parity as a ground for gestural theories of speech perception

Notion of parity seems to play an important role in the arguments used by the proponents of gestural theories of speech perception. As Liberman and Whalen wrote in their 2000 paper, interlocutors must agree on which information present in the speech is considered linguistic, that "/ba/ counts, but a sniff does not". Goldstein and Fowler (2003) consider it important that articulatory phonology, whose very basics they outline, is a "parity-fostering" system and is thus more "natural" than other approaches to speech perception.

But is it really so important and if it is - does it really call for a gestural approach? In a revision of the motor theory of speech perception Galantucci, Fowler and Turvey (2006) regard two subtypes of parity requirments
- there must be an agreement between language users on what count as a communicative message and what does not
- sent and received messages must be the same.
and they tightly connect it with a claim that perception and production should have coevolved to allow the first two. While second claim is obviously true, though with one important correction, the first claim seems to be unnecessary.

What is the final goal of a speech perception act? To capture the message, i.e. the concepts and relations between them the talking person is trying to convey. We are not transmitting phonemes, syllables, or individual words, we are transmitting ideas. The ideas indeed should end up being more or less the same, but any congruence below that level is of zero importance, at least there's no special reason for that.

Imagine a dialog of a non-signing individual with a dumb but hearing one, with the latter drawing pictures to tell, say, what is to be bought from the shop or what animals he has seen on a trip to Africa. Can such a dialog be successful? Sure. Is there any common ground that both persons use to exchange their messages? Nothing below the concept level. If the parity is to be supported only at the level of the concepts it can not suggest what kind of representations are used at the lower levels, it doesn't care.

The first claim, that there should be an agreement on what count as a communicative message hardwired in our speech perception circuitry seems unnecessary. Artficial situations like sinewave speech aside there's no necessity for the speaker to raise a sign "Me is talking speech" every time he is trying to communicate. The listeners might simply try to guess "what could it be?", build the models of what's happening and assign specific probabilities to them every time they hear something.

There is no need to use exactly the same units in communication. The only thing that matters is that listener should be able to build a model spanning from sound to meaning that would be more adequate than any non-linguistic model of the situation.

Sniffs, grunts, and mehs sometimes contribute to our perception of the situation, sometimes don't, but there's absolutely no necessity to invent a limitation that would prevent us from trying to incorporate them in our interpretative model. A sniff presumably does not have any correlate in the mental lexicon of an average english speaker - very well then, it will be discarded in the next iteration of the model. Since it has no correlate it can not be misleading, just some noise that can't corrupt such a superfluous code as speech. Moreover, after certain speech exposure irrelevant parts of the signal might get ignored on the simple statistical basis. Why should it be hardwired from the beginning?

So, parity in form of an agreement between speakers seems to be unnecessary, while "sameness" of the message is several levels higher than the speech gestures. No parity argument for the gestural theories then.


Galantucci, Fowler, Turvey. 2006. The motor theory of speech perception reviewed.
Goldstein, Fowler. 2003. Articulatory phonology: a phonology for public language use.
Liberman, Whalen. 2000. On the relation of speech to language.