Blind test

From Hydrogenaudio Knowledgebase
Revision as of 01:17, 21 August 2005 by Foot (Talk | contribs)

Jump to: navigation, search

The concept of blind test was first introduced in medicine, in order to determine the effect of drugs on test subjects without coloring the results with subjective responses from both the subjects and experimenters.

A blind medical test is a test where the patient doesn't know if he is taking a real product or placebo pills.

A double blind medical test is a test in which the patient doesn't know if he is taking a real product or placebo pills, nor does the physician who gives them to him.

Such tests are required in order to prove that when someone takes a new product, the product itself has an action on the disease, and that if a healing occurs, it is not completely caused by the natural defenses of the human body or by the placebo effect.

In order to adapt this definition in audio, we can say that:

A blind listening test is a listening test in which the tester doesn't know what source he is listening to.

A double blind listening test is a test in which nobody in the room know which source is being played.

Blinds listening tests are required in order to eliminate any bias that would lead the listener to prefer one source over another for personal reasons or preferences, that would have nothing to do with the sound. People also often imagine hearing differences between, say, a CD and an MP3 coming from this CD, just because knowing that they are listening to an MP3 disturbs them. When the test is blind, the listener doesn't know if he is listening to a CD or an MP3, so he can't tell that the MP3 sounds bad just because he doesn't like MP3. It can only tell it if it does sound bad. In Hydrogenaudio, we use blind test mostly in order to judge the sound quality of codecs, but they can also be used to listen to any piece of audio device. The prefered tests for audio codecs are ABX tests.

Double blind tests are prefered in medicine. There are evidences that when the physician knows what he is giving, the effect on the patient is affected. The situation in audio tests is a bit different, but a double blind test is always more rigorous than a simple blind test, because any influence from the operator on the listener is ruled out.

Setting up double blind listening tests is very difficult in hifi. Either it requires very much money, for example to compare two amplifiers, 8 hidden units of each model would be connected to a switch, and no one in the room would know which of the 16 possible sources is what. Either it is very complicated and time consuming: in order to switch between the two amplifiers, the listeners would have to leave the room, and the operator would come in, close the door, choose to switch or not the devices, leave, and tell the listeners to come in again.

Fortunately, if we compare codecs on a computer, double blind tests are very easy to perform. When a program like ABC/HR plays a random file for us to identify, since we have no way to know if it is the original file or the encoded file, we can say that the test is double blind.