10.04.2013 Views

The contrastive hierarchy in phonology 2009 Dresher.pdf - CUNY ...

The contrastive hierarchy in phonology 2009 Dresher.pdf - CUNY ...

The contrastive hierarchy in phonology 2009 Dresher.pdf - CUNY ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

as non<strong>contrastive</strong> for [back], the advantage shifts back to the MCS analysis. For<br />

<strong>in</strong> the MCS theory, it is a given that learners must arrive at some <strong>contrastive</strong><br />

<strong>hierarchy</strong> for their language. Moreover, this same <strong>hierarchy</strong> must account not<br />

just for the behaviour of velar consonants and the feature [back], but also for the<br />

other phonemes and features as well. Thus, multiple sources of evidence exist<br />

bear<strong>in</strong>g on the construction of the <strong>contrastive</strong> <strong>hierarchy</strong> of a language (see<br />

further §8.7).<br />

It is not clear whether there are comparable limits on DT theory. <strong>The</strong><br />

idealizations made to account for post-velar front<strong>in</strong>g are particular to that<br />

problem. Noth<strong>in</strong>g <strong>in</strong> the theory as presented <strong>in</strong> Padgett 2003a would prevent us,<br />

for example, from group<strong>in</strong>g together East Slavic velars and labials aga<strong>in</strong>st<br />

coronals if that suited a different problem.<br />

8.4. Structured Specification <strong>The</strong>ory and contrast via natural classes<br />

Structured Specification, or natural classes, theory (Broe 1993, Frisch 1996, Frisch,<br />

Pierrehumbert and Broe 2004) proposes that phonological processes are sensitive<br />

to similarity: harmony processes and co-occurrence restrictions affect segments<br />

that are highly similar. A similarity metric is derived from the natural classes<br />

created by the phonemes of an <strong>in</strong>ventory; it measures similarity by comput<strong>in</strong>g<br />

the feature classes shared by two phonemes, and divid<strong>in</strong>g by the number of<br />

shared plus unshared classes. This method of comput<strong>in</strong>g similarity weights<br />

features differently depend<strong>in</strong>g on how much they contribute to creat<strong>in</strong>g dist<strong>in</strong>ct<br />

natural classes. Features that uniquely dist<strong>in</strong>guish members of the <strong>in</strong>ventory will<br />

370

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!