Contrastive intonation in autism: the effect of speaker- and listener-perspective

Constantijn Kaland, Emiel Krahmer, Marc SwertsInterspeech 2012

TL;DR

This study investigates how individuals with autism spectrum disorder (ASD) use contrastive intonation, a way of emphasizing words to highlight differences. The research found that both typical speakers and those with ASD adjust their intonation based on whether the information is new to the listener, suggesting a shared ability to account for listener perspective in this specific linguistic function. However, differences emerged in more general prosodic features like speech dynamism and speaker commitment, where individuals with ASD were perceived as less dynamic and committed.

What Problem Does This Paper Address?

When speakers want to highlight a difference between something they're talking about and something previously mentioned, they often use "contrastive intonation." This involves emphasizing certain words and de-emphasizing others. For example, in "Last week I saw a yellow bird, but yesterday I saw a red bird," the word "red" would likely be emphasized. This technique is usually employed when the prior information is shared with the listener.

However, in real conversations, interlocutors don't always share the same information. Previous research showed that typically developing speakers can adjust their contrastive intonation based on whether the listener already knows the antecedent information. This adjustment involves making the emphasized word less prominent and the de-emphasized word more prominent when the antecedent is not shared.

Individuals with autism spectrum disorder (ASD) are known to have difficulties with "Theory of Mind" (ToM), which is the ability to understand others' mental states, beliefs, and perspectives. This can lead to communication challenges, and atypical prosody (the rhythm, stress, and intonation of speech) is often observed in ASD. While previous studies have noted general prosodic differences in ASD, there's a lack of research directly examining how functional aspects of prosody, like contrastive intonation, are produced by speakers with ASD, particularly concerning their ability to account for listener perspective. This paper aims to fill that gap by investigating whether speakers with ASD adjust their contrastive intonation based on listener knowledge, and how this compares to typical speakers.

Methods

The study employed a two-part experimental design:

  1. Production Experiment: Participants acted as speakers in a referential communication task. They were instructed to guide listeners in placing figures on a bingo card. The task was designed to elicit contrastive intonation by presenting scenarios where a target figure could be distinguished by color or shape, and where this information was either shared with the listener (antecedent shared) or not (antecedent different). The key manipulation was whether the speaker was instructing the same listener sequentially or different listeners. Speech was recorded for later analysis.
  2. Prosodic Analysis:
    • Prominence Analysis: Recorded noun phrases (NPs) containing target figures were analyzed for prominence using a perception rating task by three intonation experts. They rated the strength of accent on a three-point scale. A "prominence difference score" was calculated by subtracting the prominence of the unfocused word from the focused word. This score indicates how much more prominent the focused word was.
  3. Perception Experiment: A separate group of participants listened to NPs from both typical and ASD speakers and judged them on "speech dynamicity" (how monotonous or dynamic the speech sounded) and "speaker commitment" (how committed the speaker seemed). These judgments were made on a five-point scale.

Statistical analysis involved Repeated Measures ANOVAs to compare prominence difference scores and perception ratings across different conditions (developmental group, listener perspective, focus type).

Datasets

The study utilized NPs collected from both typical speakers (data from a previous study [3]) and speakers with High-Functioning Autism (HFA) in the current production experiment. These NPs were then used as stimuli in the perception experiment. No new datasets were introduced; rather, existing data was analyzed and expanded upon.

Key Insights and Ideas

  • Shared Perspective-Taking in Contrastive Intonation: Both typical and ASD speakers demonstrate an ability to adjust contrastive intonation based on whether the antecedent information is shared with the listener. This suggests that, for this specific linguistic function, individuals with ASD can account for listener perspective, contrary to some broader theories of ToM deficits.
  • Focus on Functional vs. Formal Prosody: The study highlights a crucial distinction between the functional use of prosody (like contrastive intonation for information highlighting) and its formal aspects (like the acoustic qualities of speech, e.g., dynamism, pitch variation). While functional aspects showed surprising similarity, formal aspects revealed notable differences.
  • Subtle Differences in Focus Marking: While generally similar, ASD speakers showed a slight tendency to over-emphasize adjectives when the antecedent was not shared, a deviation not observed in typical speakers. This suggests a less refined control over focus marking in specific contexts.
  • Perceived Dynamism and Commitment: Listeners consistently perceived speakers with ASD as less dynamic and less committed compared to typical speakers. This aligns with common observations of atypical prosody in ASD.
  • Interplay between Dynamism and Commitment: The perception experiment revealed a strong correlation between speech dynamism and speaker commitment, suggesting that a more dynamic speech pattern leads listeners to perceive the speaker as more committed.

Key Results

  • Prominence Difference Scores:
    • Both groups produced significantly larger prominence difference scores when the antecedent was shared with the listener compared to when it was not.
  • Speakers produced larger differences when the focused word was an adjective compared to a noun.
  • A trend indicated that typical speakers produced larger differences than HFA speakers, though this was not statistically significant.
  • Perception of Dynamicity and Commitment:
    • Typical speakers were perceived as more dynamic (M=3.22) than HFA speakers (M=3.14).
  • Typical speakers were perceived as more committed (M=2.95) than HFA speakers (M=2.85).
  • Speakers producing focused adjectives were perceived as more committed than those producing focused nouns.
  • This commitment effect interacted with development, with typical speakers showing a larger difference in perceived commitment between adjective and noun focus than HFA speakers.

Limitations and Future Work

  • Subtle Differences in Production: While the study found no major differences in the functional use of contrastive intonation, the subtle tendency for HFA speakers to over-emphasize adjectives when the antecedent is not shared warrants further investigation. More data is needed to assess if this difference is truly meaningful.
  • Explanation for Production Differences: The observed differences in speech dynamicity and commitment in HFA speakers might contribute to their overall less clear contrastive intonation, as a flatter intonation could lead to smaller prominence differences. This connection could be explored further.
  • ToM Interpretation: The findings challenge a strict interpretation of impaired ToM as the sole explanation for prosodic difficulties in ASD, as HFA speakers did account for listener perspective in contrastive intonation. Future research could explore how different aspects of ToM might relate to specific prosodic production and perception abilities.
  • Generalizability: The study focused on a specific linguistic phenomenon (contrastive intonation) and a specific population (HFA). Future work could investigate other functional prosodic features and a broader range of ASD presentations.

Summary updated

BibTeX
@inproceedings{kaland2012contrastive,
  author = {Constantijn Kaland and Emiel Krahmer and Marc Swerts},
  title = {{Contrastive intonation in autism: the effect of speaker- and listener-perspective}},
  booktitle = {Interspeech},
  year = {2012},
}
Figure