Pronoun Comprehension in Tagalog-Speaking Children with Autism by
Patricia B. Amillos S3431622
A Master’s thesis
Submitted to the University of Groningen In Partial Fulfilment of the Requirements
for the Degree of European Masters in Clinical Linguistics August 2019
Supervisors:
Dr. Gerard Bol Dr. Nick Riches
University of Groningen Newcastle University
Abstract
Autism Spectrum Disorder is widely known for its characteristic deficits in social communication and social interaction, yet recent studies show that grammatical impairments are also present. Pronoun comprehension requires both functional syntactic and pragmatic knowledge, making difficulties likely in autism. In typical development, most investigations are on the binding principles which adhere to the generativist framework. There does exist an equivalent, more functional explanation to pronoun comprehension that only relies on rules of discourse to explain pronoun comprehension. Thus, this study investigated the presence of pronoun comprehension difficulties in autism and tried to differentiate whether a syntactic or pragmatic framework best explains their performance. A total of 22 Tagalog-speaking children, 11 with autism and 11 controls, participated in a two-alternative forced choice task and picture verification task. An online survey with 18 young adults was also performed to represent data at the end of linguistic development. The outcomes of this study show that children with autism exhibit decreased performance on sentences that involve pronouns. However, the question regarding a syntactic or pragmatic-based deficit remains open for further investigation. What is known from present findings is that the ASD group shows a different pattern from the TD group in deciding pronoun coreference. The role of similar degrees of prominence between pronoun and antecedent is discussed. Finally, evidence was also found that sensitivity to cues affecting pronoun coreference continue to develop in adulthood.
Acknowledgements
I would like to express my sincerest gratitude to my supervisors who guided me throughout this project. First, I thank Dr. Nick Riches who generously filled gaps in my knowledge from the most basic of linguistic concepts to more complicated matters of interpreting my data. I am also very grateful to Dr. Gerard Bol for kindly agreeing to supervise so that I could fulfill the last requirement in my EMCL+ journey. I would also like to thank Dr. Wim Tops for his consistently constructive feedback that has improved my research abilities. The entirety of the EMCL+ professors and coordinators are also worth mentioning here. Thank you for investing time and resources into a fulfilling educational experience.
My deepest thanks also go to the institutions, families, and children who agreed to participate in this study. This includes students in Immaculate Conception Cathedral School, Core Skills Therapy Center, Haven of Learning Therapy Center, and University of the Philippines Integrated School. Meeting you all reminded me of the value of the path I chose to take.
My friends and family deserve to celebrate this milestone as much as I do. To friends old and new alike, thank you for the loyalty you’ve shown. Thank you to my father and mother who have instilled the value of perseverance yet had unconditionally supported me through the times I failed. I am most grateful to my sister Suzanne who helped me discover my purpose. Lastly, I am grateful to Justin Toribio who put his trust in me so that I could learn to trust myself.
Table of Contents Abstract………..……… Acknowledgements………..……….. List of Tables……….……… List of Figures……….………... 1 Introduction………
1.1 Language Impairments in Autism Spectrum Disorders
1.2 The Binding principles………. 1.2.1 Principle A ……….…. 1.2.2 Principle B ……….. 1.2.3 Principle C ………... 1.3 The Binding Principles in Autism ……….. 1.4 Pragmatic Framework of Pronoun Comprehension ……… 1.5 Comparing the Syntactic and Pragmatic Accounts ………. 1.6 The Tagalog Language ...……… 1.7 Research Questions ……….
2 Method……...………
2.1 Participants ……….. 2.2 Materials ………...
2.2.1 Sentence Repetition Task ……… 2.2.2 Ravens Coloured Progressive Matrices ……….. 2.2.3 Two-alternative forced choice task ………. 2.2.4 Picture verification task………... 2.2.5 Online Survey ………. 2.3 Procedure ……… 2.3.1 Sentence Repetition Task ……… 2.3.2 Ravens Coloured Progressive Matrices ……….. 2.3.3 Two-alternative forced choice task……….………. 2.3.4 Picture verification task ……….……. 2.3.5 Online Survey ……….………… 2 3 6 7 8 9 12 13 14 18 19 22 27 31 33 35 35 35 35 36 36 42 42 43 43 43 43 45 46
2.4 Data Analysis ……….…….
3 Results ……….………..
3.1 Linguistic Abilities and Non-Verbal Intelligence ………... 3.2 Comprehension of Principle A, B, C, and Filler Sentences ……… 3.3 Preference for Syntactic vs. Pragmatic Interpretations ………...……… 3.4 Acceptance of Syntactic vs. Pragmatic Interpretations ……….…... 3.5 Adult Preference for Syntactic vs. Pragmatic Interpretations ………..………...
4 Discussion ………..…...
4.1 Pronoun Comprehension in Children with Autism ………. 4.2 Comparison of the Syntactic and Pragmatic Accounts ………... 4.2.1 Preference for a coreferential vs non-coreferential interpretation …….. 4.2.2 Acceptance of a coreferential vs non-coreferential interpretation……... 4.2.3 Preferred interpretations in adults …..……….… References ………. Appendices ………... 47 48 48 48 51 52 57 58 58 61 61 61 63 66 71
List of Tables
Table 1 SRep Task and RCPM scores across groups………...……… Table 2 Results of accuracy ~ condition * group + (1|subjects) + (1|items) model……... Table 3 Results of acceptance ~ condition*group +
(1+subject|items) + (1|isubject) + (1|items) model ……….. 48 50
List of Figures
Figure 1 Exemplification of c-command……….
Figure 2 Exemplification of sentence (1)……… Figure 3 Exemplification of Sentence (2a)……….. Figure 4 Comparison of pronoun position in English (left)
versus Italian (right) adapted from McKee (1992)……… Figure 5 Exemplification of Sentence (6)……… Figure 6 Exemplification of Sentence (13a)……… Figure 7 Exemplification of Sentence (15)………..
Figure 8 Procedure for Choice Task………
Figure 9 Procedure for Verification Task……… Figure 10 Mean accuracy of groups across conditions of the
choice task………. Figure 11 Odds ratios for the fixed effects in Table 2………... Figure 12 Group preferences for coreferential interpretation
across sentence types……….……… Figure 13 A’ scores on verification task filler items……….. Figure 14 B”D scores between showing response bias in groups……….. Figure 15 Mean acceptance for each of the possible interpretations
of C-asp sentences………. Figure 16 Odds Ratios for the model in Table 3……… Figure 17 Adult acceptance of coreference for Principe C vs C-asp
sentences……….... 12 13 15 18 19 28 30 45 46 49 50 52 53 54 55 56 57
Introduction
Communication flows when a topic is presented, referred back to, and built upon. Having to restate points in full each time would leave discourse sounding pedantic. It is for this purpose that referential expressions, such as pronouns, exist. Pronouns are substitutes to nouns or noun phrases whose referents can easily be established with sufficient context in discourse. Children understand this relationship easily enough that by age six they can reliably identify the referents of both reflexive and personal pronouns (Chien & Wexler, 1990). The process is not quite as straightforward in children with autism spectrum disorder (ASD) who show marked difficulty with pronouns. Pronouns have semantic, syntactic, and pragmatic rules that govern their usage. In typical development, most research has favored a syntactic approach to explain the emergence of pronoun comprehension. This is represented by the binding principles (Chomsky, 1986). On the other hand, the major social communication deficits in ASD encouraged a pragmatic analysis of their pronoun difficulties (Martin and McDonald, 2003). What remains scarce is seeing if pronoun comprehension in autism can be attributed to grammatical deficits distinct from their social communication issues. The first aim of this study is to test the binding principles in children with autism.
The binding principles explain how pronouns and noun phrases become coreferential (i.e. refer to the same entity). These are the three rules, concisely known as Principle A, B, and C which rely on the notion of c-command to establish coreference. C-command is a hierarchical relationship among nodes in a phrase structure tree. Children are believed to possess an innate knowledge of this abstract syntactic relation that accounts for their understanding of pronouns and coreference. Research has shown that these principles are operative as early as three and across a variety of languages such as English (Chien & Wexler, 1990), Dutch (Philip & Coopmans, 1996), Russian (Avrutin & Wexler, 1992), Icelandic (Sigurjónsdóttir, 1992), and Italian (McKee, 1992) to name a few. However, the same interpretations of pronoun coreference can be achieved through the rules of discourse alone (Ambridge, Pine, & Lieven, 2014). This brings forth the broader question of whether it is necessary to assume an innate set of rules when it is possible for children can develop pronoun comprehension through experience. This debate is parallel to that of the nativists versus the constructivists, which can be explored through comparing the binding principles to the more functional account of pronoun coreference.
The debate seems moot under the circumstances that both accounts predict the same interpretation of coreference. They do diverge, however, with a sentence like “He was juggling when Pinnochio dropped the ball”. “He” as referring to Pinnochio would not be acceptable according to the binding principles but is valid under the discourse rules. This is due to the presence of progressive aspect in the main clause that encourages a coreferential reading (Harris & Bates, 2002). Thus, progressive grammatical aspect will be key in this study’s attempt to compare the binding principles against the pragmatic account of coreference. Doing provides perspective on which account better explains children’s acquisition of pronoun coreference.
Previous research into pronouns and ASD highlights the specific cognitive phenotype of autism. These are weak central coherence (Happe & Frith, 2006) and theory of mind (Baron-Cohen, 2000) deficits. Comprehension of pronouns requires sharing a mental space with a communication partner as well as integrating multiple communicative acts with context. However, these views are insufficient in explaining grammatical deficits found in autism. Initial investigations show that children with autism perform worse than controls on reflexive pronouns (Perovic, Modyanova, & Wexler 2013a). This signals that both syntactic and pragmatic domains are affected in pronoun comprehension in autism. Thus, comparing syntactic and pragmatic theories of coreference can also inform understanding of clinical populations.
On a final note, it is worth mentioning that this study explores the binding principles in Tagalog. There is no shortage of crosslinguistic research on the binding principles, yet most can still be traced back to an Indo-European origin. As an Austronesian language, Tagalog is both typologically-different and understudied. This is of value to research on the binding principles as their innateness presumes them to be universal across languages.
Language Impairments in Autism Spectrum Disorders
Currently, the diagnosis of autism relies on the presence of observed social communication and social interaction deficits along with repetitive and stereotyped behaviors (American Psychological Association, 2013). The Diagnostic and Statistical Manual Fifth Edition (DSM-5) states that social deficits manifest on the whole communication process, emphasizing conversation, affect, body language, and eye contact as relevant areas. This is different from including language impairment as a diagnostic feature. Communication, although mostly involving language, encompasses more than the linguistic domains of phonology, semantics, syntax, and pragmatics.
While a portion of individuals with autism will exhibit problems in these areas, language impairment is much more variable in its manifestation along the spectrum (Tager-Flusberg, 2006). This has led to a shift in defining autism based on more consistent characteristics such as those found in the DSM-5 (Tager-Flusberg, 1996).
Despite the variability, language functioning is central to our understanding of autism. For one, speech delays are the first symptom that alert parents of developmental delays related to autism (Kim, Paul, Tager-Flusberg, & Lord, 2014). Language is also an important prognosticating factor for autism diagnosis. Children initially diagnosed with autism can receive a milder classification of Pervasive Developmental Disorder-Not Otherwise Specified (PDD-NOS) if they make sufficient gains in expressive and receptive language at 5 years old (Lord, Risi, & Pickles, 2004). It was also once proposed that autism arose from the same “general organic damage” to the brain that brought about childhood aphasia (equivalent to specific language impairment or SLI) (Rutter, 1965). Bartak, Rutter, and Cox (1975) found no support for this theory. Comparing the ASD group with a sample of children with only language impairments, they reported that half of the children in the latter group did not qualify as having autism. Furthermore, the autism group showed more extensive linguistic impairment and socio-emotional disturbances absent in the other group.
The early findings of Bartak, Rutter, and Cox (1975) do not sever the link between autism and SLI completely. In fact, recent studies show a stronger link between the two disorders. Both autism and SLI have high heritability estimates, with family members of those with autism showing significant histories of language delay (Bolton et al., 1994). Meanwhile, siblings of children with SLI also show increased risks for autism (Tomblin, Hafeman, and O’Brien, 2003). In terms of brain morphology, both autism and SLI groups show differences with controls in areas such as the inferior frontal gyrus and planum temporale (Tager-Flusberg, 2006). In non-language impaired right-handed controls, the left inferior frontal gyrus corresponding to Broca’s area is larger than the analogous area in the right hemisphere. Children with SLI and autism show an asymmetry that is either reduced or reversed. Apart from these neurobiological similarities, behavioral data from language tests support an overlap between the two diagnoses.
Kjelgaard and Tager-Flusberg (2001) conducted a battery of standardized tests to comprehensively describe the linguistic profile of 89 children with autism from ages 4 to 14 years.
There were measures for receptive vocabulary, expressive vocabulary, articulation, morphosyntax, semantics, and non-word repetition. Results show that there are no differences for receptive and expressive vocabulary knowledge and a positive relationship between overall IQ and language scores. More interestingly, they were able to divide their sample into three subgroups based on their receptive vocabulary scores and comprehensive language battery scores. These groupings were the: 1) normal language group (scores within normal range), 2) borderline group (scores between one to two standard deviations below the mean), and 3) impaired group (scores below two standard deviations below the mean). Articulation was within normal range for all three groups, but the borderline and impaired groups showed poorer semantic and syntactic knowledge compared to their vocabulary skills. Through this, it is established that language impairment is not universal in autism but is present in a specific subgroup. Tager-Flusberg and Joseph (2003) also support the existence of a language impaired subgroup, but instead divide individuals into autism with normal language (ALN) and autism with language impairment (ALI). The ALI group includes individuals in the borderline and impaired groups in the Kjelgaard and Tager-Flusberg (2001) study. Children with ALI omitted past tense and third-person singular morphology similar to children with SLI (Roberts, Rice, & Tager-Flusberg, 2004). However, error patterns such as echolalia were unique to the autism group.
This existence of an ALI subgroup motivates an investigation of linguistic domains apart from pragmatics in autism. Although overlaps in ASD and SLI were presented as justification, this study does not aim to prove that the link exists. A comparison was included to illustrate how linguistic profiling and linguistic probes commonly used to describe children with SLI can be informative in describing a specific subset of the autism spectrum. Deficits in vocabulary, semantics, and morphosyntax in the ALI group could not be sufficiently described by usual measures such as pragmatic checklists and discourse sampling used to examine social communication deficits. Grouping children with autism based on linguistic functioning could lead to more homogenous subgroups in research, as well as more specialized assessments and treatments. Pronouns are an example of research in autism wherein majority of explanations are pragmatic in nature (Baltaxe & D’Angiola, 1996; Charney, 1980). An experiment of the binding principles would complement this knowledge resulting in a more holistic understanding of pronoun comprehension deficits in autism.
The Binding Principles
As introduced earlier, the binding principles refer to Principles A, B, and C that govern the coreference between pronouns and referential expressions (i.e. nouns or noun phrases). Each principle is specific to the type of coreference it delimits. Principle A refers to reflexive pronouns, Principle B is for personal pronouns, while Principle C is for referential expressions. Central to all of these is c-command, which describes the relationships contained within the phrase structure trees of sentences.
Figure 1 illustrates that nodes B and C are sibling nodes as they are both daughters of A. Therefore, B c-commands the encircled area that includes C and all its daughter nodes D, E, F, G, and H. C also c-commands B as its sister. However, A does not c-command any node because it dominates all other nodes. This c-command relationship is what establishes a binding relationship. Binding is defined as coreference under command, which means that the binder (node B) c-commands the bindee (node C and its daughters). Another important concept is that of the governing category, which for these purposes can be simply thought of as the clause (node A). The relevance of c-command is discussed in the next sections.
Figure 1. Exemplification of c-command. The figure is a representation of the underlying structural relationships in an independent clause. Letters are used in place of the actual constituents to emphasize form rather than meaning.
Principle A. This principle is stated as “an anaphor is bound within its governing category”. An example of an anaphor or reflexive pronoun is himself used in (1). Note that beside the words John and himself is the index “i" which means that the two entities corefer. Coreference is expressed as coindexation in the given examples. Since (1) is a sentence comprised on only one clause, its governing category is quite clear to be the entire sentence.
Principle A states that himself must be bound within this entire clause and its tree structure will help determine if it indeed is. Figure 2 shows the relationship of himself to the rest of the nodes. Here, himself is in the lowest level of the tree and is a daughter to the T’ node. The T’ node is analogous to the position of node C in Figure 1. If so, John known as DP1 is sister to T’ and
effectively c-commands T’ and all of its daughter nodes. Therefore, John c-commands himself and a binding relationship exists satisfying Principle A. This leads listeners to comprehend that himself refers to John in the sentence. Principle A demands that the antecedent for reflexive pronouns come from within a noun phrase in the clause.
(1) Johni washed himselfi.
Figure 2. Syntax tree of sentence (1). This sentence has a similar form to Figure (1) but with the actual constituents of the sentence provided. Although words are produced in a linear fashion, the diagram shows that meaning is extracted through existing abstract hierarchies.
Children reliably adhere to this constraint across languages such as English, Italian, Icelandic, Russian, and Hebrew (Avrutin & Wexler, 1992; Mckee, 1992; Sigurjonsdottir & Hyams, 1992; Grodzinsky & Kave, 1993). While Chien and Wexler (1990) state that mastery of Principle A occurs at around 4 years of age, Mckee (1992) showed that English-speaking children as young as 2;9 years are consistent in their comprehension of reflexive pronouns. The difference in findings may be due to difference in methodologies. Chien and Wexler (1990) used an act-out task in the form of a “Simon-Says” game wherein children were read a sentence containing either a reflexive or pronoun and are asked to perform the action using puppets. The youngest group of participants (2;6 years) had an accuracy of only around 13%. Performance increased steadily across groups and the oldest age group (6 years) had a mean performance of around 90%. In contrast, Mckee (1992) tested children via a truth-value judgment task. While one experimenter manipulates characters performing different actions, another voices a puppet who describes the events. The child is instructed to reward the puppet if its description is correct (match condition), but not if its description is incorrect (mismatch condition). Results showed that children in the younger group (2;9 to 4;7 years) were accurate on sentences containing reflexives 88% of the time. The discrepancy in performance might be due to the more demanding nature of the act-out task for younger children. In summary, children learning different languages show consistent understanding of reflexive pronouns starting 3 to 4 years.
Principle B. Principle B states that “a pronoun must be free (i.e., not bound) within its governing category”. Stated differently, it means that a pronoun cannot share its meaning with an antecedent that c-commands it and is in the same clause. This requirement is the exact opposite of what Principle A demands for reflexive pronouns, making the distribution of reflexives and pronouns complementary.
This constraint is illustrated in sentences (2a) to (2c). The phrase structure tree of (2a) is provided in Figure (3). The pronoun him has the entire clause (node TP) as its governing category. The structural relationship of John to node TP is similar to the prior example in that it c-commands node T’ and all its daughters. The difference is that the expression in the lowest level is a pronoun and not an anaphor. Thus, we have an existing c-command relationship between John and him within the governing category. Due to this, Principle B disallows coreference and effectively prevents him from referring to John. The correct interpretation is expressed in (2b) where John
and him have different indices. A Principle B violation is equivalent to letting John be co-indexed with him despite the presence of c-command in the governing category, which is equivalent to (2c). A listener mistaking (2c) as the correct meaning thus commits a Principle B violation.
(2) a. John washed him. b. Johni washed himj.
c. *Johni washed himi.
Figure 3: Exemplification of Sentence (2a). The structure of the tree is similar to prior examples, yet the presence of a pronoun rather than the reflexive constrains coreference differently.
In typical development, there exists a contrast in how easily children comprehend reflexive pronouns versus personal pronouns. Children mistakenly allow pronouns to corefer with local antecedents until 6 years of age. Using the same paradigm as the experiment with reflexives, participants in the oldest age group had only 64% accuracy in sentences containing a non-reflexive pronoun (Chien & Wexler, 1990). Children can easily identify what John admires himself means, but mistakenly provide the same interpretation to John admires him. This relatively late acquisition of personal pronouns is called the “Delayed Principle B Effect” (DPBE) and it has been observed in languages such as Russian (Avrutin & Wexler 1992) and Dutch (Philip & Coopmans, 1996). This asymmetry in Principle B performance poses an issue because it implies that one of the innate
set of principles is seemingly “acquired” later. Instead of accepting this explanation, authors have turned to the property of pronouns being able to form anaphoric relations even without binding.
Binding was earlier described as coreference under c-command. While binding requires c-command for two expressions to corefer, pragmatically-motivated coreference does not. Coreference simply means two expressions have the same referent, which can be established without binding through deixis, emphatic stress, and contextual information (Guasti, 2002). These occur in Evans-style contexts shown in (3) to (5), first discussed in literature by Evans (1980). Evans-style contexts license a coreferential interpretation otherwise prohibited by Principle B. The examples below are adapted from Chapter 8 of Guasti (2002). Capitalization indicates stress.
(3) You know what Wendy, Ariel, and Peter Pan have in common. WENDY admires Peter Pan, ARIEL admires Peter Pan, and PETER PAN admires him too.
(4) When Aladdin looks in the mirror, he doesn’t see Jasmine. Aladdin sees HIM.
(5) It is not true that no one has invited Michael. Michael has invited HIM [pointing at Michael].
In the above sentences, Principle B blocks coindexation between each of the pronouns and the referential expressions Peter Pan, Aladdin, and Michael due to the presence of c-command. However, the coreferential interpretation is accessible to listeners due the use of pragmatic devices. Evans-style contexts establish that pronouns can be interpreted coreferential either through binding or pragmatically-motivated coreference. The presence of both possibilities apparently contributes to the chance performance of children below 6 years.
Chien and Wexler (1990) performed an additional set of experiments wherein sentences with a referential NP (e.g. the bear) as subjects compared to those with quantified NPs (e.g. every bear). The rationale for this is that quantified NPs, as opposed to referential NPs, can only obtain a coreferential interpretation through binding. Children who were in the 3-year-old group obtained around 70% accuracy in the quantified NP condition. Chien and Wexler (1990) take this as evidence of Principle B being present at the same time as the other binding principles. Thus it is not syntactic knowledge that is at fault, but rather children have difficulty differentiating when licensed exemptions to Principle B apply to sentences with referential NPs as subjects (i.e. Evans-style contexts). Chien and Wexler (1990) attribute this to the late emergence of the pragmatic skill
“Principle P”. The development of Principle P at around 6 years of age increases the performance of children above chance level.
An alternative view is provided by Grodzinsky and Reinhart (1993) with reference to Reinhart’s Rule I, simplified, states: if the speaker has the option to express coreference with and without binding, the form involving binding is always preferred. If a child hears sentence (2) repeated below, the first step would be to form a representation wherein a coreferential interpretation can be achieved with binding. This is indeed possible with a sentence like (1). Since a form with binding to express coreference is available, then the speaker must have chosen sentence (2) to express a non-coreferential meaning. In Evans-style contexts, the listener must decide that the coreferential interpretation is only achievable without binding due to the pragmatic contexts involved.
(1) John washed himself. (2) John washed him.
The fact that listeners have to remember the target sentence, form two distinct representations, and compare and decide among the options creates a bottleneck in processing in young children. Grodzinsky and Reinhart (1993) state that this is the root of chance-level performance in children younger than 6 years. They also maintain that children, in fact, have a knowledge of Principle B and when to apply it.
The issue with both of these explanations is that the DPBE is notably absent in Romance languages (McKee, 1992), Norwegian (Hestvik & Philip, 2000), Greek (Varlokosta, 2000), and Hungarian (Margócsy, 2000). McKee (1992) compared children learning both Italian and English and found that the Italian-speaking group had 90% accuracy on Principle B sentences, while the English-speaking group only had 61% accuracy. The explanations mentioned above would predict no difference between groups, which is not the case. This phenomenon is termed the “Clitic Exemption Effect” or “CEE” (Bauuw, 2000), wherein languages who do not exhibit the DPBE have pronouns that are syntactic clitics. According to McKee (1992), clitic pronouns in Romance languages are found outside the VP node unlike in English (Figure 4) which affects the child’s interpretation of the governing category. Clitics are found higher up in the T’ node when compared to full pronominal forms which are found within the VP. This may confuse children that the
governing category of full pronominal forms is only the VP, incorrectly identifying the subject NP as a non-local antecedent. This is less likely with the syntactic structure of clitics, where the most immediate dominating node is the entire clause itself. As to why this is involves a new definition of the governing category and specifying different domains for reflexives and pronouns, which is outside the scope of this paper. This is also true for other explanations of the CEE such as “Scope Economy” (Di Sciullo & Agüero-Bautista, 2008) and the Guise Theory (Avrutin, 1994) which are explained in the respective references.
Figure 4: Comparison of pronoun position in English (left) versus Italian (right) adapted from McKee (1992). The clitic (cl) is found at the level of the “T” node which is immediately dominated by the entire clause corresponding to the governing category.
Principle C. The last of the binding principles, this states that “an R-expression must be free (i.e., not bound)”. This is similar to Principle B in its requirement for freedom, but this time attention should be paid to the noun phrase within the clause. In parallel to the examples above, sentence (6) has the expressions John and he only that the pronoun is now in the subject position. Listeners interpret he and John as referring to separate entities (hence the different indices). This sentence would have a similar phrase structure tree to the other two sentences, and through this it is established that he is in a position that c-commands John as we see in Figure 4. Allowing he to bind John would constitute a Principle C violation. Hence, the correct interpretation is the non-coreferential one where he refers to another person.
Figure 5: Exemplification of Sentence (6). This sentence also involves a pronoun, but the fact that the pronoun is in the subject position leads it to c-command the R-expression. Thus, sentences like these are interpreted through Principle C which would also disallow coreference.
Unlike Principle B, there is no delay in the acquisition of Principle C. Crain and McKee (1985) used a truth value judgment task to test TD children, and at 3 years old they correctly identify sentences that adhere to Principle C 73% of the time. Complementary to this finding is that of Guasti and Chierchia (2002) who found that Italian-speaking children can also correctly judge these sentences at around 3;10 years.
The evidence coming from early acquisition and cross-linguistic studies have lent much support for the innateness of the binding principles. In ASD where linguistic development is disrupted, does the extent go as far as affecting principles that are “universal” to the development of language? The following studies in the next section shed light on this query.
The Binding Principles in Autism
Initial investigations of the binding principles in this population were performed by Perovic, Modyanova, and Wexler (2013a) through the use of a two-choice picture selection task. Apart from testing pronoun comprehension, the study also aimed to distinguish whether the issues with pronoun comprehension arise from syntactic or pragmatic deficits in ASD. To answer this, performance on Principle A sentences was contrasted to that in Principle B sentences. Principle A
is assumed to be a purely syntactic principle, whereas Principle B interacts with pragmatic development. Thus, issues with only Principle A point to a purely syntactic deficit while issues with Principle B suggest an influence of pragmatic issues. In this study, 14 children with autism aged 6;6-17 (M = 11.6) were tested along with controls matched in linguistic and non-verbal abilities. There were four conditions: name-reflexive, name-pronoun, control-name, and control possessive. The name-reflexive and name-pronoun conditions were single-clause sentences that used a possessive DP as a subject and had a reflexive or pronoun in the object, respectively (e.g. Bart’s dad is touching himself/ Bart’s dad is touching him). Possessive DPs were used as it allows for two possible antecedents (i.e. Bart or dad), one c-commanding and one not. These were the critical conditions to test knowledge of Principle A and B. The control-possessive condition used single-clause sentences without any pronouns in the object position (e.g. Bart’s dad is petting a dog) and were used to test knowledge of just the possessive structure. The control-name condition had no possessive structures nor pronouns (e.g. Bart is pointing to Dad). The ASD group was found at par with the control group in their comprehension of pronouns but had a persisting difficulty with reflexives. They made errors in allowing reflexive pronouns to co-refer with a non-local antecedent. This pattern deviates from typical-development. Authors conclude that a distinct grammatical deficit exists in ASD, and possibly an impaired understanding of c-command relations in binding.
Elaborating the design and stimuli used in the prior study is necessary as those that followed it made use of the same method and stimuli. One such investigation was also done by Perovic, Modyanova, and Wexler (2013b) wherein they compared groups of children with ASD with language impairment (ALI) and ASD with normal language (ALN) to two other groups matched in age and NVIQ as well for some. The first group had William’s Syndrome (WS), while the other was a control group of typically-developing (TD) children. The purpose of these comparisons is to understand how issues in pronouns vary across different neurodevelopmental disorders and how they interact with language delays and intellectual abilities. Results showed that the ALI group was the only one who showed a specific problem with reflexive pronouns. The authors take this as evidence of a specific issue with grammatical knowledge in a subtype of autism with language disorders, unrelated to NVIQ deficits also seen in ASD and WS, as well as pragmatic deficits prevalent across the autism spectrum.
A similar methodology was adapted in Dutch by Geutjes (2014). A group of 13 children with ASD aged 6-14 years old were tested on translated versions of the sentences used by Perovic and colleagues. Standardized language testing was done to ensure that none of the participants were considered language impaired. When compared to age-matched TD controls, the ASD group did not differ in performance for both reflexive and personal pronouns. Despite this, the oldest child in the ASD group continued to make mistakes whereas the TD group had ceiling performance at age 10. The author takes this as evidence of pronoun comprehension issues in ASD.
So far, the studies discussed take the performance of children on Principle B sentences as similar to TD children (Perovic, Modyanova, & Wexler, 2013a; Perovic, Modyanova, & Wexler, 2013b; Geutjes, 2014). While statistically there is indeed no difference between controls and the ASD group, it is worth mentioning that the “typical” DPBE persists in the ASD group past the age of six (Perovic, Modyanova, & Wexler, 2013a; Perovic, Modyanova, & Wexler, 2013b). This prolonged delay might signal that there is something atypical in their comprehension after all. In a study conducted on 20 Greek-speaking children with autism, results showed that the ASD group was less accurate in their comprehension of clitic pronouns (Terzi, Marinis, Kotsopoulou, & Francis, 2014). On the other hand, reflexives and strong pronouns were not affected. Comprehension was tested using a picture-selection task, while production was tested via an elicitation task. While the earlier discussion on clitics only mentioned their influence on the syntactic structure, clitics also have specific semantic and pragmatic factors that license their use. Terzi and colleagues conclude that children with ASD likely have issues with the pragmatic conditions that require a clitic rather than a full pronoun. In typical development, clitics are seen to alleviate the DPBE through the CEE due to their morphosyntactic properties. This study on Greek shows that clitics are problematic in ASD due to the pragmatic requirements defining their proper use. On a larger scale, this emphasizes the need to examine languages with different pronominal systems as the unique characteristics may be variably affected in ASD.
A 2017 study by Khetrapal and Thornton utilized a dynamic truth-value judgment task (TVTJ) to examine knowledge of c-command and Principle A in 12 high-functioning children with autism aged 5;4 to 12;7 years. In this two-part experiment, the dynamic TVTJ was used. The test of c-command was performed using sentences that involved negation and disjunction, wherein they found that knowledge of c-command is intact in the autism group. Of greater interest is the test of Principle A, which differs from prior studies discussed. This involved presenting a
video-taped story. At the end, a puppet manipulated by the first experimenter asked another puppet to describe how the story ended. The second puppet would reply with a description which was also the target sentence. There were two conditions similar to the prior studies (name-reflexive and control-possessive) and children were provided four stories under each condition. The target sentences were similar to those used by Perovic and colleagues, but with a prepositional phrase added to the end to make it sound more natural (e.g. Bart’s dad washed himself with soap). In the name-reflexive condition, a rejection of the test sentence was always the correct answer as authors claimed that it would require more effort for the participant to override the “Principle of Charity” (i.e. finding an interpretation where the target sentence is true). The control-possessive condition had counterbalanced “yes” and “no” answers.
Both knowledge of c-command and Principle A were found to be similar across the TD and ASD groups. It shows that English-speaking children on the higher functioning end of the spectrum may not have syntactic issues. In an unpublished study by Khetrapal (2015), Principle B and C were also tested using the dynamic TVTJ on the same participants. Despite the lack of group differences found, the raw scores show that the ASD group scored 15-20 % lower. Perhaps a task emphasizing only grammatical features of sentences without the role of narrative comprehension would be sensitive enough to detect group differences.
Pragmatic Framework of Pronoun Comprehension
The earlier discussion frames pronoun coreference purely through the lens of the binding reflects the reality that not much work is done outside of this (Ambridge & Lieven, 2011). The dearth of literature, however, does not represent complete agreement that the binding principles are the best explanation of pronominal coreference. Van Hoek (1998) elaborates on a framework of coreference that adheres to cognitive grammar. Alongside this are several criticisms about the binding principles. First of which is the arbitrary notion that c-command determines pronominal coreference and not some other grammatical relation. Furthermore, syntactic trees are not representative of actual meaning or pronunciation which are the actual structures present in communication. While the binding principles predict coreference in single-clause sentences well enough, they do not explain coreference across discourse. It seems that the pure reliance on the binding principles relegates nuance and discourse to pragmatics instead of treating these as equally
relevant to interpreting coreference. Van Hoek’s deviates from this to focus more on the active conceptualization in the minds of communicators.
Fundamental to Van Hoek’s (1998) explanation are the concepts of accessibility and conceptual distance. Simply put, accessibility determines the use of a noun or a pronoun to refer to the concept the speaker is trying to reference. Accessibility is operationalized as the amount of effort required to retrieve the referent. Full noun phrases are less accessible as their meaning is not within the shared context. On the other hand, pronouns are more accessible as their meanings are already shared between the interlocutors. Complementary to this is the notion of conceptual distance. Full noun phrases are “farther” from both speaker and listener given that they are not shared in the immediate context. Conversely, using pronouns bring the concept “closer” as they refer to something already familiar to the participants. There is also variation of distance among pronouns parallel to person. The first and second person pronouns I and you referring to the participants are conceptually closer than the third person he/she which refers to an entity who is not involved in the exchange. Therefore, a concept that is accessible is also conceptually close and one that is less accessible is conceptually further away.
Using accessibility and conceptual distance, it is possible to explain why coreference is possible in (7) but not (8). In (7), the subject Suzy is not accessible and conceptually distant which is why the full noun phrase is required. It then becomes natural to subsequently refer to Suzy as her as she is now part of the shared context. The same logic applies as to why she and Suzy cannot co-refer in (8). The subject pronoun “she” implies accessibility and closeness, while Suzy does not. The conflicting discourse status of the two nominal expressions make coreferential interpretation impossible.
(7) Suzy saw a rat near her. (8) She saw a rat near Suzy.
Children strongly make use of accessibility information to decide on coreference even if it violates the binding principles. Sekerina, Stromswold, and Hestvik (2004) tested comprehension of short-distance pronouns (SDPs) with 16 children of 4;9 to 7;10 years and compared their performance to adults. SDPs are found in sentences like (9). Coreference would theoretically be disallowed by Principle B due to the subject c-commanding in the pronoun in the
preposition phrase. It would only allow a sentence-external referent to be the correct interpretation. In reality, the correct interpretation is ambiguous and processing theories would predict that the coreferential interpretation is actually preferred.
(9) The boyi has put the box behind himi/j.
SDPs are another exception to the binding principles. This is because the sentence-internal referent (the boy) is the more recent and more accessible referent than an entity not mentioned in the sentence. Both Sekerina, Stromswold, and Hestvik’s (2014) behavioral and eye-tracking data support this prediction. Adults found SDPs ambiguous, but had a higher preference for the sentence-internal referent. Children’s preference for the sentence-internal referent was much higher than that of adults, and they are able to detect the ambiguity albeit at a much slower rate. This shows that children are implicitly able to detect the ambiguity but are unable to utilize this as explicit knowledge in the picture-selection task. This study illustrates that coreference can be decided primarily using pragmatic factors instead of considering them as a supplement to syntactic principles.
Returning to the case of illicit coreference in (8), one could wonder if transferring the prepositional phrase to the beginning of the sentence such as in (10) would affect the interpretation. Despite the noun phrase being introduced first, the coreferential interpretation remains impossible. Van Hoek’s (1998) explanation of conceptual reference points is relevant to this phenomenon.
(10) Near Suzy, she saw a rat.
Conceptual reference points are starting points in which the interpretation of sentences or discourse can be made. Phrased differently, reference points serve as “topics” or prominent information. Once the listener adopts a reference point, then he/she uses that to recreate a context for interpreting the rest of the utterance. The portion of the discourse understood in the context of the reference point is called the dominion. Cases of illicit coreference are when full noun phrases, which have referents that are inaccessible, appear in contexts where the referent is in fact highly accessible. In other words, a full noun phrase cannot appear in the dominion of a reference point which it corresponds to. As for pronouns, it is important that they are contained within dominions to obtain meaning from a reference point. There are three important factors to determine reference points in discourse. These are prominence, linear order, and connectivity.
Prominence refers to the figure/ground asymmetry present within sentences, where some parts are more prominent than others. The subject is the most prominent element and serves as the reference point. This ties in with linear order, wherein the subject is mentioned first because of its prominence in discourse. This is mostly true for English, but it is important to note that cross-linguistically there are languages with free word order which may signal prominence differently. Returning to (7) as an example, Suzy is the reference point by virtue of being subject. The phrase near her is considered part of the dominion. The pronoun her is then interpreted in the context of the reference point Suzy, hence we have a licensed coreferential interpretation. In contrast to this is example (10). The subject she is the reference point. The use of a pronoun in the subject position implies a highly accessible referent from discourse context. Meanwhile, the full noun phrase in near Suzy implies that the full noun phrase is not immediately accessible. The noun phrase does not obtain its meaning from the context provided by the reference point, hence blocking coreference.
The next concept is connectivity. This refers to the degree of connectedness of concepts in sentences or across discourse. Concepts can be tightly connected within a sentence by a verb. As seen in (10), the verb saw forms a direct connection between the subject and direct object. This connectedness decreases if coreference spans across sentences such as in (11). The subject, the man, is established as the reference point. The following sentence which contains the pronoun he is part of the dominion. Due to this, coreference remains accessible given the context. Sometimes connectivity disappears, such as in paragraphs, and requires that the full noun phrase be repeated once more.
(11) The man with the umbrella crossed the street. He was in a hurry.
Young children are able to make use of prominence, linear order, and connectivity to aid their judgments of coreference. Children as young as 2.5 years prefer to interpret ambiguous pronouns as coreferential to the more prominent antecedent (i.e. mentioned first, subject position, mentioned more often) in the prior context (Song & Fisher, 2007; Song & Fisher, 2005). It was seen that this subject bias is correlated to increased linguistic awareness in adults (Arnold, Strangmann, Hwang, Zerkle, & Nappa, 2018) and higher print awareness in children (Arnold, Castro-Schilo, Zerkle, & Rao, 2019). These studies show that although there is a continuity in
adult and child pronoun comprehension, environmental influences can affect its development in each individual.
Finally, adults and children also make use of connectivity and prefer to maintain the same semantic roles of entities (Marastos, 1973). In (12a), there are two possible referents for him, but the preferred interpretation is to have him co-refer with Harry since Harry would maintain the object status in both clauses. However, when stress is applied to the pronoun (12b), this signals a change in roles and would encourage him to co-refer to John instead. While adults detect this function of stress, this shift is difficult for children to process. Only the oldest group of children aged 5 years were able to score above chance on sentences like (12b). This study shows that the development of supralinguistic cues may appear later than those found explicitly in utterances.
(12) a. John hit Harry, and then Sarah hit him. b. John hit Harry, and then Sarah hit HIM.
The fact that pragmatics plays a larger role in pronoun usage justifies why it is a widely-examined area in ASD. Research into pronouns in autism has largely been focused on reversals or sparsity in discourse. Pronoun reversals, where children with ASD confuse you and I, were one of the first symptoms described by Kanner (1943) upon the conception of the diagnosis. Early explanations of this phenomenon ranged from incomplete self-differentiation, avoidance of the pronoun I, learning language as “frozen forms”, and lack of attention to adult models (Oshima-Takane & Benaroya, 1989; Chiat, 1992; Charney, 1980; Fay, 1979).
Current explanations of the pragmatic difficulties are summarized by Martin and McDonald (2003), two of which are discussed here. The first of which is the social inference theory or better known as theory of mind. With this skill, one is able to represent the mental states of others and employ this during communication. This skill is clearly necessary in pronoun comprehension, as the prior discussion emphasized the need for speaker and listener to share reference points, assumptions of accessibility, and conceptual closeness of different entities. The second explanation offered is that of weak central coherence (WCC) which is characterized by the inability to use context to derive meaning. This is relevant in pronoun comprehension as listeners have to consistently infer meaning through the combination of speech, gesture, and facial expression. The plethora of information requires combining all knowledge into a cohesive whole,
which is a struggle when one has weak central coherence. Martin and McDonald (2003) do clarify, however, that WCC is not a deficit per se but just a difference in processing style. Whereas TD individuals are inclined to see the whole, persons with autism are adept at recalling seemingly random pieces of information (e.g. patterns, strings of letters, dates). This shows greater attention to the individual parts.
Either account seems to be suitable for the overall pragmatic difficulties seen in ASD (i.e. sarcasm, humor, narratives, mental state language) (Perkins, 2009). However, these all-encompassing explanations are missing how specific cognitive styles of ASD interact with specific features of language. Meanwhile, research in TD has the opposite problem of purely focusing on the structural and neglecting the context. As there have been few studies directly comparing both syntactic and pragmatic theories of coreference, knowledge in both typical and atypical development can be gained through directly comparing how well each theory predicts patterns of pronoun comprehension.
Comparing the Syntactic and Pragmatic Accounts
In the previous section, we have encountered criticisms against the binding principles that are also said about the larger generativist framework under which they belong. As Ambridge, Pine, and Lieven (2014) put forth, binding principles are mere byproduct of the existing rules of discourse. To insist on the need for innate syntactic principles would be redundant, especially given the numerous exceptions to these rules that children have to learn individually.
Principle C is of more interest as anaphora in multiclausal utterances better contrast the two accounts. Sentence (13a) provides an example of the Principle C constraint on binding between he and Pinnochio. It is comprised of two parts, a main clause (he plays the guitar) and a subordinate clause (when Pinnochio dances) (Figure 6). Principle C requires that he refers to an entity not mentioned in the sentence (exophoric reading such as in 13b). In Figure (6), we see the source of this interpretation. The subject of the main clause he c-commands the r-expression Pinnochio in the subordinate clause. Hence if he enters a binding relation with Pinnocchio such as in (13c), a Principle C violation is committed.
(13) a. He plays the guitar when Pinnochio dances. b. Hei plays the guitar when Pinnochioj dances.
c. *Hei plays the guitar when Pinnochioi dances.
Figure 6 Exemplification of Sentence (13a). Two clauses are shown in one tree, with the subordinate clause contained within the VP of the main clause
Meanwhile, the pragmatic account of Van Hoek (1998) offers an explanation based on the concepts of prominence, reference point, and dominion to interpret (13a). It was established that pronouns are often used to refer to entities accessible to the listeners and are considered old information. Furthermore, the subject position in the sentence is reserved for the current focus/topic of the discourse as it indicates prominence. The pronoun he, as subject, is established as the reference point. As was stated, unlicensed coreference occurs when a full noun phrases occur in contexts where the referent is highly accessible. The subordinate clause containing Pinnocchio cannot be considered part of the domain, because it conflicts with the discourse status of he. Thus, pragmatic account also determines an exophoric referent for he in (13a).
Despite both accounts preferring an exophoric interpretation, it is not the case that a pronoun and an r-expression can never co-refer when they are in a complex sentence. In fact, both accounts actually license binding and coreference if the pronoun is simply placed in the subordinate clause such as in (14a) and (14b). Whether the subordinate clause comes before or after the main clause is not of importance. The changes made in the location of the pronoun allows he to obtain its meaning from Pinnochio (i.e. anaphora). The position of the subordinate clause only changes whether we have backward anaphora (14c) or forward anaphora (14d), but both are
legitimate instances of binding. It should also be noted that the exophoric readings, wherein he refers to an unmentioned entity in the sentence, is also still possible (14e and 14f).
(14) a. While he plays the guitar Pinnochio dances. b. Pinnochio dances while he plays the guitar. c. While hei plays the guitar Pinnochioj dances.
d. Pinnochioi dances while hei plays the guitar.
e. While hei plays the guitar Pinnochioj dances.
f. Pinnochioi dances while hej plays the guitar.
Although both accounts make the same interpretations, they again differ as to how it is possible. The syntactic account claims that when the pronoun is placed in the subordinate clause, it no longer c-commands the r-expression. Hence, the r-expression is free, and no Principle C violation is committed. Using Van Hoek’s (1998) explanation, Pinnochio is consistently the reference point as the subject in the main clause. This makes the subordinate clauses part of the dominion allowing he to corefer with Pinnochio. Alternatively, Harris and Bates (2002) attribute this change to the “backgrounding” function of subordinate clauses in discourse. Subordinate clauses are said to express different contexts (e.g. reason, purpose, cause, manner, or means) in which the information in the main clause should be interpreted. All prominent information is contained in the main clause. Given this knowledge, listeners understand that information from both subordinate and main clauses should be integrated or that a relevant connection exists between them. Included in this integration is interpreting the pronoun in the subordinate clause and r-expression in the main clause coreferential.
So far, the sentences provided are unhelpful in determining whether the syntactic or pragmatic account drives pronoun comprehension in typical and atypical language development. This is where the role of grammatical aspect emerges. Harris and Bates (2002) contrast sentences wherein the verb in main clause is changed to have the past perfect tense or progressive aspect. The exact example they provided is seen in (15). In this sentence, Principle C would disallow he in the main clause to refer to Jack in the subordinate clause due to the presence of c-command (i.e. r-expression is bound) (Figure 7). The syntactic account would then predict an exophoric reading of he.
(15) He had been staring at the control panel for over an hour when Jack received a message from his commander.
Figure 7 Exemplification of Sentence (15) This has a similar syntactic structure with prior sentences, but the key contrast is the progressive aspect of the verb in the main clause.
However, the pragmatic account claims that it is possible that he refers to Jack in this sentence. That is, despite the alleged constraints set in the syntactic tree, the coreferential meaning is still available to the listener. Harris and Bates (2002) attribute this possibility to the “backgrounding” function of the progressive aspect. Initially, it was mentioned that subordinate clauses signal background information or represent the dominion of an utterance. If the progressive aspect also achieves this effect, then the main clause in (15) would be “backgrounded” or considered as the dominion of the utterance. This creates a different context wherein we would consider Jack as the reference point despite it being in the subordinate clause. He, as an anaphor, would need to obtain its meaning from the reference point. This would result in a coreferential interpretation of he and Jack.
Hopper (1982) explains the origins of the backgrounding function of grammatical aspect. In his analysis of narrative discourse, he states two kinds of events universal in narrative discourse. These are the foreground or “main line” events and the background or “shunted” events. Foregrounded events are sequential, punctual, and maintain a consistent topic throughout. In contrast, backgrounded events occur simultaneously with foreground events and often shift in topic. Therefore, while foregrounded clauses narrate, background clauses comment on the narration instead. Due to this, main events are more appropriately expressed with the perfective form, while
background events more frequently produced with the progressive aspect. Returning to discussion on multiple-clause sentences, progressive aspect in the main clause would convert its contents to a comment or elaboration on the assertion in the subordinate clause.
Harris and Bates (2002) substantiated Hopper’s (1982) this theory on the “backgrounding” function of grammatical aspect to in a series of four experiments in adult listeners. Their results showed that an anaphoric reading was allowed more in sentences with progressive aspect as compared to sentences where only the simple past tense was used the main clause. However, sentences that made use of a subordinate clause to present background information still yielded the highest number of coreferential interpretations. It appears that in English, syntactic subordination is the strongest cue for backgrounding information and grammatical aspect is a weaker cue compared to this. In conclusion, the observed effects of grammatical aspect on coreferential interpretation provides us an avenue to compare the two different accounts of pronoun comprehension
The Tagalog Language
Tagalog is an Austronesian language whose standardized form, Filipino, is the national language of the Philippines. It is a predicate-initial language and known for its “voice system” wherein the verb always has an affix to mark its relationship with the “privileged syntactic argument (PSA)”. The term PSA is used here as the question grammatical subjecthood in Tagalog is still debated (Kroeger, 1993). The affix on the verb expresses the thematic role of the PSA. If the PSA is a noun or noun phrase, it bears the marker ang before it to mark its relationship with the verb. All other remaining arguments are also preceded by a marker depending on their thematic roles.
In sentence (16), the agent babae (woman) is marked with ang which makes it the PSA. The verb is affixed with <um> to reflect the agent voice (AV). The argument isda (fish) is marked with ng to reflect that it is the patient (P). The remaining argument palengke (market) is marked with sa and is an oblique (OBL) in this sentence.
(16) B<um>ili [ang babae] [ng isda] [sa palengke].
AV.buy PSA woman P. fish OBL market
The order of arguments after the verb is relatively free. Sentences (17), (18), and (19) all mean the same as Sentence (16). Changing the word order also does not affect the case markings of the arguments.
(17) B<um>ili [ng isda] [ang babae] [sa palengke].
AV.buy P. fish PSA woman OBL market
The woman bought fish from the market.
(18) B<um>ili [sa palengke] [ng isda] [ang babae]
AV.buy OBL market P. fish PSA woman
The woman bought fish from the market.
(19) B<um>ili [ang babae] [sa palengke] [ng isda]
AV.buy PSA woman OBL market P. fish
The woman bought fish from the market.
Any of the arguments in Sentence (20) can also be promoted to the PSA. This is accompanied by a corresponding change in the affix of the verb. This reflects different thematic roles of the PSA.
(20) B<in>ili [ng babae] [ang isda] [sa palengke].
P.V.buy A. woman PSA fish OBL market
The fish was bought by the woman in the market.
(21) Binil<han> [ng babae] [ng isda] [ang palengke].
Locative.buy A. woman P. fish PSA market
The woman bought fish from the market.
(22) <Ipinag>bili [ng babae] [ng isda] [ang lalake] [sa palengke]. Benefactive.buy A. woman P.fish PSA man OBL market
The woman bought fish for the man in the market.
(23) <Ipinang>bili [ng babae] [ng isda] [ang pera] [sa palengke]. Instrumental.buy A. woman P. fish PSA money OBL market The woman bought fish using money from the market.
Tagalog personal pronouns can replace noun phrases in sentences. Pronouns are marked for person, case, number, but not gender. Pronouns are not preceded by the case marker as the pronominal form is already marked for case. In this study, only the third-person singular pronoun is used. This would be siya which would be equivalent to “he/she” in English. If the entity is not the PSA in the sentence, the form niya would be used for the thematic role of patients. Tagalog personal pronouns are said to be enclitics due to their obligatory placement after a word which they form an immediate part (Schachter & Otanes, 1972). In this case, siya always occupies the second position in the clause, usually after the verb being an immediate part of the entire sentence. This could be seen in the Tagalog versions of the sentences we used to explain the binding principles earlier.
(24) Pinaliguan siya ni John.
PV.wash 3P.SG.PSA A.John
John washed him.
(25) Pinaliguan niya si John.
PV.wash 3P.SG.A PSA.John
He washed John.
In both sentence (24) and (25), the pronouns occupy the place after the verb. Otherwise, the sentences would be ungrammatical. Sentence (24) is a Principle B sentence, and here the pronoun siya is the patient and is marked as the PSA. John, as the non-PSA agent, is preceded by ni which is a version of the case marker ng but for proper nouns. The sentence in (25) is a Principle C sentence and niya is the agent and is not the PSA, as John is which is shown by the si case marker (i.e. a proper noun version of ang).
As for the progressive aspect, Tagalog has agglutinative characteristics in terms of morphology. Apart from the voice affix, progressive aspect can be expressed through reduplication of the first syllable of the verb stem. For example, pinaliguan comes from ligo and marking the progressive aspect requires it to become pina<li>liguan.
Research Questions and Predictions
The first research question aims to answer whether there is a difference in accuracy, particularly for the Principle B sentences between TD children and those with ASD. Previous
findings show that although no group differences were found, the ASD group still did not reach ceiling performance at later ages and appear to have an extended DPBE effect as was seen in the studies by Perovic and colleagues (2013a, 2013b). Thus, a group by condition interaction is expected wherein children with ASD would show decreased accuracy for when asked to identify antecedents of personal pronouns in Principle B sentences.
The second research question is specific to comparing the syntactic and pragmatic accounts of coreference. Adults in the Harris and Bates studies were seen to be sensitive to the progressive aspect and allowed a coreferential interpretation in sentences like (6) when it was present. Typically-developing children are expected to mirror this performance with increased preference or acceptance for a coreferential interpretation (i.e. pragmatic account) for sentences with progressive aspect in the main clause verb. On the other hand, the pragmatic difficulties of children with autism may make them less sensitive to this backgrounding effect of grammatical aspect, so it is expected that children with autism will prefer the non-coreferential interpretation (i.e. syntactic account) for these kinds of sentences.
Method Participants
A total of 40 participants were recruited for this study. There were 11 participants each in the Autism (ASD) and typically-developing (TD) groups. Recruitment was performed via convenience sampling wherein invitations were extended to therapy centers, elementary schools, special education schools, and other personal contacts located within Metro Manila. To be included in the Autism group, children required a diagnosis of Autism Spectrum Disorder given by a developmental pediatrician, speak Tagalog as a first language or at a similar fluency to their first language (i.e. English), and are within 6 to 21 years old. Exclusion criteria were the presence of any sensory deficits and any other co-existing neurodevelopmental disorders. Criteria were confirmed only through the participants’ therapists or parents since access to medical records was not possible. Meanwhile, the typically-developing children had no existing diagnosis of any neurodevelopmental disorders, speak Tagalog as a first language, and are within 6 to 21 years of age. This was also based on reports of parents or teachers. The ASD group’s ages ranged from 8.2 to 21.5 years (M = 15.01, SD = 3.44) and 9 participants were male. Participants in the TD group were 8.0 to 20.3 years old (M = 12.41, SD = 3.56) and 3 were male. The recruitment of children with and without disabilities for this study has been approved by the ethics committee of the University of Groningen (Appendix G).
The 18 adults who participated in the online survey were also recruited via convenience sampling. The ages ranged from 22 to 27 years old (M = 23.94, SD = 1.30). There were 16 respondents who reported Tagalog as their first language and English as a second language, while two reported that Tagalog was second to English. Participants also spoke other Philippine languages such as Ilokano (n = 2) and Cebuano (n = 1), while one also spoke Mandarin Chinese in addition to Tagalog and English.
Materials
Sentence repetition task. Currently, there are no standardized versions of the sentence repetition (SRep) tasks available to test linguistic abilities in Tagalog. Therefore, an adaptation of the Recalling Sentences subtest of the Clinical Examination of Language Fundamentals 5 (CELF-5) was performed. This resulted in a test with two practice items and 26 target items grouped into
