The Effects of Mothers’ Directive Utterances and Language-Modelling Utterances on the Verbal Behavior of Children with Autism Spectrum Disorder —A Study Based on Lag Sequence Analysis

Parent-child verbal interactions are among the most critical factors influen-cing the conversational behavior of children with autism spectrum disorder (ASD). In this study, we investigated the frequency and sequential patterns of conversational behavior between children with ASD aged 3 - 6 years (n = 40) and their mothers in natural situations. The findings showed that mothers of Chinese children with ASD had 2.06 times more directive utterances than language-modeling utterances. The dominant child behaviors were answers, non-answer, and idiosyncratic conversations. In addition, when mothers used directive utterances, the conversational atmosphere was relatively tense, and it received more non-answer and idiosyncratic conversations along with responses from the children or even diversions to new topics. When mothers used language-modeling utterances, the conversational atmosphere was relatively relaxed, and children’s responses were more varied and freer of idiosyncratic conversations. The Effects of Mothers’ Directive Utterances and Language-Modelling Utterances on the of Children


Introduction
Autism spectrum disorder (ASD) is a complex neurological pervasive develop-

Language-Modeling Utterances
Language-modeling utterances refer to utterances that is responsive to children's utterances (Girolametto et al., 2000) and is also referred to as reactive utterances (Girolametto, Sussman, & Weitzman, 2007). These utterances are characterized by input behaviors that are child-centered (Matychuk, 2005;Snow, 1972Snow, , 1995. These behaviors include imitations, recasts, descriptions or comments, labels, and affirmation or support from the mother. Some research has found a positive correlation between most language-modeling utterances and language outcomes in children with ASD (McDuffie & Yoder, 2010;Siller & Sigman, 2002). The effects of these subcategories of language-modeling utterances on the language abilities of children with ASD are described below.
When mothers use labels (tags), descriptions, or comments on objects in the environment (without requiring children to respond), they can provide children with rich language input that allows children to make connections between words and objects (Girolametto & Weitzman, 2002;Hoff & Naigles, 2002). In Psychology McDuffie and Yoder's (2010) and Sigman's (2002, 2008) study, it was found that when parents used more follow-up comments (describing the child's focus of attention but not asking the child to respond) predicted the language ability of children with ASD after three years.
There are some similarities between imitations and recasts. Simply imitating the child's previous words may help maintain the child's attention and support comprehension (Sokolov, 1993). Recasts are adding to the child's previous utterances and structurally changing one or more components and may involve rearranging or expanding the child's utterances (Fey & Proctor-Williams, 2014;Nelson, Camarata, Butkovsky, & Camarata, 1995), which allows children to develop more advanced language models by contrasting their own utterances with more advanced adult utterances. In addition, another reason why recasting is thought to help children's language development is that it does not force children to respond. This gives children time to adjust rather than spending all their attentional resources on answering (Fey & Proctor Williams, 2000).
Maternal affirmation or support may help create a warm environment conducive to language learning, but to our knowledge, the link with language outcomes in children with ASD has not been investigated. Praise is also an expression of maternal affirmation or support, and parental praise has been used as a marker of positive parenting behaviors in many studies (Breitenstein et al., 2012;Chorpita, Daleiden, & Weisz, 2005). Although there is some debate about whether excessive use of praise negatively affects children's intrinsic motivation (Owens, Slep, & Heyman, 2012), there is now a large body of research showing that the strategic use of praise enhances children's feelings of confidence (Brummelman et al., 2014;Cimpian, 2010;Henderlong & Lepper, 2002;Zentall & Morris, 2010).
In summary, the significant difference between directive and language-modeling utterances is that the latter is child-directed, consistent with the child's current interests and language skills, and does not require an explicit response or "performance" from the child. (Girolametto et al., 2000;Murray & Hornbaker, 1997;Roberts & Kaiser, 2011). By reviewing the above literature, it is clear that directive utterances may further limit the opportunities for children with ASD to improve their communication skills, so children with ASD may require a higher proportion of language-modeling utterances to achieve maximum improvement in communication skills compared to normally developing children.

Deficiencies of Previous Studies
Although there are many pioneering studies in this area, they all suffer from the following two problems.
First, no causal conclusions can be drawn. Most of the parent-child studies we reviewed fall into two broad categories. In one category, only the mother's interactive behaviors were observed and then correlated with the child's linguistic productivity (language productivity), as measured by specialized scales: the MCDI (McDuffie & Yoder, 2010), PPVT-R (Chang & Luo, 2020), MSEL (Venuti Psychology et al., 2012); test scores (receptive and expressive language) (Drew et al., 2002) or expressive language skills (volume of words, total vocabulary, the average length of utterances, et al.) (Blom et al., 2020;Venuti et al., 2012) Another category is that both mother's and child's behaviors are counted or coded and correlated with each other, possibly also with other measures and variables (Jones & Schwartz, 2009). A problem common to both types of studies is that data analysis using correlational methods does not allow causal conclusions to be drawn.
Second, the interactivity between parents and children is ignored. The earliest studies of parent-child interaction primarily examined parents' influence on their children. However, current theory increasingly emphasizes that children bring important input to the interaction, suggesting that the relationship is not static but communicates in a mutually regulated, dynamic, and adaptive dialogue (Bretherton, 1992). Considering that both parents and children can play the role of initiator and responder in a dialogue (Cohn & Tronick, 1988;Fogel, 2009;Kelly & Barnard, 2000), therefore, it is vital to use a tool to measure each other's specific behaviors and to fully describe what each interaction partner brings to the engagement and how these characteristics influence the trajectory of the engagement. Reviewing the published and unpublished behavioral observation tools used to measure parent-child interactions from the 1970s to the present, there is a distinct lack of a tool that provides a comprehensive measure of parent-child interactions.

Current Study
To address the above questions, the present study used Behavior Sequence Analysis (BSA) to investigate the effects of mothers' verbal interaction behaviors on the conversational behaviors of children with ASD. Behavior sequence analysis, also known as Lag Sequence Analysis (LSA), is an effective method for understanding the dynamic interactions between complex behavioral chains. In LSA, behavioral chains are parsed to identify individual behaviors and behavioral shifts (Bakeman & Gottman, 1997;Clarke & Crossland, 1985). These behaviors are designated as specific codes that are recorded in the order in which the events occur (Keatley, 2018). The sequence of behavioral codes is then statistically analyzed to determine how often behavioral shifts occur and to calculate whether these shifts are significant and occur with a higher than expected probability. Currently, LSA is mainly applied to educational research, such as exploring the action paths of online learners (Hou, 2012), knowledge construction patterns in online forums (Wu & Hou, 2015), or criminal psychology (Ellis, Clarke, & Keatley, 2017;Longridge et al., 2020) and so on. The integration of LSA with language research is still rare. For example, LSA has been used to investigate the association between which repair strategies of mothers and the responses of children with language impairment (Barachetti & Lavelli, 2011;Lavelli et al., 2018) or Molinari, Mameli and Gnisci (2013) using LSA to explore the three-stage sequence of teacher-student initiation-feedback. Thus, this study Psychology aimed to examine the causal effects of mothers' directive and language-modeling utterances on the verbal behaviors of children with ASD using LSA.

Subjects
The participants in this study were 40 mother-child pairs of children with ASD who lived in China for long periods and used Mandarin as their daily language of communication and exchange. Demographic information of mothers and children with ASD is detailed in Table 1.
The inclusion of children with ASD was subject to the following criteria: 1) all participants had an operational IQ of at least 70, consistent with the criteria of other relevant studies (Siegel, Minshew, & Goldstein, 1996); 2) all participants had no vocal difficulties and had language; 3) the children were diagnosed in the appropriate department of a regular hospital and were diagnosed prior to the study according to DSM-IV criteria (APA, 2013). A validation diagnosis was completed and then judged by at least two experienced exceptional education practitioners, consistent with other relevant research criteria. All families participating in the study had signed an informed consent form.

Corpus Collection
This study used video or audio recording to record the interaction process. Most children used the video method, and individual mothers and children with ASD who did not accept the video method adopted the audio recording method. Most of the situations for the corpus collection were familiar to the children, which was to reduce their anxiety easily induced in the laboratory. In order to ensure the comparability of the interactive data of different children, the timing and content of the video recording were semi-structured. However, the verbal and interactional styles of the interaction were not limited to reflecting the natural daily conversation.
The video time was divided into a warm-up and a formal video phase. Warm-up phase: One of the researchers interacted with the child in the child's preferred College and above 5 5 5 5 Psychology manner to familiarize the child with the researcher until the child's wariness was reduced; the parents were informed to use their usual mode of interaction as much as possible throughout the process. Formal videotaping phase: Formal filming begins when the child and mother are ready. During the formal filming segment, the researcher did not enter into the interaction in any way, ensuring that each mother and child had more than 5 minutes of contextual material in each situation, with 20 minutes of effective material.
For the video (audio) content, the researcher designed three contexts, namely, the free talk context, the drawing context, and the play context. These three contexts basically cover the interaction between mothers and children in their daily lives in terms of play, interaction, and learning, which are the typical basic types of activities between children and mothers. Therefore, the research conducted in these three contexts is representative.

Transcription
The research assistant transcribed the video recordings using Computerized Language Analysis (CLAN) software following the "Codes for the Human Anal-

ysis of Transcripts" (CHAT) format in the international Child Language Data
Exchange System (CHILDES) (MacWhinney, 2000). This process involved transcribing the language of each child and mother in the video in textual form, marking each utterance individually, and classifying the lexical elements (Chang, 1997) until the entire sample was transcribed. Except for Chinese characters, all symbols were displayed in English and converted to text files that could be run in CLAN.
Some modifications were made and are listed in Table 2.
Children's language development requires acquiring conversational skills such as initiating, maintaining and repairing topics. (Capps et al., 1998;Loveland et al., 1988;Volden, 2004). The present study considers conversational skills from the perspective of conversation skills. In this study, children's behaviors were coded from the perspective of conversational skills: invitation, answers, expands, and repairs. In addition, considering that children with ASD have difficulties in all of these areas, that is, children with ASD often fail to cooperate with conversations and appear to respond inappropriately or not, Non-answer and idiosyncratic conversation (idiosyncratic answers, idiosyncratic expands, idiosyncratic repairs) in children with ASD were also coded. (Table 3)

Data Analysis
We used LSA to explore patterns of mother-child interactions. LSA is a helpful Psychology Words that direct the child's attention to an object or activity, or calling the child's name to get the child's attention" (Girolametto et al., 2000).

M2 verbal prohibitions or reprimands
Utterances that uses negative commands to interfere with or prevent behavior has already begun (Tulkin & Kagan, 1972); and words that scold, condemn, punish, criticize, warn, or exhort the child (Girolametto et al., 2000).

M3 follow-in directives
Words that command, suggest, or instruct a child to play in a certain way or demonstrate a specific behavior (Girolametto et al., 2000;Tulviste, 2004).

M4 questions
The mother asks the child a question to elicit an answer (Konstantareas et al., 1988) "Do you like to swim?"/ "What color is this?"

M5 request for clarification
Asking the child to clarify the meaning in the preceding utterances (Girolametto et al., 2000).
That's X/This is X/Red X/You have X.

M8 recasts
The mother repeats the preceding child's approximation or verbalization and completes the pronunciation by adding one or more morphemes or words (Girolametto et al., 1999).

M9 descriptions or Comments
Words describing the child's or mother's own activities (Girolametto et al., 2000).
Example 1, C: Little brother going to sleep -M: going to sleep -C: eyes closed (expanded).
Example 2, C: This is a fish -C: Small fish swimming (expanded).

C4 repairs
The child adds to his or her own or his or her mother's words.
Example 1, M: Where is the kitchen? -C: Over here -C: This side is all kitchen (repair).
C5 non-answer The mother asks a question, and the child does not answer within five seconds without answering (Grelle, 2013).
Example 1, M: What are you doing? -C: 0. (child does not answer within five seconds)

C6 idiosyncratic conversation
Words that do not fit the subject matter or are incomprehensible (Zambrana, 2007). tool for researchers to analyze linear relationships of behaviors and extract patterns. It also allows researchers to examine whether specific behavioral sequences reach a statistically significant level (Bakeman & Gottman, 1997). This study used a tool developed by Bakeman and Quera (Bakeman & Quera, 1995), a LSA tool called General Sequential Querier (GSEQ5.1), to analyze the data.
The process of using GSEQ is divided into the following steps: 1) Enter the encoded data into the GSEQ in a specific format.
2) The data is compiled with MDS files.
3) After the behavioral series analysis, the frequencies and adjusted residual tables are given.

Coding Reliability
All transcripts were coded by the author. 20% of the transcripts were randomly selected and independently coded by another coder. Cohen's Kappa statistic was Psychology used to calculate inter-rater reliability for all codes of verbal interaction. The mean of these indexes was 0.96.

Behavior Frequency Analysis
When analyzing mother-child interaction behaviors based on the coding system, a total of 26,278 mother-child interaction behaviors were obtained, and the specific distribution is shown in Table 4. As can be seen from Table 4, the behavior with the highest frequency during mother-child interaction was M3 (4504, 21.12%), followed by M4 (3395, 17.14%), and the results strongly indicate the importance of these two behaviors in mother-child interactions. The three behaviors with the highest frequency in children were C2 (3132, 11.92%), C5 (2784, 10.59%), and C6 (1130, 4.3%), and these results showed the characteristics of children with ASD when communicating to some extent. Based on the frequency of coded behaviors in Table 4, further statistics were done on mother-child interaction behaviors, as shown in Table 5. First, the results showed that the ratio of mothers' verbal behavior to children's verbal behavior is about 1.88:1, and the ratio of mothers' verbal behavior to children's nonverbal behavior is even 2.72:1. This showed an unbalanced nature of parent-

LSA of Mother-Child Verbal Interaction
A total of 26,237 mother-child verbal interaction behavior sequence relationships emerged during the conversation between mother and child with ASD, and the distribution is shown in Table 6. Each column indicates the beginning behavior that emerged, and each row indicates the behavior that occurred subsequently. The numbers indicate the frequency of the following behavior occurring after starting one. The table contents were converted to standard scores to obtain the residual table in Table 7. According to the theory of LSA, if the Z-score value > 1.96, the behavior sequence reaches the level of significance (p < 0.05).
Behavior sequences with Z-score values > 1.96 were filtered according to the residual table, and behavior transitions were plotted. The boxes in the figure indicate various mother-child speech acts, the connecting lines indicate that the transition between acts is significant, the arrows represent the direction of the behavior transition, and the data on the lines are the adjusted residual values (Z-score). To more clearly demonstrate the effects of maternal conversational behavior on children's conversational behavior, we plotted two patterns of behavioral sequences separately, according to the mother's utterances.

LSA of Directive Utterances
Pattern A is a pattern of verbal interaction based on the mother's directive utterances. The following findings were made regarding this part of the sequence (See Figure 1).
First, the M4 (questions)-C2 (answers)-M4 (questions), M4 (questions)-C5 (non-answer)-M4 (questions), and M4 (questions)-C6 (idiosyncratic conversation)-The highly significant relationship and one or even multiple cycles between the three sequences of M4 (questions) suggested that the mother's constant questioning and the switching of conversational rounds may have been Psychology effective. However, the utterances initiative was in the mother's hands. Second, the sequences C5 (non-answer)-M1 (lead attentional directives)-C1 (invitation), C5 (non-answer)-M3 (follow-in directives)-C1 (invitation) showed that when the child non-answer, if the mother tried to attract the child's attention by lead attentional directives to attract the child's attention or follow-in directives to the activity the child is doing, the child may start a new topic based on the scene at hand.
Finally, the sequences C1 (invitation)-M5 (request for clarification) and C3 (expands)-M5 (request for clarification) showed that if the mother's request for clarification followed the child's speech, there were repairs, non-answers, idiosyncratic conversations or answers. These results suggested that children's responses to their mothers' requests for clarification are diverse. It suggested that mother's requests for clarification may require specific strategies to be taught to receive repairs or answers from the child (Figure 1).

LSA of Language-Modelling Utterances
Model B is a model of interaction based on the mother's language-modeling utterances, and the following findings are specific to this part of the sequence (See Figure 2).
Overall, it can be found that in addition to M6 (labels), M7 (imitation), M8 (recasts), M9 (descriptions or comments), and M10 (affirmation or support) of the mother may elicit C1 (Invitation) and C3 (expands) of the child.
Second, the mother's labeling behavior did not impact the child, but the child may also benefit from the mother's labeling. It just did not manifest itself through immediate behavior.
Third, the sequences C5 (Non-answer)-M9 (Description or comments)-C3 (expands), C5 (Non-answer)-M9 (Descriptions or comments)-C1 (Invitation) suggested that description or comment by the mother when the child does not res-Psychology  Fourth, the sequences M10 (Affirmation or support)-C3 (Expansion) and M10 (Affirmation or support)-C1 (Invitation) suggested that the effect of maternal affirmation or support on children with ASD is uncertain.
Finally, regarding the sequence M8 (recast)-C3 (expands) and the sequence M8 (recasts)-C1 (Invitation), we can see one or more cycles between M8 and C3, suggesting a more significant role of maternal recasting in the expansion of the child.

Discussion
The results of this study demonstrate a new measure of mother-child interaction by showing the differences in the behavior of children with ASD between the two modes of mothers' use of directive utterances and language-modeling utterances. In addition, this study also used behavioral frequency analysis to demonstrate the frequency of use of directive utterances and language-modeling utterances by mothers and the child's conversational behaviors. The following discussion was conducted from two dimensions: directive utterances and language-modeling utterances.

Discussion of Directive Utterances
Through behavioral frequency statistics, we found that the ratio of mothers' directive utterances to language-modeling utterances was 2.06:1. Moreover, lead attentional directives (4505, 21.12%) were the most frequent of all behaviors among the directive utterances, about which two reasons can be explained. First, the strategies used by any parent in interacting with their child may depend on socioeconomic status and cultural norms (Gogate, Bahrick, & Watson, 2000;O'Neil-Pirozzi, 2006). This is unconsciously reminiscent of the cultural influences on mothers in China. According to reports, Chinese mothers use more instructional language (or utterances with instructional intent) with their children than American mothers (Kita, 2009). Jose et al.'s (2000 study showed that Chinese parents in America use more instructions when interacting with their children compared to European mothers in America. Second, according to Gillum, Nelson and Camarata (2003) mothers of children with ASD behave more directively during interactions than mothers of regular children, which was not included in this study, so there is no way to compare, but directive utterances did occur more often than demonstrative utterances. This may be because children with ASD face significant challenges in engaging in interactions and maintaining conversations. One study reported that children with ASD do not have frequent eye contact, turn-of-talk skills, and referential eye contact from a very young age (Wimpory et al., 2000). This is consistent with Krupa et al. (2019), i.e., parents constantly give instructions or commands to their children because they assume that such sessions will produce more joint participation and attention in their children.
Questions have been considered a strategy to encourage children to talk or Psychology elicit their language (Howlin et al., 1973), but sometimes it may not be as effective as they should be. In the present study, the questions (3395, 17.14%) behavior occurred second in frequency. It cycled once or even more in the behavioral sequence with the behavioral sequence of non-answer (z = 90.81, p < 0.05), idiosyncratic conversation (z = 17.88, p < 0.05). Such behavioral sequences suggest that children with ASD feel great pressure when their mothers constantly ask questions but fail to give timely verbal support, at which point they tend to fulfill their session obligations with idiosyncratic conversation or non-answer. When mothers use requests for clarification behaviors, children are asked to repair failures in conversation. According to previous research, the ASD group can respond to such communication failures; they employ various techniques to respond to conversational failures and add more information as the failures persist while producing more inappropriate language (Volden, 2004). In the present study, it was found that when the mother asked for clarification, the children showed a non-answer (z = 16.08, p < 0.05), answers (z = 6.26, p < 0.05), an idiosyncratic conversation (z = 7, p < 0.05), repairs (z = 22.27, p < 0.05), and one or even multiple cycles between them. This is consistent with Volden's (2004) study. More specific results were obtained in the current study. The behavioral sequences showed that child's responses to their mothers' Requests for clarification were diverse, suggesting that mothers should increase their sensitivity to capture children's conversations in time during the sessions.
Finally, the sequences in this study suggest that children's responses to maternal verbal prohibitions or condemnations and lead attentional directives are Invitations. Maternal verbal prohibitions or condemnations can protect children from harm (e.g., "Don't go there"), manage behavior (e.g., "Don't throw toys"), or instill good play behavior in children (e.g., "Don't hit"). In contrast, the directing attention command is designed to distract or engage the attention of the child with ASD, and the mother may go on to call their name more frequently. In addition, it serves a similar but different purpose than the follow-in directives. It is possible that the mother's purpose of lead attentional directives is to guide the behavior rather than elicit a response from the child and therefore does not contribute as much to the child's conversational development. According to previous studies, follow-up instructions (follow-in directives) are associated with better language outcomes in children with ASD (McDuffie & Yoder, 2010), so we argue that in addition to guiding behavior, mothers' lead attentional directives may facilitate the mapping between labels and objects or events in a similar way to follow-up comments (Mccathren, Yoder, & Warren, 1995;McDuffie & Yoder, 2010), just not manifested through immediate behaviors.

Discussion of Language-Modeling Utterances
By counting the frequency of behaviors of language-modeling utterances, we found that mothers used more descriptions or comments (3135, 11.93%) compared to labels (440, 1.67%), imitation (882, 3.36%), recasts (205, 0.87%), and affirmation or support (948, 3.61%) among all behaviors classified as language-modeling Psychology utterances. This suggests that using descriptions or comments may be a more naturally occurring interactive behavior for mothers of children with ASD than other language-modeling utterances. Therefore, when implementing a parent-centered intervention program, therapists should encourage parents to engage in more language-modeling utterances based on descriptions or comments. The following explains the sequence that unfolds around the mother's language-modeling utterances.
First, suppose mothers describe children's non-answer behavior (non-answer-descriptions or comments-expands or invitation). In that case, maternal descriptions or comments are thought to provide children with essential connections between words and objects in their environment, thus facilitating language learning in a responsive manner (Girolametto & Weitzman, 2002). In other words, descriptions or comments provide children with opportunities for verbal input, which also gives children the initiative to take control of the conversation. Another explanation, in conjunction with the findings of this study, is that if children are attentive and interested in the topic, it is likely that children will expand on the mother's topic if the mother describes the current event. For example, *M: Let the rabbit lean against the window, and then it can look outside the view. *C: You can see the view outside. In contrast, when children's attention is shifted, they may ignore the mother's descriptions or even initiate new topics, leaving the mother baffled. For example, *M: The little monkey is so poorly pinned down. *C: The man. *M: Who? Second, if the mother can imitate the child's idiosyncratic conversation (idiosyncratic conversation-imitation-expands or invitation), this may give the child the option to continue to expand the topic or invite a new one. Following Sokolov's (1993) suggestion, imitation or partial imitation lets children know that they are interested in their activity and is considered to provide them with "linguistically relevant information." In this way, children actively express themselves without limiting their responses, so they are free to invite or expand the topic of their choice.
Again, given that mothers' labels during interactions are thought to facilitate language learning because children make connections between words and objects when mothers provide markers (Girolametto & Weitzman, 2002). However, the results of this study are puzzling in that maternal labeling did not affect children's behavior. We speculate that perhaps the mothers used this opportunity as a teaching activity and did not care if the children responded. However, the labels may still have provided helpful information to the children and even reinforced existing knowledge about them.
Finally, affirmation or support creates a warm environment for children to have conversations. For example, *C: Get on the six-zero-two bus (C says while drawing the light). Psychology When the child was drawing, the mother affirmed or supported his behavior, the child described his own drawing, the mother gave affirmation or support, and the child initiated a new topic. For example, *M: You want to draw a bus, don't you? *C: Yes. *M: Okay (M looks at C in the drawing and says). *C: Door.
It is evident that the mother's affirmation or support initiates the child into new topics, which can sometimes be a good thing; the mother's affirmation or support lets the child with ASD know that they are on the right path, thus increasing their motivation and interest in generating more language.

Limitations and Future Directions
The current study uses LSA to support a causal explanation of mother-child with ASD interactions. There were far more boys than girls in the current study.
While this reflects the reality that more boys than girls are diagnosed with ASD, efforts should also be made to involve more girls with ASD if it is hoped that future research will benefit a larger group. In addition, this study focused on recording verbal communication behaviors during mother-child interactions.
However, nonverbal communication skills, including gestures, are also an essential part of mother-child with ASD interactions (Manwaring et al., 2017). Therefore, future studies need to examine the relationship between other verbal and nonverbal behaviors and children's communication skills. By doing so, we could understand how children with ASD benefit from interventions designed to increase the frequency and types of gestures.

Conflicts of Interest
The author declares no conflicts of interest regarding the publication of this paper.