Questions: Constraint-Based Phonology: Formal Foundations

Question 1

In a language, the markedness constraint *Complex (no complex onsets) is ranked above the faithfulness constraint Dep-IO (don't insert segments). The underlying form /bla/ is the input. What does EVAL select as the optimal output?

• A. /bla/ — because the input form should always be preserved when possible
• B. /əbla/ — because inserting a vowel before the cluster avoids the complex onset at no deletion cost
• C. /la/ — because deleting the /b/ avoids the complex onset
• D. /bəla/ — because inserting a vowel between consonants satisfies *Complex while violating only the lower-ranked Dep-IO

Answer: (D)

When *Complex outranks Dep-IO, the grammar tolerates faithfulness violations (insertions) to avoid complex onsets. The output /bəla/ violates only Dep-IO (one insertion), which is the lower-ranked constraint. The faithful output /bla/ violates the higher-ranked *Complex and is therefore suboptimal. This demonstrates the core OT insight: the winning form is not the one that satisfies all constraints, but the one that only violates lower-ranked ones. Different languages with different rankings produce different optimal outputs from the same input.

Question 2

A language ranks *NoCoda (no coda consonants) above Max-IO (don't delete input segments). Given the underlying form /kant/ as input, what output does OT predict?

• A. /kant/ — because faithfulness constraints should always be satisfied in the optimal output
• B. /kan/ — because deleting the coda /t/ satisfies the higher-ranked constraint at the cost of the lower-ranked one
• C. /kanto/ — because epenthesis satisfies both constraints simultaneously
• D. /ant/ — because onset consonants are less marked than coda consonants

Answer: (B)

/kan/ is optimal because *NoCoda outranks Max-IO: the language tolerates deletion (violating Max-IO) to avoid a coda consonant (satisfying *NoCoda). The faithful output /kant/ would violate the higher-ranked constraint. This is the OT architecture in action: ranked constraints compete, the higher-ranked one wins, and the output is the candidate that minimizes violations starting from the top. A language with the opposite ranking (Max-IO >> *NoCoda) would preserve /kant/ intact.

Question 3

In Optimality Theory, every candidate output violates at least one constraint — the 'optimal' output is optimal not because it's perfect, but because it violates only lower-ranked constraints.

• T. True
• F. False

Answer: True

True. This is the most important conceptual shift from rule-based phonology. There is no candidate in OT that satisfies every constraint, because markedness constraints and faithfulness constraints are in permanent tension: satisfying markedness (simplify the output) necessarily violates faithfulness (preserve the input), and vice versa. The optimal candidate is the one that loses to no other candidate in the constraint hierarchy — not the one that violates nothing. A candidate wins by violating fewer highly-ranked constraints than any competitor, even if it violates lower-ranked constraints freely.

Question 4

In Optimality Theory, the set of constraints (CON) varies across languages — different languages have different constraints, which is what produces surface differences in pronunciation.

• T. True
• F. False

Answer: False

False. This is the opposite of OT's central claim. CON is universal — the same set of constraints is shared across all human languages. What varies between languages is the *ranking* of those constraints. This universality is theoretically significant: it means cross-linguistic variation is constrained — you can only produce a language by ranking the universal constraints differently, not by inventing new constraints. This predicts that language types will cluster around the possible rankings rather than vary arbitrarily.

Question 5

How does Optimality Theory explain why languages differ from each other in phonology, but only in constrained, predictable ways — not arbitrarily?

Model answer: OT explains cross-linguistic variation through ranking permutation over a universal constraint set. All languages have the same constraints (CON); they differ only in how those constraints are ordered relative to one another. The space of possible human language grammars is bounded by the possible rankings of the universal constraints — large but structured. Languages cannot differ arbitrarily because they cannot invent new constraints; they can only reorder existing ones. Typological patterns — the clustering of features across languages — directly reflect constraint interaction properties.

This is OT's main theoretical advantage over rule-based accounts. Rule-based phonology stipulates different rules for each language with no principled reason why language A has certain rules and language B others. OT provides a unified explanation: the same constraints, differently ranked, generate both A and B as predicted outcomes. When a typological generalization holds — 'no language has voiced obstruents only in coda position' — OT explains it as a consequence of constraint interaction, not a coincidence requiring separate stipulation in each language's grammar.