the relevance of syntactic structures for syntactic variation; or, ���

the interaction of effects���on the particle verb alternation

•  Daniel Ezra Johnson – Lancaster University – d.e.johnson @ lancaster.ac.uk

•  Thanks to: – Bill Haddican – Kyle Gorman, Laurel MacKenzie, Joel Wallenberg

a variationist’s view of the world

S P

V

vers une architecture

G L

C

A P C

plumbing the depths

•  economy principles – some variable alternations are like categorical ones so don’t duplicate the machinery (Guy & Boberg 1997) – some variable alternations aren’t like categorical ones these must be outside the grammar (MacKenzie 2012) – can we further justify these w/ appeal to learnability?

•  theoretically and/or empirically-motivated architectural constraints – ‘phonetics can’t read’ – ‘grammars can’t count’ so if something is counting, it can’t be the grammar – if effects interact, they must apply at the same stage

the variable

•  one of a few English ‘alternations’ •  long history of study •  but what are ‘alternants’ theoretically? •  using VOP vs. VPO – agnostic labels •  no referential meaning difference •  little or no social meaning difference •  compare to ditransitive (‘dative alternation’) •  what is beyond the ‘envelope of variation’? – idioms (fixed and favoring) – VOPP constructions

the syntax

•  does syntax just give options, or does variation arise there? •  if the latter, potential for duplication of machinery •  if the former, in the syntax you have X[+F] or you don’t •  but does the lexicon work the same way?

the data

•  two acceptability judgment experiments •  subjects judged sentences ‘bad’ (0) to ‘good’ (1) •  experiment 1: 297 subjects, US/Can./UK/Ire. •  object weight: ‘the (lumpy 10-pound) pumpkin’ •  object oldness: via cataphoric pronoun

Because she had no money… vs. Because it tasted funny… … Susan spit the conference dinner out.

•  32 stimuli, all compositional (Lohse et al. 2004) •  32 fillers/normalizers – treated as fillers here •  able to look at VPO and VOP separately

the data

•  experiment 2: 125 subjects from USA •  object length is now fixed •  four topic/focus conditions via question prompt – Q1. What did the friends do? (VP focus) – Q2. What did the friends pass around? (object focus) – Q3. What happened? (wide focus) – Q4. What happened to the beer? (object topic) – A1-4. The friends passed the beer around.

•  worked much better than the cataphoric pronoun •  won’t discuss today, very similar analytical issues

the effects

•  ‘social’: time, register/style, variety: US vs. UK •  ‘individual’: interacts with all the below •  prosodic: object weight – affected by processing constraints – never represented in syntax

•  information-structural: old/new, topic/focus – affected by processing constraints – sometimes represented in syntax

•  lexical: verb, particle, V-Prt pair: not today

the predictions

•  ~100 years of research – corpora that must treat VOP/VPO as a choice – experiments that treat VOP/VPO as a choice

•  a heavy object should… – make VOP order worse – have no effect on VPO order (make slightly worse?)

•  a discourse-old (or topic) object should… – make VOP order better – make VPO order worse (probably both?)

object weight

5

67

89

10

VOP - light VOP - heavy VPO - light VPO - heavy

8.21 7.97 8.05 8.21

-0.24 +0.17

p = .002 p = .02

response ~ order * weight + (order * weight | subject) + (order * weight | stimulus)

object weight

-2.0 -1.5 -1.0 -0.5 0.0 0.5

-0.5

0.0

0.5

1.0

effect of heavy object - VOP

effe

ct o

f hea

vy o

bjec

t - V

PO

b = -0.096r = -.497p ~= 0

response ~ order * weight + (order * weight | subject) + (order * weight | stimulus)

-2.0 -1.5 -1.0 -0.5 0.0 0.5

-0.5

0.0

0.5

1.0

effect of heavy object - VOP

effe

ct o

f hea

vy o

bjec

t - V

PO

b = -0.096r = -.497p ~= 0

object weight

response ~ order * weight + (order * weight | subject) + (order * weight | stimulus)

UK / IRL (152) Canada (32) USA (113)

object newness

response ~ order * focus + (order * focus | subject) + (order * focus | stimulus)

56

78

910

VOP - old VOP - new VPO - old VPO - new

7.91 8.27 7.89 8.37

+0.36 +0.49

p = .12

56

78

910

VOP - old VOP - new VPO - old VPO - new

8.07 8.39 7.87 8.26+0.32 +0.39+.077.62 8.07 7.87 8.50+0.45 +0.63+.18

object newness

response ~ country * order * focus + (order * focus | subject) + (order * focus | stimulus)

UK / IRL (152) USA (113)

-0.5 0.0 0.5 1.0 1.5 2.0

0.0

0.5

1.0

1.5

effect of new object on VPO order

effe

ct o

f new

obj

ect o

n VO

P or

der

object newness

response ~ country * order * focus + (order * focus | subject) + (order * focus | stimulus)

UK / IRL (152) USA (113)

-0.5 0.0 0.5 1.0 1.5 2.0

0.0

0.5

1.0

1.5

effect of new object on VPO order

effe

ct o

f new

obj

ect o

n VO

P or

der

object newness

response ~ country * order * focus + (order * focus | subject) + (order * focus | stimulus)

UK / IRL (152) USA (113)

effect of new object on VPO order

effe

ct o

f new

obj

ect o

n V

OP

ord

er

-0.25 0.00 0.25 0.50 0.75

-0.25

0.00

0.25

0.50

0.75

object weight * object newness ?

UK / IRL (61) USA (39)

object weight * object newness ?

0 1 2 3 4 5 6 7 8 9 10

050

100

150

200

mean: 8.60order: VOPobject: old, light

1 2 3 4 5 6 7 8 9 10

050

100

150

200

mean: 8.35order: VOPobject: old, heavy

0 1 2 3 4 5 6 7 8 9 10

050

100

150

200

mean: 8.57order: VOPobject: new, light

0 1 2 3 4 5 6 7 8 9 10

050

100

150

200

mean: 8.34order: VOPobject: new, heavy

object weight * object newness !

0 2 3 4 5 6 7 8 9 10

050

100

150

200

mean: 8.42order: VPOobject: old, light

0 1 2 3 4 5 6 7 8 9 10

050

100

150

200

mean: 8.59order: VPOobject: old, heavy

0 1 2 3 4 5 6 7 8 9 10

050

100

150

200

mean: 8.54order: VPOobject: new, light

0 1 2 3 4 5 6 7 8 9 10

050

100

150

200

mean: 8.64order: VPOobject: new, heavy

conclusions

•  prosodic and information-structural effects on

word order variation can be elicited experimentally •  measuring acceptability on 11-point Likert scale,

then treating it as linear – while not ideal! – gave better results than attempts at normalization

•  subjects vary along every dimension you measure •  subject random effects are very valuable data •  two word order ‘alternants’ can appear linked or

‘yoked’ together, cf. quantum entanglement •  object weight and information status can interact

more discussion and questions

•  constituent length can be extragrammatical,

but it can also affect ‘grammar competition’ •  can ‘the same’ effect be in 2 places (economy)? •  despite (arguably) being represented in syntax,

information structure also affects competition •  does any variation arise in the syntax proper? •  can we constrain ‘grammar competition’ to make

it distinguishable from ‘lower-level’ variation? •  whether 2 or 3 levels, syntax = phonology?

references (1)

•  Adger, David and Jennifer Smith. 2010. Variation in agreement:

A lexical feature-based approach. Lingua 120: 1109–1134.

•  Bates, Douglas, Martin Maechler, Ben Bolker and Steven Walker. 2013. lme4: Linear mixed-effects models using Eigen and S4. R package version 1.0-5. http://cran.R-project.org/package=lme4

•  Bresnan, Joan and Marilyn Ford. 2010. Predicting syntax: Processing dative constructions in American and Australian varieties of English. Language 86(1): 168-213.  

•  Cappelle, Bert. 2009. Contextual cues for particle placement: multiplicity, motivation, modeling. In Bergs and Diewald (eds.), Context in Construction Grammar. John Benjamins. 145-192.

references (2)

•  Guy, Gregory R. and Charles Boberg. 1997. Inherent variability

and the obligatory contour principle. Language Variation and Change 9: 149-164.

•  Kroch, Anthony S. 1989. Reflexes of grammar in patterns of language change.  Language Variation and Change 1: 199-244.  

•  Lavandera, Beatriz. 1978. Where Does the Sociolinguistic Variable Stop? Language in Society 7(2): 171-182.

•  Lohse, Barbara, John A. Hawkins and Thomas Wasow. 2004. Domain minimization in English verb-particle constructions. Language 80(2): 238-261.