24.961
•
Serial Derivations in OT
In Classical OT the input is mapped to the output in one step with no intermediate stages
(fully parallel)
•
This factor allows constraints to perform global comparisons referring to information
that only becomes available later in the derivation in the traditional ordered rule
framework: top down effects, remote structures
•
but there are situations where valid generalizations require reference to intermediate
stages between the input and the output
•
two alternative approaches are explored in the current generative literature to capture
this phenomenon: Stratal Optimality Theory (Kiparsky, Rubach, Anttila, Bermudez-
Otero, Jones, …) and Harmonic Serialism (McCarthy, Pater, Elfner, Pruitt, …)
Harmonic Serialism
•
input mapped to output via a Gen-Eval loop; Gen is restricted to minimal modifications
of the input: add or delete a single segment, change a single feature coefficient. What
counts as a minimal change is a research question (e.g. reparsing of syllable structure
comes for free)
•
the candidates are submitted to the ranked constraint set and an optimal candidate
emerges through normal evaluation, which is then sent back to Gen for another round of
candidate creation by minimal changes
•
when the Gen-Eval loop leads to no change, the derivation is said to “converge” on the
output
•
derivations thus display monotonic improvement, given the M > F constraint ranking
Example 1 Macushi Carib (McCarthy 2010)
•
stress is parsed by left-to-right iambs and then the weak vowel of the foot is deleted
/wanamari/
->
/u-wanamari-ri/ ->
•
•
wnámrí
wánmárri
‘mirror’
‘my mirror’
insertion and deletion of a foot is assumed to be an elementary operation
constraints and ranking:
Parse-syllables: penalize a syllable that is not parsed by a foot
*Weak vowel in foot: penalize an unstressed vowel in the foot
Max-V
ranking: Parse-syll » *Weak-Vowel » Max-V
1
4>!'3>!03=))&$#!/>%%'4%$+!
'+!ৄ!
MM!
M!
!
:*'3J<1'6.!
!
"
:;<
B&$2!8!)=!H*'3'1'6.H!ৄ!I:*3J<:16K<L!
4+!
MMMMS!
!
!
*'3'1'6.!
!
*'3'1'6.! ?@ABCDBEFF@GFC MND?F@OC*$'P Q@R
7+!
MMMS!
!
M!
*'31'6.!
#$%$&$!'!()*$%+!,-./!'//012&.)3!'4)0&!56'#0'%3$//!1$'3/!&-'&!&-$!7'3#.#'&$!/$&!'&!/&$2!
'+!ৄ! :*'3J<1'6.!
MM!
M!
!
89!/-)*3!.3!&'4%$'0!:;<9!.37%0#$/!7'3#.#'&$/!*.&-!=))&!2'6/.35!)6!/>37)2$!40&!3)&!4)&-+!
,'4%$'0!:;<!/-)*/!-)*!/&6$//!'//.531$3&!./!.3&6.3/.7'%%>!)6#$6$#!4$=)6$!/>37)2$+!
4+!
MMMMS!
! @R!2).3&%$//%>9!/.37$!MND?
! DBEFF@GFC9!*-.7-!./!(.)%'&$#!
*'3'1'6.!
@ABC
,-$!.12$6'&.($!&)!2'6/$!./!26)(.#$#!4>!&-$!7)3/&6'.3&!?
B>37)2$!26.)6!&)!=))&!2'6/.359!'/!.3!:;<79!(.)%'&$/!Q
F@OC*$'P!./!
03=))&$#!/>%%'4%$+!
4>!'3>!
('70)0/%>!
4>! 03=))&$#!MMMS!
()*$%/+! T3#$$#9!! 4$7'0/$!M!=))&! 2'6/.35! .3&6)#07$/! '!
7+!/'&./=.$#!
*'31'6.!
:;<
B&$2!8!)=!H*'3'1'6.H!ৄ!I:*3J<:16K<L!
(.)%'&.)3!)=!MND?
F@OC*$'P9!'/!.3!:;<'9!&-$!=))&D2'6/.35!.12$6'&.($!?@ABCDBEFF@GFC!10/&!4$!
,'4%$'0!:;<!/-)*/!-)*!/&6$//!'//.531$3&!./!.3&6.3/.7'%%>!)6#$6$#!4$=)6$!/>37)2$+!
U
@R
!
*'3'1'6.! ?@ABCDBEFF@GFC MND?
6'3P$#!-.5-$6!)6!$%/$!=))&%$//!:;<4!*)0%#!*.3+
! F@OC*$'P@RQ
!2).3&%$//%>9!/.37$!MND?
F@OC*$'P!./!
B>37)2$!26.)6!&)!=))&!2'6/.359!'/!.3!:;<79!(.)%'&$/!Q
@ABCDBEFF@GFC!./!/&.%%!&-$!26.1$!#.6$7&.($9!/)!/>37)2$!./!
@&!/&$2!V9!/'&./='7&.)3!)=!?
('70)0/%>!
/'&./=.$#!
4>!
03=))&$#!
()*$%/+!
T3#$$#9!
4$7'0/$!
=))&!
2'6/.35!
.3&6)#07$/!
'!
'+!ৄ! :*'3J<1'6.!
MM!
M!
!
)37$!'5'.3!2)/&2)3$#W!
(.)%'&.)3!)=!MND?
F@OC*$'P9!'/!.3!:;<'9!&-$!=))&D2'6/.35!.12$6'&.($!?@ABCDBEFF@GFC!10/&!4$!
4+!
MMMMS!
!
!
*'3'1'6.!
:"<6'3P$#!-.5-$6!)6!$%/$!=))&%$//!:;<4!*)0%#!*.3+
B&$2!V!)=!H*'3'1'6.H!ৄ!I:*3J<:16K<L!
U
!
! *$'P QM!@R
! 7+!@&!/&$2!V9!/'&./='7&.)3!)=!?
*'3'1'6.!
?@ABCMMMS!
DBEFF@GFC
MND?F@OC
*'31'6.!
@ABCDBEFF@GFC
!./!/&.%%!&-$!26.1$!#.6$7&.($9!/)!/>37)2$!./!
,'4%$'0!:;<!/-)*/!-)*!/&6$//!'//.531$3&!./!.3&6.3/.7'%%>!
)6#$6$#!4$=)6$!/>37)2$+!
)37$!'5'.3!2
)/&2)3$#W!
'+!ৄ!
!
MM!
!
:*'3J<:1'6K<!
@R!2).3&%$//%>9!/.37$!MND?F@OC*$'P!./!
B>37)2$!26.)6!&)!=))&!2'6/.359!'/!.3!:;<79!(.)%'&$/!Q
:"<
B&$2!V!)=!H*'3'1'6.H!ৄ!I:*3J<:16K<L!
4+!
MMS!
M!
!
:*'3J<1'6.!
('70)0/%>!
/'&./=.$#!
4>!
03=))&$#!
()*$%/+!
T3#$$#9!
4$7'0/$!
=))&! 2'6/.35! .3&6)#07$/! '!
!
*'3'1'6.!
?@ABCDBEFF@GFC MND?F@OC*$'P Q@R
7+!
MMS!
!
M!
:*3J<1'6.!
F@OC*$'P9!'/!.3!:;<'9!&-$!=))&D2'6/.35!.12$6'&.($!?
@ABCDBEFF@GFC!10/&!4$!
(.)%'&.)3!)=!MND?
'+!ৄ! :*'3J<:1'6K<!
!
MM!
!
U
!
6'3P$#!-.5-$6!)6!$%/$!=))&%$//!:;<4!*)0%#!*.3+
T3! '! *)6#! )=! &-./! /.X$9! =0%%! /'&./='7&.)3! )=! ?@ABCDBEFF@GFC! -'/! 4$$3! '7-.$($#! 4>!
4+!
MMS!DBEFF@GFC!./!/&.%%!&-$!26.1$!#.6$7&.($9!/)!/>37)2$!./!
M!
!
:*'3J<1'6.!
@ABC
@&!/&$2!V9!/'&./='7&.)3!)=!?
&-$!$3#!)=!/&$2!V9!/)!.&!./!=.3'%%>!2)//.4%$!&)!'&&$3#!&)!&-$!6$Y0.6$1$3&/!)=!MND?
F@OC*$'P9!
)37$!'5'.3!2)/&2)3$#W!
*-.7-!./!&-$!3$Z&!1'6P$#3$//!7)3/&6'.3&!.3!&-$!6'3P.35+![3$!)=!&-$!03/&6$//$#9!=))&$#!
MMS!
!
M!
7+!
:*3J<1'6.!
:"< #$%$
B&$2!V!)=!H*'3'1'6.H!ৄ!I:*3J<:16K<L!
'&!
!U9!*.&-!&-$!
#$%$&.35!'&!
\+!
()*$%/!
&$/!
/&$2
)&-$6!
/&$2!
T3! '! *)6#! )=! &-./! /.X$9! =0%%! /'&./='7&.)3! )=! ?@ABCDBEFF@GFC! -'/! 4$$3! '7-.$($#! 4>!
!
*'3'1'6.!
?@ABCDBEFF@GFC MND?F@OC*$'P Q@R
:]<&-$!$3#!)=!/&$2!V9!/)!.&!./!=.3'%%>!2)//.4%$!&)!'&&$3#!&)!&-$!6$Y0.6$1$3&/!)=!MND?
B&$2!U!)=!H*'3'1'6.H!ৄ!I:*3J<:16K<L!
F@OC*$'P9!
!
:*'3J<:1'6K<!
?
@ABC
DB
EFF@GFC
MND?
F@OC
Q
@R
'+!ৄ! :*'3J<:1'6K<!
!
MM!*$'P
!
*-.7-!./!&-$!3$Z&!1'6P$#3$//!7)3/&6'.3&!.3!&-$!6'3P.35+![3$!)=!&-$!03/&6$//$#9!=))&$#!
()*$%/!#$%$&$/!'&!/&$2!U9!*.&-!&-$!)&-$6!#$%$&.35!'&!/&$2!\+!
'+!ৄ!
!
M!M!
M! !
4+! :*3J<:1'6K<!
MMS!
:*'3J<1'6.!
:]<
B&$2!U!)=!H*'3'1'6.H!ৄ!I:*3J<:16K<L!
4+!
!
MMS!!
! M!
7+!! :*'3J<:1'6K<!
:*3J<1'6.!
DBEFF@GFC MND?F@OC
:*'3J<:1'6K<! ?@ABCMMS!
*$'P Q@R
@/! .&!
-'22$3/9!
^-.7-! ()*$%!
#$%$&$/!
./! 03.12)6&'3&9!
/.37$!)=!
0%&.1'&$%>!
4)&-! #$%$&$+!
! -'/! 4$$3!
'7-.$($#!
4>!
T3! '! *)6#!
)=! =.6/&!
&-./! /.X$9!
=0%%! /'&./='7&.)3!
?@ABCDBEFF@GFC
'+!ৄ!
!
M! WorksFT_`
M!
:*3J<:1'6K<!
McCarthy, John
J. "An
Introduction to Harmonic Serialism."
From the Selected
of John
J. McCarthy, University
&-$!7)3/&6'.3&!6$/2)3/.4%$!=)6!%$=&D&)D6.5-&!=))&!2'6/.359!@
DF:=))&9!*)6#<9!'%/)!='()6/!
F@OC*$'P9!
&-$!$3#!)=!/&$2!V9!/)!.&!./!=.3'%%>!2)//.4%$!&)!'&&$3#!&)!&-$!6$Y0.6$1$3&/!)=!MND?
of Massachusetts-Amherst. January 2010. © John J. McCarthy. All rights reserved. This content is excluded from
#$%$&.35!=6)1!%$=&!&)!6.5-&!a!-$37$!I:*3J<:1'6K<L!6'&-$6!&-'3!I:*'3J<:16K<L!./!/-)*3!
4+!
MMS!
!
:*'3J<:1'6K<!
*-.7-!./!&-$!3$Z&!1'6P$#3$//!7)3/&6'.3&!.3!&-$!6'3P.35+![3$!)=!&-$!03/&6$//$#9!=))&$#!
our Creative
Commons
license. For more information,!see http://ocw.mit.edu/help/faq-fair-use/.
'/!&-$!*.33$6!.3!:]<+!
()*$%/!#$%$&$/!'&!/&$2!U9!*.&-!&-$!)&-$6!#$%$&.35!'&!/&$2!\+!
^-.7-! ()*$%! #$%$&$/! =.6/&! ./! 03.12)6&'3&9! /.37$! 0%&.1'&$%>! 4)&-! #$%$&$+! @/! .&! -'22$3/9!
b.3'%%>9!
&-$! #$6.('&.)3!
7)3($65$/!
'&! /&$2!of;9!
'3#!
*.33$6!
.#$3&.7'%! &)!
So
Parse-Syll
» *V-Place/weak
enforces
multiple footing
the*.&-!
word .320&!
before
can occur;
:]<
B&$2!U!)=!H*'3'1'6.H!ৄ!I:*3J<:16K<L!
FT_`deletion
DF:=))&9!*)6#<9!'%/)!='()6/!
&-$!7)3/&6'.3&!6$/2)3/.4%$!=)6!%$=&D&)D6.5-&!=))&!2'6/.359!@
)3$!
'3)&-$6+!
@%&$63'&.($/!
&)!
&-$!
.3&$3#$#!
*.33$69!
/07-!
'/!
&-)/$!
.3!
:c<4!
'3#! :c<79!
then
the !weak vowel
of (marí) deletes
toDB
give
(wná)(mrí).
this Q
output
#$%$&.35!=6)1!%$=&!&)!6.5-&!a!-$37$!I:*3J<:1'6K<L!6'&-$6!&-'3!I:*'3J<:16K<L!./!/-)*3!
EFF@GFC
MND?When
F@OC*$'P
@R is submitted again to
:*'3J<:1'6K<!
?@ABC
6$.3&6)#07$!(.)%'&.)3/!)=!&-$!&)2D6'3P$#!1'6P$#3$//!7)3/&6'.3&/!)6!(.)%'&$!='.&-=0%3$//!
'/!&-$!*.33$6!.3!:]<+!
Gen-Eval,
all syllables
are parsed, no weak!vowels appear, and
'+!ৄ!
M! no further
M! deletion will occur so
:*3J<:1'6K<!
7)3/&6'.3&/!56'&0.&)0/%>W!
b.3'%%>9!
&-$!output
#$6.('&.)3!
7)3($65$/!
'&!and
/&$2!
;9!(wná)(mrí)
*.&-! .320&!
'3#!
*.33$6!
Eval converges
on same
as in the
previous loop
thus
is the
final
output..#$3&.7'%! &)!
4+!
! .3&$3#$#! *.33$69!
MMS!
! '/! &-)/$! .3! :c<4! '3#! :c<79!
:*'3J<:1'6K<!
)3$!
'3)&-$6+!
@%&$63'&.($/!
&)!
&-$!
/07-!
An alternative derivation that deletes the weak vowel of the second foot and then the first also
^-.7-!
()*$%! #$%$&$/! =.6/&! ./! 03.12)6&'3&9! /.37$! 0%&.1'&$%>! 4)&-! #$%$&$+! @/! .&! -'22$3/9!
6$.3&6)#07$!(.)%'&.)3/!)=!&-$!&)2D6'3P$#!1'6P$#3$//!7)3/&6'.3&/!)6!(.)%'&$!='.&-=0%3$//!
converges on (wná)(mrí).
FT_`DF:=))&9!*)6#<9!'%/)!='()6/!
&-$!7)3/&6'.3&!6$/2)3/.4%$!=)6!%$=&D&)D6.5-&!=))&!2'6/.359!@
7)3/&6'.3&/!56'&0.&)0/%>W!
#$%$&.35!=6)1!%$=&!&)!6.5-&!a!-$37$!I:*3J<:1'6K<L!6'&-$6!&-'3!I:*'3J<:16K<L!./!/-)*3!
Classical OT has no intermediate stage and thus has no easy way to distinguish the intended
'/!&-$!*.33$6!.3!:]<+!
winner from
losing&-$!
competitors
based7)3($65$/!
solely on the'&!
output
plus .320&!
Max-V:'3#!
it is only
by going
b.3'%%>9!
#$6.('&.)3!
/&$2!forms
;9! *.&-!
*.33$6!
.#$3&.7'%! &)!
through
the intermediate
step of creating
iambic (sˈs)
that the deleting
weak
vowels
)3$! '3)&-$6+!
@%&$63'&.($/!
&)! &-$!an.3&$3#$#!
*.33$69!
/07-! '/!
&-)/$!
.3!are
:c<4! '3#! :c<79!
defined
6$.3&6)#07$!(.)%'&.)3/!)=!&-$!&)2D6'3P$#!1'6P$#3$//!7)3/&6'.3&/!)6!(.)%'&$!='.&-=0%3$//!
7)3/&6'.3&/!56'&0.&)0/%>W!
/wanamari/
Parse-syll
*weak
Max-V
>(wná)(mrí)
(wá)(nmá)(rí)
**
*
/u-wanamari-ri/
>(má)(nrí)(rí)
(ú)(mná)(rrí)
***
***
2
Example 2: opaque stress in Levantine Arabic (Elfner 2009)
/katab-at/
->
/katab-na/
/katab-t/
(káta)bat
‘she wrote’
ka(táb)na
‘we wrote’
katábit
‘I wrote’
1. derivational model (Abu-Salim 1982)
/katab-at/
/katab-t/
kátabat
katábt
stress
----------
katábit
epenthesis
2. OT constraints:
*Complex onset,coda: penalize syllables with a complex onset or coda
PwHd: penalize a Prosodic word that does not parse a foot (and therefore have stress)
Foot-Bin: penalize a foot that is not bimoraic (one heavy or two lights)
Non-Fin: penalize a foot that parses a word-final syllable
Parse segments into syllables: penalize a segment not in a syllable
Align-Rt: penalize feet that are not aligned with right edge of word
3. ranking: PwHd, Non-Fin, Ft-Bin » Align-Rt
/katab-at/
PwHd
Ft-Bin
Non-Fin
Align-Rt
> (ká.ta).bat
ka.ta.bat
*
*!
ka(tá.bat)
*!
ka(tá)bat
4. ranking:
*!
*
PwHd » *Complex » Parse-seg
/katab-t/ -> ka(tá)bit: step 1
/katab-t/
PwHd
*Complex
> ka(táb)<t>
ka.ta.bit
ka.(tábt)
Parse-seg
*
*!
*!
3
/katab-t/ -> ka(tá)bit: step 2
/ ka(táb)<t>/
PwHd
*Complex
Parse-seg
> ka(tá)bit
ka(táb)<t>
*!
5. On the next Gen-Eval loop, ka(tá)bit is still mapped to ka(tá)bit and so this is the output.
6. The serialization of Stress before Epenthesis depends crucially on allowing candidates with
just one change per loop (either epenthesis or foot parsing but not both) and relying on
constraint ranking to choose foot parsing over epenthesis as the optimal output; on the second
loop the requirement to have a stress is satisfied and so the candidate with epenthesis plus the
earlier introduced stress can emerge as the winner.
7. Classical OT cannot (easily) choose (kata)bat over ka(ta)bat while also choosing ka(ta)bit over
(kata)bit without referring to the derivational status of the vowels as underlying (templatic) vs.
inserted.
Example 3: Salayarese stress and epenthesis
/kartu/ -> kará:tu but /kikir/ -> kí:kiru
ranking hierarchy: *rC » *r# » *rV
(this is a familiar hierarchy when we see Donca’s work)
ranking: *rC » PWrHd » *r# » *rV
/kartu/ -> kará:tu
/kartu/
Step 1
*rC
kartu
*W
(kár)tu
*W
/karatu/
Step 2
*rC
> ka(rá)tu
karatu
/kartu/ -> kará:tu
/ ka(rá)tu /
> ka(rá:)tu
ka(rá)tu
*r#
*
> karatu
/kartu/ -> kará:tu
PWrHd
PWrHd
*r#
*W
Step 3
StoW
*W
4
/kikir/ -> kí:kiru Step 1
*rC
/kikir/
PWrHd
> (kíki)r
kikir
*r#
*W
*W
kikiru
/kikir/ -> kí:kiru Step 2
*rC
/(kíki)r/
PWrHd
> (kíki)ru
(kí:ki)r
*r#
*W
/kikir/ -> kí:kiru Step 3
/(kíki)ru /
> (kí:ki)ru
/(kíki)ru
Str-to-W
*W
The Serial OT model will not be able to treat all cases of derivational opacity the same way;
future research will determine which provides an overall better typology of the data
Stratal OT
In this model the input is mapped to the output through a series of sub-grammars of classicalformat parallel OT constraint rankings. The subgrammars reflect successively more inclusive
morphological and phrasal domains: e.g. stem > word > phrase. The constraint rankings may
differ (minimally) from one stratum to another.
Example 1 Levantive Arabic stress and epenthesis
1. derivational model:
/katab-at/
/katab-t/
kátabat
katábt
stress
----------
katábit
epenthesis
word-level: Dep-V » Parse-Segment; PwHd, Non-Fin, Ft-Bin » Align-Rt
/katab-t/
> ka(táb)t
(káta)bit
Dep-V
Parse-Seg
*
*!
phrase-level: Parse-Seg » Dep-V
/ ka(táb)t /
> ka(tá)bit
ka(táb)t
(káta)bit
Ident-stress
Parse-Seg
Dep-V
*
*!
*!
5
Stratal OT appears to be able to describe more cases of opacity than Harmonic Serialism if one is
allowed to freely assign different alternations to different strata, sometimes without independent
evidence from the morphology. At least for the Arabic case there is some independent evidence
for the phrasal (postlexical) status of the epenthesis since if the next word begins with a vowel
then there is no epenthesis: katábt il-kútub ‘I wrote the books’ vs. šírib il-ʔáhwa ‘he drank the
coffee’
Example 2: North German (Ito & Mester 2002)
Meere [me:Rə]
Licht [ɕ]
Meer [me:ɐ]
Buch
[x]
Chemie [ɕ]
solch [ɕ]
Kirche [kɪɐɕə]
rules:
/bu:ɕ/
/ kɪrɕə /
bu:x
--------
ɕ -> x / [+back]----
-------
kɪɐɕə
r -> ɐ in coda
Stratum 1
*[+back, -cons] ɕ » Ident-[back] (M > F)
*[+back, -cons] ɕ
/buɕ /
Ident-[back]
*
> bux
*!
buɕ
Ident-cons » *coda-rhotic (F > M)
Ident-[cons]
/me:R/
*coda-rhotic
*
> me:R
*!
me:ɐ
Stratum 2
Ident-back > *[+back, -cons] ɕ (F > M)
*coda-r > Ident-cons
/me:R/
*coda-rhotic
Ident-[cons]
*
> me:ɐ
me:R
(M > F)
*!
6
Ident-[back]
/kɪRɕəә/
*[+back, -cons] ɕ
*
> kɪɐɕəә
*!
kɪɐxəә
Bloomfield (1933) observed that diminutive suffix –chen does not shift to a back consonant after
a back vocoid (Kuh, Kuhchen [ku:ɕən]) and suggested that there was a # juncture in this
structure /ku:#ɕən/, supporting the idea that the Ident-[back] » *[+back, -cons] ɕ ranking
characterizes the outer layer of word structure
See the recent thesis of Patrick Jones for an in-depth analysis of the complex tonal patterns of
Kinande cast with the Stratal OT model.
References
Abu-Salim, Issam M. (1982). A reanalysis of some aspects of Arabic phonology: a metrical
approach. PhD dissertation, University of Illinois.
Jones, Patrick. 2014. Verbal Tone in Kinande: Prosody, Opacity, and Morphosyntactic Structure.
MIT Ph.D. dissertation.
Bloomfield, Leonard. 1933. Language: New York: Holt.
Elfner, Emily. 2010. Stress-Epenthesis interactions in Harmonic Serialism.
Kiparsky, Paul. 2000. Opacity and cyclicity. The Linguistic Review 17, 351-67.
McCarthy, John J. (2008a). The serial interaction of stress and syncope. NLLT 26. 499-546.
7
MIT OpenCourseWare
http://ocw.mit.edu
24.961 Introduction to Phonology
Fall 2014
For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.