Generative vs.
structuralist
approaches to alternations
LING 451/551
Spring 2011
Generative view of phonology
• Hayes 6.1.1
– „The morphology of a language places morphemes
in different phonological contexts…‟
• Different pronunciations of same morpheme
can be a source of information about
phonology
• Not always thought to be the case...
Structuralist phonology
• Dominant model of phonology prior to the
Chomskyan “revolution” (late 1950s, early
1960s)
• Who were the structuralists?
Eugene Nida
Mary Haas
Rulon Wells
Kenneth Pike
Zellig Harris
Leonard Bloomfield
Charles Hockett
And others...
• Bernard Bloch, George Trager, Henry Lee Smith,
Archibald Hill, Martin Joos, Morris Swadesh, Stanley
Newman, Carl Voegelin, Charles Fries, ...
Structuralist phonology
• A.k.a. “taxonomic” phonology, i.e. „not
explanatory‟
“its striking reliance, in almost all versions,
on procedures of segmentation and
classification (identification of variants)”
(Chomsky 1964)
Structuralist model of grammar
• Levels of representation
Morphological component
(morphological rules)
Morphophonemic component
(morphophonemic rules)
Phonemic component
(phonological rules)
Phonetic component
Structuralist prohibition against
“mixing levels”
• Only “purely” phonological information could be
used to abstract away from phonetic rep
• Alternations---evidence from morphophonemic
component---could not be a source of evidence for
phonological component
• Why mixing of levels outlawed
– assumptions about how linguistic data is processed, like
descriptive linguist working from the “bottom up”
Biuniqueness requirement
• „each sequence of phones is represented by a
unique sequence of phonemes, and ... each
sequence of phonemes represents a unique
sequence of phones.‟
• not okay: /t/ /d/
[th t d]
• okay: /t/ /d/
[th t d]
How structuralists handled
neutralization
• “complete overlapping”
• German Final Devoicing
– [bunt] „colorful‟ [buntes] „colorful‟ (gen.)
– [bunt] „federation‟ [bundes] „federation‟ (gen.)
• Most abstract phonological representations possible:
– /bunt/ /buntes/
– /bunt/ /bundes/
• Since [t d] contrast ([buntes] vs. [bundes]), by
Biuniqueness [t] cannot belong to both /t/ and /d/
Morphophonemic level of
representation
• A.k.a. dictionary level of representation | |
• Represents relationship between forms of
„federation‟
– |bunt| |bunt-es| „colorful‟
– |bund| |bund-es| „federation‟
Argument against biuniqueness
• Made by Morris Halle, 1957 Linguistic Society of
America meeting
Russian obstruent voicing assimilation
• Obstruents agree in voicing with following
obstruent
−sonorant
[-sonorant] [ voiced] / ___
voiced
• [t] [d], [tj] [dj] contrast before vowels (and
sonorants)
– [tam] „there‟
– [dam] „I‟ll give‟
– [xo|tjitje] „you want‟
– [xo|djitje] „walk!
Voicing alternations involving /t d/
• Structuralist analysis
• Voicing Assimilation (for [t]~[d], [tj]~[dj]) must be a
morphophonemic rule (by Biuniqueness, since [t], [d]; [tj]
[dj] contrast)
phonetic phonemic morphophonemic
[datjlji] /datjlji/ |datj-lji| „whether to give‟
[dadjbɨ] /dadjbɨ/ |datj-bɨ| „give‟ (subjunctive)
[godlji] /godlji/ |god-lji| „whether a year‟
[godbɨ] /godbɨ/ |god-bɨ| „year‟ (subjunctive)
More alternations
• In Russian, [x] [ɣ], [ts] [dz], [ʧ] [ʤ] are in
complementary distribution, also participate in
obstruent voicing assimilation
• Structuralist analysis
phonetic phonemic morphophonemic
[ʒeʧlji] /ʒeʧlji/ |ʒeʧ-lji| ‘whether to burn’
[ʒeʤbɨ] /ʒeʧbɨ/ |ʒeʧ-bɨ| ‘burn’ (subjunctive)
Halle‟s observation
• Russian Voicing Assimilation is a
morphophonemic rule when segments contrast
([d]~[t]), a phonemic rule when segments in
complementary distribution ([ʧ]~[ʤ])
• Structuralists‟ theory therefore leads to loss of
generalization
Result
• General abandonment of biuniqueness as
constraint on phonological representations.
• Collapse of phonological and morphophonemic
levels of representation in generative
phonological approaches.
– Generative “underlying representation” =
structuralists‟ morphophonemic level.
Development of classical generative
phonology
• Chomsky and Halle 1968 and later work
– Phonological representations consist of features
(< Trubetzkoy)
– Restrict architecture for description
• formal simplicity should reflect phonological
naturalness
– Source-oriented model
• Rules generate phonetic representations from more
abstract phonological ones.
Abstract analyses on the rise
• Underlying representations in generative
phonology can be more abstract---further
removed from surface pronunciation---than a
phonemic representation that abstracts away
from complementary distribution
• (Later backlash against abstractness)
