Children`s Interpretation of Ambiguous Focus in Sentences With “Only”

LANGUAGE ACQUISITION, 13(3), 253–284
Copyright © 2005/2006, Lawrence Erlbaum Associates, Inc.
Children’s Interpretation of Ambiguous
Focus in Sentences With “Only”
Kevin B. Paterson
School of Psychology
University of Leicester
Simon P. Liversedge
Department of Psychology
University of Durham
Diane White
Department of Psychology
University of Derby
Ruth Filik
Department of Psychology
University of Glasgow
Kristina Jaz
School of Psychology
University of Leicester
We report 3 studies investigating children’s and adults’ interpretation of ambiguous
focus in sentences containing the focus-sensitive quantifier only. In each experiment, child and adult participants compared sentences with only in a preverbal position and counterpart sentences without only against a series of pictures depicting
events that matched or mismatched with the sentence meaning. The sentences with
only were ambiguous between an analysis with contrastive focus assigned to the
verb phrase (VP) and one with contrastive focus assigned to the direct object. The
results indicate that both children and adults interpreted sentences with only as excluding the possibility of events that formed a contrast with VP constituents.
Children also appeared to interpret sentences without only as excluding the possibility of these events despite the absence of grammatical cues that might indicate
contrastive focus. We consider these results in relation to a processing account of focus interpretation (Crain, Ni, and Conway (1994)).
Correspondence should be sent to Kevin Paterson, School of Psychology, Henry Wellcome Building, University of Leicester, Leicester, LE1 9HN United Kingdom. E-mail: [email protected]
254
PATERSON ET AL.
1.
INTRODUCTION
The assignment of focus in a sentence has been recognized as an important aspect
of semantic interpretation (e.g., Jackendoff (1972), Krifka (1992), Rooth (1992)).
Focus is used to syntactically mark newly asserted information or to indicate that
a contrast is to be made between current information and its alternatives. Focussensitive quantifiers, such as only and even, are associated with the assignment of
contrastive focus, meaning that these words often indicate that a contrast is to be
made between the referent of a focused sentential constituent and its alternatives.
However, the choice of focus in a sentence frequently is ambiguous. Consequently, sentences that contain a focus-sensitive quantifier often are ambiguous
between qualitatively distinct interpretations in which the quantifier associates
with a different syntactic constituent. Thus, part of the comprehension process for
such sentences will involve the reader or hearer identifying which syntactic constituent is focused and using this information to determine the nature of the contrast that is to be instantiated.
Some researchers have used the focus sensitivity of only to investigate whether
the choice of focus in a sentence can guide the processing of syntactic ambiguities
(e.g., Clifton, Bock, and Rado (2000), Filik, Paterson, and Liversedge (2005),
Liversedge, Paterson, and Clayes (2002), Ni, Crain, and Shankweiler (1996), Paterson, Liversedge, and Underwood (1999), Sedivy (2002)). Others have used it to
investigate the development of linguistic constraints on sentence interpretation
(Crain, Ni, and Conway (1994), Crain, Philip, Drozd, Roeper, and Matsuoka
(1992), Drozd (2001), Drozd and van Loosbroek (1998), Gualmini, Maciukaite,
and Crain (2002), Halbert, Crain, Shankweiler, and Woodams (1995), Paterson,
Liversedge, Rowland, and Filik (2003), Philip and Lynch (1999)). In the present
experiments, we used the focus sensitivity of only to investigate which interpretations children and adults assign to a sentence with ambiguous choice of focus.
Several factors are believed to influence the choice of focus in sentences containing a focus-sensitive quantifier. Foremost among these are syntactic constraints on the quantifier’s range, where range (or scope) refers to the set of
syntactic constituents that the quantifier can in principle operate on (Jackendoff
(1972)). Jackendoff did not define the rules governing the range of focussensitive quantifiers, but Reinhart (1999) proposed that only is syntactically restricted to taking scope over those constituents that it c-commands in a parse
tree. Thus, when only is included in a sentence, it will associate with a constituent that it c-commands. The sentences in (1) and (2) illustrate how this ccommand constraint can result in a reader or hearer assigning qualitatively
different interpretations to sentences that differ only in terms of the focussensitive quantifier’s location.
(1) Only the fireman is holding a hose.
INTERPRETATION OF AMBIGUOUS FOCUS
255
(2) The fireman is only holding a hose.
For the sentence in (1), only occurs in a presubject position, whereas for the sentence in (2), it occurs in a preverbal position. When only is in the presubject position, it c-commands those constituents that form the subject noun phrase (NP).
Consequently, the reader or hearer should interpret the sentence in (1) with only
associating with this constituent (i.e., the fireman). The nature of the contrast that
is represented will depend on the lexical characteristics of the focus-sensitive
quantifier. Some quantifiers, such as only, specify an exclusive contrast, meaning
that the referent of the focused constituent does not share an attribute with any of
its alternatives, whereas others, such as even, specify an additive contrast,
whereby the attribute assigned to the referent of the focused constituent is shared
with at least one alternative (König (1991)). Because the lexical characteristics of
only specify an exclusive contrast, the sentence in (1) should be understood as
meaning that no one other than the fireman is holding a hose.
When only occurs in a preverbal position, as it does for the sentence in (2), it ccommands those constituents that form the verb phrase (VP). Therefore, the reader
or hearer should interpret this sentence with only associating with one of these constituents. It is important to note that whereas the requirement for only to focus on a
constituent within its range disambiguates the sentence in (1), the choice of focus is
still ambiguous for the sentence in (2) because the quantifier can associate either
with the VP or with the direct object (i.e., a hose).1 When it associates with the VP,
the reader or hearer will understand the sentence as meaning that the fireman is
holding a hose and not doing anything else, whereas when it associates with the direct object, the sentence will be understood as meaning that the fireman is holding a
hose and not holding anything else. Thus, although syntactic restrictions on a quantifier’s range can restrict it to associating with a constituent within a particular syntactic domain, this may not fully disambiguate the sentence.
Sentence prosody may also help the reader or hearer to identify the focus of a
sentence. Within the formal linguistics literature, it is widely argued that prosody
is used to mark the focused constituent in a sentence, although the accounts differ
over the precise characteristics of these cues (e.g., Jackendoff (1972), Kadmon
(2001), Selkirk (1995)). According to Jackendoff, when prosody marks the focused constituent in a sentence, a focus-sensitive quantifier will associate with
this constituent if it is within its range; if prosody marks a focused constituent that
1
1There is another possible analysis of ambiguous sentences like (2), with only associating with the
verb, thereby indicating that the fireman was holding a hose and not doing anything else with a hose.
However, this analysis often has not been featured in discussions of focus identification (e.g., Crain et
al. (1994), Frazier (1999)), and data obtained in our experiments suggest that participants rarely assign
it to ambiguous sentences (< 2% of trials for ambiguous sentences with preverbal only). Consequently,
for reasons of clarity, we concentrate on interpretations with only associating with either the direct object or the VP and consider data that have a bearing on the processing of this ambiguity.
256
PATERSON ET AL.
is outside of the quantifier’s range, then the sentence is likely to be perceived as
unacceptable.2 If correct, this suggests that prosodic cues may not override syntactic restrictions on focus but that they may help to disambiguate the choice of
focus in a sentence. However, and very importantly, explicit prosodic cues are absent in silent reading (but see Fodor (2002)). Consequently, this mechanism for
the assignment of focus may be unavailable, and readers may rely on other
sources of information to identify the sentence focus.
Having identified the focus of a sentence, the reader or hearer must also determine the nature of its alternatives. Jackendoff (1972) proposed that the alternatives to the denotation of a focused constituent are computed at a semantic level of
analysis rather than as part of the sentence’s syntactic analysis and are derived
from background or presuppositional information about the discourse. According
to Jackendoff, substituting an “appropriate semantic value” for the focused information creates an alternative. Thus, if the sentence in (1) is analyzed with focus
on the subject NP, the alternatives will comprise propositions about other persons
who might have been carrying a hose. That is to say, in a context containing a fireman and policeman, the sentence will be interpreted as meaning that although the
fireman is holding a hose, the policeman is not. Jackendoff was not explicit about
the nature of the psychological processes that enable a reader or hearer to identify
the appropriate alternatives to focused information. This is entirely understandable because the question of how a reader or listener might compute alternatives is
beyond the scope of Jackendoff ’s formal approach to focus interpretation. However, determining the precise nature of this inferential process represents a significant challenge and will be an important step in understanding the psycholinguistic
processes that are associated with focus interpretation.
It is perhaps worth noting that researchers have made several attempts at producing formal theories of focus interpretation, and such accounts provide more
detailed theoretical explanations of how contrasting alternatives are computed.
2
2There
is a debate concerning whether it is possible to interpret a sentence with a focus-sensitive
quantifier associating with a constituent that is outside of its range (e.g., see discussions in Kadmon
(2001)). To illustrate this debate, consider the question and answer pair shown below in which
prosodic focus is indicated using capitals (we thank Ken Drozd for providing this example):
Did John only buy an iPod?
No, BILL only bought an iPod.
For this sentence pair, only should associate with a VP constituent. However, some readers might consider that the quantifier instead associates with Bill in the answer despite this constituent being outside
the quantifier’s scope. Our view is that this is a matter for empirical investigation. Paterson,
Liversedge, and Filik (2005) recently presented some adult data that have a bearing on this issue.
These data suggest that whereas an interrogative context can disambiguate the choice of focus within a
focus-sensitive quantifier’s range, it cannot enable readers to interpret the quantifier as associating
with a constituent that is outside of this range. The results suggest that syntax places strong constraints
on focus interpretation and that these constraints are not easily overridden in adult sentence comprehension.
INTERPRETATION OF AMBIGUOUS FOCUS
257
Such theories include the Structured Meaning Theory (e.g., Cresswell and von
Stechow (1982), Jacobs (1986), von Stechow (1981)) and the Alternative Semantics Theory (Rooth (1992)). However, all of these accounts, in fundamental terms
at least, adhere to Jackendoff ’s (1972) basic approach. From this discussion, it
should be clear that the computation of contrastive focus is a relatively complex
psychological process requiring a number of representational commitments
within the semantic representation of the text. For focus computation to occur
successfully, the reader or hearer must first identify a focused constituent that is
within the range of the focus-sensitive quantifier. They must then compute and
instantiate a set of alternatives to the focused constituent through the use of background and presuppositional knowledge about the discourse.
Despite the computation of contrastive focus being recognized as an important
aspect of language processing that might have a direct influence on the nature of
the semantic representation during comprehension, there have been relatively few
empirical investigations of the psychological processes by which it is computed.
The work of Crain and his colleagues is an important exception to this. Crain et al.
(1994) presented the most detailed processing account of focus interpretation to
date. It requires that adults resolve linguistic ambiguities, including an ambiguous
choice of focus, using a combination of syntactic knowledge and processing principles originally outlined within the Referential Theory of sentence processing
(e.g., Crain and Steedman (1985)). According to this account, readers and hearers
use syntactic knowledge to restrict a focus-sensitive quantifier to taking scope
over constituents in a particular syntactic domain. Recall that syntactic restrictions on the range of only require it to associate with a constituent that it ccommands in a parse tree. Thus, if adults employ syntactic knowledge to restrict
the range of only, they should interpret a sentence with presubject only (e.g., (1))
by assigning contrastive focus to the subject NP but interpret a sentence with
preverbal only (e.g., (2)) by assigning contrastive focus to a VP constituent.
Recall also that syntactic restrictions on a quantifier’s range do not necessarily
disambiguate its focus. For this reason, Crain et al. (1994) proposed that adults
also employ two general processing principles to resolve focus assignment ambiguities. The Principle of Referential Success stipulates that when an ambiguity
occurs within a disambiguating referential context, adults will favor an interpretation that refers to entities from that context. Thus, the resolution of focus assignment ambiguities should be guided by the availability of referents for the focused
constituent and its alternatives in the prior discourse context. The second principle, the Principle of Parsimony, governs the processing of an ambiguity when it
occurs outside of a disambiguating discourse context. In its original formulation
(i.e., Crain and Steedman (1985)), the Principle of Parsimony stipulates that the
parser incorporates the fewest unsupported referential presuppositions about entities that are not made explicit in the sentence. However, Crain et al. argued that
the principle forms part of a broader strategy for minimizing the incorporation of
unsupported presuppositional information. On this view, a reader or hearer should
258
PATERSON ET AL.
adopt the analysis of an ambiguity that makes the fewest assumptions about information that is not given in the sentence. Crain et al. referred to this broader strategy as one that involves avoiding incorporating presuppositions about the
meaning of a sentence that might need to be changed later. Furthermore, Crain et
al. considered that the processing of focus assignment ambiguities in sentences
with only in a preverbal position (e.g., (2)) provides a good example of how this
strategy might guide ambiguity resolution.
Consider the sentence in (2). This sentence is ambiguous between an analysis
with contrastive focus assigned to the direct object alone or to the VP. Crain et al.
(1994) observed that there is a particular semantic relationship between the interpretations that correspond to these two analyses such that an analysis with contrastive focus on the VP is true in a subset of the circumstances in which an analysis
with contrastive focus on the direct object is true. When contrastive focus is assigned to the direct object (i.e., a hose) for the sentence in (2), the sentence is true in
circumstances in which the only thing the fireman is holding is a hose. By contrast,
when contrastive focus is assigned to the VP, the sentence is true when the only
thing the fireman is doing is holding a hose. For the fireman not to be doing anything other than holding a hose entails that he is also not holding anything other than
a hose, whereas the reverse does not hold. Thus, an analysis of this sentence with
contrastive focus on the direct object is true in a larger number of circumstances
than with contrastive focus on the VP. Crain et al. argued that when faced with an
ambiguity of this form, the parser preferentially adopts the analysis that is true in
the largest set of circumstances because this requires it to make fewest presuppositions about information that is not given in the sentence. Thus, when an ambiguous
sentence like (2) occurs outside of a disambiguating discourse context, adults
should favor the analysis with contrastive focus on the direct object.
In developing their processing account of focus interpretation, Crain et al.
(1994) also considered how children might acquire the alternative interpretations
of a focus assignment ambiguity. Crain et al. started from the assumption that as
children do not receive explicit instruction or feedback about how to interpret ambiguous sentences, they must employ strategies that enable them to acquire the alternative analyses without instruction. Therefore, Crain et al. proposed that
children acquire the alternative analyses of a linguistic ambiguity, including focus
assignment ambiguities, by employing a learnability principle that orders the acquisition of hypotheses concerning the interpretation that is assigned to an ambiguity. Crain et al. referred to this principle as the Semantic Subset Principle.
According to this principle, if an ambiguity has two different interpretations, but
one interpretation is true in a subset of the circumstances in which the other is
true, learners will initially hypothesize the interpretation that is true in the smallest set of circumstances. In this way, the learners are assured of formulating
falsifiable hypotheses about the sentence meaning, and as they encounter situations in which the initially formulated hypothesis proves to be incorrect, they will
accumulate evidence for the existence of the alternative analysis.
INTERPRETATION OF AMBIGUOUS FOCUS
259
Crucially, the Semantic Subset Principle and the Principle of Parsimony make
different predictions concerning the preferred interpretation of certain ambiguities including the focus assignment ambiguities that occur when only is in a
preverbal position in a sentence. Consider again the sentence in (2). Recall that an
analysis with contrastive focus on the entire VP is true in any circumstance in
which the fireman is not doing anything other than holding a hose, whereas the
analysis with contrastive focus on the direct object alone is true in only those circumstances in which the fireman is not holding anything other than a hose. Recall
also that an analysis with contrastive focus on the VP is true in a subset of those
circumstances in which an analysis with contrastive focus on the direct object is
true. According to Crain et al.’s (1994) version of the Principle of Parsimony,
when this ambiguity occurs outside of a disambiguating context, the adult reader
or hearer should preferentially adopt the analysis that is true in the largest set of
circumstances, which, in this case, is the analysis with contrastive focus on the direct object. However, according to the Semantic Subset Principle, learners should
initially formulate hypotheses concerning the meaning of a sentence that are true
in the smallest set of circumstances. Thus, the principle predicts that young children will preferentially adopt the analysis with contrastive focus on the VP because this analysis is true in a smaller set of circumstances than an analysis with
focus on the direct object. Crain et al. (1994) argued that as children encounter situations in which this initially formulated hypothesis is incorrect, they will acquire
evidence for the existence of alternative analyses and ultimately the alternative
sentential meanings.
It is important to note that Crain et al. (1994) did not claim that young children
employ different sentence processing strategies from adults. In fact, they considered that children and adults employ the same sentence processing strategies. Instead, they claimed that during language learning, other principles govern the
order of acquisition of the alternative analyses or interpretations of an ambiguity.
Thus, even though processing may proceed according to adult processing strategies, particular analyses of an ambiguity should not be available to the parser at
certain points in development.
Crain et al. (1994) also proposed that young children do not initially employ
syntactic restrictions on the range or scope of focus-sensitive quantifiers when
formulating hypotheses concerning the interpretation of a sentence. They argued
that children instead generate a single interpretation with the same focus assignment for sentences with only irrespective of the surface position of the quantifier.
As with the acquisition of the alternative analyses of an ambiguity, it is assumed
that as children encounter situations in which the meaning they assign to the sentence is incorrect, they will gather evidence for the existence of alternative, correct interpretations.
Several studies (Crain et al. (1992), Drozd and van Loosbroek (1998), Philip
and Lynch (1999)) have produced results that appear to support the claim that
young children give the same focus assignment to sentences with only regardless
260
PATERSON ET AL.
of its surface position. Such data are consistent with the claim that children do not
employ syntactic rules to restrict the range of the quantifier. Crain et al. examined
how 3- to 6-year-olds compared sentences like (3) and (4) with a picture of a cat
holding a flag, a duck holding both a flag and a balloon, and a frog holding a balloon.
(3) Only the cat is holding a flag.
(4) The cat is only holding a flag.
If participants used syntactic rules to restrict the quantifier’s range, then they
should have judged (4) to be a true description of events and judged (3) to be false.
However, a majority of participants judged both sentences to be true. Crain et al.
(1992) took this to indicate that they adopted an analysis with contrastive focus on
the VP regardless of the quantifier’s surface position. Philip and Lynch ((1999);
see also Drozd and van Loosbroek (1998)) examined how adults and preschoolers
compared a sentence with presubject only, for example, Only the dog is holding
an octopus, with a picture of a dog holding an octopus and a starfish and two cats
holding nothing. Adults judged the sentence to be a true description of events,
whereas over one third of preschoolers judged it to be false. Thus, it appeared that
some children misanalyzed the sentence by assigning contrastive focus to the direct object rather than to the subject NP.
Paterson et al. (2003) proposed an alternative explanation of these findings.
They argued that children experience difficulty in processing contrastive focus
rather than failing to apply syntactic restrictions on the range of the focussensitive quantifier and that they may process sentences with only without
computing an explicit set of alternatives. That is to say, they may understand the
sentence with only as asserting that an attribute belongs to the denotation of the
focused constituent without explicitly representing the alternatives that do not
have this attribute. Paterson et al. suggested that failing to compute a set of alternatives might result in children making errors in many experimental tasks by producing the same responses for sentences with only as they would for counterpart
sentences without only. Paterson et al. noted that in the Crain et al. (1992) study,
children compared sentences with only against a picture depicting events that
were consistent with contrastive focus on the direct object and inconsistent with
contrastive focus on the subject NP. However, the events also were consistent
with the analysis that would be assigned to counterpart sentences without only, for
example, The cat is holding a flag.3 Thus, although there was evidence that children did not process sentences by assigning contrastive focus to the subject NP,
3
3Crain
et al. (1994) first discussed a study in which sentences with only in either the presubject or
preverbal position were compared with a picture that correctly depicted events described by the sentence with preverbal only but not those described by the sentence with presubject only. This experiment does not appear to have included other pictures that correctly depicted events described by the
INTERPRETATION OF AMBIGUOUS FOCUS
261
the study did not unequivocally show that the children instead processed the sentences by assigning contrastive focus to the direct object. Paterson et al. argued
that the children might not have interpreted sentences by constructing an explicit
set of alternatives to the focused constituent. Instead, they may have constructed a
representation containing exclusively the focused information. That is to say, they
might have mentally instantiated the focused information, comprising a cat holding a flag, without computing an explicit set of alternatives. If this was the case,
then they may have judged a sentence like (4) to be true of any state of affairs that
is consistent with the events represented by the focused information regardless of
where only occurs in the sentence. Paterson et al. argued that a more thorough test
of children’s use of syntactic information when processing sentences with only
must include control conditions that enable the experimenter to determine if participants use information about both the focused information and its alternatives.4
To test this possibility, Paterson et al. (2003) conducted three studies in which
children and adults interpreted sentences with only in presubject and preverbal
positions, such as the sentences in (1) and (2), and counterpart sentences without
only, such as The fireman is holding a hose. Participants compared these sentences with pictures depicting events that matched or mismatched with possible
interpretations of the sentences, as shown in Figure 1.
For Experiment 1 (Paterson et al. (2003)), the critical comparison concerned
adults and 6- to 7-year-olds. Participants indicated whether sentences matched
with each of the corresponding pictures viewed separately. Children and adults
produced an equivalent high frequency of correct responses for sentences without
only, but children made more errors for sentences with only than adults did. When
the nature of the errors was examined, it was clear that most errors were made besentence
4
with presubject only, nor does it appear to have included sentences without only. A second
experiment involved six “VP-oriented” child participants: that is, participants who systematically
misanalyzed sentences with presubject only as having the same meaning as counterparts with
preverbal only. In this second study, Crain et al. (1994) investigated whether the VP-oriented participants misanalyzed sentences with presubject only by identifying either the direct object or the VP as
the sentence focus. In classifying these participants, Crain et al. (1994) included three picture contexts
that correctly depicted events described by sentences with presubject only. Participants were classified
as VP-oriented if they rejected such sentences on all three trials.
4
Paterson et al. (2003) also argued that aspects of the experiment design employed by Philip and
Lynch (1999) undermined their confidence in the conclusions reached in this study. Paterson et al.
noted that participants in Philip and Lynch’s task first compared the target sentence containing
preverbal only with two other pictures before comparing it with the critical picture and that Philip and
Lynch only reported data for participants who responded correctly to the first two pictures (without reporting error rates). Paterson et al. argued that it would not be surprising if preschoolers found this task
confusing, and its complexity may well have impaired their performance. Paterson et al. also noted
that the study lacked control conditions that were needed to show that preschoolers performed the task
satisfactorily. Because Philip and Lynch did not include sentences without only along with their sentences with only, Paterson et al. considered that it was not possible to demonstrate that preschoolers
made errors exclusively for sentences with presubject only and therefore made errors by associating
the focus quantifier with the wrong syntactic constituent.
262
PATERSON ET AL.
FIGURE 1 Examples of picture materials used by Paterson, Liversedge, Rowland, and Filik
(2003). Reprinted from Cognition 89, Paterson, K. B., S. P. Liversedge, C. Rowland, and R.
Filik, “Children’s Comprehension of Sentences with Focus Particles,” 263–294, Copyright
(2003), with permission from Elsevier.
cause children had produced similar responses for sentences with and without
only rather than because they had produced similar responses for sentences with
only appearing in different syntactic positions. Thus, the results were consistent
with an account in which children made errors because they misinterpreted sentences with only by failing to take account of contrastive information rather than
by failing to apply syntactic restrictions on its range. Paterson et al. also noted that
it was not simply the case that the children lacked any understanding of the lexical
INTERPRETATION OF AMBIGUOUS FOCUS
263
meaning of only because the data indicated that on many trials, they were able to
correctly evaluate the match between the sentences with only and the pictures depicting contrast information. Paterson et al. replicated their principal findings in a
second study in which participants simultaneously viewed all of the pictures before indicating which picture matched with the sentence meaning. In a third study,
they obtained a similar pattern of results for 4- to 5-year-old preschoolers in a
modified version of the original experiment.5
In our experiments, we investigated children’s and adults’ interpretations of
sentences with preverbal only (e.g., (2) and (4)) when these sentences have an ambiguous choice of focus. Our principal aim in conducting these experiments was
to evaluate Crain et al.’s (1994) claim that when sentences of this type are presented outside of a disambiguating discourse context, then adults will prefer to interpret them by assigning contrastive focus to the direct object, whereas children
will prefer to interpret them by assigning contrastive focus to the VP. We conducted three experiments using a sentence–picture verification task in which child
and adult participants compared ambiguous sentences with only in a preverbal position and unambiguous sentences without only against pictures depicting events
that either matched or mismatched with the meaning of the sentence. Figure 2
shows an example sentence and its corresponding pictures.
We expected participants to judge sentences without only to be true for those
pictures depicting events described in the sentence. That is, participants should
not interpret the sentence as indicating that a contrast is to be made between the
denotation of a sentential constituent and some alternatives to it. Thus, the sentence without only in Figure 2 should be judged to be true for pictures B, C, and D.
We also expected that adult participants would interpret sentences with only by
assigning contrastive focus. If they assigned contrastive focus to the direct object,
then they should judge the sentences to be true of each picture that depicts the
events described in the sentence unless it also depicts a contrast based on the direct object. That is, participants should judge the sentence with only in Figure 2 to
be a true description of pictures B and D, but critically, they should judge it to be a
false description of picture C, as this picture shows the woman walking something
5
5Kiss (1998) argued for a distinction between identificational focus and informational focus.
Identificational focus is equivalent to contrastive focus. It distinguishes between the denotation of a
particular syntactic constituent for which a particular predicate holds and a set of contextually relevant
alternatives for which the predicate does not hold. Informational focus, by contrast, marks that the focused information is new information without requiring that it be contrasted with some alternatives.
Kiss argued that only phrases are associated with identificational focus. However, it is at least possible
that in language processing, participants have both identificational focus and informational focus interpretations available to them. For example, it might be the case that when the reader or hearer cannot
identify any alternatives to the focused constituent that they assign an informational focus interpretation to the sentence. Whether this provides an explanation of the effects obtained by Paterson et al.
(2003) is an open question.
264
PATERSON ET AL.
FIGURE 2
Example sentence and picture materials used in these experiments.
other than a dog. By contrast, if participants assign contrastive focus to the VP,
then they should judge sentences with only to be true of each picture that depicts
the events described in the sentence unless it depicts a contrast based on the VP.
Thus, if participants assign contrastive focus to the VP for the sentence in Figure
2, they should judge it to be a true description of picture B but a false description
of both pictures C and D because these pictures show the woman doing something
other than walking a dog. Table 1 illustrates the predicted pattern of responses for
sentences that do not specify contrastive focus and those that are interpreted with
contrastive focus on either the direct object or VP.
INTERPRETATION OF AMBIGUOUS FOCUS
265
TABLE 1
Predicted Pattern of Responses for Sentences That Do Not Specify Contrastive
Focus or That Receive an Interpretation With Contrastive Focus on Either the
Direct Object (DO; i.e., DO-Contrast) or the Verb Phrase (VP; i.e., VP-Contrast)
Pictures
Focus
A
B
C
D
No contrast
DO contrast
VP contrast
✕
✕
✕
✔
✔
✔
✔
✕
✕
✔
✔
✕
Note. Tick marks (✔) indicate pictures that match with an analysis, and crosses (✕) indicate pictures that mismatch with the analysis.
2.
EXPERIMENT 1
In this experiment, we examined the performance of 7- to 8-year-olds and 9- to
10-year-olds and two groups of adults on a sentence–picture verification task. We
presented participants with a sentence and required them to indicate whether it
was a true description of events depicted in each of four pictures, with each picture viewed individually. Each sentence was shown to participants on a card
(which remained visible throughout the trial) and was read aloud to child participants and to one group of adult participants at the start of each trial and prior to being shown each of the four pictures.
Prosody was likely to play an important role in disambiguating the choice of
focus in a sentence, and it was possible that the experimenter would provide unintentional prosodic cues to focus assignment when reading aloud the ambiguous
sentences. For this reason, we included a second group of adult participants in the
experiment who did not hear the sentences read aloud. If these adults produced a
different pattern of responses from adults who heard the sentences read aloud,
then we could conclude that prosody had biased sentence interpretation. We did
not include a control group of children who did not hear the sentences read aloud,
principally because we were concerned that the performance of children who only
read the sentences might be contaminated by individual differences in their reading ability. Furthermore, studies have suggested that children are insensitive to
prosodic cues to focus assignment (e.g., Gualmini et al. (2002)).
Figure 2 gives an example of the sentences and the corresponding set of pictures used in this experiment. We used sentences that included only in a preverbal
position and counterpart sentences without only. The sentences with only could be
interpreted with contrastive focus on either the direct object or the VP. When the
direct object is the focus, this indicates that a contrast is to be made between the
referent of the direct object (e.g., a dog) and some alternatives (e.g., a cat). By
266
PATERSON ET AL.
contrast, when the VP is focused, this indicates that a contrast is to be made between the act it described (e.g., walking a dog) and some alternatives (e.g., throwing a ball).
We expected adult and child participants to interpret sentences without only
without assigning contrastive focus. If adults processed the ambiguous sentences
with only in line with Crain et al.’s (1994) account, then they should prefer a direct-object- (DO-) contrast analysis of the ambiguity. By contrast, Crain et al.’s
account predicts that child participants will produce a nonadult pattern of responses for sentences with only by preferentially adopting a VP-contrast analysis
of the ambiguity. Paterson et al. (2003) had shown that children between 4 and 10
years of age produce nonadult patterns of response when comparing sentences
with only in different syntactic positions with pictures that matched with the
meaning of alternative analyses of the sentence. Similarly, we might expect children in the age groups we studied in this experiment to produce nonadult patterns
of response.
2.1.
Method
2.1.1. Participants. Four groups participated in this experiment: fourteen
7- to 8-year-olds and fourteen 9- to 10-year-olds from a school in Derbyshire and
two groups of fourteen adults who were undergraduates at the University of
Derby.
2.1.2. Materials and design. In this experiment, we employed a sequential sentence–picture verification task in which participants heard (or read) sentences and indicated whether they correctly described the events depicted in each
of four pictures. Pictures were viewed one at a time.
We used 10 sentences with a NP-V-NP structure. An example sentence is
shown in Figure 2, and the full set of sentences is included in the Appendix. The
direct object of the sentence was always indefinite. One version of each sentence
included only in an auxiliary position, and the other version did not include only.
Sentences with only were ambiguous between DO-contrast and VP-contrast analyses. For each sentence, the corresponding pictures depicted events that matched
or mismatched with its possible interpretations.
Picture A always showed the character from the sentence but not performing
the described action (e.g., the woman is not walking a dog). This picture was included to make participants aware that it was possible for sentences to be false descriptions of the depicted events. Picture B always showed the character
performing the described action and doing nothing else. Pictures C and D discriminated between the alternative focus assignments. Picture C always depicted the
character performing the described action but also performing another action that
supplied a contrast for the direct object (e.g., the woman is also walking a cat).
Picture D always depicted the character performing the described action and an-
INTERPRETATION OF AMBIGUOUS FOCUS
267
other action that supplied a contrast with the VP (e.g., the woman is also throwing
a ball). It is important to note that picture D was true for the DO-contrast analysis
of sentences with only, as this picture does not depict a contrast based on the direct object.
There were two independent variables: participant group and sentence type.
The dependent variable was the frequency of different categories of responses:
Responses were consistent either with participants interpreting sentences without
specifying contrastive focus (no contrast) or by adopting a DO-contrast or VPcontrast analysis. Table 1 illustrates the pattern of responses corresponding to
these alternative analyses.
2.1.3. Procedure. We tested participants individually, each test session
lasting approximately 20 min. Sentences were divided into two lists, with each list
including one version of each sentence: either with or without only. Each participant viewed the sentences from one list. Thus, each participant viewed five sentences with only and five without only and no participant viewed more than one
version of each.
On each trial, participants first viewed the sentence on a card. Then the sentence was read aloud to the children and to one group of adults. The second adult
group viewed the sentences without them being read aloud. Participants were then
shown the corresponding set of pictures one at a time and asked to indicate if the
sentence was a true or false description of depicted events. Sentences were presented in a fixed order to half of the participants in each group, with the order of
presentation reversed for the other participants. We presented pictures to participants in order A to D for half of the trials and in order D to A for the other half.
2.2.
Results
Table 2 shows the mean percentage frequency of no-contrast responses, DOcontrast responses, and VP-contrast responses by 7- to 8-year-olds, 9- to 10-yearolds, and adults who either heard or did not hear sentences read aloud for
sentences with and without only.
As the response categories were not statistically independent, we conducted separate 4 (participant group) × 2 (sentence type) mixed design analyses of variance
(ANOVAs) treating each category as a separate dependent variable. Effects were
considered to be significant when p < .05. For no-contrast responses, there was no
effect of participant group, F < 1.3; significantly fewer no-contrast responses for
sentences with than without only, F(1, 56) = 101.9, p < .001; and a significant interaction, F(3, 56) = 11.4, p < .001. The 7- to 8-year-olds did not produce significantly
different amounts of no-contrast responses for sentences with and without only, F <
1. However, 9- to 10-year-olds made more no-contrast responses for sentences
without only than for ones with only, F(1, 56) = 14.2, p < .001; as did adults in the
read-aloud condition, F(1, 56) = 52.7, p < .001; and adults in the not-read-aloud
268
PATERSON ET AL.
TABLE 2
Mean Percentage Frequency of Responses From Experiment 1 That Were
Consistent With Participants Representing the Sentence With and Without
Only Without Including Contrast Information, Basing the Contrast
on the Direct Object or Basing the Contrast on the Verb Phrase (VP)
Response
Trial
Without only
7–8 years
9–10 years
Adults (spoken)
Adults (read)
With only
7–8 years
9–10 years
Adults (spoken)
Adults (read)
No Contrast
Information
Contrast Based
on Direct Object
Contrast Based
on VP
31
35
67
69
12
1
7
8
36
59
13
12
23
4
3
0
15
18
20
11
45
71
47
73
condition, F(1, 56) = 67.7, p < .001. Thus, there was some evidence that the youngest children made errors by failing to process contrastive information for sentences
with only, consistent with Paterson et al.’s (2003) account, but in this experiment,
this accounted for only a small proportion of the data.
Participants produced more DO-contrast responses for sentences with than
without only, F(1, 56) = 7.3, p < .01, with no difference between participant
groups and no interaction, F < 1.3. Thus, there was no evidence that in contrast
with children, adults preferred to assign focus to the direct object of the ambiguous sentences. Turning now to the VP-contrast responses, we found that participants made more VP-contrast responses for sentences with than without only,
F(1, 56) = 55.9, p < .001. There was also a main effect of participant group, F(3,
56) = 3.4, p < .05. The 9- to 10-year-olds produced more VP-contrast responses
than did 7- to 8-year-olds ( p < .05) or either group of adults ( p < .05), and adults
produced more VP-contrast responses than did 7- to 8-year-olds ( p < .05).
Finally, there was a significant interaction, F(3, 56) = 9.5, p < .001. Although
adults produced significantly more VP-contrast responses in the read-aloud, F(1,
56) = 17.8, p < .001, and not-read-aloud conditions, F(1, 56) = 61.1, p < .001, no
significant differences were found for either 9- to 10-year-olds, F < 2.8, or 7- to 8year-olds, F < 1.5. There were two important features to these results. First, children, unlike adults, often assigned sentences with and without only a VP-contrast
interpretation. Second, both the children and adults were most likely to assign a
VP-contrast rather than a DO-contrast analysis to sentences with only.
Planned comparisons indicated that adults in the read-aloud condition produced fewer VP-contrast responses than adults in the not-read-aloud condition for
269
INTERPRETATION OF AMBIGUOUS FOCUS
sentences with only, t(27) = 2.53, p < .05, with no difference for sentences without
only, t < 1. This effect was consistent with unintentional prosodic cues reducing
the likelihood of adult participants forming a contrast based on the VP. Thus, it
appeared that although prosodic cues affected adults’ interpretation of the ambiguity, their effect was to bias participants against assigning what appears to be the
preferred interpretation.
The classification of participant responses into no-contrast responses, DOcontrast responses, and VP-contrast responses was a conservative approach that
could be considered disadvantageous in that it led to a high rate of unclassified responses. It was possible that these unclassified responses might include data that
would be informative about the interpretations assigned to the sentences. Therefore, to avoid the possibility that we had failed to detect systematicities across all
our data, we also examined how frequently participants judged each picture as being a true depiction of the events described by the sentences with and without
only. Inspecting the raw frequency data enabled us to examine the dominant pattern of responses across the entire data set, including those responses that formed
unclassified patterns of responses. These frequencies are shown in Table 3.
An inspection of the data in Table 3 indicates that participants always judged
Picture B to depict the events described by the sentences and that they almost always judged Picture A to not depict these events. The 7- to 8-year-olds and 9- to
10-year-olds often rejected Pictures C and D as depictions of events in the
sentences without only. This was consistent with these participants assigning a
VP-contrast interpretation to sentences without only despite the absence of grammatical cues to this effect. Adults were much more likely to judge Pictures C and
D to be consistent with the meaning of sentences without only, but they nevertheTABLE 3
Percentage Frequency of Participants Judging Sentences With and Without
Only to Be a True Description of Each Picture for Experiment 1
Picture
Trial
A
(Incorrect)
B (No
Contrast)
C (Direct
Object Contrast)
D (VP
Contrast)
0
1
1
0
100
100
100
100
33
36
83
77
53
36
86
84
0
0
0
0
100
100
100
100
30
6
3
0
46
21
40
16
Without only
7–8 years
9–10 years
Adults (spoken)
Adults (read)
With only
7–8 years
9–10 years
Adults (spoken)
Adults (read)
Note.
VP = verb phrase.
270
PATERSON ET AL.
less rejected these pictures on a small proportion of trials, suggesting that they had
a residual tendency to interpret sentences without only as having a VP-contrast interpretation.
The 7- to 8-year-olds produced a very similar pattern of responses for sentences with and without only, frequently judging Pictures C and D to be consistent
with the meaning of both sentence forms. The results for the 9- to 10-year-olds
and the adults were most consistent with the view that these participants sometimes interpreted sentences with only by assigning a DO-contrast interpretation
but most frequently assigned a VP-contrast interpretation to them.
Finally, adult participants who heard the sentences spoken aloud were more
likely than those who did not to accept sentences with only with respect to Picture
D. Because adults rarely judged Picture C to depict the events in these sentences,
this pattern of results was consistent with adults who heard the sentences being
more likely than those who did not to assign a DO-contrast analysis to sentences
with only.
2.3.
Discussion
The study produced several key findings. The first was related to the sentences
without only. Child participants produced an unexpected pattern of responses
for these sentences, which they assigned a VP-contrast analysis. This was entirely unexpected because these sentences did not include any syntactic cues to
contrastive focus and therefore should not have received an interpretation that
specified such a contrast. We considered that there were at least two possible explanations for this effect. First, it may be that unintentional prosodic cues caused
participants to assign focus to the VP. Alternatively, it may be that the participants opted for the picture that they considered to best represent the meaning of
the sentence, which may have been the one that depicted only the events
described in the sentence. However, the adult data suggested that the former explanation was less likely to be correct. Although we found that adults predominantly analyzed sentences without only without specifying contrastive focus, on
a small proportion of trials they also produced responses that were consistent
with a VP-contrast interpretation. Exactly the same effect was obtained for
adults who did and did not hear the sentence spoken aloud. Therefore, it does
not appear that this effect is attributable to prosodic cues causing participants to
focus on the VP. This certainly appears to be the case for adult participants and
may also have been the case for the children.
The second key finding related to the sentences with only. Both child and adult
participants preferred the VP-contrast analysis of the ambiguity in these sentences. Thus, although the children interpreted the sentences in line with Crain et
al.’s (1994) account, the adults did not. Moreover, contrary to Crain et al.’s ac-
INTERPRETATION OF AMBIGUOUS FOCUS
271
count, the results suggest that children and adults gave the same focus assignments to the ambiguous sentences. However, before concluding that both children
and adults favored a VP-contrast analysis of the ambiguity, we must consider another possibility, namely, that the task used in the present experiment was insensitive to default interpretation preferences and instead encouraged participants to
consider the alternative possible analyses of the ambiguity.
Recall that we presented participants in our experiment with a sentence, and
they had to judge whether it matched in meaning with a sequence of individually
viewed pictures. For our results to reflect default interpretation preferences, it was
necessary for participants to adopt a preferred analysis of an ambiguity and to adhere to this analysis during the course of each trial either by maintaining it in
memory as they evaluated each picture or by readopting the same analysis as they
compared the sentence with each picture. However, it was not possible for the experimenter to determine whether participants did adhere to a single analysis during each trial or whether at any point they abandoned it in favor of an alternative
analysis. Indeed, it was possible that the sequential nature of the sentence–picture
comparisons encouraged participants to either consider the possible alternative
analyses of a sentence within each trial or modify their initial interpretation of a
sentence depending on the visual context. If this were the case, then we would expect participants often to judge sentences with only to be false for pictures depicting a contrast based on either the direct object or the VP, which is consistent with
the results we obtained. Consequently, we cannot unambiguously interpret the results of this experiment as demonstrating that adults have a preference for adopting a VP-contrast analysis of the ambiguity. Obviously, the participants in
Experiment 1 also compared sentences without only and each of the corresponding pictures. As the sentences without only were unambiguous with respect to
contrastive focus, we did not consider it likely that the sequential nature of the
task would cause participants to consider alternative interpretation of these sentences in the course of each trial.
In Experiment 2, we used a task that encouraged participants to maintain a single analysis during the course of each trial and therefore might provide further information concerning participants’ default interpretation preferences. For this
task, we presented participants with a sentence, and then they viewed the corresponding pictures simultaneously and were required to select those pictures that
matched with its meaning.
3.
EXPERIMENT 2
In the second experiment, we employed a modified version of the sentence–picture verification task that might provide further information concerning participants’ default interpretations of the focus assignment ambiguity. We presented
272
PATERSON ET AL.
participants with the sentence while they viewed all of the corresponding pictures,
and we requested that they select those pictures that matched the meaning of the
sentence. We reasoned that if participants had a strong preference for adopting
one interpretation of an ambiguous sentence with only over another, then they
would adopt the preferred interpretation and select those pictures that are consistent with its meaning.
We examined the same predictions that were tested in Experiment 1. Thus, we
investigated whether children and adults would interpret sentences with and without only either by not assigning contrastive focus or by assigning contrastive focus to either the direct object or the VP. The pattern of responses corresponding to
each of these analyses is shown in Table 1.
3.1.
Method
3.1.1. Participants. Four groups participated in this experiment: ten 7- to
8-year-olds and ten 9- to 11-year-olds from a school in Derbyshire. Two groups of
ten adult participants were undergraduates at the University of Derby.
3.1.2. Materials and design. Experiment 2 employed the same sentence
and picture materials and manipulated the same experimental variables as Experiment 1 and differed only in the experimental task. For Experiment 2, we employed a version of the sentence–picture verification task in which participants
heard or read the sentence and then simultaneously viewed all four versions of the
corresponding pictures. Participants selected those pictures that matched with the
meaning of the sentence.
3.1.3. Procedure. We tested participants individually, with each test session lasting approximately 20 min. We divided sentences into two lists, with each
list including one version of each sentence: either with or without only. Each participant viewed sentences from one list. Each participant viewed 10 sentences, 5
sentences with only and 5 without only. No participant viewed more than one version of each.
For each trial, participants first heard or read a sentence and then compared it
with four simultaneously viewed pictures. Child participants viewed the sentence
on a card; the sentence was also read aloud by the experimenter at the beginning
of the trial and on viewing the pictures and the card was visible throughout the
trial. We employed an identical procedure for one group of adults. The other
group of adults read the sentences from the card, which was visible throughout the
trial, without the experimenter reading the sentence aloud. We presented sentences in a fixed order to half of the participants in each group, with the order of
presentation reversed for the other participants.
273
INTERPRETATION OF AMBIGUOUS FOCUS
3.2.
Results and Discussion
Table 4 shows the percentage frequency of responses that were consistent with interpreting sentences without specifying a contrast, specifying a DO contrast, or
specifying a VP contrast. Other responses were due to errors or participants making a combination of responses that did not fit the preceding categories.
The results from this experiment are remarkably clear, rendering statistical
analyses unnecessary. All participants showed a clear preference for selecting
only Picture B regardless of whether the sentence included only. This picture portrayed those events described by the sentence and depicted no other contrasting
events within the picture. This result was particularly unexpected for sentences
without only. We considered that there were two possible explanations for this result. First, it was possible that it reflected a preference for processing sentences
with and without only as being true in circumstances that did not include contrasting events based on the VP constituents. Alternatively, the result might have
arisen from a misunderstanding of the experimental task. Participants may have
misunderstood the instruction to select those pictures that matched with the meaning of the sentence and instead understood the task as requiring them to select the
picture that best matched with the meaning of the sentence. For both sentences
with and without only, the picture that best matched the meaning of the sentence
was the picture that depicted events from the sentence without including any contrasting events. Our data cannot discriminate between these two possible accounts. Note also that whatever the reason for the results obtained for sentences
without only under these experimental conditions, the preference that occurred
was clearly not attributable to any differences in child and adult comprehension
TABLE 4
Mean Percentage Frequency of Responses From Experiment 2 That Were
Consistent With Participants Representing the Sentence With and Without
Only Without Including Contrast Information, Basing the Contrast
on the Direct Object or Basing the Contrast on the Verb Phrase (VP)
Response
Trial
Without only
7–8 years
9–10 years
Adults (spoken)
Adults (read)
With only
7–8 years
9–10 years
Adults (spoken)
Adults (read)
No Contrast
Information
Contrast Based
on Direct Object
Contrast Based
on VP
0
0
0
0
3
2
0
0
87
85
84
82
0
0
0
0
1
0
2
4
91
90
84
88
274
PATERSON ET AL.
because it occurred not only for the children but also (admittedly to a lesser extent) for the adult participants.
4.
EXPERIMENT 3
To rule out the possibility that participants selected the picture that best matched
the sentence meaning, we conducted a third experiment in which we presented
participants with a sentence, and then they simultaneously viewed all the corresponding pictures, and we asked them to indicate which pictures did not match the
meaning of the sentence. As with the version of the sentence–picture verification
task used in Experiment 2, we considered that this task would be informative
about default processing preferences, and unlike the task employed in Experiment
1, it would not encourage participants to consider alternative possible analyses of
the sentence or to modify their interpretation depending on the visual context provided by each picture. However, in contrast with the task employed in Experiment
2, this task encouraged participants to consider whether each of the pictures
matched or mismatched with the participant’s understanding of the meaning of
the sentence, thus discouraging participants from simply identifying one picture
that best matched with the meaning of the sentence. That is, whereas the Experiment 2 task may have encouraged participants to identify only the exemplar
matching picture, the Experiment 3 task would encourage participants to identify
mismatching pictures and therefore to retain all of the matching pictures. If participants had a strong default preference for adopting one possible analysis of the
sentence rather than another, then participants should reject pictures that were inconsistent with that preference. We examined the same predictions as were examined in Experiments 1 and 2.
4.1.
Method
4.1.1. Participants. Three groups participated in this experiment: twenty
7- to 8-year-olds from a school in Norfolk and two groups of 7 adult participants
who were undergraduates at the University of Derby.
4.1.2. Materials and design. In Experiment 3, we employed the same set
of sentence and pictures as used in Experiment 2 and manipulated the same experimental variables. This experiment differed only in terms of the experimental task
because it required participants to indicate which pictures mismatched with the
sentence meaning rather than indicating which matched with its meaning.
4.1.3. Procedure. We tested participants individually, with each test session lasting approximately 20 min. Sentences were divided into two lists, with
each list including one version of each sentence: either with or without only. Each
275
INTERPRETATION OF AMBIGUOUS FOCUS
participant viewed sentences from one list. Thus, each participant viewed five
sentences with only and five without only. No participant viewed more than one
version of each.
For each trial, the child participants first viewed the sentences on a card with
the sentence also read aloud by the experimenter at the beginning of the trial and
on viewing the pictures with the card visible throughout the trial. For one group of
adult participants the sentences were shown on a card and they were read aloud by
the experimenter; the other group viewed the sentences on a card without hearing
them read aloud. On each trial, the four pictures were presented simultaneously
for both child and adult participants. Participants indicated which pictures did not
match with the meaning of the sentence. Sentences were presented in a fixed order
to half of the participants in each group, with the order of presentation reversed
for the other participants.
4.2.
Results and Discussion
Table 5 shows the percentage frequency of no-contrast, DO-contrast, and VPcontrast responses by 7- to 8-year-olds and adults in read-aloud and not-readaloud conditions for sentences with and without only. Other responses were due to
errors or combinations of responses that did not fit the preceding categories.
As in Experiments 1 and 2, we conducted separate 4 (participant group) × 2
(sentence type) mixed design ANOVAs treating each category as a separate dependent variable. Participants produced more no-contrast responses for sentences
without only than for sentences with only, F(1, 31) = 82.4, p < .001. A main effect
of participant group, F(2, 34) = 25.3, p < .001, was due to the children producing
fewer no-contrast responses than adults in the read-aloud condition ( p < .001) or
TABLE 5
Mean Percentage Frequency of Responses From Experiment 3 That Were
Consistent With Participants Representing the Sentence With and Without Only
Without Including Contrast Information, Basing the Contrast on the Direct
Object or Basing the Contrast on the Verb Phrase (VP)
Response
Trial
Without only
7–8 years
Adults (spoken)
Adults (read)
With only
7–8 years
Adults (spoken)
Adults (read)
No Contrast
Information
Contrast Based
on Direct Object
Contrast Based
on VP
8
77
74
8
0
0
65
14
14
8
0
0
5
0
14
67
100
71
276
PATERSON ET AL.
adults in the not-read-aloud condition (p < .001). No difference was found between the two adult groups (p > .05). There also was a significant interaction, F(2,
31) = 25.3, p < .001. The 7- to 8-year-olds produced the same number of nocontrast responses for sentences with and without only, F < 1, but adults in the
read-aloud group produced more no-contrast responses for sentences without only
than for sentences with only, F(1, 31) = 48.3, p < .001, as did adults in the notread-aloud group, F(1, 31) = 44.8, p < .001.
No main effects or interactions were found for DO-contrast responses, Fs <
1.8. However, more VP-contrast responses were produced for sentences with than
without only, F(1, 31) = 86.4, p < .001, with no effect of participant group, F <
1.6, although there was a significant interaction, F(2, 31) = 29.4, p < .001. The 7to 8-year-olds produced the same number of VP-contrast responses for sentences
with and without only, F < 1. Adults in the read-aloud condition produced more
VP-contrast responses for sentences with than without only, F(1, 31) = 71.1, p <
.001, as did adults in the not-read-aloud condition, F(1, 31) = 31.6, p < .001.
The results were clear. As in Experiment 1, the child participants tended to assign a VP-contrast interpretation to sentences with and without only. Adult participants were much less likely to interpret sentences without only by assigning a
VP-contrast interpretation. Instead, adults predominantly interpreted these sentences without assigning contrastive focus. By contrast, adults assigned a VPcontrast analysis to sentences with only. Thus, adults and children showed the
same bias for adopting the VP-contrast analysis of sentences with only; therefore,
there was no evidence that they adopted different analyses of the ambiguity.
As in Experiment 1, to check for hidden patterns in unclassified responses, we
examined how often each picture was retained (not rejected) in Experiment 3. Table 6 shows the frequency of responses in which participants retained each picture
as matching with the meaning of sentences with and without only.
TABLE 6
Percentage Frequency of Participants Retaining Sentences With and
Without Only as a True Description of Each Picture for Experiment 3
Picture
Trial
A
(Incorrect)
B (No
Contrast)
C (Direct
Object Contrast)
D (VP
Contrast)
0
0
0
88
100
100
12
97
83
28
100
86
0
0
0
89
100
86
12
0
3
33
0
17
Without only
7–8 years
Adults (spoken)
Adults (read)
With only
7–8 years
Adults (spoken)
Adults (read)
Note.
VP = verb phrase.
INTERPRETATION OF AMBIGUOUS FOCUS
277
These results are similar to those obtained in Experiment 1. In both experiments, 7- to 8-year-olds tended to reject sentences with only and without only as
descriptions of Pictures C and D as expected under a VP-contrast interpretation.
Adults in both experiments tended to reject sentences with only with respect to
Pictures C and D but to accept sentences without only as descriptions of these pictures. These results suggest that children, unlike adults, assigned VP-contrast interpretations to sentences without only regardless of differences in task demands.
To summarize, the results obtained in Experiment 3 were consistent with those
obtained in Experiment 1. We found that young children were most likely to interpret sentences without only by assigning a VP-contrast analysis. They were also
most likely to interpret sentences with only by assigning this analysis. Adults did
not assign contrastive focus when they interpreted sentences without only but assigned a VP-contrast interpretation to sentences with only. The pattern of results
was different from those obtained in Experiment 2 because in this experiment, we
encouraged the participants to evaluate whether the pictures were inconsistent
with the meaning of a sentence rather than whether they best matched with its
meaning.
5.
GENERAL DISCUSSION
Our studies provide novel findings concerning children’s and adults’ processing
of contrastive focus. Perhaps the most important of these findings was that children and adults both assigned VP-contrast interpretations to sentences with
preverbal only. Thus, the results suggest that both child and adult participants had
a strong preference for analyzing the ambiguous sentences with only by assigning
contrastive focus to the VP.
On the face of it, this interpretation of the children’s performance is consistent
with Crain et al.’s (1994) account of how learnability conditions might direct children’s interpretations of sentences that are made ambiguous by the inclusion of a
focus-sensitive quantifier such as only (see Introduction). According to this account, children should assign a VP-contrast interpretation to ambiguous sentences
with preverbal only not because this is their preferred interpretation of the ambiguity but because this is the only interpretation the Semantic Subset Principle
makes available to the parser. Our data suggest, as this account predicts, that children are most likely to assign VP-contrast interpretations to sentences with
preverbal only.
However, it would only be possible to interpret our results as providing support for this account if it could be shown that children and adults favor different
analyses of the ambiguity because it is only possible to ascribe a nonadult interpretation to children’s performance if they do in fact exhibit different behavior to
that of adults. The results of all three experiments suggested that adults, as well as
children, preferred to assign a VP-contrast analysis to sentences with preverbal
278
PATERSON ET AL.
only. Therefore, the children in our experiments did not produce nonadult responses. Consequently, it would be unparsimonious on one hand to attribute their
interpretative preferences to the operation of learnability constraints and on the
other hand to attribute identical adult interpretative preference to sentenceprocessing strategies. Instead, these data are most parsimoniously interpreted as
showing that children and adults used the same strategies to resolve the ambiguity. Thus, despite the fact that our results were consistent with Crain et al.’s (1994)
predictions concerning children’s interpretations of a focus assignment ambiguity, our data do not support the view that learnability constraints directed them to
adopt a nonadult interpretation of the ambiguity.
As we noted previously, it was the adult participants rather than the children
whose performance appeared to be inconsistent with Crain et al.’s (1994) account
of the interpretations given to a focus assignment ambiguity. According to this account, if the sentence is presented outside of a disambiguating context, the adult
parser constructs alternative analyses in parallel and, in accordance with the Principle of Parsimony, settles on the analysis that presupposes the least amount of semantic information or, alternatively, the analysis that makes the sentence true in
the largest set of circumstances. For a sentence like The fireman is only holding a
hose, this analysis is the DO-contrast analysis. In conducting our experiments, we
had assumed that showing the participants a sentence containing only before allowing them to view the pictures would encourage them to assign an initial default DO-interpretation to the sentence in the absence of disambiguating context
in accordance with the Principle of Parsimony and to hold that interpretation in
their memory during the course of a trial. However, adults strongly preferred a
VP-contrast interpretation of these sentences in apparent contrast to Crain et al.’s
predictions.
However, we believe that there may be an alternative explanation of our results
that is consistent with Crain et al.’s (1994) account. Participants in our experiment
may have abandoned any default interpretation they had initially built for sentences with only when they saw the pictures. Under this view, the Principle of
Referential Success rather than the Principle of Parsimony would have been operative and guided the parser to adopt a VP-contrast interpretation of the sentences
with only. This is a potentially important observation, because it suggests that participants were either able to access the alternative possible interpretations of the
ambiguities in parallel or able to switch between these alternatives relatively cost
free in the course of evaluating each sentence. It is also possible that children, like
adults, showed a strong preference for the VP-contrast interpretation because this
interpretation was the most consistent with the Principle of Referential Success
whether the sentence included only or not and not because they were led by the
Semantic Subset Principle to adopt interpretations consistent with the smallest set
of circumstances. Clearly, further research is needed to determine whether readers
and hearers have a default preference for adopting a particular analysis of a focus
INTERPRETATION OF AMBIGUOUS FOCUS
279
assignment ambiguity and to investigate the effects of referential context on the
resolution of this form of ambiguity.
One aspect of the children’s data was particularly surprising. We found that the
child participants also had a clear preference for interpreting the sentences without only as excluding the possibility of events that formed a contrast with the VP.
Thus, it appeared that the children had interpreted sentences without only as if
they included contrastive focus on the VP despite the absence of grammatical
cues to this effect. This was a robust effect that was consistently obtained across
all three experiments.
One possible explanation for this unanticipated finding is that children expected sentences to be maximally informative about the depicted events (e.g.,
Grice (1975)) and judged sentences without only to be false for pictures depicting
anything in addition to the described events. This account is consistent with the
findings from Experiment 2. In this experiment, participants almost always
judged sentences with and without only to be true for just one picture, which depicted events from the sentence without including any additional contrasting
events. It seemed likely that participants in this experiment had misunderstood the
instructions as requiring them to select one picture and picked the one that best
matched with the meaning of the sentence as per the Gricean maxim of quantity.
Furthermore, when required to select just one picture, adult responses were similar to children’s responses, again suggesting that they favored a maximally informative reading of the sentences. It is also possible that presenting participants
with sentences that differed by the presence or absence of a focus-sensitive quantifier and requiring them to compare the sentences with pictures highlighting potential contrasting events may have encouraged child participants to process
sentences without only by including contrastive focus. Further investigations
clearly are required to determine the precise conditions under which children
compute contrastive focus, particularly when those conditions do not potentially
promote it.
We must briefly consider one possible objection to the effects we have obtained for child participants and the conclusions that we have drawn. First, it may
be of concern to some readers that the child participants in our study were between 7 and 10 years old. Such readers might consider that children of this age
would already have acquired the alternative possible interpretation of the ambiguity that we examined, and therefore our studies might not have been sensitive to
the influence of language acquisition principles on sentence interpretation. However, we note that Crain et al.’s (1994) original studies examined the interpretative
preferences of children between 3 and 6 years of age and found that almost all of
these children produced nonadult patterns of response when evaluating the meaning of sentences with presubject only. Furthermore, Paterson et al. (2003) found
that children between 4 and 10 years produced nonadult responses when asked to
evaluate the meaning of sentences with presubject and preverbal only. Thus, al-
280
PATERSON ET AL.
though our youngest age group was a little older than the participants in Crain et
al.’s studies, they were within the age range examined by Paterson et al. It is
therefore entirely conceivable that the child participants in our study might have
produced nonadult responses. Furthermore, the children’s performance on the experimental task was exactly as Crain et al. had predicted. In our experiments, it
was the adults rather than the children who produced results that were contrary to
Crain et al.’s predictions. Thus, it is not possible to discount our results on the
grounds that the children were of an inappropriate age to observe learnability effects on sentence interpretation because the children’s responses matched with
Crain et al.’s account of these putative effects.
Finally, we must consider how these results relate to the findings obtained by
Paterson et al. (2003). Crain et al. (1994) proposed that young children might
make errors in interpreting sentences with only by failing to apply syntactic restrictions on the range of the quantifier and therefore interpreting a sentence with
only in one syntactic position (e.g., a presubject position) as if it had the same
meaning as a counterpart sentence with only in another syntactic position (e.g., a
preverbal position). By contrast, Paterson et al. proposed that young children
might make errors when interpreting sentences with only by failing to mentally
instantiate an explicit set of alternatives. Consequently, these children might
make errors in many experimental tasks by producing the same pattern of responses for sentences with only as for counterpart sentences without only. Paterson et al. compared these alternative accounts of children’s errors in three
experimental studies and claimed that their results supported their view that children principally make errors by failing to mentally represent an explicit set of alternatives to the focused information in a sentence with only rather than by failing
to apply syntactic restrictions on the quantifier’s range. Given these findings, we
might have expected child participants in our experiments to also fail to mentally
represent an explicit set of alternatives when interpreting sentences with only and
evaluate sentences with and without only in similar ways. However, although the
youngest participants seemed to assign similar interpretations to sentences with
and without only, older children and adults typically did not. Overall, the kinds of
errors expected from Paterson et al.’s (2003) study occurred much less frequently
in our data.
One possible reason for this effect is that the picture material used in our materials provided a strong cue to participants to take account of contrasting information when evaluating the sentences. In these experiments, participants examined
four pictures on each trial. Of these pictures, two always depicted information that
formed a contrast with the VP constituents. It is therefore possible that the picture
materials used in this experiment cued the participants to consider possible contrasts based on these constituents independently of the presence of only and reduced the likelihood of participants processing sentences with only without
specifying an explicit set of alternatives. One consequence of being cued into taking account of contrast information may have been that it reduced the likelihood
INTERPRETATION OF AMBIGUOUS FOCUS
281
of participants processing sentences with only without specifying an explicit set
of alternatives. Another possible consequence might have been that it cued participants into evaluating those sentences that did not include only by applying
contrastive focus to the VP despite the absence of explicit grammatical cues that
would support such an analysis.
A related point concerns the pragmatic felicitousness of our materials.6 It could
be argued that although the pictures depicting events that formed a contrast with
constituents of the VP were effective at falsifying different focus interpretations,
these pictures did not supply conditions in which the different interpretations of
the sentences might be judged to be true. Consider the sentence and pictures
shown in Figure 2. Picture D is intended to be true for the sentence with preverbal
only when contrastive focus is assigned to the direct object. However, this picture
does not provide a contrast that supports this analysis. That is, it does not include
another entity, such as a cat that is not being walked, that would form a contrast
with the direct object. This contrast must be included if the DO-contrast interpretation of the sentence is to be felicitous. Crain et al. (1996) argued that children
may make nonadult responses on tasks involving a nonfelicitous presentation of
material. Although it is far from clear whether the nonfelicitous nature of the pictures used in this study contaminated either the child or adult responses, there is
certainly merit in future research that investigates this issue.
Other research might consider the role of prosody. Although it is well known
that prosody influences adult focus assignments (e.g., Carlson (2004), Gennari,
Meroni, and Crain (2004), Halbert et al. (1995)), some investigations suggested
that it has little influence on children’s focus interpretation (Gualmini et al.
(2002)). In our experiments, the experimenter avoided providing explicit prosodic
cues to focus assignment because the aim of the experiments was to determine the
focus assignment preferences in the absence of prosodic cues. We included additional conditions in which the adult participants read the experimental sentences
without hearing them spoken aloud to provide a control condition. By comparing
conditions in which adults either heard the sentences spoken aloud or not, we
hoped to determine whether the results from our studies were contaminated by unintentional prosodic cues. Because adult participants showed the same interpretative bias irrespective of whether they heard the sentences spoken aloud or not, we
were able to conclude that unintentional prosodic cues did not unduly affect our
results with adults. However, as we neither systematically manipulated prosody
nor directly controlled it, these experiments do not enable us to form any conclusions concerning the role of prosodic cues in focus assignment.
We believe that these data have an important bearing on the development of
processing accounts of focus interpretation. As we noted in the Introduction, there
are well-developed theoretical accounts of focus interpretation in the formal linguistics literature but a relative scarcity of psycholinguistic processing accounts.
6
6We
are grateful to an anonymous reviewer for bringing this issue to our attention.
282
PATERSON ET AL.
We believe that these results will contribute to the development of psycholinguistic theories by highlighting the nature of focus assignment ambiguities and
their implication for sentence comprehension processes.
ACKNOWLEDGMENTS
We are grateful to Susan Dorsett for collecting data for Experiment 2 and the
adult data for Experiment 3 and to Tom Warne for drawing the pictures used in
these experiments. We also are grateful to two anonymous reviewers and to Ken
Drozd for providing detailed feedback on an earlier version of this article. S. P.
Liversedge acknowledges the support of Grant S19168 from the Biotechnology
and Biological Sciences Research Council (United Kingdom).
REFERENCES
Carlson, K. (2004) “Syntactic vs. Prosodic Focus Effects in Parsing,” poster presented at the 17th Annual CUNY Conference on Human Sentence Processing, College Park, Maryland.
Clifton, C., Jr., J. Bock, and J. Rado (2000) “Effects of the Focus Particle Only and Intrinsic Contrast
on Comprehension of Reduced Relative Clauses,” in A. Kennedy, R. Radach, D. Heller, and J.
Pynte, eds., Reading as a Perceptual Process, Elsevier, Amsterdam, The Netherlands.
Crain, S., W. Ni, and L. Conway (1994) “Learning, Parsing and Modularity,” in C. Clifton, L. Frazier,
and K. Rayner, eds., Perspectives on Sentence Processing, Lawrence Erlbaum Associates, Inc.,
Hillsdale, New Jersey.
Crain, S., W. Philip, K. Drozd, T. Roeper, and K. Matsuoka (1992) “Only in Child Language,” ms.,
University of Connecticut, Storrs.
Crain, S. and M. Steedman (1985) “On Not Being Led Up the Garden Path: The Use of Context by the
Psychological Syntax Processor,” in D. R. Dowty, L. Kartunnen, and A. M. Zwicky, eds., Natural
Language Parsing: Psychological, Computational, and Theoretical Perspectives, Cambridge University Press, Cambridge, England.
Crain, S., R. Thornton, C. Boster, L. Conway, D. Lillo-Martin, and E. Woodams (1996) “Quantification Without Qualification,” Language Acquisition 5, 83–153.
Cresswell, M. J. and A. von Stechow (1982) “De re Belief Generalized,” Linguistics and Philosophy 5,
503–535.
Drozd, K. (2001) “Children’s Weak Interpretations of Universally Quantified Questions,” in M.
Bowerman and S. C. Levinson, eds., Language Acquisition and Conceptual Development, Cambridge University Press, Cambridge, England.
Drozd, K. and E. van Loosbroek (1998) “Dutch Children’s Interpretations of Focus Particle Constructions,” poster presented at the 23rd Annual Boston University Conference on Language Development, Boston, Massachusetts.
Filik, R., K. B. Paterson, and S. P. Liversedge (2005) “Parsing With Focus Particles in Context: Evidence From Eye Movements in Reading,” Journal of Memory and Language 53, 473–495.
Fodor, J. D. (2002) “Prosodic Disambiguation in Silent Reading,” NELS 32, 113–132.
Frazier, L. (1999) On Sentence Interpretation, Reidel, Dordrecht, The Netherlands.
Gennari, S., L. Meroni, and S. Crain (2004) “Rapid Relief of Stress in Dealing With Ambiguity,” in
J. C. Trueswell and M. K. Tanenhaus, eds., Approaches to Studying World-Situated Language Use:
INTERPRETATION OF AMBIGUOUS FOCUS
283
Bridging the Language-as-Product and Language-as-Action Traditions, MIT Press, Cambridge,
Massachusetts.
Grice, H. P. (1975) “Logic and Conversation,” in P. Cole and J. Morgan, eds., Syntax and Semantics,
Vol. 3, Academic Press, New York.
Gualmini, A., S. Maciukaite, and S. Crain (2002) “Children’s Insensitivity to Contrastive Stress in
Sentences With Only,” Proceedings of the 25th Annual Penn Linguistics Colloquium, University of
Pennsylvania Press, Philadelphia. (Penn Linguistics Club, http://www.ling.upenn.edu/papers/
pwpl.html)
Halbert, A. M., S. Crain, D. Shankweiler, and E. Woodams (1995) “Interpretive Use of Emphatic
Stress by Children,” poster presented at the 8th Annual CUNY Conference on Human Sentence
Processing, Tucson, Arizona.
Jackendoff, R. S. (1972) Semantic Interpretation in Generative Grammar, MIT Press, Cambridge,
Massachusetts.
Jacobs, J. (1986) “The Syntax of Focus and Adverbials,” in W. Abraham and S. de Meij, eds., Topic,
Focus, and Configurationality, John Benjamins, Amsterdam, The Netherlands.
Kadmon, N. (2001) Formal Pragmatics, Blackwell, Oxford, England.
Kiss, K. É. (1998) “Identificational Focus Versus Information Focus,” Language 74, 245–273.
König, E. (1991) The Meaning of Focus Particles, Routledge, New York.
Krifka, M. (1992) “A Compositional Semantics for Multiple Focus Constructions,” in J. Jacobs, ed.,
Informationsstruktur und Grammatik, Westdeutscher Verlag, Opladen, Germany.
Liversedge, S. P., K. B. Paterson, and E. Clayes (2002) “The Influence of Only on Syntactic Processing of ‘Long’ Relative Clause Sentences,” Quarterly Journal of Experimental Psychology 55A,
225–240.
Ni, W., S. Crain, and D. Shankweiler (1996) “Sidestepping Garden Paths: The Contribution of Syntax,
Semantics and Plausibility in Resolving Ambiguities,” Language and Cognitive Processes 11,
283–334.
Paterson, K. B., S. P. Liversedge, and R. Filik (2005) “Focus Assignment in Reading,” poster presented at the Architectures and Mechanisms for Language Processing Conference, Ghent, Belgium.
Paterson, K. B., S. P. Liversedge, C. Rowland, and R. Filik (2003) “Children’s Comprehension of Sentences With Focus Particles,” Cognition 89, 263–294.
Paterson, K. B., S. P. Liversedge, and G. Underwood (1999) “The Influence of Focus Operators on
Parsing of Short Reduced Relative Clause Sentences,” Quarterly Journal of Experimental Psychology 52A, 717–738.
Philip, W. and E. Lynch (1999) “Felicity, Relevance, and Acquisition of the Grammar of Every and
Only,” in S. C. Howell, S. A. Fish, and T. Keith-Lucas, eds., Proceedings of the 24th Annual Boston
University Conference on Language Development, Cascadilla Press, Somerville, Massachusetts.
Reinhart, T. (1999) “The Processing Cost of Reference-Set Computation: Guess Patterns in Acquisition,” ms., Utrecht University, Utrecht, The Netherlands.
Rooth, M. (1992) “A Theory of Focus Interpretation,” Natural Language Semantics 1, 75–116.
Sedivy, J. C. (2002) “Invoking Discourse-Based Contrast Sets and Resolving Syntactic Ambiguities,”
Journal of Memory and Language 46, 341–370.
Selkirk, E. O. (1995) “Sentence Prosody: Intonation, Stress and Phrasing,” in J. Godsmith, ed., Handbook of Phonological Theory, Blackwell, Oxford, England.
Von Stechow, A. (1981) “Topic, Focus, and Local Relevance,” in W. Klein and W. Levelt, eds.,
Crossing the Boundaries in Linguistics, Reidel, Dordrecht, The Netherlands.
Submitted 4 June 2004
Final version accepted 24 December 2005
284
PATERSON ET AL.
APPENDIX
Sentences Used in Experiments
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
The
The
The
The
The
The
The
The
The
The
woman is (only) walking a dog.
boy is (only) pulling a cart.
boy has (only) caught a fish.
woman is (only) watering a flower.
boy is (only) throwing a ball.
nurse is (only) carrying a bucket.
boy is (only) painting a chair.
woman is (only) pushing a pram.
postman is (only) holding a letter.
girl is (only) cutting a cake.