Comparison of Disfluent and Ungrammatical Speech of Preadolescents with and without ASD

The aim of this paper is to analyse disfluencies and ungrammatical expressions in the conversational speech of 11–13-year-old Finnish-speaking boys who have a high-functioning autism spectrum disorder (ASD) in comparison to age- and gender-matched controls. The participants with ASD were all diagnosed with Asperger Syndrome (AS). Our analyses were quantitative and qualitative: We first identified and measured the duration of speech including disfluencies and ungrammatical expressions, and then analysed the impact of disfluent speaking turns on interaction using conversation analytic methods. The aim of the quantitative analysis was to compare the proportion of speech disfluencies in the ASD group with that of their age-matching control group. The aim of the qualitative analysis, in turn, was to examine the types of disfluencies and ungrammatical expressions in more detail and to observe how other participants in the interaction reacted to them.

Both speech disfluencies and ungrammatical expressions interrupt the continuous flow of speech and may thus present challenges for conversational interactions. When speech is fluent, speech production flows smoothly both in terms of sound and information (e.g., Manning 2010). There are few interruptions, and listeners can concentrate on the content of speech and are not disturbed by the manner in which the speech is produced. Disfluency refers to the interruption of on-going speech and exhibits features such as silent and filled pauses, sound prolongations, repetitions and cut-off utterances (e.g., Shriberg et al. 2001). Disfluencies have been connected to speech planning and processing, and they are relevant for listeners’ comprehension of speech (e.g.,Wiklund and Laakso 2019). Furthermore, normally fluent speakers present disfluencies, i.e., they hesitate, repeat, and interrupt their speech in order to formulate their expressions (e.g., McDougall and Duckworth 2017). These speakers also display considerable individual variation in the rate and types of disfluency features they produce. Disfluencies are phenomena that assist speakers in planning and articulating their speech and allow time for the listeners to understand what was stated. Disfluencies can also be more disturbing and can be divided into stuttering-like disfluencies (SLD) and other disfluencies (OD) (Ambrose and Yairi 1999; Byrd et al. 2012; Tumanova et al. 2014). Ambrose and Yairi (1999, p. 899) include in the category of SLD (a) part-word repetitions (“b-but,” “thi-thi-this”), (b) single-syllable word repetitions (“you,” “and and”), (c) disrhythmic phonations, such as prolongations (“mmmmy,” “cooookie”), blocks (“#toy”), and broken words (“o#pen”). The OD group, in turn, includes (d) interjections (“um”), (e) revisions/abandoned utterances (“Mom ate/Mom fixed dinner,” “I want/Hey look at that”), and (f) multisyllable/phrase repetitions (“because because,” “I want I want to go”). Disfluencies have been defined in varying ways using different terminology, and researchers have not always held similar views on what can be defined as stuttering-like and other disfluencies.

Several previous studies have demonstrated that many speakers with ASD, and especially AS, produce disfluent speech (e.g., Shriberg et al. 2001; see Table 1 that presents the main results of a number of previous studies related to disfluencies in speech of adults and school-aged children with AS or high-functioning autism, HFA). Although the definitions of disfluencies differ, the studies report that speakers with AS have been found to produce numerous stuttering-like disfluencies such as sound, syllable, or word repetitions but also atypical disfluencies such as breaks, insertions and word-final sound prolongations.

Two of the studies on adult speakers with AS that include a control group demonstrated that speakers with AS produce more disfluencies than the controls (Shriberg et al. 2001; Lake et al. 2011). Shriberg et al. (2001) observed that when compared to controls, significantly more participants in both the AS and HFA groups had non-fluent phrasing that included sound-, syllable-, or word-level repetitions or part-word revisions, also with multiple occurrences of these behaviours within one utterance. Lake et al. (2011) detected both quantitative and qualitative differences: Adult speakers with AS produced fewer listener-oriented disfluencies such as filled pauses and revisions and more silent pauses and repetitions that reflected the speaker’s speech processing.

Three of the studies compare the disfluencies in AS with stuttering and report both similarities and differences. Studies by Scott et al. (2006) and Sisskin (2006) presented each two cases with AS without matched controls, and Scaler Scott et al. (2014) examined a group of school-aged children with AS and age-matched control groups of children who stutter (CWS) and children with no diagnosis (ND). The results of Scaler Scott et al. (2014) reveal that the three participant groups displayed differences that were statistically significant in terms of the percentage of words that contained stuttering-like disfluencies. CWS produced a higher percentage of stuttering‐like disfluencies (41% for CWS; 21% for children with AS), while the children with AS produced a higher percentage of word-final disfluencies¹ (5% for children with AS and 1% for CWS). Word‐final disfluencies were present in the speech sample of eight out of 11 children with AS. By comparison, word‐final disfluencies were present in four out of 11 children of the CWS group and three out of 11 children of the ND group. The speech of two young adults with AS analysed by Scott et al. (2006) included both stuttering-like disfluencies (part-word repetitions and blocks), as well as non-stuttering-like disfluencies (phrase repetitions and interjections). In Sisskin’s (2006) two case studies, the disfluencies of the participants with AS were not typical of a developmental fluency disorder.

The definition and classification of disfluencies differed in the studies presented in Table 1, but all studies found that speakers with AS frequently have disfluent speech with at least repetitions, and some studies also suggest that speakers with AS are more disfluent than the controls. Several studies also indicate that speakers with AS revise and self-repair their speech (Shriberg et al. 2001; Sisskin 2006; Wiklund and Laakso 2019), although contradictory findings have also been reported (Lake et al. 2011). Insertions (Lake et al. 2011), interjections (Scott et al. 2006; Sisskin 2006), pauses (Lake et al. 2011) and word-final prolongations (Scaler Scott et al. 2014; Sisskin 2006) have also been found to occur. Stuttering-like disfluencies (e.g. part or whole word repetititions and blocks) were observed in three studies (Scaler Scott et al. 2014; Scott et al. 2006; Sisskin 2006), but school-aged children with AS nonetheless differed from children who stutter. In short, all seven studies presented in Table 1 have had small samples ranging from 2 to 15 speakers with ASD and only three studies have presented control group data. Furthermore, only the study by Scaler Scott et al. (2014) examined preadolescents with a mean age of 11 years. Thus, additional research is needed to compare the disfluencies of preadolescents with ASD with an age-matched control group. Furthermore, as disfluency may disturb conversation in a similar manner as stuttering, we need to determine the interactional consequences for social conversation caused by the disfluent speech production of preadolescents with ASD.

Ungrammatical expressions also create an impression of disfluency in the speech of people with ASD (Wiklund and Laakso 2019). The prior study demonstrated that aspects that make the speech flow disfluent and difficult to follow are incorrect case endings, ambiguous pronominal references, grammatically incoherent syntactic structures and inaccurate tenses. Indeed, syntactic impairments have been found in individuals with ASD (Cummings 2014). For example, children with ASD may align their use of syntactic structure to that of a conversational partner (Allen et al. 2011). They also encounter difficulties in their grammatical comprehension of instructions (Saalasti et al. 2008) and tend to map verbs onto causative actions (Naigles et al. 2011). McGregor et al. (2012) concluded that only those children with ASD who do not have syntactic deficits demonstrate an age-appropriate knowledge of words. ASD children’s word learning has also been found to be compromised owing to their reduced sensitivity to the social information of gaze cues (Norbury et al. 2010). Thus, children with ASD may have more profound linguistic difficulties underlying their challenges in discourse and social interaction (see also Saalasti et al. 2008; Wiklund 2016; Wiklund and Laakso 2019).

Even though disfluencies by people with ASD have been studied before, the current analysis is methodologically innovative in that it adopts the framework of conversation analysis (CA) (see e.g. Heritage 1984). CA has already been used by researchers when conducting research with individuals with ASD (see for example Journal of Autism and Developmental Disorders Vol. 46, Issue 2, Feb. 2016), but a combination of CA, quantitative measures and the use of a control group in the study of disfluencies of persons with ASD is a new approach. Overall, CA is the study of recorded, naturally occurring talk-in-interaction. The aim of studying these interactions “is to discover how participants understand and respond to one another in their turns at talk, with a central focus on how sequences of actions are generated” (Hutchby and Wooffitt 2008, p. 12). Concerning atypical and asymmetric interactions (such as the therapeutic conversations of our data), in contrast to the research that focusses on individuals’ impairments, CA research has the potential of revealing participant collaboration and resources during interaction. Therefore a study adopting the methods of CA may lead to an increased understanding of the causes and interactional effects of the disfluency phenomena.

The main research objectives of this study are (1) to compare the quantity of disfluencies by preadolescents with ASD and typical development, and (2) to examine the interactional consequences of disfluencies in more detail in order to determine whether differences can be detected between the groups.

In our study, the durational mean proportions of disfluencies and ungrammatical expressions were greater in the speech of participants with high-functioning ASD/AS (26.4%) than in the control group (15.5%). Thus, our study confirms the findings by Lake et al. (2011), Scaler Scott et al. (2014) and Shriberg et al. (2001) that more disfluencies occur in the speech of individuals with ASD than in the speech of matched controls. Individual variation in the amount of disfluency was high within both groups, which is in line with the results reported on large individual ranges of speech disfluency in adult speakers (e.g., McDougall and Duckworth 2017). Furthermore, a qualitative difference was noted: The ASD group produced long and complex disfluent turns with word searches, self-repairs, false starts, fillers, sound prolongations, inconsistent syntactic structures and grammatical errors, whereas the control group mainly had disfluencies that were fillers and sound prolongations. Previously, a qualitative difference has been reported by Lake et al. (2011), who demonstrated that adult speakers with AS produced fewer listener-oriented disfluencies such as filled pauses and revisions and more silent pauses and repetitions that reflect the speaker’s speech processing. This result is in line with our findings in the sense that the disfluencies produced by the speakers with AS in our study also seem to reflect the speaker’s own grammatically disturbed speech processing, whereas the disfluencies (fillers and sound prolongations) produced by the control group members were more listener-oriented. When the percentages of speaking turns with disfluencies were compared, control speakers produced more disfluent speaking turns with one or two mild disfluencies such as a filler particle niinku (appr. “like”) and the vocalization öö (“er” or “um”) than preadolescents with ASD, who produced more speaking turns with grammatical errors and complex disfluencies. A qualitative difference in the types of disfluencies have also been observed previously between 4- and 8-year-old children with ASD and their age-matched controls: Control children produced more fillers (e.g. um, uh) (MacFarlane et al. 2017). It is also important to note that disfluencies and grammatical incoherence tended to co-occur in the speech of preadolescents with ASD, whereas similar co-occurrences were not observed in the control group.

In addition, due to the quantitative and qualitative difference, the disfluencies and the ungrammatical expressions that arose in the ASD participants’ interactions also caused comprehension problems during conversational interaction (see also Wiklund and Laakso 2019), whereas the control group did not experience comprehension problems. The control group’s disfluencies were similar to the hesitations that normally fluent speakers (c.f., McDougall and Duckworth 2017) produce in order to plan expressions and to help listeners project that a continuation is to come. The participants with ASD often had a combination of disfluencies and grammatical errors and this caused more profound problems for the listeners to understand what was said. As a consequence, the disfluency of participants with ASD had more serious effects for the interactional flow of their conversations in comparison to the age- and gender-matched controls. Our study points out that it would be very important to study in a more extensive manner the conversational interactions of participants with ASD. Further studies may reveal how the interactional difficulties arise from the morpho-syntactic level of constructing coherent speaking turns and finding correct lexical items, challenging the view of autism as a disorder of the pragmatic use of language. In the future, it would also be important to study mixed groups including participants with and without ASD in order to examine if the quantitative and qualitative difference discovered in this study can also be found in mixed groups.

The disfluencies produced by preadolescents with ASD in our data were predominantly more complex than mere repetitions. This finding is in contrast with some prior studies that have reported syllable, word or phrase repetitions that occur frequently in the speech of adult speakers with AS (e.g., Lake et al. 2011; Scott et al. 2006; Shriberg et al. 2001; Sisskin 2006). Instead, in the current study, the disfluencies by the preadolescents with ASD appeared to result from their problems in constructing grammatically coherent utterances, and displaying a difficulty in finding and/or selecting correct lexical items, both words and case endings. This difference as compared to prior studies of adults may reflect the age of the participants (i.e., the lexical and grammatical abilities by the preadolescents with ASD were not yet fully developed) or the grammatical structure of the highly inflected language, Finnish, which the preadolescents in our data were speaking. To examine the latter, it would be necessary to conduct comparative studies between data from different languages, and this is one direction for future research. In addition, the age and development of lexical and grammatical abilities may play a role, as some previous studies on disfluencies by children with ASD (e.g., MacFarlane et al. 2017; Plexico et al. 2010) have reported that English-speaking children with ASD produce atypical disfluencies related to revising the content of speech instead of plain fillers, which are produced by children with typical development. Atypical disfluencies are not observed in children with developmental stuttering either. The comparison of our results to previous studies is however difficult because the definitions and terminology related to disfluency differs and the grammatical construction of utterances is seldom examined at the same time. In future studies, it would be interesting to count the numbers of occurrences of different types of disfluencies and ungrammatical utterances in detail and to compare ASD and control groups from this point of view in larger data sets allowing statistical analyses. In particular, our study suggests that it would be important to conduct longitudinal studies or cross-sectional studies of different age groups of children with ASD in which both the lexico-grammatical development and disfluency features would be examined.

The speech of the participants with ASD in our study was characterised by morpho-syntactic problems such as incorrect case endings, ambiguous pronominal references, disconnected syntactic structures and verb tense problems. These difficulties in constructing fluent utterances can also be connected to the previous findings of problems in higher-level cognitive processes such as attention, working memory, and executive function, which have been observed by individuals with high-functioning autism (e.g Joseph et al. 2005; Kenworthy et al. 2008). Thus, future research could also address the connections between these cognitive abilities and the lexico-grammatical construction of conversational speech by individuals with ASD.

As our data constitutes authentic interaction, we believe that the results also apply at least in a part to other types of spontaneous conversation. However, in the future, it would be essential to study disfluencies, grammatical problems and their interactional consequences by individuals with ASD in different types of everyday conversational interactions. The current multi-case study can be seen as a pilot that will hopefully be followed up in the future by research based on more data from everyday social encounters of a larger group of participants with and without ASD.