Your final project for LSCI $119/219$ will be a brief squib $($a short paper$),$ which is due June $14$th at

$11$:$59$ pm$.$ This paper should be at least four pages, and should consist of the following

components:

$1.$ Create or find a dataset that has some interesting property with respect to theory

comparison. Some of the properties we$$ve focused on in class so far are:

a$.$ Opacity $($counterfeeding$,$ counterbleeding$)$

b$.$ Lookahead$/$sour grapes

c$.$ Myopia $($the opposite of lookahead$)$

d$.$ Multi$-$site optionality

e$.$ Conspiracies

$2.$ Perform an analysis of this data in at least two of the frameworks listed below.

a$.$ SPE $($with simultaneous rule application$)$

b$.$ SPE $($with directional iterative rule application$)$

c$.$ Optimality Theory

d$.$ Harmonic Serialism

e$.$ Harmonic Grammar $($covered in Week $9)$

f$.$ Maximum Entropy OT $($covered in Week $10)$

Your analyses should follow the same format as the assignments: that is$,$ clearly define your

rules$/$constraints$,$ choose derivations$/$tableaux that demonstrate necessary

orderings$/$rankings$.$

$3.$ Compare the analyses in the two frameworks. For example:

a$.$ Are both frameworks able to capture the patterns in the data $($e$.$g$.$ for an opaque

pattern, can SPE and OT both predict the correct outcomes$)?$

b$.$ Do you need to make certain assumptions to account for the data under one of the

models $($e$.$g$.$ did you need to use conjoined constraints to make an OT analysis of an

opaque phenomenon work? Some other type of constraint we haven$$t discussed in

class?$)$

c$.$ If both analyses work, can you think of any reasons to prefer one over the other?

In essence, this is like the homework assignments but you$$re producing two analyses of a single

data set and comparing them.

Tips on finding data

There are a few options for where you can find an appropriate data set:

$$ Put a new one together based on a language you or someone you know speaks $($or based on

reference materials$).$ This is the most work, but likely to be the most interesting! If you are

interested in this and don$$t have a good idea of how to approach it$,$ message Connor for

help. This is a great option if you$$re interested in grad school $($or you are currently a grad

student$).$

$$ Find an existing dataset that we haven$$t talked about in class by looking at papers,

textbooks, etc. I$$ve uploaded Bruce Hayes$$ phonology textbook, which has some examples

of several relevant problem sets if you search for terms like $$counterfeeding$$ etc.

$$ Use a dataset that we have talked about in class. The key thing here is that you cannot

compare models we have already explicitly compared in class: so e$.$g$.,$ we compared

SPE and OT analyses of the Lena Asturian data set in class, so this would not be appropriate

for your squib.

A few concrete suggestions:

$$ Take $($a subset of$)$ the Uyghur data from HW$1,$ analyze it in OT $($or some other framework$)$

and compare against your rule$-$based analysis. If you didn$$t do well in HW$1,$ this could be an

opportunity to revise your analysis.

$$ The Palauan data from Lecture $4,$ p$.8($don$$t worry about accounting for stress position$).$

$$ The Polish data from Lecture $4,$ pp$.11-12$

$$ The Southern Kikuyu data from Lecture $10,$ pp$.6-7($we already compared this in SPE and OT$,$

but HS could be interesting$)$

$$ Tianjin tone sandhi from Lecture $10,$ pp$.12-13$

Collaboration and citation

As usual, it$$s fine to collaborate so long as you $($a$)$ write up your individual paper; and $($b$)$ list your

collaborators at the top.

If you rely on existing work for either datasets or details of your analysis, you must cite your

sources. It$$s fine to use elements of a published analysis in your write$-$up$,$ but you need to be clear

when you$$re doing so vs$.$ when you$$re proposing your own analysis