The Weekly Newsletter of MIT Linguistics

Phonology Circle 9/19 - Erin Olson

Speaker: Erin Olson (MIT)
Title: Intermediate Markedness and its consequences for the GLA
Date/Time: Monday, September 19, 5:00-6:
Location: 32-D831

In the phonological acquisition literature, it has been observed that children sometimes acquire marked structures of the target language in a two-step fashion: they go through a stage in which they produce the marked structure only in some privileged position(s) within the word, before producing that structure in the full range of positions found in the target language. These stages have primarily been analyzed as being due to the ranking schema in (1) (Tessier 2009).

(1) Positional Faithfulness >> Markedness >> General Faithfulness

These stages have been shown to be problematic for gradual OT learning algorithms such as the GLA (Boersma 1997; Magri 2012), as these algorithms do not predict that children should ever go through such a stage (Jesney and Tessier 2007, 2008; Tessier 2009). As such, Jesney and Tessier (2007, 2008) advocate for using an HG-based learner, which is capable of predicting these stages.

In this talk, I will review Jesney and Tessier’s (2007, 2008) claim that the GLA is incapable of predicting intermediate stages, and I will show that this claim is premature. The GLA is capable of predicting such stages under the following conditions: a) the intermediate stage can be characterized by the ranking in (2):

(2) Positional Markedness >> Faithfulness >> General Markedness

and b) if it cannot be characterized in this way, then the Positional Faithfulness constraint that decides the error is ranked low in the grammar. I will also discuss how multiple, successive intermediate stages can be predicted, and set out a typology of possible intermediate stage orders.