<OT> New Posting: ROA-1036

[roa at ruccs.rutgers.edu]

ROA 1036-0609

The naturalness of product-oriented generalizations

Michael Becker <michael at linguist.umass.edu>
Lena Fainleib <fainleib at post.tau.ac.il>

Direct link: http://roa.rutgers.edu/view.php3?roa=1036


Abstract:
Allomorph selection by markedness constraints, which assess
outputs only, is standard practice in Optimality Theory
(Mester 1994 and onwards). This practice predicts that allomorph
selection will be sensitive only to properties of the affixed
output form; i.e. allomorph selection is product-oriented
(Bybee & Slobin 1982, Albright & Hayes 2003). We present
evidence from an artificial language experiment, showing
that speakers are biased to prefer product-oriented generalizatio
ns.

Our language is based on Hebrew, where the plural affix
[-ot] is preferred over [-im] when the stem contains [o].
In real Hebrew, all plural stems with [o] also have an [o]
in the singular stem. In an experiment where [o] was present
only in the singular or only in the plural, participants
preferred pairing the allomorph [-ot] with the plural stem
[o], showing a bias for a product-oriented generalization
in the absence of positive evidence for it from real Hebrew.

Comments: 
Keywords: allomorph selection, product-oriented / source-oriented, cloning, artificial language experiment
Areas: Phonology,Morphology,Learnability
Type: Manuscript

Direct link: http://roa.rutgers.edu/view.php3?roa=1036