Title: Li6 Phonology and Morphology
1Li6 Phonology and Morphology
- Representation of segments
2Todays topics
- feature geometry
- autosegments
- the skeleton
3Fundamental points about FG
- All features are not created equal certain sets
of features pattern together, certain features
never interact, and so on. - Not all rule types allowed by the formal
machinery of traditional linear phonology are
attested/possible. - Our hypothesis to account for these
generalizations - features are organized in an invariant
hierarchical tree structure provided by UG.
4Traditional feature theory
5Traditional segmental representation
- coronal
- -voice
- -cont /t/
- ant
- dist
6Formal elegance
- Consider the widespread phenomenon of nasal place
assimilation. - inept, impossible, incomplete
- How does a traditional linear model (i.e.
pre-1976) account for this process? - Rules of this type (and in fact assimilation
rules in general) have no special status in a
linear model of phonology.
7A stronger hypothesis
- Phonol. has only two basic operations at its
disposal - spreading
- delinking
- Phonol. has only two basic constraints at its
disposal - the Line-Crossing Constraint
- the OCP (Obligatory Contour Principle)
- Desiderata/postulates
- common rule types involve simple machinery
- phonological rules perform single operations only
- feature organization is universally determined
X X F
X F
X X X F F
X X F F
8The Place Node
- Assume that assimilation and dissimilation/OCP,
which are extremely common, operate only on
single elements. - If this is true, then natural classes (sets of
sounds that pattern together with respect to
phonological rules) should be reflected in
representations. - What implication does nasal place assimilation
have for our representation of segments, then? - It suggests that the places of articulation are
grouped under a common node within the tree
structure that makes up a segment. - We call this particular node the Place node.
9The feature tree
back to laryngeal neutralisation
10Evidence for the Place node
- English nasal assimilation
- Sanskrit s-assimilation
UR SR gloss
indras Suras indraS Suras Indra the hero
tas ?a? ta? ?a? those six
divas putras diva? putrah son of the goddess
nalas kamam nalax kamam at will
11How do we represent Place assim?
- The Sanskrit case
- X X
-
- cons cons
- cont
- Place
12Laryngeal spreading
- Sanskrit coda neutralization (cf. Korean)
- nominative accusative gloss
- marut marutam wind god
- suhrt suhrdam friend
- agnimat agnimatham near the fire
- kakup kakubham region
13Laryngeal delinkingNew Julfa future marker
- a. kERtHAm I will go
- k?tAm I will give
- k?kiEnAm I will exist
- b. g?b?z?Am I will buzz
- g?lAm I will cry
- g?z?rAm I will bray
- c. kH?tHo?niEm I will allow
- kH?tSHApHiEm I will measure
- kH?X?ndAm I will laugh
- kH?sAvoRiEm I will grow accustomed to
- d. gH?bHiERiEm I will carry
- gH?gHom I will come
- gH?dH?niEm I will put
- gH?dzHiEviEm I will form
14Autosegments
- phonemes are not simply lists of feature
specifications - features and segments are not necessarily in a
one-to-one relationship (the relationship is
nonlinear) - individual features enjoy a certain amount of
autonomy with respect to other features
characterizing the segments to which they are
associated (they behave as autosegments) - Today well focus on tone vowel harmony in
lecture 8 (round, back, and ATR as autosegments)
15Typology of autosegmental relationships
- a. one to one X X
-
- F F
- b. one to many X X
- F
- c. many to one X
- F F
- d. bare anchor X
- e. floating feature F
1611 and many1
- Margi (Kenstowicz 1994312)
- base form definite gloss
- a. sál sál-árì man
- kùm kùm-árì meat
- b. ?ímí ?ímy-árì water
- kú kw-árì goat
- tágú tágw-árì horse
- c. tì ty-arì morning
- hù hw-arì grave
- ú?ù ú?w-arì fire
H L ˆ HL ( falling) ? LH (rising)
17More Margi
- a. à sá gU? you go astray
- à tsú gU? you beat
- b. á wì gU? you run
- á dlà gU? you fall
- c. á v??l gU? you fly
181many
- Margi (Kenstowicz 1994319)
- a. tSU speak tSí-bá tell
- ghà reach ghà-bá reach
- f ?? swell f ì-bá make swell
- b. sá go astray sá-ná lead astray
- dlà fall dlà-nà overthrow
- bdlU? forge bdl??-ná forge
19Universal Association Convention
- Match tones and tone-bearing units one-to-one,
L?R or R?L.
20R?L mapping Kikuyu
- subject marker object marker root tense suffix
- tò we mò him ròr look at íré
- má they má them tóm send
- tò ròr ìré má rór ìré
- tò mò ròr ìré má mó ròr ìré
- tò mà rór ìré má má rór ìré
- tò tòm íré má tóm íré
- tò mò tòm íré má mó tòm íré
- tò mà tóm íré má má tóm íré
- How do these forms illustrate the following?
- R?L mapping
- one-to-many association
- The OCP
21Floating features
- Japanese mimetic palatalization
- palatalization targets the rightmost non-r
coronal consonant - If there are none, the palatalizing feature links
to the leftmost segment
UR mimetic gloss
a. /dosa/ doša-doša in large amounts
b. /toko/ trotting coko-coko childish small steps
c. /poko/ up and down movement pyoko-pyoko (pokyo) flip-flop, jumping around imprudently
d. /koro/ kyoro-kyoro look around indeterminately
221many pt 2 the skeleton
- Bakwiri syllable reversing language game (Hombert
1986) - kwélì falling ? líkwè
23Geminates
- Patient LBs errors when spelling geminates
(Caramazza and Miceli 1990)
Stimulus Response
pezzo zeppo
cellula leccula
blocco bcollo
24Conclusions
- Nonlinear, autosegmental representation of
segmental material enables us to account
satisfyingly for a wide range of phenomena that
would be difficult or impossible to account for
with non-autonomous linear representations.