國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
type when restructuring occurs. This is because eat or type are not the termination of the NP.
This thesis will make use of Sense Unit Condition and restructuring to interpret how an IP is formed.
2.3 Optimality Theory 2.3.1 Basic Concepts
Prince and Smolensky (1993, 2004) propose a constraint-based framework, Optimality Theory (henceforth “OT”), to account for processing of human languages.
Unlike Generative Theory which uses rules to derive the surface forms step by step, OT is a device in which output candidates are evaluated in a parallel fashion. (17) presents how OT maps an input to a output. The sources of the inputs are not
restricted to the underlying representations in a language. All possible combinations of phonological elements can be the inputs, which property is called Richness of the base (Prince and Smolensky, 1993/2004). The input is submitted to GEN. GEN generates all possible output candidates, called Freedom of Analysis (Kager, 1999).
Then, these candidates are assessed by EVAL, in which the criterion of harmony is applied in the selection of the actual output. The candidate having the least violations of a set of ranked constraints is chosen as the optimal or most harmonic output. EVAL consists of two types of constraints, markedness and faithfulness constraints, which are responsible for the evaluation. It is assumed that these constraints are universal and violable. Universality means that languages in the world share the same set of constraints. The different rankings among the constraints result in language diversity or variations. The use of the concept of Violability, which was not applied in the theories proposed in the past, permits that constraints are violable, but that the
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
15
violations are required to be minimal.
(17) The processing schema for OT
GEN EVAL
>> >> … Candidate a
Candidate b input Candidate c
Candidate d output Candidate e
The selection process is represented by a tableau, as in (18).
(18) Tableau
Con1, Con2 >> Con3 >> Con4
Input Con1 Con2 Con3 Con4
→ a. Candidate1 *
b. Candidate2 *!
c. Candidate3 *! *
d. Candidate4 *!
(18) is a violation tableau. In (18), the input is put in the top left position and the candidates are displayed at the left side. The constraints are ranked from left to right on the top of the line. The dotted line demonstrates that no domination relationship is found among the constraints. The solid line indicates that the constraints are in conflict and that one constraint outranks the following one. Therefore, the ranking in (18) is Con1, Con2 >> Con3 >> Con4. The notation → indicates the winner. A
C1 C2 Cn
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
violation mark is represented by the asterisk *. The exclamation mark ! indicates that the candidate is ruled out because of a fatal violation. The shaded cell means that a constraint is irrelevant to select this candidate as the winner because this candidate is removed by a higher-ranked constraint.
In addition to the violation tableau, a combination tableau is another format used to display the selection process. It provides two sorts of information: the numbers of the violations and the comparative symbols, W and L. The comparative symbols are restricted to the loser lows, since they function as presenting how a loser is assessed in comparison with the winner with respect to every constraint, as exemplified in (19).
(19) Combination tableau
Input Con1 Con2
→ a. Candidate 1 *
b. Candidate 2 * W L
In (19), Con1 favors the winner, Candidate 1, over the loser, Candidate 2. As a result, W is put into the cell of Candidate 2. L is entered into the cell to the right of W because Con 2 favors the loser, Candidate 2, over the winner, Candidate 1. This is the evidence that Con 1 dominates Con 2. Therefore, a combination tableau like (19) is a good tool to present ranking arguments.
Chapter Four in this thesis will make use of combination tableaux to
present the ranking arguments in dealing with Yinping tone sandhi in terms of OT.
2.3.2 Generalized Alignment
McCarthy and Prince (1993) introduce Generalized Alignment, a constraint family, to analyze the phenomena of the sharing of an edge in morphological
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
17
constituents and prosodic constituents, as described in (20).
(20) Generalized Alignment (McCarthy & Prince, 1993:80) Align (Cat1, Edge1, Cat2, Edge2) = def
∀Cat1∃Cat2 such that Edge1 of Cat1 and Edge2 of Cat2 coincide.
Where
Cat1, Cat2 ∈ PCat ∪ GCat Edge1, Edge2 ∈ {Right, Left}
PCat represents prosodic categories and GCat grammatical categories. From the definition in (20), a designated edge of a morphological or prosodic constituent (Cat1) corresponds to a designated edge of a specific morphological or prosodic constituent (Cat2). For example, Align (PPh, R, XP, R) requires that the right edge of the
phonological phrase be aligned with that of a maximal projection. The Generalized Alignment constraints are related to End-based Theory (Selkirk, 1986), since End- based Theory deals with how the edge of a prosodic constituent corresponds to the edge of a syntactic constituent. Prince and Smolensky (1993) extend Generalized Alignment to all kinds of edges of grammatical and prosodic constituents, and so provide a more general tool to account for edge-sharing phenomena in world languages.
2.3.3 Cophonology Theory
The cophonology approach (Orgun, 1996; Anttila, 1997; Inkelas and Zoll, 2007) deals with language-internal diversity in the light of the re-ranking of a set of
unspecified constraints. In this model, constraints are general and purely phonological and are not indexed for specific contexts. This model is schematized in (21).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
(21) Schema for Cophonology
Master Ranking Con1 >> {Con2, Con3}
Cophonology A Cophonology B Con1 >> Con2 >> Con3 Con1 >> Con3 >> Con2
In (21), the Master Ranking, a superordinate node, consists of a set of unspecified general constraints, that is, Con2 and Con3 within the braces. They are specified in the individual cophonologies A and B due to morphological constructions or register.
That is, Con2 is ranked above Con3 in Cophonology A, but the opposite ranking, where Con3 is higher-ranked than Con2, forms the Cophonology B. Therefore, the language-internal diversity is accounted for by virtue of the re-ranking of a set of unspecified constraints.