The Scope of Generalization in Phonology

1
The Scope of Generalization in Phonology

• Gregory R. Guy
• New York University
• VGFP Workshop, Stanford, July 07

2
Generalization in Phonology

• Identify (and explain?) phonological patterns
  that are prevalent across some domain

3
Maximum generality phonological universals

• For all human speakers (of all languages),
• in all linguistic contexts,
• in all lexical items,
• x is always true.

4
Non-universal generalizations

• Involve limits on either
• the SCOPE of one of domains (the
  all quantifiers)
• OR
• the PREVALENCE of the pattern (the
  always quantifier)
• or both

5
Scope Social domain, contextual
domain, lexical domainPrevalence frequency or
probability

• For all human speakers (of all languages),
• in all linguistic contexts,
• in all lexical items,
• x is always true.

6
Quantifying social scope(e.g. language-specific
generalizations)

• For speakers in some social domain i
• e.g., a speech community, dialect, language,
• OR
• a social group defined by age, class, gender,
  ethnicity, etc.

7
Quantifying contextual scope e.g.,
context-sensitive generalizations, gradience

• .. in some linguistic context j

8
Quantifying lexical scope e.g., lexical
frequency, lexical exceptions

• .. in some lexical domain k ..

9
Quantifying prevalence e.g., variable,
stochastic, or probabilistic generalizations

• .. x is true with a probability p.

10
Quantified Generality

• For speakers in some social domain i,
• in some linguistic context j,
• in some lexical domain k,
• x is true with a probability p
• .. where, typically, p is a function of i, j, k

11
Social scope

• For speakers in some social domain i

12
Social proximity implies linguistic similarity

• Speech community members share grammatical
  properties
• Contrasting Constraints Hypothesis Different
  speech communities may have contrasting values
  for the probabilistic constraints on variable
  processes.
• Shared Constraints Hypothesis The members of a
  speech community share common values for the
  probabilistic constraints on variable processes.

13

• Contrasting constraints

14
Communities differ Following context effect on
coronal stop deletion in two cities

• Speech speakers with Community
• Community constraint ranking preference
• CgtV Cgt0 Vgt0
• Philadelphia 89 100 95 CgtVgt0
• (N19)
• New York 100 50 0 C0gtV
• (N4)
• Cconsonant, Vvowel, 0pause

15
Communities differ Final -s deletion in four
Brazilian cities
16

• Shared constraints

17
Within communities speakers share constraint
rankings and values

• In a study of coronal stop deletion in 16
  Philadelphian speakers, looking at 8 constraints
  (3 morphological and 5 phonological), individual
  results are distributed as follows

18
Shared constraint rankings Coronal stop deletion
in 16 Philadelphians

• number of speakers ()
• deviations from --number of tokens per
  speaker-- random
• community order gt170 100-170 lt100
  distribution
• 0 5 (100) (0.1)
• 1 3 (60) 1 (17) (2.8)
• 2 2 (40) 4 (67) (17.4)
• 3 (39.8)
• 4 (25.0)
• 5 1 (17) (8.5)
• 6 (5.7)
• all 8 (0.1)

19
Shared values with sufficient data, speakers
converge

20

21
Contextual scope

• in some linguistic context j

22
Contextual scope gradient effects on variable
processes

• OCP (Obligatory contour principle) is a general
  phonological constraint against sequences of
  adjacent identical elements.
• In many languages it categorically prohibits
  certains sequences.
• e.g., English affix allomorphy
• cats vs. glasses, backed vs. batted

23
OCP effects are gradient in variable processes

• Place effect on deletion of final coronal cor,
  -ant consonants in three languages
• Percent deleted Factor Weight
• Place Port Span Eng Port
  Span Eng
• Coronal
• cor, ant 21 31 44 .66 .57 .65
• Labial
• -cor, ant 14 32 .53 .56
• Velar
• -cor, -ant 6 16 34 .31 .38
  .35

24
English coronal stop deletion by preceding
context (Guy Boberg 1995)

• Preceding Context N Factor
  weight
• Identity with deletion target
• /t,d/ cor, -son, -cont (categorical
  absence, i.e., 1.00)
• Two shared features
• /s,z,?,z/ cor, -son 276 49 .69
• /p,b,k,g/ -son, -cont 136 37 .69
• /n/ cor, -cont 337 46 .73
• One shared feature
• /f,v/ -son 45 29 .55
• /l/ cor 182 32 .45
• /m,?/ -cont 9 11 .33
• No shared features
• /r/ 86 7 .13
• vowels (nearly categorical retention,
  i.e., 0.00)

25
Conclusion OCP is gradient

• The disharmony of an OCP violation increases in
  proportion to the phonological similarity between
  adjacent elements.

26
Lexical scope

• .. in some lexical domain k ..

27
Lexical issues for phonology

• Lexical exceptions
• Lexical frequency
• Historical borrowings with distinct phonology
  (e.g., Latinate vocabulary of English,
  Chinese-origin vocabulary of Japanese)
• Recent borrowings
• Proper names

28
Defining lexical scope generalizations over part
of the lexicon

• Two strategies for handling lexically-restricted
  properties
• Tweak the phonology
• Tweak the underlying representations

29
Tweaking the phonology

• Exception features co-index phonological rules
  with lexical items they apply to (cf. Chomsky
  Halle)
• Co-phonologies, lexical classes different
  constraints or constraint rankings for different
  subsets of the lexicon (cf. Inkelas, Ito
  Mester)

30
Tweaking underlying representations

• The (lexically partial) generalization is already
  encoded in the UR, not generated by the phonology
• Items that fail to show some generalization get
  URs that block that outcome
• Variable lexical class membership (cf. Coetzee,
  this afternoon)

31
Example English plurals with f-v alternations

• Regular pattern final C is invariant in plural
• cat-cats, chief-chiefs, puff-puffs, etc.
• Exceptional pattern final fgtv in plural
• leaf-leaves, wife-wives, loaf-loaves, etc.

32

• Tweak the phonology
• Special rule for fgtv in plurals
• Exception feature specifies all the words that
  undergo this rule
• Tweak the lexicon
• URs of leaves, wives, loaves have /v/
• URs of leaf, wife, loaf, etc. are under-specified
  for voice, with appropriate conventions to fill
  in specification.

33
Lexical exceptions in variation

• Many variable processes are known to exhibit
  unusual frequencies of occurrence in particular
  lexical items.
• e.g., coronal stop deletion in English is
  exceptionally frequent in and
• (Exceptional because deletion occurs
  significantly more often in and than in
  phonologically comparable words like sand, band,
  hand, etc.)

34
The two strategies applied to lexical exceptions
to variable processes

• Phonological tweak exceptional lexical items
  have a feature that raises or lowers the
  probability of a given phonological process
  occurring in that word.
• e.g., and is associated with an exception
  feature that raises the probability of coronal
  stop deletion.

35

• Lexical tweak exceptional lexical items have
  alternate entries that pre-encode the output of
  the process.
• e.g., and has an alternate entry an. When
  this form is selected, it always surfaces without
  a final /d/, thereby boosting the apparent rate
  of coronal stop deletion.
• (cf. rock n roll, an orthographic
  representation of this underlying form?)

36
Testing the strategiesVariation as a window
into phonological organization

• The two strategies for handling lexical
  exceptions may not be decidable on
  obligatory/categorical data because of absence of
  constraint interaction
• But variation data, showing constraint
  interaction, allows a test of the models.

37
The two strategies make different quantitative
predictions

• Exception feature approach simply boosts the
  overall probability of deletion in and, leaving
  other constraint effects unchanged.
• Hence, effect of following C vs. V should be the
  same in exceptional and unexceptional words
• Cheese n crackers is always deleted more than
  ham n eggs

38

• The lexical entry approach achieves elevated
  surface rates of -d absence in and by selection
  of UR an, which does not undergo coronal stop
  deletion, and is therefore insensitive to
  constraints on that process.
• Hence, lexical exceptions show reduced effect
  of following C vs V
• Cheese n crackers is as likely as ham n eggs

39

• The specific quantitative effect
• A surface corpus of exceptional words is a
  mixture of two sets of foms
• -some are derived from underlying full forms
  (e.g. and) and show the effects of constraints
  on the process,
• -others are derived from underlying reduced
  forms (an) and are not affected by constraints
  on the process

40

• The mixture of the two sets has the quantitative
  effect of attenuating the effect of constraints
  on the process.
• -in a multivariate analysis, this attenuation
  should be manifested as a smaller range of values
  for a factor group measuring a constraint on the
  process (e.g., the following segment effect on
  coronal stop deletion).

41
Predictions

• Exception feature approach constraint effects
  should be equivalent in exceptional and
  nonexceptional corpora
• Multiple underlying entries constraint effects
  should appear to be weaker in exceptional than
  nonexceptional corpora.

42
Data English coronal stop deletion and
exceptional and

• Non-exceptional Exception
  (and)
• words
• N del N del
• __C 572 39.3 441 95.7
• __V 495 15.8 312 82.1
• Range 23.5 gt 13.6
• (Source Neu 1980)

43
Lexical exceptions in Brazilian Portuguese -s
deletion

• Features of following C Non-exceptions
  Lexical exceptions
• (-mos forms)
• Voice/Manner sonorant .69 .49
• voiced obstruent .44 .58
• voiceless obstruent .36 .44
• Range .33 gt .14
• Place labial .32 .58
• coronal .61 .53
• velar .44 .39
• Range .29 gt .19
• N 5880 1225
• Log likelihood -704.8 -791.5

44
-s deletion in Salvadoran Spanish (Hoffman 2004)

• Non-exceptional words Lexical
  exceptions
• Following context (entonces, digamos,
  pues)
• sonorant .60 .63
• voiced obstruent .75 .55
• voiceless obstruent .33 .38
• vowel .36 .38
• pause .44 .56
• Range .42 gt .25
• Syllable Stress
• stressed .38 .42
• unstressed .62 .58
• Range .24 gt .16

45
Summary In 5 constraints (factor groups) on 3
processes in 3 languages

• Magnitude of constraint effect is always weaker
  for exceptional lexical items
• This is consistent with predictions of the
  lexical entry (lexicon tweaking) strategy
  contradicts exception feature (phonology
  tweaking) strategy.

46
Conclusion Speakers tweak the lexicon

• Lexical exceptions to variable processes are
  accomplished by alterations to the underlying
  representation and the existence of multiple
  representations
• (cf. Kiparskys treatment of -t,d deletion in
  stratal OT).

47
Another prediction

• Exception feature approach permits both positive
  and negative exceptions (lexical items that
  undergo a process at a higher or lower
  probability than other words)
• Underlying form approach allows only positive
  exceptions, with higher probabilities (cant
  block -t,d deletion)

48
Impressionistic confirmation

• All lexical exception cases in variation studies
  I am familiar with involve elevated rates of
  occurrence of a variable process, never reduced
  rates.
• This confirms the prediction of the lexical entry
  approach.

49
The Paninian nature of partial generalizations

• Variation involves the quantification of
  prevalence
• Non-universal generalization involves
  quantification of scope, in social, contextual,
  and lexical domains.