Linguistic and Prosodic Correlates of Perceived Social Skills in Conversation

Background:

Autism spectrum disorder (ASD) is characterized by deficits in social communication, and even fluent children with ASD are often perceived by conversational partners as more socially awkward than typically-developing (TD) peers, even after very brief exposure (Grossman, 2015). However, little research has investigated how such rapid social judgments are made, nor if there are specific conversational (linguistic or prosodic) cues that are associated with such judgments.

Objectives:

To determine if specific linguistic and prosodic patterns are associated with positive or negative social perceptions of speakers during conversation.

Methods:

Thirty-four children and adolescents (mean age = 11.7 years, SD = 2.27 years) with ASD (N=12), autism (N=11), and typical development (N=11) participated in a short (2-5 minute) semi-structured conversation with an adult speaker. The first 30 seconds of each conversation were rated by a large number (M = 72.9, SD = 1.2, range=70-76) of native American English-speaking Mechanical Turk workers from 0 to 100 on five social dimensions: likability, outgoingness, social skillfulness, responsiveness, and fluency. The same 30 seconds of each conversation were transcribed and analyzed for linguistic characteristics (pauses during and between conversational turns, length and number of speech turns, errors, stutters, echoes, self-repairs, overlaps, “filler” words such as ah, eh, em, er, oh, um, and like) and robust prosodic features (median fundamental frequency (f₀), f₀range, and median syllable rate).

Results:

A one-way ANOVA found no differences between the ASD, autism, and TD groups on any of the five social dimensions. In fact, the ASD group was rated non-significantly higher than both the TD and autism groups on all dimensions (Figure 1). However, several linguistic and prosodic variables were associated with perceived social skills. The total amount of time that the child was silent throughout the interaction (total pauses) as well as the amount of time that the child was silent immediately after his partner’s conversational turn (no responses) were strongly negatively correlated with perceptions of social skills (Table 1). Conversely, the quantity of child speech (total speech) was strongly positively correlated with social perceptions on all five dimensions (Table 1). Of the prosodic characteristics, median f₀ was moderately positively correlated with likability only; this relation is maintained after controlling for gender, age, and ASD diagnosis; r = 0.52 (p< 0.01).

Conclusions:

Contrary to our hypothesis, we did not find that children with ASD/autism were perceived more negatively than TD children on five social dimensions. Rather, our research suggests that conversational partners are sensitive to the quantity of speech and silence produced during a conversation, as well as the timingof such responses. These factors, which may be indicative of individual style, personality, and temperament, could drive rapid judgments of social skills more than diagnostic classification.

Table 1. Pearson’s r values between language sample characteristics and perceived social skills (p < .001 for all correlations).

	Likable	Outgoing	Socially skilled	Responsive	Fluent
Total pauses	-0.66	-0.70	-0.74	-0.75	-0.84
No responses	-0.73	-0.70	-0.72	-0.78	-0.75
Total speech	0.56	0.68	0.67	0.74	0.66

Figure 1. Mean ratings of the three groups on five social dimensions.

See more of: Communication and Language
See more of: Communication and Language