and Canada it sits at exactly 60 Hz. This is one of the most common anomalies people notice, and it has a completely ordinary cause.

Does the anomaly appear as a brief flash of colour across all frequencies at once? That is a broadband burst — typically caused by a phone receiving a notification, a device vibrating, or a nearby electrical switch.

Is the sound concentrated in the low-frequency range, smearing across the bottom of the display? That is wind or mechanical rumble — heating systems, pipes, or traffic outside.

Does the sound fall in the range of roughly 85 to 3000 Hz with layered bands? Human voices occupy this range. But so do many other things — including interference artifacts from compressed audio files.

Before deciding the sound is unexplained, account for everything that was physically present. Ask yourself:

• What electrical devices were running nearby — a fridge, a boiler, a router, a fan?
• Was the building old? Older timber and masonry creaks and settles as temperature changes, especially at night.
• Could there be wildlife? Mice and rats behind walls, birds in roof spaces, and insects produce sounds that are easy to misidentify indoors.
• Was the recording device near its own speaker, charger, or another phone? These introduce interference directly into the recording.
• Was the audio file compressed — sent via a messaging app, or saved as a heavily compressed MP3? Compression algorithms can introduce "pre-echo" and smearing effects in quiet passages. These artefacts sometimes sound like whispers or voices.

Audio pareidolia is the tendency to hear meaningful sounds — especially words or voices — in random noise. It is a well-documented feature of how the brain works. Because the brain is highly tuned to detect speech, it will sometimes impose the pattern of a voice onto low-level noise, particularly if you are already expecting or hoping to hear something.

The test is simple and important. Find someone who has not heard the recording. Play them the clip with no explanation and no prompting. Ask: "What do you hear?" Do not tell them what you think is there.

• If they independently describe the same sound, the same word, or the same pattern — that is more significant.
• If they only hear it after you tell them what to listen for — that is strong evidence of audio pareidolia. This does not mean you imagined it. It means the brain filled in a pattern from ambiguous noise. It happens to everyone.

Low-quality recordings produce more artefacts. Check: was the device at its recording limit (very close to a loud sound, or the input level too high)? Clipping — where the microphone is overloaded — produces distortion that can create voice-like effects. Was the device old or damaged? Worn microphone membranes introduce noise. Was the file converted, edited, or resaved multiple times? Each conversion step can degrade quality and introduce new artefacts.

If the anomaly only appeared after the file was shared or converted, that is important context.