theories and models of language change
TRANSCRIPT
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Theories and Models of Language ChangeSession 5: Models I - Social Factors and Structure
Roland Mühlenbernd
May 19, 2015
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Review: Universal Darwinism
Mechanisms of universal evolution:
1. variation: continuing abundance of different elements
2. selection : number/probability of copies of elements -depending on interaction between element features andenvironmental features
3. replication: reproduction/copying of elements
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Review: Evol. Approaches to Language Change
Evolution and Language Change:
1. variation: how come new variants to a speechcommunity?
I innovation/actuationI language contact: borrowing, structure primingI transformation of already existing material:
recombination, reanalysis, adaptation2. selection : how are (competing) variants selected
I substitution vs differential replicationI selection vs evolutionary driftI social vs functional factors
3. replication: how do linguistic variants reproduce?I types vs tokensI replication units and replicatorsI mechanism of replication: e.g. social learning,
imitation, priming
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Language Change - Broad and Narrow Sense
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Methods
What is the Scientific Method?I research, data acquisitionI theorizeI testI evaluate
TheoryField Work,Experiments
Comp. Model,Simulation
structure, behaviour
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Language Change: Linguistic Items
”Languages are evolving populations of linguistic items.”(Nettle, 1999)
I isolatable elements of phonological and grammaticalstructure
I that can be individually learned and changed in languagehistory
I cultural traits, linguistic memes, linguemes
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Exercise I
What is according to Nettle (1999) the source of variation inlanguage-related cultural evolution?
I imperfect learning√
I language contactI speaker innovation
The source of variation in cultural evolution modelsis of the imperfect-learning type, and for convenienceI will adopt the same strategy here...
David Nettle (1999): Using Social Impact Theory to simulatelanguage change
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Biological, Cultural and Linguistic Evolution
I in biological evolution an individual inherits genes fromits parents with even probability getting the variant ofmother or father
I cultural evolution models assume infinite populationswithout particular structure (mean field)
I in linguistic evolutionI an individual ”inherits genes” (acquires linguistic items)
from a wider range of sourcesI the speech communities are finite and socially and
spatially structuredI new variants cannot be expected to cause change as the
next generation will simply discount it in favor of the morecommon old pendant (threshold problem)
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
The Threshold Problem (Exercise II)
What is Nettle’s description of the ’threshold problem’?I In language change a new variant can only spread, if the
innovator of it passes a threshold of social status andinfluence, and this threshold is unrealistically high.
I In language change the new variant can only spreadsociety-wide, if the the degree of ’locality’ or’parochialism’ of the social structure passes a particularthreshold.
I In language change new variants can only establish if theypass the threshold of frequency, which in the beginningthey never have.
√
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
The Threshold Problem
I if the learner always adopts the common norm, the resultafter a few generations is homogeneity
I if the learner is biased towards new variants, those have achance of overcoming the threshold
I two possibilities of such a bias:1. social: influential people are bearers of new variants2. functional: new variant has attribute that makes it easier to
acquire or use
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Exercise III
What is according to Nettle (1999) the principal feature thatdefines the functional bias of a linguistic item?
I learnability√
I brevityI specificity
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Social Impact Theory
Social Impact Theory (Latane 1981): The likelihood that aperson will respond to social influence will increase with:
I Strength: how important the influencing group of peopleare to you.
I Immediacy: how close the group are to you (in space andtime) at the time of the attempted influence.
I Number: How many people there are in the group.
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Social Impact Theory
I Idea: a speaker will adopt the variant with the greatestimpact on him
I the impact of variant p is influenced by social andfunctional parameters: Ip = bp × f (Sp,Dp,Np), whereby
I bp represents acquisitional bias for using pI Np is the number of individuals using p (number)I Sp is the status of a individual using p (strength)I Dp is the social distance to a individual using p
(immediacy)
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Social Impact TheoryNettle’s formula:
Ip = bp × Nap ×
∑i si × 1
d2i
Np
0 1 2 3 4 5 6 7 80
0.2
0.4
0.6
0.8
1.0
di
1d2
0 1 2 3 4 5 6 7 80
1
2
3
4
Np
Nap
a = .7a = .5a = .3
I Why 1/d2? Answer: established in SIT, apparently by analogywith gravitation laws, where effects decrease with the square ofdistance
I Why exponent a? Answer: The persuasiveness of a belief doesnot rise linearly with the number of people holding it.
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Exercise IV
Ip = bp × Nap ×
∑i si × 1
d2i
Np
Given the ’Social Impact Theory’ formula (3) (on page 102) forthe impact of a variant p. Given the following values:
I variant p has a functional bias b of 0.5I the parameter a is 1I the language learner X is influenced by 3 persons using
variant p:I social status 5, social distance 2I social status 3, social distance 2I social status 8, social distance 4
What is the impact value of variant p on language learner X?
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Simulation Models
A simulation model
I takes the main relationships of real world situations andexplore effects across a range of conditions
I can cover complex iterated processes that cannot bepredicted by simple thought of deduction
I helps to explore under what assumptions and initialconditions does the system behave in the desired way
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
The Simulation ModelNettle’s social map:
5 4 3 2 1 1 2 3 4 5 5 4 3 2 15 4 3 2 1 1 2 3 4 5 5 4 3 2 15 4 3 2 1 1 2 3 4 5 5 4 3 2 1
I horizontal: familial social ties (parents, children...)I vertical: horiz. social ties (own peer group, friends...)I In each life stage (simulation step) a whole cohort along
the vertical goes through language acquisitionI age 1-3: for possible variants p and q compute Ip and Iq;
acquire variant with higher impact valueI age 1 has no influenceI age 4-5: stay at acquired variantI age 1-2: 5% mutation, due to imperfect learning
I After each life stage: increase ages by one and replace age5 by age 1 (replace dead by newborn agents)
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
The Simulation Model
Nettle’s toroidal social map:
5 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 55 4 3 2 1 1 2 3 4 5 5 4 3 2 1 1 2 3 4 5
Table: Nettle’s 20× 20 social map with initial age structure.
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Simulation ResultsParameter: ∀i, j : si = sj, a = 1
Percentage of individuals having p over 200 lifestages Simulation output after 100 lifestages
Parameter: ∀i, j : si = sj, a < .5
Percentage of individuals having p over 200 lifestages Simulation output after 100 lifestages
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Simulation ResultsChanging the effect of distance:
Ip = bp × N0.8p ×
∑i si × 1
d4i
Np
Percentage of individuals having p over 400 lifestages, with the variable of distance raised to the power 4 and threedifferent values of mutation rate.
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Simulation Results
Note:I a < 0.5 or distance raise to power 4 both pushed the
system to a 50/50 equilibrium (socially specializedvariants / differential replication)
I variable a > 0.5 is desirable to achieve communityconsensus
I the induced parochialism implied by raising power of thedistance variable reduces SIM to a cultural evolutionmodel with small number of cultural parents→ power 2 isrealistic assumption
Under what circumstances can the new variant replace the oldone totally? (threshold problem)
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Exercise 3
Nettle is discussing the distribution of features in a population.He mentions that many attributes are normally distributed(Gaussian distribution), but some are distributed according to aPoisson curve, like...
I weightI wealth
√
I height
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Simulation Results
Testing Social Selection:I A: different values of status, according to a Poisson curveI B: additional hyperinfluential agents (s = 100 with
probability .025)I result: with superinfluential agents the new variant
occasionally replaces the established oneagents with s
0.1
0.2
0.3
0.4
0.5
1-4 5-8 9-12 13-16 17-20s
Poisson distribution for status groups Percentage of individ. having p over 400 lifestages
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Simulation Results
Testing Social Selection:I different functional biases for new variant: 0.5, 1 and 2.0I result: only for similar biases variants replace each other -
but how similar?I note: functional bias has only impact in combination with
social selection!
Percentage of individuals having p over 400lifestages for different biases of q
Mean percentage having p averaged over 400lifestages for different biases of q
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Overall ResultsNettle’s conclusion:
I Emergence of separate dialects by decreasing theimportance learners give to conformity (decreasingparameter a, increasing power of distance factor)
I Changes are adopted if there are speakers which are muchmore influential than others (superinfluential agents)
I Functional biases determine the fate of competinglinguistic variants (bq >> 1)
Exercise 6:
“Without the potential for change provided by differences insocial influence, functionally favoured variants might neverovercome the threshold required to displace prior norms.Functional biases may therefore affect the direction of lan-guage change rather than providing sufficient conditions forit to occur.”
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Further ’Spatial’ Simulation Models
Language change and social networks (Ke, Gong, Wang 2008)
I more realistic network structures: small world structuresI adoption of Nettle’s age structureI simplification of SIT formula: F(U) = fu × qu
(functional bias and frequency, no status)I result: change can happen without superinfluential agents
Center and peripheries: Network roles in language change(Fagyal, Swarup, Escobar, Gasser, Lakkaraju 2010)
I directed networks (scale-free small world structures)I more than two competing variants (probabilistic adoption)I agents’ influence value (status)I result: i) peripheral low-connected agents preserve;
ii) central influential agents induce propagations
RolandMühlenbernd
Review: Evolut.Approaches toLanguage Change
A Simulation Modelof Language ChangeLinguistic Items
The Threshold Problem
Social Impact Theory
The Simulation Model
Simulation Results
Further ’Spatial’Simulation Models
Homeworks
Homeworks
I Read the article ‘Self-organization in vowel systems’ (deBoer, 2000)
I solve the appropriate exercises given on ILIAS