Participants and procedure
One hundred forty-two children with ASD participated in this study (de Bruin, Verheij, Wiegman, & Ferdinand, 2007; Eussen et al., 2015) of which 62 were selected for analysis in this article (see Fig. 1). They visited the outpatient clinic at the Department of Child and Adolescent Psychiatry/Psychology of Erasmus Medical Center – Sophia’s Children Hospital Rotterdam, Netherlands between July 2002 to September 2004. Inclusion criteria were: (1) meeting the clinical diagnosis of Diagnostic and Statistical Manual of Mental Disorders(DSM-IV-TR; American Psychiatric Association, 2000) classification of ASD and (2) parents being able to communicate in the Dutch language. The exclusion criteria in this study were (1) children aged below 7 years old because cut-off points of ILL and LA were not available for children younger than 7 years, (2) the presence of severe neurological or physical problems, such as epilepsy or muteness. The total group (n = 62) children had a mean age of 9.41 years (SD = 1.65 years). 84% of the participants were boys (n = 52). All children with complete data on Kiddie Formal Thought Disorder Scale (KFTDS), the Children’s Communication Checklist (CCC), and the first-order ToM task were included in the analyses.
The clinical diagnostic assessment was obtained by a multi-disciplinary team that consisted of a parental interview on the child’s early developmental history, medical history, and current functioning. Parents of the participating children additionally filled out a questionnaire, e.g., the CCC. The children also underwent psychological assessments e.g., the KFTDS and the ToM task (de Bruin, 2006). Parents of the participants had signed informed consent forms.
The Medical Ethics Committee of the Erasmus Medical Centre approved this study.
Kiddie-Formal Thought Disorder Rating Scale (KFTDS) was used to assess disorganized speech, quantified as the presence of loose associations (LA) and illogical thinking (ILL) (Caplan et al., 1989; Caplan et al., 2000). The validity of these KFTDS scales has been established in children with ASD (Solomon et al., 2008; Van der Gaag, Caplan, van Engeland, Loman, & Buitelaar, 2005). Children were asked to answer standard questions after listening to two audio-taped stories (i.e., “What did you like about this story?” or “Do you think this a true story?”). Children were then asked to make up his/her own story about one of four given topics (the incredible hulk, a witch, a disobedient child, or an unhappy child). This test took 20 to 30 minutes, and the child’s speech was audiotaped. All stories were scored according to the KFTDS guidelines (Caplan et al., 1989) by summing frequency counts for illogical thinking and loose associations. Total raw scores were corrected for the variability of speech by converting the number of utterances per minute. These continuous KFTDS scores were divided into different ordinal categories as falling above or below the cut-off points, in which a score above the cut-off point indicates a higher likelihood of pathology. Using cut-off points, continuous KFTDS scores become dichotomous as falling above or below the cut-off point. Cut-off points were not calculated for children younger than 7 years old because KFTDS is a reliable measure of disorganized speech in children ages 7–18 years old (Caplan et al., 1989; Caplan et al., 2000).
The inter-rater reliability for the total KFTDS score was good, with a kappa of 0.77 (Caplan et al., 1989). The kappa for ILL was 0.78 and for LA 0.71 (Caplan et al., 1989). The validity of the KFTDS has been established in children with schizophrenia spectrum disorders and ASD (Caplan et al., 2000).
In our sample of children aged 7–12 (n = 62), 71% of children scored above the cut-off point for ILL (n = 44), while 19.4% of children scored above the cut-off point for LA (n = 50). Based on these results, in the next step, we combine these into one integrated/combined grouping variable. A score above the cut-off point indicated a higher likelihood of disorganized speech. By using the cut-off point value, continuous KFTDS scores were dichotomized as falling above or below the cut-off point. Children with a score above the cut-off point for ILL and LA were assigned to the group ‘severe disorganized speech’, children with a score above the cut-off point for ILL and below the cut-off point for LA were assigned to the group ‘mild disorganized speech’, and children below the cut-off point of ILL and LA were assigned to the group ‘no disorganized speech’.
The rationale for dividing the children into three groups is based on the study by Eussen et al. (2015), who found that ILL predicted the severity of autistic symptoms in adolescence, while LA did not. Those findings indicated that children with ASD show ILL from early childhood onwards, while LA deteriorates when they reach the age of 7 years old. In our study, we included children aged between 7 and 12 years and found that 70% of the children scored above the cut-off point for ILL, while 76% of the children scored below the cut-off point for LA. Based on these results, we combined ILL and LA into one integrated/combined grouping variable, as having both LA and ILL is reflective of the presence of more types of disorganized speech than only having one or the other.
Social pragmatic communication difficulties
The Children's Communication Checklist – CCC (Bishop, 1998) was administered to the parents of the included children. The CCC assessed pragmatic aspects of a child’s social pragmatic communication difficulties. The questionnaire consists of 70 multiple choice questions on a 3-point scale (0 = definitely applies, 1 = applies somewhat, 2 = does not apply). The lower the score on the CCC, the more impaired the child is. The questions are divided into nine subscales; five subscales for social pragmatic communication skills (coherence, inappropriate initiation, stereotyped language, use of context, and conversational rapport), two subscales to measure formal language (speech production and syntax), and the remaining two subscales assess characteristic behaviors of ASD (social relationships and interests). In this study, we used five social pragmatic communication skills subscales. There are 38 social pragmatic communication items in total, with the sum of the five subscales ranging between 88 to 162. The CCC was translated into Dutch using a two-way translation procedure (Hartman et al., 1998). The internal consistency for the Pragmatic Composite Score was 0.76 (Geurts et al., 2009). Meanwhile, the internal consistency in our dataset is .80.
Social cognition problems
Theory of mind (ToM) was measured using a Dutch ToM task which was based on the classic Sally–Anne false-belief task (Baron-Cohen et al., 1985). A child was told the following story: Ben places a bone in a barrel and goes out to play; while he is gone, Max (Ben's brother) takes the bone from the barrel and hides it in another barrel. After telling this story, children were asked a control question (Where is the bone really?). A wrong answer for this question represents an insufficient understanding of the story and understanding is necessary to perform well on the first-order false-belief tasks. One child who answered this question wrong was excluded from the analyses. Then, children were asked a false-belief question (Which barrel will Ben choose?). The child was given a score of 1 for a correct answer and a score of 0 for the wrong answer.
Individual factors were taken into consideration as possible confounding variables. First, sex may have influenced the result because females were usually more verbally oriented (Walker, 2005). Despite both sex have similar social understanding difficulties, males showed less restricted, repetitive patterns of behaviors (Lai, Lombardo, Auyeung, Chakrabarti, & Baron-Cohen, 2015). Intelligence Quotient (IQ) was taken into account as a possible covariate based on Van der Gaag et al. (2005), who found a significant negative correlation between Total IQ with illogical thinking and loose association on the KFTDS. A lower IQ is also associated with more difficulties in social pragmatic communication (Schirvar, 2013) and false-belief task (Baren-Cohen, 1989). In this study, IQ was measured using the Wechsler’s Intelligence Scale for Children-Revised version (WISC-R). Finally, age was taken into account as a covariate because developmental aspects play a vital role in the manifestation of DS and therefore scores could be correlated with age. Age also affects social pragmatic communication (Norbury et al., 2014; Tager-Flusberg, Paul, & Lord, 2005) and ToM (Wimmer & Perner, 1983). The older the child is, the better his/her social pragmatic communication skills are, both for TD children and children with ASD (Simmons et al., 2014).
Of the 142 children recruited, 89 children had complete KFTDS data, while 53 children had incomplete KFTDS data. Six children were younger than 7 years and three children with high LA and low ILL scores were excluded from the analyses. The group of children with scores above the cut-off point of LA and below the cut-off point of ILL was too small to be included separately in the analyses. n = 18 children who did not have complete data on the outcome measures, i.e., SPCD and SCP, were also excluded from the analyses (Fig. 1).
Figure 1 Sample description
The effects of potential attrition (sex, age, and IQ) were checked for the group with complete data (N = 89) versus the group with incomplete data (N = 53). Then, the descriptive characteristics (sex, age, and IQ) of the children were provided for the total group (n = 62) and compared between three subgroups i.e., children with severe (n = 12), mild (n = 32) and no DS (n = 18).
Then, assumptions of normality, independence, and homogeneity of variance were checked for the CCC scores. Also, the assumptions of Fisher-Freeman-Halton Exact test were checked for the ToM data. Pearson correlations were used to estimate the influence of covariates (age and IQ) concerning the main variables (KFTDS groups, CCC and ToM). A Spearman correlation was performed to investigate the influence of sex to KFTDS groups, CCC and ToM. In case one of the covariates showed a significant correlation with both the predictor and the outcome variable, we included this covariate in the subsequent analysis of covariance (ANCOVA) and logistic regression analysis for ToM.
A one-way analysis of variance (ANOVA) was used to explore between-group differences for CCC. We examined whether the groups differed significantly on the subscales, i.e., inappropriate initiation, coherence, stereotyped language, use of content, and rapport (Aim 1). Post hoc pairwise tests were carried out to compare between two groups. In this study, we chose Games-Howell procedure as it remains accurate when sample sizes are unequal (Field, 2009). To examine whether disorganized speech influenced their first-order false-belief task (Aim 2), the study used a Fisher-Freeman-Halton Exact Test. We looked at the cross-tabulation and observed the number of children with disorganized speech pass the first-order false-belief task in each group.
In this study, a p-value of < .05 (confidence interval of 95%) is considered to be significant. The statistical analyses have been performed using IBM SPSS Statistics 21.