Please post a comment on something from the readings from week 5. If it’s on Bermudez-Otero, please post by the end of the day Sunday, and if it’s on McCarthy, please post by the end of the day Wednesday.
Please post a comment on something from the readings from week 5. If it’s on Bermudez-Otero, please post by the end of the day Sunday, and if it’s on McCarthy, please post by the end of the day Wednesday.
RE: Canadian Raising.
I remember that we tried to make the opaque interaction work in HS by hypothesizing that flapping happens in two steps. I don’t remember what the two steps proposed were, but I just wanted to mention something that I think might suggest that flapping is indeed a two-stage process at some level of description.
I’ve noticed that older recordings of American English speakers speaking in a formal register sometimes have almost voiceless (often even aspirated), but still significantly shortened [t]s in the flapping context. It sounds as if they try to undo flapping, but don’t quite manage to. An example of this is (the composer) John Cage’s speech in this extended fragment:
http://www.youtube.com/watch?v=Wrah3bEEiaU
The oiceless flap is noticeable in, for instance, the word matting at 0:38. There are many examples in of this in the rest of the sound file (the technique of composition of this piece of music required that each story he told be stretched or compressed to fit 1 minute).
This pronunciation variant may suggest that flapping occurs in two stages: shortening/weakening of the closure gesture, then voicing. I don’t think that having these two steps in an HS derivation will solve the problem (at least, not if voicing is a [-voice] to [+voice] change), but at least, I think that the dissociation of voicing from shortening may provide an interesting insight into the raising/flapping interaction.
I apologize – I gave the wrong youtube link. This is the correct one:
http://www.youtube.com/watch?v=AJMekwS6b9U
Can we discuss a concrete example of the application of “Archiphonemic Prudence” and “quarantining”? I’m wondering how computational/psychologically viable we can consider a learning algorithm that lets the learner “quarantine” the choice between two competing hypotheses and continue on with the learning process. In particular, what prompts the learner to revisit the “quarantined” decision? That is, if the decision is “quarantined” how does the learner know when she/he has gathered sufficient information to revist the quarantined decision and choose between the two options? In addition, is there any way we can test (experimentally or through a corpus study) the existence of the “quarantining” procedure? For example, what production errors would we expect to see a child (or computational model) make if he or she is forced produce an output whose realization depends crucially on the selection between two competing hypotheses that are currently “quarantined”?
In footnote 1, BO mentions that “In Stratal OT, only the highest grammatical level is subject to Richness of the Base. The input to anon-initial stratum n will possess systematic properties enforced by the constraint hierarchy”. I don’t understand what does he mean by that. He also mentions “identity map” (Yip 1996) . Could we please have a couple-of-minutes background on these two when we’ll discuss the article in class?
B-O:
1) p. 3:
B-O doesn’t like that IO-Faith is bidirectional and Sympathy theory is only in a single direction. Does this actually make sense? Is the “other direction” of IO-Faith as used in input learning really distinct, or is it just the same thing, playing a distinct role because the algorithm using it is different?
2) p. 4:
B-O finds fault with Sympathy because of a stipulation it makes. He also dislikes that it does not predict “non-paradigmatic non-vacuous Duke-of-York gambits”. The former was created entirely to cause the latter, so I’m failing to see a real objection here…
3)
Throughout B-O takes the non-existence of good learning procedures for Sympathy theory as evidence that they *cannot* exist. Is there any reason to think this is true? How strong was the effort?
—
I’m generally a little confused about whether the learning procedure outlined can really perform as well as advertised without stipulations about the levels of processes. This is probably just due to not working it through enough yet…