AI summary by TwinMind
Lecture 8: Phonology, Part 1
MIT OpenCourseWare
1h 10m 39s
English Plurals Review
- Descriptive facts: English plurals typically spelled with 's' but pronounced three ways
- /z/ after voiced sounds (dogs)
- /s/ after voiceless sounds (cats, giraffes)
- /əz/ after stridents (bridges, brushes, messes)
- Stridents: fricatives with high frequency noise (/s/, /ʃ/) and affricates ending in those fricatives
- Two theoretical approaches:
- Suffix underlyingly /s/, voices to /z/ after voiced sounds
- Suffix underlyingly /z/, devoices to /s/ after voiceless sounds
- Evidence supports /z/ underlying form:
- English allows words ending voiced+/s/ (ribs, tense)
- English doesn't allow words ending voiceless+/z/ (no words like *cats)
- General English phonotactic constraints:
- No voiced obstruent after voiceless sound at word end
- No two stridents in sequence at word end
- Phonological rules as repairs:
- Obstruents devoice after voiceless sounds
- Schwa insertion between stridents (∅ → /ə/ between stridents)
- Rule ordering matters: schwa insertion before devoicing prevents *brushəs
- Minimal repair principle: one operation fixes multiple violations
Phonotactics vs. Repairs Distinction
- Phonotactics: conditions on how sounds can combine
- Repairs: specific operations to fix phonotactic violations
- Both components needed for complete analysis
- Broad vs. narrow transcription terminology (not "loose/tight")
- Broad: predictable details omitted
- Narrow: all phonetic details represented
Cross-linguistic Application
- Rules apply to borrowed words with non-English sounds
- Bach /bax/ → Bachs /baxs/ (devoicing after voiceless)
- Rouge /ruʒ/ → rouges /ruʒəz/ (schwa insertion after strident)
- Shows algorithmic nature of plural formation
Yokuts (Northern California) Phonological Conspiracy
- General constraint: no three consonants in sequence
- Multiple morphological contexts show same avoidance pattern:
Gerund Formation (-taw suffix)
- Regular: moot-taw, swear-taw, eat-taw (two consonants acceptable)
- With vowel insertion: ilk → ilik-taw, log → logo-taw (three consonants avoided)
- Rule: ∅ → /e/ between first and second consonants when three consonants would result
Desiderative Formation (-hatin suffix)
- Regular: know-hatin, sink-hatin
- H-deletion: speak-atin, lift-atin (after verbs ending in two consonants)
- Rule: /h/ deletion when preceded by two consonants
Passive Formation (-hnel suffix)
- Regular: tie-hnel, hit-hnel
- H-deletion: help-nel, hold-nel (after verbs ending in consonants)
- Rule: /h/ deletion when surrounded by consonants
Analysis Options
- Individual morpheme-specific rules vs. general conspiracy
- Conspiracy analysis: all rules serve same function (avoid three-consonant sequences)
- Question raised: is H-deletion specific to suffix-initial H or general?
Japanese Pitch Accent System
- Tokyo dialect accent realization rules:
- Unaccented words: L-H-H pattern (fish = sakanawa)
- Accented words: H on accented syllable, L thereafter
- Examples: pillow (makura), heart (kokoro), head (atama)
- Noun vs. verb accent differences:
- Nouns: accent can be on any syllable or absent
- Verbs: only accented/unaccented distinction, no position choice
- Minimal pairs: rain vs. candy (both ame, different accent patterns)
Japanese Accent Conspiracy
- Fundamental constraint: maximum one accent per word
- Multiple repair strategies depending on context:
Postposition Accent Deletion
- "From" (kara): unaccented, doesn't affect noun accent
- "To" (made): accented, loses accent after accented nouns
- Evidence: sakanamade (to fish) shows made accent, makurakara shows no made accent
Prefix Accent Dominance
- "Even" (gurai): accented prefix
- Deletes accent on following noun: makura → gurai-makura (unaccented)
- Repair: noun accent deletion before accented prefix
Compound Accent Placement
- Single accent required in compounds
- Neither component accent necessarily preserved
- Accent placement rules:
- Near morpheme boundary
- Avoid final syllable
- Prefer second component when possible
- Examples: Chiba-ken → Chiba-ken (accent on second component first syllable)
- Complex cases: Kagawa-ken → Kagawa-ken (accent on first component final when second too short)
Priority Ranking Analysis
- Japanese accent system as ranked constraints:
- One accent per word (highest priority)
- Keep accent on "even" (gurai)
- Keep accent on nouns
- Keep accent on "to" (made)
- Optimality Theory preview: languages choose optimal outputs satisfying highest-ranked constraints
- Compound accent priorities:
- One accent only
- Accent near boundary
- Avoid final accent
- Prefer second component accent
