**Changes in Developmental Trajectories of Preschool Children with Autism Spectrum Disorder during Parental Based Intensive Intervention**

### **Arianna Bentenuto \*, Giulio Bertamini, Silvia Perzolli and Paola Venuti**

Department of Psychology and Cognitive Science, Laboratory of Observation, Diagnosis and Educational (ODFLAB), University of Trento, 38068 Rovereto, Italy; giulio.bertamini@unitn.it (G.B.); silvia.perzolli@unitn.it (S.P.); paola.venuti@unitn.it (P.V.) **\*** Correspondence: arianna.bentenuto@unitn.it

Received: 11 April 2020; Accepted: 11 May 2020; Published: 12 May 2020

**Abstract:** Background: Research highlights the positive effects of early intensive intervention with parent and school involvement for preschool children with Autism Spectrum Disorder (ASD) on general developmental outcomes and social skills in randomized controlled trials. However, given the inter-individual variability in the response to treatment, it is necessary to investigate intervention effects in terms of mediators and moderators in order to explain variability and to highlight mechanisms of change. Methods: 25 children in the experimental group were exposed to early intensive intervention and 14 children in the control group were subjected to "as usual" intervention. The initial assessment was obtained at the time of diagnosis (T1) and the follow-up assessment was conducted after 15 months of intervention (T2) in both groups. Results: Participants in the experimental group achieved more prominent gains in both cognitive and socio-interactive skills. The role of specific factors able to predict general quotient and language quotient after intervention were investigated, pointing out the contribution of personal–social and performance abilities. Conclusions: The findings support the importance of parental involvement in targeting ASD core symptoms. Further, results informed our understanding of early predictors in order to identify specific elements to be targeted in the individualized intervention design.

**Keywords:** Autism Spectrum Disorder (ASD); early intensive intervention; developmental trajectories; moderators and mediators of intervention.

### **1. Introduction**

Autism Spectrum Disorder (ASD) is defined as a set of neurodevelopmental disorders (DSM-5) that impact on children's development by disrupting socioemotional reciprocity and producing a set of restricted repetitive patterns of behaviours and interests [1]. According to the Centres for Disease Control, about 1 of 59 children were diagnosed with ASD [2]. Psychoeducational intervention for children with Autism Spectrum Disorder (ASD) currently represents a main strategy to achieve symptoms reduction, promoting better adaptation and developmental outcomes [3]. Therefore, the increased prevalence of ASD led to a growing attention to early intervention research.

Different models of intervention started to prove their efficacy in randomised controlled clinical trials, together with longitudinally stable and generalizable outcomes [4–8]. Further, in line with this, a recent study review underlines how developmental interventions improve some specific areas, particularly socio-communicative domain in children with ASD [9]. Considering both efficacy and effectiveness of intervention, areas of improvement include IQ scores, verbal and non-verbal communication measures, adaptive behaviour and social and self-skills but there is less evidence of a significant impact on core autistic symptoms [10,11]. In line with this, specific improvement of

core autistic symptoms has rarely been reported, mainly due to the lack of scalable and quantifiable autism-specific treatment response measures, and due to the fact that standardized diagnostic instruments are not sensitive enough to detect changes after intervention [12–15]. While overall group improvements may be evident, the rate and the nature of these improvements is highly variable across individual differences in children with ASD [16]. Studies on efficacy show, in fact, great inter-individual variability in the response. Some children respond well to treatment (high-responders), whereas other children respond less to the same model of intervention (low- or non- responders) [17,18]. Variability in ASD in fact, not only concerns clinical expressions but also intervention outcomes [19]. Hence, it is difficult to identify one kind of intervention with the highest degree of efficacy compared to others, given that a specific intervention can be useful for specific domains and patients but not for others [12,13,20]. Despite this, treatments share some common principles: precocity, intensity, individualisation and integrated work [20–23].

To conclude, a great amount of research reported the efficacy of different kinds of intervention, underlying improvement of specific skills and highlighting the fundamental role of personalisation. For this reason, current research is focused on developmental trajectories of children with ASD during intervention [24–27]. The role of specific factors influencing intervention response need further investigation [28]. Some evidence indicates that factors associated with different responses include pre-treatment cognitive abilities [10,19,29,30], symptoms severity [31], adaptive skills [30,32], younger age [33], communication abilities [34], play skills [35,36], interest in objects [37], joint attention [36] and imitation [31].

Overall, studies on developmental trajectories focused on cognitive and/or adaptive functioning and symptoms severity pointing out different trends. Cognitive and/or adaptive skills showed major improvements compared to symptoms severity that are demonstrated to be more persistent [19,38,39]. Further, there is consensus regarding the importance, as prognostic factors, of IQ and speech level measured at the beginning of intervention. The level of language development is an important variable that has long been considered a predictive factor of child's outcomes [40,41].

In particular, children who received an intervention targeting early social intersubjective abilities have shown greater long-term language improvements than children in a control group [42]. Recent literature on developmental early intensive intervention focused mainly on interactive pleasure and exchange as a fertile ground to acquire competencies. In line with this, intervention intensity into the therapy room is not able to guarantee generalization of competencies if family and school are not encouraged to take an active role. Parents and school educators are, therefore, involved into the intervention program in order to generalize acquired competencies in more naturalistic settings. Further, there is some evidence that only children without intellectual disabilities at baseline were able to transfer the acquired socio-communication skills into daily life, therefore generalizing them [19]. In the Italian context, school represent a social opportunity in order to increase appropriate stimulations

In order to investigate developmental trajectories, we considered the learning rates, calculated as the difference between mental ages before intervention and after intervention and the time elapsed. It represents an alternative tool to measure change in studies of early intervention [43]. Through these indexes, it is possible to compare developmental profiles throughout time, not only at an absolute level but also taking into account the time elapsed between the two assessments with regard to the typical developmental trajectory. It clearly represents changes in age-equivalents over time and it is more appropriate when intervention lengths of time are similar, but not perfectly equal. Further, it represents an advantage when children functioning's are compared at different chronological ages. In fact positive learning rates mean that the child is narrowing the developmental gap. On the contrary, negative learning rates indicate a wider gap in the developmental trajectory. Learning rates may be useful for both outcome studies and progress representation of specific children functions [44], given that the value can be easily compared among them.

For the reasons expressed above, the purposes of the present work were: (1) to compare developmental trajectories for children receiving a parental based intensive intervention that provides 5–6 h per week, with both family and school involvement, with children exposed to "as usual" intervention, that provides 2–3 h per week of rehabilitative activities delivered by community services (see Methods' section for details); (2) to compare developmental trajectories of children with cognitive functioning equal or above 70 points at general quotient with children with cognitive functioning below 70 points at general quotient in both groups (3) to investigate the relationship between child pre-treatment characteristics and developmental trajectories. We had the following hypotheses in relation with the described objectives. First, we expected to find an overall increased level in cognitive abilities in both groups, however, we hypothesized a greater increase considering children exposed to early intensive intervention with family and school involvement, compared to children exposed to "as usual" intervention. Specifically, in relation to the intervention principles we hypothesized an increased level of linguistic skills. Secondly, we tried to identify a decreased level of autistic symptomatology, in particular considering the socio-communicative area, given the stability throughout the development of the restrictive and repetitive behavioural pattern [27,45]. Thirdly, consistently with previous studies [19,39], we expected that children without cognitive impairment showed major improvements in the developmental trajectory, compared to children with cognitive impairment. Finally, we hypothesized that specific child's variables might influence the developmental trajectory, specifically the chronological age and linguistic abilities at the beginning of the intervention were considered.

### **2. Materials and Methods**

### *2.1. Participants*

This study involved 25 children with Autism Spectrum Disorder (ASD) (M chronological age = 39.76 months, SD = 10.22; M mental age = 27.92 months, SD = 9.19) exposed to early intensive treatment with parent and school involvement delivered by ODFLab and 12 children with ASD (M chronological age = 45.33 months, SD = 8.34; M mental age = 33.17 months, SD = 12.80) subjected to "as usual" treatment delivered by community services in other regions after a diagnostic assessment at ODFLab (Table 1). All participants were recruited at ODFLab, a clinical and research centre of the Department of Psychology and Cognitive Science—(University of Trento) specialised in functional diagnosis of neuro developmental disorders, especially ASD, where families usually turn to in order to assess children's clinical profile. Moreover, the laboratory employed and currently delivers early intensive intervention with a developmental perspective in the local community [46]. Families coming from other regions usually turn to ODFLab only for the first assessment and monitoring of developmental trajectories every year. The intervention is therefore carried out in their local community services. All families involved in this project were adequately informed about procedure and agreed with a written informed consent. They were also aware of the possibility to drop out from the study in every moment.



The diagnosis of ASD was confirmed through clinical judgment by an independent clinician based on the DSM-5 criteria for Autism Spectrum Disorder, as well as through the Autism Diagnostic Observation Schedule (ADOS-2) [47].

The linguistic mental age was assessed through "Language and Communication subscale" of the Griffith Mental Development Scales. Considering the intervention group the average is 22.76 months (SD = 14.16) and for the control group the average is 27.75 months (SD = 13.51).

The socioeconomic status (SES) of the families, calculated with the Four-Factor Index of Social Status [48], indicated a middle status in the intervention group and a middle-high status in the control group.

### *2.2. Procedure*

All procedures of our study were in accordance with the ethical standards of the Italian Association of Psychology (AIP) and with the ethical standards of the Ethics Committee of the University of Trento (Italy) and the last version of Declaration of Helsinki [49]

In order to determine children's developmental level, the Griffith Mental Development Scale-Edition Revised [50] was administered to all children. Children were classified as "children without cognitive impairment" if they received a score equal or above 70 on the general developmental quotient and as "children with cognitive impairment", if they received a score lower than 70. In the experimental group, fourteen children (56%) were classified as children without cognitive impairment and 11 children (44%) were classified as children with cognitive impairment. Considering the control group, six children (50%) were classified as children without cognitive impairment and six children (50%) were classified as children with cognitive impairment. Taking into account the level of language development and the chronological age of children, ADOS Toddler, Module 1 and Module 2 were used to certify the presence of Autism Spectrum Disorder and to specify the severity level.

These measures (see measures' section for details) were applied before intervention (T1), during the first diagnostic and functional assessment. After intervention (T2), children were re-assessed in order to investigate developmental trajectories pre- and post-intervention, considering both cognitive and socio-interactive aspects. For participants in the experimental group (M = 14.72 months, SD = 4.36) and participants in the control group (M = 16.67 months, SD = 4.47) the amount of elapsed time, around fifteen months, is comparable.

### *2.3. Measures*

### 2.3.1. Griffiths Mental Development Scales-Edition Revised

The Griffiths Mental Development Scale, Edition Revised [50] was used to assess children's mental development level. The GMDS-ER are developmental scales normalized also in an Italian sample and are administered by trained psychologists to the child in a laboratory setting through standardized activities designed to evaluate different aspects of mental development in infants and children, providing scores relative to 6 subscales: Locomotion; Personal–Social; Communication and Listening; Eye–Hand Coordination; Performance; and Practical Reasoning. This scale provides a global quotient and a developmental age-equivalent—allowing to detect developmental delays—as well as specific quotients and developmental age-equivalents for each of the 6 subscales. Both global score and subscale scores were taken into account for the purposes of the present study.

### 2.3.2. Autism Diagnostic Observation Schedule-2 (ADOS-2)

In the present study, we used the Autism Diagnostic Observation Schedule-2 (ADOS-2) [47] both to confirm participants' diagnosis, to measure symptoms severity, and to investigate patterns of change before and after intervention. The administration of this tool is carried out by trained psychologists after an official ADOS course. For the purposes of this study, we used Toddler Module, Module 1 and Module 2. Each module gives a final score that classifies the child into mild, moderate or severe form of symptoms. Both global score and scores considering social-affect area and restricted, repetitive behaviours area are taken in consideration for the purposes of the present study.

### *2.4. Models of Intervention*

### 2.4.1. Parental Based Intensive Intervention

ODFLab (Observation, Diagnosis and Educational Laboratory) proposes and currently applies an "Italian Model of Intervention" which integrates empirically validated scientific principles together with guidelines in accordance to the Italian sanitary system and organization of educative system that guarantees a specialized educator for classrooms with children with special needs [22,46,51]. The intervention is individualized, comprehensive and integrates behavioural, developmental and relationship-based principles, according to the basic concepts of the Early Start Denver Model [10,13]. This intervention promotes Intentionality by giving to a child behaviour a communicative value so that he/she experiments that an action influences others behaviour and Reciprocity, starting from child behaviour to build up exchanges based on shift alternation. The therapist's goal is, therefore, to facilitate intentionality and reciprocity for children and share them with parents and educators. Further, intervention goals are constantly monitored and changed depending on the child's developmental improvements. Trained therapists aim constantly to create pleasant relationships starting from a child's own pleasure during shared activities [22].

The intervention is focused on the activation of interactive circuits during communication and on acquisition of specific functional competencies through psycho-educative activities. The intervention identifies key target areas and comprises specific activities and related objectives that are progressively adapted based on a specific observational schedule. This is regularly filled in by the psychologists to monitor the learning trajectory and disclose emergent abilities to be targeted during the intervention. Hence, the activities are highly integrated into playful routines to promote the development of specific objectives (e.g., language) by means of a comprehensive work on emergent abilities (e.g., communicative gestures or imitation). These principles are in line with Early Start Denver Model and more generally with Naturalistic developmental behavioral interventions [9,10]. In order to strengthen the generalization of child competencies it is fundamental to involve caregivers into the therapeutic setting from the beginning. In fact, caregivers represent a child's main interactive partners who, if they adequately learn appropriate interactive strategies, may effectively exploit them in more naturalistic settings. To this end, caregivers are involved in a child's social routines as an active part during intervention. For the same reason, they are fundamental to help school educators in understanding and responding to child behavior and structuring adequate activities. Moreover, in the educational context, it is possible to implement peer-mediated routines to promote appropriate social exchanges with peers that usually are not included in rehabilitative and psycho- educative activities. The intervention comprises:


The focus of the proposed intervention is mainly on building the "net"; in fact, given the pervasiveness of the disorder, the treatment necessarily has to be multimodal, integrated, rooted in the community and it should provide the fundamental involvement of both family and subsequently of school. In order to promote generalization of child's competencies, the network is aimed at providing appropriate strategies to detect and promptly respond to the child's needs, decreasing the child's frustration and boredom.

The intervention is delivered by licensed psychologists after receiving specific training on developmental models of intervention for children with ASD. The team is regularly supervised at least once every three weeks by an expert psychotherapist and all the psychologists have completed the introductory course to the Early Start Denver Model. Further, some of them attended the advanced course.

### 2.4.2. "As Usual" Intervention

With the term "as usual" intervention, we refer to specific rehabilitative activities such as psychomotricity and speech therapy employed by local community services. In particular, psychomotricity comprises a set of activities to promote communicative and relational abilities by means of body awareness and body movement. Psychomotricity is performed by professionals with a specific bachelor's degree. Moreover, speech therapy directly targets receptive and expressive language without a specific focus on socio-communicative routines. These specific activities represent effective strategies for intervention with preschool children with ASD [46] The intensity is generally from one to three hours per week, calibrated according to child's needs by the reference developmental neuropsychiatrist [46]. In the community services, no active involvement of caregivers and school is provided, but meetings for parents are planned if requested by them and two institutional meetings per year are planned with school educators to monitor the child's schooling.

From the two interventions' description, we would like to underline that the core difference regards the degree of involvement of social context families and school and not the specific rehabilitative activities known to be effective in dealing with children with ASD.

### **3. Results**

### *3.1. Analytic Plan*

The data were controlled for normality and homoscedasticity through the Shapiro–Wilk normality test and Levene test for homogeneity of variances. Parametric inferential tests (T test) were used when appropriate to identify group differences before the intervention (T1) and after the intervention (T2), as well as for investigating longitudinal changes. Otherwise, non-parametric tests were performed (Wilcoxon–Mann–Whitney test). Effect sizes were calculated using r2. Linear Regression models were implemented to test for predictors of change, and checked for assumptions. Repeated Measures Analyses of Variance (ANOVA) were performed to check for Group differences. Data were analysed using R statistical software [52].

### *3.2. Preliminary Analysis*

At T1, there were no significant differences in chronological ages between the intervention group (M = 39.76 months; SD = 10.22) and the control group (M = 45.33 months; SD = 8.84), and the time passed between the first and the second assessment was not significantly different between the two groups (t(35) = 1.26 ; *p* = 0.215; r2 = 0.044). Further, no significant differences (t(31) = 1.630; *p* = 0.113; r<sup>2</sup> = 0.08) emerged between the intervention and the control group regarding the socio-economic status of the families.

There were no significant differences at T1 and T2 between the two groups also regarding age equivalents of all the subscales of the Griffiths Mental Development Scales, as well as standardized quotients and the Autism Diagnostic Observation Schedule-Second Edition scores (Tables 2 and 3). Therefore, the whole sample was included to fit linear models. Then, paired T tests in both groups were performed to identify longitudinal changes.


**Table 2.** Developmental quotients in the two groups at T1 and T2.

\* *p* < 0.05; \*\* *p* < 0.01.

**Table 3.** ADOS scores in the two groups at T1 and T2.


\* *p* < 0.05; \*\* *p* < 0.01; \*\*\* *p* < 0.001.

### *3.3. Longitudinal Changes*

### 3.3.1. Cognitive Profile

Paired T-tests for dependent samples revealed a significant (t(24) = <sup>−</sup>2.320; *p* = 0.029; r<sup>2</sup> = 0.18) change in the General Quotient of the Griffiths Mental Development Scales between T1 (M = 73.64; SD = 15.84) and T2 (M = 79.12; SD = 22.02) for the intervention group. Children in the intervention group had a mean difference of 5.48 (SD = 11.81). The control group showed a non-significant (t(11) <sup>=</sup> <sup>−</sup>1.52; *p* = 0.156; r<sup>2</sup> = 0.17) longitudinal change between T1 (M = 74.08; SD = 19.5) and T2 (M = 69.50; SD = 18.28) in the General Quotient, with a mean difference of 4.58 (10.43).

Regarding the longitudinal changes for Locomotor, Personal-Social, Performance and Practical Reasoning subscales, no significant differences emerged between the intervention and control groups. However, the control group showed a significant (t(11) = <sup>−</sup>2.434; *p* = 0.033; r<sup>2</sup> = 0.350) improvement

in the Eye and Hand Coordination subscale between T1 (M = 64.00; SD = 17.73) and T2 (M = 73.25; SD = 17.32). The change between T1 (M = 72.80; SD = 18.87) and T2 (M = 78.12; SD = 22.43) resulted to be non-significant (t(24) <sup>=</sup> <sup>−</sup>1.77; *<sup>p</sup>* <sup>=</sup> 0.089; r<sup>2</sup> <sup>=</sup> 0.115) in the intervention group.

The Language Quotient showed a significant (t(24) <sup>=</sup> <sup>−</sup>3.387; *<sup>p</sup>* <sup>=</sup> 0.002; r2 <sup>=</sup> 0.32) change between T1 (M = 58.00; SD = 28.97) and T2 (M = 75.32; SD = 35.34) in the intervention group with an effect size indicating a strong effect in this subscale. Children in the intervention group had a mean difference of 17.32 (SD = 25.57), showing strong improvements in the Language domain. The difference was significant (t(11) = <sup>−</sup>2.59; *p* = 0.02; r<sup>2</sup> = 0.38) also for the control group, showing a mean difference of 9.58 (SD = 12.82), lower than the intervention group. (Table 2)

### 3.3.2. Socio-Communicative Profile

A significant (t(24) = 4.50; *p* = 0.0001; r2 = 0.46) difference in the ADOS-2 Total Score emerged between T1 (M = 16.20; SD = 4.15) and T2 (M = 13.60; SD = 4.33) in the intervention group, indicating a strong effect size. The difference was resulted to be non-significant (t(11) = 1.73; *p* = 0.112 r<sup>2</sup> = 0.21) in the control group, with a mean difference of -1.58 (SD = 3.18) and a lower effect size. Regarding the intervention group, a significant (t(24) = 4.08; *p* < 0.001; r2 = 0.41) difference in the Social Affect area between T1 (M = 12.32; DS = 3.18) and T2 (M = 10.04; DS = 3.35) emerged, indicating a strong effect and a mean reduction of -2.28 (SD = 2.79). A significant (t(11) = 2.80; *p* = 0.017; r<sup>2</sup> = 0.42) difference between T1 (M = 11.75; SD = 3.55) and T2 (M = 10.08; SD = 3.48) was also found in the control group, with a mean difference of −1.67 (SD = 2.06). (Table 3)

### *3.4. Children with and without Intellectual Impairment*

Afterwards, to further investigate trajectories of change, the sample was differentiated in terms of cognitive functioning between the two groups. Coherently with literature and clinical standards, the threshold of 70 was considered in the General Development Quotient of the Griffiths Mental Development Scales. The filter yielded 14 children with General Quotient above 70 in the intervention group (11 children with General Quotient equal to or below 70) and six children above 70 in the control group (six children equal to or below 70).

Regarding the General Quotient, the children without intellectual impairment in the intervention group showed a significant (t(13) = <sup>−</sup>3.71; *p* = 0.003; r<sup>2</sup> = 0.51) longitudinal difference between T1 (M = 84.64; SD = 10.87) and T2 (M = 96.14; SD = 8.69) indicating a strong effect with a mean difference of 11.5 (SD <sup>=</sup> 11.61). This difference was resulted to be non-significant (t(5) <sup>=</sup> <sup>−</sup>1.41; *<sup>p</sup>* <sup>=</sup> 0.219; r2 <sup>=</sup> 0.28) in the control group between T1 (M = 82.83; SD = 6.77) and T2 (M = 87.67; SD = 9.77), with a mean difference of 4.83 (SD = 8.42) and a lower effect size.

Focusing on the Language subscale, children in the intervention group with a General Quotient above 70 at T1 showed a significant (t(13) <sup>=</sup> <sup>−</sup>4.00; *<sup>p</sup>* <sup>=</sup> 0.002; r<sup>2</sup> <sup>=</sup> 0.55) longitudinal difference between T1 (M = 73.79; SD = 29.54) and T2 (M = 102.14; SD = 17.84), indicating a strong effect with a mean difference of 28.36 (SD = 26.53). Children in the control group who had a General Quotient above 70 showed a non-significant (t(5) = <sup>−</sup>1.97; *p* = 0.106; r<sup>2</sup> = 0.44) difference between T1 and T2 in the Language Quotient. The effect size was still relevant, but the mean difference was 11.67 (SD = 14.50). The difference was resulted to be non-significant between the two groups with respect to children with intellectual impairment.

With respect to the ADOS-2, a significant (t(13) = 4.09; *p* = 0.001; r2 = 0.56) longitudinal difference emerged in the Total Score in the intervention group without intellectual impairment between T1 (M = 11.29; SD = 3.00) and T2 (M = 8.14; SD = 2.60), indicating a strong effect with a mean difference of −3.14 (SD = 2.88). The difference was not significant in the control group of children without intellectual disability (t(5) = 1.6; *p* = 0.17; r<sup>2</sup> = 0.34) with a mean difference of <sup>−</sup>2.67 (SD = 4.08) and a lower effect size.

No significant differences emerged with respect to the Repetitive Restricted Behaviors area in both children with and without intellectual disability.

Furthermore, considering the Social Affect area, a significant longitudinal difference (t(13) = 3.69; *p* = 0.003; r2 = 0.51) emerged for children without intellectual impairment in the intervention group between T1 (M = 11.29; SD = 3.00) and T2 (M = 8.14; SD = 2.60), indicating a strong effect and a mean difference of -3.14 (SD = 3.18). The difference was not significant for children without intellectual impairment in the control group (t(5) = 2.15; *p* = 0.08; r2 = 0.48), with a mean difference of <sup>−</sup>2.33 (SD = 2.66). With respect to children with intellectual impairment, no significant differences emerged between the two groups.

### *3.5. Predictor Analysis*

In the analysis of predictors of outcomes, all participants were considered without group distinction, given that all children received some form of intervention. Linear Regression Models were fitted in order to test the goodness of different sub quotients at T1 in predicting the General Quotient and the Language Quotient at T2.

The General quotient at T2 was predicted by the combination of Personal-Social (β = 0.46; *p* = 0.006) and Performance Quotients (β = 0.21; *p* = 0.041) and the Chronological age (β = −0.60; *p* = 0.003) at T1. The model was significant (F(4,32) = 27.38; *p* < 0.001; Adjusted R2 = 0.75) and explained a significant proportion of the variance. The Language Quotient term resulted to be not significant (β = 0.18; *p* = 0.082) in this model.

Then, the Language Quotient at T2 was considered as a dependent variable and possible predictors among the subquotients at T1 were investigated. The Language quotient at T2 was predicted by the Language Quotient (β = 0.67; *p* < 0.001), the Personal-Social Quotient (β = 0.50; *p* = 0.036) and the Chronological age (β = −1.12; *p* = 0.001) at T1. The model resulted to be significant (F(3,33) = 28.74; *p* < 0.001; Adjusted R<sup>2</sup> = 0.70) and explained a significant proportion of the variance.

### *3.6. Responders and Non-Responders*

The 41% of the total sample responded to the interventions with a recovery in the age-equivalent, having a positive learning rate. This group was defined as "responders". In particular, in the intervention group, there was a percentage of 44% of responders, while the control group had a 25% of responders.

To investigate the baseline characteristics of children who positively responded to the intervention, differences at T1 between the responders and non-responders groups were examined.

The General Quotient of the responders group (M = 79.36; SD = 8.85), was significantly (t(35) <sup>=</sup> <sup>−</sup>2.12; *p* < 0.05; r<sup>2</sup> = 0.11) higher than the General Quotient of the non-responders group (M = 68.00; SD = 18.70) at T1.

Considering the sub quotients, only the Language Quotient of the responders group (M = 69.86; SD = 24.27) was significantly (W = 80; *p* = 0.012) higher than the Language Quotient of the non-responders group (M = 52.00; SD = 27.71) at T1.

Considering the age-equivalents learning rate in the Personal-Social domain, a significant (t(35) = 3.90; *p* < 0.001; r<sup>2</sup> = 0.30) difference emerged at T1 in the ADOS-2 score of Repetitive Restricted Behaviors between the responders and non-responders groups. Responders group started with a mean score of 2.31 (SD = 1.49), while the non-responders group had a mean score of 4.54 (SD = 1.74).

Moreover, a significant (t(35) = <sup>−</sup>2.25; *p* < 0.05; r<sup>2</sup> = 0.13) difference emerged at T1 in the Personal-Social Quotient between children who showed improvements in the age-equivalents learning rate of the Language subscale. The responders group started with a mean Personal–Social Quotient of 76.94 (SD = 16.81), while the non-responders group had a mean of 62.14 (SD = 21.80) at T1.

Finally, considering the age-equivalents learning rate in the Performance subscale, children who improved in time (responders) started with a mean Personal-Social Quotient of 77.19 (SD = 16.98), whereas non-responders group had a mean quotient of 61.95 (SD = 21.56) at T1. The difference was significant (t(35) <sup>=</sup> <sup>−</sup>2.33; *<sup>p</sup>* <sup>&</sup>lt; 0.05; r<sup>2</sup> <sup>=</sup> 0.13).

### **4. Discussion**

Given the complexity of evaluating treatment efficacy and the importance of individualized treatment for children with ASD, the main purpose of the present study was to analyse developmental trajectories of preschool children with ASD in order to understand how specific developmental areas evolve in time. As a way to do so, we took into consideration two groups of children exposed to two different kinds of intervention. On the one hand, an intensive intervention focused on the involvement of family with a specific work on wide-range socio communicative abilities and on the other hand, a rehabilitative "as usual" intervention. The results of the empirical research underline how early intensive intervention with parent involvement promotes better results and generalization of a child's competencies [4,5].

Regarding our first aim concerning the differences in the trajectories, our results are in line with the previous literature [5,6,10,22]. In fact, a significant improvement in the general quotient of children exposed to the early intensive intervention emerged, compared to children receiving the rehabilitative "as usual" intervention.

In particular, analysing the specific subscales, it came to light that linguistic-communication abilities present major improvements compared to the other subscales of the general quotient for both groups. In fact, the significantly increased level in the control group is not surprising given that specific rehabilitative activities provided also by local community services improve child linguistic abilities, especially considering both receptive and expressive language. In line with this, the recent literature, using a different measure for investigating the general quotient (Mullen Scales of Early Learning, Communication and Behavior Scales), reported major improvements in linguistic and communicative areas, particularly in both expressive and receptive language after 9 months [45].

Further, our results support the ground idea of developmental models of treatment for ASD that, unlike specific rehabilitative speech therapy-centred treatment, focus on wide-range socio communicative abilities. Developmental models of intervention [4,43] are based on the exploitation of communicative nonverbal behaviours, gestures and their integration together with intentionality and reciprocity to promote the development of language skills through generalization and reduction of avoidance of social interactions. Interestingly, in our intervention group, the mean difference in language skills between the two assessments was greater than the mean difference of the control group. One possible explanation of this result derives from theoretical principles of the intervention that focus on developmental phases with the major aim of promoting intersubjectivity during the exchanges with the other (e.g., supporting non-verbal communication and the correct understanding of social signals). To this end, intentionality and reciprocity are promoted given their importance for language development [40,41]. Further, these results could also be explained by the specific features of the intervention proposed. In fact, the intervention design is aimed to impact the most possible different contexts in the daily living of the child, and greatly extends the possibilities to experiment effective social interplays in a wider range of contexts. In our idea, participating at a major numbers of more appropriate social interactions could lead to better outcomes for children.

From the analysis of the cognitive profile, a significant increase in eye-hand motor coordination for the "as usual" intervention group also emmerged. In fact, a possible explanation could be that rehabilitative interventions such as psychomotricity comprise focused and specific motor activities, involving both gross and fine motor skills. From a clinical point of view of integrating different modalities in order to reach major outcomes, it is important also for networking interventions to comprehend rehabilitative activities to support these aspects.

Concerning the socio-communicative area, that is our second hypothesis, our analysis shows that the general behavioural expressions of ASD decreased significantly in both groups. In fact, some atypical behaviours tended to diminish after the intervention. In particular, children showed improved competencies in the socio-communicative area [9]. These gains were more prominent in the early intensive intervention group, probably thanks to active involvement in the social context that guarantees a generalization of competences. Furthermore, in line with the literature, the area of

restrictive and repetitive behaviours tends to be more stable. In fact, previous studies did not find significant modifications regarding this area after intervention [45,53]. Interventions generally support specific cognitive and social abilities that do not directly impact the area. Furthermore, the specific trends of this domain appear to be under-investigated [54]. However, slight modifications in this area, like the reduction emerged in the intervention group in our results, could be related to the specific work on anxiety reduction, emotions and self-regulatory mechanisms.

Taken together, these results highlight how specific work on a wide range of socio communicative abilities could promote better linguistic gains together with a reduction in symptoms severity with respect to the Social Affect area of the ADOS-2. Interestingly, this area of the ADOS-2 focuses on communicative abilities and social affect, considering different modalities and their integration. These results support the idea that intervention impacts developmental trajectories improving a large spectrum of socio communicative abilities, including receptive and expressive communication but also those important precursors of verbal communication like gestures, imitation and joint attention, fundamental elements to initiate or respond adaptively to the social exchange.

There is great consensus regarding the importance of cognitive level as a prognostic factor considering the developmental trajectory of children with ASD. References [38,39] also pointed out that children with cognitive level equal or above 70 points at the general quotient tend to improve more rapidly over time. In line with this, cognitive abilities are associated with different outcomes. For example, [19] found out that only children without impairment gained significant improvement in adaptive skills after 2 years of treatment, compared to children with intellectual disability. Further, only the first group of children was able to transfer the acquired socio-communication abilities into daily life after 1 year of treatment, showing generalization of competencies. On the contrary, this was not found for children with intellectual impairment. In line with these findings, our results show that children without intellectual impairment in the intervention group reached major gains in the general quotient after intervention. Particularly, the same pattern emerged considering the linguistic quotient, in which children without intellectual disability in the intervention group showed major improvements compared to the other group. With regard to children with cognitive impairment, no differences in both early intensive intervention and "as usual" groups were found.

Another key aspect focusing on developmental trajectories of symptom severity revealed that children without intellectual impairment show a more relevant increase in socio-communicative competencies compared to children with intellectual impairment [38,39]. In line with the analysis considering symptomatology of ASD, our results point out two different trajectories in the group exposed to early intensive intervention with parent involvement: less variability in symptoms expression was found considering children with cognitive impairment, and more gains were found regarding children without intellectual disability. With respect to the group exposed to the early intensive intervention, we found a specific trajectory that characterized children without intellectual impairment: increased level of cognitive abilities, specifically concerning linguistic skills, and reduced levels of symptoms expression. This specific outcome profile was coherent with one specific trajectory defined by [38].

A debate is still open on the identification of pre-treatment variables associated with different response outcomes.

With respect to our third aim, chronological age at the beginning of the intervention had an important role in predicting developmental outcomes, strongly supporting the idea of early intervention with children with ASD. Further, the analysis of pre-treatment variables pointed out the personal–social and performance areas as important predictors of the general quotient after intervention. In our analysis, younger children with better nonverbal intelligence skills, assessed by the Performance subscale, and personal autonomies (assessed by the Personal-Social subscale), showed better developmental outcomes. To our knowledge, no previous studies investigated the relationship between different domains of development and subsequent outcomes. Interestingly, our results highlighted the association of two specific developmental areas as possible prognostic markers of better developmental trajectories.

A wide consensus is present concerning chronological age, supporting early intervention [6,23]. However, the relation with cognitive functioning appears to be more complex, with controversial evidence. On the one hand, lower cognitive skills are found to be associated with larger improvements [55], pointing out the possibility of substantial improvements for children starting below the average. On the other hand, other authors found out that higher cognitive skills predicted better outcomes on child developmental trajectory [39,56], suggesting a complex relationship that needed to be further investigated. More interestingly, sub-components of the general intelligence were investigated to identify markers in the neurodevelopmental profile and early neurodevelopmental milestones that could predict later cognitive functioning and the acquisition of language [40].

With the aim to deeply analyze developmental domains and given the significant improvement concerning language skills in our results, we focused the analysis on the Language Quotient after intervention, showing that pre-treatment language skills and personal-social abilities, together with age, predict better linguistic outcomes. This could underlie how, in the development of language, an important role is played by nonverbal communicative aspects [57]. In fact, the Personal–Social subscale investigates the development of a wide range of nonverbal communicative and social signals (e.g., social smile, showing, orienting the others' gaze and communicative gestures) whereas in the Language subscale, besides the verbal skills, another set of communicative behaviours (e.g., pointing) are investigated, supporting the idea that the association between these two factors could represent possible prognostic markers specific for language development.

Taken together, and in line with other recent research works [40], these results seem to support the impact of wide-range of socio communicative behaviours and skills on developmental trajectories, regarding both the general cognitive skills and, more specifically, on language development [58]. Further, despite previous research depicting the role of symptom severity on intervention outcome, our analysis suggests that developmental areas were more predictive of outcomes than symptom severity before the intervention [26,40].

On the basis of these results, the analysis of responders focused on differentiating children who recovered in age-equivalents, narrowing the gap between their chronological and mental age, from children who seem to remain more stable. Interestingly, the responders showed a higher cognitive functioning before intervention and, in particular, greater language skills, coherently with our previous results. Furthermore, children who narrow the linguistic gap started with higher personal–social abilities and, interestingly, children who closed the performance gap also started with higher personal–social abilities. These results highlight the role of some cognitive factors (in particular, personal–social skills) not only in predicting outcomes after intervention but also in differentiating children who showed significant recovery from those characterized by more stable response trajectories.

Finally, concerning the trajectory of symptoms severity, our results evidence a significantly higher proportion of children who showed a reduction in symptomatology in the intervention group. Unexpectedly, a significant difference in restricted and repetitive behavioural pattern before intervention emerged between children who show a better recovery in personal–social skills, being characterized by lower symptom severity, and children who show a more stable outcome in this cognitive domain. This result may point out a potential role of this area in supporting or impeding the development of personal–social abilities and require further investigation in order to better understand its impact on the developmental trajectory.

This knowledge may have important implications for clinical practice, providing clinicians more information about specific areas to be targeted by the intervention and disclosing the importance of specific behaviours for subsequent language outcomes.

### *Limitations*

This study presents some limitations. First of all, despite our results being in line with previous literature, a main limitation of the present work is represented by the small sample size, and hence, results should be replicated in studies with larger samples. Further, sample size is important with respect to the high variability reported in the literature concerning different response trajectories. A small sample size reduces the possibility to investigate clusters of response profiles [39]. Moreover, the sample is unbalanced with respect to gender, thus reducing the possibility to investigate gender differences in the response trajectory, as emerged by recent literature [59]. In addition, our sample was not randomized. However, our aim was to understand intervention outcomes guarantying to patient better opportunities with respect to the specific intervention offered by the local territories. Children were assessed by independent examiners that were aware of their local origin but blinded to this study and not involved in children's therapeutic intercourse. The presence of only two assessments represents a limitation in order to better evaluate the response trajectory. Thus, an additional point to address in our further studies will be to measure children's developmental profiles in other time points in order to trace the response during time evidencing improvements and tendencies towards the stabilization of the profiles. Another future perspective is represented by a detailed analysis of specific socio-communicative elements evaluated by the ADOS-2. As an example, social affect behaviours such as pointing, showing and quality of social overtures could be important markers of change to be investigated, as pointed out by some research results [58], and could play a role in the response. Finally, characterizing children who narrow the gap and those displaying more stable trajectories could better inform about prognostic markers associated with better outcomes. In addition, it could disclose new features to be taken into account in order to explain the variability in the response and improve developmental outcomes of more persistent profiles.

### **5. Conclusions**

Identifying early trajectories of children with ASD has both theoretical and clinical implications. From a theoretical perspective, it can inform our understanding of early predictors and mediators of change in order to identify specific elements to be targeted in the intervention design. Further, this type of perspective enhances knowledge about ASD according to a developmental perspective.

From a clinical standpoint, careful attention to developmental trajectories may help in structuring individualized intervention based on a child's specific competencies in every phase of development. Finally, it is important to emphasize the fundamental role of social context in order to guarantee generalization of child competencies and better outcomes over time.

To conclude, the importance of networking intervention on child cognitive and social development led us to exploit online technologies in order to support social context through regular meetings to build up a valid online network.

**Author Contributions:** Conceptualization and Methodology: P.V., A.B. Formal Analysis: G.B. Patient recruitment: A.B. Data Curation: G.B., S.P. Writing—Original Draft Preparation: G.B., S.P., A.B. Writing—Review and Editing: P.V., G.B., S.P., A.B. Supervision: P.V. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** We gratefully acknowledge the families participating in our research and all the clinical psychologists and psychotherapists of ODFLAB.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **The Source of Palm Orientation Errors in the Signing of Children with ASD: Imitative, Motoric, or Both?**

### **Aaron Shield 1,\*, Megan Igel 1, Kristina Randall <sup>1</sup> and Richard P. Meier <sup>2</sup>**


Received: 20 February 2020; Accepted: 28 April 2020; Published: 30 April 2020

**Abstract:** Palm orientation reversal errors (e.g., producing the 'bye-bye' gesture with palm facing inward rather than outward as is customary in American culture) have been documented in the signing of deaf and hearing children with autism spectrum disorder (ASD) and in the imitation of gestures by signing and non-signing children with ASD. However the source of these unusual errors remains opaque. Given that children with ASD have documented difficulties with both imitation and motor skills, it is important to clarify the nature of these errors. Here we present a longitudinal case study of a single child with ASD, a hearing, signing child of Deaf parents. Samples of the child's signing were analyzed at ages 4;11, 6;2, 10;2, and 14;11. Lexical signs and fingerspelled letters were coded for the four parameters of sign articulation (handshape, location, movement, and palm orientation). Errors decreased for handshape, location, and movement after age 4;11, but increased on palm orientation from 4;11 and remained high, exceeding 55% of signs by 14;11. Fingerspelled letters contained a large proportion of 180-degree reversals, which suggest an origin in imitation differences, as well as midline-facing errors, suggestive of a motor origin. These longitudinal data suggest that palm orientation errors could be rooted in both imitation differences and motoric difficulties.

**Keywords:** autism spectrum disorder; sign language; imitation; cognition; language acquisition

### **1. Introduction**

We previously presented the first report [1] on an aspect of the language development of native-exposed signing children with autism spectrum disorder (ASD). In that paper, we showed that three young children with ASD who had been exposed to American Sign Language (ASL) from birth by their deaf parents exhibited an unusual formational pattern in their signing: the reversal of the palm orientation parameter, such that signs normally produced with an outward-facing palm were produced with an inward-facing palm, or vice versa. Since such errors are not known to occur in the typical development of ASL beyond an early age, we speculated that such reversals could be unique to signing children with ASD and as such might be included in clinical criteria adapted for sign-exposed children. Interestingly, to-date signing children with ASD have not been found to produce pronoun reversals [2] like those characteristically found in the speech of some hearing children with ASD [3–5] as well as very young typically-developing hearing children [6–8], raising the possibility that the documented palm reversals could be a sign language analog to pronoun reversals in speech—that is, errors that occur due to the child's difficulty understanding how linguistic forms shift between speakers/signers.

These palm reversal errors have thus provided an opportunity to speculate about the kinds of cognitive, linguistic, or motoric differences that might underlie their production by signers with ASD. At the time of our initial report, we followed the interpretation used in a review of 21 studies of imitation by children with ASD [9] which described difficulties with "self-other mapping", the translation of others' movements onto one's own body. Particularly strong evidence of this interpretation came

from a number of studies [10–13] which had found reversal errors in gesture imitation by hearing children with ASD that were of the same type as those we later documented in signing children with ASD. Although the errors we previously reported [1] were not errors of imitation, but rather of sign production, both elicited and spontaneous, we found it plausible that differences in imitation style could contribute to erroneous phonological representations of signs, thus accounting for the reversed palm orientation parameter in sign language production. We later elaborated on this hypothesis [14], describing a "visual matching strategy" in imitation that is characteristic of some learners with ASD, in which signs are imitated as they appear from one's own perspective, resulting in palm orientation reversals and other erroneous sign productions, such as reversals of the direction of movement.

Despite the reasonable conjecture that such errors could be the result of an imitation difference, motor issues cannot be excluded as a possible cause of palm orientation errors. From 50 to 80% of children with ASD exhibit motor impairments [15–18], including basic motor skill deficits in reaching and walking [19,20], gross and fine motor incoordination [15,17,21], as well as deficits in praxis/motor planning [22–26], and such deficits have been found to extend to deaf, signing children with ASD [27]. Children with motor issues with the articulation of signs might execute signs with the palm facing the midline of the body, which is the default resting position of the palm when the arms are hanging at one's sides. Producing inward- or outward-facing palm orientations requires the supination and pronation of the forearm, respectively. The ability to pronate and supinate the forearm develops throughout early childhood, with about 90% of typical children reaching mastery by age 6.5 [28]. Signers with motor disorders resulting from Parkinson's Disease have been shown to neutralize the palm orientation parameter by producing signs toward the midline rather than with inward or outward orientation as a result of reduced motoric effort [29].

Given that children with ASD have documented difficulties with both imitation and motor skills, it is important to clarify the nature of the unusual sign articulation errors that we have documented in signing children with ASD. In particular, longitudinal data on the developmental trajectory of such errors in comparison with the other articulatory parameters of sign could be illuminating. In this regard it is possible to make predictions about what the developmental trajectory of articulatory parameters would look like under two hypotheses:

*Motor origin hypothesis*: Motor difficulties are predicted to result in the palm facing the midline (default resting position) rather than outward or inward (pronated or supinated). Furthermore, if motor issues are the sole or primary cause of palm orientation errors, then the error rate in palm orientation is predicted to: (a) mirror that of the other sign language parameters (handshape, location, and movement) and (b) decrease over time as motor skills improve.

*Imitation hypothesis*: Differences in imitation are predicted to result in 180-degree reversal errors (signs specified for outward orientation produced with inward-facing orientation and vice versa); see Figure 1. Furthermore, if differences in imitation are the sole or primary cause of palm orientation errors, then the error rate in palm orientation is predicted to: (a) diverge from that of the other articulatory parameters (which are less affected by the visual matching imitation style), and (b) could remain relatively stable over time, as imitations solidify into mental (phonological) representations.

(**a**) (**b**)

**Figure 1.** (**a**) Example of how the fingerspelled letter t is typically produced; and (**b**) How the fingerspelled letter t would be imitated with 180-degree reversal.

A large number of the palm orientation reversal errors documented in our prior report [1] were produced on fingerspelled letters rather than on lexical signs: we reported 50 reversal errors on fingerspelled letters and five reversal errors on lexical signs. Fingerspelling is a system whereby the written alphabet of a spoken language is represented by different hand configurations. The fingerspelling system in ASL is one-handed; that is, each letter of the written alphabet is represented by a unique hand configuration (see Appendix A). Signed languages differ from each other in how they represent written alphabets as well as in the extent to which fingerspelling plays a role in the larger signed language. It is conventionally understood that the American Deaf community employs fingerspelling to a greater extent than in most other Deaf communities around the world [30].

Fingerspelling is most often used for proper names or for technical or novel terms for which a conventional lexical sign is lacking. Unlike lexical signs, which only employ one or two different handshapes [30], fingerspelling requires the signer to execute a series of different handshapes, one for each letter of the word being spelled. Lexical signs can be specified for different locations from the head to the waist or can be made in neutral space (e.g., the sign mother on the chin, the sign father on the forehead), [Links to video examples from the SignBank database [31] are provided for all lexical signs in the online version of the paper.] In contrast, fingerspelling in ASL is performed in a relatively small neutral space in front of the signer's torso. Most fingerspelled letters are static handshapes without movement, with the exceptions of the letters j and z (Appendix A), while lexical signs draw from an extensive set of possible movements. Finally, while palm orientations of lexical signs can be specified to face up, down, toward the midline, to the sides, or toward or away from the signer's body, fingerspelled letters in ASL all face outward from the signer's body with a pronated forearm, with the exception of the letters g, h, p, and q. The letters g and h face inward, with the forearm rotated inward (supinated), while the letters p and q face downward, with a flexed wrist and pronated forearm (Appendix A), though note that there is a variant production of p with only very slight flexion of the wrist and supination of the forearm, resulting in inward palm orientation [32], but the participant in this study did not produce any tokens of this variant.

In the sections that follow we distinguish between lexical signs and fingerspelled letters and analyze them separately. We do so for the following reasons: (1) we observed a large number of fingerspelling errors in our previous work [1]; and fingerspelled words (2) require the execution of a series of hand configurations in sequence; (3) are uncomplicated by changes in location; (4) are largely uncomplicated by changes in movement; and (5) present frequent opportunities for 180-degree palm reversals given their specification for outward-facing palm orientations.

This study presents a longitudinal case study of a single native signer with ASD, a hearing son of two Deaf parents, and analyzes the four articulatory parameters of his signs over a 10-year period, in order to shed light on the nature of palm orientation errors in ASD.

### **2. Materials and Methods**

The parents of the participant gave their informed consent before including their child in the study. The study was conducted in accordance with the Declaration of Helsinki, and procedures were prospectively approved by the Institutional Review Board of the University of Texas at Austin (at ages 4;11 and 6;6; Protocol #2007-08-0022), Boston University (age 10;2; Protocol 2471E) and Miami University (age 14;11; Protocol 01375).

### *2.1. Participant*

The child described here was one of the three natively sign-exposed children with ASD described previously [1]; in that work he was referred to as "Child 3". He is a left-handed hearing male, diagnosed with ASD at age 2;6 by a licensed clinical psychologist. He has two Deaf parents who communicate primarily through ASL and a younger hearing brother. His parents indicated that he has received occupational therapy for low muscle tone affecting his fine motor skills. (While handedness is certainly a relevant factor in considering how children might imitate signs [14,33], the palm orientation parameter is unaffected by handedness. Therefore the child's left-handedness is not considered further in our analyses.)

In addition to the data collected at age 6;6 reported previously [1], we visited the child at three different times over the course of ten years: at ages 4;11, 10;2, and 14;11. Over the course of those ten years we collected a number of standardized measures of language (both English and ASL), nonverbal intelligence, and ASD; these are reported below. He exhibits moderate ASD symptoms, by behavioral observation (ADOS-2) and by parent report (SCQ and AQ-Adolescent). He scores in the impaired range on nonverbal intelligence (TONI-4) and on receptive language for English (CELF-5; PPVT-4) and ASL (ASL RST).

### 2.1.1. Autism

The Autism Diagnostic Observational Schedule, Second Edition (ADOS-2; [34]) was administered at age 9;11 by a clinician who had attained research reliability on the instrument and was fluent in English and ASL. The child's total score of 15 (corresponding to a severity score of 6 on a scale of 1–10) indicated moderate ASD symptoms and was above threshold for autism classification. His mother completed the Social Communication Questionnaire (SCQ; [35]), Lifetime Form, at 10;2 and 14;11; his total score was above threshold for ASD risk at both ages (raw score of 14 at 10;2 and raw score of 16 at 14;11). Additionally, his mother completed the Autism Quotient (AQ; [36]), Adolescent Version, at 14;11. His score of 34 was above the threshold score for ASD of 32.

### 2.1.2. Intelligence

We administered the Test of Nonverbal Intelligence, Fourth Edition (TONI-4; [37]) at age 10;2 and 14;11. At 10;2 he achieved a raw score of 6, which translates into a standard score of 69, just under 2 SD below the mean. At 14;11 he achieved a raw score of 25, corresponding to a standard score of 86, or 17th percentile for his age and an age equivalent of 9;0.

### 2.1.3. Language

Our participant is the bimodal bilingual child of Deaf parents. It is important to note that there is no established profile for bimodal bilingual children exposed to both a signed language and a spoken language, such as the hearing children of Deaf adults [38]. However, hearing children of Deaf adults typically have speech that is equivalent to monolingual hearing children by about age 7 [39]. At age 6;6 our participant's mother filled out the Language Proficiency Profile, Second Edition (LPP-2; [40]), a parent report measure to estimate global communication skills. His total score of 26 indicated language well below his chronological age. At age 10;2 we administered both the Peabody Picture Vocabulary Test, Fourth Edition (PPVT-4; [41]) and the American Sign Language Receptive Skills Test (ASL RST; [42]) to obtain measures of his receptive skills in English and ASL. He obtained a standard score of 46 on the PPVT-4 (1st percentile), corresponding to an age equivalent of 4;1. On the ASL RST, he obtained a raw score of 6, corresponding to an age equivalent of under age 3, the youngest age for which norms on this test are given. At 14;11 we repeated the ASL RST and added the Receptive Language Index subtests of the Clinical Evaluation of Language Fundamentals, Fifth Edition (CELF-5; [43]). On the ASL RST he achieved a raw score of 12, corresponding to a standard score of 71, about 2 SD below the mean. On the CELF-5, he achieved a standard score of 45, more than 2 SD below the mean.

### 2.1.4. Prior report

In our previous report [1], we described data collected from the child when he was age 6;6. At 6;6, the child produced 59 signs, of which 35 (59.3%) contained one or more articulatory errors. For a summary of the child's articulation errors at that age, see Table 1.


**Table 1.** Articulation errors previously reported at age 6;6.

### *2.2. Procedure*

The child was observed at home at all four time points. At age 4;11, he was observed in an unstructured, naturalistic interaction with his Deaf father, who engaged with him while reading to him from a picture book. At 6;6 and 10;2, he was observed interacting with the first author, a hearing researcher fluent in ASL, who performed a series of experimental tasks, including eliciting fingerspelled words and lexical signs. At 14;11, he was observed interacting with his Deaf mother, who asked him a series of questions in ASL about friends, school, books, and movies.

### *2.3. Coding*

Using ELAN (EUDICO Linguistic Annotator) multimodal coding software [44], we coded 12 continuous minutes from each time point (ages 4;11, 6;6, 10;2, and 14;11) for all signs produced. For age 6;6, previously reported [1], we coded a new 12-minute span of video. Each letter of a fingerspelled word was coded and counted as an individual sign. Each sign was coded for handshape, location, movement, and palm orientation (inward, outward, upward, downward, or midline-facing). The coded value for each parameter was scored as being produced correctly or as an error based on standard citation forms; we used the ASL Signbank as a reference (see https://aslsignbank.haskins.yale.edu/) [31]. Where movement segments were deleted, resulting in a missing second location, errors were coded as movement errors only. Errors were qualitatively described so as to allow for further analysis.

### *2.4. Reliability*

To ensure the reliability of the coding system, two 5-minute segments (one from age 4;11 and one from 14;11) were blindly recoded by a second trained coder experienced in the coding of ASL. Differences in coding were discussed by both coders and disagreements were resolved through consensus. The main coder then adjusted the rest of the coding to reflect the decisions made through consensus discussion with the second coder.

### **3. Results**

Table 2 presents a comparison of the overall number of signs produced and the number of signs produced per minute. Overall sign production increased over time, although note that we have counted individual fingerspelled letters as separate signs. Importantly, the child's fingerspelling increased over time, which could account for the greater number of signs produced, especially at 14;11. This increase in fingerspelling is in line with other reports of the developmental trajectory of fingerspelling, which have shown that fingerspelling to children by adults increases as the children mature, and the fingerspelling produced by such children increases in turn [30].


**Table 2.** Comparison of quantity of signs produced and error rates across time points

Table 3 presents the total number of lexical signs and fingerspelled letters produced at each time point, and the number of errors on each of the four sign parameters produced for both types of signs. Note that fingerspelled letters are all produced in neutral space and generally do not exhibit movements, except for the letters j and z (see Appendix A); thus location and movement errors are unlikely on fingerspelled letters.


**Table 3.** Errors on lexical signs and fingerspelled letters at each age, classified by parameter.

Figure 2 shows the child's error rates on the four sign articulation parameters across the four time points, collapsing lexical signs and fingerspelled letters. Error rates for handshape, location, and movement decrease over time, while the palm orientation parameter shows the opposite trend, increasing to an error rate of over 50% at age 14;11. By comparison, studies of the early acquisition of phonological parameters in ASL have found that the handshape parameter is the most error-prone early in development, while location is acquired earliest, as appears to be the case for this participant at age 4;11. Most studies of phonological development in ASL have primarily focused on children who are much younger than the participant in this study, i.e., under age 2 [45–47].

**Figure 2.** Proportion of signs exhibiting errors in the four sign parameters at four ages.

We distinguished three different types of palm orientation errors: 180-degree reversals (substitutions of inward palm orientation for outward or vice versa), midline errors (neutralization of the palm orientation parameter such that the palm faced toward the midline rather than inward, outward, up, or down), and other errors (e.g., substitutions of an upward- or downward-facing palm for inward or outward). Table 4 reports the frequency of each error type of error at each age.


**Table 4.** Palm orientation errors by type at each age.

Looking across all of the palm orientation errors produced in our sample, 159 of 182 errors (87.3%) were produced on fingerspelled letters while the remaining 23 errors (12.6%) were produced on lexical signs. We report all fingerspelled letters in Table 5 below. Note that a number of fingerspelled names produced at age 14;11 were redacted to protect the participant's identity. In these redacted fingerspelled names, the participant produced 7 names: four 5-letter names and one 4-letter name with all letters produced facing the midline; one 4-letter name with all letters reversed, and one 4-letter name with the first three letters reversed and the last letter produced with correct outward orientation. There were no instances of g, h, p, or q in these names, so the target orientation for all letters was outward.

It is clear from Table 5 that the child produced fingerspelled letters with inconsistent palm orientation. Indeed, he produced certain fingerspelled handshapes with different palm orientations during the same session (e.g., with in/mid/outward palm orientation on e and o at 6;6 and d, a, and y at 14;11) and varied the palm orientation of fingerspelled letters at different ages (e.g., l outward at 10;2 but inward and midline-facing at 14;11).

In order to understand the inconsistency exhibited in palm orientation, we examined how palm orientation errors occurred within fingerspelled words. First, some words maintained a consistent palm orientation, be it correct (outward) as in #door. at 6;6, or incorrect such as the midline orientation exhibited in #swing at 14;11 or reversed (inward) as in the #yoda example illustrated in Figure 3. [As is conventional in the literature, fingerspelled words are denoted by a preceding pound sign (#)]. However, we also found instances in which the child switched between (correct) outward palm orientation and (incorrect) inward palm orientation within the same fingerspelled word. Words that follow this pattern of inconsistency include #teach, #phone, #mother, and #father (Figure 4) at 10;2. Recall that all fingerspelled letters are typically produced with the palm facing outwards (with pronated forearm) except for g and h (which face inward with supinated forearm), and p and q (which face downward, with pronated forearm and flexed wrist). This difference in specification for palm orientation means that in words that contain these four letters, the signer must switch between outward, inward, and downward palm orientations in the course of normal signing, which requires the pronation, supination and re-pronation of the forearm (as well as wrist flexion for p and q). The child in this study also produced words without errors in which he successfully switched between inward, outward, and downward palm orientations, such as the word #telephone at 6;6, where p and h were produced with correct downward and inward orientations, respectively, and all other letters with correct outward orientation (though note the substitution of i for l). Other examples of this include the words #school (produced without the c), #girl, #chair, and #bug produced at 10;2. However there are also examples in which the child produced a reversed palm orientation on letters that are adjacent to h. Examples that follow this pattern include #chair at 6;6, #teach, #phone, #mother, and #father (Figure 4) at 10;2, and #school and #theincredibles at 14;11.

**Table 5.** Fingerspelled letters produced at each age. Letters produced with correct outward palm orientation (all letters except g, h) are represented in plain font, letters produced with correct inward palm orientation (g, h) are underlined, while letters that exhibited 180-degree reversals are bolded and letters produced with midline errors are italicized. Some tokens contain spelling errors produced by the child, e.g., the substitution of i for l.


**Figure 3.** The fingerspelled word #yoda produced with reversed, inward palm orientation on each handshape at 14;11. The word was produced rapidly and fluently, unlike the labored production of #father in Figure 4.

**Figure 4.** The fingerspelled word #father produced on the left hand at 10;2 with correct outward palm orientation on the letter f, mid-facing orientation on the letter a, correct outward palm orientation on the letter t, correct inward orientation on the letter h, incorrect reversed palm orientation on the letter e, and correct outward palm orientation on the letter r. Note the lack of inhibition of movement of the non-signing right hand, indicative of a lack of motor control.

### **4. Discussion**

We have documented the development of the four parameters of sign articulation over a period of ten years in a single child with ASD, a natively sign-exposed hearing child of two Deaf parents. This is the first time that the sign development of a native signer with ASD has been studied longitudinally. We had hypothesized that the palm orientation errors documented previously [1] could have imitative or motoric origins, and that the developmental trajectory of the palm orientation parameter, in comparison with the other parameters of sign language development, could shed light on this question. Here we evaluate the evidence for both hypotheses.

Is there evidence in favor of the motor origin hypothesis? Yes. The strongest evidence is the occurrence of palm orientation errors produced toward the midline rather than inward or outward. Such errors accounted for 47.8% of the palm orientation errors in our sample, occurred at all ages studied, and reflect the neutralization of the palm orientation parameter toward a default resting position [29]. The fact that errors on the three other parameters (handshape, location, and movement) decrease over time suggests a developmental trajectory of improvement in motor skills that does not extend to palm orientation; in particular, the handshape parameter, which has the highest error rate at age 4;11, decreases nearly to zero by 10;2, and remains stable at 14;11. Numerous studies have found that, of the three major parameters of handshape, location, and movement, handshape is the parameter that is mastered latest in typical development [45,46,48–53], probably due to the late development of the fine motor control required to produce handshapes accurately (though note that not all of these studies examined palm orientation as a separate parameter).

A second source of evidence in favor of the motor hypothesis is the fact that the child sometimes reversed palm orientation on letters that were adjacent to the inward-facing letters g and h. This suggests that the child anticipated the switch in palm orientation on a subsequent letter (as in the c in #chair at 6;6 or the t in #mother at 10;2), or failed to reorient his palm to face outward (i.e., re-pronate the forearm) following one of the inward facing letters (as in the o in #phone, the e in #father, and the e in #mother at 10;2). Additional evidence of motor control issues include the lack of inhibition of the non-dominant hand shown in Figure 4, and the unusually high signing shown in Figures 3 and 4.

Given the evidence for motor impairment causing palm orientation errors, is there also evidence in favor of the imitation hypothesis? Here, too, the answer appears to be yes. Unlike the other parameters of sign formation (handshape, location, and movement), which show a clear decrease in error rate over the ten-year period, palm orientation errors increase over time, to above 50% at age 14;11, and reversal errors made up nearly half of all palm orientation errors documented in this study (82 out of 182 errors; 45.1%). Particularly striking are fingerspelled words that do not include the letters g and h but which were produced with consistently inward palm orientation, as in #park, #paris, and #yoda (Figure 3) at 14;11. It is unlikely that such 180-degree reversal errors would result from motoric difficulties, since the supination of the forearm entailed in the production of inward palm orientations is as motorically difficult to execute as the pronation of the forearm entailed in outward palm orientations. Instead, these reversal errors are suggestive of differences in imitation during the sign learning process in which the child reproduces what he sees from his perspective ("visual matching"), yielding forms with reversed palm orientation. It is worth noting that most of the 180-degree reversal errors described here involve the substitution of an inward-facing palm orientation (supination) for an outward-facing palm orientation (pronation); 75 of the 82 (91.5%) 180-degree reversal errors described here fall into this category. We believe that this finding is again due to the fact that nearly all fingerspelled letters are typically produced with an outward-facing palm, and the child reported on here tended to reverse palm orientation on fingerspelled letters. Despite this trend, a minority of errors (7 of 82 or 8.5%) involved the substitution of an outward-facing palm for an inward-facing palm, showing that reversal errors can replace inward with outward palm orientations as well as outward with inward palm orientations. Other errors of this type, such as the production of the lexical sign butterfly with outward-facing rather than inward-facing palms, have been reported before [33]. More study is warranted to better understand which lexical signs could be subject to palm reversals of this type.

Why should palm orientation errors increase over time? For this question, too, there appears to be a clear answer: palm orientation errors surface most often in fingerspelling, and fingerspelling increases with age, as children become more literate and incorporate more English words into their vocabulary [30]. Indeed, fingerspelling accounted for 110 of the 112 (98.2%) palm orientation errors produced by this child at 14;11. Fingerspelled letters require the pronation of the forearm such that the signer's palm faces outward on all letters except for g and h (produced with supination such that the palm faces inward) and p and q (produced with pronation and wrist flexion such that the palm faces downward). The tendency to reverse palm orientation on fingerspelled letters was previously observed for a different child at age 7;5 [1], who produced 61 palm orientation errors, 50 of which were fingerspelled letters produced with inward palm orientation rather than outward. The other 11 errors were midline errors, confirming the patterns observed here: reversals and midline substitutions on fingerspelled letters. Similar to the case discussed here, this child produced a low rate of errors on the other sign parameters (6 movement errors, 1 handshape error, and 0 location errors out of 94 sign tokens), suggesting overall good motor control.

It is important to note that fingerspelling is typically directed toward an interlocutor. In this sense, palm orientation in fingerspelling could also reflect pragmatic competence: the signer must understand that their signing should be produced facing in the direction of their interlocutor. Typically-developing signing children do not produce reversal errors of this type on fingerspelled words; in our previous work no such errors were produced by a sample of 12 deaf children of deaf parents between age 3;7 and 6;9 [1], and to our knowledge there are no other instances of such errors in the literature. The idea that difficulties with pragmatics could underlie the palm reversals documented in the signing of children with ASD suggests a parallel between these errors and the pronoun reversals documented in the speech of hearing children with ASD, as the latter errors have been interpreted as evidence of challenges with understanding how discourse roles shift between interlocutors during conversation [6,54,55].

Despite the frequency with which palm orientation reversal errors occurred on fingerspelled letters, palm orientation reversals also occur somewhat infrequently on other types of signs, such as lexical signs. In our previous work [1], we noted the reversal of palm orientation from inward to outward on the lexical sign flashing-light, and outward to inward on the sign bye-bye, both produced by the child described here. In that work as well as in this study, we also find evidence of reversed palm orientation on number signs (which in ASL are formationally similar to fingerspelling), though these errors should be interpreted with caution since there is variability in the production of these signs depending on whether numbers are ordinal, cardinal, or part of a series such as a postal code [56].

One puzzling result is that the palm orientation value for individual signs was variable and unstable. That is, the same sign—especially the same fingerspelled letter—was produced with up to three different palm orientations, and this variability occurred both within the same session as well as across different sessions. If the child had a fixed mental representation of the palm orientation parameter for a given sign, then we would expect him to produce the sign with the same palm orientation value every time he produced that sign. We had hypothesized that differences in imitation style (such as the visual matching strategy, in which the child reproduces signs exactly as they appear from his perspective) could result in mental representations with incorrect palm orientation values [14]. Instead of a fixed but erroneous representation of palm orientation, we propose that the variability of input results in an *unstable* or *underspecified* mental representation of the palm orientation parameter. Children exposed to signs observe signs produced from various angles: whether facing the adult signer head-on, or from the side, or from behind a parent as that adult signs to others, or from every conceivable angle in between. This variability in sign input could result in an unstable value for the palm orientation parameter in the child's mental representation of the sign, or indeed no palm orientation value at all. As it happens, the palm orientation parameter may carry a low functional load compared to the other parameters that signs are composed of. Minimal pairs for palm orientation are signs which differ only in their palm orientation value, proving the phonological contrastiveness of the palm orientation parameter; e.g., children versus things [57]. Although minimal pairs for palm orientation do exist in ASL, they appear to be rare, especially compared to minimal pairs for location, handshape, and movement.

We must caution that these results may not be reflective of all signers with ASD. Indeed, ASD is characterized by its diversity of presentation, and this is true both in terms of language ability and motor skills. However, it appears clear that differences in imitation lead both hearing and deaf children with ASD to imitate gestures and produce signs in ways that are unlike typical children, and this paper argues that although motor difficulties appear to be an important factor in the production of palm orientation errors, motor impairment alone cannot account for all of the errors observed. We would not predict or expect, however, that all signing children with ASD would produce palm orientation reversals. Indeed, such reversals may occur within a subset of children whose language or ASD severity fits a specific cognitive profile, though such a profile has not yet been identified. It is worth noting that the child described here has significant intellectual disability. In contrast, the two children described in our previous report who produced palm orientation reversals did not exhibit intellectual disability [1]; both children were in the average range of intelligence but in the below-average range of language ability. It thus appears plausible that palm orientation reversal errors are unrelated to overall intelligence but could be linked to lower language abilities. A related issue is whether the differences observed could manifest in linguistic structures other than the phonological form of the sign (e.g., role-shift requiring the assumption of different perspectives, or various types of path movements entailed in agreement verbs). We have not observed either of these phenomena due to the overall low level of expressive sign language exhibited by this participant, but future research should investigate whether signing children with ASD experience difficulty with other aspects of the linguistic system that are rooted in motoric, imitative, or other cognitive challenges or differences.

Although the quantity of signs produced in the 12-minute segments increased over time (from 6.33 signs/minute to 16.42 signs/min), the differences in procedures at each time point do not permit direct comparisons. In particular, the increased fingerspelling produced at age 14;11 was largely responsible for the increase in total number of signs produced at the later age. Gains were also observed on the only language measure that was administered at two time points, the ASL RST (at ages 10;2 and 14;11), on which the child increased his raw score from 3 (age equivalent < 3 years) to 12 (age equivalent of approximately 4.5 years). Thus the increase in palm orientation errors occurred *despite* evidence of gains in both receptive and expressive language.

As palm orientation reversals have been documented in a variety of contexts (spontaneous signing, elicited signing, and in the imitation of gestures) and in a variety of populations of children with ASD (hearing children, deaf children, non-signers, and signers), we suggest that such reversals be considered a red flag for ASD diagnosis if they occur past the early developmental period. In particular, if diagnostic and screening instruments are adapted for signing children, we believe that it would be important to include items probing whether or not children produce palm reversals, as such errors rarely occur in typical development beyond the first two years of age.

### **5. Conclusions**

This study represents the first longitudinal study of a signer with ASD. It demonstrates that palm orientation errors (both 180-degree palm reversals as well as midline-facing errors) can persist beyond childhood and into adolescence. Such errors are notable for both clinical and theoretical reasons: they could serve as a modality-specific marker of ASD, and as such could be incorporated into adapted diagnostic and screening instruments for signing children, and they also provide insight into the mechanisms (both imitative and motoric) that lead to such errors. Future studies are needed to help clarify how frequently such errors occur in the population of signing children with ASD, and whether there is a specific cognitive profile of children who produce them. These studies will be crucial for a better understanding of why some children with ASD produce these unique errors, and what kinds of differences could lead to their production.

**Author Contributions:** Conceptualization, A.S. and R.P.M.; methodology, A.S. and R.P.M.; data coding and reliability, M.I. and K.R.; writing—original draft preparation, A.S.; writing—review and editing, A.S. and R.P.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by a Doctoral Dissertation Research Improvement Grant BCS-0746009 from the National Science Foundation to R.P.M. and A.S., Autism Speaks Predoctoral Fellowship #4721 to A.S., grant 1F32-DC0011219 from the National Institute on Deafness and Other Communication Disorders to A.S., and a Dean's Scholar Award from the College of Arts and Sciences and Miami University to K.R.

**Acknowledgments:** We thank the research participant and his family for their collaboration over the course of ten years. We also thank Reverb Art + Design for creating the illustrations in Figures 3 and 4.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

### **Appendix A**

**Figure A1.** ASL Fingerspelling Handshapes.

### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
