wavoman
Headphoneus Supremus
- Joined
- Jan 19, 2008
- Posts
- 1,873
- Likes
- 45
Quote:
I do not think this is so. ABX is only one of many protocols, and there are good and bad things about it. Check the literature on sensory testing GIYF.
To determine in a blind test if A and B can actually be differentiated, there are other protocols -- my fave does false ("swindle") comparisons (presenting both A's or both B's, but claiming to be presenting A and B ... lots of fun!).
You can certainly ask the direct question: "Can you perceive any difference betwen this ... and this". You don't need to ask the somewhat artificial question posed by ABX. In fact I have argued elsewhere that asking the more complex question "do you like this better than ... this" can actually lead to more accurate results than the difference question, or the (strange and off-putting IMHO ABX question). It has to do with eliminating certain kinds of response bias.
We agree on this -- you must make proper statistical tests. Adding swindles (both A or both B) actually ups the statistical power of these tests in the "can you tell A from B" protocol. And you have to randomize the order presented, etc. etc.
The food industry (taste tests) almost never uses ABX, they use one of about 10 direct choice protocols. Audio for some reason -- not a good reason I'll bet, just history -- uses ABX. No real need ... many ways to do blind testing.
Originally Posted by royalcrown /img/forum/go_quote.gif It would be much more preferable if they did a proper ABX test and asked the panel to identify which component is which, rather than just asking them what sounds better, although you'd still need to test properly for statistically significant results. |
I do not think this is so. ABX is only one of many protocols, and there are good and bad things about it. Check the literature on sensory testing GIYF.
To determine in a blind test if A and B can actually be differentiated, there are other protocols -- my fave does false ("swindle") comparisons (presenting both A's or both B's, but claiming to be presenting A and B ... lots of fun!).
You can certainly ask the direct question: "Can you perceive any difference betwen this ... and this". You don't need to ask the somewhat artificial question posed by ABX. In fact I have argued elsewhere that asking the more complex question "do you like this better than ... this" can actually lead to more accurate results than the difference question, or the (strange and off-putting IMHO ABX question). It has to do with eliminating certain kinds of response bias.
We agree on this -- you must make proper statistical tests. Adding swindles (both A or both B) actually ups the statistical power of these tests in the "can you tell A from B" protocol. And you have to randomize the order presented, etc. etc.
The food industry (taste tests) almost never uses ABX, they use one of about 10 direct choice protocols. Audio for some reason -- not a good reason I'll bet, just history -- uses ABX. No real need ... many ways to do blind testing.