Using grammaticalization theory for reconstruction Uniformitarianism The precise that we cannot postulate types of structure that are different from what we observe today. or applying this to processes of change:

The processes of change we postulate in evolution are the same as processes of change observed today or in the recorded past. Beginning with initial symbols (words) as soon as symbols can be combined, grammaticalization (and constructionalization) can begin.

No need for special genetic mutation to create grammar. Vast memory storage Highly developed sequencing (hierarchical)abilities Finely tuned perception Highly developed motor ability (oral and manual)

Inferencing mechanisms Theory of mind Social awareness; joint attention

All of these abilities can evolve and increase while language is evolving. Heine and Kuteva 2007, 2012 A theory of evolution Language starts with nouns

Expands to other lexical categories: verb, adjective and adverb…

Then, presumably once word combinations are possible, the lexicon can expand and grammaticalization begins. Diachronic typology addresses phoneme inventories

Joan Bybee University of New Mexico Greenberg and emergence

Greenberg 1969 proposes that language comparison can proceed by the comparison of diachronic changes, which occur ‘in historically independent cases’ (p. 151)

‘…the true regularity lies in the dynamic tendency rather than in the static situation. It is not merely that an exception is “explained” by reference to historical process. The implication is that the valid generalization pertains to the form and conditions in which a historical change occurs.’ (p. 171) Grammaticalization

Very similar changes across related and unrelated

Paths of change

Mechanisms of change Synchronic crosslinguistic generalizations

Universals paths of change

Universal mechanisms of change This is the basic outline of a complex adaptive system

Language is like other non-linear dynamic systems:

Like the weather, language never settles down to a fixed equilibrium state.

But as languages unfold over time, like other non- linear dynamic systems, they exhibit regularities, while also permitting variation. Attractors

Lorenz (1993: 39-41) explains that in dynamic systems, certain patterns tend to recur, whereas others tend not to occur.

Lorenz, Edward N. 1993. The Essence of Chaos. Seattle: University of Washington Press. Attractors

‘…the states of any system that do occur again and again, or are approximated again and again, more and more closely, therefore belong to a rather restricted set. This is the set of attractors.’

A simple attractor may be a fixed point; the lowest point in a bowl to which a marble rolls. Attractors in language and language change

Gram-types that occur in unrelated languages Future, perfective, plural, 1st person

Attractors are determined by initial states dynamic processes Attractors in language and language change

Initial conditions include Properties of the social, cultural and physical environment, including the human body Cognitive predilections: what people want to talk about and how they structure their conversations Language-specific properties such as particular lexical items and grammatical constructions Attractors in language and language change

Processes that occur repeatedly as language is used chunking automatization habituation and bleaching inference-making generalization and schematization • For gram-types and construction-types we have developed a general model in accord with complex adaptive systems. Synchronic crosslinguistic generalizations

Universals paths of change

Universal mechanisms of change What about phonology?

Phonetic tendencies become phonological processes in a language.

They continue their development by entering into morphological and lexical domains.

But the domain that has actually been subject of modeling in this framework, still presents problems. Phoneme inventories

Modeling of phonemic systems Liljencrants and Lindblom 1972 (testing the hypothesis that repel one another and therefore are maximally dispersed in perceptual space)

Lindblom, MacNeilage and Studdert-Kennedy 1984 (deriving onset-nucleus units via constraints on articulation and perception) finding that segment, feature and rule are emergent properties Phoneme inventories

De Boer 2000 derives inventories with a feed- back model which uses both perceptual and production constraints. Phoneme inventories

Question:

constraints = ? Vowel inventories

Can we derive vowel systems from known sound change types?

Are sound changes affecting vowels motivated by perception or production or both? Simplified schema of the of English (1400- 1600) Vowel shifts

By definition, chain shifts among vowels preserve phonemic distinctions.

Chain shifts provide evidence that categorization of vowels is active in sound change, keeping vowels perceptually distinct. Maximal dispersion in vowels shifts

Bad news: even maximally dispersed vowels undergo shift.

If [i:] is an optimal vowel, why does it ever change?

In the GVS [i:] and [u:] are believed to be the first to shift.

Weakening of the initial portion of the extreme tongue gesture. Maximal dispersion in vowel shifts

Good news:

After the high vowels diphthongize and shift downward, the mid vowels rise to take up the maximal position once more. Simplified schema of the Great Vowel Shift of English (1400- 1600)

1 1 2 2

3 3

4 The usual account of this process is that the gap left by the diphthongization of high vowels allowed the mid vowels more space for variation.

Perhaps to enhance distinctiveness the higher variants were favored in production. Articulatory processes in vowel shifts

Some vowel changes may start as to the consonantal context.

U- in American and British English is strongest after a palatoalveolar consonant (Harrington et al. 2008, Koops et al. 2013).

Vowel changes in the Northern Cities Shift show considerable effects of surrounding consonants, suggesting articulatory factors (Labov 1994). Conclusion about vowels

Vowel systems are probably shaped by both perceptual and articulatory processes or sound changes.

More research is needed to make the ‘constraints’ used in modeling vowel system emergence reflect what really happens in sound change. Emergence in consonant systems Lindblom and Maddieson 1984 (based on 317 languages of the UPSID database, Maddieson 1984)

Basic small

Elaborated large

Complex larger

Observation: small inventories contain a core set of consonants that are also present in the larger inventories, along with phonetically more complex segments. Consonant systems do not randomly select Cs No systems with many complex segments and few basic ones:

p d ts’ x w z G ł Ɂ Basic Consonant Inventory

p t k

b d g

h s

z

m n ŋ

ɾ l

j

Boro, Tibeto-Burman, India (Maddieson 1984:353) Elaborated Consonant Inventory ( and fricatives)

p t ts k

b d dz g

h f θ s x

v ð z γ

m n

ɾ l

j

Greek, Indo-European (Maddieson 1984:263) Elaborated Consonant Inventory (points of articulation)

p t c k kp

b d ɟ g gb

h f s ʃ v z ʒ

m n ŋ

ɾ l

j w

Senadi, Niger-Congo, Côte d’Ivoire (Maddieson 1984: 290) Elaborated Consonant Inventory (secondary articulation)

p pj t tj k kj

b bj d dj g gj ts tsj tʃ h dz dʒ s sj ʃ ʃj x v vj z ʒ ʒj m m n ŋ j r rj l lj

j

Lithuanian, Indo-European, Europe (Maddieson 1984:264) Complex Consonant Inventory

p t k kw q qw

b d g gw

ɓ ɗ ħ h ʔ f s: s ts’ tɬ’ ʃ x xw z

m n ŋ ŋw

rr l

j w

Iraqw, Cushitic, Tanzania (Maddieson 1984:315) Despite complexities and elaborations these consonant inventories include the basic Cs.

What is the source of complexity and elaboration? Consonant systems

The elaborations on the core system include many features created by well-known sound change types:

Added points of articulation (palatal, uvular) Secondary articulations (palatalization, ) Fricatives and affricates (from of stops) Voiceless sonorants (devoicing in voiceless contexts) Elaborations from less well-known sources:

Prenasalization (from voiced plosives) Ejectives (from voiceless obstruents, Greenberg 1970) Implosives (from voiced stops, Greenberg 1970) For elaborated and complex consonant systems

Constraints used in modeling (could) =

Known mechanisms found in sound changes Basic consonant systems

How are they created and maintained?

Maddieson 1984:16 asks

‘Is maximization of distinctiveness the principle upon which inventories are constructed?’

NO ‘certain dimensions of contrast are preferentially used before others in ways that do not seem related to salience’ ‘Basic Consonants (Lindbom and Maddieson 1988) Properties of the L&M Basic Consonant System 1. Three points of articulation: labial, dental/alveolar, velar (Glottal stop)

2. Voiceless stops 3. Voiced stops 4. Voiceless fricatives 5. Nasal stops

6. Two liquids (dental/alveolar) 7. Two glides (labial and palatal) Sound changes affecting the basic consonant system

Articulatory or perceptual?

Do consonants undergo chain shifts?

Yes, for manner and phonation type (Romance):

pp > p p > b b > β β > ø

Here Cs are remaining distinct; OR lenition is affecting them all at the same time. Sound changes affecting the basic consonant system

Consonant do not undergo chain shifts for place of articulation! p > t t > k k > p

No pressure comparable to that found in vowel systems to fill out the space, OR such changes are not articulatorily possible since the space is not continuous. How is the basic system maintained?

Consider the basic obstruents: /p t k s/

It is much more likely for a sound change to affect and change one of these segments than for a sound change to produce one of these segments. Segment Input to process Output from process /p/ 73 13 /t/ 88 17 /k/ 111 17 /s/ 38 15

Using a database (Allophon) of 800+ phonetic processes described in 82 languages (Bybee and Easterday) New sources of /p t k/ Changes within consonant clusters (10) place assimilation (Latin: septem > sette(m)) voicing assimilation (Uighur: C > vl / vlC __) continuants > stops (Greek: θ> t /C__)

Initial and final devoicing (4) maintaining POA

Other word-final changes

Heavily context-dependent, not a robust source of new plain consonants. New sources of voicelessness

Initial or final devoicing (of all obstruents) (4 Lgs)

Devoicing in C cluster (6 Lgs) Assimilation to voiceless C

Not a robust source of voiceless stops New sources of stops

In a C cluster (4 Lgs.) continuants > stops

Modern Greek: p, b, f, v > p /__s k, g, x, ɣ > k /__s f > p in 3 languages New sources of labial, dental/alveolar and velar

Where /p t k/ are created, the POA was already present in the segment or in environment.

Assimilation in a cluster: Latin: septem > sette(m)

Except: k > t / __# (Cantonese) t t’ > k k’ /__ front V (Slave) *Palatals > alveolars / __ɨ (Lahu) New sources of /p t k/ Heavily context-dependent

features of voicelessness stopness three POA are present in the environment

Processes / sound changes that create new labial, dentals / alveolars or velars are extremely rare. Processes operating on /p t k/ Voicing Fricativization Labialization Palatalization Retroflexion

Many, many more ways to modify these basic Cs than to create them. Segment Input to process Output from process /p/ 73 13 /t/ 88 17 /k/ 111 17 /s/ 38 15

Using a database (Allophon) of 800+ phonetic processes described in 82 languages (Bybee and Easterday) New sources of /s/

Assimilation in consonant clusters (8)

Processes related to palatalization (3)

Lenition from affricates

Final devoicing

Perhaps more sources of new /s/ Sources for nasal stops at the basic POA Assimilation in POA to following C (18 Lgs.)

Oral stops become nasal in a nasal context (POA stays the same) (13 Lgs.) n > ŋ before or after a C, before # or pause (4 Lgs.)

This last type of process gives new instances of ŋ ‘Basic Consonants (Lindbom and Maddieson 1988)

p t k > ʔ

p > f > h

t > ts > s > h

t k > tʃ / __i, j, e

t d > l r

p k > w

k g > j Sources exist for all basic Cs except the following:

p t k

(b d g)

m n ŋ puzzle

For stops at the three basic places of articulation, the question is how do they become established and remain unchanged?

Lenition processes / sound changes should destroy these stops. puzzle

Rather than a continuous cyclical process or destroying and re-creating these consonants, the evidence mostly supports ways of changing them.

So why do they persist?

Could it be that their stability in word-initial position allows them to survive over a long period of time? puzzle 1

Could it be that their stability in word-initial position allows them to survive over a long period of time?

When other Cs assimilate to them, does that strengthen them enough to allow them to survive? puzzle 2 That leaves open the question of where they came from.

We know something about the sources of other points of articulation and other manners and phonation types, but we have very little evidence of sound change sources for /p t k b d g m n ŋ/

Or to put it another way, for the three basic points of articulation. A starting point for the evolution of speech sounds?

MacNeilage and Davis on babbling: Three CV patterns:

Coronal C + front V Dorsal C + back V Labial C + central V A starting point for the evolution of speech sounds?

MacNeilage et al. 2000: Given mandibular oscillation, coronal + front and dorsal + back can be achieved with the tongue in a non-rest position along the front-back dimension and the third with the tongue in rest position.

‘Such a movement pattern seems likely to have been part of the earliest speech of hominids, for it can be considered the most basic movement pattern of the most basic oral articulator’ (p. 158) Evolutionary different sources for basic and non-basic consonants

p t k m n ŋ and possibly voiced counterparts come out of non- linguistic oral movements, while the non-basic Cs arise through processes occurring in the context of language use.