Optimality Theory

Optimality theory (OT) is a constraint-based theory of grammar that was introduced in the early 1990s by Alan Prince and Paul Smolensky in their [Page 1284]seminal work Optimality Theory: Constraint Interaction in Generative Grammar. The theory represents a radical departure from the generative rule-based theories that have dominated linguistics since the 1960s, and the subsequent research that has emerged from OT investigations has yielded answers to several long-standing questions, especially in the domain of phonology and acquisition.

One of the nagging questions that had eluded a satisfactory answer is why different, seemingly unrelated languages exhibit many of the same phonological processes. An equally perplexing question has been why children in the course of acquisition exhibit phonological processes that are not evident in the target language. OT addressed questions of this sort by adopting a novel architecture that employs universal constraints, eliminating any appeal to rules. The constraints of the theory are violable, and they can conflict with one another. Conflict is resolved by language-specific rankings (or weighting) of the constraints. The constraints evaluate all potential output candidates in parallel for any given input (underlying) representation and assign violation marks to candidates who violate a constraint. The output candidate who best satisfies the constraint hierarchy (i.e., incurs the least serious violations) is selected as optimal. The constraints are of two fundamental types: markedness and faithfulness. Markedness constraints impose well-formedness restrictions on potential output (surface) structures and are formulated exclusively in terms of output properties. Notions of naturalness are incorporated directly in the markedness constraints and are based on typological properties of language. Some examples of well-supported markedness constraints along with their definitions are given in (1).

• (1)Markedness constraints

  *Coda: Coda consonants are banned.

  *CC: Consonant clusters are banned.

  *Fricatives: Fricatives are banned.

Faithfulness constraints, on the other hand, demand identity between an input (underlying) representation and a corresponding output representation. Such constraints are the antithesis of rules, militating against changes between underlying and phonetic representations. There are, of course, many different ways that input and output representations might differ from one another. The constraints in (2) represent just two examples of the many faithfulness constraints that can limit the difference between the two levels of representation.

• (2)Faithfulness constraints

  MAX: Every input segment must have a corresponding output segment (no deletion).

  IDENT: Corresponding input and output segments must be identical (no change in feature specification).

The notions of conflict and violability can be illustrated by considering the role of the two conflicting constraints *Coda and MAX for those many children who omit word-final consonants in the early stages of acquiring English. While final consonant deletion is not a process that any of these children would have observed in English, the process follows naturally from *Coda being ranked over MAX. Focusing on two potential output candidates for an English word ending in a consonant (i.e., one with and the other without a final consonant), the target-appropriate (faithful) candidate incurs a fatal violation of the top-ranked constraint *Coda and is eliminated from the competition. On the other hand, the child’s erred phonetic output is selected as optimal because it complies with *Coda and only violates lower-ranked MAX, making its violation less serious. For these children, it is more important to comply with the markedness constraint *Coda than it is to be faithful to the input representation. This underscores another important assumption of OT, and that is, that markedness constraints by default dominate faithfulness constraints in the early stages of acquisition. Acquisition, thus, proceeds by the child’s phonetic representations becoming more and more faithful (target-like) over time. This means that markedness constraints will be gradually demoted below faithfulness constraints based on the positive (observable) evidence in the language to which children are exposed. MAX will eventually come to outrank *Coda, resulting in target-appropriate productions and the loss of final consonant deletion. [Page 1285]Other common error patterns in early acquisition such as stopping of fricatives and cluster reduction receive a similar account, following from *Fricatives and *CC being ranked over IDENT and MAX. The loss of these error patterns follows from the reverse ranking of these constraints.

...