1
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Optimality Theoretic Pragmatics
Maria AloniReinhard Blutner
University of Amsterdam
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Part I Optimality Theoretic Pragmatics
The Cognitive Grounding
Part IIFossilized Pragmatics
Focus and accent
2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Outline (Part I)
1 Grice and his Followers Global and Local Theories of Pragmatics
2 Cognitive Motivation of Symmetryand Bidirection
3 The Idea of Fossilization
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
1Grice and his Followers
Global and Local Theories of Pragmatics
3
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Grice (1975)
Normative Stance Naturalistic Stance
Neo-GriceanTheories (Horn Atlas)
Relevance Theory
Presumptive Meanings
OT-Pragmatics
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Rejecting the doctrin of literalmeaning
Semantic Underdetermination (Atlas)ndash Linguistic meanings underdetermine the truth-
conditional contentndash Ambiguity vs semantic non-specificity
Contextualism ndash the suggestion that the mechanism of pragmatic
interpretation is crucial both for determining what the speaker says and what he means
Linguistic meaning ne what is said
4
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Global and local theories of NL interpretation
bull Global theories have a holistic character Often they refer to normative theories They are problematic as incremental processing models
bull Local theories can account for an incremental interpretation mechanism Often they refer to compositional automatized projection routines
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zeevat (discourse particles)
Van der Sandt Geurts
Presuppos
Neo-Gricean theories
Pragmatics
Relevance Theory Levinson 2000 Chierchia
Optimality Theoretic
Implicature
Early structuralism amp lexical field theoriesBidirectional optimi-zation
Montague semantics
Interpretive optimization (Hendriks amp de Hoop)
Semantics
OT syntaxconnectionism
Traditional generative syntax
Syntax
GlobalLocal
5
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Relevance Theory
Communicative Principle of Relevance
Utterances convey a presumption of their own optimal relevance
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
What does it mean
Any given utterance can be presumed
ndash to be at least relevant enough to warrant the addresseersquos processing effort
ndash to be the most relevant one compatible with the speakerrsquos current state of knowledge and her personal preferences and goals
6
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Procedure
ndash test possible inter-pretations in their order of accessibility
ndash stop once the expectation of opti-mal relevance is satisfied
[EFFORT]
[EFFECT]
1234567
EFFORTEFFECT
middotmiddot
(cf Sperber Cara amp Girotto 1995 95)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Optimal InterpretationHendriks amp de Hoop The integration of pragmatic and syntacticsemantic informationin a system of ranked constraints in order to correctly derive the optimal interpretations
Suggestion by RT EFFECT gtgt EFFORT
ZeevatFaith gtgt Consistence gtgt Do not accommodate gtgt Strength
7
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Neo-Gricean Theories (Atlas Horn)
bull The Q-Principle (Hearer-based)ndash Say as much as you can (modulo R)
(Gricersquos first quantity maxim and the first two manner maxims)
bull The R-Principle (Speaker-based)ndash Say not more than you must (modulo Q)
(Gricersquos second quantity maxim relation maxim and the second two manner maxims)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak Bidirectionalitylangf mrang is weakly optimal (= super-optimal) iff
a langf mrang isin GENb there is no weakly optimal langfrsquo mrang isin GEN
such that langfrsquo mrang gt langf mrangc there is no weakly optimal langf mrsquo rang isin GEN
such that langf mrsquo rang gt langf mrang
This is an abstract scheme (Jaumlger 2002) The content of the ordering relation ldquogtrdquo is
determined by the system of constraints
8
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Example
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
siMsiF
siF prefer short formssiM prefer stereotypical meanings
f1
f2
m1 m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Levinsonlsquos Presumptive Meaningsbull Presumptive meanings are a matter of
preferred interpretation calculated by a particular default mechanism
bull Presumptive meanings are localbull Three heuristics
ndash Q-heuristic What isnrsquot said is not the casendash I-heuristic What is expressed simply is
stereotypically exemplifiedndash M-heuristic Whatrsquos said in an abnormal way
isnrsquot normal
9
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
What are the heuristics
siM
FrarrM
FrarrM
FrarrM
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
FrarrMsiF
siF prefer short formssiM prefer stereotypical meanings
I-heuristic What is expressed simply is stereotypically exemplifiedM-heuristic Whatrsquos said in an abnormal way isnrsquot normal
I-heur
M-heurAnti-I
Anti-M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
ComparisonNeo-Griceanglobal theory
bidirection optimizationsiFsiM
f1
f2
m1 m2f1
f2
m1 m2
Levinson (2000)local theory
unidirectional optimizationI M
10
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Conclusions
Relevance TheoryLevinson (2000)
Neo-GriceanTheory
Unidirectional OTBidirectional OT
Local TheoriesGlobal Theories
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
2Cognitive motivation of
symmetry and bidirection
11
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetric OT-systemsbull Symmetric system If f m (optimal
interpretation) then m f (optimal expression)
bull For symmetric systems unidirectionaloptimization gives the same solutions as bidirectional optimization and vice versa
f1f2
m1m2
f1f2
m1m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The symmetry testbull A set of pairs of patterns (Ai Bi) are
repeatedly presented When one member of the pair is presented the subject has to learn to produce the other Assume a 1-1 correspondence between A and B
bull If subjects are qualified to match Stimulus A to B and then without further training match B to A they have passed a test of symmetry
lop
raf
kas
lop
raf
kas
12
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Findings
bull Children as young as 2 years pass the symmetry test (Green 1990)
bull Chimps do not show symmetry (see Savage-Rumbaugh 1984)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetry between languageproduction and comprehension
bull Normally we can understand the sentences we producendash Exceptions are very rare
bull In most cases we can produce the sentences we understandndash Typical exceptions in language acquisition Childrenlsquos
ability in production lags dramatically behind their ability in comprehension
13
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Two ways of deriving symmetrybull Symmetry as a result of the
network architecturendash In symmetric networks one
and the same pattern can be produced starting from different inputs
bull Symmetry as a result of learning ndash Tesarrsquos and Smolenskyrsquos
learning theory
A rarr BA larr B
Difference A-A triggers learning
∆
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
14
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak bidirection and symmetrybull Rather than seeing weak bidirection as a online
interpretationproduction mechanism it shouldbe understood in terms of (iterated) learning(resulting in symmetric OT systems)
bull Suggestionndash Unidirectional OT (local theories) hArr Synchronic perspective
ndash (Weak) bidirection OT (global theories) hArr Diachronic perspective
15
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zipf 1949bull Two basic and competing forces
ndash Speakerrsquos economy Force of unification Rndash Hearerrsquos economy Force of diversification Q
bull The two opposing economies are evolutionary forces ie they are balanced during language evolutionndash Languages are evolving via cultural rather than
biological transmission on a historical rather than genetic timescale
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
3The idea of fossilization
16
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
FossilizationGlobal Theories
bidirectional optimization
f1
f2
m1 m2f1
f2
m1 m2
Local Theoriesunidirectional optimization
Fossilization
Markedness constraintssiFsiM
Linking constraintsI M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
(Iterated) Learning
m f mrsquoSpeaker Hearer
m = mrsquo
If yes nothing happens
If no adjustment- All constraints that favour (f m) over (f mrsquo) are promoted- All constraints that favour (f mrsquo) over (f m) are demoted
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Outline (Part I)
1 Grice and his Followers Global and Local Theories of Pragmatics
2 Cognitive Motivation of Symmetryand Bidirection
3 The Idea of Fossilization
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
1Grice and his Followers
Global and Local Theories of Pragmatics
3
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Grice (1975)
Normative Stance Naturalistic Stance
Neo-GriceanTheories (Horn Atlas)
Relevance Theory
Presumptive Meanings
OT-Pragmatics
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Rejecting the doctrin of literalmeaning
Semantic Underdetermination (Atlas)ndash Linguistic meanings underdetermine the truth-
conditional contentndash Ambiguity vs semantic non-specificity
Contextualism ndash the suggestion that the mechanism of pragmatic
interpretation is crucial both for determining what the speaker says and what he means
Linguistic meaning ne what is said
4
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Global and local theories of NL interpretation
bull Global theories have a holistic character Often they refer to normative theories They are problematic as incremental processing models
bull Local theories can account for an incremental interpretation mechanism Often they refer to compositional automatized projection routines
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zeevat (discourse particles)
Van der Sandt Geurts
Presuppos
Neo-Gricean theories
Pragmatics
Relevance Theory Levinson 2000 Chierchia
Optimality Theoretic
Implicature
Early structuralism amp lexical field theoriesBidirectional optimi-zation
Montague semantics
Interpretive optimization (Hendriks amp de Hoop)
Semantics
OT syntaxconnectionism
Traditional generative syntax
Syntax
GlobalLocal
5
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Relevance Theory
Communicative Principle of Relevance
Utterances convey a presumption of their own optimal relevance
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
What does it mean
Any given utterance can be presumed
ndash to be at least relevant enough to warrant the addresseersquos processing effort
ndash to be the most relevant one compatible with the speakerrsquos current state of knowledge and her personal preferences and goals
6
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Procedure
ndash test possible inter-pretations in their order of accessibility
ndash stop once the expectation of opti-mal relevance is satisfied
[EFFORT]
[EFFECT]
1234567
EFFORTEFFECT
middotmiddot
(cf Sperber Cara amp Girotto 1995 95)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Optimal InterpretationHendriks amp de Hoop The integration of pragmatic and syntacticsemantic informationin a system of ranked constraints in order to correctly derive the optimal interpretations
Suggestion by RT EFFECT gtgt EFFORT
ZeevatFaith gtgt Consistence gtgt Do not accommodate gtgt Strength
7
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Neo-Gricean Theories (Atlas Horn)
bull The Q-Principle (Hearer-based)ndash Say as much as you can (modulo R)
(Gricersquos first quantity maxim and the first two manner maxims)
bull The R-Principle (Speaker-based)ndash Say not more than you must (modulo Q)
(Gricersquos second quantity maxim relation maxim and the second two manner maxims)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak Bidirectionalitylangf mrang is weakly optimal (= super-optimal) iff
a langf mrang isin GENb there is no weakly optimal langfrsquo mrang isin GEN
such that langfrsquo mrang gt langf mrangc there is no weakly optimal langf mrsquo rang isin GEN
such that langf mrsquo rang gt langf mrang
This is an abstract scheme (Jaumlger 2002) The content of the ordering relation ldquogtrdquo is
determined by the system of constraints
8
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Example
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
siMsiF
siF prefer short formssiM prefer stereotypical meanings
f1
f2
m1 m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Levinsonlsquos Presumptive Meaningsbull Presumptive meanings are a matter of
preferred interpretation calculated by a particular default mechanism
bull Presumptive meanings are localbull Three heuristics
ndash Q-heuristic What isnrsquot said is not the casendash I-heuristic What is expressed simply is
stereotypically exemplifiedndash M-heuristic Whatrsquos said in an abnormal way
isnrsquot normal
9
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
What are the heuristics
siM
FrarrM
FrarrM
FrarrM
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
FrarrMsiF
siF prefer short formssiM prefer stereotypical meanings
I-heuristic What is expressed simply is stereotypically exemplifiedM-heuristic Whatrsquos said in an abnormal way isnrsquot normal
I-heur
M-heurAnti-I
Anti-M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
ComparisonNeo-Griceanglobal theory
bidirection optimizationsiFsiM
f1
f2
m1 m2f1
f2
m1 m2
Levinson (2000)local theory
unidirectional optimizationI M
10
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Conclusions
Relevance TheoryLevinson (2000)
Neo-GriceanTheory
Unidirectional OTBidirectional OT
Local TheoriesGlobal Theories
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
2Cognitive motivation of
symmetry and bidirection
11
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetric OT-systemsbull Symmetric system If f m (optimal
interpretation) then m f (optimal expression)
bull For symmetric systems unidirectionaloptimization gives the same solutions as bidirectional optimization and vice versa
f1f2
m1m2
f1f2
m1m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The symmetry testbull A set of pairs of patterns (Ai Bi) are
repeatedly presented When one member of the pair is presented the subject has to learn to produce the other Assume a 1-1 correspondence between A and B
bull If subjects are qualified to match Stimulus A to B and then without further training match B to A they have passed a test of symmetry
lop
raf
kas
lop
raf
kas
12
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Findings
bull Children as young as 2 years pass the symmetry test (Green 1990)
bull Chimps do not show symmetry (see Savage-Rumbaugh 1984)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetry between languageproduction and comprehension
bull Normally we can understand the sentences we producendash Exceptions are very rare
bull In most cases we can produce the sentences we understandndash Typical exceptions in language acquisition Childrenlsquos
ability in production lags dramatically behind their ability in comprehension
13
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Two ways of deriving symmetrybull Symmetry as a result of the
network architecturendash In symmetric networks one
and the same pattern can be produced starting from different inputs
bull Symmetry as a result of learning ndash Tesarrsquos and Smolenskyrsquos
learning theory
A rarr BA larr B
Difference A-A triggers learning
∆
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
14
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak bidirection and symmetrybull Rather than seeing weak bidirection as a online
interpretationproduction mechanism it shouldbe understood in terms of (iterated) learning(resulting in symmetric OT systems)
bull Suggestionndash Unidirectional OT (local theories) hArr Synchronic perspective
ndash (Weak) bidirection OT (global theories) hArr Diachronic perspective
15
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zipf 1949bull Two basic and competing forces
ndash Speakerrsquos economy Force of unification Rndash Hearerrsquos economy Force of diversification Q
bull The two opposing economies are evolutionary forces ie they are balanced during language evolutionndash Languages are evolving via cultural rather than
biological transmission on a historical rather than genetic timescale
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
3The idea of fossilization
16
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
FossilizationGlobal Theories
bidirectional optimization
f1
f2
m1 m2f1
f2
m1 m2
Local Theoriesunidirectional optimization
Fossilization
Markedness constraintssiFsiM
Linking constraintsI M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
(Iterated) Learning
m f mrsquoSpeaker Hearer
m = mrsquo
If yes nothing happens
If no adjustment- All constraints that favour (f m) over (f mrsquo) are promoted- All constraints that favour (f mrsquo) over (f m) are demoted
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
3
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Grice (1975)
Normative Stance Naturalistic Stance
Neo-GriceanTheories (Horn Atlas)
Relevance Theory
Presumptive Meanings
OT-Pragmatics
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Rejecting the doctrin of literalmeaning
Semantic Underdetermination (Atlas)ndash Linguistic meanings underdetermine the truth-
conditional contentndash Ambiguity vs semantic non-specificity
Contextualism ndash the suggestion that the mechanism of pragmatic
interpretation is crucial both for determining what the speaker says and what he means
Linguistic meaning ne what is said
4
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Global and local theories of NL interpretation
bull Global theories have a holistic character Often they refer to normative theories They are problematic as incremental processing models
bull Local theories can account for an incremental interpretation mechanism Often they refer to compositional automatized projection routines
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zeevat (discourse particles)
Van der Sandt Geurts
Presuppos
Neo-Gricean theories
Pragmatics
Relevance Theory Levinson 2000 Chierchia
Optimality Theoretic
Implicature
Early structuralism amp lexical field theoriesBidirectional optimi-zation
Montague semantics
Interpretive optimization (Hendriks amp de Hoop)
Semantics
OT syntaxconnectionism
Traditional generative syntax
Syntax
GlobalLocal
5
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Relevance Theory
Communicative Principle of Relevance
Utterances convey a presumption of their own optimal relevance
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
What does it mean
Any given utterance can be presumed
ndash to be at least relevant enough to warrant the addresseersquos processing effort
ndash to be the most relevant one compatible with the speakerrsquos current state of knowledge and her personal preferences and goals
6
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Procedure
ndash test possible inter-pretations in their order of accessibility
ndash stop once the expectation of opti-mal relevance is satisfied
[EFFORT]
[EFFECT]
1234567
EFFORTEFFECT
middotmiddot
(cf Sperber Cara amp Girotto 1995 95)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Optimal InterpretationHendriks amp de Hoop The integration of pragmatic and syntacticsemantic informationin a system of ranked constraints in order to correctly derive the optimal interpretations
Suggestion by RT EFFECT gtgt EFFORT
ZeevatFaith gtgt Consistence gtgt Do not accommodate gtgt Strength
7
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Neo-Gricean Theories (Atlas Horn)
bull The Q-Principle (Hearer-based)ndash Say as much as you can (modulo R)
(Gricersquos first quantity maxim and the first two manner maxims)
bull The R-Principle (Speaker-based)ndash Say not more than you must (modulo Q)
(Gricersquos second quantity maxim relation maxim and the second two manner maxims)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak Bidirectionalitylangf mrang is weakly optimal (= super-optimal) iff
a langf mrang isin GENb there is no weakly optimal langfrsquo mrang isin GEN
such that langfrsquo mrang gt langf mrangc there is no weakly optimal langf mrsquo rang isin GEN
such that langf mrsquo rang gt langf mrang
This is an abstract scheme (Jaumlger 2002) The content of the ordering relation ldquogtrdquo is
determined by the system of constraints
8
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Example
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
siMsiF
siF prefer short formssiM prefer stereotypical meanings
f1
f2
m1 m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Levinsonlsquos Presumptive Meaningsbull Presumptive meanings are a matter of
preferred interpretation calculated by a particular default mechanism
bull Presumptive meanings are localbull Three heuristics
ndash Q-heuristic What isnrsquot said is not the casendash I-heuristic What is expressed simply is
stereotypically exemplifiedndash M-heuristic Whatrsquos said in an abnormal way
isnrsquot normal
9
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
What are the heuristics
siM
FrarrM
FrarrM
FrarrM
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
FrarrMsiF
siF prefer short formssiM prefer stereotypical meanings
I-heuristic What is expressed simply is stereotypically exemplifiedM-heuristic Whatrsquos said in an abnormal way isnrsquot normal
I-heur
M-heurAnti-I
Anti-M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
ComparisonNeo-Griceanglobal theory
bidirection optimizationsiFsiM
f1
f2
m1 m2f1
f2
m1 m2
Levinson (2000)local theory
unidirectional optimizationI M
10
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Conclusions
Relevance TheoryLevinson (2000)
Neo-GriceanTheory
Unidirectional OTBidirectional OT
Local TheoriesGlobal Theories
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
2Cognitive motivation of
symmetry and bidirection
11
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetric OT-systemsbull Symmetric system If f m (optimal
interpretation) then m f (optimal expression)
bull For symmetric systems unidirectionaloptimization gives the same solutions as bidirectional optimization and vice versa
f1f2
m1m2
f1f2
m1m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The symmetry testbull A set of pairs of patterns (Ai Bi) are
repeatedly presented When one member of the pair is presented the subject has to learn to produce the other Assume a 1-1 correspondence between A and B
bull If subjects are qualified to match Stimulus A to B and then without further training match B to A they have passed a test of symmetry
lop
raf
kas
lop
raf
kas
12
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Findings
bull Children as young as 2 years pass the symmetry test (Green 1990)
bull Chimps do not show symmetry (see Savage-Rumbaugh 1984)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetry between languageproduction and comprehension
bull Normally we can understand the sentences we producendash Exceptions are very rare
bull In most cases we can produce the sentences we understandndash Typical exceptions in language acquisition Childrenlsquos
ability in production lags dramatically behind their ability in comprehension
13
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Two ways of deriving symmetrybull Symmetry as a result of the
network architecturendash In symmetric networks one
and the same pattern can be produced starting from different inputs
bull Symmetry as a result of learning ndash Tesarrsquos and Smolenskyrsquos
learning theory
A rarr BA larr B
Difference A-A triggers learning
∆
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
14
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak bidirection and symmetrybull Rather than seeing weak bidirection as a online
interpretationproduction mechanism it shouldbe understood in terms of (iterated) learning(resulting in symmetric OT systems)
bull Suggestionndash Unidirectional OT (local theories) hArr Synchronic perspective
ndash (Weak) bidirection OT (global theories) hArr Diachronic perspective
15
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zipf 1949bull Two basic and competing forces
ndash Speakerrsquos economy Force of unification Rndash Hearerrsquos economy Force of diversification Q
bull The two opposing economies are evolutionary forces ie they are balanced during language evolutionndash Languages are evolving via cultural rather than
biological transmission on a historical rather than genetic timescale
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
3The idea of fossilization
16
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
FossilizationGlobal Theories
bidirectional optimization
f1
f2
m1 m2f1
f2
m1 m2
Local Theoriesunidirectional optimization
Fossilization
Markedness constraintssiFsiM
Linking constraintsI M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
(Iterated) Learning
m f mrsquoSpeaker Hearer
m = mrsquo
If yes nothing happens
If no adjustment- All constraints that favour (f m) over (f mrsquo) are promoted- All constraints that favour (f mrsquo) over (f m) are demoted
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
4
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Global and local theories of NL interpretation
bull Global theories have a holistic character Often they refer to normative theories They are problematic as incremental processing models
bull Local theories can account for an incremental interpretation mechanism Often they refer to compositional automatized projection routines
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zeevat (discourse particles)
Van der Sandt Geurts
Presuppos
Neo-Gricean theories
Pragmatics
Relevance Theory Levinson 2000 Chierchia
Optimality Theoretic
Implicature
Early structuralism amp lexical field theoriesBidirectional optimi-zation
Montague semantics
Interpretive optimization (Hendriks amp de Hoop)
Semantics
OT syntaxconnectionism
Traditional generative syntax
Syntax
GlobalLocal
5
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Relevance Theory
Communicative Principle of Relevance
Utterances convey a presumption of their own optimal relevance
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
What does it mean
Any given utterance can be presumed
ndash to be at least relevant enough to warrant the addresseersquos processing effort
ndash to be the most relevant one compatible with the speakerrsquos current state of knowledge and her personal preferences and goals
6
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Procedure
ndash test possible inter-pretations in their order of accessibility
ndash stop once the expectation of opti-mal relevance is satisfied
[EFFORT]
[EFFECT]
1234567
EFFORTEFFECT
middotmiddot
(cf Sperber Cara amp Girotto 1995 95)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Optimal InterpretationHendriks amp de Hoop The integration of pragmatic and syntacticsemantic informationin a system of ranked constraints in order to correctly derive the optimal interpretations
Suggestion by RT EFFECT gtgt EFFORT
ZeevatFaith gtgt Consistence gtgt Do not accommodate gtgt Strength
7
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Neo-Gricean Theories (Atlas Horn)
bull The Q-Principle (Hearer-based)ndash Say as much as you can (modulo R)
(Gricersquos first quantity maxim and the first two manner maxims)
bull The R-Principle (Speaker-based)ndash Say not more than you must (modulo Q)
(Gricersquos second quantity maxim relation maxim and the second two manner maxims)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak Bidirectionalitylangf mrang is weakly optimal (= super-optimal) iff
a langf mrang isin GENb there is no weakly optimal langfrsquo mrang isin GEN
such that langfrsquo mrang gt langf mrangc there is no weakly optimal langf mrsquo rang isin GEN
such that langf mrsquo rang gt langf mrang
This is an abstract scheme (Jaumlger 2002) The content of the ordering relation ldquogtrdquo is
determined by the system of constraints
8
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Example
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
siMsiF
siF prefer short formssiM prefer stereotypical meanings
f1
f2
m1 m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Levinsonlsquos Presumptive Meaningsbull Presumptive meanings are a matter of
preferred interpretation calculated by a particular default mechanism
bull Presumptive meanings are localbull Three heuristics
ndash Q-heuristic What isnrsquot said is not the casendash I-heuristic What is expressed simply is
stereotypically exemplifiedndash M-heuristic Whatrsquos said in an abnormal way
isnrsquot normal
9
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
What are the heuristics
siM
FrarrM
FrarrM
FrarrM
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
FrarrMsiF
siF prefer short formssiM prefer stereotypical meanings
I-heuristic What is expressed simply is stereotypically exemplifiedM-heuristic Whatrsquos said in an abnormal way isnrsquot normal
I-heur
M-heurAnti-I
Anti-M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
ComparisonNeo-Griceanglobal theory
bidirection optimizationsiFsiM
f1
f2
m1 m2f1
f2
m1 m2
Levinson (2000)local theory
unidirectional optimizationI M
10
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Conclusions
Relevance TheoryLevinson (2000)
Neo-GriceanTheory
Unidirectional OTBidirectional OT
Local TheoriesGlobal Theories
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
2Cognitive motivation of
symmetry and bidirection
11
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetric OT-systemsbull Symmetric system If f m (optimal
interpretation) then m f (optimal expression)
bull For symmetric systems unidirectionaloptimization gives the same solutions as bidirectional optimization and vice versa
f1f2
m1m2
f1f2
m1m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The symmetry testbull A set of pairs of patterns (Ai Bi) are
repeatedly presented When one member of the pair is presented the subject has to learn to produce the other Assume a 1-1 correspondence between A and B
bull If subjects are qualified to match Stimulus A to B and then without further training match B to A they have passed a test of symmetry
lop
raf
kas
lop
raf
kas
12
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Findings
bull Children as young as 2 years pass the symmetry test (Green 1990)
bull Chimps do not show symmetry (see Savage-Rumbaugh 1984)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetry between languageproduction and comprehension
bull Normally we can understand the sentences we producendash Exceptions are very rare
bull In most cases we can produce the sentences we understandndash Typical exceptions in language acquisition Childrenlsquos
ability in production lags dramatically behind their ability in comprehension
13
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Two ways of deriving symmetrybull Symmetry as a result of the
network architecturendash In symmetric networks one
and the same pattern can be produced starting from different inputs
bull Symmetry as a result of learning ndash Tesarrsquos and Smolenskyrsquos
learning theory
A rarr BA larr B
Difference A-A triggers learning
∆
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
14
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak bidirection and symmetrybull Rather than seeing weak bidirection as a online
interpretationproduction mechanism it shouldbe understood in terms of (iterated) learning(resulting in symmetric OT systems)
bull Suggestionndash Unidirectional OT (local theories) hArr Synchronic perspective
ndash (Weak) bidirection OT (global theories) hArr Diachronic perspective
15
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zipf 1949bull Two basic and competing forces
ndash Speakerrsquos economy Force of unification Rndash Hearerrsquos economy Force of diversification Q
bull The two opposing economies are evolutionary forces ie they are balanced during language evolutionndash Languages are evolving via cultural rather than
biological transmission on a historical rather than genetic timescale
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
3The idea of fossilization
16
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
FossilizationGlobal Theories
bidirectional optimization
f1
f2
m1 m2f1
f2
m1 m2
Local Theoriesunidirectional optimization
Fossilization
Markedness constraintssiFsiM
Linking constraintsI M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
(Iterated) Learning
m f mrsquoSpeaker Hearer
m = mrsquo
If yes nothing happens
If no adjustment- All constraints that favour (f m) over (f mrsquo) are promoted- All constraints that favour (f mrsquo) over (f m) are demoted
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
5
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Relevance Theory
Communicative Principle of Relevance
Utterances convey a presumption of their own optimal relevance
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
What does it mean
Any given utterance can be presumed
ndash to be at least relevant enough to warrant the addresseersquos processing effort
ndash to be the most relevant one compatible with the speakerrsquos current state of knowledge and her personal preferences and goals
6
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Procedure
ndash test possible inter-pretations in their order of accessibility
ndash stop once the expectation of opti-mal relevance is satisfied
[EFFORT]
[EFFECT]
1234567
EFFORTEFFECT
middotmiddot
(cf Sperber Cara amp Girotto 1995 95)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Optimal InterpretationHendriks amp de Hoop The integration of pragmatic and syntacticsemantic informationin a system of ranked constraints in order to correctly derive the optimal interpretations
Suggestion by RT EFFECT gtgt EFFORT
ZeevatFaith gtgt Consistence gtgt Do not accommodate gtgt Strength
7
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Neo-Gricean Theories (Atlas Horn)
bull The Q-Principle (Hearer-based)ndash Say as much as you can (modulo R)
(Gricersquos first quantity maxim and the first two manner maxims)
bull The R-Principle (Speaker-based)ndash Say not more than you must (modulo Q)
(Gricersquos second quantity maxim relation maxim and the second two manner maxims)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak Bidirectionalitylangf mrang is weakly optimal (= super-optimal) iff
a langf mrang isin GENb there is no weakly optimal langfrsquo mrang isin GEN
such that langfrsquo mrang gt langf mrangc there is no weakly optimal langf mrsquo rang isin GEN
such that langf mrsquo rang gt langf mrang
This is an abstract scheme (Jaumlger 2002) The content of the ordering relation ldquogtrdquo is
determined by the system of constraints
8
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Example
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
siMsiF
siF prefer short formssiM prefer stereotypical meanings
f1
f2
m1 m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Levinsonlsquos Presumptive Meaningsbull Presumptive meanings are a matter of
preferred interpretation calculated by a particular default mechanism
bull Presumptive meanings are localbull Three heuristics
ndash Q-heuristic What isnrsquot said is not the casendash I-heuristic What is expressed simply is
stereotypically exemplifiedndash M-heuristic Whatrsquos said in an abnormal way
isnrsquot normal
9
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
What are the heuristics
siM
FrarrM
FrarrM
FrarrM
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
FrarrMsiF
siF prefer short formssiM prefer stereotypical meanings
I-heuristic What is expressed simply is stereotypically exemplifiedM-heuristic Whatrsquos said in an abnormal way isnrsquot normal
I-heur
M-heurAnti-I
Anti-M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
ComparisonNeo-Griceanglobal theory
bidirection optimizationsiFsiM
f1
f2
m1 m2f1
f2
m1 m2
Levinson (2000)local theory
unidirectional optimizationI M
10
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Conclusions
Relevance TheoryLevinson (2000)
Neo-GriceanTheory
Unidirectional OTBidirectional OT
Local TheoriesGlobal Theories
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
2Cognitive motivation of
symmetry and bidirection
11
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetric OT-systemsbull Symmetric system If f m (optimal
interpretation) then m f (optimal expression)
bull For symmetric systems unidirectionaloptimization gives the same solutions as bidirectional optimization and vice versa
f1f2
m1m2
f1f2
m1m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The symmetry testbull A set of pairs of patterns (Ai Bi) are
repeatedly presented When one member of the pair is presented the subject has to learn to produce the other Assume a 1-1 correspondence between A and B
bull If subjects are qualified to match Stimulus A to B and then without further training match B to A they have passed a test of symmetry
lop
raf
kas
lop
raf
kas
12
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Findings
bull Children as young as 2 years pass the symmetry test (Green 1990)
bull Chimps do not show symmetry (see Savage-Rumbaugh 1984)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetry between languageproduction and comprehension
bull Normally we can understand the sentences we producendash Exceptions are very rare
bull In most cases we can produce the sentences we understandndash Typical exceptions in language acquisition Childrenlsquos
ability in production lags dramatically behind their ability in comprehension
13
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Two ways of deriving symmetrybull Symmetry as a result of the
network architecturendash In symmetric networks one
and the same pattern can be produced starting from different inputs
bull Symmetry as a result of learning ndash Tesarrsquos and Smolenskyrsquos
learning theory
A rarr BA larr B
Difference A-A triggers learning
∆
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
14
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak bidirection and symmetrybull Rather than seeing weak bidirection as a online
interpretationproduction mechanism it shouldbe understood in terms of (iterated) learning(resulting in symmetric OT systems)
bull Suggestionndash Unidirectional OT (local theories) hArr Synchronic perspective
ndash (Weak) bidirection OT (global theories) hArr Diachronic perspective
15
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zipf 1949bull Two basic and competing forces
ndash Speakerrsquos economy Force of unification Rndash Hearerrsquos economy Force of diversification Q
bull The two opposing economies are evolutionary forces ie they are balanced during language evolutionndash Languages are evolving via cultural rather than
biological transmission on a historical rather than genetic timescale
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
3The idea of fossilization
16
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
FossilizationGlobal Theories
bidirectional optimization
f1
f2
m1 m2f1
f2
m1 m2
Local Theoriesunidirectional optimization
Fossilization
Markedness constraintssiFsiM
Linking constraintsI M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
(Iterated) Learning
m f mrsquoSpeaker Hearer
m = mrsquo
If yes nothing happens
If no adjustment- All constraints that favour (f m) over (f mrsquo) are promoted- All constraints that favour (f mrsquo) over (f m) are demoted
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
6
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Procedure
ndash test possible inter-pretations in their order of accessibility
ndash stop once the expectation of opti-mal relevance is satisfied
[EFFORT]
[EFFECT]
1234567
EFFORTEFFECT
middotmiddot
(cf Sperber Cara amp Girotto 1995 95)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Optimal InterpretationHendriks amp de Hoop The integration of pragmatic and syntacticsemantic informationin a system of ranked constraints in order to correctly derive the optimal interpretations
Suggestion by RT EFFECT gtgt EFFORT
ZeevatFaith gtgt Consistence gtgt Do not accommodate gtgt Strength
7
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Neo-Gricean Theories (Atlas Horn)
bull The Q-Principle (Hearer-based)ndash Say as much as you can (modulo R)
(Gricersquos first quantity maxim and the first two manner maxims)
bull The R-Principle (Speaker-based)ndash Say not more than you must (modulo Q)
(Gricersquos second quantity maxim relation maxim and the second two manner maxims)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak Bidirectionalitylangf mrang is weakly optimal (= super-optimal) iff
a langf mrang isin GENb there is no weakly optimal langfrsquo mrang isin GEN
such that langfrsquo mrang gt langf mrangc there is no weakly optimal langf mrsquo rang isin GEN
such that langf mrsquo rang gt langf mrang
This is an abstract scheme (Jaumlger 2002) The content of the ordering relation ldquogtrdquo is
determined by the system of constraints
8
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Example
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
siMsiF
siF prefer short formssiM prefer stereotypical meanings
f1
f2
m1 m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Levinsonlsquos Presumptive Meaningsbull Presumptive meanings are a matter of
preferred interpretation calculated by a particular default mechanism
bull Presumptive meanings are localbull Three heuristics
ndash Q-heuristic What isnrsquot said is not the casendash I-heuristic What is expressed simply is
stereotypically exemplifiedndash M-heuristic Whatrsquos said in an abnormal way
isnrsquot normal
9
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
What are the heuristics
siM
FrarrM
FrarrM
FrarrM
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
FrarrMsiF
siF prefer short formssiM prefer stereotypical meanings
I-heuristic What is expressed simply is stereotypically exemplifiedM-heuristic Whatrsquos said in an abnormal way isnrsquot normal
I-heur
M-heurAnti-I
Anti-M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
ComparisonNeo-Griceanglobal theory
bidirection optimizationsiFsiM
f1
f2
m1 m2f1
f2
m1 m2
Levinson (2000)local theory
unidirectional optimizationI M
10
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Conclusions
Relevance TheoryLevinson (2000)
Neo-GriceanTheory
Unidirectional OTBidirectional OT
Local TheoriesGlobal Theories
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
2Cognitive motivation of
symmetry and bidirection
11
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetric OT-systemsbull Symmetric system If f m (optimal
interpretation) then m f (optimal expression)
bull For symmetric systems unidirectionaloptimization gives the same solutions as bidirectional optimization and vice versa
f1f2
m1m2
f1f2
m1m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The symmetry testbull A set of pairs of patterns (Ai Bi) are
repeatedly presented When one member of the pair is presented the subject has to learn to produce the other Assume a 1-1 correspondence between A and B
bull If subjects are qualified to match Stimulus A to B and then without further training match B to A they have passed a test of symmetry
lop
raf
kas
lop
raf
kas
12
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Findings
bull Children as young as 2 years pass the symmetry test (Green 1990)
bull Chimps do not show symmetry (see Savage-Rumbaugh 1984)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetry between languageproduction and comprehension
bull Normally we can understand the sentences we producendash Exceptions are very rare
bull In most cases we can produce the sentences we understandndash Typical exceptions in language acquisition Childrenlsquos
ability in production lags dramatically behind their ability in comprehension
13
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Two ways of deriving symmetrybull Symmetry as a result of the
network architecturendash In symmetric networks one
and the same pattern can be produced starting from different inputs
bull Symmetry as a result of learning ndash Tesarrsquos and Smolenskyrsquos
learning theory
A rarr BA larr B
Difference A-A triggers learning
∆
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
14
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak bidirection and symmetrybull Rather than seeing weak bidirection as a online
interpretationproduction mechanism it shouldbe understood in terms of (iterated) learning(resulting in symmetric OT systems)
bull Suggestionndash Unidirectional OT (local theories) hArr Synchronic perspective
ndash (Weak) bidirection OT (global theories) hArr Diachronic perspective
15
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zipf 1949bull Two basic and competing forces
ndash Speakerrsquos economy Force of unification Rndash Hearerrsquos economy Force of diversification Q
bull The two opposing economies are evolutionary forces ie they are balanced during language evolutionndash Languages are evolving via cultural rather than
biological transmission on a historical rather than genetic timescale
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
3The idea of fossilization
16
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
FossilizationGlobal Theories
bidirectional optimization
f1
f2
m1 m2f1
f2
m1 m2
Local Theoriesunidirectional optimization
Fossilization
Markedness constraintssiFsiM
Linking constraintsI M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
(Iterated) Learning
m f mrsquoSpeaker Hearer
m = mrsquo
If yes nothing happens
If no adjustment- All constraints that favour (f m) over (f mrsquo) are promoted- All constraints that favour (f mrsquo) over (f m) are demoted
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
7
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Neo-Gricean Theories (Atlas Horn)
bull The Q-Principle (Hearer-based)ndash Say as much as you can (modulo R)
(Gricersquos first quantity maxim and the first two manner maxims)
bull The R-Principle (Speaker-based)ndash Say not more than you must (modulo Q)
(Gricersquos second quantity maxim relation maxim and the second two manner maxims)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak Bidirectionalitylangf mrang is weakly optimal (= super-optimal) iff
a langf mrang isin GENb there is no weakly optimal langfrsquo mrang isin GEN
such that langfrsquo mrang gt langf mrangc there is no weakly optimal langf mrsquo rang isin GEN
such that langf mrsquo rang gt langf mrang
This is an abstract scheme (Jaumlger 2002) The content of the ordering relation ldquogtrdquo is
determined by the system of constraints
8
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Example
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
siMsiF
siF prefer short formssiM prefer stereotypical meanings
f1
f2
m1 m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Levinsonlsquos Presumptive Meaningsbull Presumptive meanings are a matter of
preferred interpretation calculated by a particular default mechanism
bull Presumptive meanings are localbull Three heuristics
ndash Q-heuristic What isnrsquot said is not the casendash I-heuristic What is expressed simply is
stereotypically exemplifiedndash M-heuristic Whatrsquos said in an abnormal way
isnrsquot normal
9
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
What are the heuristics
siM
FrarrM
FrarrM
FrarrM
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
FrarrMsiF
siF prefer short formssiM prefer stereotypical meanings
I-heuristic What is expressed simply is stereotypically exemplifiedM-heuristic Whatrsquos said in an abnormal way isnrsquot normal
I-heur
M-heurAnti-I
Anti-M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
ComparisonNeo-Griceanglobal theory
bidirection optimizationsiFsiM
f1
f2
m1 m2f1
f2
m1 m2
Levinson (2000)local theory
unidirectional optimizationI M
10
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Conclusions
Relevance TheoryLevinson (2000)
Neo-GriceanTheory
Unidirectional OTBidirectional OT
Local TheoriesGlobal Theories
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
2Cognitive motivation of
symmetry and bidirection
11
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetric OT-systemsbull Symmetric system If f m (optimal
interpretation) then m f (optimal expression)
bull For symmetric systems unidirectionaloptimization gives the same solutions as bidirectional optimization and vice versa
f1f2
m1m2
f1f2
m1m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The symmetry testbull A set of pairs of patterns (Ai Bi) are
repeatedly presented When one member of the pair is presented the subject has to learn to produce the other Assume a 1-1 correspondence between A and B
bull If subjects are qualified to match Stimulus A to B and then without further training match B to A they have passed a test of symmetry
lop
raf
kas
lop
raf
kas
12
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Findings
bull Children as young as 2 years pass the symmetry test (Green 1990)
bull Chimps do not show symmetry (see Savage-Rumbaugh 1984)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetry between languageproduction and comprehension
bull Normally we can understand the sentences we producendash Exceptions are very rare
bull In most cases we can produce the sentences we understandndash Typical exceptions in language acquisition Childrenlsquos
ability in production lags dramatically behind their ability in comprehension
13
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Two ways of deriving symmetrybull Symmetry as a result of the
network architecturendash In symmetric networks one
and the same pattern can be produced starting from different inputs
bull Symmetry as a result of learning ndash Tesarrsquos and Smolenskyrsquos
learning theory
A rarr BA larr B
Difference A-A triggers learning
∆
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
14
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak bidirection and symmetrybull Rather than seeing weak bidirection as a online
interpretationproduction mechanism it shouldbe understood in terms of (iterated) learning(resulting in symmetric OT systems)
bull Suggestionndash Unidirectional OT (local theories) hArr Synchronic perspective
ndash (Weak) bidirection OT (global theories) hArr Diachronic perspective
15
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zipf 1949bull Two basic and competing forces
ndash Speakerrsquos economy Force of unification Rndash Hearerrsquos economy Force of diversification Q
bull The two opposing economies are evolutionary forces ie they are balanced during language evolutionndash Languages are evolving via cultural rather than
biological transmission on a historical rather than genetic timescale
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
3The idea of fossilization
16
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
FossilizationGlobal Theories
bidirectional optimization
f1
f2
m1 m2f1
f2
m1 m2
Local Theoriesunidirectional optimization
Fossilization
Markedness constraintssiFsiM
Linking constraintsI M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
(Iterated) Learning
m f mrsquoSpeaker Hearer
m = mrsquo
If yes nothing happens
If no adjustment- All constraints that favour (f m) over (f mrsquo) are promoted- All constraints that favour (f mrsquo) over (f m) are demoted
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
8
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Example
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
siMsiF
siF prefer short formssiM prefer stereotypical meanings
f1
f2
m1 m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Levinsonlsquos Presumptive Meaningsbull Presumptive meanings are a matter of
preferred interpretation calculated by a particular default mechanism
bull Presumptive meanings are localbull Three heuristics
ndash Q-heuristic What isnrsquot said is not the casendash I-heuristic What is expressed simply is
stereotypically exemplifiedndash M-heuristic Whatrsquos said in an abnormal way
isnrsquot normal
9
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
What are the heuristics
siM
FrarrM
FrarrM
FrarrM
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
FrarrMsiF
siF prefer short formssiM prefer stereotypical meanings
I-heuristic What is expressed simply is stereotypically exemplifiedM-heuristic Whatrsquos said in an abnormal way isnrsquot normal
I-heur
M-heurAnti-I
Anti-M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
ComparisonNeo-Griceanglobal theory
bidirection optimizationsiFsiM
f1
f2
m1 m2f1
f2
m1 m2
Levinson (2000)local theory
unidirectional optimizationI M
10
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Conclusions
Relevance TheoryLevinson (2000)
Neo-GriceanTheory
Unidirectional OTBidirectional OT
Local TheoriesGlobal Theories
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
2Cognitive motivation of
symmetry and bidirection
11
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetric OT-systemsbull Symmetric system If f m (optimal
interpretation) then m f (optimal expression)
bull For symmetric systems unidirectionaloptimization gives the same solutions as bidirectional optimization and vice versa
f1f2
m1m2
f1f2
m1m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The symmetry testbull A set of pairs of patterns (Ai Bi) are
repeatedly presented When one member of the pair is presented the subject has to learn to produce the other Assume a 1-1 correspondence between A and B
bull If subjects are qualified to match Stimulus A to B and then without further training match B to A they have passed a test of symmetry
lop
raf
kas
lop
raf
kas
12
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Findings
bull Children as young as 2 years pass the symmetry test (Green 1990)
bull Chimps do not show symmetry (see Savage-Rumbaugh 1984)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetry between languageproduction and comprehension
bull Normally we can understand the sentences we producendash Exceptions are very rare
bull In most cases we can produce the sentences we understandndash Typical exceptions in language acquisition Childrenlsquos
ability in production lags dramatically behind their ability in comprehension
13
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Two ways of deriving symmetrybull Symmetry as a result of the
network architecturendash In symmetric networks one
and the same pattern can be produced starting from different inputs
bull Symmetry as a result of learning ndash Tesarrsquos and Smolenskyrsquos
learning theory
A rarr BA larr B
Difference A-A triggers learning
∆
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
14
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak bidirection and symmetrybull Rather than seeing weak bidirection as a online
interpretationproduction mechanism it shouldbe understood in terms of (iterated) learning(resulting in symmetric OT systems)
bull Suggestionndash Unidirectional OT (local theories) hArr Synchronic perspective
ndash (Weak) bidirection OT (global theories) hArr Diachronic perspective
15
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zipf 1949bull Two basic and competing forces
ndash Speakerrsquos economy Force of unification Rndash Hearerrsquos economy Force of diversification Q
bull The two opposing economies are evolutionary forces ie they are balanced during language evolutionndash Languages are evolving via cultural rather than
biological transmission on a historical rather than genetic timescale
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
3The idea of fossilization
16
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
FossilizationGlobal Theories
bidirectional optimization
f1
f2
m1 m2f1
f2
m1 m2
Local Theoriesunidirectional optimization
Fossilization
Markedness constraintssiFsiM
Linking constraintsI M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
(Iterated) Learning
m f mrsquoSpeaker Hearer
m = mrsquo
If yes nothing happens
If no adjustment- All constraints that favour (f m) over (f mrsquo) are promoted- All constraints that favour (f mrsquo) over (f m) are demoted
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
9
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
What are the heuristics
siM
FrarrM
FrarrM
FrarrM
langf2 m2ranglangf2 m1rang
langf1 m2ranglangf1 m1rang
FrarrMsiF
siF prefer short formssiM prefer stereotypical meanings
I-heuristic What is expressed simply is stereotypically exemplifiedM-heuristic Whatrsquos said in an abnormal way isnrsquot normal
I-heur
M-heurAnti-I
Anti-M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
ComparisonNeo-Griceanglobal theory
bidirection optimizationsiFsiM
f1
f2
m1 m2f1
f2
m1 m2
Levinson (2000)local theory
unidirectional optimizationI M
10
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Conclusions
Relevance TheoryLevinson (2000)
Neo-GriceanTheory
Unidirectional OTBidirectional OT
Local TheoriesGlobal Theories
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
2Cognitive motivation of
symmetry and bidirection
11
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetric OT-systemsbull Symmetric system If f m (optimal
interpretation) then m f (optimal expression)
bull For symmetric systems unidirectionaloptimization gives the same solutions as bidirectional optimization and vice versa
f1f2
m1m2
f1f2
m1m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The symmetry testbull A set of pairs of patterns (Ai Bi) are
repeatedly presented When one member of the pair is presented the subject has to learn to produce the other Assume a 1-1 correspondence between A and B
bull If subjects are qualified to match Stimulus A to B and then without further training match B to A they have passed a test of symmetry
lop
raf
kas
lop
raf
kas
12
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Findings
bull Children as young as 2 years pass the symmetry test (Green 1990)
bull Chimps do not show symmetry (see Savage-Rumbaugh 1984)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetry between languageproduction and comprehension
bull Normally we can understand the sentences we producendash Exceptions are very rare
bull In most cases we can produce the sentences we understandndash Typical exceptions in language acquisition Childrenlsquos
ability in production lags dramatically behind their ability in comprehension
13
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Two ways of deriving symmetrybull Symmetry as a result of the
network architecturendash In symmetric networks one
and the same pattern can be produced starting from different inputs
bull Symmetry as a result of learning ndash Tesarrsquos and Smolenskyrsquos
learning theory
A rarr BA larr B
Difference A-A triggers learning
∆
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
14
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak bidirection and symmetrybull Rather than seeing weak bidirection as a online
interpretationproduction mechanism it shouldbe understood in terms of (iterated) learning(resulting in symmetric OT systems)
bull Suggestionndash Unidirectional OT (local theories) hArr Synchronic perspective
ndash (Weak) bidirection OT (global theories) hArr Diachronic perspective
15
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zipf 1949bull Two basic and competing forces
ndash Speakerrsquos economy Force of unification Rndash Hearerrsquos economy Force of diversification Q
bull The two opposing economies are evolutionary forces ie they are balanced during language evolutionndash Languages are evolving via cultural rather than
biological transmission on a historical rather than genetic timescale
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
3The idea of fossilization
16
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
FossilizationGlobal Theories
bidirectional optimization
f1
f2
m1 m2f1
f2
m1 m2
Local Theoriesunidirectional optimization
Fossilization
Markedness constraintssiFsiM
Linking constraintsI M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
(Iterated) Learning
m f mrsquoSpeaker Hearer
m = mrsquo
If yes nothing happens
If no adjustment- All constraints that favour (f m) over (f mrsquo) are promoted- All constraints that favour (f mrsquo) over (f m) are demoted
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
10
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Conclusions
Relevance TheoryLevinson (2000)
Neo-GriceanTheory
Unidirectional OTBidirectional OT
Local TheoriesGlobal Theories
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
2Cognitive motivation of
symmetry and bidirection
11
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetric OT-systemsbull Symmetric system If f m (optimal
interpretation) then m f (optimal expression)
bull For symmetric systems unidirectionaloptimization gives the same solutions as bidirectional optimization and vice versa
f1f2
m1m2
f1f2
m1m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The symmetry testbull A set of pairs of patterns (Ai Bi) are
repeatedly presented When one member of the pair is presented the subject has to learn to produce the other Assume a 1-1 correspondence between A and B
bull If subjects are qualified to match Stimulus A to B and then without further training match B to A they have passed a test of symmetry
lop
raf
kas
lop
raf
kas
12
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Findings
bull Children as young as 2 years pass the symmetry test (Green 1990)
bull Chimps do not show symmetry (see Savage-Rumbaugh 1984)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetry between languageproduction and comprehension
bull Normally we can understand the sentences we producendash Exceptions are very rare
bull In most cases we can produce the sentences we understandndash Typical exceptions in language acquisition Childrenlsquos
ability in production lags dramatically behind their ability in comprehension
13
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Two ways of deriving symmetrybull Symmetry as a result of the
network architecturendash In symmetric networks one
and the same pattern can be produced starting from different inputs
bull Symmetry as a result of learning ndash Tesarrsquos and Smolenskyrsquos
learning theory
A rarr BA larr B
Difference A-A triggers learning
∆
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
14
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak bidirection and symmetrybull Rather than seeing weak bidirection as a online
interpretationproduction mechanism it shouldbe understood in terms of (iterated) learning(resulting in symmetric OT systems)
bull Suggestionndash Unidirectional OT (local theories) hArr Synchronic perspective
ndash (Weak) bidirection OT (global theories) hArr Diachronic perspective
15
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zipf 1949bull Two basic and competing forces
ndash Speakerrsquos economy Force of unification Rndash Hearerrsquos economy Force of diversification Q
bull The two opposing economies are evolutionary forces ie they are balanced during language evolutionndash Languages are evolving via cultural rather than
biological transmission on a historical rather than genetic timescale
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
3The idea of fossilization
16
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
FossilizationGlobal Theories
bidirectional optimization
f1
f2
m1 m2f1
f2
m1 m2
Local Theoriesunidirectional optimization
Fossilization
Markedness constraintssiFsiM
Linking constraintsI M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
(Iterated) Learning
m f mrsquoSpeaker Hearer
m = mrsquo
If yes nothing happens
If no adjustment- All constraints that favour (f m) over (f mrsquo) are promoted- All constraints that favour (f mrsquo) over (f m) are demoted
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
11
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetric OT-systemsbull Symmetric system If f m (optimal
interpretation) then m f (optimal expression)
bull For symmetric systems unidirectionaloptimization gives the same solutions as bidirectional optimization and vice versa
f1f2
m1m2
f1f2
m1m2
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The symmetry testbull A set of pairs of patterns (Ai Bi) are
repeatedly presented When one member of the pair is presented the subject has to learn to produce the other Assume a 1-1 correspondence between A and B
bull If subjects are qualified to match Stimulus A to B and then without further training match B to A they have passed a test of symmetry
lop
raf
kas
lop
raf
kas
12
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Findings
bull Children as young as 2 years pass the symmetry test (Green 1990)
bull Chimps do not show symmetry (see Savage-Rumbaugh 1984)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetry between languageproduction and comprehension
bull Normally we can understand the sentences we producendash Exceptions are very rare
bull In most cases we can produce the sentences we understandndash Typical exceptions in language acquisition Childrenlsquos
ability in production lags dramatically behind their ability in comprehension
13
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Two ways of deriving symmetrybull Symmetry as a result of the
network architecturendash In symmetric networks one
and the same pattern can be produced starting from different inputs
bull Symmetry as a result of learning ndash Tesarrsquos and Smolenskyrsquos
learning theory
A rarr BA larr B
Difference A-A triggers learning
∆
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
14
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak bidirection and symmetrybull Rather than seeing weak bidirection as a online
interpretationproduction mechanism it shouldbe understood in terms of (iterated) learning(resulting in symmetric OT systems)
bull Suggestionndash Unidirectional OT (local theories) hArr Synchronic perspective
ndash (Weak) bidirection OT (global theories) hArr Diachronic perspective
15
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zipf 1949bull Two basic and competing forces
ndash Speakerrsquos economy Force of unification Rndash Hearerrsquos economy Force of diversification Q
bull The two opposing economies are evolutionary forces ie they are balanced during language evolutionndash Languages are evolving via cultural rather than
biological transmission on a historical rather than genetic timescale
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
3The idea of fossilization
16
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
FossilizationGlobal Theories
bidirectional optimization
f1
f2
m1 m2f1
f2
m1 m2
Local Theoriesunidirectional optimization
Fossilization
Markedness constraintssiFsiM
Linking constraintsI M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
(Iterated) Learning
m f mrsquoSpeaker Hearer
m = mrsquo
If yes nothing happens
If no adjustment- All constraints that favour (f m) over (f mrsquo) are promoted- All constraints that favour (f mrsquo) over (f m) are demoted
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
12
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Findings
bull Children as young as 2 years pass the symmetry test (Green 1990)
bull Chimps do not show symmetry (see Savage-Rumbaugh 1984)
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Symmetry between languageproduction and comprehension
bull Normally we can understand the sentences we producendash Exceptions are very rare
bull In most cases we can produce the sentences we understandndash Typical exceptions in language acquisition Childrenlsquos
ability in production lags dramatically behind their ability in comprehension
13
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Two ways of deriving symmetrybull Symmetry as a result of the
network architecturendash In symmetric networks one
and the same pattern can be produced starting from different inputs
bull Symmetry as a result of learning ndash Tesarrsquos and Smolenskyrsquos
learning theory
A rarr BA larr B
Difference A-A triggers learning
∆
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
14
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak bidirection and symmetrybull Rather than seeing weak bidirection as a online
interpretationproduction mechanism it shouldbe understood in terms of (iterated) learning(resulting in symmetric OT systems)
bull Suggestionndash Unidirectional OT (local theories) hArr Synchronic perspective
ndash (Weak) bidirection OT (global theories) hArr Diachronic perspective
15
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zipf 1949bull Two basic and competing forces
ndash Speakerrsquos economy Force of unification Rndash Hearerrsquos economy Force of diversification Q
bull The two opposing economies are evolutionary forces ie they are balanced during language evolutionndash Languages are evolving via cultural rather than
biological transmission on a historical rather than genetic timescale
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
3The idea of fossilization
16
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
FossilizationGlobal Theories
bidirectional optimization
f1
f2
m1 m2f1
f2
m1 m2
Local Theoriesunidirectional optimization
Fossilization
Markedness constraintssiFsiM
Linking constraintsI M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
(Iterated) Learning
m f mrsquoSpeaker Hearer
m = mrsquo
If yes nothing happens
If no adjustment- All constraints that favour (f m) over (f mrsquo) are promoted- All constraints that favour (f mrsquo) over (f m) are demoted
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
13
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Two ways of deriving symmetrybull Symmetry as a result of the
network architecturendash In symmetric networks one
and the same pattern can be produced starting from different inputs
bull Symmetry as a result of learning ndash Tesarrsquos and Smolenskyrsquos
learning theory
A rarr BA larr B
Difference A-A triggers learning
∆
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
14
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak bidirection and symmetrybull Rather than seeing weak bidirection as a online
interpretationproduction mechanism it shouldbe understood in terms of (iterated) learning(resulting in symmetric OT systems)
bull Suggestionndash Unidirectional OT (local theories) hArr Synchronic perspective
ndash (Weak) bidirection OT (global theories) hArr Diachronic perspective
15
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zipf 1949bull Two basic and competing forces
ndash Speakerrsquos economy Force of unification Rndash Hearerrsquos economy Force of diversification Q
bull The two opposing economies are evolutionary forces ie they are balanced during language evolutionndash Languages are evolving via cultural rather than
biological transmission on a historical rather than genetic timescale
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
3The idea of fossilization
16
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
FossilizationGlobal Theories
bidirectional optimization
f1
f2
m1 m2f1
f2
m1 m2
Local Theoriesunidirectional optimization
Fossilization
Markedness constraintssiFsiM
Linking constraintsI M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
(Iterated) Learning
m f mrsquoSpeaker Hearer
m = mrsquo
If yes nothing happens
If no adjustment- All constraints that favour (f m) over (f mrsquo) are promoted- All constraints that favour (f mrsquo) over (f m) are demoted
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
14
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Mirror neurons in the premotor cortexA mirror neuron is a neuron which fires bothwhen performing an actionand when observing the same action performedby another creature
bull Monkey grasp a nutbull Monkey sees how
another creaturegrasp a nut
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Weak bidirection and symmetrybull Rather than seeing weak bidirection as a online
interpretationproduction mechanism it shouldbe understood in terms of (iterated) learning(resulting in symmetric OT systems)
bull Suggestionndash Unidirectional OT (local theories) hArr Synchronic perspective
ndash (Weak) bidirection OT (global theories) hArr Diachronic perspective
15
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zipf 1949bull Two basic and competing forces
ndash Speakerrsquos economy Force of unification Rndash Hearerrsquos economy Force of diversification Q
bull The two opposing economies are evolutionary forces ie they are balanced during language evolutionndash Languages are evolving via cultural rather than
biological transmission on a historical rather than genetic timescale
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
3The idea of fossilization
16
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
FossilizationGlobal Theories
bidirectional optimization
f1
f2
m1 m2f1
f2
m1 m2
Local Theoriesunidirectional optimization
Fossilization
Markedness constraintssiFsiM
Linking constraintsI M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
(Iterated) Learning
m f mrsquoSpeaker Hearer
m = mrsquo
If yes nothing happens
If no adjustment- All constraints that favour (f m) over (f mrsquo) are promoted- All constraints that favour (f mrsquo) over (f m) are demoted
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
15
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Zipf 1949bull Two basic and competing forces
ndash Speakerrsquos economy Force of unification Rndash Hearerrsquos economy Force of diversification Q
bull The two opposing economies are evolutionary forces ie they are balanced during language evolutionndash Languages are evolving via cultural rather than
biological transmission on a historical rather than genetic timescale
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
3The idea of fossilization
16
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
FossilizationGlobal Theories
bidirectional optimization
f1
f2
m1 m2f1
f2
m1 m2
Local Theoriesunidirectional optimization
Fossilization
Markedness constraintssiFsiM
Linking constraintsI M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
(Iterated) Learning
m f mrsquoSpeaker Hearer
m = mrsquo
If yes nothing happens
If no adjustment- All constraints that favour (f m) over (f mrsquo) are promoted- All constraints that favour (f mrsquo) over (f m) are demoted
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
16
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
FossilizationGlobal Theories
bidirectional optimization
f1
f2
m1 m2f1
f2
m1 m2
Local Theoriesunidirectional optimization
Fossilization
Markedness constraintssiFsiM
Linking constraintsI M
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
(Iterated) Learning
m f mrsquoSpeaker Hearer
m = mrsquo
If yes nothing happens
If no adjustment- All constraints that favour (f m) over (f mrsquo) are promoted- All constraints that favour (f mrsquo) over (f m) are demoted
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
17
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Population in pairwiseinteraction
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
All possible strategies
Smolensky
AntiHorn
Horn
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
18
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
General Observationsbull Horn and Anti-Horn are the only strategies (OT-
systems) that are stable
bull Starting with a uniform Smolensky population the system will always move into
ndash a pure Horn population supposed P(m1) gt P(m1)
ndash a pure Anti-Horn population supposed P(m1) lt P(m1)
bull The same holds for mixed populations
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Joost Zwarts round in Englisha The postman ran round the block (in a circle)b The burglar drove round the barrier (to the opposite side)c The steeplechaser ran round the corner (to the other side)d The captain sailed round the lakee The tourist drove round the city centre
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
19
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
Strongest Meaning Hypothesis
detourorthogonalityinversion
completeness
StrengthConsistenceround the door
Lexicon round rarr Approx [Circle]
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
JZwaarts om and rond in Dutch
Ze zaten rond (om) de televisie
Een man stak zijn hoofd om(rond rondom) de deur
De auto reed om (rond rondom) het obstakel heen
het gebied rondom (om) hetstadje
a They sat round the television
b A man put his head round the door
c The drove round the obstacle
d the area round the little town
DETOUR ------------------------------------------------ CIRCLEom hellip strengthening rarr hellip larr weakening hellip rond
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
20
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The interplay between broadeningand narrowing
Zwartrsquos (2005)
bull If rond has some inter-pretation m then it has each stronger inter-pretation
bull If om has some inter-pretation m then it has each weaker interpretation
bull there is some overlap between om and rond
rond
om
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
The puzzle
bull the marked form (rond) conforms to the stronger (= preferred) meanings [hellip Circle]
bull the unmarked form (om) conforms to the weaker meanings [hellip DeTour]
bull This conflicts with weak bidirection and iconicity
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions
21
ZAS Workshop Berlin
Reinhard Blutner University of Amsterdam May 2007
bull Local theories ndash unidirectional optimizationndash Synchronic view
bull Global theories ndash bidirectional optimizationndash Diachronic view
bull Weak bidirection as (lexical) fossilizationndash Fossilization = lsquoroutinizationrsquo of implicatures
4Conclusions