When Pragmatics Helps Syntax: An Eye Tracking Study on
Scope Ambiguity Resolution in 4- to 5-Year-Old Children
Karoliina Lohiniva and Daniele Panizza
1. Introduction*
1.1. Theoretical Background
Sentences with multiple scope-taking operators often lead to
interpretational ambiguity. Consider the sentence in (1).
(1) All pirates did not return to the ship
The sentence in (1) contains a subject-position universal quantifier all and
negation, and it can be assigned two interpretations. The surface scope
interpretation is exemplified in (2a), and corresponds to a reading where the
operators are interpreted in the order in which they appear in the sentence.
Under the inverse scope interpretation, the operators are interpreted in an
inverse order with respect to the sentence, as exemplified in (2b).
(2) a. surface scope: All pirates are such that they did not return to the ship
∀x [pirate(x) → ¬ return_to_the_ship(x)]
b. inverse scope: It is not the case that all pirates are such that they
returned to the ship
¬∀x [pirate(x) → return_to_the_ship(x)]
While (2a) is only true in a context where none of the pirates returned to the
ship, (2b) is true when no pirates or some but not all of the pirates returned.
Notice that (2a) asymmetrically entails (2b): in each context in which (2a) is
true, (2b) will be true as well, but not vice versa. The surface scope
interpretation is thus logically stronger than the inverse scope interpretation.
The sentence in (1) may also receive a third interpretation that is
semantically distinct from (2a) and (2b), and comes about through pragmatic
strengthening. As the interpretation in (2b) enters in a scalar relationship with
the logically stronger alternative (3a) in which the quantifier no is substituted for
not all, given the negative scale <not all, no> (cf. Horn 1989, Levinson 2000),
Karoliina Lohiniva, Université de Genève, and Daniele Panizza, Georg-August-*
Universität Göttingen, CRC “Textstrukturen” (daniele.panizza@gmail.com).
© 2016 Karoliina Lohiniva and Daniele Panizza. Proceedings of the 40th annual Boston
University Conference on Language Development, ed. Jennifer Scott and Deb Waughtal,
216-228. Somerville, MA: Cascadilla Press.
the scalar algorithm can generate the strengthened interpretation in (3b) via an
indirect scalar implicature (Chierchia 2004).
(3) a. stronger alternative of (2b):
No pirate returned to the ship
¬∃x pirate(x) → ¬ return_to_the_ship(x)
b. strengthened interpretation of (2b):
It is not the case that all pirates are such that they returned to the
ship, but some of them did.
¬∀x [pirate(x) → return_to_the_ship(x)] &
∃x [pirate(x) & return_to_the_ship(x)]
Critically, the strengthened interpretation in (3b) is only true in a scenario in
which some but not all of the pirates returned to the ship. It is logically stronger
than (2b), but semantically distinct from the surface scope reading in (2a): the
truth of the strengthened (3b) entails the falsity of (2a).
1.2. Previous developmental studies
Children’s interpretation of scopally ambiguous sentences has been the
topic of extensive research in the last two decades, and the questions concerning
children’s capacity to access both scopal interpretations of (1) and their possible
preferences for one or another interpretation have received different answers.
Early studies (Musolino 1998, Musolino, Crain and Thornton 2000) found that
English-speaking children almost uniquely adopt surface scope interpretations
of sentences including subject-every and negation. Musolino et al. (2000)
investigated 5-year-old children’s interpretation of sentences such as (1) using
the Truth-Value Judgment Task (TVJT, Crain and Thornton 1998). In the TVJT,
an experimenter-controlled puppet describes a scene acted out with props and
toys. Participants must judge whether the puppet's description of the visual
scene is right or wrong. Musolino et. al found that when the experimental
sentence was used as a description of a story in which the surface scope
interpretation was false and the inverse scope interpretation was true, adults
accepted the sentence uttered by the puppet, while children consistently rejected
it. For example, in a story where three horses had debated jumping over a fence,
and only two of them consequently jumped, the subjects had to decide whether
the sentence ‘Every horse didn’t jump over the fence’ was true or not. Children
uniformly rejected the puppet’s statement as a true description of what happened
in the story, which Musolino et al. took as evidence for them not being able to
access the inverse scope interpretation. The reasoning behind this conclusion
was based on the Principle of Charity (Grice 1975, Davidson 1984): if both
scopal interpretations are available and one of them makes the sentence false,
the interpreter will tend to select the one making the sentence true. Therefore, as
217
children rejected the sentence, they could not access the interpretation that
would have made the sentence true.
Further experimental work revealed that under certain contextual
manipulations, children fare better at accessing inverse scope interpretations.
Gualmini et al. (2008) and Musolino and Lidz (2006) first showed that while
children are in fact capable of assigning an inverse scope interpretation to
sentences like (1) and (3), their doing so depends on the experimental context.
Gualmini et al. argued that the previous results were due to the infelicity of the
negative sentences used in these experiments, as they did not entail an answer to
the Question Under Discussion (QUD). They ran an experiment using the TVJT
where both surface and inverse scope interpretations were made salient and
constituted good answers to the QUD, while using a prosodic contour that is
naturally associated with inverse scope interpretations. They found that children
accepted the inverse scope scenario in 80% of the trials. Musolino and Lidz
(2006) also obtained a boost in the inverse scope acceptance rate from 15% to
55% by presenting affirmative statements involving a universal quantifier before
the test sentences, and Viau, Lidz and Musolino (2010) saw a rise from 20-40%
to 80% of inverse scope interpretations by semantically priming inverse scope.
Finally, Conroy, Lidz and Musolino (2009) reported a split pattern of results
between 4- and 5-year-old children with the rate of inverse scope interpretations
being 81% in the younger group vs. 44% in the older, and suggested that while
both groups struggle with revising the initial parse, only 5-year-old children
show an early preference for surface scope. For this reason, younger children
displayed adult-like behaviour and the developmental trajectory of scope
assignment was argued to follow a developmental U-curve.
1.3. Four factors that can influence scope ambiguity resolution
If children are able to derive an inverse scope interpretation, why do they
fail at deriving it in certain contexts? Which factors affect their comprehension
of scopally ambiguous sentences? Several answers to these questions have been
put forward in the literature. Gualmini et al. (2008) maintain that such failure is
due to children’s pragmatic limitations, which can be overcome by designing
experiments so that the QUD is sufficiently salient. Musolino and Lidz (2006)
and Viau, Lidz and Musolino (2010), on the other hand, propose that neither
grammatical competence nor pragmatic competence on their own can be the
source of the observed behaviour. Limited processing sources may prevent
children from accessing an inverse scope interpretation, and contextual
facilitation may help in overcoming this limit. Given these considerations, we
aim to test the following hypothesis through the current study:
(4) The Processing Limitations Hypothesis:
The cognitive capacities (e.g. working memory) required for structural
218
revision are limited in children. Since inverse scope interpretations require a
revision of the initial parse, children can only access an inverse scope
interpretation if they are provided with some kind of facilitation.
The hypothesis above relies on two assumptions. First, it is assumed that
the surface scope interpretation is derived first, and the inverse scope
interpretation is derived through a re-analysis of the initial parse. Second, it is
assumed that children struggle at performing structural revisions, as was found
in other developmental studies (cf. Trueswell et al. 1999). This is why they only
derive inverse scope interpretations when provided with pragmatic or semantic
facilitation.
A second factor that has been argued to influence children and adults’ scope
choices is prosody. Büring (1998) provides a description of the intonation
assigned to the readings in (2a) and (2b) where the former is associated with
falling intonation and a single main stress, and the latter presents a fallingraising
contour with two stresses. Syrett, Simon and Nisula (2014) found that
prosody affects adults’ interpretation of scopally ambiguous sentences, while
children have been shown to be insensitive to contrastive stress (McDaniel and
Maxfield, 1992; Gualmini, Maciukaite and Crain, 2003) .
Finally, a third factor that could influence children’s performance in
interpreting sentences like (1) is the relationship between accessing vs.
preferring a given reading. Take, for instance, Musolino’s (1998) original
findings. If we assumed that children can in fact access both scopal
configurations but they have a strong surface scope preference for some reason
(e.g. it is easier to process, more frequent in the linguistic input, etc.), the
expected behaviour in the critical trials where two out of three horses jump over
the fence would be identical to that reported in the study: a consistent rejection
of the puppet’s statement. Therefore, disentangling access vs. preference of each
possible reading is critical in order to rule out this alternative interpretation.
Summing up, we identify four possible factors influencing children’s
interpretation of scopally ambiguous sentences such as (1): i) possible
processing limitations and available parsing strategies, ii) the saliency of the
QUD, iii) prosody, and iv) the influence of an interpretation preference on the
final response. In the next section, we illustrate the present experiment, designed
to take these factors into account.
2. The current experiment
The goal of the present study is to explore 4- to 5-year-old German
children’s interpretation of (1) through an experimental design that allows
controlling for the factors discussed in the previous section. To achieve this
goal, we employed a modification of the TVJT that we called the Semantic
Choice Task (SCT). The main difference consists in the simultaneous
219
presentation of two contextual scenarios (SCT) instead of one (TVJT). The
felicity of the negative sentence is controlled for, as more than one good answer
to the QUD is visually prominent in the setup and corresponds to an actual
choice option. The development of a sharp experimental design with the SCT
permitted investigating separately the issues of access vs. preference of a given
interpretation, and the presentation of previously recorded test sentences
allowed controlling for prosody. By recording the participants’ eye movements
using eye-tracking, we could observe whether they display an initial preference
and whether one interpretation requires more cognitive effort to be derived.
2.1. Procedure and participants
The participants were 45 4- to 5-year-old German speaking children (mean
age = 61.6 months) and 50 adult controls, who took part in a computer version
of the experiment. At the beginning of each session, which lasted about 45
minutes, a German-speaking experimenter introduced the children to the
experimental setup. A second experimenter was responsible for acting out the
story using toys. The participants were told that they were about to hear a story
about some pirates and their adventures on the Fantastic Island. They were then
informed that the pirates were actually actors at a theatre, belonging to two
different groups: the Red pirates and the Green pirates. The task of the subject
was to help the theatre director, a character that communicated with the
experimenter and the subject via the telephone, in rewarding the group of actors
whose acting matched more closely with the instructions of the director. The
pre-recorded experimental sentences were uttered by the director during his
calls and played to the participants through speakers.
The experiment was conducted on an experimental table, divided into two
“theatre scenes” on which the actors played. The eye-tracker was positioned in
the middle, in front of the subject sitting by the table with the experimenter.
Before the start of the session, a 3-points calibration of the eye-tracking system
was performed. After calibration, the experiment started with the director calling
the experimenter to present himself to the subject and the experimenter. Two
warm-up trials were then presented to the participants. The first warm-up trial
followed the Red and Green pirates’ ships being suddenly caught in a strong
storm. The shaking of the ships caused all of the Green pirates to fall into the
sea, while only three out of five of the Red pirates fell. The director called and
uttered a warm-up sentence with a subject-position universal quantifier alle but
no negation (“All of the pirates fell from the ship”). The experimenter then
asked the participant to give a coin reward to the group of pirates whose acting
corresponded better to the sentence the director uttered. Regardless of whether
the child correctly rewarded the Green pirates or incorrectly rewarded the Red
pirates, the experimenter then asked the subject how many pirates in each group
fell into the sea, repeated the warm-up sentence, and asked again which team
220
performed better. When the subject rewarded the correct team, acknowledging
that all of them fell, the experimenter moved on to the second warm-up trial.
After the warm-up trials, the subjects saw 16 experimental trials where the
same procedure was used, except that the experimental sentence was only
repeated if the child refrained from providing an answer (which happened very
rarely). The computer version of the experiment also included a no-reward
option, so as to verify whether the matching between sentences and scenarios
was judged felicitous by adult controls. At the end of the experiment, the subject
identified the winning group of pirate actors by counting the teams’ coins.
Children received a book for participating, and adults received 5€.
2.2. Experimental Design
The experimental design consists of three test conditions involving
sentences with negation (nicht) and a subject-position universal quantifier (alle),
as in (5), and one control condition involving sentences with an existential
quantifier einige ‘some’ but no negation.
(5) Alle Piraten sind nicht auf das Schiff zurückgekehrt.
‘All of the pirates did not return to the ship.’
Two conditions investigate access to the surface, inverse and strengthened
inverse scope interpretations of (1) (described in (2a), (2b) and (3b),
respectively). The NONE-FALSE condition includes a NONE scenario (Fig.
1a), in which no pirate returned to the ship, and a FALSE scenario (Fig. 1b), in
which all pirates did. The NONE scenario is compatible with surface and nonstrengthened
inverse scope interpretations, and a correct reward response means
that either of those interpretations was adopted. The SOME-FALSE condition,
in contrast, includes the same FALSE scenario vs. a SOME scenario (Fig. 1c), in
which some but not all of the pirates returned. The SOME scenario is
compatible with an inverse scope interpretation with or without a scalar
implicature, but not with a surface scope interpretation. Thus, correctly choosing
Figure 1a (NONE) Figure 1b (FALSE) Figure 1c (SOME)
221
the SOME scenario means an inverse scope interpretation was adopted. Finally,
the NONE-SOME condition investigates the participants’ preference for one of
the two possibly true scenarios. If children only access the surface scope
interpretation of (1), they should consistently prefer the NONE scenario over the
SOME scenario. If they prefer the SOME scenario or display no preference,
however, we have evidence that they were able to access an inverse scope
interpretation.
The experimental sentences were recorded by a German-speaking actor
with a prosody that is not biased towards any of the scopal interpretations. This
intonation was generally flat, involving a mild accent on the two nouns (i.e.
Piraten and Schiff) and, critically, no focal stress on either negation (nicht) or
the universal quantifier (alle). Adults reported that the sentences sounded quite
natural despite the adopted unbiased prosody.
Eye movements were monitored through an Eye-link remote system with a
1000 Hz tracking resolution. Stimuli presentation, data processing and statistical
analysis were performed with Python (www.python.org) and R (www.R-
project.org).
2.3. Predictions
First of all, the experiment was designed to test whether German 4- and 5year-old
children can access inverse scope interpretations of sentences like (5)
in the absence of contextual, semantic or prosodic bias. If children perform
above chance in the SOME-FALSE access condition, the results provide further
empirical evidence that inverse scope is accessible to young learners.
Second, if German children prefer surface scope over inverse scope, as
found for English children in many previous studies, a preference for the NONE
scenario over the SOME scenario is expected in the preference condition.
Moreover, we expect to find an overall facilitation for surface scope responses
throughout the conditions. That is, we predict to obtain higher accuracy in the
NONE-FALSE access condition compared to the SOME-FALSE condition, for
the former is compatible with both surface and inverse scope interpretations.
Third, we expect to find faster on-line target disambiguation for NONE
scenarios vs. SOME scenarios either because surface scope is derived first (cf.
the Processing Limitations Hypothesis) or because whichever interpretation is
accessed first, the NONE scenarios but not the SOME scenarios are compatible
with both interpretations.
In general, the Processing Limitations Hypothesis predicts higher off-line
accuracy and faster on-line disambiguation in the NONE-FALSE condition
compared to the SOME-FALSE condition, as well as a preference for the NONE
scenario in the NONE-SOME condition.
Finally, although our experiment was not designed to contrast 4- to 5-yearold
children, the children’s performance might be found to vary across age with
222
respect to access to inverse scope interpretations. If younger children perform
better than older children, the hypothesis advanced by Conroy et al. 2009 will
find empirical support.
3. Results
3.1. Off-line results: Semantic choice
Adults’ semantic choice results show an overall preference for the NONE
scenario, compatible with both surface and inverse scope. In the NONE-FALSE
condition, adults selected the NONE scenario in 80% of the trials, the no-reward
option in 12% of the trials and the wrong FALSE scenario in 8% of the trials. In
the SOME-FALSE condition, they selected the SOME scenario in 70% of the
trials, the no-reward option in the 25% of the trials and the FALSE scenario in
5% of the trials. The difference between the two access conditions was tested in
a Generalized Linear Mixed Model (GLMM) and resulted in a significant main
effect of condition (p=.035). In the preference condition, adults showed a slight
preference for the NONE scenario (48% vs. 37% of choices) and selected the
no-reward option in 15% of the trials. In sum, the controls display a generally
high accuracy and found the test sentences more felicitous when evaluated in a
NONE scenario compared to a SOME scenario.
Strikingly, children display a rather different pattern of results. Contrary to
our expectations and in contrast with what was found in the control group, they
show an overall facilitation and preference for the SOME scenario, which is
only compatible with an inverse scope interpretation. As children were forced to
make a choice, we only report accuracy for the access conditions. Children’s
reward responses were correct in 65% of the trials in the NONE-FALSE
condition, and 79% of the trials in the SOME-FALSE condition (GLMM main
effect of condition: p=.000). Thus the rate of FALSE answers amounts to 35%
and 21 % in the NONE-FALSE and SOME-FALSE access conditions
respectively. Furthermore, in the preference condition, they showed a slight
preference for the SOME scenario (56% of SOME answers).
To investigate whether the profile of decision choices varied across age, we
performed a median split into two groups of children (younger than 62 months
(n=22) and older than 62 months (n=23)). Both groups showed higher accuracy
for the SOME vs. NONE scenario in the access conditions (young: NONE =
57% vs. SOME = 72%; old: NONE = 73% vs. SOME = 84%). In the preference
condition, the younger group selected the SOME scenario in 60% of the trials,
while the older group did so in 54% of the trials.
Finally, to determine whether these results were uniformly determined by
all participants regardless of their comprehension rate, we conducted a further
analysis by setting up a criterion of at least 3/4 correct answers in both access
conditions, excluding children who did not fully understand the experimental
sentences (17 out of 45). After exclusion, accuracy rose to 92% in the NONE-
223
FALSE access condition and to 93% in the SOME-FALSE access condition,
whereas the preference condition results showed no preference at all for either
scenario (49% of NONE answers). The facilitation effect for the SOME scenario
was thus due to the poor comprehenders. In contrast, the 28 children with good
comprehension did not show any facilitation or preference for either scenario.
3.2. On-line results: eye movement data
The aim of the eye movement analysis was to investigate how quickly
participants identify the correct scenario (i.e. the NONE/SOME scenario over
the FALSE one) and whether they display any difference in the two access
conditions. We will first report the target preference analysis, which considers
the proportion of fixations to the target (i.e. the correct scenario) in a given time
window divided by the total fixations towards both target and distractor (i.e. the
false scenario). This analysis was performed only on correct answers. In Figure
2 and 3, the target preference for the young (<62 months) and old (>62 months)
children is plotted over the relevant time windows. When the target preference
departs from the 0.5 line, representing the chance level, the participants
developed a preference for the target in their on-line gaze.
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
NONE scenario
SOME scenario
all the pirates have not to the ship returned ....
Figure 2. Target preference in children older than 62 months
0.7
0.6
0.5
0.4
0.3
0.2
0.1
NONE scenario
SOME scenario
all the pirates have not to the ship returned ....
Figure 3. Target preference in children younger than 62 months
224
As shown by the graphs above, from the time region that includes the
prepositional phrase (PP, i.e. ‘to the ship’), which immediately follows negation
(nicht), the older children display a steady preference for the target only in the
NONE-FALSE condition. The same group in the SOME-FALSE condition, as
well as the younger children in both conditions, display a consistent preference
for the target only from the following region that includes the past participle
(zurückgekehrt ‘returned’). This difference results in a significant interaction
between age and condition (p=.34) as well as a main effect of condition (p=.43)
in the PP region. These data suggest that the on-line disambiguation of the
NONE scenario, compatible with both surface and inverse scope interpretations,
took place more quickly than that of the SOME scenario.
The second analysis that we present is the shift latency analysis, which
computes the delay in making a shift from the FALSE to the NONE/SOME
scenario after hearing the critical word nicht. Older children took on average
1218 ms to shift their looks towards the target in the NONE-FALSE condition
vs. 1777 ms in the SOME-FALSE condition. Younger children took 1714 ms to
shift their looks towards the target in the NONE-FALSE condition vs. 1895 ms
in the SOME-FALSE condition. The statistical analysis yields significant main
effects of age (p=.37) and condition (p=.42), but no significant interaction. The
shift latency analysis confirms what was found in the target preference analysis,
and brings further support to the idea that a) there is an on-line facilitation for
the NONE scenario over the SOME scenario, and b) this effect is brought by
both subgroups of children, although it is more prominent in the older children.
4. Discussion
We illustrated an experimental study investigating the off-line interpretation
and on-line processing of structurally ambiguous German sentences including a
subject-position universal quantifier and negation in 4- to 5-year-old children
and adults. This experiment attempts to overcome the limits of previous studies
in the following respects: a) it controls for prosody by pre-recording test
sentences with unbiased intonation, b) it presents two visual scenarios
simultaneously and allows for separate investigation of accessing vs. preferring
different scope interpretations, and c) it provides simultaneously collected offline
and on-line data, crucial for addressing processing-related questions.
The first aim of this study was to investigate whether German children
access inverse scope interpretations of sentences like (5) in the absence of
contextual, semantic or prosodic bias. The answer to this question is affirmative.
In the SOME-FALSE access condition, children displayed a very high accuracy
rate, comparable to the adult rate, in rewarding the scenario that was only
compatible with an inverse scope interpretation. The possibility that this
outcome be due the prosody that we used, which could have favoured an inverse
scope interpretation, can be ruled out: as the adult rate of no-reward answers
225
was higher in the SOME-FALSE condition compared to the NONE-FALSE
condition, which means that the sentence was judged less felicitous in the
former condition, the adults' responses attest that the adopted prosody did not
bias towards an inverse scope interpretation. These results provide empirical
evidence in favour of the claim put forward by Gualmini et al. (2008), who
stated that children can access an inverse scope interpretation if the QUD is
made salient by the experimental context. Indeed, we demonstrated the validity
of this claim even in the absence of biased prosody towards inverse scope.
The second main finding is that, contrary to our expectations, the scenario
that is only compatible with an inverse scope interpretation had a higher overall
accuracy rate than the one compatible with both surface and inverse scope.
Remarkably, the higher accuracy for SOME-FALSE vs. NONE-FALSE access
conditions was significantly present in both age groups. This result cannot be
explained in terms of higher visual prominency or easier processing of the
SOME scenario: as witnessed by the eye movement analyses, the NONE
scenario was disambiguated faster than the SOME scenario by both groups of
participants. Why is it, then, that the SOME scenario was more difficult to
identify, but at the same time it’s selection was less error-prone in poor
comprehenders?
The answer to this question is by no means a simple one if we consider that
the NONE scenario is always compatible with a non-strengthened inverse scope
interpretation. The only reading of (5) that is true in the SOME scenario but
false in the NONE scenario is the inverse scope interpretation enriched by an
indirect scalar implicature, namely, ‘not all the pirates returned to the ship but
some of them did’. The hypothesis that we advance to explain this finding is that
the scalar implicature provided a pragmatic boost, facilitating comprehension
and identification of the correct scenario in poor comprehenders. While we
know from a large experimental literature that children are generally poor at
calculating scalar implicatures (cf. Noveck 2001), some findings suggest that
children are indeed able to derive scalar implicatures if the alternatives are made
salient in the context (Papafragou and Tantalou 2004). Furthermore, recent
findings by Bill et al. (2014) suggest that children’s comprehension of indirect
implicatures (e.g. ‘not all’ implicates ‘not all but some’) is better than that of
direct scalar implicatures (e.g. ‘some’ implicates ‘some but not all’). Thus, a
pragmatic boost might well be the source of this finding.
The plausibility of this explanation can be further investigated by
examining what happens on-line. Eye movement results show that the SOME
scenario is disambiguated generally later than the NONE scenario. The target
preference analysis shows this effect only for older children, whereas the shift
delay analysis suggests that both age groups manifest this behaviour although it
is more prominent in older children. If we assume that pragmatic strengthening
requires the derivation of an inverse scope interpretation, and that this derivation
results from the revision of a surface scope interpretation, we find that the on-
226
line results support this explanation. If so, the facilitating effect of pragmatic
strengthening must take place late, when the difference between the SOME vs.
NONE scenarios is revealed in the participants. Under an alternative
explanation, surface and inverse scope interpretations are both immediately
available, and the delay displayed by the SOME scenario reflects the cognitive
cost of deriving a scalar implicature. In this case, the hypothesis that pragmatic
strengthening facilitates the comprehension process is supported as well.
A second goal of this study was to seek evidence in favour or contra the
Processing Limitations Hypothesis which assumes that surface scope is derived
first and inverse scope requires a structural revision of the parse. This
assumption first seems to find support in our results, given that the NONE
scenario is disambiguated more rapidly than the SOME scenario. However,
greater ease of disambiguation with the NONE scenario does not lead to higher
accuracy in semantic choice. Instead, we find the opposite pattern, with access
to the SOME scenario resulting in a facilitation for poor comprehenders, while
the good comprehenders display neither facilitation nor preference for any of the
two scenarios. Thus, not only does the Processing Limitations Hypothesis not
find full support in our results, but we might also end up questioning the
assumption that surface scope is derived first. As mentioned above, there could
be another factor slowing down the target identification in the SOME scenario
vs. the NONE scenario, which is the derivation of a scalar implicature.
Finally, our results do not align with those reported by Conroy et al. 2009,
who found 4-year-old children to perform better than 5-year-old children at
deriving inverse scope. While our younger subjects show a high accuracy rate
for the scenario that is only compatible with inverse scope (the SOME scenario)
in the off-line choices, the older subjects’ accuracy rate was even higher.
To conclude, we demonstrated that 4- and 5-year-old German-speaking
children access both surface and inverse scope interpretations of (5) in the
absence of contextual, semantic or prosodic bias. Furthermore, we found that
inverse scope interpretations boost the accuracy of poor comprehenders,
whereas scenarios compatible with both surface and non-strengthened inverse
scope are faster to disambiguate, especially in older children. We propose that
these effects can be explained in terms of a pragmatic boost carried over by the
derivation of an indirect scalar implicature. Further research is needed to
provide more evidence for this hypothesis and to determine whether the results
reported in this study are replicable in other languages, as would be expected.
References
Bill, Cory, Jacopo Romoli, Florian Schwarz and Stephen Crain. 2014. Scalar
Implicatures Versus Presuppositions: The View from Acquisition [Special Issue].
Topoi: Presuppositions: Philosophy, Linguistics and Psychology. In Press.
227
Büring, Daniel. 1997. The Great Scope Inversion Conspiracy. Linguistics and Philosophy
20:2, 175–194.
Chierchia, Gennaro. 2004. Scalar implicatures, polarity phenomena, and the syntax/
pragmatics interface. Structures and beyond: The Cartography of Syntactic
Structures, ed. by Adriana Belletti, 39–103. New York: Oxford University Press.
Conroy, Anastasia, Jeffrey Lidz and Julien Musolino. 2009. The Fleeting Isomorphism
Effect. Language Acquisition 16:2, 106–117.
Crain, Stephen and Rosalind Thornton. 1998. Investigations in Universal Grammar: A
Guide to Research on the Acquisition of Syntax and Semantics. Cambridge, MA:
MIT Press.
Davidson, Donald. 1984. Inquiries into truth and interpretation. Oxford: Clarendon
Press.
Grice, Paul. 1975. Logic and conversation. Syntax and Semantics, ed. by Peter Cole and
James Morgan, 41–58. New York: Academic Press.
Gualmini, Andrea, Simona Maciukaite and Stephen Crain. 2003. Children's insensitivity
to contrastive stress in sentences with only. University of Pennsylvania Working
Papers in Linguistics 8:1, article 8.
Gualmini, Andrea, Sarah Hulsey, Valentine Hacquard and Danny Fox. 2008. The
Question-Answer Requirement for scope assignment. Natural Language Semantics
16:3, 205–237.
Horn, Laurence. 1989. A natural history of negation. Chicago: University of Chicago
Press.
Levinson, Stephen. 2000. Presumptive meanings. Cambridge, MA: MIT Press.
McDaniel, Dana and Thomas Maxfield. 1992. Principle B and Contrastive Stress.
Language Acquisition 2, 337–358.
Musolino, Julien. 1998. Universal Grammar and the acquisition of semantic knowledge:
an experimental investigation of quantifier-negation interaction in English. Doctoral
dissertation, University of Maryland.
Musolino, Julien, Stephen Crain and Rosalind Thornton. 2000. Navigating negative
quantificational space. Linguistics 38, 1–32.
Musolino, Julien and Jeffrey Lidz. 2006. Why children aren't universally successful with
quantification. Linguistics 44:4, 817–852.
Noveck, Ira. 2001. When children are more logical than adults: Experimental
investigations of scalar implicature. Cognition 78:2, 165–188.
Papafragou, Anna and Niki Tantalou. 2004. Children's Computation of Implicatures.
Language Acquisition,12:1, 71–82.
Syrett, Kristen, Georgia Simon and Kirsten Nisula. 2014. Prosodic disambiguation of
scopally ambiguous quantificational sentences in a discourse context. Journal of
Linguistics 50:2, 453–493.
Trueswell, John C., Irina Sekerina, Nicole M. Hill and Marian L. Logrip. 1999. The
kindergarten-path effect: studying on-line sentence processing in young children.
Cognition 73, 89–134.
Viau, Joshua, Jeffrey Lidz and Julien Musolino. 2010. Priming of Abstract Logical
Representations in 4-Year-Olds. Language Acquisition 17:1, 26–50.
228
Proceedings of the 40th annual
Boston University Conference
on Language Development
edited by Jennifer Scott
and Deb Waughtal
Cascadilla Press Somerville, MA 2016
Copyright information
Proceedings of the 40th annual Boston University Conference on Language Development
© 2016 Cascadilla Press. All rights reserved
Copyright notices are located at the bottom of the first page of each paper.
Reprints for course packs can be authorized by Cascadilla Press.
ISSN 1080-692X
ISBN 978-1-57473-066-1 (2 volume set, paperback)
ISBN 978-1-57473-166-8 (2 volume set, library binding)
Ordering information
To order a copy of the proceedings or to place a standing order, contact:
Cascadilla Press, P.O. Box 440355, Somerville, MA 02144, USA
phone: 1-617-776-2370, sales@cascadilla.com, www.cascadilla.com