Re: Speech Perception: Chinese Whispers

From: Harnad, Stevan (
Date: Sun Mar 09 1997 - 13:50:48 GMT

> From: Briony Care <>
> In relation to the perception of speech, and therefore the perception
> of phonemes; is one way of our recognition failing in chinese whispers?

Hi Briony; sorry for the delay in responding to this one, but I had to
find out from somebody what "chinese whispers" meant!

> When we can't articulate phonemes (as when we are whispering), would we
> infer what they would be in relation to the context of other phonemes
> and the sentence meaning, and therefore make wrong inferences? This
> would then result in a slightly altered sentance; the failure to
> articulate phonemes would continue as the sentance was passed on, and
> so the complete meaning and words change. Therefore we would have a
> chinese whisper.

Yes, that would be part of the reason that the message changed (but
the same effect occurs in experiments on the propagation of "rumors"
even when they are not whispered, but passed from person to person).

But whispering is more impoverished then speaking aloud, partly because
it's quieter, but partly also because the voicing cues are gone, so that
it's harder to tell the difference "bad" and "pad," or "dot" and "tot."

You're also right (though I didn't mention it in the lecture) that it
is not only the phoneme context that helps us understand speech, but
also the meaning of the utterance: Whispering, if you heard "He did
something very pad" you would perceive it as "bad."

Besides, voiced and unvoiced consonants still have some different
features even when you whisper them. The difference is in how "airy"
the explosion is. Whispered, the difference between "pig" and "big"
sounds like p-hig vs. pig. A harder one is voiced vs. voiceless "th" as
in "thistle" versus "this'l" (contraction of "this will"): You can
choose which is which if you say them one after the other, but in
isolation it's harder to say -- but the meaning is there to help you

This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:51 GMT