COMPREHENSION AND PRODUCTION OF RELATIVE CLAUSES IN BRAZILIAN-PORTUGUESE SPEAKING CHILDREN
by
Júlia Escalda Mendonça
A Master’s thesis submitted in partial fulfillment of the requirements for the degree of Master of Science
(Clinical Linguistics)
at the Joint European Erasmus Mundus Master’s Programme in Clinical Linguistics (EMCL)
UNIVERSITY OF POTSDAM February, 2011
Comprehension and production of relative clauses in Brazilian Portuguese speaking children
Júlia Escalda Mendonça
Under the supervision of Professor Dr. Flavia Adani, University of Potsdam and Professor Dr.
Gerard Bol, University of Groningen.
Abstract
The aim of the present study was to investigate the comprehension and production of subject and object relative clauses by Brazilian Portuguese-speaking children. Sixty-eight monolingual Brazilian Portuguese-speaking children aged 3;3 to 9;10 participated in the study.
An expressive vocabulary test was used as one of the inclusion criteria. Participants were presented with an agent selection comprehension test and an elicitation production test. Results showed that the abilities of production and comprehension do not overlap in typical development.
Children were able to understand relative clauses before they could consistently produce them.
There was an asymmetry in the comprehension of subject and object relative clauses. Right- branching object relatives were more demanding sentences for children to interpret when compared to right-branching subject relatives. Children consistently comprehend subject and object relatives by the age of three but are able to produce them from the age of five.
Comprehension error analysis showed that there is more than one cause of difficulty by young children in the comprehension of object relative clauses. Production error analyses provided evidence that children gradually learn the use of more complex constructions as they grow older.
Keywords: relative clauses, comprehension, production, language development.
Acknowledgments
I would like to express my great gratitude to my supervisor, Prof. Dr. Flavia Adani for her invaluable comments and feedback and also to my co-supervisor Prof. Dr. Gerard Bol.
I greatly appreciate Marisa Fonseca for the help with the data collection and Carla Fonseca for all the help and support provided at the school.
I am very grateful for the opportunity of studying in such a wonderful program as the EMCL and for that I thank all the professors at the course of the program as well as my fellow colleagues.
Many friends have contributed to my academic and personal development during this Master’s period and I would like to give many thanks to all of them, especially to Brigitta Keij, Vanessa Wolz and Alokika Fernandes.
Millions of thanks to my dear mother, Patricia, for all her love, patience, support and for her marvelous sense of humor in difficult times.
Table of contents
ABSTRACT ... i
ACKNOWLEDGMENTS ... ii
LIST OF TABLES ... vi
LIST OF FIGURES ... vii
1 Theoretical Background...1
1.1 Comprehension of relative clauses... 2
1.2 Production of relative clauses... 6
2 Methods ... 11
2.1 Participants ... 11
2.2 Materials and design ... 12
2.2.1 Vocabulary and Working Memory... 12
2.2.2 Test of Comprehension and Production of Relative Clauses... 13
2.3 Procedure ... 15
2.3.1 Test of Comprehension of Relative Clauses... 16
2.3.2 Test of Production of Relative Clauses... 17
3 Results ... 18
3.1 Test of Comprehension of Relative Clauses ... 19
3.1.1 Target responses... 19
3.1.2 Error Analysis ...20
3.2 Test of Production of Relative Clauses... 21
3.2.1 Target responses... 21
3.2.2 Error analysis ... 24
3.3 Comprehension versus Production... 28
3.4 Test of Non-Word Repetition ... 29
4 Discussion ... 31
4.1 Comprehension of relative clauses ... 31
4.2 Production of relative clauses ... 35
4.3. Comprehension versus Production ... 38
5 Conclusion ... 39
REFERENCES... 40
APPENDIX... 42
Appendix A. Response forms of the Test of Non-Word Repetition ... 43
Appendix B. Examples of pictures used in the experiment ... 44
Appendix C. Response form of the Test of Comprehension of Relative Clauses .. 45
Appendix D. Response forms of the Test of Production of Relative Clauses... 46
Appendix E. Individual scores for the Tests of Comprehension and Production of Relative Clauses and the Non-Word Repetition Test ... 47
Appendix F. Results of chance analysis from the experimental items of the
comprehension ...49 Appendix G. Results of chance analysis from the experimental and control items of the production test ... 50
List of tables
Table 2.1. Participant’s details...12 Table 2.2. Examples of each sentence type of the experiment...14 Table 3.1. Mean and percentage correct in the comprehension test by group and sentence-type
in each experimental condition, S.D. in parentheses...19 Table 3.2. Percentage of error-types in the experimental items of the comprehension test
by age group ... 21 Table 3.3. Mean and percentage correct in the production test by group and sentence-type in
each experimental and control items, S.D. in parentheses...22 Table 3.4. Examples of error-types in elicited responses to the production test in experimental
items... 25 Table 3.5. Percentage of error-types in the experimental items of the production test by
sentence-type... 26 Table 3.6. Percentage of error-types in the NR control items of the the production test by age
group...27
Table 3.7. Examples of error-types in elicited responses to the production test in N.R.
control items ...27 Table 3.8. Summary of the results of the comprehension and production tests of experimental
items...29 Table 3.9 Mean, range and percentage correct in the non-word repetition test by age group,
S.D. in parentheses...30
List of figures
Figure 2.1: Sample of an experimental test picture: “Show me the gorilla that is
chasing the lion.”...16 Figure 2.2: Sample of an experimental test picture: “Who is the arrow showing?” “Star with:
The gorilla…”...17
Chapter 1
Theoretical Background
Relative clauses are syntactic complex structures that include embedding and movement of a noun phrase from within the embedded clause. It is well documented that children have difficulties in producing and understanding constructions that involve wh-movements. Relative clauses are derived by movement either from a subject or an object position. In subject relatives there is movement of an element from a subject position and in object relatives there is movement from an object position. In both subject and object relatives the moved element leaves a trace in its original position. Depending on the position of the relative clause in relation to the main clause, subject and object relatives can be right-branching when attached to an object NP or center-embedded, when attached to a subject NP. The present study presents data from subject and object right-branching relatives, examples of these sentence types are given in (1) and (2).
(1) Subject relative clause (OS): O cachorro que está empurrando o gato.
The dogthat is pushing the cat.
(2) Object relative clause (OO): O cachorroque o gato está empurrando.
The dog that the cat is pushing.
The structure of a relative clause is composed by two features: the syntactic role of head of the relative clause and the syntactic role of the element that is relativized inside of the relative clause (the gap). The manipulation of these two features results in four types of relative clauses, namely (a) SS relatives (the main-clause subject is modified by a relative and the subject is relativized); (b) SO relatives (the main-clause subject is modified by a relative and the object is relativized); (c) OS relatives such as the example in (1) (the main-clause object is modified by a relative and subject is relativized) and (d) OO relatives such as the example in (2) (the main- clause object is modified by a relative and object is relativized). The following examples, adopted from Corrêa (1995), exemplify subject and object or center-embedded relative clauses:
(3) Subject relative clause (SS): O cavalo que pulou a cerca derruba a galinha.
The horse that jumped over the fence knocks down the chicken.
(4) Object relative clause (SO): O carneiro que o porco empurrou come o capim.
The sheep that the pig pushed eats some grass.
1.1 Comprehension of relative clauses
With regards to the comprehension of relative clauses, previous research have used mainly methods involving toy manipulation usually called “act-out tasks”, in order to assess children’s ability to comprehend this constructions. Examples of such experimental designs were reported by Sheldon (1974) and de Villiers et al. (1979). Their results demonstrated that the performance of children until five years of age was still at chance level in the comprehension of relative clauses. More recent studies, however, suggested that act-out tasks posed some
difficulties in the assessment of the comprehension of relative clauses by children due to methodological flaws in these early experimental sets (Hamburger & Crain, 1982; Corrêa, 1995).
Hamburger and Crain (1982) criticized act-out tasks by saying that it violated pragmatic and semantic aspects of language use. The authors tested the comprehension of OS relatives by three to five-year-old children using a picture preference task. They showed that performance improved when the number of animals in the pictures in the experimental set was reduced from three to two. Their results demonstrated that the five-year-olds have already mastered comprehension of the OS relative clauses. They also reported that when the experimental setting was pragmatically appropriate, four-year-old English-speaking children’s comprehension of relative clauses was above chance and some of the three-year-olds already understood subject relatives.
Corrêa (1995) tested an alternative act-out task to investigate the comprehension of relative clauses in English and Brazilian Portuguese-speaking children from three to six years of age. The stimuli used were manipulated for embeddedness, gap, and animacy. The results suggested that both English and Brazilian Portuguese-speaking children had mastered the comprehension of relative clauses by five years of age. The OS construction was found to be the easiest to comprehend, followed by the SS construction, the OO construction, and the SO construction, across all age-groups and across both languages. There was a main effect of animacy (structures with more [+animacy] NPs were more difficult for children to process than structures with less [+animacy] NPs) across ages. Object relatives were the most demanding sentences for children to comprehend regardless of embedding. The author refers to two acquisition periods: (1) from three to four years of age, in which an improvement in performance on the experiment was attributed to children’s capacity to keep the stimulus in immediate
memory and (2) from four and five years of age, in which the improvement in the performance was attributed to children’s ability to cope with the internal processing of the relative clause.
Using the same task proposed by Corrêa (1995), Kidd and Bavin (2002) also tested English-speaking children’s comprehension of relative constructions manipulated for embeddedness and gap. The authors tested English-speaking children from three to five years of age. Their results showed a main effect for embeddedness suggesting that center-embedded structures were more difficult than right-branching structures and a significant main effect for age. Contrary to the findings of Corrêa (1995) the main effect for gap was not significant.
Friedmann and Novogrodsky (2004) investigated the comprehension of relative clauses in three groups of Hebrew-speaking children, one group with Specific Language Impairment (SLI) and two groups of typically developing children. Three types of sentences were tested using a binary sentence-picture matching task: simple subject-verb-object (SVO) sentences, right-branching subject relatives, and right-branching object relatives. The results of the study demonstrated that typically developing children at the age of six, performed above chance on all three sentence types. In contrast, both the SLI group and the four-year-old group performed still at chance level in object relatives. The performance of the children without language impairment suggested that between the ages of four and five, the abilities that are required to understand relative clauses are not mastered yet. According to the authors, it is only by the age of six that children already seem to master the necessary syntactic analysis to correctly comprehend right- branching object relatives.
Arnon (2005) used a modificated version of Friedmann & Novrogodsky’s task, in which children were asked to choose a referent in the picture instead of a whole picture. She tested 14 Hebrew speakers (mean age of 4; 7) in this modified task and used seven of this children in a
further experiment including resumptive pronouns. The new experimental setting enabled to detect different error-types than previously reported in the literature. The author reported a new response pattern besides the Reversal error, the novel Agent error, that consists of choosing the agent of the relative clause instead of the clausal head. The author argues that the existence of two error-types suggests that there is more than one difficulty in the comprehension of relative claused, specially for object relatives, which seem to be associated not only with the thematic assignment to the clausal head (Reversal errors) but also with the modifying nature of the relative clause (Agent errors).
Adani (2010) tested three restrictive right-branching relative clauses types, OS, OO with the subject in pre and post verbal position. The author tested 116 Italian monolingual children between three and seven years of age. The main findings of the study were that subject relatives produced greater accuracy in comprehension than object relatives. The object relatives with pre verbal subjects produced greater accuracy rates than object relatives with the subject in post verbal position. Three-year-olds were also capable of understanding relative clauses, although not yet in a consistent way. The error analysis revealed that overall, children mostly responded with the interpretation where arguments were reversed especially for object relatives. The author proposes an explanation for her findings based on the intervention effect (Garrafa & Grillo, 2008;
Friedmann, Belletti & Rizzi, 2009) due a temporary computation overload of the immature language system that affected the comprehension object relative clauses at three years of age.
Taken together, the studies revised show that gap, embeddedness and animacy play a role in the comprehension of relative clauses. There is a well documented asymmetry between the comprehension of object and subject relatives, the later being easier to understand than the first.
Center-embedded relatives are more difficult to understand than right-branching structures and sentences with less animated nouns are easier to comprehend.
1.2 Production of relative clauses
With regards to the production of relative clauses, Diessel & Tomasello (2000) reported an investigation of children’s spontaneous productions in naturalistic settings. Their study examined the development of relative clauses in the speech of four English-speaking children between 1;9 and 5;2 years of age of the CHILDES database (MacWhinney & Snow, 1990). The authors reported that the earliest relative clauses used by English-speaking children are semantically simple, occur in presentational constructions that express a single proposition in two finite clauses and usually includes an intransitive verb. As children grow older, their utterances become more grammatically complex. The authors found examples of the earliest natural attempts of relative constructions from two years onwards. The results of this investigation showed that the most frequent relative constructions were right-branching. In addition, children produced and understood right-branching before center-embedded relative clauses. Finally, the authors suggest five factors that might contribute to the development of relative clauses in spontaneous child speech: (1) the ambient language, (2) the formulaic character of the main clause, (3) the information structure of the whole utterance, (4) the communicative function of presentational relatives, and (5) the limited processing capacity of young children.
Novogrodsky and Friedmann (2006) investigated the production of relative clauses in typically developing and SLI Hebrew-speaking children. Participants in the typically developing
group were 7;6 to 11;0 years-old and in the SLI group were aged 9;3 to 14;6 years-old. A preference task and a picture description task were used to elicit right-branching subject and object relative clauses. The children in typically developing group scored 98% correct on subject relatives and 94% correct on object relatives. The authors compared three age subgroups and demonstrated that age was not a factor in the production of both subject and object relatives which they took as evidence that the production of relative clauses was already mastered by seven-year-olds. Their findings also showed that none of the participants in either of the age groups avoided relatives. The SLI group showed a deficit in the production of subject and object relatives when compared to the performance of the typically developing children.
Zukowski (2009) elaborated an experimental task to elicit the production of both right- branching and centre-embedded relative clauses in a group of participants with Williams Syndrome (ages from 10 to 16), a group of typically developing children (ages from 4 to 7) and a control group of adults. Typically developing children performed above 80% correct in the elicited production of subject relatives and above 20% correct in the production of object relatives. Overall, all the three groups were highly successful the production of subject relatives (rates between 77% and 95%). In contrast, rates of production of object relatives were much lower (rates between 10% and 54%). Object relatives elicited contextually inappropriate responses in both the group with Williams Syndrome and in the typically developing group.
The two following studies used tasks to compare both comprehension and production of relative clauses in the same population, one in Swedish-speaking children and another in Italian- speaking children.
Hakansson & Hansson (2000) tested typically developing and SLI Swedish-speaking children in comprehension and production tasks. Comprehension of relative clauses was tested
by picture pointing, act-out and oral response and production of relatives was tested with elicited imitation and sentence completion. Typically developing Swedish-speaking children (aged 3;1 to 3;7) did not have a significant difference between the proportion of correct responses on the comprehension and on the production tests on the first testing. However, they had a significantly higher proportion of correct responses on the production tests than on the comprehension tests (average of 85% correct responses on the production tests and 64% on the comprehension tests), six months later. Their findings showed a positive correlation between the proportion of correct responses on comprehension and on production of relative clauses. The authors considered that the results suggest that the relationship between comprehension and production is different at different stages in development.
Contemori & Garrafa (2010) studied comprehension and production of relative clauses of preschool-aged Italian children, typically developing (aged 3;7 to3;10 and 4;5 to 5;5) and SLI children (four children aged 4;5 to 5;9). Comprehension of relative clauses was tested with a binary picture comprehension task adapted from Friedmann and Novogrodsky (2004).
Production of relative clauses was tested with two tasks: a picture description and a preference task, both adapted from Novogrodsky and Friedmann (2006). In the comprehension task, subject relatives were better understood than object relatives by children with SLI and by typically developing children of the older group (aged 4;5 to 5;9). As for the elicited production task, unimpaired children produced significantly more subject relative targets than object relative targets. The performance of children with SLI was poorer than that of the control groups with both sentence-types. The findings of the study showed that children’s abilities in comprehension and production of relative clauses do not overlap in typically developing and children with SLI.
Typically developing children adopt many different strategies to avoid the more complex relative clauses, especially object relatives.
Data from the studies revised give controversial results about the way children produce relative clauses. In naturalistic settings, production of relative clauses was reported from two years onwards. In two studies, object relatives were more likely to elicit incorrect (i.e. non- relatives) target responses than subject relatives. Furthermore, the relation of comprehension and production is still unclear, since it has been reported that comprehension emerged after production and also the opposite pattern in different studies.
Although the comprehension and production of relative clauses has been documented in typically developing populations, it is hard to find studies in the literature in which both comprehension and production are tested in the same population and in a relatively large sample of children. Some questions remain unclear regarding the development of these two modalities in children’s language. Furthermore, there are very few studies with Brazilian Portuguese-speaking children. The aim of the present study is to investigate the comprehension and production of relative clauses, both subject and object relatives, by typically developing Brazilian-Portuguese speaking children.
The questions to be addressed in the present study are:
How does comprehension and production of relative clauses emerge in Brazilian Portuguese-speaking children?
Since Brazilian-Portuguese is a basically an SVO language, will there be a difference between sentence-types, i.e. non-canonical word order sentences (object relatives) and canonical word order sentences (subject relatives) in this population?
In typical development, do these two modalities (comprehension and production) overlap?
What types of errors do children make while understanding and producing relative clauses?
Based on the literature, our expectations are that subject relatives will be easier to understand than object relatives. Given the controversial data on production of relative clauses it is not possible to make ascertain predictions about the outcome of this task in the present study.
With this study, we expect to shed some light on the theory about production and comprehension of syntactically complex sentences by young children.
Chapter 2 Methods
2.1 Participants
Sixty-eight monolingual Brazilian Portuguese-speaking children aged 3;3 to 9;10 participated in this study. They were recruited in a school in Contagem, a city in the surrounding area of Belo Horizonte, Minas Gerais, Brazil. The children were divided into seven aged based groups as shown in Table 2.1. The inclusion criteria used to select the children was that they presented normal vocabulary in an expressive vocabulary test (ABFW Vocabulary Test) and did not presented evidences of language or cognitive problems. Six children were excluded from the sample because they refused to participate in the activities or did not respond and one child was language impaired.
Table 2.1. Participant’s details
Age Group n Mean Age Age Range
3-year-olds 12 3;7 3;3 – 3;11
4-year-olds 11 4;6 4;1 – 4;10
5-year-olds 11 5;4 5;1 – 5;9
6-year-olds 13 6;6 6;0 – 6;11
7-year-olds 13 7;6 7;0 - 7;11
8-year-olds 4 8;6 8;1 – 8;10
9-year-olds 4 9;6 9;1 – 9;10
2.2 Materials and design
2.2.1 Vocabulary and Working Memory
The ABFW Vocabulary Test (Befi-Lopes et. al, 2000) was used to assess the expressive vocabulary of the children. The ABFW Vocabulary Test consists of 118 pictures divided by semantic category which are presented in sequential order as follows: clothing, animals, food, transportation, furniture, professions, locations, shapes and colors, toys and musical instruments.
The pictures are presented one by one to the child who is asked to name them. The test enables the analysis of the processes used by children to achieve the correct name of the words (usual designations for words, the non-designations for words, and substitution processes). For each of the analysis there is a reference for normality for Brazilian-Portuguese-speaking children.
The Test of Non-Word Repetition (Appendix A) was proposed by Kessler (1997) and consists of 30 non-words formed by a simple syllabic structure, following the phonological structure of Brazilian Portuguese (consonant-vowel). The words are arranged in the six lists with
five non-words based on the number of syllables, ranging from one to six syllables. The participants were invited to speak a “crazy” language. They were asked to repeat the non-word they were presented immediately after the experimenter. The response was considered correct if the child was able to repeat the item identically as presented, and incorrect if there was any omissions, substitutions or if the child did not respond.
2.2.2 Test of Comprehension and Production of Relative Clauses
The materials and procedure used on the present study were adapted from the study by Özge, Marinis and Zeyrek (2009) which was based on the experimental set by Adani (2010) (Appendices B, C and D). The tests consisted of a total of 51 sentences: 24 experimental and 27 control items. One factor was manipulated, the relative clause type in which half of the experimental items were subject relatives (SR) and half were object relatives (OR).
Experimental items consisted of semantically reversible relative clauses and control items consisted of semantically non-reversible subject and object relative clauses with animate agents and inanimate objects and subject relative clauses with intransitive verbs1. Examples each sentence type used in the two conditions of the experiment are shown in Table 2.2.
1 The following abbreviations will be used: NR to refer to semantically non-reversible control items and INT to control itemswith subject relative clauses with intransitive verbs.
Table 2.2. Examples of each sentence type of the experiment
SR OR
Experimental Items (n= 24)
Aponte o elefante que está perseguindo o urso.
Show me the elephant that is chasing the bear.
(n =12)
Aponte o carneiro que o macaco está lambendo.
Show me the lamb that the monkey is licking.
(n=12)
Control Items (n= 27)
Aponte o menino que está comendo a maçã.
Show me the boy that is eating the apple.
(NR control items; n =8)
Aponte o sorvete que o menino está segurando.
Show me the ice-cream that the boy is holding.
(NR control items; n =8) Aponte o pássaro que está dormindo.
Show me the bird that is flying.
(INT control items; n = 11)
Children were assessed in a comprehension and in a production test. The comprehension test consisted of an agent selection task, in which the participants were required to choose which one of the referents was correct for the sentence they heard. The production test consisted of an elicitation task in which the participants were required to respond verbally to the question of which one of the referents a red arrow was pointing to, creating a sentence based on the two initial words given by the experimenter.
The experimental items consisted of 14 animals (bear, elephant, cat, dog, chicken, duck, horse, sheep, monkey, lamb, gorilla, lion, mouse and rabbit) and 6 action verbs (to bite, to kick, to lick, to push, to kiss and to chase). Pictures of the NR control items consisted of one animate noun (boy, girl, man or woman) and one inanimate object (apple, ice-cream, book, newspaper, flower, fish, cake or loaf) and 4 verbs (eat, hold, smell, read, cut). The lexical items in the INT control sentences were 6 animals (fish, turtle, bee, bird, donkey and goat) and 6 action verbs (to
cry, to laugh, to run, to jump, to fly and to sleep). The sentences were recorded by a native Brazilian-Portuguese speaking female in a soundproof room and were paired with the pictures and randomized.
Each pair of animals was used with two different actions in order to avoid using the same animals and action verbs in more than one sentence. For example, in one picture a lion was biting a gorilla and in another picture a gorilla was chasing a lion. The pictures were controlled for size in order to prevent a bias for bigger pictures to be identified as correct responses.
Moreover, when a picture was used in the comprehension test as a subject relative target sentence, the same picture was used in the production test as an object relative target. The number of correct answers was also controlled and equally distributed on the three possible positions in the pictures (center, right or left) in the comprehension task as well as in the production task. To control for potential order effects on trials, two lists were generated in which the presentational order of trials was reversed, half of the participants were assessed with list A and half with list B.
2.3 Procedure
Participants were tested in a quiet room by two experimenters during school hours. The tests were administrated in one or two sessions depending on the collaboration of the child. The order of the testing was 1)Vocabulary Test; 2) Test of Production of Relative Clauses; 3) Test of Non-Word Repetition and 4) Test of Comprehension of Relative Clauses.
2.3.1 Test of Comprehension of Relative Clauses
Participants were explained they were to participate on an activity in which they would be shown some pictures and hear some sentences on the computer. Then they would have to point to the correct animal/person/object referred to in the sentence. Before the test, all the animals, persons, objects and verbs were shown in boards and elicited by the children and/or the experimenter to guarantee that the participants were familiar with all the lexical items in the test.
Two practice items, one with a subject and one with an object relative were presented before the trials.
Participants were presented with a sentence accompanied by a picture with three animals involved in an action and asked to point to the animal that corresponded to the one heard in the sentence. Figure 2.1 shows one example of the pictures used in the test.
Figure 2.1: Sample of an experimental test picture: “Show me the gorilla that is chasing the lion.”
Possible responses were choosing the correct animal by assigning correct theta roles, choosing the animal with the reversed theta role (role-reversal), or choosing the wrong animal.
2.3.2 Test of Production of Relative Clauses
Participants were presented first with a picture accompanied by two declarative sentences describing the actions in the picture. For example, participants were presented with Picture 2.1 and then heard “The gorilla is chasing the lion”, “The lion is chasing the gorilla”.
Then a red arrow appeared indicating one of the two animals (e.g. the gorilla) as showed in Figure 2.2. The participants were asked by the experimenter “Who is the arrow showing?” and the beginning of the target answer was also given “Star with: The gorilla…” Participants were required to respond verbally to the question creating a sentence based on the two initial words given by the experimenter. Target responses were either subject relatives or object relatives. Two practice items, one with a subject and one with an object relative were presented before the trials.
Figure 2.2: Sample of an experimental test picture: “Who is the arrow showing?” “Star with:
The gorilla…”
Chapter 3 Results
In this chapter are presented the results regarding the responses obtained from the Test of Comprehension and Production of Relative Clauses as well as the results from the Test of Non- Word Repetition described in Chapter. Groups results are presented in this chapter, individual scores can be found in Appendix E. The results are presented by test: Test of Comprehension of Relative Clauses, Test of Production of Relative Clauses and Test of Non-Word Repetition, respectively. Because group sizes and number of items across conditions differ, results are also presented in percentages to make comparisons between results possible.
To investigate effects across age groups, one-way ANOVA’s with post-hoc procedures were conducted. Paired samples t-tests were used for comparisons within groups between sentence-types, and single sample t-tests for comparisons to chance. As mentioned by Arnon (2005), in this kind of experimental setting, chance level is 33% since children consider all the three referents in the picture set . To compare results from the Test of Non-Word Repetition and the Tests of Comprehension and Production of Relative Clauses a correlation analysis was performed.
3.1 Test of Comprehension of Relative Clauses
3.1.1 Target responses
All participants performed at ceiling level in all the control items of the comprehension test. As seen on Table 3.1 the performance in comprehension of object relatives was lower than in subject relatives until children were five-years-old. From six years of age, the performance in both subject and object relatives is virtually the same and children have an accuracy rate above 90% correct in the experimental items.
Table 3.1. Mean and percentage correct in the comprehension test by group and sentence-type in each experimental condition, S.D. in parentheses
Age Group Subject Relatives Object Relatives
Mean (SD) % Mean (SD) %
3-year-olds 9.9 (1.4) 82.5 7.1 (2.6) 59.0
4-year-olds 10.5 (1.7) 87.5 7.3 (2.2) 61.3
5-year-olds 11.2 (1.6) 91.7 8.9 (3.2) 74.2
6-year-olds 12.0 (0) 100 11 (1.6) 91.7
7-year-olds 11.8 (0.5) 98.3 11.1 (1.9) 92.9
8-year-olds 12 .0 (0) 100 12 .0 (0) 100
9-year-olds 12 .0 (0) 100 12 .0 (0) 100
Single-sample t-tests were used for comparison to chance in the group level analysis. All groups performed above chance level in the comprehension of both subject and object relatives.
More detailed analysis can be found in Appendix F.
To investigate the comprehension of relative clauses across age groups, a one-way ANOVA was conducted. The results showed that there was a significant effect of group in the comprehension of both subject (F (6, 61) = 5.43, p < .001) and object relatives (F (6, 61) = 7.83, p < .001). Post-hoc procedures (Tukey’s and Hochberg’s) showed that age groups from three to five differ significantly in the comprehension of subject and object relatives when compared to the other age groups. Age groups from six to nine performed equally in the comprehension of subject and object relative clauses.
Paired samples t-tests were used for comparison between sentence-types within groups.
The results showed a significant difference in the comprehension of object and subject relatives.
Participants demonstrated significant greater difficulty in the comprehension of object relatives than to subject relatives from three to five years of age (3-year-olds: t(12) = 3.74, p<0.01; 4- year-olds: t(11)= 4.11, p<0.01; 5-year-olds: t (11)= 2.56, p<0.05). Age groups from six to nine had a similar performance in the comprehension of object and subject relatives.
3.1.2 Error Analysis
In the Test of Comprehension of Relative Clauses three responses were possible: correct answer, when the child had a target response by choosing the correct animal, role-reversal (SR reverse, OR reverse), when the child chose the animal with the reversed thematic-role or an incorrect response (SR Incorrect, OR Incorrect), when there was a choice for the wrong animal.
An overview of the error-types from the comprehension test is presented in Table 3.2.
Percentages of error-types are given based on the total of trials.
Table 3.2. Percentage of error-types in the experimental items of the comprehension test by age group.
Age Group Subject Relatives Object Relatives
Reverse Incorrect Reverse Incorrect
3 y.o. 5.6 11.9 29.9 11.1
4 y.o. 8.3 4.1 28.5 10.5
5 y.o 4.2 4.2 18.8 6.9
6 y.o. . . 8.3 .
7 y.o. . 1.4 4.3 2.7
In summary, the analysis of non-target responses revealed that when children did not get the target answer, they showed a non-specific response pattern for subject relatives, sometimes choosing more frequently the animal with the reversed thematic-role and sometimes choosing the incorrect animal. In contrast, in the comprehension of object relatives there was a more distinct pattern in which children tend to choose more frequently the type of response in which arguments are reversed. The occurrence of this error-type is reduced by the increment of age until children are seven-years-old when they stop making errors in the comprehension task.
3.2 Test of Production of Relative Clauses
3.2.1 Target responses
The results indicated that the production of relative clauses is a very difficult task to children in a large age range in both sentence-types. Three- and four year-old participants were not able to produce relatives in any of the test conditions (experimental and control sentences).
Table 3.3 shows the performance of five- to nine-year-olds in the production test for experimental and control items.
Table 3.3. Mean and percentage correct in the production test by group and sentence-type in each experimental and control items, S.D. in parentheses
Experimental Items Control Items
Non-reversible Intransitive
Age Group SR OR SR OR
Mean (SD) % Mean (SD) % Mean (SD) % Mean (SD) % Mean (SD) % 5-year-olds 4.5 (5.3) 37.8 1.3 (2.2) 15.9 3.5 (3.9) 43.2 2.4 (2.8) 29.6 4.0 (5.5) 36.4 6-year-olds 5.0 (4.6) 41.6 2.4 (2.8) 29.8 3.3 (2.9) 41.3 4.7 (2.7) 59.6 3.0 (4.2) 27.2 7-year-olds 4.4 (4.7) 37.2 5.0 (3.0) 62.5 3.3 (3.3) 41.3 6.3 (2.0) 78.8 3.7 (4.0) 34.3 8-year-olds 8.3 (5.7) 68.7 2.3 (3.3) 28.1 4.7 (3.6) 59.4 2.7 (3.7) 34.4 7.7(5.2) 70.4 9-year-olds 11.7 (0.5) 97.9 5.2 (3.4) 65.6 8.0 (0) 100 5.5 (3.7) 68.7 11 (0) 100
One-way ANOVA showed that there was a significant effect of group in the production of both subject (F (6, 61) = 7.66, p < 0.01) and object relatives (F (6, 61) = 8.32, p < 0.01) in the experimental items. Similarly there was also a significant effect of group in the production of subject (F (6, 61) = 7.39, p <0.01) and object (F (6, 61) = 13.78, p <0.01) relatives in the non- reversible (NR) control items as well as in the control intransitive (F (6, 61) = 7.24, p <0.01) items.
Regarding the control items, group analyses showed that children from five to eight years of age performed at chance level in the production of NR subject relatives and INT relatives. In the NR OR condition, six and seven-year-olds showed an above chance performance, in contrast to eight and nine-year-olds that performed at chance level in the production of NR object relatives. Nine-year-olds had an above chance performance in the production of NR subject relatives and INT relatives.
Concerning the experimental items, children from five to eight years of age performed at chance level in the production of both sentence-types, with two exceptions. In the group analysis, seven-year-olds showed an above chance performance in the production of object relatives and also the nine-year-olds showed an above chance performance in the production of subject relatives. Results of comparisons to chance of experimental and control items can be found in Appendix G.
Taking in consideration the results from chance analysis in the production of the experimental and the control items it is plausible to say that, as a group, children from six years of age are capable of consistently produce relative clauses. In addition, individual results (Appendix E) show that children from five years of age already show an above chance performance in the production of subject and object relatives in all test conditions. However, it is important to point out that the performance of three and four-year-olds was surprisingly low, since they did not produce any relative clauses. This could possibly be due to a methodological shortcoming in the production test that could have presented a task that was too cognitively costly for children this young.
Comparisons between test conditions (experimental versus control NR items) within groups with paired samples t-test showed that there was no significant differences between the production of relative clauses in these two conditions in most groups. The six-year-old group was the only to perform better in one of the control conditions than in the experimental condition.
In this group, participants performed better in the production of NR object relatives (M= 4.77, SD= 2.77) than in the production of reversible subject relatives (M=2.38, SD= 2.78, t(13)= -2.4, p< .05).
Paired-samples t-tests were conducted to test whether there were differences in performance in the production of subject and object relatives within participants in each of the groups. The results showed that there was no difference between sentence-types within participants in any of the experimental items and in most groups in the control NR items. Only the seven-year-old group had a better performance in the production of object relatives (M= 6.31, SD = 3.32) than in the subject relatives (M = 6.31, SD= 2.09, t(13) = - 3.19, p< .05) in the NR control condition.
3.2.2 Error analysis
In the Test of Production of Relative Clauses a diverse amount of response types were observed. The responses were divided into categories determined after the analysis of the tests response forms. Experimental items with subject relatives targets elicited mainly the complementizer omission (C.O) error-type in which the particle “that” was not produced in the response sentence. Regarding the object relatives targets, they elicited the majority of incorrect responses and four main error-types were identified: sentences in direct subject-verb-object order (S.V.O) with the correct referent or the reversed referent (S.V.O (r)), sentences in the passive voice, sentences with the copula DO and other error-types that were computed together.
Examples of each of the error-types are presented in Table 3.4.
Table 3.4. Examples of error-types in elicited responses to the production test in experimental items
Experimental Items
Error-type Target answer Example of elicited responses
SR
Complementizer Omission (C.O.)
O elefante que está perseguindo o urso.
The elephant that is chasing the bear.
O elefante (está) perseguindo o urso.
The elephant (is) chasing the bear.
Other
O pato que está beijando a galinha.
The duck that is kissing the chicken.
O pato e a galinha.
The duck and the chicken.
OR
S.V.O.
O carneiro que o macaco está lambendo.
The lamb that the monkey is licking.
O carneiro (está) lambendo o macaco.
The lamb (is) licking the monkey.
S.V.O. (r)
O carneiro que o macaco está lambendo.
The lamb that the monkey is licking.
O macaco(está) lambendo o carneiro.
The monkey (is) licking the lamb.
Passive Voice
O leão que o gorila está mordendo.
The lion that the gorilla is biting.
O leão (está) sendo mordido pelo gorila.
The lion (is) being bitten by the gorilla.
DO
O gato que o cachorro está lambendo.
The cat that the dog is licking
O gato não está fazendo nada.
The cat is not doing anything.
Other
O cavalo que a ovelha está chutando.
The horse that the sheep is kicking.
Pro cavalo.
To the horse.
Error analysis in the experimental condition of the production test is shown in Table 3.5.
Percentages of error-types are given based on the total of trials.
Table 3.5. Percentage of error-types in the experimental items of the production test by sentence-type.
Age Group SR OR
C.O. Other S.V.O. S.V.O (r) Passive Voice + DO Other
3-year-olds 93.7 6.3 34.4 26 . 17.7 21.9
4-year-olds 93.2 6.8 18.3 10.3 . 40.9 30.5
5-year-olds 59.9 2.3 29.5 13.6 9.1 27.3 4.6
6-year-olds 57.1 1.3 16.2 . 7.5 43.2 3.3
7-year-olds 62.2 0.6 15.4 . 14.4 7.7 .
8-year-olds 31.2 . . . 71.9 . .
9-year-olds 2.1 . . . 34.4 . .
With regards to the control items, the only possible error-type in the target sentences in the intransitive condition was the complementizer omission. When the target was a subject or object relative in the N.R. control condition, subject and object relative targets elicited the same error-types as the ones observed in the experimental condition. Sentence-types in the N.R.
condition were not susceptible to the S.V.O. (r) or to the +DO type of response. Error analysis in the control conditions of the production test is shown in Table 3.6. Percentages of error-types are given based on the total of trials. Examples of each of the error-types are presented in Table 3.6.
Table 3.6. Percentage of error-types in the NR control items of the the production test by age group
Age Group Control Items
SR OR
C.O Other S.V.O. Passive Voice Other
3-year-olds
91.6 8.4 71.9 . 28.1
4-year-olds
89.7 10.3 69.3 . 30.7
5-year-olds
53.4 3.4 46.6 9 14.8
6-year-olds
55.8 2.9 21.2 6.7 12.5
7-year-olds
58.7 . 6.7 9.7 4.8
8-year-olds
37.5 3.1 . 65.6 .
9-year-olds
. . . 31.3 .
Table 3.7. Examples of error-types in elicited responses to the production test in N.R. control items
Control Items
Error-type Target answer Example of elicited responses
SR
Complementizer Omission (C.O.)
O menino que está segurando o sorvete.
The boy that is holding the ice-cream.
O menino (está) segurando o sorvete.
The boy (is) holding the ice-cream.
Other
O homem que está lendo o livro.
The man that is reading the book.
Lendo o livro Reading the book.
OR
S.V.O.
A maçã que o menino está segurando.
The apple that the boy is holding.
O menino (está) segurando a maçã.
The boy (is) holding the apple.
Passive voice
A flor que a mulher está cheirando.
The flower that the woman is smelling.
A flor(está) sendo cheirada pela mulher.
The flower (is) being smelled by the woman.
Other
O pão que a mulher está cortando.
The woman that is cutting the bread.
O pão (está) cortando.
The bread is cutting,
3.3 Comprehension versus Production
Paired samples t-tests were used for comparison between the performance in the production and in the comprehension tests of the experimental items within groups. The results showed that for subject relatives, children’s performance in comprehension and production tests significantly differ until they are seven-years-old. Participants demonstrated significant greater difficulty in the production of object relatives than in the comprehension of subject relatives from three to seven years of age (3-year-olds t(12) = -23.80 , p<0.01; 4-year-olds:, t(11)= - 20.63 , p<0.01; 5-year-olds: t (11)= -4.60, p<0.01; 6-year-olds: t (13)= -5.38, p < 0.01; 7-year- olds: t (13)= -5.38, p < 0.01). At the ages of eight and nine, children had similar performance in these two tasks.
The results of the analysis for object relatives also showed a difference between children’s performance in their production until they are nine-years-old. Participants demonstrated significant greater difficulty in the production of object relatives than in the comprehension subject relatives from three to nine years of age (3-year-olds: t(12) = -9.27, p <
0.01; 4-year-olds: t(11) = -10.86, p< 0.01; 5-year-olds: t(11) = -7.06, p<0.01; 6-year-olds: t(13) = -9.94, p<0.01; 7-year-olds: t(13) = -5.67, p<0.01; 8-year-olds: t(4) = -5.90, p<0.05; 9-year-olds:
t(4) = -3.96, p<0.05).
The results from the experimental results presented in this chapter can be summarized as the following:
Table 3.8. Summary of the results of the comprehension and production tests of experimental items
Age Comprehension Production Comp X Prod
SR OR SR OR SR OR
3-year-olds > chance > chance SR > OR < chance < chance SR = OR C > P C > P 4-year-olds > chance > chance SR > OR < chance < chance SR = OR C > P C > P 5-year-olds > chance > chance SR > OR = chance = chance SR = OR C > P C > P 6-year-olds > chance > chance SR = OR = chance = chance SR = OR C > P C > P 7-year-olds > chance > chance SR = OR = chance > chance SR = OR C > P C > P 8-year-olds > chance > chance SR = OR = chance = chance SR = OR C = P C > P 9-year-olds > chance > chance SR = OR > chance = chance SR = OR C = P C > P
Note: The greater than sign (>) implies “easier to comprehend/produce than”, the lower than sign implies “easier to comprehend/produce than” (<). The equal sign (=) implies that tasks are of equal difficulty to comprehend/produce.
The same holds for comparisons to chance. C holds for comprehension, P holds for Production.
3.4 Test of Non-Word Repetition
The test of non-word repetition was composed by 30 non-words arranged in the six lists with non-words that could have from one to six syllables. Table 3.9 shows the performance of three to nine-year-olds in the non-word repetition test.
Table 3.9. Mean, range and percentage correct in the non-word repetition test by age group, S.D. in parentheses
Age Group NWR
Mean (SD) Range Mean (SD) (%) Range (%)
3-year-olds 21.6 (5.8) 7 to 30 72.2 (19.3) 23.3 to 100
4-year-olds 23.4 (4.8) 17 to 30 77.9 (16.0) 56.6 to 100
5-year-olds 21.5 (9.8) 3 to 30 71.8 (32.7) 10.0 to 100
6-year-olds 23.3 (4.8) 15 to 30 77.7 (16.1) 50.0 to 100
7-year-olds 28.0 (2.6) 20 to 30 93.3 (8.9) 66.7 to 100
8-year-olds 27.7 (4.5) 21 to 30 92.5 (15.0) 70.0 to 100
9-year-olds 27.0 (4.8) 20 to 30 90.0 (15.9) 66.7 to 100
A significant correlation was found between the performance in the non-word repetition test and the comprehension of subject relatives (Pearson= .039) and the comprehension of object relatives (Pearson =.013) in the three-year-old group. There was also a significant correlation between the non-word repetition test and the production of subject relatives (Pearson = .032) and INT relatives (Pearson= .016) by eight-year-olds.
Chapter 4 Discussion
The aim of the present study was to compare the performance in the comprehension and production of different sentence types with Brazilian Portuguese-speaking children in different age groups (three- to nine-years-old). The results of the study demonstrated that there was a significant effect of age across conditions, showing a developmental pattern for comprehension and production of relative clauses. Comprehension and production of object relatives was found to be more difficult than of subject relatives. Across age groups there was a discrepancy between comprehension and production of both sentence types in which production of relative clauses was found to be more difficult than comprehension.
4.1 Comprehension of relative clauses
Regarding the comprehension of relative clauses, previous research has shown that subject relatives are easier to understand than object relatives in different languages. Results from the test of comprehension of relative clauses confirmed that subject relatives produce greater accuracy than object relatives. These findings are in line with most of the research in the
acquisition of relative clauses and provide more evidence for the asymmetry between subjects and object relatives (Corrêa,1995; Kidd and Bavin, 2002; Friedmann and Novogrodsky, 2004;
Adani, 2010). The results of the present study suggest that children master the comprehension of subject and object relatives at three years of age.
Studies in different languages have shown a similar pattern for the comprehension of relative clauses in early development. In English, Hamburger & Crain (1982) reported that three- year-olds were able to understand right-branching subject relatives. Three-year-old Swedish- speaking children performed at chance in the comprehension of subject relatives (Hakansson &
Hansson, 2000). Friedmann & Novogrodsky (2004) demonstrated that at the age of four, Hebrew-speaking children performed above chance in the comprehension of subject relatives and around six years of age, they master the construction of object relatives. In addition, in Italian children performed at ceiling in the comprehension of subject relatives from three years of age (Adani, 2010) and had an above chance performance from four-years-old in the comprehension of object relatives (Adani, 2010; Contemori & Garrafa, 2010) . These studies also indicate that children understand subject relatives before object relatives.
Specifically with Brazilian Portuguese-speaking children the results here presented suggest that there was a developmental effect in subject and object relatives seen by an increase of accuracy with age. Brazilian Portuguese-speaking children had mastered the comprehension of subject and object relatives at three years of age. In the present study, the comprehension of subject relatives was achieved in an earlier age than previously reported in the literature (Corrêa, 1995) in which the comprehension of relative clauses by Brazilian Portuguese-speaking children was only fully achieved at five years of age.
In the study by Corrêa (1995) animacy was one of the factors manipulated in the experimental design. Her results showed a better performance in sentences with less animated NPs. In the current study, animacy was not measured directly but the sentences used in the control conditions had an inferior number of animated NPs than the sentences in the experimental condition. In this case, it is plausible to say that there was an indirect measure of animacy. As the results show, experimental sentences were harder for children to understand than control sentences, in which all the participants performed at ceiling level. In this sense, it would be possible to infer that for the experimental sentences in the current study, there was also an effect of animacy that could have contributed to children’s difficulty in comprehension of relative clauses in the experimental sentences.
Error analysis has shown that there was not a consistent error pattern of response when children did not get the target answer for subject relatives. On the other hand, for object relatives, the reverse type of error was more common than the incorrect type. Novogrodsky & Friedmann (2006) referred in their study that in order to process an input sentence that involves wh- movement, it is necessary to reactivate the head at the trace position so it can receive its thematic role (of agent or theme of the verb). In this manner, to correctly comprehend an object relative sentence it would be necessary to construct the relation between the moved element and its original position properly. From the results of the error analysis in the present study, it would be possible to consider that the establishment of such syntactic relation poses a difficulty in the comprehension of relative clauses, more specifically object relatives, by children until the age of five.
Importantly, although the reversal error was the most common, another type of answer was also observed, namely the incorrect error-type. Arnon (2005) also reported this error-type
the “Agent error”, in which children choose the agent of the relative clause instead of the clausal head. According to the author, while the reversal error is associated to a difficulty in understanding the thematic-role assignment, the agent error reveals that children are not able to understand that the relative clause adds information about the clausal head. In this manner, the results in the present study, in line with the findings by Arnon (2005), also suggest that there is more than one cause of difficulty by young children in the comprehension of object relative clauses.
It is possible to consider that three- and four-year-olds show a more prominent difficulty in the comprehension of object relatives than in subject relatives due to an intervention effect (Garrafa & Grillo, 2008; Friedmann, Belletti & Rizzi, 2009). The intervention effect refers to the difficulty to establish the correct syntactic dependency needed to understand chains in which there is an intervener between its elements. Children’s lower performance in object relatives could be explained by a temporary computation overload. As suggested by Adani (2010) and Contemori & Garrafa (2010) because of their immature cognitive system there is a temporary computation overload. It is hard for children to keep in memory arguments involved in the thematic assignment since the feature distinctions are not yet available. Interestingly, the results from the working memory test demonstrated a positive correlation between test results in the non-word repetition test and in the comprehension of subject and object relatives in the experimental items in the three-year-old group. This might suggest a possible relation between working memory and the comprehension of relative clauses.
