PosterPDF Available

Acceptability of relative clause extraposition in English: Effects of predicate type and givenness

Authors:

Abstract

Abstract: In English, subject-modifying restrictive relative clauses (RCs) are typically adjacent to their antecedents (e.g. some swimmers who had no adult supervision appeared), but can also be extraposed, appearing after the verb phrase (e.g. some swimmers appeared who had no adult supervision). Using informal judgments, Guéron (1980) and Rochemont and Culicover (1990) argue that relative clause extraposition (RCE) sentences are most acceptable when serving a presentational function, i.e. when describing the appearance of some entity into the discourse. They observe that intransitive verbs of appearance (arrive) are generally more acceptable with RCE than other intransitive verbs (smile). Transitive verbs (break something) are claimed to be even less acceptable due to their typical assertive function (Kuno & Takami, 1997). Rochemont and Culicover (1990) further propose that non-appearance verbs in RCE sentences are rendered more acceptable when the predicate is discourse given. Using a judgement task, Walker (2013) showed that RCE sentences in isolation were more acceptable with appearance verbs compared to other intransitive verbs; however, acceptability of RCE with transitive verbs, and the interaction between givenness and predicate type, has not been formally tested. The current study examines the effects of predicate type and discourse status on RCE using an acceptability judgment task. Hypotheses: (H1) Canonical sentences > RCE sentences in all conditions (Francis, 2010). (H2) In the RCE order, appearance verbs will receive higher ratings than intransitive non-appearance verbs, which will receive higher ratings than transitive-reflexive verbs (hurt themselves) (Walker, 2013). (H3) In the RCE order, acceptability for intransitive non-appearance and transitive-reflexive verbs will improve in the given condition (Rochemont & Culicover, 1990). Forty-eight participants rated sentences containing a subject-modifying RC presented with a preceding context. Three factors were manipulated resulting in 12 conditions: word order (RCE, canonical), verb type (appearance, intransitive non-appearance, transitive-reflexive) and discourse status of the predicate (new, given). Each participant rated one sentence from each condition, five practice items, and 24 filler sentences. Order of presentation was pseudorandom. Data were analyzed using a linear mixed model. H1 was supported: canonical sentences were rated higher than RCE sentences (F = 86.32 (1, 506), p < .0001). H2 was partially supported: RCE sentences containing appearance verbs were rated higher than those containing other verb types (F = 7.83 (2, 506), p = .0004); However, transitive-reflexive verbs were no less acceptable with RCE than intransitive non-appearance verbs. The latter finding is unexpected given the assertive function typically associated with transitivity (Guéron, 1980). Following Kuno and Takami (1997), we argue that this finding can be best explained by the reflexive object’s discourse givenness and its reduced semantic content compared to a typical non-reflexive object. H3 was not supported: RCE acceptability did not improve in the given context. Our findings fail to support the proposal that non-appearance verbs become more acceptable in RCE when previously mentioned but extend previous findings on predicate type by showing that transitive-reflexive verbs are moderately acceptable with RCE. Furthermore, effects of predicate type similar to those found by Walker (2013) are also found for contextualized sentences.
§Subject-modifying relative clauses are typically adjacent to their antecedent (canonical
order) (1) but can also be extraposed (relative clause extraposition (RCE) order) (2)
1) Some swimmers who had no adult supervision appeared.
2) Some swimmers appeared who had no adult supervision.
§Acceptability and occurrence of RCE is constrained by factors including predicate
type and givenness of the predicate.
§Predicate type: RCE is highly acceptable with predicates of appearance (3) (Walker,
2013; Guéron 1980) and typically less acceptable with intransitive non-appearance
predicates (4) and transitive predicates (5)
3) Some archeologists emerged who had explored the tomb.
4) ?Some archeologists slipped who had explored the tomb.
5) ??Some archeologists introduced themselves who had explored the tomb
§Givenness: Non-appearance verbs may be ‘coerced’ into a presentational reading when
the predicate is discourse given (6), predicting that RCE will be more acceptable when
the predicate has been previously mentioned (Francis & Michaelis, 2014; Rochemont
and Culicover, 1990)
6) a. ?A man screamed who wasn’t wearing any clothes
b. Suddenly there was the sound of lions growling. Several women screamed.
Then a man screamed who was standing at the very edge of the crowd.
(Rochemont and Culicover, 1990, p. 65)
§Acceptability of RCE with transitive verbs and discourse given predicates has not
been formally tested.
Acceptability of relative clause extraposition in English: Effects of predicate type and givenness
Joshua D. Weirick • Elaine J. Francis
Purdue University
UK Cognitive Linguistics Conference 2020
UK Cognitive Linguistics Conference, July 27th 29th , 2020 jweiric@purdue.edu
Background
Methods
DISCUSSION
ScreenHunter_20 Sep. 10 00.48
[+voice] [-voice]
VOT
f0
VOT
f0
English
short lag or
lead/prevoiced
(depending on
the speaker)
lower f0
long lag
higher f0
French
lead/prevoiced
lower f0
short lag
higher f0
)
Initial Stops second set
Discussion
Research Questions and Hypotheses
§RCE sentences with appearance verbs were significantly more acceptable than those
with other verb types (Fig. 2&5), consistent with results from Walkers (2013)
judgement experiment on RCE sentences in isolation.
§Acceptability of RCE did not improve in the given condition (Fig. 3&6), though
there was a significant interaction between givenness and verb type.
§In sum, no consistent effect of givenness was observed in the RCE order. Our
findings fail to support the proposal that RCE sentences with non-appearance verbs
become more acceptable when the predicate has been previously mentioned (at least,
not when other factors, such as prosodic cues, are absent).
Experiment 2
Figure 3: Mean ratings, verb type
by givenness in the RCE order only.
There was no significant main
effect of givenness and no
interaction between givenness and
word order.
n.s.*** ***
Lifeguards at the neighborhood pool are very busy during the summer. Kids will
run across the slippery pool deck, scrape their knees on the rough edges of the pool,
and appear at the pool alone. Just last week, a lifeguard had her hands full when
[some swimmers appeared who had no adult supervision].
Predicate
givenness: given
Verb type: ‘appearance’ Word order: RCE
The company executives were sitting nervously in the board room when suddenly
the director of the legal department arrived.
Then a lawyer arrived who represented the company whistleblower.
Word order: RCE
Verb type:
‘appearance
Predicate givenness:
given
Figure 2: Mean ratings, word order
by verb type. Significant word
order*verb type interaction
(F=7.8336, p=0004462). Transitive-
reflexive and intransitive non-
appearance verbs were significantly
less acceptable in RCE order
compared to canonical order, while
there was no difference in
acceptability between the canonical
and RCE orders for sentences with
appearance verbs.
Task
Likert
-style acceptability judgement on a 7-point scale
Participants
Ex1 (n=48): English speakers
recruited from Purdue
Ex2 (n=60): English speakers recruited online via Prolific
Factors and
levels (fully
crossed)
Word order
: RCE, canonical (CAN)
Verb type:
appearance, intransitive non-appearance, transitive-
reflexive (+unergative in Ex2)
Givenness of the predicate:
new, given
Ex1 verb
tokens
Appearance Non-appearance Transitive-reflexive
Appear
Arrive
enter
emerge
arise
materialize
fall
remain
survive
slip
stumble
depart
remove
protect
hurt
identify
Ex2 verb
tokens
Appearance
emerge
arrive
materialize
appear
Non
-
appearance
slip
depart
stumble
fall
Transitive
-reflexive
identify
excuse
introduce
remove
Unergative
laugh
scream
wave
smile
Statistical
analysis
Linear mixed model using lme4 (Bates et al., 2015) in R (R Core
Team, 2019). Independent variables and their interactions as main
effects with random intercepts for participants and items. (the data
from Ex1 were also analyzed using an ordinal logistic regression,
which produced the same significant main effects and interaction).
Experiment 1
Research Questions:
1. What are the effects of word order, discourse givenness, and predicate type on sentence
acceptability ratings?
2. Are there interactions among word order, discourse status and predicate type in their
effects on overall sentence acceptability ratings?
Hypotheses:
1. Canonical sentences will be more acceptable than RCE sentences across conditions (Francis,
2010).
2. In the RCE order, appearance verbs (e.g. appear, arrive) will receive higher ratings than
intransitive non-appearance verbs (e.g. fall, remain) and (in Ex2) unergative verbs, which will
receive higher ratings than transitive-reflexive verbs (e.g. hurt themselves) (Walker, 2013).
3. In the RCE order, acceptability for intransitive non-appearance, transitive-reflexive, and (in
Ex2) unergative verbs will improve in the given condition (Rochemont & Culicover, 1990).
Figure 1: Ex1 sample item and factor manipulations
Figure 5: Mean ratings,
word order by verb type.
Significant word
order*verb type
interaction (F=6.0188,
p=.000465). Transitive,
non-appearance and
unergative verbs
significantly less
acceptable in RCE order
compared to CAN.
Figure 6: Mean ratings,
verb type by givenness in
the RCE order only. No
significant advantage for
given predicates in the
RCE order. In fact, there
was a significant verb
type*givenness
interaction such that
appearance verbs and
unergative verbs were
more acceptable in the
new condition. This can
be seen in the RCE order
in Fig. 6, contrary to
prediction.
Figure 4: Ex2 sample item and factor manipulations
**
***
***
n.s.
* denotes < .05
* denotes < .05
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.