Transcript Lexicalised instances of productive constructions

Contrasting two potentiality constructions in Dutch
Maria Mos and Ad Backus ([email protected]; [email protected])
Tilburg University, The Netherlands
Case study: -BAAR and IS TE
BAAR:
•Derivational affix, more or less equivalent to –able
•E.g. leesbaar – read-able (legible)
Dit handschrift is leesbaar
This handwriting is legible
IS TE:
• Modal infinitive construction, no direct English equivalent
(compare: X is hard to find)
• E.g. is te verdedig-en - is to defend-INF (is defendable)
Deze opinie is te verdedig-en
This opinion is to defend-INF
This opinion is defendable
Case study: -BAAR and IS TE
• Very similar meaning: X can be V-ed
• Questions:
> Are they really synonyms?
> Are they productive?
-BAAR and IS TE: corpus
• Corpus Gesproken Nederlands (Corpus of
Spoken Dutch, CGN)
• 10 million words
• Contemporary Dutch
• 2/3 Netherlands, 1/3 Flanders (Belgium)
• 14 genres, ranging from telephone
conversations to official lectures
-BAAR and IS TE: corpus
-BAAR:
• Search for word class = Adj. & form = *BAAR
• 261 different types
• 3908 tokens
The most frequent types (obviously fixed units):
Type
blijkbaar
openbaar
beschikbaar
middelbaar
zichtbaar
onvoorstelbaar
schijnbaar
haalbaar
bereikbaar
dankbaar
frequency Translation
1134
evidently
306
public
173
available
115
secondary
103
visible
81
unimaginable
74
apparently
71
feasible
69
within reach
63
grateful
-BAAR and IS TE: corpus
-BAAR:
In infrequent instantiations (N<5, 50 types in the CGN):
 The meaning is ‘non-Agent argument of the Verb can
be V-ed’ (passive, potential)
 This can be the traditional object in transitive verbs
(drinkbaar -drinkable, ondoorprikbaar –unprickable,
said of a blister)
 Or a (usually) implicit argument (adembaar breathable, roddelbaar -gossipable)
 The stem is verbal (48/50)
 The stem is transitive (45/50)
-BAAR and IS TE: corpus
IS TE:
• Search for IS (+optional 0=3 words) + TE + INFINITIVE
• 120 different types
• 710 tokens
The most frequent verbs:
doen ‘do’
zien ‘see’
hopen ‘hope’
zeggen ‘say’
geloven ‘believe’
vergelijken ‘compare’
vinden ‘find’
merken ‘notice’
spreken ‘talk’
lezen ‘read’
161
105
77
67
23
18
15
12
10
7
This hides the existence of
several more inclusive
units, especially with niet:
is niet te doen
is niet te geloven
is niet te vinden
is niet te spreken
is ver te zoeken
can’t be done
is unbelievable
is hard to find
is not happy
is hard to find
-BAAR and IS TE: corpus
IS TE:
In infrequent instantiations the meaning of
the construction is:
 A predicative comment (the contribution of
the copula) on the relative difficulty or
ease with which an action can be done
 with the connotation that this difficulty or
ease exceeds what could be expected.
 stated as the personal opinion of the
speaker
 In many cases, the difficulty or ease is
explicitly indicated through an adverb of
degree.
Corpus findings: summary
For both constructions we find
A. Large number of types (suggests productivity)
B. Large number of tokens
C. Unequal distribution of tokens over types
(some frequent, many occasionally)
Entrenchment of both template and many units
Challenge
 Observation: many instantiations are
familiar (most?)
 Question: are they ever productively
used?
 Tentative answer: yes, occasionally.
 Basic observation: we can make up
novel instantiations.
 Question: when do people use the
templates? If they have any?
Productivity vs. lexicalisation
A productive construction is
A pattern with one or more open
slots that are available for new
forms
The combination of this pattern
with (a) new word(s) is a
structurally acceptable and
meaningful unit
E.g.: The X-er, the Y-er
un-ADJ
Productivity vs. lexicalisation
• Tension between elegant (=abstract)
description and psychological reality
• Do speakers have these schema’s in their
‘constructicon’? Or are most instantiations
lexicalized expressions?
> That’s the question
• Are regular forms always formed by
productive use of the schema?
> Answer: No
-BAAR and IS TE: experiment
Magnitude estimation:
• Comparable to grammaticality judgment
task, but
• Without a fixed scale: participants assign a
number to each stimulus, reflecting its
acceptability
• Advantages over traditional grammaticality
judgment:
> No fixed number of points on the scale
> No ‘middle’ point which may reflect
either average acceptability or lack of
opinion on a stimulus
-BAAR and IS TE: experiment
• Are these constructions productive?
If:
We make stimuli (novel forms) productively,
that differ with respect to verb category,
And
Participants distinguish consistently in the
acceptability between items of different
categories
Then we know that
A. they have some form of (abstract) mental
representation)
B. the categories are psychologically real.
-BAAR and IS TE: experiment
Magnitude estimation:
• Item variables:
 CONSTRUCTION TYPE:
IS TE (N=24), -BAAR (N=24), fillers (N=36)
 VERB CATEGORY:
Semantic roles and argument structure (5 different
categories)
-BAAR and IS TE: experiment
• V opt. transitive Agent Patient (N=8)
drogen-dry
E.g.: Een wollen trui is niet droogbaar in de machine
(a wool sweater is not dry-able in the machine)
-BAAR and IS TE: experiment
• V opt. transitive Agent Patient (N=8)
drogen-dry
E.g.: Een wollen trui is niet droogbaar in de machine
(a wool sweater is not dry-able in the machine)
• V transitive Agent Patient (N=12)
maaien-mow
E.g.: Die jungle die de buren hun achtertuin noemen is niet te
maaien
(that jungle that the neighbours their backyard call is not to
mow –it’s impossible to mow that jungle the neighbours call
their back garden)
-BAAR and IS TE: experiment
• V opt. transitive Agent Patient (N=8)
drogen-dry
E.g.: Een wollen trui is niet droogbaar in de machine
(a wool sweater is not dry-able in the machine)
• V transitive Agent Patient (N=12)
maaien-mow
E.g.: Die jungle die de buren hun achtertuin noemen is niet te
maaien
(that jungle that the neighbours their backyard call is not to
mow –it’s impossible to mow that jungle the neighbours call
their back garden)
• V intransitive, implied obj./patient (N=8)
zingen-sing
E.g.: De tekst van dit liedje is zo lastig dat het bijna niet zingbaar
is
(the text of this song is so difficult that it almost not sing-able
is)
-BAAR and IS TE: experiment
• V transitive Stimulus Experiencer (N=16)
two subgroups:
passive ungrammatical (N=8)
lukken-succeed
E.g. Het is een ambitieus plan, maar als iedereen helpt is het zeker
lukbaar
(it is an ambitious plan, but if everyone helps it is sure succeedable)
Het is me gelukt vs. *Ik word gelukt.
(Es ist mir gelungen vs. *Ich werde gelungen)
-BAAR and IS TE: experiment
• V transitive Stimulus Experiencer (N=16)
two subgroups:
> passive ungrammatical (N=8)
lukken-succeed
E.g. Het is een ambitieus plan, maar als iedereen helpt is het zeker
lukbaar
(it is an ambitious plan, but if everyone helps it is sure succeed-able)
Het is me gelukt vs. *Ik word gelukt.
(Es ist mir gelungen vs. *Ich werde gelungen)
> passive marginally acceptable (N=8) afschrikken-deter
E.g. Ongewenst bezoek is af te schrikken met een alarmsignaal
(unwanted visit is off to scare with an alarm signal unwanted visitors can be deterred with an alarm signal)
De menigte schrikte me af vs. ?Ik werd afgeschrikt door de menigte
(The crowd scared me off vs I was scared off by the crowd.)
-BAAR and IS TE: experiment
Setup of the experiment:
• Procedure: online experiment, introduction
+ test phase. Duration main experiment:
15-20 minutes
• Participants:
> 69 adults
> 138 children, grade 6 (11-12 yrs)
-BAAR and IS TE: experiment
Main results: adults
• Reliable test (Cronbach’s α = .85)
• Difference between construction types:
IS TE
> BAAR (Anova, df = 2, p.39),
fillers not
different from either type
• Difference between verb categories:
Experiencer verbs < all Agent-Patient verbs (Anova, df
= 4, p.000. Post-hoc Bonferroni: V Exp < others,
p<.007)
1. V opt. transitive Agent Patient (mean Z =.391)
2. V transitive Agent Patient (mean Z =.229)
3. V intransitive, implied obj./patient (mean Z = .201)
4. V transitive Experiencer subject –pass (mean Z =-.705)
5. V transitive Experiencer subject +pass (mean Z =-.533)
-BAAR and IS TE: experiment
IS TE and BAAR over different verb types: adults
-BAAR and IS TE: experiment
Main results: children
• Reliable test (Cronbach’s α = .93)
• No significant difference between construction
types (Anova, df = 2, p.39),
• Difference between verb categories:
Experiencer verbs < all Agent-Patient verbs
• Mean Z-scores:
> V opt. transitive Agent Patient :
.296
> V transitive Agent Patient :
.087
> V intransitive, implied obj./patient .091
> V transitive Experiencer –pass:
-.374
> V transitive Experiencer +pass:
-236
-BAAR and IS TE: experiment
IS TE and BAAR over different verb types: children
-BAAR and IS TE: experiment
IS TE and BAAR over different verb types: adults
Acceptable passive = higher acceptability of IS TE
(compared to BAAR)
Experiment: summary
• Stimulus Experiencer verbs are clearly
worse than other categories
> Speakers have a mental representation
of the constructions, which includes
information about the types of verbs
preferred.
> Both constructions are productive
• On Stimulus Experiencer verbs with a
(marginally) acceptable passive, IS TE
scores much higher than BAAR
> Constructions are not entirely
synonymous
Thank you!
If you have any questions, comments or
suggestions, do not hesitate to get in
touch
[email protected], [email protected]