[PLUG-TALK] Yanny/Laurel

John Jason Jordan johnxj at gmx.com
Sun Sep 8 10:03:40 PDT 2019


On Sun, 8 Sep 2019 08:12:34 -0700
Denis Heidtmann <denis.heidtmann at gmail.com> dijo:

>John-I would like your comments:
>https://www.theguardian.com/technology/2018/may/16/yanny-or-laurel-sound-illusion-sets-off-ear-splitting-arguments

I read the article and my comment is that all the observations made by
the various people cited were valid. 

One of the points I would like to emphasize is what linguists refer to
as 'top-down' processing v. 'bottom-up' processing of input information.
According to the article Prof Hugh McDermott alluded to this, but it
bears elaborating.

Bottom-up processing (in phonetic input) is the actual sound heard by
the ears. Most commentators in the article concentrated on variation in
the input - frequency, dialect of the speaker, quality of the output
device, etc. These are all bottom-up arguments.

To-down processing happens when the brain uses previously acquired
knowledge to interpret the sounds. As McDermott put it “If you heard a
conversation happening around you regarding ‘Laurel’ you wouldn’t have
heard ‘Yanny’. In his example, context is key to understanding. But
context is not all that we use when we interpret what we hear - even
things as basic as our knowledge of the language make a difference. For
example if you hear the word 'the' you know that the following word
must be a noun or a nominal modifier. Even interpreting the sounds that
you hear is impacted by top-down processing. For example, if you speak
English fluently then you have acquired the very complex set of English
phonological rules - e.g., if you hear an [s] then you know the next
sound must be a vowel or a stop, but not an [f] or a [v] (not counting
foreign words).

In my case I heard 'Laurel,' but I must point out that I have never ever
heard 'Yanny.' And I must also emphasize that I have taken graduate
level courses in both articulatory and acoustic phonetics, where I had
to train my ears to listen acutely. A research rule in linguistics is
never to use data from a linguist.


More information about the PLUG-talk mailing list