Linguistic and cognitive abilities in children with Autism and a Language Impairment as compared to children with Specific Language Impairment

(1)

Linguistic and cognitive abilities in children with Autism and a

Language Impairment as compared to children with Specific

Language Impairment

Lindsey Ciebrant

Studentnumber: 10633545

Master Thesis: General Linguistics (Clinical Track) University of Amsterdam

Date: June 2, 2018

Supervisor: Jeannette Schaeffer Second reader: Laura Bos

(2)

Abstract

This study investigates the (morpho)syntactic, pragmatic, and extra-linguistic abilities of children with autism and language impairment (ALI). Eight children with ALI were matched on age and gender to eight typically developing (TD) children, and eight children with specific language impairment (SLI). The participants were experimentally tested on components of (morpho)syntax, pragmatics, and extra-linguistic cognition. The results show that the children with ALI performed poorly on the tasks related to syntax, while the children with SLI performed poorly on all linguistic tasks. Furthermore, both the ALI and SLI groups performed poorly on all extra-linguistic tasks. To investigate potential relations between linguistic and extra-linguistic skills, correlations were computed between all linguistic tasks and tasks that measure phonological memory and executive functions. With resulting (near-) significant correlations, we propose that the syntactic deficits in children with ALI are due to weak memory/executive functions, whereas the morphosyntactic deficits in children with SLI are due to weak memory/executive functions and a morphosyntactic deficit.

(3)

Acknowledgements

I can still remember the day my thesis subject started to take shape: it was the afternoon of 17 April, after Anne and I left a draughty room and said goodbye to Jeannette Schaeffer after our first meeting. Before the conversation, I never thought I would write a thesis related to children with autism and language impairment. Not because I did not have an interest in this topic, but simply because I did not know anything about it. In the tight months that followed, I knew that this topic would be the last thing I would think about before I fell asleep, and the first thing I would think about when I woke up.

Writing this thesis was not easy, but luckily I had the support of the people whom I want to thank in this section. Firstly, I would like to thank Jeannette Schaeffer for the support and supervision of my thesis. I am grateful that Anne and I were allowed to use the collected data from the database that belongs to you and Iris Duinmeijer, and I appreciate all your insights and feedback. Thank you for your confidence that things would end well; I consider it very inspiring to have met you.

Secondly, I would like to express my gratitude to Iris Duinmeijer for letting me use the dataset. Iris, although you were busy, you always patiently replied to all my e-mails about the dataset. In addition, you looked up results, audio files, and movie files and entrusted Anne and me with them.

My thanks also go to Laura Bos. I want to thank Laura not only for being the second reader on such short notice, but also for being my teacher for the last two years. In this time, I have learned a great deal about clinical linguistics, and I want to continue doing so.

I would also like to express my gratitude to my student colleague Anne Nagtzaam. We spent many hours in the university cafeteria together, where we thought about our hypotheses, transcribed and scored audio- and movie files, and executed our statistics. I am very grateful that I was allowed to go through this process with you, because without you, it would have been much lonelier.

Last but not least, I want to thank my family and friends for their endless support. Although you have no idea what I have been doing since the beginning of my Bachelor’s in Sign Linguistics in 2012, you have always supported me. My special thanks go to Jean (the father of my son). You always tell me to do what makes me happy and support me. Together with Keano (our stubborn two-year-old), you have taught me that life is not just about your education or career, and that you can achieve anything in life if you believe in it. Thank you.

(4)

1. Introduction

In the last few years, many researchers have examined the language abilities of children with specific language impairment (SLI) and autism spectrum disorder (ASD), and have compared the two clinical groups with each other. Typically, children with SLI show difficulties with (morpho)syntax, whereas children with ASD struggle with pragmatics. However, it has been demonstrated that the impairments experienced by children with SLI or ASD are not that unambiguous.

SLI is a disorder in language acquisition in children whose deficit in spoken language abilities cannot be attributed to neurological damage, hearing impairments, or cognitive deficits (Leonard, 2009; Leonard, 2014). However, recent studies suggest that many children with SLI have several additional impairments in extra-linguistic cognition (working memory and other executive functions) (Vugs et al., 2014; Torrens & Yagüe, 2016; Montgomery, 2000; Baddeley, 2003; Weismer et al., 1999; Henry et al., 2012; Weismer et al., 2000) and pragmatics (Steegs et al., 2010; Schaeffer, 2018; and Bishop, 2000; Bishop et al., 2000; Botting & Conti-Ramsden, 2003).

Most research on extra-linguistic cognition in children with SLI has been conducted on working memory (WM), and has found verbal WM impairments in individuals with SLI

(Gabig, 2008; Steele et al., 2007; and Montgomery et al., 2010; Archibald & Gathercole, 2006; Atkins & Baddeley, 1998; Daneman & Carpenter, 1980; King & Just, 1991; McDonald, 2008; Finney et al., 2014; Hill, 2015; Weismer et al., 1999; and Weismer et al., 2000 in Weismer et al., 2017), whereas deficits in non-verbal WM and visuospatial WM are scarcer (Archibald & Gathercole, 2006; Archibald & Gathercole, 2007; Vugs et al., 2014; Henry et al., 2012; and Bavin et al., 2005; Marton, 2008; and Im-Bolter et al., 2006 in Weismer et al., 2017).

Turning now to autism, according to the Diagnostic and Statistical Manual of Mental Disorders (DSM-V), children with autism spectrum disorder (ASD) are defined as having persistent deficits in social communication and social interaction across multiple contexts (pragmatics and discourse functioning) which cannot be explained by developmental delays or intellectual disabilities (American Psychiatric Association, 2013).

While children with ASD usually have difficulties with pragmatic phenomena, recent studies show that there are also subsets of children who display similar grammatical impairments as those reported for SLI (Kjelgaard et al., 2001; Roberts et al., 2004; Botting & Conti-Ramsden, 2003; Modyanova et al., 2017; Tuller et al., 2017).

(7)

In addition, WM deficits are also seen in subgroups of children with ASD. While some studies have found evidence of WM deficits (Joseph et al., 2005; Luna et al., 2002 and Schuh & Eigsti, 2002 in Weismer et al., 2017), others have reported normal performance (Dawson et al., 2002 and Koshino et al., 2005 in Weimer et al., 2017).

Based on the literature, it is only logical that the question has arisen of whether SLI and ASD are part of the same continuum (Bishop, 2010). It is not clear yet whether the pragmatic impairments in SLI are caused by weak WM and/or other executive functioning (EF), and whether the grammatical impairments in children with ASD are caused by pragmatic impairments.

Although many studies have compared the linguistic abilities of children with SLI and ASD, little research has examined the linguistic profiles of children with autism and a language impairment. The term ‘autism plus language impairment’ (ALI) will be used to describe this group. The studies that have examined the extra-linguistic cognition of children with ALI suggest that this group performs similarly on verbal WM and phonological memory tasks as children with SLI (Hill et al., 2015, Tager-Flusberg, 2015). Therefore, the question remains: do children with ALI perform poorly on both the (morpho)syntactic and the pragmatic level?

This study provides insight into the (morpho)syntactic, pragmatic, and cognitive profiles of children with ALI. In addition, it compares the (morpho)syntactic, pragmatic, and cognitive abilities of children with ALI to those of children with SLI, and of typically developing (TD) children. Our main research question is: How do children with ALI perform on (morpho)syntactic, pragmatic, and extra-linguistic cognitive tests in comparison to their TD and SLI peers? An additional focus of this study is the role of extra-linguistic cognition. Furthermore, the potential relation between weak memory/EF and (morpho)syntactic impairments is also investigated.

To answer the research question, this study used data from a large test battery of 16 tests designed and conducted by Jeannette Schaeffer and Iris Duinmeijer at the University of Amsterdam. The data served to analyse the language and cognitive profiles of eight children with ALI.

(8)

2. Theoretical background

2.1. (Morpho)syntactic phenomena

In the present study, four (morpho)syntactic tasks were used to investigate the linguistic abilities of children with ALI. This section presents the theoretical background of mass-count distinction, subject-verb agreement, object relative clauses, and sentence repetition. Every section includes an introduction to the (morpho)syntactic phenomena in adult language, followed by previous findings regarding the acquisition of the phenomena in TD children, children with SLI, and children with (high functioning) autism.

2.1.1. Mass-count distinction

One of the ways to distinguish between mass and count in Dutch is plural morphology on the noun. As illustrated in example (1), count nouns (e.g., dog, book, ball, shoe, and chair) can be used in combination with plural morphology, but mass nouns cannot (Chierchia, 1998). Mass nouns (e.g. water, sand, slime, and silverware) need classifiers or a measure phrase to demonstrate a change in volume:

(1) Morphosyntactic properties of count and mass nouns

Count Mass

Plural morphology dogs, books *waters, *sands

Classifier or measure phrase a bottle of water, three cups of sand

What the schema in (1) does not show is that besides mass nouns and count nouns, flexible nouns (e.g., rope/s, pizza/s, stone/s, paper/s, string/s) also exist. Flexible nouns can behave both as mass nouns and as count nouns. For example, the noun chocolate can be used as a mass noun when the noun is quantized by a measure phrase, as in (2), and as a count noun when it has a plural morpheme, as in (3). Therefore, the distinction between flexible mass and count nouns is marked morphosyntactically.

(2) Three pieces of chocolate are required for the recipe. (3) My brother brought me chocolates from Norway.

Finally, there are object-mass nouns. Object-mass nouns (e.g., furniture, mail, jewellery, footwear, clothing) have the morphosyntactic distribution of mass nouns, but seem to refer to a set of individual objects (examples of furniture: tables, chairs, sofas, beds).

(9)

Previous research on the acquisition of the mass-count distinction shows that TD Dutch children start making clear distinctions between mass nouns and count nouns at the age of six (van Witteloostuijn, 2013, van Witteloostuijn & Schaeffer 2014, van Witteloostuijn & Schaeffer, 2018).

Barner and Snedeker (2005) were the first to distinguish all four noun types as described above in the investigation of child language (mass, count, flexible, object-mass). They developed a quantity judgement task (QJT) to investigate the mass-count distinction in a group of English children and adults. In subsequent studies, the QJT was adapted. Among them, Witteloostuijn (2013) and van Witteloostuijn and Schaeffer (2018) used an (improved) QJT for Dutch.

In their studies, the participants were presented with pictures of two characters. In all pictures, one of the characters always had two large objects, while the other character always had four, five, or six smaller objects (in volume and surface) of the same kind. Subsequently, the participants were asked which one had more X. The experiment contained five conditions: classical mass, classical count, flexible count, flexible mass, and object-mass. Each contained four experimental items, except for flexible mass and count, which contained six items each.

The results of van Witteloostuijn’s (2013) study show that children from the age of four (4;1-12;6) were able to distinguish count nouns from mass nouns, and that they based their quantity judgement on number more often for count nouns (81%) than for mass nouns (4%). For the flexible count and mass nouns, the results show that the distinction between flexible mass and flexible count become clearer when children are around the age of six (6;2-7;11). At this age, 82% of the quantity judgements on number were based on flexible count nouns, whereas 23% were based on flexible mass nouns.

In conclusion, van Witteloostuijn’s (2013) study demonstrates that TD Dutch-speaking children acquire count nouns before mass nouns (overt morphosyntax), and that flexible nouns are acquired at a later stage than count and mass nouns.

Van Witteloostuijn (2013), and van Witteloostuijn and Schaeffer (2014; 2018) also investigated the mass-count distinction in children with SLI. The results of van Witteloostuijn’s (2013) study show that children with SLI (10;0-12;6) based their answers on number for flexible count nouns 81% of the time, while this only occurred 38% of the time for flexible mass nouns. For the classical mass-count distinction, the results of the SLI group correspond with those of the TD children: children with SLI (from the age of 6;2) based their

(10)

quantity judgement on number more often for count nouns than for mass nouns (65% versus 21%).

In summary, van Witteloostuijn’s (2013) results show that children with SLI mostly have problems with flexible nouns, because this group encounters difficulties with distinguishing the presence or absence of plural morphemes (e.g., pizza versus pizzas). Dutch children with SLI are unable to make the distinction between flexible mass and flexible count until they are ten years old. This is four years later than TD Dutch children.

Schaeffer (2018) compared the mass-count distinction in flexible nouns in 27 children with SLI and 27 children with high functioning autism (HFA) aged 6-14 (and a group of 27 TD peers) with an (improved) QJT based on Barner and Snedeker’s (2005). The results of this QJT show that unlike the children with SLI, the children with HFA performed according to the TD norm on the test, and did not seem to have problems with nominal morphology (collapsed flexible mass and count score: TD 73.50, HFA 46.89, SLI 27.05).

In addition, in the same study a Spearman’s rank-order correlation revealed a significant correlation between the WM scores and scores on the mass and count nouns in the SLI group. This provides evidence that WM abilities (verbal and non-verbal WM) are linked to (morpho)syntactic abilities, and that weak WM negatively impacts (morpho)syntactic tests (Schaeffer, 2016). However, these correlations were not found for the HFA group.

2.1.2. Subject-verb agreement

In Dutch, agreement between the subject and verb is expressed by attaching a suffix to the verb which expresses person and number. Therefore, subject-verb agreement is a morphosyntactic phenomenon (verbal morphology). Regarding subject-verb agreement, Dutch is not a rich morphological language, and has a limited number of verb endings (Spoelman & Bol, 2012). For an illustration of the Dutch verb paradigm for the verb juichen (to cheer), see table 1 (example from Spoelman & Bol).

(11)

Table 1. Subject-verb agreement in Dutch for the verb juichen (to cheer) Person Present Past

1sg (ik) juich juich-te 2sg (jij) juich-t juich-te 3sg (hij/zij/het) juich-t juich-te 1pl (wij) juich-en juich-t-en 2pl (jullie) juich-en juich-t-en 3pl (zij) juich-en juich-t-en

Previous literature on the acquisition of subject-verb agreement in Dutch TD children shows that verbal agreement is expressed correctly from the moment children produce two-word utterances (2;0-3;0 years old) (Blom, 2003; Polišenká, 2010). In Polišenská’s (2010) study, 12 monolingual Dutch children (3;0) participated in an elicitation task in which they were asked to finish sentences by the experimenter. Polišenká found that the agreement suffixes 3SG (-t) and 3PL (-en) were accurately applied in 93% and 100% of existing verbs, and in 100% and 93% of nonsense verbs, respectively. She suggests that the agreement suffixes 3SG and 3PL in Dutch monolinguals is productive at three years of age.

Whereas subject-verb agreement in Dutch TD children is productive at the age of three (Polišenká, 2010 & Blom, 2003), studies in children with SLI report that this group continues to produce root infinitives (mama lopen, “mommy walk”) until the age of eight (de Jong, 1999). De Jong’s (1999) work demonstrates that Dutch monolingual children with SLI have difficulties with the suffixation of 3SG and 3PL. In this study, 35 Dutch children with SLI (7;8) were matched with 35 TD children on age (7;7), and with 20 TD children on language (4;9). The outcome of the elicited production task (Pingu story retelling) shows that the children with SLI used the agreement suffix 3SG in 61% of the obligatory contexts, whereas the TD language- and age-matched groups used it in 87% and 89% of the obligatory contexts, respectively. In the SLI group, the agreement suffix 3PL was accurately used in 69% of the obligatory contexts, whereas the TD language- and age-matched groups had an accuracy of 95% and 97%, respectively. De Jong’s (1999) study suggests that children with SLI have impaired verb inflection representations, and still struggle with subject-verb agreement at the age of eight.

Little research has investigated subject-verb agreement in monolingual Dutch children with ASD. In Schaeffer’s (2018) work, the results of an elicited production task of subject-verb agreement (1SG, 2SG, 3SG, 1PL, and 2PL) in children with SLI and HFA (27

(12)

children per group aged 6-14) show that the children with SLI performed significantly worse in the subject-verb agreement task than the TD children, while the children with HFA performed significantly better than the children with SLI, and performed according to the TD norm. The results of the test revealed an accuracy score of 60.31% for the TD group, 51.2% for the HFA group, and 23.13% for the SLI group. In addition, within the SLI group, correlations were also found between phonological memory scores (NWR) and scores on the subject-agreement verb task.

Schaeffer’s (2018) results suggest that children with HFA do not have impairments in (morpho)syntax, whereas children with SLI do struggle with morphosyntactic processes required for subject-verb agreement.

2.1.3. Object relative clauses

An object relative clause (ORC) is a relative clause whose head serves as the object of the verb in the clause (see example 4b from Schaeffer & Siekman, 2016:138). Therefore, the ORC is derived by movement from the object position (Schaeffer & Siekman, 2016). To correctly produce an ORC, the speaker has to make use of his or her morphosyntactic (subject-verb agreement in singular and plural contexts) and syntactic knowledge (the placement of the object in the sentence structure).

(4) a. The girl pushes the boy

b. This is the boy that the girl pushed Object relative clause

Besides ORCs, there are also relative clauses in which the head is the subject of the embedded verb, also known as subject relative clauses (SRCs). As illustrated in (5a) and (5b), both Dutch ORCs and SRCs can be ambiguous between a subject and object reading (examples from Duinmeijer, 2017:178):

(5) a. ik ben liever de jongeni [diei ti (S) de vader (O) roept] Subject reading

“I would rather be the boy that calls the father.”

b. ik ben liever de jongeni [diei (O) de vader ti (S) roept] Object reading

“I would rather be the boy that is called by the father.”

(13)

Dutch relative clauses follow a subject-object-verb (SOV) word order, which complicates distinguishing SRCs from ORCs. In (4a), the head jongen is the subject of the relative clause, whereas in (4b) the head jongen is the object of the relative clause. This ambiguous reading can be solved by strategies like passivization and semantic factors such as animacy.

Studies on the acquisition of ORCs in Dutch show that TD children acquire ORCs later than SRCs (Thiel et al., 2014; Friedmann & Novogrodsky, 2004). In addition, research shows that the acquisition of ORCs (production and comprehension) does not occur before the age of six (Schouwenaars et al., 2014, Duinmeijer, 2017; Rademaker, 2014; Thiel et al., 2014). This corresponds to the results of the relative clause elicited production task in Duinmeijer’s (2017) work. The latter reports that TD children (6;4-11) produce adult-like ORCs with or without strategies like passivization (Duinmeijer, 2017).

In the past, research has examined the production, comprehension, and judgement of relative clauses of children with SLI (Friedmann & Novogrodsky, 2004; Novogrodsky & Friedmann, 2006; van der Lely, 1998; Zwitserlood et al., 2015). However, the present work only focuses on the production of ORCs in Dutch. In Duinmeijer’s (2017) aforementioned study, three different tasks were designed to test the knowledge and production of relative clause structures in 62 TD children and 63 children with SLI (two groups of young and older children aged from 6;4-11 and 12-15;10). The results of the relative clause elicited production task show that the children with SLI (young group: 63.55%, older group: 77.99%) produced less accurate ORCs than the TD children (young group: 95.27%, older group: 100%).

Little research has investigated ORCs in children with HFA. In 2016, Schaeffer and Siekman investigated the production, comprehension, and judgement of ORCs in 25 Dutch-acquiring children with HFA (aged 6-14). The results of the elicited production task show that the children with HFA did not differ from their TD peers in the ORC task (accuracy score: HFA 53%, TD 37%), and that TD children produced more correct singular ORCs than plural ORCs (p= 0.007).

2.1.4. Sentence repetition

Sentence repetition tests have been shown to be an efficient and reliable clinical marker to identify children with SLI in both the spoken and signed modality (Chiat & Roy, 2008; Conti-Ramsden, 2003 in Marshall et al, 2014). If the sentences that have to be recalled are long enough, children must draw on their (morpho)syntactic skills and memory system (phonological memory and verbal WM) to repeat them (Marinis & Armon-Lotem, in press).

(14)

Because a sentence rehearsal task involves language processing at all levels of representation (in comprehension, production, and WM components), a deficit in one or more of those domains and processes may affect the performance on the task (Marinis & Armon-Lotem, in press).

Because it is not fully clear what causes sentence repetition to fail, it is unknown whether sentence repetition demands successful comprehension (which requires semantic skills, (morpho)syntactic skills, and vocabulary knowledge) (Lust et al.,1987), or whether this failure is due to an impairment in (morpho)syntax, language processing, memory, and production (Polišenská et al. 2015).

Cross-linguistic studies show that sentence repetition tasks are successfully used with young children. Unfortunately, however, little research has investigated sentence repetition in monolingual Dutch children. Friedmann (2007) observed that children aged from two to four (2;11-3;11) responded very well to a sentence repetition task in Hebrew, whereas the test appeared to be inadequate for younger children (2;3-2;8). This suggests that three- and four-year-old children are able to rehearse sentences based on their (morpho)syntactic and memory skills.

In the past decades, many researchers have investigated sentence repetition in children with SLI. Schaeffer’s study examined linguistic and cognitive abilities in children with SLI and HFA, using the sentence repetition task as background measurement. In this study, 27 children with SLI and 27 children with HFA participated (6;-14) in the sentence repetition task from the Dutch version of the Clinical Evaluation of Language Fundamentals (CELF-4-NL) (Semel et al., 2008). The results of this test show that the children with SLI performed worse on the sentence repetition task in comparison with their TD peers, whereas only the children with HFA who failed on the pragmatic tasks (article choice, and direct object scrambling) showed weak sentence repetition performance.

In addition, in the same study, one significant correlation was found in the SLI group between the score of the sentence repetition task and the score of the digit span forward task. This provides evidence that WM abilities are linked to grammatical abilities, and that weak WM negatively impacts grammatical tests (Schaeffer, 2018). However, this correlation was not found for the HFA group.

In Riches et al.’s (2010) study, three groups of adolescents (TD, ALI, and SLI) participated in the English version of the Clinical Evaluation of Language Fundamentals -3 UK (CELF-3-UK; Semel et al., 2000). In this study, 17 TD children (mean age 14;4), 16

(15)

children with ALI (mean age 14;8), and 14 children with SLI (mean age 15;3) were asked to recall 24 sentences of the CELF. The results of this study show that both the children with SLI (mean score 22.1) and the children with ALI (mean score 22.2) performed worse on the sentence repetition task than the TD children (mean score 28.2). According to the authors, the results indicate that the group with ALI presented a similar kind of syntactic impairment as the group with SLI.

2.2. Pragmatic phenomena

In the present study, two pragmatic tasks were used to investigate the pragmatic abilities of children with ALI. This section presents the theoretical background of the pragmatic phenomena of article choice, and direct object scrambling. Every section includes an introduction to the pragmatic phenomenon in adult language, followed by previous findings of the acquisition of the phenomenon in TD children, children with SLI, and children with HFA.

2.2.1. Article choice

According to Stalnaker (1974), Heim (1982), Schaeffer (1999), Schaeffer and Matthewson (2005), and Schaeffer et al. (2018), the choice between a definite and an indefinite article requires the knowledge of speaker/hearer assumptions regarding the referent of the noun. Therefore, article choice is suggested to be a phenomenon driven by pragmatics (Schaeffer, 2018).

(6) Gisteren zag ik een (bepaalde) aap in de dierentuin. De aap was klein.

“Yesterday I saw a (certain) monkey in the zoo. The monkey was small.” (7) Ik heb zin om een film te kijken (wat voor film dan ook).

“I want to see a movie (whatever movie it may be).”

In the Dutch sentences above, in (6) the noun aap (monkey) is introduced by the speaker while the referent is unknown to the hearer. This context (known as context B) demands the use of the indefinite article een (a). In the second sentence in (6), the referent is known by both the speaker and hearer (also known as common ground; context A), requiring the use of the definite article de (the). In the sentence in (7), the referent of the noun film (movie) is unknown to both the speaker and hearer, also leading to the choice of an indefinite article (context C).

(16)

As illustrated in the Dutch sentences above, within the Dutch article system, there are three possible assumption states. These are proposed by Schaeffer et al. (2018) and are schematized in table 2.

Table 2. The Dutch article system (table from Schaeffer et al., 2018:93)

context assumed by label common ground Dutch article A speaker and

hearer

definite referential part of common ground

de B speaker only indefinite referential not part of

common ground een C neither speaker

nor hearer

indefinite non-referential not part of common ground

een

Previous research on the acquisition of article choice shows that young monolingual TD children often overgenerate indefinite articles in definite conditions (Keydeniers et al., 2017; Van Hout et al., 2010).

Keydeniers et al. (2017) report the results of a group of 14 Dutch-speaking children (2;1-3;7) who participated in an article choice sentence elicitation task. These results show that the children produced indefinite responses in a definite condition (16.7%) and definite responses in an indefinite referential condition (11.9%), while the overuse of definite responses in a non-referential condition (1.9%) was less common.

In the literature, the errors regarding article choice are either explained by either the lack of the concept-of-non-shared-assumptions (CNSA), or the failure to draw scalar implicatures. The CNSA appears when a child attributes his or her own (speaker) beliefs to the hearer. If the CNSA is absent, the speaker and hearer assumptions are not always independent, which prevents the child from distinguishing context A from context B; the child then groups context A and B as opposed to C (Schaeffer and Mathewson, 2005). Once the CNSA is acquired (3;0-4;0), the overgeneration of definite articles disappears.

The other explanation for article choice errors is the failure of scalar implicatures. Scalar implicature is “an inference that goes beyond the explicit meaning of an utterance, and is due to pragmatic factors” (Schaeffer et al. 2018: 95). The articles a and the are scalar terms and part of the entailment scale <a - the>, in which a is ordered as the weaker member, and the is the stronger and most informative member. Article errors appear because children do not always draw scalar implicatures, but arbitrarily choose an article for the use of either a

(17)

referential or non-referential indefinite article (van Hout et al., 2010). As children acquire scalar terms, errors regarding article choice decrease.

Turning now to previous research on article choice in children with SLI and HFA, Schaeffer et al. (2018) report that children with SLI and HFA have difficulties with article choice: both groups overgenerate indefinite articles in the definite condition (HFA 15%, SLI 13%). Schaeffer et al. (2018) suggest that a weak, underdeveloped, or impaired phonological memory is responsible for losing relevant information about the discourse, making it impossible to calculate scalar implicatures, and causing overgeneration of indefinite articles in both populations. Moreover, the authors discuss the possibility that an underdeveloped WM may be responsible for the pragmatic errors made by children with SLI and/or HFA. Although no correlation was found in this study between WM scores and overgeneration of indefinite articles, the children with SLI showed weaker WM than the children with HFA, suggesting that for the children with SLI, weak WM may cause pragmatic overgeneration of the indefinite article in the SLI group.

2.2.2. Direct object scrambling

In Dutch, the position of the object can occur before or after (sentential) negation or adverbs. This movement of object placement is typically referred to as object scrambling. Before an adult or a child can correctly scramble the direct object, the speaker has to take into account the beliefs of both the speaker and hearer to establish referentiality. Schaeffer (2000) defines referentiality as follows: “a nominal expression is referential if it has a ‘fixed referent’ implying that it is known to the speaker and/or to someone whose propositional attitudes are being reported” (Schaeffer, 2000:24). Since referentiality refers to the way a nominal expression is interpreted or understood, direct object scrambling requires (1) semantic knowledge, i.e. knowledge about definiteness and referentiality; (2) pragmatic knowledge, i.e. knowledge of speaker/hearer beliefs, and information structure; and (3) syntactic knowledge, i.e. knowledge of positions for direct objects in the sentence structure (Schaeffer, 2000; 2017). Example (8) illustrates two sentences of direct object placement. In sentence (8a), we see that a referential direct object tends to be scrambled, and thus precedes an adverb or sentential negation, whereas a non-referential direct object as in (8b) must remain unscrambled and thus follows an adverb or sentential negation.

(18)

(8) a. Jan heeft een boek niet gelezen (referential) (scrambled) Jan has a book not read “Jan didn’t read (a certain) book”

b. Jan heeft niet een boek gelezen (non-referential) (non-scrambled) Jan has not a book read “Jan didn’t read (any) book”

Schaeffer (1997; 2000) investigated direct object scrambling in 49 monolingual TD Dutch-acquiring children between two and three years old (2;4-3;11). The study tested direct object scrambling over negation and different types of adverbs (manner adverb, temporal/locative adverb) with an elicited production task. The results show that the children aged from 2;4 (70%) to 3;11 (28%) often failed to scramble over negation in referential contexts. Schaeffer (2000) concludes that underdeveloped pragmatics cause the failure to distinguish between different types of referentiality.

In 2017, Schaeffer conducted the elicited production task of direct object scrambling in a group of children with SLI (28 children aged 6-14) and HFA (28 children aged 5-14), to investigate whether direct object scrambling errors can be caused by impaired grammar or pragmatics. The results showed that both the children with SLI (45%) and HFA (63%) tended to scramble direct objects in the (definite and indefinite) referential conditions less often than their TD peers (84%), and that both the children with SLI (40%) and HFA (27%) produced more non-scrambled responses in the referential conditions than the TD children (8%).

Schaeffer (2017) indicates that the direct object scrambling errors made by both groups could not be accounted for by problems with referentially (ToM and CNSA), as was suggested for younger Dutch TD children (Schaeffer, 2017). Instead, Schaeffer (2017) argues that the direct object scrambling difficulties of the SLI group stemmed from a syntactic weakness (syntactic object placement), whereas the difficulties of the HFA group were due to a problem with consistently integrating the different language components required for direct object scrambling – namely, syntax, semantics, and pragmatics (underdeveloped pragmatic knowledge related to speaker/hearer assumptions (Schaeffer, 2018).

Furthermore, Schaeffer (2018) calculated whether the direct object scrambling errors of the children with SLI and HFA could be explained by cognitive functions such as non-verbal reasoning, non-non-verbal inhibition, and theory of mind. However, no correlations were found between the direct object scrambling task and cognitive tasks.

(19)

3. Hypotheses and predictions

3.1 Hypotheses regarding (morpho)syntactic tasks

To provide insight into the linguistic and cognitive abilities of children with ALI, this study investigates the (morpho)syntactic, pragmatic, and extra-linguistic abilities of these children, and compares them to those of TD children and children with SLI.

As discussed above, children with SLI show poor performance on (morpho)syntactic tasks (section 2.3.1, 2.3.2, 2.3.3, and 2.3.4), whereas children with HFA do not show any deficits in these tasks. Based on the previous findings and the assumption that the language impairment in children with ALI cause difficulties with (morpho)syntax, we hypothesize that children with ALI perform significantly worse on (morpho)syntactic tasks in comparison to their TD peers. We also hypothesize that children with ALI perform similarly on (morpho)syntactic tasks as children with SLI.

If the results show that the ALI group perform significantly worse on morphosyntax in comparison to the TD group, the first hypothesis will be verified. In contrast, if the results of children with ALI on morphosyntax do not significantly differ from those of the TD group (performance according to the TD norm), this hypothesis will be falsified.

Furthermore, if the children with ALI perform significantly differently on morphosyntax than children with SLI, the second hypothesis will be falsified. Similar performance on morphosyntax of children with ALI and children with SLI would constitute a null-result, rendering no strong conclusions.

3.1.1. Predictions: mass-count

The distinction between mass nouns and count nouns requires knowledge about (morpho)syntax. Based on the previous studies on Barner and Snedeker’s (improved) QJT (van Witteloostuijn 2013; van Witteloostuijn & Schaeffer, 2014; Schaeffer, 2018), we have seen that particularly children with SLI struggle with the production of flexible nouns, whereas children with HFA do not seem to have problems with nominal morphology, and perform according to the TD norm on tasks involving the flexible mass-count distinction. Based on the previous results and the assumption that the language impairment in children with ALI cause difficulties with (morpho)syntax, it is predicted that the children with ALI will perform worse on the mass-count task than TD children. We also predict that the children with SLI will score with less accuracy than the TD group.

(20)

3.1.2. Predictions: subject-verb agreement

As shown above, in Dutch, agreement between the subject and verb is expressed by attaching a suffix to the verb which expresses (singular) person and number. Therefore, subject-verb agreement is a grammatical phenomenon including morphosyntax and verbal morphology. Previous research (Blom et al., 2014, De Jong, 1999, Verhoeven et al., 2011, Schaeffer, 2018) shows that children with SLI have difficulties with subject-verb agreement, whereas children with HFA perform according to the TD norm (no grammatical impairment). Based on the previous results and the assumption that the language impairment in children with ALI causes problems with morphosyntax and morphology, we predict that these children will score with less accuracy on the subject-verb agreement task than their TD peers.

3.1.3. Predictions: object relative clauses

Previous studies on ORCs show that TD children produce more correct singular ORCs than plural ORCs (Duijmeijer, 2017; Schaeffer & Siekman, 2016). In addition, we have seen that children with HFA do not have difficulties with any ORC tasks (Schaeffer & Siekman, 2016), whereas children with SLI do show impairments in the production of these clauses. Based on the previous results, and the assumption that weak morphosyntactic and syntactic skills cause children with SLI to struggle with the production of ORCs, we predict that the same morphosyntactic and syntactic impairments in children with ALI will cause a lower performance on the ORC production task in comparison with TD children. In accordance with the findings reported by Duinmeijer (2017), and Schaeffer and Siekman (2016), we also predict that TD children will produce more correct singular than plural ORCs.

3.1.4. Predictions: sentence repetition

In line with Riches et al. (2010), we predict that children with ALI will have similar syntactic impairments as children with SLI. Therefore, we also predict that both groups will perform worse on the sentence repetition task in comparison with their TD peers.

(21)

3.2. Hypotheses regarding pragmatic tasks

As shown in sections 2.3 and 2.4, children with SLI perform worse on all linguistic tasks, whereas children with HFA only struggle with pragmatic-driven tests. Based on the literature and the assumption that children with ALI have both (morpho)syntactic and pragmatic impairment, we hypothesize that children with ALI perform significantly worse on the pragmatic tasks in comparison to their TD peers. We also hypothesize that children with ALI perform worse than children with SLI on the pragmatic tasks, because children with ALI are assumed to have problems with both (morpho)syntax and pragmatics, while children with SLI are assumed to have problems with (morpho)syntax only.

3.2.1. Predictions: article choice

Based on the previous findings that both children with SLI and HFA provide fewer correct definite articles in definite contexts, and overgenerate indefinite articles in definite contexts (Schaeffer, 2017;2018), it is predicted that children with ALI will also provide fewer correct definite articles in the definite context than TD children (similar to children with SLI), and will overgenerate indefinite articles. Based on the previous results and the assumption that the language impairment in children with ALI cause difficulties with (morpho)syntax, we also predict that children with ALI will perform worse on the article choice test in comparison with children with SLI.

3.2.2. Predictions: direct object scrambling

As discussed above, direct object scrambling is the movement of an object placement before or after sentential negation or adverbs that requires semantics, pragmatics, and syntactic knowledge. Schaeffer’s (2017; 2018) results regarding direct object scrambling show that both children with SLI and HFA tend to scramble direct objects in the (definite and indefinite) referential conditions less often than their TD peers, and that both groups produce more non-scrambled responses in the referential conditions. Based on these results, we predict that children with ALI will provide fewer correctly scrambled direct objects than TD children in referential conditions, and more non-scrambled responses in the referential condition.

We also predict that children with ALI will perform worse than children with SLI on direct object scrambling, because the previous findings and the assumption that children with ALI have problems with both (morpho)syntax and pragmatics, while children with SLI are assumed to have problems with (morpho)syntax only.

(22)

3.3. Hypotheses regarding extra-linguistic cognitive tasks and correlations

In the literature, mixed results are reported for various types of WM and other EF in children with SLI and (high functioning) autism. Therefore, we only make general hypotheses for extra-linguistic abilities that are frequently investigated: verbal WM and phonological memory. In line with the results of Hill et al. (2015) and Tager-Flusberg (2015), we hypothesize that children with ALI perform poorly in comparison to TD children on verbal WM and phonological memory tasks. Secondly, we hypothesize that performance on verbal WM tasks by children with ALI is similar to that of children with SLI. The first hypothesis will be verified if the ALI group scores significantly worse on verbal WM and phonological memory tasks in comparison to the TD group. In contrast, if the results of children with ALI on extra-linguistic cognition do not significantly differ from the TD group (performance according to the TD norm), this hypothesis will be falsified. Moreover, the second hypothesis will be falsified if children with ALI perform significantly differently on extra-linguistic cognition than children with SLI. Similar performance on extra-linguistic cognition of children with ALI and children with SLI would constitute a null-result, rendering no strong conclusions.

To determine whether there is a relationship between weak memory/ EF and (morpho)syntactic impairments in this study, we calculate the correlation between the scores on the extra-linguistic cognitive tasks and the scores on the (morpho)syntactic tasks. Based on Schaeffer’s (2018) findings, we hypothesize that there is a correlation between the scores of the verbal WM, and those of the phonological memory and (morpho)syntactic tests. The hypothesis will be verified if positive correlations are found between the verbal WM, and phonological memory and (morpho)syntactic tasks are found. If no such correlations are found, no strong conclusions can be made.

(23)

4. Methodology

4.1 Participants

For this study, eight Dutch-speaking monolingual children with HFA and a language impairment were selected from the database of Schaeffer and Duinmeijer, who recruited participants with HFA and SLI to investigate the grammatical, pragmatic, and cognitive development of children with HFA and SLI at the University of Amsterdam for their respective studies in 2016 (Schaeffer) and 2017 (Duinmeijer). The majority of the children selected in the present study were not included in Schaeffer and Duinmeijer’s earlier studies, which excluded children with additional impairments or diagnoses. The selected children were matched on age and gender with a group of TD children (control group) and a group of children with SLI, using the same database. In total, data from 27 Dutch-speaking monolingual children was used for this study (for an overview of the group information, see table 3).

Table 3. List of group information

The children with SLI were recruited from special schools for children with speech and language problems in the Netherlands, and were diagnosed with SLI by a certified speech-language pathologist. The children with HFA and a speech-language impairment were recruited either from special schools for children with speech and language problems, or from Dutch organizations for autism, autism groups on Facebook, and other contacts. The children with HFA and a language impairment all had a language impairment but were not all necessarily diagnosed with SLI. Three of the participants were diagnosed with SLI and ASD, three participants were diagnosed with ASD and dyslexia, one participant had ASD and a speech and language disorder, and one participant had ASD with a reading disorder and a learning disability (see table 4). Children with an IQ of less than 85 and/or who were officially diagnosed with any additional disorder (this only applied to the SLI group) were not included.

Group N Age (Mean)

ALI 8 10.14

TD 8 10.17

(24)

Table 4. List of participant information

To further confirm the SLI disorder of the impaired population, age-normalized scores of expressive and receptive linguistic ability were obtained from the Dutch version of the Clinical Evaluation of Language Fundamentals (CELF-4-NL) (Semel et al., 2008). Whereas the SLI group performed far below the norm score of the 50th_{percentile (Mean 15.88, SD 6.56) (as}

expected), the TD group performed around or above the norm score (Mean 49.75, SD 8.1), and the ALI group performed slightly below the norm score (Mean 30.38, SD 13.05). In addition to the experimental tasks, scores of non-verbal intelligence were obtained using Raven’s Progressive Matrices (Raven, 1976). The Raven test showed no difference in non-verbal intelligence between the groups (TD Mean 37, SD 13.87, ALI Mean 36, SD 11.99, SLI Mean 28, SD 12.63), suggesting that if the results of the morphosyntactic, pragmatic, and extra-linguistic cognitive tasks show a difference of performance between the groups, this difference cannot be due to non-verbal reasoning.

4.2 Materials and procedure

The tasks used for the present study are part of a much larger battery of 16 tests designed by Jeannette Schaeffer and Iris Duinmeijer at the University of Amsterdam. The tasks are

sex diagnosis age

ALI_01

m PDD-NOS, dyslexia 13.54

ALI_02 _m _{Asperger’s, dyslexia, highly intelligent} _13.3 ALI_03

m

PDD-NOS, Asperger's, learning disability (specific: reading disorder), idiopathic generalized epilepsy

12.44

ALI_04 _m _{PDD-NOS, ADHD, dyslexia} _11.7

ALI_05

m SLI, developmental disorder on the autism

spectrum (unknown) 8.35

ALI_06

m SLI, developmental disorder on the autism

spectrum (unknown) 8.34

ALI_07

m PDD-NOS, speech and language disorder 6.79

ALI_08

(25)

designed to investigate the grammatical, pragmatic, and cognitive development of children with HFA and SLI. All tasks were executed in Dutch.

4.2.1. (Morpho)syntactic tasks 4.2.1.1 Mass-count

The mass-count task, based on a quantity judgement task by Barner and Snedeker, is designed to investigate the distinction between mass and count nouns (2005). In this task, the participants were simultaneously presented with pre-recorded utterances and pictures of two characters (in Microsoft PowerPoint). In all pictures, one of the characters always had two large objects, while the other character always had between four and six smaller objects of the same kind. In this test, the volume and the surface of the two large objects was greater than that of the smaller objects combined (see figure 1).

The flexible nouns used for this test functioned either as mass nouns (without plural marking), e.g. ropes, or as count nouns (with plural marking), e.g. rope. In this test, the participants were asked which of the characters had more X. Nouns (X) represented by count syntax should elicit a response based on the number of individual items, whereas nouns represented by mass syntax should elicit a response based on volume. The mass-count task consisted of 12 experimental items per condition (mass and count), and 8 fillers. The fillers consisted of count nouns for which the objects only differed in number of items. For the results, the collapsed accuracy scores of the flexible mass-count nouns were used. Figure 1 presents a test item for count syntax (9) and mass syntax (10).

Figure 1. Example of test item in the mass-count task

(9) Wie heeft er meer pizza’s? “Who has more pizza-PL?”  Target response: the horseman   (10) Wie heeft er meer pizza?  “Who has more pizza?” Target response: the cowboy

(26)

4.2.1.2. Subject-verb agreement

The subject-verb agreement task consisted of an elicitation game (based on Blom et al., 2008 and Unsworth et al., 2014) designed by Duinmeijer (2012) to elicit the production of subject-verb agreement in the first, second, and third person singular and the first and second person plural. The task was presented as a game where the participants were asked to describe actions presented on cards. At the beginning of the game, the participant, the experimenter, and Kim (a doll) each received a pile of 30 cards illustrating a person acting out six transitive verbs (clean, read, drink, bake, film, and comb). The cards were presented upside down, and in every round the top cards were turned over and the participant was asked to say what everyone was doing according to the cards (i.e., What are you doing? What am I doing? What is Kim doing? What are we doing?). The production of the five conditions (as mentioned above) was presented by 12 items each, using the six transitive verbs twice. The produced items were scored as either correct or incorrect, based on which the accuracy scores for subject-verb agreement were calculated.

Figure 2. Example of the set-up of the subject-verb agreement task: elicitation production game (image from Duinmeijer, 2017:124)

Figure 3. Examples of the questions and target items during the practice trial (image from Duinmeijer, 2017:125)

(27)

4.2.1.3. Object relative clauses

Object relative clause production was tested using an elicited production task designed by Iris Duinmeijer, based on Novogrodsky and Friedmann (2006). The test elicits the production of relative clauses in three conditions (subject relative clause irreversible, subject relative reversible, and ORC reversible). Because the present study only focuses on the ORC (reversible) condition, we only elaborate on this condition here. Prior to the task, the participants were asked to help the experimenter with a list to gain more insight into what children like and do not like. During the test, the participants were presented with short stories about two children (girls or boys), and were asked to choose which of the two characters they would rather be (see example 11 for an illustration).

(11) There are two boys and a father and a mother. The father calls a boy and the mother calls a boy. Which boy would you rather be?

Target response: ik ben liever de jongen die de vader roept  “I would rather be the boy that the father calls”

Target response: ik ben liever de jongen die wordt geroepen door de vader “I would rather be the boy that is called by the father”

In the items that elicited object relatives, the two characters either underwent two actions or the same action executed by different subjects that were always animate. The items of the ORC condition were provided in the singular (as shown in example 11) and plural.

(28)

Importantly, singular subjects were ambiguous or reversible, while the use of plural objects is was not. The test consisted of 30 randomized items (of which 12 were object relatives), and started with a practice trial. The responses of the participants were either scored as correct (scrambled in the referential conditions or non-scrambled in the non-referential condition), incorrect (scrambled in the non-referential conditions or non-scrambled in the referential conditions), or irrelevant. For the results, we calculated the accuracy score of the total number of correct object relatives (or passives) for the singular or plural items in the collapsed referential conditions, and the non-referential condition.

4.2.1.4. Sentence repetition

Sentence repetition was assessed using the sentence repetition task of the CELF-4-NL, in which morphological, syntactical, and verbal WM skills are tested. In this test, the participant was asked to listen to spoken sentences of increasing length and complexity, and to repeat the sentence as faithfully as possible without changing the semantic, syntactic, or morphological structure. The test included 31 sentences, which were scored according to the CELF-4-NL. In this study, the scores are reported as the norm scores that were calculated according to the CELF-4-NL manual.

4.2.2. Pragmatic tasks 4.2.2.1. Article choice

Article choice was tested using an elicited production task based on Schaeffer and Matthewson (2005). In this test, an experimenter (B) sat next to the participant while the latter was asked to describe an event in a picture or a short video clip displayed on a computer screen (Microsoft PowerPoint) to an experimenter (A) who could not see the screen. The experiment consisted of 18 items in three conditions (six items in a definite condition, six items in an indefinite referential condition, and six items in an indefinite non-referential condition), and tested the participants’ article choice. For the filler items, 18 items from the direct object scrambling test were used (see 4.2.2.2.). The produced items were scored as either correct (definite or indefinite article), substitutions (definite or indefinite article), or irrelevant. Figure 4 provides an example of the definite condition of the task (for examples of the other conditions, see figure 22-23 in the appendix).

(29)

Figure 4. Example of the definite condition of the article choice task

Picture 1 Picture 2

Picture 1: Experimenter A (who cannot see the screen): “Hey, who do you see in the picture?”

Participant: [Name of the puppet]

A: “And what else?”

Participant “A bear!”

[the image (picture 1) changes into a short movie clip in which the puppet hugs the bear (picture 2)]

A: “And what did [name] just do?”

Target response: “He hugged the bear.”

4.2.2.2. Direct object scrambling

Direct object scrambling was tested using the same elicited production task used in the article choice task. The direct object scrambling part was based on methods used in a study by Schaeffer (2000), in which the test items elicited direct object scrambling with respect to sentential negation, rather than the use of adverbs (Schaeffer, 2018). Prior to the task, the participant was told that experimenter A was not always paying attention, and that if this experimenter said something wrong, he or she needed to be corrected. During this task, the participant listened together with experimenter A (who could not see the screen) to a story told by experimenter B (while looking at a picture). Subsequently, experimenter A asked a question or made a note which elicited the use of direct object scrambling in three conditions (six items in a definite condition, six items in an indefinite referential condition, and six items in an indefinite referential condition). The produced items were scored as scrambled, non-scrambled, or irrelevant. Figure 5 provides an example of the definite condition of the task (for examples of the other conditions, see figures 24 and 25 in the appendix).

(30)

Figure 5. Example of the definite condition of the direct object scrambling task

B: Patrick verveelt zich en kijkt of er iets leuks te doen is. “He”, zegt Patrick, “een boek! Maar ik houd niet van boeken.”

“Patrick is bored and he is looking for something to do. ‘Hey,’ says Patrick, ‘a book! But I don’t like books.’

“Dus dat ga ik NIET1_lezen”.

So that go I not read 

“So I’m not going to read it.”

A: Ik weet het. Het boek gaat Patrick WEL lezen. I know it. The book goes Patrick yes read “I know. Patrick is going to read the book.” Participant: Nee!

“No.”

B: Nee, he? Wat gebeurt er echt? No huh? What happens there really? “No? What’s really happening?” 

Target: Patrick gaat het boek NIET lezen. (scrambled) Patrick goes the book not read

“Patrick is not going to read the book.”

Non-target: Patrick gaat NIET het boek lezen. (non-scrambled) Patrick goes not the book read

4.2.3. Extra-linguistic cognitive tasks

4.2.3.1. Non-verbal working memory: odd-one-out

In this task, the participants were presented with a sequence of pictures with three geometrical figures, and asked to point out the one that was different from the other two. In addition, the participants had to remember the spatial locations of the geometrical figures, and were asked to indicate the location of the “odd-one-out” over trials. The task increased in complexity, starting with sequences of one item and continuing with longer sequences up to six items. The

1_{capitals indicate stress}

(31)

test consisted of six trials in which four sequences were presented. The test was broken off if two out of four sequences in a trial were incorrect. Figure 6 presents an example of the task.

Figure 6. Example of the odd-one-out task, and the picture recall (Henry, 2001 in Duinmeijer, 2017)

The task was administered on a laptop with a touch screen, and the responses were coded automatically in E-prime. This study used the accuracy score of the total number of correct items and the memory level (number of correctly remembered items).

4.2.3.2. Non-verbal inhibition: Luria’s hand-game

In this task, the participants’ inhibition was tested by using the motor part of an inhibition test battery (VIMI Hand-Fist game, Henry et al., 2012). This task consisted of two conditions (each of which comprised 20 trials). In the first condition, the participant was asked to copy the handshape (finger or fist) that the experimenter was presenting, and in the second condition the participant was asked to inhibit the copy response, and to show the alternative handshape (illustrated in figure 7). After the first condition, a second condition (flat horizontal handshape or flat vertical hand handshape) was tested. The items of the test were scored as either correct or incorrect, in which the copy and inhibition accuracy were separately scored. This study only used the accuracy scores of the inhibition responses.

Figure 7. Example of Luria’s hand-game with on the left the copy condition, and on the right the inhibition condition (Henry et al., 2012 in Duinmeijer, 2017)

(32)

4.2.3.3. Verbal working memory: digit span backward task

For the digit span backward task, auditory verbal WM was measured using the Wechsler Intelligence Scale for Children (WISC-III- NL). In this task, the participant was asked to repeat sequences of digits of increasing length in the same order. The digits were orally presented by the experimenter at a reading pace of one second per digit. After a practice trial of three items of a sequence of two-three digits, the participant started by recalling sequences of three digits, working to a maximum of eight digits. For each sequence two trials were administered. The test was cut off when the participant gave an incorrect response on both items of the two trials. In this study, the scores are reported as the average maximum number of digits repeated correctly.

4.2.3.4. Phonological memory: non-word repetition

In this task, the participants were asked to recall 40 non-words that varied in length (two to five syllables) and phonotactic probability (high/low according to the Dutch phonological rules) (developed by Rispens & Baker, 2012). The items ranged from non-words with two syllables with high phonotactic probability, e.g. kuimop, to items with five syllables with low phonotactic probability, e.g. geumuwoekuubir. The participants were asked to listen carefully to each word and repeat the non-words as faithfully as possible. The items of the non-words were scored as either correct or incorrect, and the scores on the task were presented as the number of correctly repeated non-words.

4.2.3.5. Verbal working memory: digit span forward task

For the digit span forward task, the same procedure was followed as in Wechlers’s digit span backward task. In this task, the participant was asked to repeat sequences of digits of increasing length in the same order.

(33)

5. Results

5.1. (Morpho)syntactic task results

This section presents the results of the (morpho)syntactic tasks. Recalling our two hypotheses that children with ALI would perform worse on morphosyntax in comparison to their TD peers, and would perform similarly on these tasks as children with SLI, we predicted significant differences between accuracy scores on the mass-count, subject-verb agreement, relative clause production, and sentence repetition tasks between the ALI group and the TD group, and between the SLI group and the TD group.

5.1.1. Mass-count results

Figure 8 presents the collapsed accuracy scores for the flexible mass and count conditions for all groups. Because not all children participated in this task, a few children were excluded from the statistical analysis. For this task, we used the results of five participants per group. A linear regression analysis using the lm() function in R calculated the difference between accuracy scores of the groups with an average score of 95 (SD 6.85) for the TD group, 83.35 (SD 17.91) for the ALI group, and 75.25 (SD 13.19) for the SLI group. The function shows that the children with ALI did not perform significantly differently than the TD children (95% c.i.= -8.12 .. 31.42; p=0.211), while the children with SLI did perform significantly worse than the latter (95% c.i.= 4.42 .. 35.08; p£0.018). Moreover, the children with ALI scored better on the mass-count task in comparison with the children with SLI (by 8.1%), but this difference was not significant (95% c.i.=-31.04 .. 14.84; p=0.439).

Figure 8. Accuracy score: mass-count. Figure 9. Accuracy score: subject-verb agreement *= significantly different from TD (p<0.05) ***= significantly different from TD (p<0.0001)

(34)

5.1.2. Subject-verb agreement results

Figure 9 displays the accuracy scores for subject-verb-agreement for all groups. The bars show differences in accuracy scores between the groups, with an average score of 100 (SD 0) for the TD group, 83 (SD 27.08) for the ALI group, and 72.4 (SD 28.66) for the SLI group. Similar to the mass-count results, the linear regression analysis revealed that the children with ALI did not perform significantly differently than the TD children (95% c.i.=-3.53 .. 37.53; p=0.098), while the children with SLI did perform significantly worse than the latter (95% c.i.= 9.5 .. 27.75 ; p£0.0006). Furthermore, the children with ALI performed better (by an average of 1.63%) than the children with SLI, but similar to the mass-count task this difference was not significant (95% c.i.=-24.09 .. 20.84; p=0.879).

5.1.3. Object relative clause production results

Figure 10 displays the accuracy of the singular and plural responses for all groups in the ORC production task. Because not all children participated in this task, a few children were excluded from the statistical analysis. For this task, we used the results of six participants per group. The light grey bars in figure 10 represent the singular responses, whereas the dark grey bars represent the plural responses. As indicated by the light grey bars, the accuracy of singular responses was higher than that of the plural responses. For both response types, the lm() function in R computed the difference between the accuracy scores of the groups. The analysis revealed a difference in responses between the TD group (sg: Mean 97.5, SD 6.94 ; pl: Mean 91.7, SD 13.87), the ALI group (sg: Mean 55.7, SD 42.81; pl: Mean 44.5, SD 44.24), and the SLI group (sg: Mean 33.3, SD 44.24; pl: Mean 33.3, SD 42.94). Contrary to the mass-count and subject-verb agreement tasks, the function showed that both the children with ALI (sg: 95% c.i.=2.05 .. 80.95, p£0.041; pl: 95% c.i.=4.99 .. 89.34, p£0.032) and the children with SLI (sg: 95% c.i.=21.65 .. 106.01, p£0.007; pl: 95% c.i.=14.76 .. 101.91, p£0.014) performed significantly worse on the ORC production task (both singular and plural responses) than the TD children. Moreover, the children with ALI scored better than the SLI group on the task, but not significantly (sg: 95% c.i.=-79.39 .. 34.73, p=0.404; pl: 95% c.i.=-69.12 .. 46.79, p=0.677).

(35)

Figure 10. Accuracy score of the ORC production task for singular and plural. *= significantly different from TD (p<0.05)

Figure 11. Number of items correct on the sentence repetition task (maximum score = 31).

**= significantly different from TD (p<0.001), ***=significantly different from TD (p<0.0001)

5.1.4. Sentence repetition results

Figure 11 presents the proportion of correct responses in the sentence repetition task for all groups. Because not all children participated in this task, a few children were excluded from the statistical analysis. For this task, we used the results of seven participants per group. As shown in the figure, the lm() function showed a difference between the correct responses of the groups with an average score of 13.57 for the TD group (SD 2.3), 7 for the ALI group (SD 3.76), and 3 for the SLI group (SD 2.44) (maximum score = 31). In line with our predictions, the function showed that both the children with ALI (95% c.i.= 3.08 .. 10.34; p£0.002) and the children with SLI (95% c.i.=7.38.. 12.90; p£0.0004) performed significantly worse on the sentence repetition task than the TD children. Moreover, the children with ALI scored better than the children with SLI on the task, with a nearly significant difference (95% c.i.=-7.12.. 0.26; p=0.066).

5.1.5. Summary

The outcomes of the (morpho)syntactic task show that the results of the children with ALI were not fully in line with our predictions. We found that the children with SLI performed worse on all morphosyntactic tasks (significantly different from their TD peers), while the children with ALI only performed poorly on the relative clause production and sentence repetition tasks. The following are the ALI group scores according to the TD norm on the mass-count and subject-verb agreement tasks (see example 1):

(12) Mass-count TD=ALI>SLI Subject-ver agreement TD=ALI>SLI Object relative clauses TD>ALI>SLI Sentence repetition TD>ALI>SLI

Linguistic and cognitive abilities in children with Autism and a Language Impairment as compared to children with Specific Language Impairment