How to Run a Blind Listening Test on Your Own Gear
I spent 180 euros on a small headphone amplifier and spent the next week telling people it had opened up the percussion in a soundtrack I know note by note. A friend who works in audio measurement let me finish, then asked me to prove it.

I spent 180 euros on a small headphone amplifier and spent the next week telling people it had opened up the percussion in a soundtrack I know note by note. A friend who works in audio measurement let me finish, then asked me to prove it. We wired both sources through a switch box, matched the levels, and she ran twelve trials while I sat with my back to the desk. I got seven right. Guessing gets you six.
That afternoon changed how I buy audio equipment, and it did not make me a cynic. Some of the differences I believed in survived the same test easily, and the ones that survived are the ones worth money. A blind test at home takes about forty minutes to set up, costs nothing, and it is the only way to tell which of your impressions belong to the music and which belong to the price tag.
Why Listening With Your Eyes Open Fails
Expectation is not weakness, it is how hearing works. Your auditory system fills gaps with prediction constantly, which is why you can follow a conversation in a noisy cafe, and that same machinery happily fills in detail you expect a 400 euro cable to deliver.
The second problem is memory. Auditory short term memory for timbre decays in roughly ten to twenty seconds, which is less time than it takes most people to unplug one device and plug in another. By the time you have swapped, you are not comparing two sounds. You are comparing one sound to your memory of a sound, and memory is where the expectation lives.
There is also order bias. The second thing you hear in any pair tends to sound slightly better, because you know the music a little better by then and your attention has settled. Audio shops have understood this for decades, and it is why the unit they want you to buy gets played second.
Level Matching Is Most of the Test
If you do one thing from this article, do this one. A difference of 1 dB in loudness is heard as more detail, more presence, better bass, tighter imaging, almost anything except what it is, which is slightly louder.
Match within 0.2 dB if you can. Play a 1 kHz sine tone through each source and measure at the same point with a phone sound meter app held in a clamp, or if the output goes to speakers, use the app on a tripod rather than in your hand. For headphone outputs, the cheap approach that works is a multimeter set to AC volts across the output with the tone playing, adjusting until both read the same voltage. Expect to be surprised: two sources at the same volume number on the dial were 1.4 dB apart in my case, which is comfortably enough to explain everything I thought I had heard.
Unmatched levels are the reason most online comparison videos are worthless, and the reason most people who try a blind test for the first time discover the difference they were sure about has disappeared.
Running the Test Without Buying Anything
For files, the simplest route is a free media player with an ABX plugin. It plays A, plays B, then plays X which is secretly one of them, and you decide which. It keeps score and reports the statistics for you, which removes the temptation to stop counting when you are ahead.
For hardware you need a person. Put both devices in place, level matched, and have someone else switch while you look away. Agree on a signal so you know when a trial has started. If nobody is available, you can still do a useful version with files alone, since most of the claims people argue about live in the file rather than the box.
Keep the samples short, five to ten seconds, and loop the same passage every trial. Long samples feel more thorough and are worse, because your memory fades across them. Choose the passage carefully: cymbal decay, a reverb tail, applause, a sustained string chord, or the high sparkle on a hi hat. These are where codecs and conversion struggle. A thick orchestral tutti hides almost everything.
Then do enough trials. Sixteen is the number I use, and you need twelve correct out of sixteen before the result means anything at the usual significance threshold. Eight or nine correct out of twelve is not evidence of anything, and this is the point where most informal tests quietly declare victory.
What Survives a Blind Test and What Does Not
The table below reflects my own results and those published by people with better rooms than mine. Treat the right hand column as a prior rather than a verdict on your ears.
| Claimed difference | Realistic odds of hearing it blind | Where it shows up first |
|---|---|---|
| Different headphones | Always, and quickly | Everywhere. This is the only purchase with a guaranteed effect |
| Eartip or earpad change | Very high | Bass level and midrange tone |
| 128 kbps MP3 against lossless | High for most trained listeners | Cymbals, applause, sibilance on vocals |
| 320 kbps or good AAC against lossless | Low, under good conditions | Pre echo on a sharp attack, dense cymbal work |
| An external converter against a modern phone output | Very low, unless the load is demanding | Loudness differences, hiss on sensitive earphones |
| Expensive analogue cables | Effectively zero | Nowhere, in any properly run test I have seen |
| A 1 dB equaliser shelf | Moderate, with practice | Perceived weight and warmth |
| Speaker position moved 20 cm | High | Bass evenness and stereo width |
The pattern is worth sitting with. Everything near the top is physical: what sits against your ear, and where sound sources are in a room. Everything near the bottom is electrical and already far below the threshold of audibility in competent modern gear. We made the same argument about one of those rows in why expensive cables genuinely do not improve your sound.
Trust Your Ears Is Not the Advice You Think It Is
The phrase gets used to dismiss blind testing entirely, usually with the argument that measurement cannot capture musical enjoyment. The first half of that is right and the conclusion does not follow.
Your ears are not the weak link. Ear anatomy resolves differences far smaller than the ones in most of these arguments. The weak links are the twenty second memory, the order effect and the knowledge of what you are listening to, and a blind test removes exactly those three things and nothing else. You still listen with your own ears, to your own music, in your own room.
One honest limit, and it matters. A failed test does not prove two things sound identical. It proves that you, in that room, on that passage, on that day, could not tell them apart. A difference below your threshold is still a difference, and some people do pass tests that most of us fail, particularly on codec artefacts. What a failed test does establish is that the difference is not large enough to justify what you were about to spend, which is the practical question. The same reasoning sits behind our note on whether headphone burn in changes the sound.
What to Do With Your Result
If you fail to hear a difference, keep the cheaper option and move the money up the chain to something in the top half of that table. Fit, pads, placement and the headphone itself are where spending reliably pays.
If you pass, you have learned something genuinely useful about your own hearing, and the passage that gave it away is the part to write down. In my case I can pass a 128 kbps test reliably on cymbal decay and fail it on anything else, which tells me what to listen for and what to stop worrying about. Keep a short text file of the results with dates. A year later it stops arguments with yourself in a shop. The rest of our equipment notes sit in the listening section.
FAQ
How many trials do I need for a blind listening test?
Sixteen trials with twelve or more correct is a sensible home standard. Ten correct out of sixteen is within the range of guessing. If you want a stricter result, do twenty five trials and look for eighteen correct, but fatigue starts working against you past about half an hour.
Can I test headphones blind on my own?
Not properly. You can feel which headphone is on your head, so there is no blind condition to create. Headphone comparisons are the one case where sighted listening is unavoidable, which is also the case where the differences are large enough that it matters least.
Does a blind test work for streaming quality settings?
Yes, and it is one of the easiest tests to set up, since you can switch quality in the app on the same track. Level matching still applies, because some services apply different loudness normalisation per tier, and that alone can decide your answer.
Why do I hear a difference until I test it blind?
Because knowing which device is playing is itself an input to perception, and because unmatched levels and swap delays both push you towards the answer you expect. Nothing is wrong with your hearing. The test simply removes the parts of the comparison that were never about sound.
More in Listening
Listening
How Eartip Fit Changes the Sound of Your Earbuds
I spent two weeks convinced my new earbuds were broken.
Listening
How to Hear Syncopation and Clap the Offbeats Right
The clavinet riff in Superstition starts half a beat before you expect it to.
Listening
How to Hear a Key Change in Music You Already Know
There is a moment in a lot of songs where everything suddenly feels taller.