Phonological trends in the lexicon — Theory Michael Becker University of Massachusetts Amherst michael.becker@phonologist.org EVELIN 2012 MIT / UNICAMP Campinas, Brazil 1 / 18
Overview • Overview • Basic mechanics of Optimality Theory (OT, Prince & Smolensky OT Basics 1993/2004) Rankings • Ranking arguments Trends with constraint • A grammatical approach to lexical trends cloning Grammatical vs. ◦ Learning lexical trends respresentational approaches ◦ Projecting trends onto novel words • Representational vs. grammatical approaches ◦ Surface-true UR’s vs. abstract UR’s Representational vs. grammatical � = OT vs. rules ◦ 2 / 18
• Overview OT Basics • Predictable distibutions • Contrast • Sources of constraints Rankings Trends with constraint cloning OT Basics Grammatical vs. respresentational approaches 3 / 18
Predictable distibutions • Overview • Complementary distribution OT Basics • OT (Prince & Smolensky 1993/2004): markedness constraints • Predictable distibutions • Factorial typology • Contrast • Sources of constraints Rankings Trends with constraint cloning Grammatical vs. respresentational approaches 4 / 18
Predictable distibutions • Overview • Complementary distribution OT Basics • Predictable ◦ [sa, S i, su], *[ S a, si, S u] distibutions • Contrast (e.g. native vocabulary of Japanese) • Sources of constraints ◦ Two generalizations: Rankings only [ S ] appears before [i], only [s] appears elsewhere Trends with constraint ◦ Rule-based analysis: cloning Grammatical vs. One phoneme: /s/, no S in underlying representations respresentational • approaches One rule: [s] → [ − ant] / i • rule application: • Underlying representation (UR) / susi / Rule application S Surface representation (SR, PR) [ su S i ] No /su S i/ in this language • • OT (Prince & Smolensky 1993/2004): markedness constraints • Factorial typology 4 / 18
Predictable distibutions • Overview • Complementary distribution OT Basics • OT (Prince & Smolensky 1993/2004): markedness constraints • Predictable distibutions • Contrast • Sources of /si/ /sa/ *si * S *si * S constraints Rankings a. ☞ sa a. si *! Trends with constraint cloning b. ☞ Si Sa * b. *! Grammatical vs. respresentational approaches /Si/ /Sa/ *si * S *si * S a. ☞ sa a. si *! b. ☞ Si * b. Sa *! No phonemes = all sounds are allowed in URs. • Factorial typology 4 / 18
Predictable distibutions • Overview • Complementary distribution OT Basics • OT (Prince & Smolensky 1993/2004): markedness constraints • Predictable distibutions • Factorial typology • Contrast • Sources of constraints /si/ /sa/ * S *si * S *si Rankings Trends with constraint a. ☞ si a. ☞ sa cloning * Grammatical vs. respresentational Si Sa b. *! b. *! approaches /Si/ /Sa/ * S *si * S *si a. ☞ si a. ☞ sa * b. Si *! b. Sa *! What kind of language did we get? 4 / 18
Contrast • Overview • English contrasts [su] ‘sue’ and [ S u] ‘shoe’ OT Basics • English also contrasts [si] ‘sea’ and [ S i] ‘she’ • Predictable distibutions • English does not contrast [ S r] ‘shrew’, ‘shriek’, etc. and *[sr] • Contrast • Sources of constraints Rankings Trends with constraint cloning Grammatical vs. respresentational approaches 5 / 18
Contrast • Overview • English contrasts [su] ‘sue’ and [ S u] ‘shoe’ OT Basics • Predictable We need to be faithful to [ ± ant]. ◦ distibutions • Contrast ◦ I DENT (ant): Assign one violation mark to every output • Sources of constraints segment that has an input correspondent, and where the Rankings [ ± ant] feature of the input doesn’t match the [ ± ant] feature of Trends with constraint the output cloning Grammatical vs. respresentational /su/ *si I DENT (ant) * S approaches a. ☞ su Su b. *! * /Su/ *si I DENT (ant) * S a. su *! b. ☞ Su * 5 / 18 • English also contrasts [si] ‘sea’ and [ S i] ‘she’
Contrast • Overview • English contrasts [su] ‘sue’ and [ S u] ‘shoe’ OT Basics • English also contrasts [si] ‘sea’ and [ S i] ‘she’ • Predictable distibutions • Contrast I DENT (ant) ≫ *si ◦ • Sources of constraints /si/ I DENT (ant) *si * S Rankings Trends with constraint a. ☞ si cloning * Grammatical vs. respresentational b. Si *! * approaches /Si/ I DENT (ant) *si * S a. si *! * b. ☞ Si * • English does not contrast [ S r] ‘shrew’, ‘shriek’, etc. and *[sr] 5 / 18
Contrast • Overview • English contrasts [su] ‘sue’ and [ S u] ‘shoe’ OT Basics • English also contrasts [si] ‘sea’ and [ S i] ‘she’ • Predictable distibutions • English does not contrast [ S r] ‘shrew’, ‘shriek’, etc. and *[sr] • Contrast • Sources of constraints /sri/, Rankings * S *sr I DENT (ant) *si /Sri/ Trends with constraint cloning a. ☞ Sri (*) * Grammatical vs. respresentational approaches b. sri *! (*) The [s]—[ S ] contrast is neutralized before [r] Compare with Arabic [ S r 1 b] ‘drink’ vs. [sr 1 q] ‘steal’ 5 / 18
Sources of constraints • Overview • Prince & Smolensky (1993/2004): all constraints are innate OT Basics • Hayes & Wilson (2008): induction of markedness constraints = all • Predictable distibutions markedness constraints are language-specific • Contrast • • Sources of As of now, nobody tried to induce faithfulness constraints. constraints Rankings Trends with constraint cloning Grammatical vs. respresentational approaches 6 / 18
Sources of constraints • Overview • Prince & Smolensky (1993/2004): all constraints are innate OT Basics • Predictable ◦ Not 100% true — Prince & Smolensky (1993/2004) also had distibutions • Contrast constraints that align specific morphemes with word edges, • Sources of constraints e.g. Tagalog’s A LIGN (um-) Rankings ◦ Prince & Smolensky (1993/2004) weren’t worried about a Trends with constraint huge Universal Grammar cloning Grammatical vs. respresentational • Hayes & Wilson (2008): induction of markedness constraints = all approaches markedness constraints are language-specific • As of now, nobody tried to induce faithfulness constraints. 6 / 18
Sources of constraints • Overview • Prince & Smolensky (1993/2004): all constraints are innate OT Basics • Hayes & Wilson (2008): induction of markedness constraints = all • Predictable distibutions markedness constraints are language-specific • Contrast • Sources of constraints ◦ Markedness constraints are generated by the learner, and Rankings kept if useful in making the data probable (what does that Trends with constraint mean?) cloning Grammatical vs. ◦ Universal Grammar is small, but important: features, respresentational constraint templates, autosegmental tiers approaches ◦ No discussion of faithfulness ◦ (How does this prevent learning of unnatural trends in Turkish, or the weaker learning of unnatural trends in Hungarian?) • As of now, nobody tried to induce faithfulness constraints. 6 / 18
• Overview OT Basics Rankings • Ranking arguments • More ranking arguments • Conflicting arguments Trends with constraint cloning Rankings Grammatical vs. respresentational approaches 7 / 18
Ranking arguments • Overview • Is a ranking needed? OT Basics • Comparative tableau 1 (Prince 2002) Rankings • Comparative tableau 2 • Ranking arguments • • More ranking Comparative tableau 3 arguments • Summary: • Conflicting arguments Trends with constraint cloning Grammatical vs. respresentational approaches 8 / 18
Ranking arguments • Overview • Is a ranking needed? OT Basics input input C1 C2 C1 C2 Rankings • Ranking arguments a. ☞ output1 a. ☞ output1 * • More ranking arguments • Conflicting arguments b. output2 * b. output2 * * Trends with constraint cloning Grammatical vs. input C1 C2 respresentational approaches a. ☞ output1 * *** b. output2 ** * input C1 C2 C3 a. ☞ output1 * b. output2 * * • Comparative tableau 1 (Prince 2002) 8 / 18 • Comparative tableau 2 Comparative tableau 3
Ranking arguments • Overview • Is a ranking needed? OT Basics • Comparative tableau 1 (Prince 2002) Rankings • Ranking arguments • More ranking Standard tableau: arguments • Conflicting arguments input C1 C2 C3 Trends with constraint cloning a. ☞ output1 * Grammatical vs. respresentational b. output2 * * approaches Comparative tableau: input C1 C2 C3 output1 ≻ output2 W W L Rule: every L must be dominated by some W • Comparative tableau 2 • Comparative tableau 3 8 / 18 • Summary:
Ranking arguments • Overview • Is a ranking needed? OT Basics • Comparative tableau 1 (Prince 2002) Rankings • Comparative tableau 2 • Ranking arguments • More ranking arguments Standard tableau: • Conflicting arguments Trends with constraint input C1 C2 C3 cloning Grammatical vs. a. ☞ output1 * * respresentational approaches b. output2 * * Comparative tableau: input C1 C2 C3 output1 ≻ output2 W L One W is more informative than 2 W’s • Comparative tableau 3 8 / 18 • Summary:
Recommend
More recommend
