ABX Semantics

[Tom Andry]

This whole discussion started as an aside in this thread. To be fair to the OP, I've split it to a different thread.

Please read this. It is important to remember - ABX is ONLY useful in ferreting out small differences. You don't need to ABX your cooking compared to a 5 star chef's. Even if you haven't eaten your cooking in years. You don't need to ABX a Yugo vs. a Porsche. The differences are so great no amount of time or personal bias can (realistically) alter the plain fact that they are different.

 

[Tomorrow]

Please read this. It is important to remember - ABX is ONLY useful in ferreting out small differences. You don't need to ABX your cooking compared to a 5 star chef's. Even if you haven't eaten your cooking in years. You don't need to ABX a Yugo vs. a Porsche. The differences are so great no amount of time or personal bias can (realistically) alter the plain fact that they are different.

ABXing is not done to prove that any two or more things are 'different', Tom. It's to aid in experiencing/describing 'how' they are different.

EDIT: Tom. You know that your wife (if you have one who'll tolerate you ) and the woman down the street are different. But until you A/B/uh...X them, you don't know how they are different. (Or do you? LOL.)

 

[Tom Andry]

You're wrong. The only reason to do a double blind test is to show differences. The researchers usually use a double tailed test which doesn't assume that one is better or worse than the other. Just different. After a significant result is found, the researchers go back and use other data to describe the differences.

 

[Tomorrow]

You're wrong. The only reason to do a double blind test is to show differences. The researchers usually use a double tailed test which doesn't assume that one is better or worse than the other. Just different. After a significant result is found, the researchers go back and use other data to describe the differences.

I'm wrong?! Many times but not on this one, ol' buddy. You have confused the two tests, as I suspected. I didn't say double blind test. I said A/B switching (X). DBT's determine human detectability of differences. A/B switching resolves the differences...not the fact of a difference (which is settled by the DBT).

And where are you getting this "better or worse" comment. Those concepts are judgemental and of personal opinion...which is irrelevant to testing of differing parameters.

By the way, look at my edit from the previous post....you'll see what I mean.

 

[Tom Andry]

ABX is a method of conducting a double blind test. The data that is gathered is used to show that there is a difference between two speakers. The user is asked to listen to two speakers playing the same passage of music (or whatever) and then asked to listen to a 3rd "unknown" set of speakers (the X in the ABX). The data is crunched to show that people could reliably differentiate between the two sets of speakers (cables/transports/whatever) by correctly matching the A/B to the X. AFTER THAT - the participants are polled to "describe" those differences. The ABX shows that there is a difference. What happens after the ABX describes those difference (and are separate from the the ABX test).

 

[Tomorrow]

ABX is a method of conducting a double blind test. The data that is gathered is used to show that there is a difference between two speakers. The user is asked to listen to two speakers playing the same passage of music (or whatever) and then asked to listen to a 3rd "unknown" set of speakers (the X in the ABX). The data is crunched to show that people could reliably differentiate between the two sets of speakers (cables/transports/whatever) by correctly matching the A/B to the X. AFTER THAT - the participants are polled to "describe" those differences. The ABX shows that there is a difference. What happens after the ABX describes those difference (and are separate from the the ABX test).

Tom, I wouldn't get so involved in this...you are usually more jovial and circumspect in your statements,....and less accusatory and incorrectly informed. I don't mind being called 'wrong', but not for the 'wrong' reason. So here goes.....

I don't know if you've ever had any formal education on scientific methodologies or psychological testing, but I think you've made some incorrect assumptions.

A/B switching need NOT be blind...and usually is not. It is a method of comparing elements of some activity/thing. (e.g. "Speaker A is dull. Speaker B is bright. I think I prefer the bright speaker.") One person can do A/B switching. It's purpose is only to compare elements. This method is great for determining one's preferences for types of sounds/music/cuts of beef, or whatever. It does not require a 3rd, or 4th, or nth speaker (though it could).

A DBT, on the other hand, is a formal test that provides information in some form from third parties who are 'blind' to the actual activity/thing being measured. (e.g. In a test of a new drug, the Rx or a placebo is given to a group of people who don't know which is which. The associated sequilae/effects are then collected/measured.) In the case of speakers, great pains are taken to construct a valid DBT, with issues such as placement, switching times, random population selection, and generally holding all audio and electronic variables as constant as physically possible. DBT usually involves the standard requirements of scientific testing, including the use of control groups, assumptions to be proved/disproved, etc.

It should also be mentioned that A/B switching can be a part of a DBT design, but need not be, depending upon what is being measured.

 

[Tomorrow]

Tom,

I see what you're saying about A/B/X being a very specific and limited form of A/B switching (which I am using in a general sense)....that is comparing 2 kinds of speakers to an unknown 3rd. I don't know why one would do that as opposed to comparing one to one, or four to one (A/B/C/D/X?), or whatever. But I see what you're referring to.

Sorry OP, this has gotten way off topic. My bad.

 

[MDS]

I thought the X in A/B/X was either A or B not an unknown third item.

The test is meant to deal with the case where one believes that A sounds superior to B (or vice-versa). If one has such a strong conviction as to which is better then when X is played he should be able to identify if it as the one that he thought was superior more often than chance alone would dictate (but it never happens that way. )

 

[Tom Andry]

I thought the X in A/B/X was either A or B not an unknown third item.

Yes, that is what I meant. It is only unknown which it is (A or B).

The test is meant to deal with the case where one believes that A sounds superior to B (or vice-versa).

Actually, the test is to detect if there is a difference between A and B. If there is truly a difference, than participants can match X to A or B correctly a certain number of times. Personal preference (likes or dislikes) don't come into play until after.

For example, if one speaker is blown and the other is working properly, the participants could likely identify X as correctly A or B nearly 100% of the time. This doesn't say that they like A or B any better. It just means that they can hear a difference. After the test, the researcher will come in and say, "So, you all could identify the correct speaker... why?" The answers will probably be something along the lines of, "Well, duh! That one buzzed and screeched and the other made beautiful music!" The researcher might even pass out a survey to gather specific qualitative judgments about the different speakers. But that is outside of the ABX test.

Of course, in this specific example, an ABX test is completely overkill. Most anyone could identify the differences without setting up so elaborate a test. That is why I said that double blind tests (ABX being a method) are really only useful to pick up small differences or to ferret out differences that are possibly being swayed by bias or othere external factors.

 

[Tomorrow]

Yes, that is what I meant. It is only unknown which it is (A or B).



Actually, the test is to detect if there is a difference between A and B. If there is truly a difference, than participants can match X to A or B correctly a certain number of times. Personal preference (likes or dislikes) don't come into play until after.

For example, if one speaker is blown and the other is working properly, the participants could likely identify X as correctly A or B nearly 100% of the time. This doesn't say that they like A or B any better. It just means that they can hear a difference. After the test, the researcher will come in and say, "So, you all could identify the correct speaker... why?" The answers will probably be something along the lines of, "Well, duh! That one buzzed and screeched and the other made beautiful music!" The researcher might even pass out a survey to gather specific qualitative judgments about the different speakers. But that is outside of the ABX test.

Of course, in this specific example, an ABX test is completely overkill. Most anyone could identify the differences without setting up so elaborate a test. That is why I said that double blind tests (ABX being a method) are really only useful to pick up small differences or to ferret out differences that are possibly being swayed by bias or othere external factors.

I...uh...er....am...have been...ambiguoused.

EDIT: And I happen to like buzzing, screeching speakers. Why the hell else would I own JBL's?

 

[Audioholics]

Tom, I wouldn't get so involved in this...you are usually more jovial and circumspect in your statements,....and less accusatory and incorrectly informed. I don't mind being called 'wrong', but not for the 'wrong' reason...

I don't know if you've ever had any formal education on scientific methodologies or psychological testing, but I think you've made some incorrect assumptions.

You have foot in mouth disease. Tom's background is in Psychology and has included studies in these styles of testing and more.

You are/were flat out wrong. What's frightening is the length to which you were determined to defend your wrongness.

We're all still "friends" here of course, but I wanted to straighten out the record.

 

[Tomorrow]

You have foot in mouth disease. Tom's background is in Psychology and has included studies in these styles of testing and more.

You are/were flat out wrong. What's frightening is the length to which you were determined to defend your wrongness.

We're all still "friends" here of course, but I wanted to straighten out the record.

Yes, Clint. We are all indeed "friends" here, and I appreciate your diagnosis of my "disease". I'm sorry I frightened you. Tom and I worked out our differences via PM last night and we have decided that we were BOTH wrong. (Note his reply to MDS about A/B/X'ing. And I misused the term to exclude all but A/B switching.)

NOW the record is straight.

 

[Audioholics]

Fair enough! .

 

[majorloser]

Tom's background is in Psychology.....

That explains a LOT!

The resident forum shrink.

 

[mtrycrafts]

The researcher might even pass out a survey to gather specific qualitative judgments about the different speakers. But that is outside of the ABX test.
.

Or, that grading of different aspects of the speakers, as Toole has done, known something like ABC/hr for some reason, is also conducted under DBT conditions. Then, the numerical grades are statistically analyzed to see if any aspect is statistically different between the speakers as in his paper:

http://www.harman.com/wp/pdf/AudioScience.pdf

specifically in figure 4, page 10.

 

[mtrycrafts]

We're all still "friends" here of course, but I wanted to straighten out the record.

Yes, of course, the record must be straight friends.

 

[mtrycrafts]

The resident forum shrink.

And, it is free Now, the fun begins, all out in the public