Joao Costa Assimetria
-
Upload
layton-hariett -
Category
Documents
-
view
226 -
download
0
Transcript of Joao Costa Assimetria
-
7/25/2019 Joao Costa Assimetria
1/18
Subjectobject asymmetries in the acquisition of Portuguese relativeclauses: Adults vs. children
Joao Costa *, Maria Lobo, Carolina Silva
FCSH-Universidade Nova de Lisboa, Portugal
1. Introduction
A-bar dependencies established via movement, like those involved in the formation of questions, relative clauses or
topicalization constructions, are a well-studied domain in language acquisition. Recent studies reveal that children are able
to perform this type of movement operation, confirming early findings by Borer and Wexler (1987). However, unlike these
authors suggest, it is not the case that all subtypes of A-bar dependencies are produced or comprehended by children at the
same time. Several studies, conducted for many different languages, lead to the conclusion that dependencies in which the
subject is moved are acquired earlier than those in which the object is moved ( Adams, 1990; Berman, 1997; Brown, 1972;
Correa, 1982, 1995; de Villiers et al., 1994; de Vincenzis, 1991; Friedmann et al., 2009; Friedmann and Novogrodsky, 2004;
Hakansson and Hansson, 2000; McKee et al., 1998; Roth, 1984; Sheldon, 1974; Tavakolian, 1981; Vasconcelos, 1991 ). Some
of these studies further show that the difficulties in production and in comprehension of object dependencies are higher
when the moved object crosses an intervening lexically overt subject, as depicted in (1):
(1) [ Obj [ Subj [ V t
[
(1) [ Obj [ Subj [ V t
If, for instance, the subject is null, no intervention effect arises, and childrens ability to produce and comprehend the
dependencies improves substantially (e.g. Friedmann et al., 2009; Friedmann and Costa, 2010).
Lingua 121 (2011) 10831100
A R T I C L E I N F O
Article history:
Received 1 June 2010
Received in revised form 2 November 2010
Accepted 3 February 2011
Keywords:
Subject relative
Object relative
Acquisition
Production
Comprehension
European Portuguese
A B S T R A C T
Several studies reveal an asymmetry between the acquisition of subject and object relative
clauses (Adams, 1990; Berman, 1997; Brown, 1972; Correa, 1982, 1995; de Villiers et al.,
1994; de Vincenzis, 1991; Friedmann et al., 2009; Friedmann and Novogrodsky, 2004;
Hakansson and Hansson, 2000; McKee et al., 1998; Roth, 1984; Sheldon, 1974; Tavakolian,
1981; Vasconcelos, 1991). The goal of this paper is to show that a similar asymmetry is
found in the acquisition of European Portuguese. FollowingBelletti and Contemori (2010),
we show that, in production, the adult control group also fails to produce object relative
clauses. However, similarly to what was found for Italian by Belletti and Contemori, a
closer look into the types of productions made by adults and children reveals that only the
latter fail to produce relatives. Adults, instead, opt for other grammatical alternative
strategies.
2011 Elsevier B.V. All rights reserved.
* Corresponding author.
E-mail addresses: [email protected](J. Costa),[email protected](M. Lobo),[email protected](C. Silva).
Contents lists available at ScienceDirect
Lingua
j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / l i n g u a
0024-3841/$ see front matter 2011 Elsevier B.V. All rights reserved.doi:10.1016/j.lingua.2011.02.001
http://dx.doi.org/10.1016/j.lingua.2011.02.001mailto:[email protected]:[email protected]:[email protected]://www.sciencedirect.com/science/journal/00243841http://www.elsevier.com/locate/linguahttp://dx.doi.org/10.1016/j.lingua.2011.02.001http://dx.doi.org/10.1016/j.lingua.2011.02.001http://www.elsevier.com/locate/linguahttp://www.sciencedirect.com/science/journal/00243841mailto:[email protected]:[email protected]:[email protected]://dx.doi.org/10.1016/j.lingua.2011.02.001 -
7/25/2019 Joao Costa Assimetria
2/18
The goal of this paper is twofold. First, we will show that the well-studied subjectobject asymmetries found in the
production of A-bar dependencies, more specifically in the production of relative clauses, are also found in European
Portuguese, providing further support to the crosslinguistic robustness of relative clauses as an area of linguistic development.
The second goal of the paper is methodological in nature. We will show that a first look at the adult control group data reveals
that, just like children, adults also exhibit a subjectobject asymmetry in the production of relative clauses. In order to
understandthis behavior, we willarguethat a qualitativeanalysis of theperformanceof bothgroups (children andadults), anda
comparison between their performances in a production task and in a comprehension task is needed. In a nutshell, following
Friedmann et al. (2009), we show that children do not produce object relatives, because they are not able to deal with the
intervention created in the syntactic dependency, whereassome adults do not produce object relatives justbecause theyprefer
to produce alternative structures, like passives. The fact that the results of the two groups are qualitatively different provides
evidence for claiming that children and adults have different underlying reasons for avoiding object relatives.
The article is organized in the following way.
In section2, we provide an overview of work on this topic as background to this study, our research questions and
hypotheses. In section 3, an experiment eliciting the production of subject and object relative clauses in European
Portuguese, borrowed from Novogrodsky and Friedmann (2006), is presented. In section4, we present the results of an
experiment on the comprehension of subject and object relatives in European Portuguese, modeled after Friedmann and
Novogrodsky (2004). Section 5 provides a discussion of the results obtained for children and for adults, which tries to explain
why adults apparently fail to produce object relatives in the elicitation experiment. Section6synthesizes the papers main
conclusions.
2. Background, research questions and hypotheses
As mentioned in section1, the acquisition of A-bar dependencies involving movement has received a lot of attention in
the literature on child language and language development (seeGuasti, 2002for a summary). Relative clauses instantiate a
case of an A-bar dependency with movement. As shown in (2), in European Portuguese, like in most worlds languages,
relative clauses may be headed by a relative pronoun which is either the subject (as in 2a) or the object (as in 2b)of its clause:
(2) a. Eu vi o rapaz que abracou a mae.
I saw the boy that hugged the mother
b. Eu vi o rapaz que a mae abracou.
I saw the boy that the mother hugged
Assuming that the two types of relative clauses involve the creation of an A-bar dependency via movement of the relative
operator, thecrucial difference between thetwo structures is that only (2b), the object relative clause, involves a dependencyin which a constituent the subject intervenes in between the moved operator and the arguments original position (the
complement of the verb). Results found for several languages suggest that it is intervention, rather than movement, that
causes childrens difficulties in comprehending and producing relative clauses, since they only fail with object relatives (e.g.
Friedmann and Novogrodsky, 2004; Friedmann et al., 2009). An interesting aspect of this type of analysis is that it is based on
a rather abstract level of representation (the nature of the dependency created), and not on any type of surface characteristic
of relative clauses. As such, it is predicted that the difficulties with object relative clauses hold crosslinguistically, whenever
the relevant abstract configuration emerges, and independently of the actual lexical material used in each language for the
creation of these structures. Based on this background, we are interested in evaluating the crosslinguistic robustness of the
findings on relative clauses, by assessing their production by children acquiring European Portuguese. We, therefore,
formulate the following hypothesis:
(3) H1: If the difficulties with object relative clauses depend on an abstract dependency in which there is intervening
material between the gap and the displaced element, and not on the specific relative pronoun or complementizerused in each language, children acquiring European Portuguese are expected to exhibit problems in the
production and comprehension of object relative clauses, as was found for other languages.
Friedmann and Novogrodsky (2004)explain the difficulties with object relative clauses in thematic terms: the intervention
of the subject makes it difficult for children to know which argument the trace position must associate with: the moved
operator or the subject. This problem with assignment of the relevant interpretation is subject to development. As such,
adults are not expected to have a problem in the interpretation and production of this type of dependencies with intervening
arguments.1 Based on this, we can formulate our second hypothesis:
(4) H2: If the problems in comprehending and producing crossing dependencies are subject to development,
children and adults are expected to diverge in the production and comprehension of object relative clauses.
1 Correa (1995)also provides evidence for the relevance of intervening lexical material in the comprehension of object relatives in Brazilian Portuguese.
J. Costa et al. / Lingua 121 (2011) 108311001084
-
7/25/2019 Joao Costa Assimetria
3/18
We will test these two hypotheses with the presentation of a production task and a comprehension test on relative clauses in
European Portuguese.
3. Experiment 1: production of subject and object relative clauses in European Portuguese
Methodology. In order to elicit subject and object relative clauses, we borrowed a test used in Novogrodsky and
Friedmann (2006), consisting of a preference test, in which children were asked to indicate their preferences after a
question posed by the experimenter. In this task, the child is invited to participate in an interview, in which s/he is
presented with two situations, and has to tell the experimenter which one s/he prefers. The elicitation of the relative
clause is guaranteed by the only rule the child has to follow: s/he must start all his/her answers by saying: I prefer to
be the child. . .. In (5a) and (5b), we provide an example of a test item for the elicitation of a subject relative and an object
relative, respectively2:
(5) a. Ha dois meninos. . . Um menino come chocolate, o outro menino come
Have-3sg two boys. . .one boy eats chocolate, the other boy eats
gelado. Que menino eque gostavas mais de ser?
Icecream. Which boy is that liked-impf-2sg more of be-inf
Comeca com: Gostava de ser o menino. . .
Start with liked-impf-1sg of be-inf the boy. . .
There are two boys. . .
One boy eats chocolate, the other boy eatsice cream. Which boy would you rather be?
Start with: Id rather be the boy. . .
b. Ha dois meninos. . . A mae penteia um menino, o vizinho penteia o
Have-3sg two boys. . .the mother combs one boy, the neighbor combs the
outro menino. Que menino eque gostavas mais de ser?
other boy. Which boy is that liked-impf-2sg more of be-inf
Comeca com: Gostava de ser o menino. . .
Start with liked-impf-1sg of be-inf the boy. . .
There are two boys. . . The mother combs a boy, the neighbor combs the
other boy. Which boy would you rather be?
Start with: Id rather be the boy. . .
In the test, twentyrelative clauses were induced, ten of each type (subject and object).The stimuli includedbothreversible and
nonreversible predicates. That is, in bothconditions, one can find relatives with a reversible predicate, as in I would liketo bethe
boythat themotherhugs (itcould be thecasethatthe motherhugsthe boy or the boyhugsthe mother),and relatives with a non-
reversible predicate, as inI would like to be the boy that the alarm clock wakes up (in this case, only the alarm clock can be the
subject). As can be seen in Appendix A, the stimuli included 8 reversible predicates (6 in the subject condition, 2 in the object
condition), and 12 non-reversible predicates (4 in the subject condition, 8 in the object condition). Controlling for
thereversibility of thepredicate is important, since we are concerned about the thematic interpretationof thearguments. In the
case of non-reversible predicates, the meaning of the verb may turn out to be an important cue, whereas with reversible
predicates there is no help given by the verb. In order to control for the effects of animacy (cf. Goodluck and Tavakolian, 1982),
some of the non-reversibleitems include animate and inanimate DPs, both in the subjectand in the objectconditions. As canbe
seen inAppendix A,the stimuli included 14 animate predicates (6 in the subject condition, 8 in the object condition) and 6
inanimate DPs (4 in thesubjectcondition,2 in the object condition).Thestimuli alsoincludedboth definiteand indefinite DPs inboth conditions. As can be seen in Appendix A, the stimuli included 10 definite DPs (2 in the subject condition, 8 in the object
condition) and 10 indefinite DPs(8 in the subject condition, 2 in the object condition). As we will show, therewas a generalized
good performance on subject relatives independently of reversibility, animacy and definiteness.
All participants were tested individually in a quiet room by two experimenters. One of the experimenters prompted the
sentences, and the other transcribed the childs response (the same role was played by the same experimenter for all
children). No time limit was imposed, and no stimulus or correction was given depending on the type of response, besides
general encouragement to pursue the task and a final reward after completion of the task (for the children only). The tests
were recorded using a digital voice recorder. Childrens productions were fully transcribed during and after the session by
the two experimenters. One transcription was made by one of the experimenters during the session, right after the childs
utterance. A second transcription was made on the basis of the audio recording by the other experimenter. In case of false
2
We refer the reader toFriedmann et al. (in preparation), in which the details of the adaptation of this test to sixteen different languages are presented.SeeAppendix Afor the list of sentences used in the elicitation task.
J. Costa et al. / Lingua 121 (2011) 10831100 1085
-
7/25/2019 Joao Costa Assimetria
4/18
starts or reformulations, everything was transcribed, but only the last utterance of the child was counted and coded. The
second transcriptions were double-checked by the two experimenters, and compared with the original transcriptions for
reliability. This means that each experimenter heard the recordings and checked the transcriptions, comparing the final
transcript and the original one. No case of disagreement between the original and the final transcript occurred.
Thesentences were coded as targetor non-target. We considered targetall theproductions includinga relative clauseof
the type being elicited,and non-target allthe productions that didnot include a relative clauseof thetypebeingelicited. All
non-target responses were scored as grammatical or ungrammatical. They were further classified according to a
description of theresponsestrategy used(e.g. passive,resumptive pronoun,omission of the relativepronoun, thematicrole
reversal, repetition of the antecedent, head change, and chopping relative). In section 5, we present the most relevant
categories used in the coding of grammatical and ungrammatical strategies, and examples for each of the strategies
referred to here.
Participants.The childrens group included 60 typically developing monolingual children acquiring European Portuguese
(25 male, 35 female), aged between 3;9 and 6;2 (mean: 5;1). No child had any history of social, emotional, cognitive or
hearing impairment. All children were from the Lisbon area. The adult control group consisted of 20 adults (9 male, 11
female), aged between 23 and 45. All adults had a university background and no specific training or education in the area
they were being tested for.
Results.The results of the elicitation experiment are presented separately for children and adults. We will examine the
case for the children first. As the graph in Fig. 1shows, there was a very significant difference between the two types of
relatives3.
The results show a clear asymmetry between subject and object relative clauses, favoring the former over the latter,
which confirms the tendency found for other languages. A look at individual results, presented in Table 1, reveals that there is
no clear age effect.
There are two important aspects worth emphasizing in the individual results. First, at all age groups, there are a few
children who are better than others in object relatives, which attests to some individual variation. Second, and more
importantly, at all age groups and for all children, subject relatives are better than object relatives. In other words, we did not
find any child who could produce object relatives better than subject relatives.
Reversibility did not play an important role in determining accurate production of subject and object relatives. As shown
inthe graph in Fig. 2, there is no relevant difference between the production of relatives with or without reversible predicates
within each condition.4
The same holds for animacy, which did not play an important role in determining accurate production of subject and
object relatives. As shown in the graph in Fig. 3, there is no relevant difference between the production of relatives with or
without animate DPs within each condition.5
A similar finding emerged for definiteness. This factor did not play an important role in determining accurate production
of subject and object relatives. As shown in the graph in Fig. 4, there is no relevant difference between the production of
relatives with or without definite DPs within each condition6.
[
Fig. 1. Experiment 1: percentage of target relatives elicited overall, subject vs. object condition (children).
3 Recall that target is meant as relative clause of the type being elicited.4 The following stimuli sentences from Appendix A were counted as reversible: Subject condition: 5,6,7,8,9and10; Object condition: 1,4,5,6,7,8,9and10.5 The following stimuli sentences from Appendix A were counted as reversible: Subject condition: 5,6,7,8,9and10; Object condition: 1,4,5,6,7,8,9and10.6
The following stimuli sentences fromAppendix Awere counted as containing a definite DP:Subject condition:6 and 7;Object condition:1, 2, 3, 4, 5, 6, 7and 10.
J. Costa et al. / Lingua 121 (2011) 108311001086
-
7/25/2019 Joao Costa Assimetria
5/18
Note that, as mentioned in the description of the methodology, the stimuli sentences include reversible and non-reversible
predicates, definite and indefinite DPs, and animate and inanimate DPs, but are not balanced for these variables. As such, the
data in the previous threegraphs maybe only be taken as preliminary indications that these variables didnot affectthe results,
and that the subjectobject difference is the crucial differentiating factor. Independent tests controlling each of these variables
independently with balanced stimuli may help determining whether these preliminary results are confirmed.
Let us now look at the performance of the adult control group. The graph in Fig. 5presents the results of adults in the
elicitation task.
Table 1
Experiment 1: percentage of target relatives elicited by participant, subject vs. object condition.
Years;months Subject relatives Object relatives
3;9 10% 0%
4;3 80% 40%
4;3 50% 3%
4;4 20% 0%
4;4 90% 0%
4;4 100% 20%4;5 60% 20%
4;5 70% 0%
4;5 80% 0%
4;6 50% 0%
4;6 100% 90%
4;6 100% 50%
4;6 100% 0%
4;6 100% 100%
4;7 100% 70%
4;8 90% 50%
4;10 20% 0%
4;11 90% 50%
4;11 60% 10%
4;11 70% 50%
4;11 100% 0%
4;11 100% 6%4;11 100% 30%
5;0 90% 40%
5;1 90% 10%
5;3 80% 20%
5;3 100% 60%
5;3 100% 90%
5;3 100% 60%
5;3 100% 100%
5;4 30% 0%
5;4 60% 20%
5;4 100% 30%
5;4 100% 80%
5;4 100% 20%
5;4 100% 40%
5;5 90% 0%
5;5 60% 0%5;5 70% 0%
5;5 100% 80%
5;6 100% 30%
5;7 70% 0%
5;7 0% 10%
5;7 100% 10%
5;7 90% 20%
5;7 80% 20%
5;8 30% 10%
5;8 70% 10%
5;8 0% 0%
5;8 100% 90%
5;9 10% 0%
5;9 90% 20%
5;10 100% 90%
5;11 70% 40%
5;11 100% 0%
6;0 100% 70%
6;0 70% 10%
6;0 90% 0%
6;2 100% 100%
6;2 100% 0%
J. Costa et al. / Lingua 121 (2011) 10831100 1087
-
7/25/2019 Joao Costa Assimetria
6/18
[
Fig. 2. Experiment 1: percentage of target subject and object relatives elicited depending on reversibility of predicates children.
[
Fig. 3. Experiment 1: percentage of target subject and object relatives elicited depending on animacy of DPs children.
[
Fig. 4. Experiment 1: percentage of target subject and object relatives elicited depending on definiteness of DPs children.
J. Costa et al. / Lingua 121 (2011) 108311001088
-
7/25/2019 Joao Costa Assimetria
7/18
Just like we observed for children, in adults too, there is an obvious asymmetry between subject and object relatives. As
shown in the graph inFig. 5,subject relatives are produced 99.5% of the time (199/200), whereas object relatives were only
produced at a rate of 49.5% (99/200). The results are, therefore, very similar to those obtained for children. However, as we
will show later, the results are only superficially similar, since adults finish the sentences by producing grammatical
alternatives to object relatives, like passives (as in (50a), contrasted to the target object relative clause in (5 0b)), whereas
children produce ungrammatical structures, like sentences in which the thematic roles are reversed (as in (6a), which
contrasts to the target object relative clause in (6b)):
(50) a. Grammatical Subject Passive:
Gostava de ser o menino que e penteado pelo avo.
Liked-impf-1sg of be-inf the boy that is combed by-the grandpa
Id rather be the boy that is combed by the grandpa.
b. Target object relative:
Gostava de ser o menino que o avopenteia.
Liked-impf-1sg the boy that the grandpa combs
Id rather be the boy that grandpa combs.
(6) a. Ungrammatical thematic role reversal:
Gostava de ser o menino que acorda o radio.
Liked-impf-1sg of be-inf the boy that wakes the radio
Id rather be the boy that wakes up the radio.
b. Target object relative clause:
Gostava de ser o menino que o radio acorda.
Liked-impf-1sg of be-inf the boy that the radio wakes
Id rather be the boy that the radio wakes up.
We will return to this matter in the discussion section where individual data will be presented (cf. Fig. 9), but the reader can
anticipate that there is a difference between adults and children, in that adults prefer not to produce object relatives,
whereas children are not able to produce them.
If we return to the hypotheses formulated at the beginning of the article, we may say that the results of Experiment 1
confirm Hypothesis 1, since it was found that, like in other languages, object relatives are produced less often than subject
relatives. On the contrary, Hypothesis 2 was not supported, since no difference was found between children and adults.
Actually, the lack of confirmation for Hypothesis 2 compromises the validity of Hypothesis 1, since the validity of the
experiment is maybe challenged by the results of the adult control group. At this moment, three interpretations of this result
are legitimate: (i) the experiment is not valid; (ii) the subjectobject asymmetry in relatives is not a marker of language
development, since it also found in adults; (iii) the similarity in results between children and adults is misleading.
In the remainder of the paper, we will provide evidence for the interpretation iii. Experiment 2, which will be presented in
the next section, will constitute the first piece of evidence in favor of this interpretation of the facts.
[
Fig. 5. Experiment 1: percentage of target relatives elicited overall, subject vs. object condition adults.
J. Costa et al. / Lingua 121 (2011) 10831100 1089
-
7/25/2019 Joao Costa Assimetria
8/18
4. Experiment 2: comprehension of subject and object relative clauses in European Portuguese
The second experiment aims at testing whether there are subjectobject asymmetries in the comprehension of relative
clauses, which would match the findings for production.
Methodology. In order to assess the comprehension of subject and object relative clauses, we used Friedmanns (1998)
relative clause comprehension sentence-picture matching task. This task consists of 40 pairs of pictures with two characters
in each of the pictures one of the characters is performing an action on the other, e.g. a boy drying a hippo or a hippo drying
a boy. The experimenter asks the child to point to the relevant picture, using sentences containing relative clauses, such as
Show me the boy that dries the hippo or Show me the boy that the hippo dries. Twenty subject relatives and twenty
object relatives were used in the experiment (cf.Appendix Bfor the list of stimuli used). Given the nature of the task, only
reversible verbs were used. InFig. 6, we provide an example of the type of pictures used, kindly provided by N. Friedmann.
All participants were tested individually in a quiet room. No time limit was imposed, and no stimulus or correction was
given depending on the type of response, besides general encouragement to pursue the task and a final reward after
completion of the task (for the children only). The responses were registered by an experimenter during the session.
Participants.The participants of Experiment 2 were exactly the same as in experiment 1, both in the childrens group and
in the adult group. The two experiments were planned together, so that we could assess potential differences between
[
Fig. 6. Example of test item from Friedmanns (1998)comprehension test.
J. Costa et al. / Lingua 121 (2011) 108311001090
-
7/25/2019 Joao Costa Assimetria
9/18
production and comprehension in the two groups of participants. It is important to emphasize, though, that Experiment 2
was run after Experiment 1, on separate days (with at least one week of interval). We wanted to make sure, this way, that the
production task was by no means influenced by the type of structures uttered by the experimenters in the comprehension
task, since the latter includes the subject and object relatives to be elicited in the production experiment.
Results.Let us first look at the results of children in the comprehension task. As shown, in Fig. 7, a clear subjectobject
asymmetry was found:
Subject relatives were comprehended almost perfectly, with a rate of success of 96%, whereas in object relatives children
only reached 68% of correct interpretations. The percentage of comprehension of object relatives is higher than the
production of object relatives, which is a normal effect of comprehension tasks. Although the performance on object relatives
is above chance, the difference between the two conditions is statistically significant (x2 = 58.48,p < .0001). As shown in
Table 2, just like in the case of production, there is some individual variation, but almost no case in which there is better
comprehension of subject relatives than object relatives (only one case, which is signaled).
Adults performed at ceiling on both conditions, reaching 100% in subject and in object relatives).
Unlikewhat wasfound in Experiment 1 forproduction, adults and childrendo not perform alikein the comprehension task.
In isolation, the results of Experiment 2 confirm the two Hypotheses being tested in this article: on the one hand, the results of
thecomprehensiontask confirm thecrosslinguistic robustness of thesubjectobjectasymmetryin relativeclauses, on theother
hand, it appears that there is a developmental effect, given the difference in results between children and adults (Fig. 8).
Combining the results of the two experiments, we are left with a new issue to address concerning the behavior of the
adult control group. Why is there an asymmetry in performance between the two experiments? In other words, how can we
understand the fact that adults behave like children in the production of object relative clauses, if they can comprehend them
with no difficulty, as shown by the results of Experiment 2?
Note that, by now, we can rule out the possibility suggested at the end of the previous section that object relatives are also
difficult for adults (and therefore not an adequate marker of language development), since the comprehension data provide
evidence that they pose no problem for the adult control group. We are thus left with two other possibilities: either the
production task is not a valid assessment tool or the production results are to be interpreted differently.
[
Fig. 7. Experiment 2: percentage of target comprehension of relatives overall, subject vs. object condition children.
[
Fig. 8. Experiment 2: percentage of target comprehension of relatives overall, subject vs. object condition adults.
J. Costa et al. / Lingua 121 (2011) 10831100 1091
-
7/25/2019 Joao Costa Assimetria
10/18
Table 2
Experiment 2: percentage of target comprehension of relatives by participant, subject vs. object condition.
Age Subject relatives Object relatives
3;11 90% 90%
4;3 80% 50%
4;5 100% 60%
4;5 95% 55%
4;5 90% 65%
4;6 100% 90%4;6 100% 20%
4;6 90% 60%
4;6 95% 45%
4;7 100% 90%
4;7 95% 85%
4;7 100% 90%
4;7 60% 65%
4;7 95% 90%
4;8 100% 40%
4;8 95% 75%
4;8 100% 50%
4;10 100% 70%
4;11 100% 80%
4;11 95% 95%
4;11 100% 55%
5;0 100% 90%5;0 100% 60%
5;0 90% 80%
5;1 100% 80%
5;1 100% 70%
5;2 100% 50%
5;2 100% 60%
5;2 85% 55%
5;4 100% 85%
5;4 90% 90%
5;5 100% 50%
5;5 100% 70%
5;5 100% 80%
5;5 90% 80%
5;5 100% 90%
5;5 100% 80%
5;5 100% 55%5;6 100% 70%
5;6 100% 75%
5;6 100% 60%
5;6 100% 90%
5;6 100% 90%
5;7 90% 20%
5;7 100% 70%
5;7 100% 100%
5;8 100% 20%
5;9 100% 50%
5;9 100% 60%
5;9 100% 20%
5;9 100% 70%
5;9 100% 60%
5;9 100% 50%
5;10 100% 20%
5;10 90% 70%
5;10 80% 70%
5;10 100% 55%
5;11 100% 70%
5;11 95% 90%
5;11 100% 70%
6;1 90% 60%
6;1 100% 70%
6;2 100% 70%
6;2 100% 80%
6;2 80% 55%
6;2 100% 25%
6;4 90% 30%
6;4 100% 80%
6;4 95% 90%
J. Costa et al. / Lingua 121 (2011) 108311001092
-
7/25/2019 Joao Costa Assimetria
11/18
5. Discussion reconciling the results of the two experiments.
The results of the two experiments only led to a partial confirmation of the two hypotheses being tested in this paper. On
the one hand, it was found that, as was found for other languages, children produce more subject relative clauses than object
relative clauses, making it legitimate to suppose that there is a subjectobject asymmetry in the acquisition of relative
clauses. On the other hand, the same asymmetry was found in adults, which makes one doubt that there is a developmental
problem and an acquisitional asymmetry. However, the results of Experiment 2 the comprehension test revealed that
there is an asymmetry between subject and object relatives for children only, since only children, and not adults, have
difficulties in the comprehension of object relatives.
Based on these results, we have to ask the following questions:
(a) Why do adults behave like children in the production task only?
(b) Given the lack of evidence for a developmental effect, can it be considered that the production of relative clauses is a
marker of linguistic development?
(c) Is the production task used for assessing knowledge of relative clauses a valid one?
As mentioned above, we contend that the lack of knowledge on object relatives that adults exhibit in the results of
Experiment 1 is just apparent, in the sense that it does not really reflect lack of competence on the production of object
relatives. In fact, a qualitative analysis of the data provides a key to reconciling the results of the two experiments. The first
question to be addressed is the following: what were adults doing when they were not producing object relative clauses? In
the graphic inFig. 9, we provide individual data for adults.
The first interesting aspect to be noted is that there is variation in the object relative condition (but not in the subject
relative condition): some participants consistently produced object relative clauses, others avoided them altogether, and a
few only produced them sometimes. Interestingly, whenever an object relative clause was not produced, one of two
alternative structures was produced: a subject relative with a passive, as in (7a), or a reduced passive (without any relative
operator), as in (7b):
(7) Passive subject relative clause:
a. Gostava de ser o menino que e penteado pela mae.
Id like to be the boy that is combed by the mother
Reduced passive:
b. Gostava de ser o menino penteado pela mae.
Id like to be the boy combed by the mother
In the following figure, we show the distribution of the two types of productions by individual adult participants:
As shown in the graph inFig. 10, some participants consistently opted for the production of passives (reduced or full
subject relatives). Crucially, all responses were fully grammatical and legitimate alternatives in the grammar.
This is different from what can be observed in childrens results. Let us first consider the rate of grammatical and
ungrammatical responses children gave in each condition (Figs. 11 and 12).
The qualitative analysis reveals that, unlike adults, children produce ungrammatical structures in the object relative
condition. An analysis of the type of target-deviant structures uttered by children revealed that the following categories of
productions emerged (an example is given for each error type for a given target response):
[
Fig. 9. Experiment 1: target production of relative clauses elicited by participant, subject vs. object condition adults.
J. Costa et al. / Lingua 121 (2011) 10831100 1093
-
7/25/2019 Joao Costa Assimetria
12/18
(8) Target: Gostava de ser o menino que o avovisita.
Id like to be the child that the grandpa visits
a. Omission of the relative pronoun:
Gostava de ser o menino o avovisita.
Id like to be the boy 1 the grandpa visits
b. Thematic role reversal:
Gostava de ser o menino que visita o avo.
Id like to be the child that visits grandpa
c. Repetition of the antecedent:
Gostava de ser o menino que o avo visita o menino.
Id like to be the child that grandpa visits the child
[
Fig. 12. Experiment 1: percentage of responses per grammaticality in object condition: children.
[
Fig. 10. Experiment 1: types of responses in object relative condition adults.
[
Fig. 11. Experiment 1: percentage of responses per grammaticality in subject condition: children.
J. Costa et al. / Lingua 121 (2011) 108311001094
-
7/25/2019 Joao Costa Assimetria
13/18
d. Insertion of resumptive clitic:
Gostava de ser o menino que o avo o visita.
Id like to be the child that grandpa him-cl visits
e. Relative clause with null argument:
Gostava de ser o menino que visita.
Id like to be the child that visits.f. Ungrammatical non-relative subordinate clause:
Gostava de ser o menino o avo a visitar.
Id like to be the child the grandpa at visit-infinitive
g. Head change:
Gostava de ser o avoque visita o menino.
Id like to be the grandpa that visits the child
Out of these patterns of response, two were found more frequently: reversal of thematic roles (as in 8b) and repetition of the
argument (with a noun phrase (as in 8c) or a clitic pronoun (as in 8d)). Reversal of thematic roles was found in 10% of the
responses.7 Repetition of the antecedent was found 15% of the time (which is the sum of cases in which the antecedent is
repeated as a DP coded as repetition of the antecedent inTable 3, as a pronoun coded as resumptive pronoun inTable
3, or as a DP in sentences with null subject coded as null subject + repetition of the antecedent in Table 3). The otherresponse patterns or combinations of these types of deviances never reached rates above 4%, as shown inTable 3(leaving
aside the category other ungrammatical strategies, which corresponds to a category in which we counted all cases with no
more than three occurrences. The value of 14% emerging there is, therefore, just the sum of many different responses found).
The two most frequent error patterns have an aspect in common: in both cases, the relative operator is not interpreted as
an object in thecase of reversal, because it is turnedinto a subject, in the case of resumption, because the argument position
is filled in with a noun phrase or with a pronoun. This gives support to the idea that the problems with object relatives are
due to the assignment of thematic roles to the arguments involved, and not to any type of structural deficit ( Friedmann and
Novogrodsky, 2004; Friedmann et al., 2009). The crossing dependency created by the movement of the object across the
subject makes it difficult to determine the interpretation of the object, and children resort to strategies by which the
interpretation of the object is satisfied in some alternative way (by insertion of an additional argument in the object position
or by eliminating the intervention configuration).
A further important piece of information comes from a comparison between children and adults in the strategies for
avoiding the use of object relative clauses. As we have shown above, some adults produced passive subject relatives orreduced passives instead of object relatives. These are legitimate alternatives, since they are grammatical structures, and
adequate in the elicitation context. As shown in Fig. 13, the two strategies that adults used when they avoided object
relatives were not very significant in childrens productions (in this figure, we present only a subset of the grammatical types
of answers given in order to allow for a comparison with the graph presented in Fig. 10for adults).
What is important to observe is that, unlike in the adult group, passives are not a strategy widely used by children.
Actually, the 8% production in the graph in Fig. 13 represents a raw total of 46 passives out of the 600 elicitations.
Interestingly, 35 of these 46 passives were produced by 4 children only, who consistently produced passives. The other 56
children avoided passives. Again, this differs from what was found for adults. Recall that all adults who did not produce
object relatives opted for a passive structure. This shows that passive is a valid option for adults, but not for the vast majority
Table 3
Experiment 1: percentage of error types in object relative condition children.
Error type
Omission of the relative pronoun 1% (5/600)
Thematic role reversal 10% (62/600)
Repetition of the antecedent 9% (56/600)
Resumptive pronoun 3% (19/600)
Relative clause with null argument 4% (24/600)
Ungrammatical non-relative subordinate clause 3% (19/600)Head change 0.3% (2/600)
Null subject + repetition of antecedent 3% (20/600)
Chopping relative 0.2% (1/600)
Other ungrammatical strategies 14% (84/600)
7 As argued in Belletti (2008), theta-role reversal may be easily confused with a grammatical object relative with subjectverb inversion. A
disambiguation test reported in Costa et al. (2009) reveals that children acquiring European Portuguese do not produce inversion in relative clauses,confirming the interpretation of these results. This confirms similar findings for Italian reported inBelletti and Contemori (2010).
J. Costa et al. / Lingua 121 (2011) 10831100 1095
-
7/25/2019 Joao Costa Assimetria
14/18
of children. Another consequence of this result is that the issue of individual variation in what concerns the choice of passive
alternatives only arises for adults, not for children, who seem not to be able to resort to passives as a valid option. Regarding
the validity of this method, this result also shows that the elicitation procedure is robust for children. As for adults, it is less
valid, since it does not make sure that a relative clause is going to emerge is the only potential output.
Bearing this description in mind, we can nowconcludethat the similarity in results between the children and theadults in
Experiment 1 was only a coincidence, not reflecting similar competence on the production of object relative clauses. A
qualitative analysis of the results revealed that there is no similarity between the two groups of participants. In fact, the
following differences emerged:
(a) Children utter ungrammatical structures in the context of relative clauses, unlike adults.
(b) Adults opt for passive structures as an alternative to object relative clauses, unlike children.
(c) Childrens productions reveal problems with the interpretation of the displaced object, unlike the productions of adults.
Based on these considerations, we can now return to the questions raised at the beginning of this section:
(a) Why do adults behave like children in the production task only?
(b) Given the lack of evidence for a developmental effect, can it be considered that the production of relative clauses is a
marker of linguistic development?
(c) Is the production task used for assessing knowledge of relative clauses a valid one?
We now have the means to provide simple answers to these three questions. As mentioned above, the similarity between
adults and children in the production task was just apparent. A qualitative analysis of the productions revealed that their
performances were quite different, and that children, unlike adults, did have problems with object relative clauses, replacing
them by ungrammatical structures. This makes the asymmetry between production and comprehension vanish, as it
becomes evident that adults do not have a problem with object relative clauses in production (since they never replace them
by ungrammatical structures in production) nor in comprehension.
As for question (b), we can now see that, in spite of prima facie evidence, there is indeed a developmental effect, and
children diverge from adults in production as well as in comprehension. As such, the production of relative clauses is a good
marker of linguistic development. An alternative explanation might consider that adults are better to children due to the
effects of schooling and because they have been exposed to object relatives in formal contexts. The research ofFontes (2008)
andValente (2008)reveals that schooling has a positive effect on the development of relatives with a PP, but not on the
development of object relatives. In fact both authors show that object relatives are not problematic for children who are
around 13 years old. InFriedmann and Costa (2010), it is shown that the poor performance on object relatives correlates
strongly with a poor performance on coordinated sentences involving a dependency (as in John saw Bill and sang). This
provides independent evidence for a problem with dependencies notcontingent on exposure to a particular type of structure
that might be acquired through formal education. Further evidence comes from the crosslinguistic similarities found in
Friedmann et al. (in press) in the response to this production task in 16 languages. It was found that, in all languages, children
produce similar types of responses, which provides good evidence to say that there is a general developmental problem and
not variation due to late exposure to these structures.
Finally, our results confirm the validity of the preference task as an assessment tool for knowledge of relative clauses, but
it also reveals that a quantitative analysis of the data is not sufficient. Our data suggest that a qualitative analysis is needed,
[
Fig. 13. Experiment 1: percentage of grammatical responses in object relative condition per type of response (subset) children.
J. Costa et al. / Lingua 121 (2011) 108311001096
-
7/25/2019 Joao Costa Assimetria
15/18
and that problems with relative clauses can be found whenever they are replaced by ungrammatical or target-deviant
structures. An adult-like performance is found in the production of object relative clauses or in their replacement by passive
subject relative clauses. Also, we saw that the production and comprehension tasks complement each other in the sense that
the results of the comprehension task confirm the poor performance on the production task.
Summarizing our results, we conclude that, in the production task, adults did not produce object relatives all the time
because they had grammatical alternatives available to them, which were legitimate surrogates. Children, on the other hand,
failed to produce object relatives, because they lacked some of the knowledge required to produce these structures, namely
the knowledge needed to ensure the right interpretation of the object moved across an intervening subject. Further, they had
no valid alternative to replace object relatives, unlike adults did.
On the basis of these conclusions, it is now possible to return to the original hypotheses presented at the beginning of this
article, which we repeat here for convenience:
(9) H1: If the difficulties with object relative clauses depend on an abstract dependency in which there is intervening
material between the gap and the displaced element, and noton the specific relative pronoun or complementizer
used in each language, children acquiring European Portuguese are expected to exhibit problems in the
production and comprehension of object relative clauses, as was found for other languages.
(10) H2: If the problems in comprehending and producing crossing dependencies are subject to development,
children and adults are expected to diverge in the production and comprehension of object relative clauses.
Both hypotheses were confirmed. The evidence found for European Portuguese confirms the crosslinguistic robustness of
object relative clauses as a marker of linguistic development, supporting Hypothesis 1. The differences between children andadults, found through a qualitative analysis of the productions of the two groups, provide the confirmation to Hypothesis 2.
6. Conclusion
In this article, we replicated previous tests on the production and comprehension of subject and object relative clauses,
and argued in favor of the following claims.
Object relative clauses are a good marker of linguistic development, since there are developmental effects and qualitative
differences between the results of children and those of adults.
The data collected for European Portuguese strengthen the claim that the difficulties with object relative clauses relate to
the type of abstract dependency created, and not to any type of surface property of this construction. Given the abstract
source of the problem, it is expected that difficulties are found crosslinguistically, and that they are not dependent on specific
properties of surface material.
The misleading similarities between the results of children and adults in the production task revealed that a qualitativeanalysis of the production of both groups is necessary in order to uncover the existence of difficulties with object relative
clauses. A mere quantitative analysis may hide a good performance on alternative legitimate structures, as also argued in
Belletti and Contemori (2010).
The types of errors produced by children in the production task lends support to the idea that childrens difficulties with
object relative clauses have to do with the interpretation of the displaced argument across the intervening subject.
The comprehension task is a good complement to the production task, since it revealed that there was indeed a
developmental effect, in spite of the appearances created by the quantitative data of the production experiment.
Appendix A. Production test
The elicitation test is the one described in Friedmann et al. (in preparation). The test consists of a preference task with the
following structure.
There are two children. One child does X, the other child does Y. Which child would you rather be? Start with: Id rather be
the child. . .
Below, we list the items used for the two conditions:
SUBJECT RELATIVES:
1. Um menino bebe coca-cola e o outro menino bebe agua.
One child drinks coke, and the other child drinks water
2. Um menino come gelado e outro menino come chocolate.
One child eats ice-cream, and the other child eats chocolate
3. Um menino recebe uma prenda e outro menino da uma prenda.
One child receives a gift, and the other child offers a gift
J. Costa et al. / Lingua 121 (2011) 10831100 1097
-
7/25/2019 Joao Costa Assimetria
16/18
4. Um menino encontra uma bola e outro menino compra uma bola.
One child finds a ball, and the other child buys a ball
5. Um menino assusta um amigo e outro menino desenha um amigo.
One child scares a friend, and the other child draws a friend
6. Um menino abraca a mae e outro menino empurra a mae.
One child hugs the mother, and the other child pushes the mother
7. Um menino visita o tio e outro menino convida o tio.One child visits the uncle, and the other child invites the oncle
8. Um menino desenha um polcia e outro menino desenha um cantor.
One child draws a policeman, and the other child draws a singer
9. Um menino encontra o professor e outro menino encontra um amigo.
One child finds the teacher, and the other child finds a friend
10. Um menino filma um cantor e outro menino filma um dancarino.
One child films a singer, and the other child films a dancer
OBJECT RELATIVES:
1. O medico trata um menino, a enfermeira trata outro menino.
The doctor treats one child, the nurse treats the other child
2. O radio acorda um menino, o despertador acorda o outro menino
The radio wakes up one child, the alarm clock wakes up the other child
3. O banho aquece um menino, o banho refresca o outro menino.
The bath warms a child, the bath cools the other child
4. O elefante molha um menino, o elefante levanta o outro menino.
The elephant wets one child, the elephant lifts the other child
5. O avoprocura um menino, o avoencontra o outro menino.
The grandpa seeks one child, the grandpa finds the other child
6. A tia fotografa um menino, a tia fotografa o outro menino.The aunt photographs one child, the aunt photographs the other child
7. O pai abraca um menino, o pai beija o outro menino.
The father hugs one child, the father kisses the other child
8. Um amigo abraca um menino, a mae abraca o outro menino.
A friend hugs one child, the mother hugs the other child
9. Um vizinho penteia um menino, o pai penteia outro menino.
A neighbor combs one child, the father combs the other child
10. O professor fotografa um menino, o avo fotografa o outro menino.
The teacher photographs one child, the grandpa photographs the other child
Appendix B. Comprehension test
The comprehension test was a translation ofFriedmanns (1998) relative clause comprehension task. Each picture is used
twice, one time for testing subject relatives, another time for testing object relatives. The test contains 20 pictures, which are
used for assessing the comprehension of 20 subject relatives and 20 object relatives. For a picture like the one in Fig. 4, the
following two sentences are used:
Subject relative: Mostra-me o menino que seca o hipopotamo.
Show me the boy that dries the hippo
Object relative: Mostra-me o menino que o hipopotamo seca.
Show me the boy that the hippo dries
J. Costa et al. / Lingua 121 (2011) 108311001098
-
7/25/2019 Joao Costa Assimetria
17/18
The following stimuli were used for other pictures:
Subject relatives
Mostra-me:
Object relatives
Mostra-me:
1 O menino que esta a molhar o pai. O menino que o pai estaa molhar.
The child that is wetting the father The child that the father is wetting
2 A girafa que esta a lamber a vaca. A girafa que a vaca estaa lamber.
The giraffe that is licking the cow The giraffe that the cow is licking3 O menino que estaa puxar o avo. O menino que o avoestaa puxar.
The boy that is pulling grandpa The boy that grandpa is pulling.
4 O menino que seca o hipopotamo. O menino que o hipopotamo seca.
The boy that dries the hippo The boy that the hippo dries
5 O medico que esta a pintar o soldado. O medico que o soldado estaa pintar.
The doctor that is painting the soldier The doctor that the soldier is painting.
6 O coelho que esta a empurrar o pinguim. O coelho que o pinguim estaa empurrar.
The rabbit that is pushing the penguin The rabbit that the penguin is pushing
7 O anao que estaa fotografar o prncipe. O anao que o prncipe estaa fotografar.
The dwarf that is photographing the prince. The dwarf that the prince is photographing.
8 A menina que estaa secar a mae. A menina que a mae esta a secar.
The girl that is drying the mother The girl that the mother is drying
9 O anao que estaa tapar o prncipe. O anao que o prncipe esta a tapar.
The dwarf that is covering the prince. The dwarf that the prince is covering
10 O menino que estaa beijar o avo. O menino que o avoestaa beijar.
The boy that is kissing grandpa The boy that grandpa is kissing11 A menina que esta a fotografar a enfermeira. A menina que a enfermeira estaa fotografar.
The girl that is photographing the nurse. The girl that the nurse is photographing.
12 O menino que esta a abracar o macaco. O menino que o macaco esta a abracar.
The boy that is hugging the monkey The boy that the monkey is hugging
13 A girafa que esta a lavar a menina. A girafa que a menina estaa lavar.
The giraffe that is washing the girl The giraffe that the girl is washing
14 O gato que estaa morder o cao. O gato que o cao estaa morder.
The cat that is biting the dog. The cat that the dog is biting.
15 A menina que estaa pintar a mae. A menina que a mae estaa pintar.
The girl that is painting the mother The girl that the mother is painting
16 A menina que estaa tapar a avo. A menina que a avo estaa tapar.
The girl that is covering grandma The girl that grandma is covering
17 O menino que esta a pentear o rei. O menino que o rei estaa pentear.
The boy that is combing the king The boy that the king is combing
18 O menino que esta a lavar o pinguim. O menino que o pinguim esta a lavar.
The boy that is washing the penguin. The boy that the penguin is washing.19 O anao que estaa filmar o prncipe. O anao que o prncipe estaa filmar.
The dwarf that is filming the prince. The dwarf that the prince is filming
20 O cao que esta a empurrar o menino. O cao que o menino estaa empurrar.
The dog that is pushing the boy. The dog that the boy is pushing.
References
Adams, C., 1990. Syntactic comprehension in children with expressive language impairment. British Journal of Disorders of Communication 25, 149171.Belletti, A., 2008. Relative clauses in Italian. In: Paper Presented at the COST A33 meeting in Warsaw.Belletti, A., Contemori, C., 2010. Intervention and Attraction. On the production of subject and object relatives by Italian (young) children and adults. In:
Costa, J., et al. (Eds.), Language Acquisition Development Proceedings of G.A.L.A 2009, Cambridge Scholars Press.Berman, R., 1997. Early acquisition of syntax and discourse in Hebrew. In: Shimron, Y. (Ed.), Psycholinguistic Studies in Israel: Language Acquisition,
Reading and Writing. Magnes Press (in Hebrew), Jerusalem, pp. 57100.Borer, H., Wexler, K., 1987. The maturation of syntax. In: Roeper, T., Williams, E. (Eds.), Parameter-setting and Language Acquisition. Reidel, Dordrecht, The
Netherlands.Brown, H., 1972. Childrens comprehension of relativized English sentences. Child Development 42, 19231936.Correa, L.M., 1982. Strategies in the acquisition of relative clauses. In: Aitchison, J., Harvey, N. (Eds.), Working Papers of the London Psycholinguistic
Research Group, vol. 4. pp. 3749.Correa, L.M.S., 1995. An alternative assessment of childrens comprehension of relative clauses. Journal of Psycholinguistic Research 24, 183203.Costa,J., Lobo, M.,Silva, C.,Ferreira,E., 2009. Producao e compreensao de oracoes relativas em portugues europeu:dados de desenvolvimento tpico,de PEDL
e do agramatismo In Textos Seleccionados. XXIV Encontro Nacional da APL. APL, Lisboa, pp. 211224.de Villiers,J.G., de Villiers,P.A., Hoban, E., 1994. The central problem of functional categories in the English syntax of oraldeaf children. In: Tager-Flusberg, H.
(Ed.), Constraints on Language Acquisition: Studies of Atypical children. Erlbaum, Hillsdale, NJ, pp. 947.de Vincenzi, M., 1991. Syntactic Parsing Strategies in Italian: The Minimal Chain Principle. Kluwer, Dordrecht, The Netherlands.Fontes, E., 2008. A producao de frases relativas restritivas no final do 18 e 28 Ciclos do ensino basico. MA dissertation, Universidade de Lisboa.Friedmann, N., 1998. BAFLAFriedmanns Battery for Agrammatism. Tel Aviv University, Tel Aviv.Friedmann, N., Belletti, A., Rizzi, L., 2009. Relativized relatives: types of intervention in the acquisition of A-bar dependencies. Lingua 119, 6788.Friedmann, N., Costa, J., 2010. The child heard a coordinated sentence and wondered: on childrens difficulty in understanding coordination and relative
clauses with crossing dependencies. Lingua 120 (6), 15021515.Friedmann, N.,Novogrodsky, R.,2004. Theacquisition of relative clausecomprehensionin Hebrew: a study of SLIand normal development.Journal of Child
Language 31, 661681.Friedmann, N., et al. in preparation. The acquisition of relative clause production in sixteen languages. COST-A33 project.Goodluck, H., Tavakolian, S., 1982. Competence and processing in childrens grammar of relative clauses. Cognition 11, 127.
J. Costa et al. / Lingua 121 (2011) 10831100 1099
-
7/25/2019 Joao Costa Assimetria
18/18
Guasti, M.T., 2002. Language Acquisition. The Growth of Grammar. MIT Press, Cambridge.Hakansson, G., Hansson, K., 2000. Comprehension and production of relative clauses: a comparison between Swedish impaired and unimpaired children.
Journal of Child Language 27, 313333.McKee, C., McDaniel, D., Snedeker, J., 1998. Relative children say. Journal of Psycholinguistic Research 27, 573596.Novogrodsky, R., Friedmann, N., 2006. The production of relative clauses in SLI: a window to the nature of the impairment. Advances in Speech-Language
Pathology 8, 364375.Roth, P.F., 1984. Accelerating language learning in young children. Journal of Child Language 11, 89107.Tavakolian, S.L., 1981. The conjoined-clause analysis of relative clauses. In: Tavakolian, S.L. (Ed.), Language Acquisition and Linguistic Theory. MIT Press,
Cambridge, MA, pp. 167187.Sheldon, A., 1974. The role of parallel function in the acquisition of relative clauses in English. Journal of Verbal Learning and Verbal Behavior 13, 272 281.
Valente, P., 2008. A produc
ao de frases relativas restritivas no final do 38 Ciclo e. do Secundario. MA dissertation, Universidade de Lisboa.Vasconcelos, Manuela, 1991. Compreensao e producao de frases com oracoes relativas: um estudo experimental com criancas dos tres anos e meio aos oito
anos e meio. Dissertacao de Mestrado, Universidade de Lisboa.
J. Costa et al. / Lingua 121 (2011) 108311001100