yidiny final deletion is not conditioned by morpheme boundaries
TRANSCRIPT
Yidiny final deletion is not conditioned by morpheme boundaries
Erich Round (U.Queensland)Claire Bowern (Yale)Barry Alpher
Yidiny
• Yidinyic subgroup of the Paman
• Rainforest region southwest of Cairns
• Dixon (1977a) A Grammar of Yidiny
• Dixon (1977b) Some phonological rules in Yidiny, Linguistic Inquiry
Deletion has received relatively less attention
• Phonological processes:
• Stress
• Vowel length
• Word final deletion
• conditioning by morpheme boundaries
Hayes 1982, 1985; Kager 1993; Crowhurst & Hewitt 1995; Hall 2001; Pruitt 2011; Kaviloda & Lunden 2014
Structure
• Yidiny
• Deletion conditioned by
• Feet
• Final consonants
• Ghosts
• Morpheme boundaries
Word final consonants are restricted
• Words can end with
• any long or short vowel — a i u a: i: u:
• certain consonants — l r ɽ y m n ɲ ŋ
• clusters
• Constraint: C-Condition (CCOND)
• enforces restrictions on word final consonants & clusters
Focus is on [root + short suffixes]
• Disyllabic suffixes → independent p-words (deletion doesn’t apply)
• Focus → p-words comprised of [root ±short suffixes]
Ghosts delete more readily than full segments
• Idiosyncratic division → segments that undergo deletion / don’t
• e.g. Afar, French, Lardil, Polish, Tiwi, Yokuts, Zulu
• Segments that delete easily → “ghosts” (vs “full” segments)
• French → ghost /t/ in /pətit/ ‘small‘ — full /t/ in /ɔnɛt/ ‘honest’
• Lardil → ghost /i/ in /malari/ ‘bream‘ — full /i/ in /ɳæcari/ ‘hit-NEG’
(Tranel 1981, Archangeli 1988, Zoll 1996, Round 2011, Sibanda 2011)
Ghosts delete more readily than full segments
• Ghost–full contrast
• like between any pair of phonemes
• only that it reveals itself in terms of deletions
• Many specific proposals (e.g. Dixon’s ‘morphophonemes’)
• I’ll take as given:
• such a distinction exists
• ghosts delete more readily than full segments
Final deletion only affects certain words
• Final vowels in some words
• Also word final vowel deletion
[1] /ɡindanu-‐ŋɡu/ → ɡindanuŋɡumoon-‐ERG/ɡubuma-‐ɡu/ → ɡubumaɡublack pine-‐PURP
[2] /ɡindanu/ → ɡinda:nmoon[ABS]/ɡubuma/ → ɡubu:mblack pine[ABS]
Vowels delete rather than go unfooted
• Feet → binary; built from left edge
• Condition A → vowels prefer to delete than go unfooted
[1] /ɡindanu-‐ŋɡu/ → (ɡin da) (nuŋ ɡu)
/ɡubuma-‐ɡu/ → (ɡu bu) (ma ɡu)
[2] /ɡindanu/ → (ɡin da:n) not (ɡin da) nu
/ɡubuma/ → (ɡu bu:m) not (ɡu bu) ma
PARSE ⨠ MAX
• PARSE — syllables must be in a foot
• MAX — don’t delete
• PARSE is the ‘more powerful force’, it’s higher ranked
TABLEAU (1) ON HANDOUT
Delete must leave permissible C’s word finally• Deletion → must leave permissible word final consonants
• CCOND >> PARSE — well-behaved final C’s is an ‘even stronger force’
[3] /waguᶁa/ → (wa ɡu) ᶁa not (wa ɡu:ᶁ)
man[ABS]
/ɡudaɡa/ → (ɡu da) ɡa not (ɡu da:ɡ)
dog[ABS]
TABLEAU (2) ON HANDOUT
Condition C — ghost OR final C–CV
• One of two things:
• EITHER final V must be a ghost
• OR word must end with V–C(C)V
• Next:
• VCV-final words (irrespective of boundaries)
• CCV-final words (irrespective of boundaries)
morpheme boundary
Ghosts delete; Full segments don’t
• Dixon 1977:58 ‘no phonological, semantic or grammatical factor’
• Rather — ghost vowels [2] versus full vowels [4]
• Preservation of full segments is ‘more powerful’ than parsing into feet
[2] /ɡindanu/ → (ɡin da:n) moon[ABS]
/ɡubuma/ → (ɡu bu:m) black pine[ABS]
[4] /mulari/ → (mu la:) ri iniIated man[ABS]
/ᶁudulu/ → (ᶁu du:) lu brown pigeon[ABS]
TABLEAU (3) ON HANDOUT: MAX-FULL >> PARSE
Dixon’s Condition C is complex
• Dixon’s condition C:
• EITHER final V is a ghost
• OR word ends with V–C(C)V
• Us:
• Ghosts delete
• Morpheme boundaries play no role
Suffixes group into three types
1
locative -‐la
1
accusative -‐ɲa
1
ablative–causal -‐mu
1 comitative -‐yi1genitive -‐ni~nu
1
comitative+imperative
-‐ŋa-‐ø
1
causal+imperative -‐ŋa-‐ø
1
past tense -‐ɲu
2
(nominal) purposive -‐gu
2 set inclusion -‐ba2
fear -‐da2
another -‐bi3 (verbal) purposive -‐na
Dixon: V–CV#Us: ghost Vdelete
remain (Condition B)
remain Us: full VDixon: exception
Roots ending in CCV don’t delete — due to CCOND
• Roots ending in CCV never delete
• Our analysis → C-cond — no word-final clusters
• Dixon → rule deletes (C)V from C(C)V#
• this would affect roots
• morpheme boundary → deletes (C)V from –C(C)V#
• Why did Dixon frame the rule as CV deletion?
Again, suffixes group into three types
1
dative -‐da
1
ablative–causal -‐mu
1comitative -‐ᶁi
1 accusative -‐ɲa1genitive -‐n{i,u}
1
ergative -‐du/-‐bu/-‐ᶁu
1
locative -‐da/-‐ba/-‐ᶁa
2dative -‐nda
2 conj.+purposive {-‐l/-‐ɽ}-‐na2conj.+lest[abs] {-‐n/-‐l/-‐ɽ}-‐ᶁi
3ergative -‐ŋgu
3 causal subordinate
-‐ɲu-‐nda3conj.+past {-‐l/-‐ɽ}-‐ɲu
Dixon: not V–CCV#Us: Ccond
delete
remain
remain
Us: exceptionDixon: V–CCV#
Us: CcondDixon: exception
Our constraint ranking already provides a solution
• Constraint ranking gives an account of suffixes in group 3
• Just one assumption: they contain ghost C’s
• Ghost C’s delete, so syllables are parsed into feet
[5] /wawa-‐l-‐ɲu/ → (wa wa:l) see-‐PAST
/marɡu-‐ŋɡu/ → (mar ɡu:ŋ) grey possum-‐ERG
ghosts
TABLEAU (4) ON HANDOUT
Yidiny’s final deletion is clarified
• Dixon’s analysis of final deletion has remained stubbornly quirky
• By extending a contrast between ghost & full segments, we account for
• all of the data
• no exceptions
• no reference to morpheme boundaries
• pattern in the lexicon regarding ghosts and full segments
• With a simple & unremarkable set of constraints
Yidiny&final&deletion& & Round,&Bowern,&Alpher&|&ALS2014&|&Dec&12&!
Page&1&of&2&!
&Yidiny&final&deletion&is¬&conditioned&by&morpheme&boundaries&
&Erich&Rounda,&Claire&Bowernb,&Barry&Alpher&
a#University&of&Queensland,&[email protected]&b#Yale&University&!!&
(t1)& & & PARSESYLL& MAX&
& /ɡindanu/&‘moon[ABS]’& & &
& ☞& (ɡin&da:n)& & *&
& & (ɡin&da:)&nu& *W& L&
& /ɡubuma/&‘black&pine[ABS]’& & &
& ☞& (ɡu&bu:m)& & *&
& & (ɡu&bu:)&ma& *W& L&
& /ɡindanuZŋɡu/&‘moonZERG’& & &
& ☞& (ɡin&da)&(nu&ŋɡu)& & &
& & (ɡin&da)&nuŋ& *W& **W&
& /ɡubumaZɡu/&‘black&pineZPURP’& & &
& ☞& (ɡu&bu)&(ma&ɡu)& & &
& & (ɡu&bu)&ma& *W& **W&!&&&
(t2)& & & CCOND& PARSESYLL& MAX&
& /waɡuᶁa/&‘man[ABS]’& & & &
& ☞& (wa&ɡu:)&ᶁa& & *& &
& & (wa&ɡu:ᶁ)& *W& L& *W&
& /ɡudaɡa/&‘dog[ABS]’& & & &
& ☞& (ɡu&da:)&ɡa& & *& &
& & (ɡu&da:ɡ)& *W& L& *W&!! !
Yidiny&final&deletion& & Round,&Bowern,&Alpher&|&ALS2014&|&Dec&12&!
Page&2&of&2&!
&&
Ghost&segments&are&underlined:&!
&&
(t3)&!& & CCOND& MAXZ&
FULL1&PARSE&SYLL&
MAX&
/ɡindanu/&‘moon[ABS]’&
☞& (ɡin&da:n)& & & & *&
& (ɡin&da:)&nu& & & *W& L&
/ɡubuma/&‘black&pine[ABS]’&
☞& (ɡu&bu:m)& & & & *&
& (ɡu&bu:)&ma& & & *W& L&
/waɡuᶁa2/&‘man[ABS]’&
☞& (wa&ɡu:)&ᶁa& & & *& &
& (wa&ɡu:ᶁ)& *W& *W& L& *W&
& (wa&ɡu:)& & **W& L& **W&
/ɡudaɡa2/&‘dog[ABS]’&
☞& (ɡu&da:)&ɡa& & & *& &
& (ɡu&da:ɡ)& *W& *W& L& *W&
& (ɡu&da:)& & **W& L& **W&
/mulari/&‘initiated&man[ABS]’&
& (mu&la:r)& & *W& L& *W&☞& (mu&la:)&ri& & & *& &
/ᶁudulu/&‘brown&pigeon[ABS]’&
& (ᶁu&du:l)& & *W& L& *W&☞& (ᶁu&du:)&lu& & & *& &
(t4)&!& & CCOND& MAXZ&
FULL1&PARSE&SYLL&
MAX&
/marɡuZŋɡu/&‘possumZERG’&
& (mar&ɡu:ŋ)&ɡu& & & *W& L&
& (mar&ɡu:ŋɡ)& *W& & & *L&
☞& (mar&ɡu:ŋ)& & & & **&
/marɡuZnda2/&‘possumZDAT’&
☞& (mar&ɡu:n)&da& & & *& &
& (mar&ɡu:nd)& *W& *W& L& *W&
& (mar&ɡu:n)& & **W& L& **W&
/wawaZlZɲu/&‘seeZPAST’&
& (wa&wa:l)&ɲu& & & *W& L&
& (wa&wa:lɲ)& *W& & & *L&
☞& (wa&wa:l)& & & & **&
/wawaZlZᶁi2/&‘seeZLEST’&
☞& (wa&wa:l)&ᶁi& & & *& &
& (wa&wa:lᶁ)& *W& *W& L& *W&
& (wa&wa:l)& & **W& L& **W&
!!
!!!!_____________________!1&Or&any&constraint,&C,&with&an&equivalent&effect,&for&example&DEPZROOT&if&ghosts&are&assumed&to&lack&root&nodes.&See&Round,&Bowern,&Alpher&(in&prep)&for&further&details.&2&Final&vowel&assumed&to&be& full,&by& lexicon&optimisation.&Positing&a&ghost&vowel&would&also&work:& just& subtract&one&violation&of&MAXZFULL&for&the&second&and&third&candidates.!