Putting together the pieces

1
Putting together the pieces

• An intra-disciplinary look at sound change

2
Outline

• 1) Background on Neogrammarian Controversy
• 2) Why the controversy is ongoing
• 3) Why the split is unnecessary and problematic
• 4) End to Neogrammarian Controversy
• 5) Proposed solution analogy all the way down
• 6) Examples and discussion of usage-based models
• 7) Conclusion

3
1) Background on the Neogrammarian Controversy
4
Neogrammarian Controversy

• Neogrammarian sound change (regular sound
  change or sound change proper)
• 1) exceptionless
• 2) phonetically (i.e., physically) motivated
• Other key players analogy and dialect borrowing

5
The Neogrammarians

• Aller Lautwandel, soweit er mechanisch vor sich
  geht, vollzieht sich nach ausnahmslosen Gesetzen,
  d.h. die Richtung der Lautbewegung ist in allen
  Angehörigen einer Sprachgenossenschaft, ausser
  dem Fall, dass die Dialektspaltung eintritt
  stetts dieselbe, und all Wörter, in denen der
  der Lautbewegung unterworfene Laut unter gleichen
  Verhältnissen erscheint, werden ohne ausnahme von
  der Änderung ergriffen.(Wilbur 1977 xl,
  reproducing Brugmanns Vorwort to Morphologische
  Untersuchungen xiii)
• Every sound change, as long as it proceeds
  physically, comes to completion by following
  exceptionless rules, i.e., the course of sound
  change is in every member of a speech community,
  unless dialect split is caused by the change, and
  all words in which the sound appears in the same
  conditioning environment, will, without
  exception, be subject to the change. (my
  translation)

6
The Neogrammarians

• Neogrammarian principles
• Invoking the uniformitarian theory, they argued
  that analogy was a natural mechanism of change in
  the present, thus also the past.
• With analogy and dialect borrowing in their
  toolkit, they argued that one didnt need a new
  sound law to describe every change or deviation
  from otherwise regular-seeming sound change.
• So they postulated that sound change
  Lautgesetze must be limited to those following
  the principles of phonetic motivation and
  exceptionlessness.
• Based on their experience, it seemed that
  regularity might be limited to phonetically
  driven mechanisch changes, and analogy could
  cover the more sporadic, cognitively-driven
  geistlich changes.

7
Arguments with mechanism of sound change

• Lexical Diffusion whether sound change can
  proceed word by word, possibly leaving some words
  unchanged
• Grammatical Conditioning whether regular sound
  change can be conditioned or blocked by
  non-phonetic factors vs analogical repair
• Language-internal vs. external factors what role
  do social factors play
• Physical vs analogical what is phonetic vs
  cognitive
• Actuation vs expansion what is sound change
  proper vs spread of sound change

8
Mechanism of sound change

• Centuries-old sound changes cannot be entirely
  proven to have developed through one or another
  process because we cant go back to the beginning
  and witness it.
• Descriptions of sound change in progress are
  subject to debate because once in progress,
  they beginning is past although we can observe
  sound change development.
• The most compelling reason for the more recent
  attempts to delineate sound change, or
  actuation, from expansion
• Expansion of sound changes in progress
  frequently do not obey the rules of Neogrammarian
  sound change because they are subject to
  processes of multiple kinds of analogy and the
  influence of social factors.

9
2) Why the controversy is ongoing
10
Why the controversy is ongoing

• To maintain hold on comparative method, which is
  less effective the less regular the
  correspondence sets are.
• Analogical forms and borrowings arent compared
  because they arent considered regular processes.
  
• We need the comparative method to establish
  language relationships.

11
Why the continued controversy?

• Neogrammarians are insistent because they want an
  explanation for what motivates regularity, some
  kind of universal, or at least default,
  principle.
• The original Neogrammarians were fighting against
  an endless supply of speculation and
  non-universals. They were trying to nail down
  something more reliable, that generalizations
  could be drawn from that might teach us something
  useful about language and allow us to be sure
  that we are reconstructing languages and
  establishing relationships based on principled
  methods rather than guesswork (Wilbur 1977).

12
3) Why the split is unnecessary and problematic
13
Why the split is unnecessary and problematic

• Plenty of processes can have regular outcome
  without being purely phonetically motivated.
• - Analogy (morphological, lexical, phonological)
• - Hypercorrection (Pargman 1998)
• Some phonetically-motivated processes discounted
  by Neogrammarians can display more-or-less
  regular outcomes
• - Dissimilation (Grassmanns Law)
• - Metathesis (Hock 1985)

14
Why the split is unnecessary and problematic

• Many of the Neogrammarians Lautgesetze (sound
  laws) such as Grimms law or the Old High German
  Consonant Shift were chain shifts or parallel
  shifts, which cant be purely phonetic because
  they involve acoustics and perception, and are
  most likely analogical because they involve
  phonemic re-categorization based on what other
  phonemes are doing.
• Even assimilation, especially anticipatory
  assimilation, must begin in the mind. Your
  articulators dont know what sounds are coming
  up, but your mind does, and prepares for them by
  making an earlier sound more like a later.

15
Why the split is unnecessary and problematic

• Plenty of phonetically-motivated changes can be
  irregular.
• - incomplete sound changes
• - near-mergers and near-splits
• Phonetically motivated mini-sound changes
  (Ohala 2003) may not spread, and usually dont.
• Variation occurs even within probabilistic
  regularity.
• Low-level phonetic variation can persist over
  long periods of time without leading to sound
  change the variation has to cause categorical
  change before it is registered as a sound change,
  which can depend on the perceptual space, lexical
  items involved, frequency, social identification,
  and other non-physical factors.

16
Why the split is unnecessary and problematic

• Having a default assumption that sound change
  will proceed exceptionlessly as long as it is
  phonetically driven leaves us with a lot of
  exceptions.
• The assumption is based on our ability to find
  more frequent regularity in phonetically
  conditioned and unconditioned sound changes.
• But exceptions abound.
• No proof of what about phonetically motivated
  sound change would make it more likely to be
  regular.
• A phonological conditioning environment is only a
  description, not an explanation.

17
Why the split is unnecessary and problematic

• We dont have to throw regularity out the window
• On the basis of finding and establishing regular
  sound correspondences, we can still find
  statistical regularity, and investigate
  exceptions.
• Statistical regularity can still give way to
  outliers - the outliers should be investigated in
  any case

18
4) Resolution of the Neogrammarian Controversy
19
Resolution of the Neogrammarian Controversy

• Neogrammarian sound change, at least as laid out
  by the Neogrammarians, cannot exist. Heres why
• 1) Neogrammarian Sound change must be
  exceptionless
• 2) Neogrammarian Sound change must proceed
  mechanically, that is purely physically
  (phonetically).

20
Resolution of the Neogrammarian Controversy

• 1) There is no such thing as exceptionlessness,
  only an endless cycle of variation and
  generalization.
• That is, even regular sound change is only
  probabilistically regular, including phonemic and
  non-phonemic variation
• Because all synchronically stable sounds also
  exhibit phonemic and non-phonemic variation
  (Peterson and Barney 1952, Ladefoged and
  Broadbent 1957, among many others)

21
Resolution of the Neogrammarian Controversy

• 2) No aspect of language can be removed from
  cognitive processes because production is
  entrenched in a feedback loop with perception
  therefore, analogical processes are at work
  before, during, and after each utterance.
• At its most basic, analogy is the association
  of some thing with some other group of things.
• Phoneme identification is the association of an
  unspecified period of sound, varying along many,
  mostly continuous dimensions, with other such
  continuously varying productions, to arrive at a
  shared category.

22
Resolution of the Neogrammarian Controversy

• Infants must learn to generalize the stream of
  random and non-random noise emanating from their
  caregivers into words and sounds before they
  begin to speak.
• Infants at 6 months and younger can distinguish
  between syllables differing by a distinctive
  feature, but lose the ability to distinguish
  within phonemic categories of their L1 after they
  have begun to generalize what phonetic
  information is relevant to their L1 by around
  10-12 months (Werker and Polka 1993)

23
5) Proposed solution

• Analogy all the way down

24
Analogy all the way down

• Variation in production of speech sounds has a
  basis in both articulatory and cognitive factors
• Regularity happens because the mind is
  predisposed to form patterns and generalizations
  from the input it receives. From a continuous
  stream of sound, the mind breaks the input down
  into components and stores the information based
  on association.

25
Analogy all the way down

• Probably the most automatic of these is the
  segmentation of the sound wave into words. Words
  are more easily and quickly recognized than
  syllables, which are more easily and quickly
  recognized than phonemes (Mehler, Dupous and
  Segui 1990 Werker and Polka 1993).
• And from the lexicon, associations of similar
  sounds into phoneme-like groupings can be derived
  (Beckman and Edwards 2000).

26
Analogy all the way down

• Each token uttered, across a range of
  variability, is generalized both by phoneme and
  by word (and across other associated elements).
• This explains why there is evidence not only for
  what one could call phonologically analogical
  sound change, and for lexical diffusion (lexical
  analogy), but also mixed cases, rare instances of
  grammatical conditioning, and association with
  social factors involving the spread of sound
  change

27
Analogy all the way down

• The reason sound changes that have their roots in
  so-called phonetic motivation (either
  phonetically conditioned or completely
  unconditioned) end up being more frequently
  regular is because they are easily phonologically
  generalizable.
• Phonological generalization generally happens
  below the level of consciousness, though it can
  become more conscious if marked by greater
  distance from the previous phonological
  perceptual structure or if marked by social
  factors.

28
Analogy all the way down

• Note understanding that all sound change is
  rooted in analogical processes does not mean that
  we cant have regularity, nor does it exempt us
  from explaining all sound changes - rather it
  holds us to a higher degree of specificity, that
  we must explain how each instance and pattern of
  analogy worked, and what it was based on.

29
6) Examples and discussion

• Usage-based (exemplar) models

30
Usage-based (exemplar) models

• Exemplar models can account for spread (social
  factors) of sound change (cf. Pierrehumbert
  2006)
• A tight network has more limited variation, and
  is more likely to reinforce whatever pattern the
  variation is taking (including change)
• A loose network has much more potential for
  variation, sometimes having a centering effect on
  the mode
• Status, affiliation, or other sociolinguistic
  factors can lend weight to exemplars with these
  features

31
Usage-based (exemplar) models

• Exemplar models can account for rate of sound
  change
• Phonetic change can proceed only as quickly as
  can be calculated by total accumulated tokens and
  frequency of tokens and acoustic distance between
  tokens
• But some factors can speed up the change, such as
  if the distance between earlier tokens and newer
  tokens increases, the new locus will shift more
  quickly.
• This explains the sigmoidal curve described by
  variationists.

32
Usage-based (exemplar) models

• If the variation is inconsistent, that is new
  tokens are also widely dispersed, then low-level,
  underlying variation can persist for a long time.
  
• Lexical items play a role in this, such that
  sounds can be recategorized due to their presence
  in a word with no minimal pair for that phoneme.
  Hence, the further the nearest phoneme and the
  fewer possible minimal pairs, the more able to
  re-generalize.
• If the difference between tokens is so
  perceptible as to be assigned social meaning, the
  change then can become a matter of higher-level
  operations (e.g., choice of variants for social
  reasons) and is also accelerated (and may be more
  prone to error).

33
Example- non-phonetically conditioned possibly
phonemic variation

• Mini-sound change -- the bang/bangs split, a very
  mixed affair
• 1) Normal raised /ae/ before /ng/
• Very raised to /e/ in some words
• 2) Dialect borrowing --
• D1 /e/, D2 /ae/
• 3) More frequent words develop into separate
  variants in free variation (dang, bang, blanket)
• 4) Infrequent words (bangs, bank shot, blankie)
  belong to one phonemic category /e/
• 5) Infrequent, but higher register (later
  learned?) words belong to one category /ae/
  (dank, vanquish, manx)
• 6) Transitive form of verbs more likely to have
  /e/ (hang, rang)

34
Example- non-phonetically conditioned possibly
phonemic variation

• Other Northern transplants in the community,
  beyond family.
• One could imagine how this might become a sound
  change.
• The split might continue and be adopted by
  others.
• Phonemic categories might emerge from various
  generalizations
• lexical (some words develop as /ae/ others /e/)
  
• grammatical (the trend for transitive verbs
  becomes more generalized)
• phonological (some generalization following
  certain classes of sounds, such as coronals, for
  instance)
• It could become regular if enough people adopted
  the split along the same lines of generalization

35
7) Conclusion
36
Conclusion

• Hans Henrich Hock (2003), also discussing the
  scope of analogy in sound change, poses a similar
  question
• where in this continuum should we draw the
  dividing line, on what grounds should we draw it,
  and should we draw one at all?

37
Conclusion

• I propose that, rather than drawing lines in the
  sand, we should be more explanatory.
• For example, we should answer these questions
• Is it sound change (is one phoneme different
  between two stages?)
• What changes? And what are the conditions for
  change?
• How regular is it (r2, what of variation is
  accounted for by the proposed conditions)?
• Explain any apparent exceptions/outliers.
• What is/are the proposed mechanism/s of change?
• Is there any synchronic evidence for the
  postulated mechanisms?

38
Conclusion

• Divisions into categories of sound change are
  similar to clusters of phonetic commonalities
  into phonemes. The categories have overlapping
  features based on association, corresponding to
  transparency of analogical processes, linguistic
  category, language intrinsic vs not, actuation vs
  expansion, etc. And we are statistically correct
  in saying that changes that emerge as
  phonologically conditioned are more frequently of
  higher regularity than others.
• Analogical processes are the default, and
  irregular and incomplete sound changes are the
  same kind of creature as regular sound changes,
  only with more exceptions, which could be argued
  in the reverse as the result of other
  regularizing processes that sound change got in
  the way of.

39
Conclusion

• If you want an absolute, undeniable basis for
  sound change, the Big Bang (Janda and Joseph
  2003), it is phonetic variation.
• The fact that no two productions are exactly
  alike means that we have to generalize what we
  hear to correspond with what we have heard
  previously and all that we associate it with.
  This means that the acoustic space is a dynamic,
  evolving perceptual system that requires
  generalization, or analogical processes, to
  derive categories of sounds, and also words,
  meaning, and all the stuff that makes language
  work. No matter what interdependent processes
  take place to create sound change, there must be
  variants (more or less differing from one
  another) and association of those variants with
  one or more derived categories, such as phoneme,
  word, register, or status.

40
Conclusion

• The Neogrammarians were correct in that regular
  sound change does have its basis in phonetics,
  and that is phonetic variation though so does
  irregular, incomplete, and every other kind of
  imaginable sound change.
• Regular sound change relies on the analogical
  workings of the inner mind.

41
References

• Beckman, M.E., Edwards, J. (2000). The ontogeny
  of phonological categories and the primacy of
  lexical learning in linguistic development. Child
  Development, 71, 240-249.
• Garner, W. (1974). The Processing of Information
  and Structure. Oxford Erlbaum.
• Hock, H. H. (1985) Regular Metathesis.
  Linguistics 23, 529-546.
• Hock, H. H. (2003) Analogy. In Joseph, B. and
  Janda, R. 2003.The Handbook of Historical
  Linguistics. Malden, MA Blackwell.
• Janda, R. and Joseph, B. (2003) Reconsidering the
  Canons of Sound-Change Towards a Big Bang
  Theory. Historical Linguistics 2001. Selected
  Papers from the 15th International Conference on
  Historical Linguistics, ed. by Barry Blake and
  Kate Burridge. Amsterdam John Benjamins,
  205-219.
• Johnson, K. (1997). Speech Perception without
  speaker normalization. IN K. Johnson and J.
  Mullenix (eds.), Talker variability in speech
  processing. San Diego Academic Press. 145-166.
• Labov, W. (1981) Resolving the Neogrammarian
  Controversy. Language 57 (2) 267-308
• Mullenix, J. And Pisoni, D. (1990) Stimulus
  variability and processing dependencies in speech
  perception. Perception and Psychophysics, 47
  (4) 379-390.
• Ohala, J. (2003). Phonetics and Historical
  Phonology. In Joseph, B. and Janda, R. 2003.The
  Handbook of Historical Linguistics. Malden, MA
  Blackwell.
• Pargman, S. (1998). On the Regularity of
  Hypercorrection in Phonological Change.
  Diachronica 15, 285-307
• Pierrhumbert, J. (2006). The next toolkit.
  Journal of Phonetics 34, 516-530
• Werker, J. F. and Polka, L. (1993). Developmental
  changes in speech perception new challenges and
  new directions. Journal of Phonetics 21, 83-101
• Wilbur, Terence (ed.) (1977). The Lautgesetz
  Controversy a documentation (1885-86).
  Amsterdam Benjamins.

42
Extra Slides
43
Warning

• The following images are highly oversimplified
  representations

44
Separate accumulation of variation of two distant
phonemes
45
Accumulation of exemplars around a mode of
variation
46
Neighboring phonemes
47
Very limited variation causes reinforcement of
phoneme categories
48
Wide range of variation causes overlap of phonemes
49
Empty acoustic space allows more room for
variation
50
Categorically different variants
51
Beginning of sound shift accumulation of new
center of variation
52
Sound shift new sound is as prevalent as old
sound
53
Sound shift new mode covers accumulated
variation, shifts toward center
54
Sound shift is complete new mode is centered
around new target pronunciation
55
Why the split is unnecessary and problematic

• To follow up on an earlier elaboration if we
  say that V1qV2 gt V1ðV2 / only if V1 is not
  accented, it is no more explanatory of sound
  change than if we said V1qV2 gt V1ðV2 /
  intransitive verbs
• A phonological conditioning environment gives us
  a chance to examine why that environment might
  motivate a change, or how such a change may
  proceed, but is not, in itself, an explanation.

56
Resolution of the Neogrammarian Controversy

• If we can bring all of the important
  characteristics of the phonetics of a vowel
  together (including duration, F0, F1, F2, F3,
  phonation type) with associated information from
  its carrier phrase (including semantics, syntax,
  intonation, other measurements of other vowels
  and consonants both near and far) and the speaker
  (including age, height, gender, dialect,
  affiliation) to determine whether we hear the
  sound in pin or in pen, for example (Garner 1974,
  Mullenix and Pisoni 1990, Johnson 1997, among
  others), why would it seem strange that we form
  associations among these factors, which sometimes
  lead to the beginning or expansion of sound
  change?

57
Example- phonetic (non-phonemic) variation

• Stable, underlying variation across the dental
  fricative in American English in Ohio.
• 1) Production (in manner) extends from
  near-vocalic approximants to plosives
• - No competition in manner at same place of
  articulation
• 2) Production extends across a range of voicing
  and sonority
• - Voicing carries virtually no functional load
• 3) Variation in place is mostly limited to
  alveolar, dental and interdental, and mostly to
  the voiced phoneme, which is more likely to
  become alveolar.
• - No other associations in competition