Re: checking the acoustical difference

[Pallier Christophe <pallier@xxxxxxxxxxxxx>]

On Mon, 7 Jun 2004, Daniel Kislyuk wrote:

> I want to prove that the MMN that I observe is of the phonetic origin 
> while the acoustic difference is so faint, that it won’t be noticed 
> and won’t elicit MMN in a non-speech context.

> What kind of approach would you consider to be more reasonable and why?

I do not like very much the idea of changing the context of
the cue because perception is holistic... Therefore, the perceptual
distance could be modified, regardless of the phonetic role of the cue.

Another possibility is to create another token of 'ba' that is at
the same acoustic distance from the standard 'ba' than deviant 'va'. If
you show less MMN with this new 'ba' as oddball, this could be
taken as evidence for a phonetic contribution to the MNN. 

Yet, again, the demonstration would not be totally convincing,
because one could still critisize your notion of acoustic
distance.  

In my opinion, the best design contrasts speakers with different phonetic 
categories. Any difference between the group must be attributed to
learned phonetic knowledge. (see, e.g., Naatanen et al. Nature. 1997;
385(6615):432-4). 

Now that I think about it, rather than trying to "fight" against the
acoustic/perceptual factor, it could be nice to show that two
*clearly distinguishable* contrasts (with "equal" acoustic distances), 
one phonemic and the other non phonemic, produce different MMN.
If one does that with several phonemic contrasts, and find 
similar MMNs when the contrasts are phonemic, then this would be good
evidence for a phonemic MMN. I don't know of such a study has been done.


Christophe Pallier
www.pallier.org