To encourage experimentation with objective testing because I think people should try it even if they don't like it or think it's futile, or will ignore the results.

There are some rules for obejective testing, such as it must be blind, level matched and there must be a series of trials (16 or so). So, if you are a subjectivist and want to show that objective testing is worthless, then these are the groundrules for getting started. I say this because there are a handful of objectivists, and some middle-ground types who haven't got the rules straight.

As an aside this topic boiled over again, when Wadia Meister (the subjectivist) alluded to a blind test at the bake-off.

Now let me clear up a couple more factual errors

Steve Toy: 13 out of 16 is better than 5 out of 5 - it's a statistical fact not an opinion..

PeteH: I can't decide which side of this debate your on  Your cetainly not help by quoting Stereophile who cynically interpreted Levinthal's paper for their own ends.

Moreover, you say,  as Leventhal demonstrates - is that the differences involved are small and not readily apparent under conditions such as hitherto normally used for ABX testing. This is NOT what Leventhal said at all - it may be what Stereophile think he said. Leventhal is talking about the probability of detection being small, not the quality difference being small.

Now, as we've heard on many occasions here and elsewhere, the subjectivists report obvious differences which are readily detectable, so that the probability of detection is high. It's about detectability. It's not the magnitude of the quality difference - which is good becuaes no one can agree on whether changes are small or large - but on how readily detectable the change is.

Now, let's turn that around for fun If we double blind test people, with say 16 trials, and they hear no difference (and I won't list a string of papers where that is the case), the consolation prize is that we may have commited a statisical error but that for that to be the case their ability to discriminate has to be moderate

Let's suppose you review what I've written and refute double blind testing on the basis that the detectability is small. Then I ask you what is the point of any kind of dealer demo? Or more interesting, if you undertake a small number of trials (AB swaps) at a dealer, that your purchase isn't based on random choice?

And this is the dilemma:

If you think the detectability is high "I can easily tell the differences", you should prove it in a blind test.

If you think the detectability is low, accept that your choosing your toys at random.

 

I would be deeply insulted by the Goodmans quip, except for the fact that my "entry level" three way active speakers were designed by an ex-Goodmans engineer called Billy Woodman