|
General Asylum General audio topics that don't fit into specific categories. |
|
|
|
General Asylum General audio topics that don't fit into specific categories. |
|
In Reply to: How about some REAL math? posted by chickenlogic on April 8, 2002 at 17:42:09:
People have done tests that would detect things like 60% chance of detection for a given stimulii.The results? It seems that the difference between "no detect" and "close to 100% detect" is very VERY small, in fact teeny enough that it's hard to hit anything between "nothing" and "nearly everything".
Your example was very carefully chosen as a straw man that creates an impression that runs counter to the experience of experimenters, in short, you built a straw man and then burnt it.
Now, how about something that describes how real subjects work? (Btw, your point that finding 60% results is hard is quite true, it's the existance of the 60% point that isn't.)
Some typical results are, say for masking, dropping by 1 dB drops you from "sure detection" to "no result", cold. No 60% chance, 50% mean until the subject revolts and chases you down the hall.
That's 1dB change in a probe signal 35dB down.
At the .5dB point, one subject was entirely reliable, the other entirely unreliable. At the 34.5 dB point both were effectively 100%, at 35.0 dB, one was close to 100% the other was random, and at 35.5 dB both were random to type 2 of about 54%.
So that's how the organism worked in at least one test. You need to account for how people do respond.
There are, btw, physiological and mathematical reasons why this happens. Effectively, one "trial" in terms of the test is NOT one trial in terms of listening, but many. ONE "detect" will suffice to sway the subject, and a few will make the subject (*&(*& certain.
What does this mean? It means that even though the subject may, on a per-chance-of-detection basis, have a low detection threshold, they may get, say, 500 chances in one trial, and if ONE of them trips, they will claim difference.So analyze THAT, now. In your 5 minute song, you've got a few chances to notice difference per second, minimum. Typically, at least one every 200 milliseconds. So, now, what detection chance will give you at least one hit in all those trials to give you your 60% chance?
Heh. Try it.
JJ
This post is made possible by the generous support of people like you and our sponsors:
Follow Ups
- Yes, let's do some math, shall we? - jj 04/8/0218:00:44 04/8/02 (2)
- Why certainly, my good chum...... - chickenlogic 00:53:57 04/9/02 (1)
- Um, you're missing my point entirely.... - jj 08:39:25 04/9/02 (0)
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2025, Audio Asylum, all rights reserved ]
