Results
Coherence of real and artificial systems
• Figures, from left to right, represent the intrinsic, interactional and global co- herences of UPSID and random systems (red and blue markers respectively).
• The third figure also indicates the most frequent systems in UPSID; they ap- pear to be the ones with the highest coherence.
Stability of real and artificial systems
• Top figures: stability of UPSID and random systems (red and blue markers respectively); Bottom figures: global coherence of UPSID systems and of the systems they became after a 50-step evolution (blue and purple markers res- pectively). From left to right: 3 different S values - 0.008 (mixed model), 0.5 (deterministic) and 0.001 (stochastic).
• Both deterministic and stochastic models led to new systems that are not consis- tent with the actual ones in terms of global coherence; the intermediate model produced more consistent systems, at least for sizes up to 11 segments.
• The comparison of the 3 S values seems to indicate that the evolution of pho- nological inventories is primarily internally driven.
• Starting from specific systems, we studied the outcome of 500 50-step evolu- tions; results show that systems with high coherence tend to be preserved whe- reas systems with low coherence will systematically get transformed.
Conclusions & Perspectives
• The notions of coherence and stability used in our model seem to capture at least part of the forces driving evolution of phonological inventories.
• More qualitative studies of the outcomes of the model are needed to test the validity of the observed changes.
• The next step is to test predictions of the model with actual known evolutio- nary paths.
References
1. Liljencrants, J., Lindblom, B. 1972. Numerical simulation of vowel quality systems: the role of perceptual contrast. Language 48, 839-862.
2. Lindblom, B., MacNeilage, P., Studdert-Kennedy, M. 1983. Self-organizing processes and the explanation of phonological. In: Butterworth, B., Comrie, B., Dahl, Ö. (eds), Explanations for language universals. Berlin: Mouton, 181-203.
3. Lindblom, B., Maddieson, I. 1988. Phonetic universals in consonant systems. In: Li, C., Hyman, L. (eds), Language, Speech and mind. London: Routledge, 62-78.
4. Labov, W. 1994. Principles of Linguistic Change. Volume 1: Internal Factors. Oxford:Blackwell.
5. Labov, W. 2001. Principles of Linguistic change. Volume II: Social Factors. Oxford: Blackwell.
6. Coupé, C., Marsico, E. & Pellegrino, F. 2009. Structural complexity of phonological systems. In Pellegrino, F., Marsico, E., Chitoran, I. & Coupé, C. (eds), Approaches to phonological complexity, Phonology & Phonetics Series vol. 16, Ber- lin, New York, Mouton de Gruyter, pp. 141-169
7. Maddieson, I. 1984. Patterns of sounds. Cambridge: Cambridge University Press.
8. Maddieson, I., Precoda, K. 1990. Updating UPSID. UCLA Working Papers in Phonetics 74, 104-111.
An evolutionary model of phonological inventories based on synchronic data
Christophe Coupé, Egidio Marsico & François Pellegrino
Laboratoire Dynamique du Langage, CNRS - Université de Lyon; Institut Rhône-Alpin des systèmes complexes
Introduction
• Phonological inventories are the results of the interaction between internal (arti- culatory, perceptual and cognitive) and external (e.g. sociolinguistic) constraints ([1], [2], [3], [4], [5]).
• A system can be said to be coherent if it satisfies internal constraints [6].
• Internal constraints lead systems to diachronically stabilize around maximal co- herence; evolution accomodating only for these constraints is deterministic.
• External constraints can disrupt this evolution toward less coherent systems;
they bring ‘‘stochasticity’’.
• The synchronic frequency of distribution of an inventory reflects i) its response to internal constraints and ii) its positioning or not at the crossroad of evolu- tionary trajectories.
Goal
• Design an evolutionary model based on the synchronic distribution of phone- mes in actual inventories.
• Analyse inventories in the light of notions of coherence and stability by conside- ring frequencies of individual segments and pairs of segments.
• Assess the weight of external constraints.
Data & methodology
Data
• The genetically and geographically balanced UPSID database ([7], [8])
• 451 languages, 833 phonemes, 96 articulatory features Building a measure of coherence for inventories
• If inventories were randomly built, all segments would have a probability of occurrence of 0.5; the intrinsic strength of a segment relates to how much it de- viates from randomness.
• If segments behaved independently, the probability of 2 segments to co-occur would be equal to the product of their respective frequencies; the interactional strength of a pair of segments relates to how much it deviates from independence
• The log of the binomial test quantifies the previous deviations.
• For individual segment, the sign tells whether this segment is favored or disfa- vored in systems; for pairs of segments, it indicates whether these two elements
‘attract’ or ‘repulse’ each other.
• For any given system, we obtain its intrinsic strength by summing the intrinsic strengths of all present segments AND absent segments.
• We compute the interactional strength of the system by adding the interactional strengths of all pairs of present AND absent segments.
• The (global) coherence of a system results in the combination of these intrinsic and interactional strengths. As it takes into account all possible segments and not just those present, it is independent of the size of the system.
Deriving stability and evolutionary trajectories from coherence
• An index of stochasticity S models deterministic vs. stochastic evolutions.
• From a given inventory, we can compute a set of new inventories differing by a few units and rank them according to their coherence.
• A deterministic model (high S) will tend to maximize coherence by choosing the highest ranked output as the next evolutionary step.
• A stochastic one (low S), incorporates ‘external factors’ by assigning to each potential new system a probability of being chosen.
Contact: Christophe COUPE, christophe.coupe@ish-lyon.cnrs.fr Financial support: CNRS; ASLAN Laboratory of Excellence
LYON UNIVERSIT DE
Institut des systmes complexes Complex Systems Institute Rhne-Alpes
IXXI
/ã/ ! /ã/
/a/ 72 320
! /a/ 11 48
/ã/ ! /ã/
/a/ 82 310
! /a/ 1 58
/a/ 392
/ã/ 83
/a/ 225
/ã/ 225
Expected distribution Actual distribution Strengths
str(/a/)=140.4 str(/ã/)=-100.0 str(/a/&/ã/)=2.2 str(/ã/&!/a/)=-8.5
25000 27000 29000 31000 33000 35000 37000 39000 41000 43000 45000
0 5 10 15 20 25 30 35
UPSID Random systems
Intrinsic coherence
-4500 -4000 -3500 -3000 -2500 -2000 -1500 -1000 -500 0
0 5 10 15 20 25 30 35
UPSID Random systems
Interactional coherence
23000 25000 27000 29000 31000 33000 35000 37000 39000 41000
0 5 10 15 20 25 30 35
UPSID Random systems Most frequent
Total coherence
31000 32000 33000 34000 35000 36000 37000 38000 39000 40000
0 5 10 15 20 25 30
UPSID S=0.008
Total coherence
0 5000 10000 15000 20000 25000 30000 35000 40000 45000
0 10 20 30 40 50 60 70
UPSID S=0.001
Total coherence
31000 32000 33000 34000 35000 36000 37000 38000 39000 40000
0 5 10 15 20 25 30
UPSID S=0.5
Total coherence
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
0 5 10 15 20 25 30 35
Evolved UPSID Random systems
Stability
Stability 0,001
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
0 5 10 15 20 25 30 35
Evolved UPSID Random systems
Stability Stability 0,5
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
0 5 10 15 20 25 30 35
Evolved UPSID Random systems
Stability
Stability 0,008
